Comprehensive molecular mechanics model for oxidized type I copper proteins: active site structures, strain energies,...

The ligand field molecular mechanics (LFMM) model has been applied to the oxidized Type 1 copper center. In conjunction with the AMBER94 force field implemented in DommiMOE, the ligand field extension of the molecular operating environment (MOE), LFMM parameters for Cu-N(imidazole), Cu-S(thiolate), Cu-S(thioether), and Cu-O(carbonyl) interactions were developed on the basis of experimental and theoretical data for homoleptic model systems. Subsequent LFMM optimizations of the active site model complex [Cu(imidazole)2(SMe)(SMe2]+ agree with high level quantum results both structurally and energetically. Stable trigonal and tetragonal structures are located with the latter about 1.5 kcal mol-1 lower in energy. Fully optimized unconstrained structures were computed for 24 complete proteins containing T1 centers spanning four-coordinate, plastocyanin-like CuN2SS' and stellacyanin-like CuN2SO sites, plus the five-coordinate CuN2SS'O sites of the azurins. The initial structures were based on PDB coordinates augmented by a 10 A layer of water molecules. Agreement between theory and experiment is well within the experimental uncertainties. Moreover, the LFMM results for plastocyanin (Pc), cucumber basic protein (CBP) and azurin (Az) are at least as good as previously reported QM/MM structures and are achieved several orders of magnitude faster. The LFMM calculations suggest the protein provides an entatic strain of about 10 kcal mol-1. However, when combined with the intrinsic 'plasticity' of d9 Cu(II), different starting protein/solvent configurations can have a significant effect on the final optimized structure. This 'entatic bulging' results in relatively large fluctuations in the calculated metal-ligand bond lengths. For example, simply on the basis of 25 different starting configurations of the solvent molecules, the optimized Cu-S(thiolate) bond lengths in Pc vary by 0.04 A while the Cu-S(thioether) distance spans over 0.3 A. These variations are the same order of magnitude as the differences often quoted to correlate the spectroscopic properties from a set of proteins. Isolated optimizations starting from PDB coordinates (or indeed, the PDB structures themselves) may only accidentally correlate with spectroscopic measurements. The present calculations support the work of Warshel who contends that adequate configurational averaging is necessary to make proper contact with experimental properties measured in solution. The LFMM is both sufficiently accurate and fast to make this practical.