Probing Allosteric Communication of Bacterial ClpP Peptidase using Dynamic Network Analysis and Machine Learning

Energy-dependent proteolysis plays an essential role in controlling metabolic pathways and the cell cycle. The ClpP peptidase oligomerizes as two stacked heptameric rings, each of which encloses a degradation chamber. To ensure selective degradation, substrate proteins' access to the chamber is controlled by a gating mechanism of its axial pore that involves conformational transition of N-terminal loops of ClpP subunits. Hexameric ring ATPases, such as ClpX and ClpA, which unfold proteins targeted for degradation, axially cap each ClpP ring and allosterically induce gate-opening. To elucidate the gating mechanism, we performed all-atom molecular dynamics simulations and normal mode analysis of both "closed" and "open" states of wild-type (wt) ClpP and several mutants. Analysis of collective variables that distinguish each state of ClpP, reveals that the backbone hydrogen bond formation in the N-terminal loops and changes in solvent accessible surface area of exposed hydrophobic residues are the largest contributors to the substrate translocation to the degradation chamber. Principal components and normal modes highlight key motions and hotspot residues for allostery. Dynamic network analysis reveals stronger coupling between N-terminal loops and ATPase binding sites in the "open" state. Our machine learning approaches, combined with SHapley Additive exPlanations (SHAP) values, identify structural features and their relative importance for characterizing wt ClpP and mutants.