Molecular determinants of avoidance and inhibition of Pseudomonas aeruginosa MexB efflux pump

Transporters of the resistance-nodulation-cell division (RND) superfamily of proteins are the dominant multidrug efflux power of Gram-negative bacteria. The major RND efflux pump of Pseudomonas aeruginosa is MexAB-OprM, in which the inner membrane transporter MexB is responsible for the recognition and binding of compounds. The high importance of this pump in clinical antibiotic resistance made it a subject of intense investigations and a promising target for the discovery of efflux pump inhibitors. This study is focused on a series of peptidomimetic compounds developed as effective inhibitors of MexAB-OprM. We performed multi-copy molecular dynamics simulations, machine-learning (ML) analyses, and site-directed mutagenesis of MexB to investigate interactions of MexB with representatives of efflux avoiders, substrates, and inhibitors. The analysis of both direct and water-mediated protein-ligand interactions revealed characteristic patterns for each class, highlighting significant differences between them. We found that efflux avoiders poorly interact with the access binding site of MexB, and inhibition engages amino acid residues that are not directly involved in binding and transport of substrates. In agreement, machine-learning models selected different residues predictive of MexB substrates and inhibitors. The differences in interactions were further validated by site-directed mutagenesis. We conclude that the substrate translocation and inhibition pathways of MexB split at the interface (between the main putative binding sites) and at the deep binding pocket and that interactions outside of the hydrophobic patch contribute to the inhibition of MexB. This molecular-level information could help in the rational design of new inhibitors and antibiotics less susceptible to the efflux mechanism. IMPORTANCE Multidrug transporters recognize and expel from cells a broad range of ligands including their own inhibitors. The difference between the substrate translocation and inhibition routes remains unclear. In this study, machine learning and computational and experimental approaches were used to understand dynamics of MexB interactions with its ligands. Our results show that some ligands engage a certain combination of polar and charged residues in MexB binding sites to be effectively expelled into the exit funnel, whereas others engage aromatic and hydrophobic residues that slow down or hinder the next step in the transporter cycle. These findings suggest that all MexB ligands fit into this substrate-inhibitor spectrum depending on their physico-chemical structures and properties.