One component of the larger protein quality control network in many bacteria species is caseinolytic peptidase P (ClpP), a tetradecamer that sequesters fourteen serine catalytic triads within a barrel-shaped quaternary structure. The indiscriminate nature of the catalytic triads to cleave peptide bonds necessitates the regulation of peptide entry into ClpP via two axial pores that are surrounded by pore loops which can adopt either a “closed” or “open” conformation. In a typical scenario, motor proteins such as ClpX dock to grooves between the pores and equator of an assembled ClpP tetradecamer, serving as both a substrate recognition and peptide threading unit to ensure that only unfolded, damaged, or defective proteins are degraded within the lumen of ClpP. Small molecules such as acyldepsipeptide (ADEP) can bind to the same grooves motor proteins, triggering the dramatic conformational change in ClpP and converting the pore loops to the “open” configuration. In this state, ClpP nonselectively degrades any peptide fragments that diffuse into its central chamber, eventually causing cell death. To investigate the structural features that evolve during the pore opening-closing transitions, we utilize molecular dynamics simulations and classification machine learning models. Seven different point mutations that affect the salt bridges, hydrogen bonds, and hydrophobic interactions that underlie the pore loop transitions are compared by computing SHAPley additive explanations (SHAP values), which reveal the impact of each structural feature in the context of every other feature. We find that the solvent-accessible surface area of the pore wall is highly perturbed by the mutations, indicated by the reliance of the classification models on that feature for accurate predictions, along with the presence or absence of intra-protomer salt bridges and native contacts.
Dennison et al. (2026) studied this question.