Computational structure-based investigation of factors governing the biological function of protein PAS domains

Abstract

Per-Arnt-Sim (PAS) domains are conserved, widely distributed, sensory and interaction modules in signal transduction proteins, involved in transcriptional regulation and response to xenobiotic and hypoxic stress, among other functions. Their dysregulation has been linked to cancer and cellular senescence. Despite showing low primary sequence conservation and diverse functionality, PAS domains display impressively high structural conservation. However, understanding the molecular mechanisms that drive their biological functions is elusive, and factors governing their regulation by small molecule ligands remain poorly understood. To address some of these knowledge gaps, I employed all-atom molecular dynamics simulations, “druggability” assessment tools and structural bioinformatic approaches in order the investigate the molecular characteristics that govern the intrinsic dynamics and ligand binding properties of a set of PAS domains, present in human proteome: transcription factors aryl hydrocarbon receptor (AHR), hypoxia-inducible factors 1α and 2α (HIF1α, HIF2α), steroid receptor coactivators 1 and 3 (NCOA1, NCOA3), and PAS kinase. In the absence of experimental structures, molecular models of these domains have been developed and validated in the works described in this thesis. One of the central problems in computational studies of the “druggability” of disease-linked PAS domains is the prediction not only of their binding affinity (potency) but also assessing their pharmacological effects (agonism and antagonism). In the scarcity of experimental structures of these proteins, the latter is very challenging. This thesis describes the deconvolution of the factors governing interactions of human AHR with validated agonists and antagonists and the protocol which allows for differentiation between ligands exerting different pharmacological effects. To enable these investigations, the multimeric AHR-HSP90 chaperone complex has been modelled and interaction with ligands and co-chaperones were investigated. The new model of early stages ligand-induced AHR activation (called “latch and twist” model) has been proposed in this work, which combines the most recent structural biology data with thermodynamics (molecular interactions) and kinetics (ligand residence times) considerations

of the events leading to AHR activation. The mechanistic “latch and twist” model presented in this chapter rationalises almost fifty years of studies of AHR and provides a framework for future structure-guided drug development. Finally, the work presented herein explored the architecture and arrangement of PAS domains of human PAS kinase (PASK), which is a conserved metabolic sensor modulating metabolism and fitness. Despite its key role in mastering metabolic regulation, the molecular mechanism of PASK’s activity remains poorly understood, and structural information is scarce. The work presented in this dissertation investigated how many regulatory PAS domains the PASK is likely to have, how those domains may modulate the kinase activity, and how those interactions could be controlled by small molecules. The results identified a plausible mechanism of autoinhibition of the kinase domain, suggesting that all putative PAS domains may be required for the inhibition. These results may serve as a starting point for the development of novel therapeutics modulating PASK at specific