Comparison of allosteric signaling in DnaK and BiP using mutual information between simulated residue conformations

Abstract

The heat shock protein 70 kDa (Hsp70) chaperone system serves as a critical component of protein quality control across a wide range of prokaryotic and eukaryotic organisms. Divergent evolution and specialization to particular organelles have produced numerous Hsp70 variants which share similarities in structure and general function, but differ substantially in regulatory aspects, including conformational dynamics and activity modulation by cochaperones. The human Hsp70 variant BiP (also known as GRP78 or HSPA5) is of therapeutic interest in the context of cancer, neurodegenerative diseases, and viral infection, including for treatment of the pandemic virus SARS-CoV-2. Due to the complex conformational rearrangements and high sequential variance within the Hsp70 protein family, it is in many cases poorly understood which amino acid mutations are responsible for biochemical differences between protein variants. In this study, we predicted residues associated with conformational regulation of human BiP and Escherichia coli DnaK. Based on protein structure networks obtained from molecular dynamics simulations, we analyzed the shared information between interaction timelines to highlight residue positions with strong conformational coupling to their environment. Our predictions, which focus on the binding processes of the chaperone's substrate and cochaperones, indicate residues filling potential signaling roles specific to either DnaK or BiP. By combining predictions of individual residues into conformationally coupled chains connecting ligand binding sites, we predict a BiP specific secondary signaling pathway associated with substrate binding. Our study sheds light on mechanistic differences in signaling and regulation between Hsp70 variants, which provide insights relevant to therapeutic applications of these proteins.

Dynamic correlations: exact and approximate methods for mutual information

“…Several computational studies have investigated mutual information in proteins using both exact ( McClendon et al 2014 , Cortina and Kasson 2016 , Long and Tian 2016 , Singh and Bowman 2017 , Sogunmez and Akten 2020 , Shao et al 2021 , Schneider and Antes 2023 ), mostly considering the magnitudes of fluctuations, and approximate calculations ( Lange and Grubmüller 2006 , Jana et al 2011 , Rennebaum and Caflisch 2012 , Invernizzi et al 2014 , Scarabelli and Grant 2014 , Köhler et al 2015 , Tiberti et al 2015 , Ahmed and Barakat 2017 , Milenkovic and Bondar 2018 , Zhang et al 2018 , Foster et al 2021 , Payne et al 2022 , Tekpinar and Yildirim 2022 ). By relying on the three approximations, calculations requiring shorter simulations or only static crystal structures may become possible.…”