Conformational variability: implications for biomolecular activity and in drug design

Abstract

[eng] Formation of a protein-ligand complex is opposed by the loss of rotational, translational and conformational degrees of freedom, but if the ligand binds in a conformation that does not corresponds to the global minimum, an energetic penalty should also be expected. Although it is well known that the bioactive conformation rarely corresponds to the global minimum identified with current small-molecule force fields, there is no consensus yet about how frequently this occurs in reality or about the maximal conformational penalty that a ligand can attain. Here we investigate this aspect of molecular recognition, using a diverse set of 92 drug-like ligands from the PDB. The global minimum is obtained minimizing a diverse set of conformations for each ligand, while the bioactive conformation results from a restrained minimization of the crystallographic structure. Optimization at B3LYP/6-31D(d) level of theory in the presence of a reaction field (PCM solvation), followed by single point RI-MP2//aug-cc-pVDZ provides average penalties of 2.5 kcal/mol, and only 15% of structures have energies above 5 kcal/mol. Using these values as reference, common force fields and simpler solvation models are evaluated. Depending on the choice of force field, the corresponding values range between 10 and 5 kcal/mol and between 85% and 40%, respectively. Interestingly, an inverse relationship is found between the level of theory and the average internal energy. Contrary to previous reports, we find that the poor performance obtained with molecular mechanics force-fields is more frequently due to inaccuracies in the free energy of solvation provided by the Generalized Born solvation method than to the in vacuum internal energy provided by the force field. This study provides the most accurate estimate of the conformational penalties of bioactive conformations to date, as well as a benchmark to assess the accuracy of theoretical formalisms. On the other hand, to understand at the molecular level how phosphorylation acts as a regulator of HDAC8 activity, we studied molecular dynamics of the same with and without phosphorylation on Ser39. Contrary to the initial expectations, introduction of the negatively charges phosphate does not result in a major structural rearrangement on the local environment. However, we observed that a loop distantly placed from the phospho-acceptor Ser39 in the 3D space of HDAC8 opens and de-associates itself from the deacetylase core in Wt within the simulation time span of 80ns, but this change does not occur when the protein is phosphorylated. The loop is held in place by the interactions between two conserved residues a pi-ring of Phe207 and the hydrophobic side chain of Lys202 (hence we call it KF-loop) and remains closed until 500ns. From this perspective we postulated that the decreased deacetylase activity obtained in the phosphorylated form of HDAC8 could be due to the lack of structural flexibility in this loop-region, which is involved in substrate recognition and binding. Additionally, the pathway involved in the allosteric transmission of the structural perturbation from the phosphorylation site to the KF-loop was investigated using the statistical Kullback-Leibler divergence methodology. To further test the relationship between enzymatic activity and mobility of the KF-loop, as well as to validate the putative allosteric pathway, several point mutations were designed and investigated both computationally and experimentally.