In silico protein model generation

We have chosen S. aureus (Sa) and P. aeruginosa (Pa) as gram-positive and gram-negative bacterial representatives for computational analyses of DnaK protein binding. 3D structures of DnaK proteins, from the aforesaid species, were generated via homology modelling using MODELLER version 9.24⁸¹. DnaK has two conformations, namely, the open or ATP-bound and the closed or ADP-bound conformation⁶⁵. Herein, we focused on the open conformation of DnaK, to identify potential competitive inhibitors of ATP to prevent the proper functioning of DnaK protein.

We have obtained the protein sequences of Sa and Pa DnaK from UniProtKB with accession IDs of {"type":"entrez-protein","attrs":{"text":"Q2FXZ2","term_id":"122539397","term_text":"Q2FXZ2"}}Q2FXZ2 and {"type":"entrez-protein","attrs":{"text":"A6VCL8","term_id":"166918241","term_text":"A6VCL8"}}A6VCL8⁸², respectively. To search for suitable homology modelling templates, we utilized both NCBI BLASTp and the MODELLER in-built build_profile.py^81,83. For Pa DnaK (PaD), the templates were full-length ATP-bound E. coli (Ec) DnaK protein structures (PDB ID: 5NRO, Chain: A, Query Coverage (QC): 94%, Percent Identity (PI): 79.50%, Resolution (R): 3.25 Å; PDB ID: 4JNE, Chain: A, QC: 94%, PI: 78.80%, R: 1.96 Å; and PDB ID: 4B9Q, Chain: A, QC: 94%, PI: 77.96%, R: 2.40 Å). For Sa DnaK (SaD), besides the afore-mentioned Ec DnaK (EcD) models, we selected one additional template, from Geobacillus kaustophilus DnaK protein (PDB ID: 2V7Y, Chain: A), due to the high percentage of sequence identity expected as per the gram-positive character of S. aureus and G. kaustophilus. As this template structure was in the closed conformation and we were only interested in the open conformation, only the Nucleotide Binding Domain (NBD, residues 1 to 350 in template model) which does not differ much in both conformations, were taken into consideration for homology modelling, and the remaining C-terminal residues modelling were guided by the Ec models to shape an open conformation. Therefore, the templates for SaD were (PDB ID: 2V7Y, Chain: A, Template Residues: 1–350, QC: 57%, PI: 83.19%, R: 2.37 Å; PDB ID: 5NRO, Chain: A, QC: 93%, PI: 56.19%, R: 3.25 Å; PDB ID: 4JNE, Chain: A, QC: 92%, PI: 55.54%, R: 1.96 Å; and PDB ID: 4B9Q, Chain: A, QC: 94%, PI: 55.43%, R: 2.40 Å). The template sequences were aligned with the target sequence for homology modelling via the built-in function of MODELLER (Fig. ^S6A). 5 homology models were generated for each protein of SaD and PaD, and the models with the lowest DOPE (discrete optimized protein energy) scores were selected for downstream virtual screening for both. We then validated the SaD and PaD homology models via Swiss-Model Structure Assessment and SAVES v5.0 servers⁸⁴ (Fig. ^S6).