Computational protocol: Structural, physicochemical and dynamic features conserved within the aerolysin pore-forming toxin family

Similar protocols

Protocol publication

[…] For the MD of the toxins monomeric active form, the following PDB entries were used, from which only one monomer was taken, and, if needed, the residues known to be cleaved for toxin activation were deleted: aerolysin wt (PDB entry 1PRE) and aerolysin Y221G ((PDB entry 3C0N, deleting residues 440–470 to obtain the active forms); ETX wt (PDB entry 1UYJ) and ETX H162A (PDB entry 3ZJX), deleting residues 250–282 to obtain the active forms); LSL (PDB entry 2Y9F); parasporin-2 (PDB entry 2ZTB); and lysenin (PDB entry 3ZX7). Moreover, as in previous studies by our group, domain 1 of aerolysin was deleted in order to reduce the computation time, as it is known to act as an independent folding unit. If needed, missing side chains were added with the Modeller program version 9.13. All of the proteins were solvated in a rectangular box of TIP3P water and neutralized by NaCl. Simulations were run using the GROMACS software version 4.8,, with the Amber99SB force field, the SHAKE algorithm on all the bonds between hydrogen and heavy atoms, and Particle-Mesh Ewald, treating the electrostatic interactions in periodic boundary conditions. We chose an integration step of 2 fs. The temperature has been controlled by means of Langevin forces, using a damping constant of 1 ps−1. 800 ns of simulation were run for each protein.For the prepore molecular dynamics simulations, we first created a model for the wild-type form using as template the prepore structure of aerolysin Y221G (PDB entry 5JZH), with the Modeller program version 9.13, following a rigid procedure (without Simulated Annealing optimization). The system was then simulated for 180 ns using the GROMACS software version 4.8,, with a similar procedure as that used for the monomeric toxins. The conformation closest to the average of the production run (last 100 ns) was similar to the prepore Y221G structure. This conformation was further used as input to build up two new simulation systems, one of the wt prepore in water, and a second one of the prepore wt in the presence of a lipid bilayer. The topology for both the prepore and the 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine (POPC) membrane were obtained from the CHARMM-GUI web server, using the charmm36 force field. For the second simulation, the protein was manually located at interacting distance to the membrane (minimum distance was 4 Å). Afterwards, the systems were solvated in a rectangular box of TIP3P water and neutralized by NaCl. Simulations were run using the GROMACS software version 4.8,, with the charmm36 force field. For the prepore without membrane, a similar procedure of that used for the monomer toxins was applied, to a final simulation time of 130 ns. For the prepore with membrane, the system was first equilibrated for 150 ns in a NVT ensemble, using a time step of 1 fs, the Berendsen thermostat for temperature coupling, and with both the protein and lipid atoms initially restrained. While the lipid restrains were gradually decreased to zero, the protein was kept fixed. Afterwards, it was simulated in the NPT ensemble, with the Nose-Hoover thermostat for temperature coupling and the Parrinello-Rahman for semi-isotropic pressure coupling, gradually releasing the protein restraints, and then leaving the free systems simulate for 130 ns.For the wt pore simulation, an already published model based of the quasipore structure was used. The protein-membrane system was built with the charmm-gui server, using the membrane positioning as suggested by the PPM server. The system was simulated using a similar procedure of that of the prepore with membrane, but simulating for a longer time of 250 ns. [...] Two methods were here used to explore the main dynamics of the studied proteins: Principal Components Analysis (PCA) and Normal Mode Analysis (NMA). PCA is a statistical method based on covariance analysis that uses an orthogonal transformation to convert a set of correlated variables into a reduced set of uncorrelated variables called principal components (PCs). In this way PCA maps high-dimensional data, MD trajectories in this work, into interesting low-frequency motions concerted over large number of atoms (first PCs, as opposed to higher-order PCs which reflect faster local fluctuations). The PCA procedure is based on the diagonalization of the covariance matrix, C(i,j):1C(i,j)=〈Δri⋅Δrj〉where Δri and Δrj indicate the displacement vectors of atoms i and j, respectively, from their average positions. For the MD trajectories of the monomeric forms, PCA was performed on 800 frames containing backbone coordinates (one frame every 1 ns of simulation) using the g_covar and g_anaeig Gromacs tools,. The two extreme projections along the aerolysin Y221G MD trajectory on the average structure, calculated with the g_anaeig Gromacs tool, were used to compare with the aerolysin structures on the different states (monomer, prepore, post-prepore and quasi-pore). For the MD trajectories of the oligomers (prepore and pore), PCA was performed with the software Prody, using 130 or 250 frames, respectively (one frame every 1 ns of simulation).NMA is another well-established method to infer global protein motions, but in this case computed from a single (static) structure. The main concerted motions predicted by the NMA model make up a set of collective variables (normal modes) that are obtained from the second derivative matrix of the potential energy. In brief, NMA is a simple tool to quickly calculate the expected main global motions from a structure, whereas PCA extracts actual global motions from a MD trajectory. In the ideal scenario, NMA normal modes and PCA principal components should be similar. NMA was used instead of PCA in the lysenin and CPE monomers, as well as for the prepore of HA3 and the lysenin pore, as we did not perform MD on them. For aerolysin, NMA was used together with PCA in the analysis of the prepore and pore dynamics, to compare the main motions predicted for the proteins with the dynamics obtained in MD in presence and absence of a lipid bilayer. NMA of both the monomers and the oligomers were here performed by the Anisotropic Network Model (ANM) with the software Prody, and the results were visualized with the VMD program. The following structures were used for other aerolysin-like proteins: CPE monomer (PBD entry 2XH6), HA3 prepore (PDB entry 4EN6), lysenin pore (PDB entry 5GAQ). The hinge regions (flexible joints) were calculated by the software Bio3D as the regions of minima of mean square fluctuations for a given mode/PC.Pairwise dynamic cross-correlation coefficients were calculated with the Bio3D software, considering only Cα carbons, and results were plot in The PyMOL Molecular Graphics System, Version 1.8 Schrödinger, LLC with the Bio3D function pymol.dccm. Briefly, the cross-correlation matrices were calculated according to the following equation, giving values in the range from 1 (correlated) to −1 (anti-correlated):2C(i,j)=〈Δri⋅Δrj〉〈Δri2〉1/2⋅〈Δrj2〉1/2The squared overlap (dot product) between two vectors was calculated with the software Prody to measure the alignment between the directions of a pair of given PCs/modes, presenting a value of 1 if they are identical and of 0 is they are orthogonal.3Oi(X)=MiX/‖Mi‖‖X‖This procedure was applied to study the overlap between different sets of NMs, between NMs and PCs, and also between NMs and a vector representing the conformational change between two different structures. […]

Pipeline specifications

Software tools GROMACS, CHARMM, PPM Server, ProDy, VMD, PyMOL
Applications Protein structure analysis, Membrane protein analysis
Chemicals Amino Acids