Fulditoxin, representing a new class of dimeric snake toxins, defines novel pharmacology at nicotinic ACh receptors

Animal toxins have contributed significantly to our understanding of the neurobiology of receptors and ion channels. We studied the venom of the coral snake Micrurus fulvius fulvius and identified and characterized the structure and pharmacology of a new homodimeric neurotoxin, fulditoxin, that exhibited novel pharmacology at nicotinic ACh receptors (nAChRs).

The short-chain α-3FNTx, erabutoxin-a (EbTx-a), and long-chain α-3FNTxs, α-cobratoxin and α-BgTx use a common core of canonical AA residues in loops I and II, in addition to specific combinations of conserved AA residues for binding to muscle and neuronal α7 nAChRs see Figure S1). Interestingly, in Type III α-3FNTxs (Jackson et al., 2013) and Ω-neurotoxins (Hassan-Puttaswamy, Adams, & Kini, 2015), two unrelated classes of nAChR antagonists, almost all conserved functional AA residues of α-3FNTxs are replaced, revealing that these toxins may utilize alternative functional sites to bind to a common molecular target within nAChRs.

What is already known
• Snake three-finger α-neurotoxins competitively and selectively inhibit muscle and neuronal nAChRs.
• Their selectivity for nAChR subtypes is defined by conserved functionally invariant amino acid residues.
What does this study add • Fulditoxin represents a novel dimeric three-finger α-neurotoxin family which lacks canonical functionally invariant residues.

What is the clinical significance
• This novel neurotoxin class offers new insights into interactions between peptide antagonists and nAChRs.
Here, we describe the identification, and pharmacological and structural characterization of a dimeric α-3FNTx, fulditoxin, from Micrurus fulvius fulvius venom. The 1.95-Å crystal structure of fulditoxin revealed a homodimer of short-chain α-3FNTxs non-covalently bound by hydrophobic interactions, with the ability to form a tetrameric complex in the presence of zinc ions (Zn 2+ ), exhibiting a unique metal-binding capability not previously reported for 3FTxs. Interestingly, despite lacking all the functionally important AA residues critical for α-3FNTx interaction with nAChRs, fulditoxin produced potent, but completely reversible, postsynaptic neuromuscular blockade, as well as broad spectrum inhibition of α1β1δε, α4β2, α7, and α3β2 nAChRs. For electrophysiological studies using the Xenopus oocyte expression system, all experiments on Xenopus laevis frogs conformed to the Geneva Canton Rules on Animal Experimentation (Accreditation # G171/3551) and the RMIT University Animal Ethics committee (Protocol # 1222(Protocol # , 1223. Female X. laevis frogs were sourced from the Centre de Resources Biologiques (Rennes, Cedex, France) and maintained in an aquarium as approved by the Ethics Committee for Animal Experimentation (Swiss Academy of Medical Sciences) and from Nasco (Fort Atkinson, WI, USA) and housed in the RMIT Aquatic Facility. A maximum of three frogs were kept in purpose-built 10-L tanks at 20-25 C with 12-hr light/dark cycle within. Electrophysiological experiments were performed using oocytes obtained from three~5-year-old frogs. Frogs were anaesthetized with 1.7 mgÁml −1 of ethyl 3-aminobenzoate methanesulfonate (pH 7.4 with NaHCO 3 ), and for post-surgery recovery, animals were placed in fresh water at a level below the nostrils. Frogs were left to recover for a minimum of 4 months between surgeries. Terminal anaesthesia with 5.0 mgÁml −1 of ethyl 3-aminobenzoate methanesulfonate (pH 7.4 with NaHCO 3 ) was performed on frogs at the sixth surgery.
Animal studies are reported in compliance with the ARRIVE guidelines (Kilkenny, Browne, Cuthill, Emerson, & Altman, 2010;McGrath, Drummond, McLachlan, Kilkenny, & Wainwright, 2010) and with the recommendations made by the British Journal of Pharmacology.

| Chromatographic purification of proteins from Micrurus fulvius venom
The isolation and purification of fulditoxin from crude Micrurus fulvius venom (MFV) was carried out using established protocols in our laboratory (Hassan-Puttaswamy, Adams, & Kini, 2015;Pawlak et al., 2009;Roy et al., 2010). Crude MFV (50 mg dissolved in 1-ml MilliQ water and filtered using a 0.45-μm syringe filter) was loaded onto a Superdex 30 Hiload size-exclusion chromatography column, equilibrated with 50-mM Tris-hydrochloric acid (Tris-HCl) buffer (pH 7.4), and proteins were eluted with the same buffer using an ÄKTA purifier system (GE Healthcare, Little Chalfont, UK). Fractions containing the protein of interest were further purified by reversed phase-HPLC (RP-HPLC) using a Jupiter C18 preparative column that was equilibrated with 0.1% (v/v) TFA. Proteins were eluted with a linear gradient of 80% (v/ v) MeCN in 0.1% (v/v) TFA. Elution of proteins was monitored at 280 and 215 nm.

| Electrospray ionization-MS
The molecular mass and homogeneity of the purified protein were determined by injecting the protein samples into an API-300 LC-tandem MS system (PerkinElmer Life Sciences, Massachusetts, USA) as described previously (Roy et al., 2010). Ion spray, orifice, and ring voltages were set at 4600, 50, and 350 V respectively. Nitrogen was used as the nebulizer and curtain gas. A Shimadzu LC-10 AD pump was at a flow rate of 40 μlÁmin −1 . Analyst software (PerkinElmer Life Sciences, MA, USA) was used to analyse and deconvolute the raw MS data. Fractions that showed the expected molecular mass of fulditoxin were pooled and lyophilized.

| Capillary zone electrophoresis
The homogeneity of purified fulditoxin was assessed by capillary zone electrophoresis which was performed on a BioFocus3000 system (Bio-Rad, Foster City, CA, USA). The native protein was dissolved in MilliQ water (1.9 μgÁμl −1 ) and injected into a 25 μm × 17 cm coated capillary under a pressure mode (5 ψÁs −1 ) and run in 0.

| Protein reduction and alkylation
Lyophilized protein (1.04 mg) was dissolved in 520 μl of denaturing buffer (0.13 molÁL −1 of Tris-HCl, 1 mmolÁL −1 of EDTA, 6 molÁL −1 of guanidine-HCl, pH 8.5). β-Mercaptoethanol (1.05 μl) was added, and the solution was incubated under nitrogen at room temperature for 2 hr. The alkylating agent, 4-vinylpyridine (4.82 μl), was subsequently added and the solution was incubated under nitrogen for another 3 hr at room temperature. After the reaction, the S-pyridylethylated protein was separated from the reaction mixture by RP-HPLC on a Jupiter C18 semi-preparative column using a linear gradient of 80% MeCN in 0.1% TFA at a flow rate of 2 mlÁmin −1 . The molecular mass of the S-pyridylethylated protein was determined by electrospray ionization-MS (ESI-MS).

| Chemical cleavage
The S-pyridylethylated protein was subjected to chemical cleavage at the C-terminus of methionine residues with cyanogen bromide (CNBr). The S-pyridylethylated protein (500 μg) was dissolved in 250 μl of 70% TFA to which 5 μl of β-mercaptoethanol was added. A 200-fold molar excess (over methionine residues) of CNBr in 70% TFA was added to make a final protein concentration of 1 mgÁml −1 .
After 24 hr of incubation in the dark at room temperature, 10 volumes of MilliQ water was added to the reaction mixture and lyophilized.
Lyophilized sample containing cleaved peptides were reconstituted in 0.1% TFA and separated by RP-HPLC on a Jupiter C18 analytical column using a linear gradient of 80% MeCN in 0.1% TFA at a flow rate of 1 mlÁmin −1 . Molecular masses of the separated peptide fragments were determined by ESI-MS and their AA sequences determined by N-terminal sequencing. The complete AA sequence of fulditoxin was determined by alignment of the sequences of the CNBr-cleaved peptides.
2.5 | Size-exclusion chromatography for determination of oligomeric states of fulditoxin The oligomeric states of fulditoxin (0.15-1.5 μM) were determined by analytical size-exclusion chromatography using a Superdex 75 column (1 × 30 cm) equilibrated with 50 mM of Tris-HCl buffer, pH 7.4, in the absence and presence of 8-M urea. Size-exclusion chromatography was carried out on an ÄKTA purifier system at a flow rate of 0.6 mlÁmin −1 . Columns were calibrated with BSA (66 kDa), carbonic anhydrase (29 kDa), cytochrome C (12.9 kDa), aprotinin (6.5 kDa), and blue dextran (200 kDa) as molecular mass markers. Native proteins as well as protein samples in 8-M urea were loaded separately onto the column.
The predominant product was isolated by preparative HPLC (2 mg) and its mass determined by ESI-MS. Analytical HPLC was carried out to compare the synthetic fulditoxin (sFulditoxin) with venom-purified native protein, nFulditoxin.

| Chick biventer cervicis muscle preparation
The chick biventer cervicis muscle (CBCM) nerve-skeletal muscle preparation (Ginsborg & Warriner, 1960)  Atropine (0.5 μM) was added to all solutions to block all activity attributed to muscarinic AChRs. ACh and nFulditoxin were prepared fresh in OR2 solution and applied to the bath by gravity-driven perfusion.
Control responses were recorded before toxin exposure and toxin incubation were carried out in the recording chamber between 3 and 5 min with agitation. All recordings were performed with a TEVC automated robot system. The data were digitized and analysed offline using MATLAB (Mathworks, Natick, MA, USA; RRID:SCR_001622).

| Electrophysiological studies on sFulditoxin
Independent of the experiments described above, functional characterization of sFulditoxin at receptor level was carried out using TEVC electrophysiology as described previously (Hassan-Puttaswamy, Adams, & Kini, 2015). Stage V-VI oocytes were harvested from mature female Xenopus laevis under anaesthesia with 0.1% Tricaine as approved by the RMIT Animal Ethics Committee (Protocol # 1222, 1223). cRNA encoding for nAChR subunits were prepared and microinjected into oocytes for nAChR expression. Briefly, cDNA encoding the human α3, α4, α9, α10, α7, β2, and β4 subunits, and rodent α1, β1, γ, δ, and ε subunits of nAChRs were sub-cloned into the oocyte expression vector pT7Ts and used for cRNA preparation using the mMESSAGE mMACHINE kit (Ambion ® , Life Technologies Australia Pty. Ltd., Mulgrave, VIC, Australia); 25-ng cRNA was microinjected into oocytes, at α:β ratio of 1:1 for the human heteromeric neuronal nAChR subtypes and α1:β1:δ:ε ratio of 2:1:1:1 for the rat muscle All TEVC recordings were done manually, at room temperature (20-23 C), using a bath solution of ND96. The oocytes were continuously perfused at a flow rate of 2 mlÁmin −1 with ND96 buffer during recordings, with toxin incubated for 5 min before ACh was added.
The concentration of ACh, calculated to be the EC 50 concentration for the respective nAChR subtype, was applied for 2 s at 2 mlÁmin −1 , with 5-min washout periods between applications. Oocytes were voltage clamped at a holding potential of −80 mV for nAChRs. Data were filtered at 100 Hz and sampled at 500 Hz.

| Data analyses
In screening of fulditoxin on various nAChR subtypes, concentration response curves for antagonists were fitted by unweighted non-linear regression to the logistic equation: where Y is the normalized response, X is the antagonist concentration, top and bottom are maximal and minimal normalized responses, respectively, and IC 50 is the antagonist concentration giving 50% inhibition of the maximal response.
EC 50 shift of concentration-response plots were used to determine the competitive or non-competitive nature of the antagonist.
The EC 50 shift concentration-response plot (Arunlakshana & Schild, 1959) was plotted using the equation: EC 50 = 10^Log EC 50 ; EC 50 is the concentration of agonist that gives half maximal response; pA 2 is the negative logarithm of the concentration of antagonist needed to shift the concentration-response curve by a factor of 2; Hill slope describes the steepness of the family of curves; Schild slope quantifies how well the shifts correspond to the prediction of competitive interaction and if the action is competitive, the Schild slope will equal 1.0, and top and bottom are plateaus in the units of the Y axis.
The data and statistical analysis in this study comply with the recommendations of the British Journal of Pharmacology on experimental design and analysis in pharmacology (Curtis et al., 2018). Given the limitations in the availability of fulditoxin due to its ultra-low yield in venom (<0.2%), and complexity in its chemical synthesis, statistical analyses have not been performed because fewer than five independent experiments were performed during the course of our studies.
The pharmacology studies were not blinded as there was no expected outcome of screening fulditoxin across nAChR subtypes for activity and determining the concentration-response relationships.
2.10 | Crystal structure determination of nFulditoxin  Table 1). The data sets were processed and scaled using the program HKL2000 (Otwinowski & Minor, 1997). The Matthews coefficient was calculated as 2.18 Å 3 /Da (Matthews, 1968) corresponding to a solvent content of 44%. The structure was solved by the molecular replacement method using the program Phaser from the Phenix suite (McCoy et al., 2007;Zwart et al., 2008;RRID: SCR_014224) with β-cardiotoxin (PDB ID 3PLC) as the search model.
The initial model was refined using Phenix-refine followed by model building using the Phenix AutoBuild program (Adams et al., 2004). The missing residues were manually built using the Coot program (Emsley & Cowtan, 2004; RRID:SCR_014222), followed by refinement with Phenix-refine (Adams et al., 2002). No non-crystallographic symmetry restrain was used in the final refinement cycle. Finally, 141 well-defined water molecules were added, and the R value converged to 20.0% (R free = 23.7%) with good stereochemical parameters (Table 1).

| Dynamic light scattering
The apparent hydrodynamic radii of fulditoxin (5  Additionally allowed regions (%) 14.7 Generously allowed regions (%) 0 Disallowed regions (%) 0 a R sym = P |I i − <I>|/ P |I i |, where I i is the intensity of the ith measurement and <I> is the mean intensity for that reflection. b R work = P |F obs − F calc |/ P |F obs |, where F calc and F obs are the calculated and observed structure factor amplitudes respectively. c R free = as for R work , but for 10.0% of the total reflections chosen at random and omitted from refinement. PHARMACOLOGY (Harding et al., 2018), and are permanently

| Fulditoxin forms a non-covalent dimer in solution
During size-exclusion chromatography, it was observed that fulditoxin was contained in peak 3 (Figure 1a), in contrast to most 3FTxs which eluted later, as observed in our previous studies (Nirthanan, Charpantier, et al., 2002;Rajagopalan et al., 2007). This suggested that fulditoxin may exist in dimeric form. In analytical size-exclusion chromatography, fulditoxin eluted as a single peak corresponding to a relative molecular mass of 3.88 kDa in the presence of 8-M urea, compared to as a single peak corresponding to a mass of 11.56 kDa in the absence of urea (Figure 1f). Hence, this purified toxin was named "fulditoxin"-Micrurus fulvius fulvius dimeric neurotoxin.

| Amino acid sequence of fulditoxin
N-terminal sequencing of fulditoxin was carried out by automated Edman degradation. The first 40 AA residues of the toxin were F I G U R E 1 Isolation and purification of fulditoxin from crude Micrurus fulvius fulvius venom. Multi-step HPLC purification of M. fulvius venom, determination of fulditoxin's MW by MS, and determination of its purity by capillary electrophoresis. (a) Size-exclusion chromatography of crude MFV (50 mgÁml −1 ) on a Superdex 30 HiLoad (16/60) column. The column was pre-equilibrated with Tris-HCl buffer (50 mM), pH 7.4. Proteins were eluted at a flow rate of 1.0 mlÁmin −1 using the same buffer. Protein elution was monitored at 280 nm. The fractions indicated by a horizontal black bar corresponding to peak 3 were pooled and subjected to RP-HPLC. (b) RP-HPLC profile of size-exclusion chromatography peak 3 on a Jupiter C18 preparative column (5 μm, 300 Å, 21.2 × 250 mm) using a linear gradient (25-50% over 108 min; dotted line) of buffer B (80% ACN in 0.1% TFA) at a flow rate of 5 mlÁmin −1 . Protein elution was monitored at 215 nm. The peak indicated by an arrow corresponding to peak 3c was rechromatographed on a shallower gradient. (c) Rechromatogram of peak 3c obtained by RP-HPLC on a semi-preparative column (5 μm, 300 Å, 10 × 250 mm) using a shallow gradient (25-45% over 80 min; dotted line) of buffer B (80% ACN in 0.1% TFA) at a flow rate of 2 mlÁmin −1 . Protein elution was monitored at 215 nm. The resulting protein peak indicated by the arrow was named fulditoxin. (d) ESI-MS profile of fulditoxin. The spectrum shows a series of multiply charged ions with mass/charge (m/z) ratios ranging from +4 to +6 charges. Inset, reconstructed mass spectrum of fulditoxin corresponding to a single homogeneous protein with MW of 6947.4 Da; CPS = counts s −1 ; a.m.u = atomic mass units. (e) Capillary electrophoresis of fulditoxin shows a single peak supporting homogeneity. (f) Analytical size-exclusion chromatography profile of fulditoxin; 1.5-μM fulditoxin was loaded onto a Superdex 75 column (1 × 30 cm) equilibrated with 50-mM Tris-HCl buffer, pH 7.4 at a flow rate of 0.6 mlÁmin −1 in the presence and absence of 8-M urea respectively. (G) MW determination on size-exclusion column. Column calibration was done using BSA (ALB; 66 kDa), carbonic anhydrase (CA; 29 kDa), cytochrome c (CYC; 12.9 kDa), aprotinin (AP; 6.5 kDa), and blue dextran (200 kDa) as molecular mass markers identified by direct sequencing of the native toxin, and the remaining residues were determined by sequencing of overlapping fragments of chemically-cleaved fulditoxin (see Figure S2). The calculated mass of 6946.85 Da of the full-length fulditoxin sequence agreed with its experimentally determined molecular mass (6947.4 ± 0.6 Da). Basic Local Alignment Sequence Tool (BLAST) search (Altschul et al., 1997;RRID:SCR_004870) showed that fulditoxin shares the highest identity Notably, none of the conserved functional AA residues experimentally shown to be critical for α-3FNTxs to inhibit nAChRs are present in the sequence of fulditoxin.

| Chemical synthesis of fulditoxin
Solid-phase assembly of full-length fulditoxin (1-63) was initially attempted using automated Fmoc (9-fluorenylmethoxycarbonyl) chemistry. While the correct mass was observed in the crude product, it was not possible to efficiently separate it from accompanying impurities. It was therefore decided to synthesize the peptide fragments that could be subsequently assembled using native chemical ligation (Dawson, Muir, Clark-Lewis, & Kent, 1994). The 1-41-thioester segment was synthesized as a single chain using Boc (tertbutyloxycarbonyl) chemistry and subsequently solubilized and purified (Alewood et al., 1997). ESI-MS of the purified product revealed a mass of 4937.2 Da. Likewise, the C-terminal 42-63 fragment was synthesized and solubilized, and the purified product showed a mass of
Due to the scarcity of nFulditoxin, the highest concentration tested was 15 μM, which produced only 45% inhibition of ACh-evoked currents in hα7 nAChRs but was adequate to completely block hα1β1δε nAChRs.
We then characterized the selectivity of sFulditoxin for seven different nAChR subtypes, namely, rodent adult muscle (rα1β1δε) and human neuronal hα7, hα9α10, hα4β2, hα4β4, hα3β2, and hα3β4 nAChRs (Figure 4a) (IC 50 = 12.6 μM; 95% CI [11.3, 13.9]) subtypes. It also weakly inhibited hα4β4 (IC 50 not determined), but not the hα9α10 and hα3β4 nAChRs, even at 30 μM (Figure 4b,c). The n H of fulditoxin binding to rα1β1δε, hα4β2, hα7, and hα3β2 receptors were −1.5, −1.1, −1.7, and −1.2 respectively, suggesting non-cooperative ligand-receptor interaction (Prinz, 2010). Importantly, it was observed that sFulditoxin blocked rα1β1δε (IC 50 = 2.6 μM) and hα7 (IC 50 = 7.0 μM) subtypes with IC 50 values similar to those for hα1β1δε (IC 50 = 2.56 μM) and hα7 (IC 50 = 6.57 μM) nAChRs, noting that the experiments with F I G U R E 2 Comparison of the amino acid sequence of fulditoxin with sequences other snake three-finger toxins. In all panels, the accession numbers and the source organism are indicated. Only the conserved cysteine residues within each group are shaded in grey. The number of amino acid (AA) residues and the percentage identity (% Id) of the respective toxins with fulditoxin is indicated at the end. The disulfide linkages and segments contributing to the three loops are also shown. (a) The AA sequence of fulditoxin was subjected to Protein BLAST ® search for sequence homology with protein databases. The top 16 protein sequences that shared the highest sequence homology with fulditoxin are shown. Identical AA residues across all 17 sequences are shaded in grey. The AA residues that contribute to hydrogen bond formation (shaded in black) and hydrophobic interactions (underlined) in the dimeric interface of fulditoxin are highlighted across the sequences. (b) The sequence of fulditoxin as compared with the sequences of other 3FTxs, each representing a different subfamily: erabutoxin-a, short-chain α-neurotoxin; α-bungarotoxin, long-chain α-neurotoxin; κ-bungarotoxin, κ-neurotoxins; candoxin, elapid non-conventional three-finger α-neurotoxin; denmotoxin, colubrid non-conventional three-finger α-neurotoxins; MT2, muscarinic three-finger toxin; cytotoxin 4, cardiotoxin; mambin, antagonist of cell-adhesion processes; Toxin FS-2, L-type calcium channel antagonist; and fasciculin-1, AChE inhibitor. (c) Comparison of the AA sequence of fulditoxin with the sequences of short-chain 3F-αNTxs. The conserved AA residues in short-chain 3F-αNTxs deemed critical for the recognition and binding to muscle-type nAChRs are highlighted in black. (d) Comparison of the AA sequence of fulditoxin with the sequences of dimeric 3F-NTxs. The AA residues in fulditoxin contributing to dimerization are highlighted in black; and His29 involved in Zn 2+ binding is italicized and underlined. The cysteine residues involved in the formation of intermolecular disulfide linkages in the α-cobratoxin dimer and irditoxin A-irditoxin B dimer are underlined. (e) Comparison of the AA sequence of fulditoxin with the sequences of reversible neurotoxins sFulditoxin and nFulditoxin were independently carried out in different laboratories. Therefore, we conclude that sFulditoxin and nFulditoxin are structurally and functionally equivalent.

| Fulditoxin competitively inhibits ACh binding to the ACh-binding pocket of muscle nAChRs
To determine whether fulditoxin's low sequence identity with other α-3FNTxs (including the absence of canonical functional AA residues) would allow it to potentially bind at a different site at the nAChR

| Fulditoxin is a unique non-covalent dimer
The crystal structure of fulditoxin was determined to 1.95-Å resolution ( Figure 5; Table 1). There were two molecules, each consisting of AA residues Leu1 to Lys58 (Figure 5a), forming a tight dimer in the asymmetric unit. Both monomers were well defined in the electron density map except for five C-terminal residues (Figure 5b). Each monomer of the asymmetric unit has the characteristic three-finger protein scaffold consisting of three β-sheeted loops extending from a central core which is stabilized by four highly conserved disulfide bridges (Cys3-Cys20, Cys13-Cys38, Cys42-Cys50, and Cys51-56).
Loop I is formed by two anti-parallel β-strands βA (Lys2-Tyr4) and βB  (Table S1). Each monomer of fulditoxin is structurally similar to short-chain α-3FNTxs such as EbTx-a (PDB code 5EBX) and Naja nigricollis toxin-α (PDB code 1IQ9; Figure 5e), but loops I and III in fulditoxin are shorter, and loop II is orientated in the opposite direction when compared to the corresponding loops in other short-chain α-3FNTxs. Fulditoxin also showed overall structural similarity with long neurotoxin-1 (PDB code 1YI5), but significant differences were observed in the length of loop I and orientation of loop II between long neurotoxin-1 and short-chain α-3FNTxs, EbTx-b (PDB code 6EBX), and fulditoxin ( Figure S6).

| Dimer interface comprised primarily of hydrophobic interactions
Approximately 514-Å 2 (~12% of the total) surface areas were involved in the dimerization as calculated by the Protein Interfaces, Surfaces and Assemblies server analysis (Krissinel & Henrick, 2007). There are 20 AA residues from both monomers involved in the dimerization; with close contacts between the monomers maintained by 29 hydrophobic interactions and three hydrogen bonds (<3.2 Å; see Tables S2A,B and Figure 5b). These three hydrogen bonds were all side chain-side chain contacts, with two of them observed between Thr37 and His43 from both monomers. The hydrogen bonding contacts were observed between βE of monomer A and βD of monomer B and vice versa. The third hydrogen bond was observed between Ser40 of both monomers located in the loop which connects βD and βE (Table S2A; Figure 5b). These observations strongly suggest the existence of fulditoxin as a non-covalently linked homodimer, which is consistent with analytical size-exclusion chromatography observations ( Figure 1f).

| Zinc-binding enables fulditoxin to form a tetrameric complex
The crystal structure of fulditoxin, determined under conditions containing Zn 2+ , revealed two Zn 2+ ions, each bound to four fulditoxin monomers arranged in a tetrahedral coordination geometry, resulting F I G U R E 4 Activity of fulditoxin at nAChR subtypes expressed in Xenopus oocytes. TEVC electrophysiological characterization of synthetic fulditoxin (sFulditoxin) on nAChR subtypes expressed in Xenopus oocytes. (a) Superimposed representative ACh-evoked currents from Xenopus oocytes expressing rαβδε, hα7, hα9α10, hα3β2, hα4β4, hα4β2, and hα3β4 nAChRs in the absence (black trace) and presence of 30-μM sFulditoxin (red trace). For the nAChR subtypes indicated, r and h represent the species, rodent and human respectively. Whole-cell nAChR-mediated currents were activated by ACh at a concentration that represented the EC 50 of ACh for the respective subtype expressed in the oocyte.  Figure S7A). The tetrameric complex is formed by four dimeric fulditoxin molecules held together by two Zn 2+ , with each Zn 2+ interacting with His29 of four different monomers from the four fulditoxin molecules ( Figure S7B). Thus, Zn 2+ appears to induce the oligomerization of fulditoxin molecules.
Dynamic light scattering experiments confirmed that fulditoxin exists as a dimer (relative molecular mass of 16.1 kDa) in the absence of Zn 2 + and as a tetramer of dimers (relative molecular mass of 34.3 kDa) in the presence of 3-mM Zn 2+ . Furthermore, in the presence of 8-M urea, reduction of the fulditoxin dimer to the monomer species (relative molecular mass of 6.18 kDa) was observed (Table S3).

| DISCUSSION
In the five decades since the discovery of α-BgTx, the scope of using animal toxins as molecular probes for localization and characterization of membrane receptors and ion channels and as therapeutic leads for drug discovery has broadened substantially (King, 2011;Lewis & Garcia, 2003). Notwithstanding the extensive structure-function analyses done on typical snake α-3FNTxs such as α-BgTx, the discovery of α-3FNTxs with novel structural characteristics (dimeric α-3FNTxs [Pawlak et al., 2009;Roy et al., 2010]; non-conventional α-3FNTxs [Nirthanan, Gopalakrishnakone, Gwee, Khoo, & Kini, 2003]; or toxins with unique sequences [Ω-neurotoxins]; Hassan-Puttaswamy, Adams, & Kini, 2015) suggests that some α-3FNTxs may interact with nAChRs via different functional sites and display selectivity for nAChR subtypes not described previously. Here, we have described the discovery, chemical synthesis, and pharmacological and X-ray crystallographic characterization of fulditoxin, a structurally unique, homodimeric neurotoxin from coral snake (Micrurus fulvius fulvius) venom, which showed high selectivity towards avian muscle nAChRs, and broad selectivity for rodent and human muscle and neuronal nAChRs while lacking all AA residues critical for binding to nAChRs. It . Both the monomers were related by a twofold symmetry and their superimposition yielded an RMSD of 0.6233 Å for 58 Cα atoms. The N-and C-termini are labelled as NH 2 and COOH respectively. (e) Superimposition of fulditoxin monomer A with short-chain threefinger α-neurotoxins. Fulditoxin monomer A is shown in blue, erabutoxin-a in yellow (5EBX; RMSD 1.86 Å for 53 Cα atoms), erabutoxin-b in pink (3EBX; RMSD 1.87 Å for 53 Cα atoms), toxinα in orange (1IQ9: RMSD 1.88 Å for 53 Cα atoms), and neurotoxin-β in cyan (1NXB; RMSD 1.98 Å for 51 Cα atoms). Loops I to III are indicated. The PDB codes are given in parentheses is also the first example of a snake dimeric neurotoxin that produced reversible neuromuscular blockade and the first 3FTx to show metal binding that enables the formation of a tetrameric complex.
The mechanism of reversible interaction of snake α-3FNTxs with muscle nAChRs is complex and not well understood. Studies on conventional monomeric long-chain α-3FNTxs have found their off-rates from nAChRs to be extremely slow, sometimes lasting days (Young, Herbette, & Skita, 2003), with a half-time for dissociation of the [ 3 H]α-toxin-Torpedo nAChR-rich membrane complex to be~60 hr (Weber & Changeux, 1974). Therefore, our experimental design using miniaturized organ baths to estimate reversal of α-3FNTx-induced neuromuscular blockade was limited, given the spontaneous decline in contractility of isolated CBCM observed after~300 min.
However, reversibility could not simply be attributed to weak or strong binding affinities of the toxin to the receptor. For instance, fulditoxin is as potent as α-BgTx and EbTx-b in producing neuromuscular blockade in avian muscle but is almost completely reversible in its action in contrast to α-BgTx and EbTx-b. Likewise, candoxin is reversible in its action at hα1β1δε, but not at hα7 receptors, while binding with nanomolar affinity to both subtypes (Nirthanan, Charpantier, et al., 2003). The reversibility of α-3FNTx action has also been attributed to a specific site of interaction on the toxin, distinct from its functional site, based on multiple sequence analyses which revealed that readily reversible α-3FNTxs lacked the conserved Asp31, postulating a role for this residue in determining reversibility (Harvey, Hider, Hodges, & Joubert, 1984;Nirthanan, Charpantier, et al., 2003). As fulditoxin binds to the ACh-binding pocket (described below) with a yet-to-be-identified, distinct functional site, the mechanism of fulditoxin's reversible neuromuscular blockade is presently unknown.
Homology modelling and overlay of their inter-subunit interfaces revealed side-chain differences between the non-conserved residues of the β2 and β4 subunits (Cuny, Kompella, Tae, Yu, & Adams, 2016).
These differences may account for fulditoxin's weak interaction with hα4β4 and hα3β4 subtypes in contrast to β2-containing nAChRs.
Interestingly, fulditoxin shares only~20% identity (including the eight conserved cysteines) with other short-chain α-3FNTxs, as well as with other classes of snake α-3FNTxs, emphasizing its structural and functional distinctiveness.
Differences in assay methods notwithstanding, the affinity of fulditoxin for human muscle nAChRs was~100 times lower compared to avian muscle nAChRs. Mammals are not the usual prey for Micrurus spp. which feeds primarily on reptiles including other snakes and lizards; and coral snakes themselves are prey for predatory birds (Jackson & Franz, 1981). Therefore, fulditoxin is likely to be optimized to target its natural prey (offence) and predators (defence) (Margres, Aronow, Loyacano, & Rokyta, 2013).

| Fulditoxin displays structural plasticity in its three-finger fold
The highly conserved "three-finger" scaffold is held together by eight conserved cysteines forming four disulfide linkages at its hydrophobic core, as well as other key AA residues such as Tyr25, Gly40, Pro44, Pro48, Arg39, and Glu58, integral to its structural stability (Kini & Doley, 2010;Ricciardi et al., 2000;Tsetlin, 2015). Fulditoxin, however, retains just the eight conserved cysteines, lacking all the other structurally invariant AA residues, suggesting that the four disulfide bonds are sufficient to retain its three-finger structure. Furthermore, structural plasticity in the loop regions of 3FTxs permits the adoption of a variety of functional conformations in order to interact with different pharmacological targets (Ricciardi et al., 2000). In this context, fulditoxin has a longer loop III (Figure 2c) and shows conformational differences at the tip of loop II (Figure 5e), suggesting that these fundamental structural differences could have important functional implications.

| Fulditoxin forms a distinct dimer primarily through hydrophobic interactions
The elution on size-exclusion chromatography (Figure 1a), dynamic light scattering ( Figure S7C), and crystal structure (Figure 5a-c) of fulditoxin indicated that it exists as a non-covalent homodimer of two short-chain α-3FNTx monomers. This dimer formation is distinct from other non-covalent α-3FNTx homodimers, such as κ-neurotoxins and haditoxin composed of long-chain and short-chain α-3FNTx monomers respectively. κ-Neurotoxins and haditoxin share a similar quaternary structure, with their two monomeric units held in an antiparallel orientation through extensive hydrogen bonding between their loop III β-strands (Figure 6; Dewan, Grant, & Sacchettini, 1994;Roy et al., 2010). In contrast, the fulditoxin dimer is held together primarily by 29 hydrophobic interactions between AA residues from loop II; and only three side chain-side chain hydrogen bonding contacts.
A BLAST search revealed that the AA residues involved in dimerization of fulditoxin are conserved among other Micrurus 3FTxs (Figure 2a). Hence, we propose that these Micrurus 3FTxs are also likely to form non-covalent dimers, and fulditoxin is the first member of this new subfamily of dimeric α-3FNTxs.

| Effects of dimerization on the pharmacology of fulditoxin
Typical monomeric short-chain α-3FNTxs like EbTx-a and EbTx-b have been shown to be selective and potent blockers of muscle nAChRs while being ineffective at neuronal nAChRs (Kessler, Marchot, Silva, & Servent, 2017;Servent et al., 1997). In contrast, haditoxin, a homodimer of short-chain α-3FNTxs, inhibits both muscle α1β1γδ as well as neuronal α7, α3β2, and α4β2 nAChRs (Roy et al., 2010). Fulditoxin, also a homodimer of short-chain α-3FNTxs, albeit with a significantly different quaternary structure, exhibits nanomolar and micromolar affinity for chick and mammalian muscle nAChRs, respectively, in addition to inhibiting several neuronal nAChR subtypes. Thus, dimerization appears to expand the selectivity of shortchain α-3FNTxs for nAChRs.

| Fulditoxin is the first metal-binding 3FTx
Four dimeric fulditoxin molecules coalesced and formed a tetrameric complex around two Zn 2+ atoms, with each Zn 2+ atom coordinated F I G U R E 6 Structures of dimeric three-finger neurotoxins. (a) Non-covalently linked homodimers -bungarotoxin (Bungarus multicinctus; Elapidae; 1KBA), haditoxin (Opiophagus hannah; Elapidae; 3HH7), and fulditoxin (Micrurus fulvius; Elapidae). (b) Covalently linked heterodimer irditoxin (Boiga irregularis; Colubridae; 2H7Z) and covalently linked homodimeric α-cobratoxin (Naja kaouthia; Elapidae; 4AEA). The figures were generated using the respective PDB structures (accession number indicated in parenthesis) of each toxin. The species and family of the source are indicated. Disulfide bonds are shown in yellow with four His29 residues at the tip of loop II ( Figure S7B). Metal-binding proteins in snake venom are predominantly enzymes whose catalytic activity involves zinc. Metals can also stabilize the conformation of proteins and contribute to their function (Moroz et al., 2009).
Fulditoxin is the first 3FTx to show metal-binding capability. Interestingly, His29 is conserved in some 3FTxs from M. fulvius (Figure 2a), and a histidine is also present at the tip of loop II of Ω-neurotoxins (Hassan-Puttaswamy, Adams, & Kini, 2015), suggesting that these 3FTXs may also undergo Zn 2+ -coordinated tetramerization.
Analytical size-exclusion chromatography of fulditoxin at pharmacologically relevant concentrations in the presence of Zn 2+ showed that it existed in the native dimeric species, suggesting that zincinduced oligomerization is likely to be concentration-dependent and may occur only at fulditoxin concentrations above 5 mgÁml −1 , which is functionally inconsequential since fulditoxin produced neuromuscular blockade at nanomolar concentrations (IC 50 values about 10 nM; 69 ngÁml -1 ).

| Fulditoxin constitutes a new class of nAChRtargeting neurotoxins
Altogether, fulditoxin is structurally and functionally distinct from other nAChR-targeting snake neurotoxins. We propose that fulditoxin Snake 3FTxs utilize a common protein scaffold to target a variety of molecular targets and exhibit diverse biological effects, underpinning functional evolutionary divergence (Kini, 2011;Kini & Doley, 2010). Interestingly, structurally distinct protein scaffolds can also exhibit functional similarity through convergent evolution (Tsetlin, 2015), including animal toxins such as monomeric/dimeric α-3FNTxs, waglerins, PLA 2 , azemiopsin, Ω-neurotoxins, and α-conotoxins, which have significantly diverse structures but bind to the ACh-binding pocket of nAChRs (Kini, 2011;Kini & Doley, 2010). Hence, Σ-neurotoxins represented by fulditoxin, while retaining the protein scaffold of α-3FNTxs, use a yet-to-be determined pharmacophore to bind to the ACh-binding pocket of the nAChR in another example of evolutionary functional convergence. Σ-neurotoxins offer new tools for studying nAChR subtypes, as well as greater insight into interactions between peptide antagonists and nAChRs.
In conclusion, fulditoxin is a novel homodimeric short-chain α-3FNTx held together primarily by hydrophobic interactions and able to form a tetrameric complex by binding Zn 2+ , the first report of metal-binding capability for the 3FTx family. Fulditoxin exhibited broad spectrum inhibition of α1β1δε, α4β2, α7, and α3β2 nAChRs and produced postsynaptic neuromuscular blockade at nanomolar concentrations, which was completely reversible in contrast to typical α-3FNTxs. Furthermore, fulditoxin competitively antagonized muscle nAChRs at the ACh-binding site despite lacking all canonical α-3FNTx functional AA residues, suggesting that this toxin used a novel and T A B L E 2 Selected dimeric snake three-finger neurotoxins that interact with nicotinic ACh receptors Long-chain α-3FNTx (or long-chain α-3FNTx and cytotoxin) Torpedo (α1β1δγ) (IC 50 = 10 nM) α7 (IC 50 = 0.2 μM) α3β2 (IC 50 = 0.15 μM)

Monocellate cobra
Naja kaouthia (Elapidae) Osipov et al., 2012 different pharmacophore for its interaction with nAChRs. This is an example of unusual functional convergence in evolution, in contrast to the characteristic functional divergence seen within the 3FTx family.