5.7. Modeling and Structural Analyses of scFv 10FG2–Toxin Complexes

In order to establish the causes of cross-neutralization of scFv 10FG2 to the toxins studied in this work (see Figure 2 and Table 3), models of scFv 10FG2 complexed with every one of the six toxins of Figure 2A were assembled based on the scFv RU1-Cn2-scFv LR ternary complex structure [23]. Also, we modeled the structure of the linker amino acid sequence between the two variable domains of the single chain (from G125 to S139), because it was not defined. We submitted this protein sequence to secondary structure prediction to the YASPIN software [31], which predicted that the linker has no defined secondary structure; thus, it was structured as an extended chain.

The six scFv 10FG2–toxin complexes prepared were submitted to molecular dynamics (MD) simulation procedures. The complexes were first prepared with the System Builder application in the Maestro program, in which the complexes were soaked in a 15-Å buffered box of water with 0.15 M of NaCl and adjusted to minimize the volume. Using Viparr utility in the Desmond program, we settled the Charmm22star force field and the space water force field to all of the systems. Every system was then submitted to MD simulation with the Desmond program [32] with the following settings: a MD simulation time of 20 ns; trajectory recording intervals of 20 ps and five ps for energy recordings; an NVT (canonical ensemble class) at a temperature of 300 K; the Nose–Hoover thermostat method; and 100 ps of relaxation time. We use an integration time step of two ps and Coulombic radius cut off of nine Å (default values).

A sample of a thousand frames was extracted from the trajectories generated by MD simulations for the analysis of the contacts in the interphase between the scFV 10FG2 and the different toxins. From this sample, we took one of every 50 frames and submitted them to the PIC (Protein Interactions Calculator) software using default values [33] for their analysis. For the calculation of the binding free energy from MD trajectories, we analyzed the whole sample of frames (1000), and then, we averaged their values. For this purpose, we used a PERL programming language software designed ad hoc to evaluate every one of the sample’s frames, and then submitted them for analysis with the FoldX program [34].