R. Declerck

The effect of the molecular environment and the temperature dependence of hyperfine parameters in first principles calculations in α-d-glucose and β-d-fructose crystal radicals have been investigated. More specifically, we show how static (0 K) cluster in vacuo hyperfine calculations, commonly used today, deviate from more advanced molecular dynamics calculations at the experimental temperature using periodic boundary conditions. From the latter approach, more useful information can be extracted, allowing us to ascertain the validity of proposed molecular models.

A method for the calculation of hyperfine parameters in extended systems under periodic boundary conditions is presented, using the Gaussian and augmented-plane-wave density functional method, and implemented in QUICKSTEP. In order to increase the efficiency in larger systems, a hybrid scheme is proposed, in which an all-electron treatment for the nuclei of interest and a pseudopotential approximation for the remaining atoms in the simulation cell are combined. The method is validated first by comparing the hyperfine parameters for a selection of atoms and small molecules (using a supercell technique) with other theoretical methods and experimental data from literature. As a typical example of a periodic system where our hybrid method can be applied, the hyperfine parameters of the well-characterized R2 L-α-alanine derived radical are evaluated, yielding results in excellent agreement with the available experimental data.

The neutral and anionic semiquinone radicals of the flavin adenine dinucleotide (FAD) cofactor noncovalently bound in glucose oxidase from A. niger are examined with the aid of QM/MM molecular modeling methods, enabling complete inclusion of the protein environment. Recently, the electron paramagnetic resonance (EPR) characteristics, the anisotropic g tensor and all the significant hyperfine couplings, of these flavoprotein radicals were determined at high resolution (J. Phys. Chem. B 2008, 112, 3568). A striking difference between the neutral and anionic radical forms was found to be a shift in the gy principal value. Within the QM/MM framework, geometry optimization and molecular dynamics simulations are combined with EPR property calculations, employing a recent implementation by some of the authors in the CP2K software package. In this way, spectroscopic characteristics are computed on the fly during the MD simulations of the solvated protein structure, mimicking as best as possible the experimental conditions. The general agreement between calculated and experimental EPR properties is satisfactory and on par with those calculated with other codes (Gaussian 03, ORCA). The protonation state of two histidines (His559 and His516) at the catalytic site of this flavoprotein is found to have a remarkable influence on the isotropic hyperfine coupling of one of the methyl groups on the neutral FAD radical, which is consistent with experimental findings in other flavoproteins (J. Biol. Chem. 2007, 282, 4738). Furthermore, the shift in the gy principal values between the neutral and anionic radicals is well reproduced by QM/MM simulations. Incorporation of at least the nearest protein environment of the cofactor radicals proves to be vital for a correct reproduction, indicating that this shift is a global feature of the protein rather than a local one. In addition, QM/MM techniques are used to make a prediction of relative angles between important spectroscopic principal directions, which are not readily determined by conventional EPR experiments. Significantly, the directions of the gx and the gy components of the g-tensor that lie in the plane of the isoalloxazine moiety are rotated by approximately 59° between the neutral and the anionic radicals.