Antigenic and Substrate Preference Differences between Scorpion and Spider Dermonecrotic Toxins, a Comparative Investigation

The Hemiscorpius lepturus scorpion and brown spider Loxosceles intermedia represent a public health problem in Asia and America, respectively. Although distinct, these organisms contain similar toxins responsible for the principal clinical signs of envenomation. To better understand the properties of these toxins, we designed a study to compare recombinant Heminecrolysin (rHNC) and rLiD1, the major phospholipase D toxins of scorpion and spider venom, respectively. Using a competitive ELISA and a hemolytic inhibition test, we come to spot a cross reaction between scorpion and spider venoms along with an epitopic similarity between rHNC and rLiD1 associated with neutralizing antibodies. Results show that the ability of the rHNC to hydrolyze lysophosphatidylcholine (LPC) is equivalent to that of rLiD1 to hydrolyze sphingomyelin and vice-versa. rHNC exclusively catalyze transphosphatidylation of LPC producing cyclic phosphatidic acid (cPA). The in-silico analysis of hydrogen bonds between LPC and toxins provides a possible explanation for the higher transphosphatidylase activity of rHNC. Interestingly, for the first time, we reveal that lysophosphatidic acid (LPA) can be a substrate for both enzymes using cellular and enzymatic assays. The finding of the usage of LPA as a substrate as well as the formation of cPA as an end product could shed more light on the molecular basis of Hemiscorpius lepturus envenomation as well as on loxoscelism.

. BLASTP analysis of the target rLiD1 and rHNC and the selected template structure.

Target Template Identity (%) Similarity (%) Cover (%) Gaps (%)
Max   [4]. b A ERRAT [5] score over 80 displays that only a few residues have an elevated error function (error > 95% confidence limit) when compared to similar experimental structures. ERRAT was calculated by theSAVES server of UCLA-DOE Lab [6]. c Verify 3D [7] result over 80% indicates that the amino acids have compatibility between the 3D model and the amino acid sequence. Verify 3D was calculated by the SAVES server of UCLA-DOE Lab [6]. d QMean [8] scores values close to zero indicate that the geometrical properties (both global and local) are similar to what one would expect from experimental structures of the same size. QMean values were obtained from Swiss model expasy.org [9]. e Z-scoreis used to test if the knowledge-based potentials could recognize a native fold, found in experimental structures, from other alternatives. The Z-score for these structures had to be within the acceptable range of −12 to 12. Z-score was calculated by the ProSA Web server [10]. f R1: Residues in most favored regions; R2: Residues in additional allowed regions; R3: Residues in generously allowed regions; R4: Residues in disallowed regions. Figure S1. Ribbon representation of the rLiD1 (A) and rHNC (B) modeled structures. A) In the rLiD1 structure, the catalytic histidines (H12 and H47) and the three residues (E32, D34, and D93) that coordinate the metal ion (orange sphere), conserved in Class II phospholipase D (PLD), are shown as sticks with carbon atoms in green. The catalytic, flexible, and variable loops are colored yellow, cyan, and salmon, respectively. The two disulfide bridges, uniquely present in Class II members, are highlighted red. B) In the rHNC structure, the catalytic histidines (H12 and H47) and the three residues (E32, D34, and D93) that coordinate the metal ion (orange sphere) are shown as sticks with carbon atoms in yellow. The catalytic, flexible, and variable loops are colored gold, cyan, and salmon, respectively. The two disulfide bridges, uniquely present in Class II members, and one additional disulfide bridge found in the rHNC sequence are highlighted blue. Three dimensional representations were built with the UCSF Chimera [11]. Leu His Figure S2. Electrostatic surface charge distribution from red (−2 kV) to blue (+2kV) of rLiD1(A) and rHNC (B). The binding site is indicated by ellipses. The electrostatic charges were calculated with APBS program in the UCSF Chimera [11].

Binding and Interaction Predictions of SM and LPC with rLiD1 and rHNC
To analyze the possible interactions of SM and LPC when bound to rLiD1 and rHNC, protein-ligand docking assays were performed using AutoDock 4.0 [12]. Molecular docking allows prediction of a potential binding mode and specific interactions between the proteins' residues and the ligand. A region that included the catalytic (residues 44 to 62), variable (residues 166 to 175), and flexible loops (residues 198 to 207), and the Mg 2+ ion, was chosen as the binding site to the assays. The selected region for docking was large enough to accommodate the ligands' long carbon tail.
SM achieved close predicted binding affinity for both rLiD1(G = −11.56 kcal/mol) and rHNC (G = −11.78 kcal/mol) ( Figure S3). The SM polar group found a similar conformation in both proteins, buried into the binding site (residues His12, Glu32, Asp34, His47, Pro50, Cys51, Asp52, Cys53, Asp91, Lys93, Pro134, Tyr135, Asp164, Ser166, Tyr228, Trp230, Thr199, Cys201) ( Figure S3A and C). In rLiD1, the polar group formed hydrogen bond interactions with His47 (one of the catalytic histidines) and Asp52, both in the catalytic loop, and with Lys93 ( Figure S3B). Attractive charge interactions with the Mg 2+ ion, Asp91 (involved in ion coordination) and Lys93 were also predicted. In rHNC, SM formed hydrogen bond interactions with Lys93 and Thr199 ( Figure S3D). Interestingly, the trimethylamine group from SM was pointed to the inside of the pocket in rHNC, while it was more solvent exposed in the pose in rLiD1, allowing the formation of interactions with more residues in rHNC. Therefore, besides the contacts with the Mg 2+ ion, Asp91, and Lys93, also present in rLiD1, interactions with Glu32 (involved in ion coordination), Asp164 and Trp230 were observed in the SM pose of rHNC. The interacting residues Glu32, His47, Asp34, Asp91, Lys93, and Trp230, previously known to affect substrate affinity and ion coordination [13], are conserved in both structures. The aliphatic tail did not find a common conformation between proteins, which might show the flexibility of this group, as expected due to their high solvent exposure.
Similar to SM, predicted binding affinity for LPC was close between rLiD1 (G = −7.20 kcal/mol) and rHNC (G = −7.91 kcal/mol) ( Figure S4A,C). Although the polar group of LPC was inside the binding site, its conformation in the rLiD1 structure was less buried than in rHNC, with the trimethylamine group of the rLiD1-LPC complex once again solvent exposed. This small discrepancy ensured that the rHNC-LPC complex achieved attractive interactions with the Mg 2+ ion, and residues Glu32, Asp91, Lys93, Asp164 and Trp230. While for rLiD1, the trimethylamine group interacted only with the Mg 2+ ion, Asp91, and Lys93. Although the aliphatic tail of LPC is smaller, it was still solvent exposed, without specific interactions with the proteins. Therefore, similarly to the observed SM, this region was not anchored at any defined region of the receptors. . Docking of SM in the rHNC active site produced more attractive contacts than with rLiD1. Attractive charge interactions were achieved with the Mg 2+ ion, Glu32, Asp91, Lys93, Asp164, and Trp230. Carbon hydrogen bond was found with Ser166 and Tyr228. The 3D surface figures were done with UCSF Chimera [11], while 2D maps were obtained with Discovery Studio Visualizer [14]. The LPC polar group achieved hydrogen bond interactions with Lys93. Docking of LPC in the rHNC active site produced more attractive contacts than the same substrate in rLiD1. Attractive charge interactions were formed with the Mg 2+ ion, Glu32, Asp91, Lys93, Asp164 and Trp230. Carbon hydrogen bond was found with Asp52, Ser166, and Tyr228. The 3D surface figures were done with UCSF Chimera [11], while 2D maps were obtained with Discovery Studio Visualizer [14].

Dynamic Behavior of rLiD1 and rHNCStructures in Apo and Bound Simulations
Due to limitations in the docking methodology, which considers a single protein conformation, we expect that differences in the stability of the docking binding modes and interactions may arise as a consequence of the rHNC distinct sequence and residue changes near the binding site when compared to rLiD1. Therefore, we performed MD simulations to consider the flexibility of the proteins. First, we analyzed the RMSD progression against the initial modeled structures during 100 ns MD simulation to see if the proteins experienced major conformational changes. All simulations were stable after 40 ns of simulation time ( Figure S5). Although flexible regions were observed, in solvent-exposed loops ( Figure S6), no major conformation changes were identified from the simulations. However, when root-mean-square fluctuation (RMSF) was analyzed, small differences in residue fluctuations were observed. The increase of flexibility of the catalytic loop and the so-called flexible loop in the unbound rHNC, when compared to unbound rLiD1, might be associated with the number of residue substitutions in both regions ( Figure S7A). There are seven and three substitutions in the catalytic and flexible loops, respectively, in rHNC when compared to rLiD1 (Table S3). On the other hand, the loop that comprises Asp91 and Lys93 was more rigid in rHNC. In this region, there are eight residue substitutions in rHNC (Supporting Info Table S3). This same loop trend could be observed in the bound, SM ( Figure S7B) and LPC ( Figure S7C) simulations.   The difference in fluctuation in the simulations of rLiD1 and rHNC bound to SM, using as reference the unbound rLiD1 simulation (∆RMSF =RMSFrLid1sm − RMSFrLiD1apo and ∆RMSF =RMSFrHNCsm − RMSFrLiD1apo). While the complex rLiD1-SM was close to the unbound simulation, the Asp1 loop of the rHNC-SM complex remained rigid. C) The difference in fluctuation in the simulations of rLiD1 and rHNC bound to LPC, using as reference the unbound rLiD1 simulation (∆RMSF =RMSFrLid1lpc − RMSFrLiD1apo and ∆RMSF =RMSFrHNClpc − RMSFrLiD1apo). While the complex rLiD1-SM was close to the unbound simulation, the Asp1 loop of the rHNC-SM complex remained rigid. LPC increased the flexibility of the catalytic and flexible loops in both proteins. However, once again the Asp91 loop had inverse flexibility, in rLiD1 simulation it was more flexible, whereas in the rHNC simulation it was more rigid. Plots were done with the R program [15].

Dynamic Behavior of Substrates Bound to rLiD1 and rHNC Structures
The substrates diverged significantly from the initial docking position in the bound simulations. SM was flexible in both rLiD1 (average RMSD of 9.8 ± 1.5 Å ) and rHNC(average RMSD of 4.9 ± 1.09 Å ) binding sites ( Figure S8A). This flexibility was due to the mobility of the aliphatic tail, since the polar group equilibrated to a position and remained close to it throughout the simulation ( Figure S8B). However, the SM polar group bound into rLiD1 deviated more from the docking position (average RMSD of 2.8 ± 0.3 Å ) than when bound to rHNC (average RMSD of 1.2 ± 0.3 Å ), due to the accommodation of the trimethylamine group from solvent-exposed in the docking pose to point to the inside of the binding pocket ( Figure S8C). On the other hand, in the rHNC-SM simulation this group remained pointed to the binding site as in the docking mode ( Figure S8D).
Although the aliphatic tail is smaller in LPC, the substrate also had a flexible behavior ( Figure  S9A) when bound to rLiD1 (average RMSD of 10.7 ± 1.9 Å ) and rHNC(average RMSD of 8.4 ± 1.8 Å ). LPC's polar group diverged from the initial docking mode more in rHNC (average RMSD of 2.6 ± 0.5 Å ) than in rLiD1 (average RMSD of 1.7 ± 0.2 Å ) ( Figure S9B). Even with this flexible behavior, the trimethylamine group of the rHNC-LPC remained pointed to the inside of the binding site throughout the simulation as displayed in the initial docking pose ( Figure S9D). In the rLiD1-LPC complex, the initial solvent-exposed trimethylamine group achieved a buried conformation that persisted throughout the simulation ( Figure S9C). These results indicate that, despite an initial difference in solvent-exposure of the trimethylamine group in the docking results, for the four systems analyzed the most stable orientation of this group is pointing towards to binding site, where it is stabilized by multiple interactions. RMSD plot of the polar group of SM. The SM polar group bound to rLiD1 deviated more from the docking position than when bound to rHNC. (C) Accommodation of the trimethylamine group from solvent exposed (green) to buried inside of the binding pocket (salmon). (D) In the rHNC simulation, the trimethylamine group remained pointed to the binding site in the simulation (purple) similar to the docking binding mode (cyan). Plots were done with the R program [15] and three-dimensional representations were built with the UCSF Chimera [11]. Overall, LPC complexed to rLiD1 was more flexible than when bound to rHNC. (B) The RMSD plot of the polar group of LPC showed that the substrate deviated more from its initial position in rHNC than when bound to rHNC. (C) Accommodation of the trimethylamine group from solvent exposed (green) to buried inside of the binding pocket (light orange). (D) In the rHNC simulation, the trimethylamine group remained pointed to the binding site in the simulation (purple) similar to the docking binding mode (cyan). Plots were done with the R program [15]and three-dimensional representations were built with the UCSF Chimera [11]. Figure S10.PairwiseSequence alignment between rLiD1 (UniProtKB/Swiss-Prot: P0CE81.1) and rHNC (UniProtKB/Swiss-Prot: A0A1L4BJ98) using Needleman-Wunsch algorithm. rLiD1 epitopes [16] are highlighted in yellow (similar to rHNC) and blue (different from rHNC). Conserved amino acids between the two sequences are also highlighted: same amino acid (|), amino acids with similar properties (:) and amino acids with weakly similar properties (.).