Nuclear hormone receptors studied by molecular dynamics computer simulations

Abstract

Nuclear hormone receptors form a super family of homologous transcription factors. The liganddependent nuclear receptors are important drug targets controlling metabolic and inflammatory disorders, amongst others. The C-terminal ligand binding domain (LBD) which binds small molecule ligands and recruit transcription cofactors, is the main focus for drug design efforts targeting nuclear receptors. high-affinity analogues to the natural hormone ligands are developed by medicinal chemistry, and theoretical modeling of their structural effects on the LBD is an integral part of the drug design process.

Two papers in this thesis concern glucocorticoid receptor (GR) point mutations which directly and indirectly influence the affinity for coactivator peptides by structural modifications. Ligands may cause similar modifications to the receptor surface, through selective stabilization of distinct receptor conformations, and thereby alter coactivator binding affinity. We show that the hGR V571M mutation increase the width of the coactivator binding pocket. In accordance with crystal structure data, where the binding pocket is wider when coactivator peptide is present, our results provide an explanation for the increased biological activity observed for this mutant. In contrast, the hGR V575M mutation is modeled to decrease coactivator affinity through unfavorable steric contacts with a coactivator peptide, which may explain budesonide resistance found in human cell fines carrying this mutation.

Influenced by the conformational selection model for molecular interaction, we have explored possible ligand unbinding pathways in retinoic acid receptor (RAR) using random expulsion molecular dynamics (REMD), and found that an agonist ligand can exit the "agonist conformation" of RAR where helix 12 covers the ligand binding pocket. Immediate stabilization of the closed conformation would shift the equilibrium towards the active state, and more quickly lead to coactivator recruitment and gene transcription than binding to an inactive state that has to undergo isomerization.

The first paper in the present thesis describes efforts to improve parameters for the modeling of hydrophobic interactions. Inclusion of solvent effects are crucial in molecular modeling, and water may be explicitly modeled by addition of water molecules, or implicitly modeled as a dielectric continuum. In the latter context it is shown how the precision of free energy perturbation (FEP) simulations of linear and cyclic alkanes can be improved by proper scaling of atomic masses. Accurate FEP simulations can be used for derivation of more accurate parameters for hydrophobic compounds in implicit solvation models.