2. Methodology
To model the all-atom structure of Nap1-nonP, an experimentally determined structure of yNap1 (PDB ID: 2Z2R)13 was solvated with the TIP3P water model14 in a rectangular box. The dimeric yNap1 molecule consists of 417 residues, of which residues 74–365 were used for the modeling of Nap1-nonP. The experimentally undetermined loop regions were modeled with a homology modeling tool (Phyre2),1515 with the N- and C-terminal coils omitted to reduce the system size. Additionally, because each nuclear localization sequence (NLS) region contains a long β-sheet protruding from the protein surface, the size of the periodic boundary box would need to be increased when these regions are modeled explicitly. Therefore, to reduce the system size, we omitted the NLS regions by replacing them with Gly, Gly, and Ser. Figure 1 shows the modeled all-atom structure of Nap1-nonP.
To address the phosphorylation effect on the accessory domain, an all-atom system of Nap1-P was also prepared. To model Nap1-P, Nap1-nonP was phosphorylated using PyTMs (a PyMol plugin) by employing a set of phosphoserine parameters proposed in a previous study.16 The Ser 140, Ser 159, and Ser 177 residues were considered as the set of phosphorylation sites. Finally, the total number of atoms, including the water molecules, was 894,611 for the Nap1-nonP system and 894,527 for the Nap1-P system.
To extend the MD simulation time step to 2.0 fs, we constrained the chemical bonds of the proteins and water molecules with the LINCS1717 and the SETTLE algorithms,18 respectively. The temperature and pressure of the systems were controlled using the modified Berendsen thermostat1919 and the Parrinello–Rahman method,20,21 respectively. The electrostatic interactions were evaluated with the particle mesh Ewald method22 using a real-space cutoff of 10.0 Å. The cutoff value for van der Waals interactions was set to 10.0 Å. All MD simulations were performed with the GPU version of the GROMACS 2019 package,23 using the AMBER 14SBonlysc force field.24 To equilibrate each solvated system, 10 ps of the NVT ensemble (T = 300 K) and 10 ps of the NPTensemble (T = 300 K and P = 1 bar) were performed sequentially after short energy minimization, where the last snapshots of the NPT simulations were employed as the starting structures of the production runs. Finally, five trials of 100-ns MD simulations were started from the relaxed structures of both systems to obtain statistically reliable trajectories.