Protocols for Metallo- and Serine-β-Lactamase\nFree Energy Predictions: Insights from Cross-Class Inhibitors

While relative binding free energy (RBFE) calculations using alchemical\nmethods are routinely carried out for many pharmaceutically relevant\nprotein targets, challenges remain. For example, open-source tools\ndo not support the easy setup and simulation of metalloproteins, particularly\nwhen ligands directly coordinate to the metal site. Here, we evaluate\nthe performance of RBFE methods for KPC-2, a serine-β-lactamase\n(SBL), and two nonbonded metal parameter setups for VIM-2, a metallo-β-lactamase\n(MBL) with two active site zinc ions. We tested two different ways\nof modeling the ligand–zinc interactions. First, a restraint-based\napproach, in which FF14SB zinc parameters are combined with harmonic\nrestraints between the zincs and their coordinating residues. The\nsecond approach uses an upgraded Amber force field (UAFF) for zinc–metalloproteins\nwith adjusted partial charges and nonbonded terms of zinc–coordinating\nresidues. Molecular mechanics (MM) and quantum mechanics/molecular\nmechanics (QM/MM) simulations show that the crystallographically observed\nzinc coordination is not retained in MM simulations with either zinc\nparameter set for a series of known phosphonic acid–based inhibitors\nbound to VIM-2. These phosphonic acid–based inhibitors exhibit\nknown cross-class affinity for SBLs and MBLs and serve as a benchmark\nfor RBFE calculations for VIM-2, after validation with KPC-2. The\nKPC-2 free energy of binding estimates are within expected literature\naccuracies for the ligand series with a mean absolute error of 0.450.280.66 kcal/mol and a Pearson’s correlation\ncoefficient of 0.930.850.98. For VIM-2, the UAFF approach has improved\ncorrelation from 0.55−0.040.88 to 0.780.380.92, compared to the restraint approach. The\npresented strategies for handling ligands coordinating to metal sites\nhighlight that simple metal parameter models can provide some predictive\nfree energy estimates for metalloprotein–ligand systems, but\nleave room for improvement in their ease of use, modeling of coordination\nsites and as a result, their accuracy.