Polarizable water networks in ligand-metalloprotein recognition. Impact on the relative complexation energies of Zn-dependent phosphomannose isomerase with D-mannose 6-phosphate surrogates.

Using polarizable molecular mechanics, a recent study [de Courcy et al. J. Am. Chem. Soc., 2010, 132, 3312] has compared the relative energy balances of five competing inhibitors of the FAK kinase. It showed that the inclusion of structural water molecules was indispensable for an ordering consistent with the experimental one. This approach is now extended to compare the binding affinities of four active site ligands to the Type I Zn-metalloenzyme phosphomannose isomerase (PMI) from Candida albicans. The first three ones are the PMI substrate β-D-mannopyranose 6-phosphate (β-M6P) and two isomers, α-D-mannopyranose 6-phosphate (α-M6P) and β-D-glucopyranose 6-phosphate (β-G6P). They have a dianionic 6-phosphate substituent and differ by the relative configuration of the two carbon atoms C1 and C2 of the pyranose ring. The fourth ligand, namely 6-deoxy-6-dicarboxymethyl-β-D-mannopyranose (β-6DCM), is a substrate analogue that has the β-M6P phosphate replaced by the nonhydrolyzable phosphate surrogate malonate. In the energy-minimized structures of all four complexes, one of the ligand hydroxyl groups binds Zn(II) through a water molecule, and the dianionic moiety binds simultaneously to Arg304 and Lys310 at the entrance of the cavity. Comparative energy-balances were performed in which solvation of the complexes and desolvation of PMI and of the ligands are computed using the Langlet-Claverie continuum reaction field procedure. They resulted into a more favorable balance in favor of β-M6P than α-M6P and β-G6P, consistent with the experimental results that show β-M6P to act as a PMI substrate, while α-M6P and β-G6P are inactive or at best weak inhibitors. However, these energy balances indicated the malonate ligand β-6DCM to have a much lesser favorable relative complexation energy than the substrate β-M6P, while it has an experimental 10-fold higher affinity than it on Type I PMI from Saccharomyces cerevisiae. The energy calculations were validated by comparison with parallel ab initio quantum chemistry on model binding sites extracted from the energy-minimized PMI-inhibitor complexes. We sought to improve the models upon including explicit water molecules solvating the dianionic moieties in their ionic bonds with the Arg304 and Lys310 side-chains. Energy-minimization resulted in the formation of three networks of structured waters. The first water of each network binds to one of the three accessible anionic oxygens. The networks extend to PMI residues (Asp17, Glu48, Asp300) remote from the ligand binding site. The final comparative energy balances also took into account ligand desolvation in a box of 64 waters. They now resulted into a large preference in favor of β-6DCM over β-M6P. The means to further augment the present model upon including entropy effects and sampling were discussed. Nevertheless a clear-cut conclusion emerging from this as well as our previous study on FAK kinase is that both polarization and charge-transfer contributions are critical elements of the energy balances.