Selective Inhibitor Design against Thymidylate Synthase of <i>Mycobacterium tuberculosis</i> Using Alchemical Simulations

Thymidylate synthase is an essential enzyme that catalyzes the conversion of deoxyuridine monophosphate (dUMP) to deoxythymidine monophosphate (dTMP). Thymidylate synthase from <i>Mycobacterium tuberculosis</i>(MtbThyX) recognizes the deprotonated substrate dUMP^(d) (ionized at N3, charge = -3) involving the cationic side chain of Arg199, whereas the human analogue (hThyA) selects the natural substrate dUMP (charge = -2) by involving the polar side chain of Asn226 in the binding pocket. Distinctly different protonation states of the substrate and the catalytic pocket architecture make MtbThyX an attractive drug target for combating <i>Mycobacterium tuberculosis</i>. Fluorodeoxyuridylate (FdUMP) is a known inhibitor of thymidylate synthase, which is severely limited by poor selectivity (it is more potent against hThyA relative to MtbThyX). Using FdUMP as a template, we designed three drug-like ligands, L1, L2, and L3, by (1) removing the proton from the Watson-Crick edge and (2) substituting the ketone/hydroxyl group with fluorine and/or a carboxylic moiety. The absence of a proton on the N3 atom of the ligand is intended to ensure selectivity by favoring MtbThyX binding (skipping the N3 ionization requirement) but penalizing hThyA binding (disrupting the interaction with Asn226). Ionization of the carboxyl group in the ligands was expected to increase the affinity in the cationic binding pocket of MtbThyX. Alchemical simulations confirmed that the designed ligands are strongly favored and disfavored relative to the substrate (dUMP) by MtbThyX and hThyA, respectively. In contrast to hThyA, the catalytic pocket of MtbThyX proved to be relatively dry and stabilized the relatively compact conformation of the ligand (which had a noticeable effect on sugar puckering). Favorable protein-ligand electrostatic interactions in the dry MtbThyX pocket strongly favored ligand binding. In contrast, the interaction between the Watson-Crick edge of the ligands and hThyA was compromised, resulting in water exposure. Ligand L2 is particularly advantageous for its highest affinity for MtbThyX and weak affinity for hThyA. The L2:MtbThyX complex is stabilized by a new salt-bridge interaction (COO^- of L2···Arg107 of protein) and a bridging water molecule (between COO^- of L2 and E92 of protein) in the binding pocket. Moreover, our estimated p<i>K</i> _a of +4.5 units of N3 (dUMP) in the MtbThyX catalytic pocket indicated the strong acidic nature of the uracil, corroborating previous experimental and computational claims. These findings provide insights into the protein-ligand binding affinity in atomic detail and a rational approach for inhibitor design against MtbThyX.