Liu Tiqing, Lovell Timothy, Han Wen-Ge, Noodleman Louis
Department of Molecular Biology, TPC-15, The Scripps Research Institute, La Jolla, California 92037, USA.
Inorg Chem. 2003 Aug 25;42(17):5244-51. doi: 10.1021/ic020640y.
To predict isomer shifts and quadrupole splitting parameters of Fe atoms in the protein active sites of methane monooxygenase and ribonucleotide reductase, a correlation between experimental isomer shifts ranging 0.1-1.5 mm s(-)(1) for Fe atoms in a training set with the corresponding density functional theory (DFT) calculated electron densities at the Fe nuclei in those complexes is established. The geometries of the species in the training set, consisting of synthetic polar monomeric and dimeric iron complexes, are taken from the Cambridge structural database. A comparison of calculated Mössbauer parameters for Fe atoms from complexes in the training set with their corresponding experimental values shows very good agreement (standard deviation of 0.11 mm/s, correlation coefficient of -0.94). However, for the Fe atoms in the active sites of the structurally characterized proteins of methane monooxygenase and ribonucleotide reductase, the calculated Mössbauer parameters deviate more from their experimentally measured values. The high correlation that exists between calculated and observed quadrupole splitting and isomer shift parameters for the synthetic complexes leads us to conclude that the main source of the error arising for the protein active sites is due to the differing degrees of atomic-level resolution for the protein structural data, compared to the synthetic complexes in the training set. Much lower X-ray resolutions associated with the former introduce uncertainty in the accuracy of several bond lengths. This is ultimately reflected in the calculated isomer shifts and quadrupole splitting parameters of the Fe sites in the proteins. For the proteins, the closest correspondence between predicted and observed Mössbauer isomer shifts follows the order MMOH(red), RNR(red), MMOH(ox), and RNR(ox), with average deviations from experiment of 0.17, 0.17, 0.17-0.20, and 0.32 mm/s, but this requires DFT geometry optimization of the iron-oxo dimer complexes.
为预测甲烷单加氧酶和核糖核苷酸还原酶蛋白质活性位点中Fe原子的异构体位移和四极分裂参数，建立了训练集中Fe原子实验异构体位移（范围为0.1 - 1.5 mm s⁻¹）与相应密度泛函理论（DFT）计算的这些配合物中Fe原子核处电子密度之间的相关性。训练集中的物种包括合成极性单体和二聚体铁配合物，其几何结构取自剑桥结构数据库。将训练集中配合物中Fe原子的计算穆斯堡尔参数与其相应实验值进行比较，结果显示出非常好的一致性（标准偏差为0.11 mm/s，相关系数为 -0.94）。然而，对于甲烷单加氧酶和核糖核苷酸还原酶结构表征蛋白质活性位点中的Fe原子，计算得到的穆斯堡尔参数与其实验测量值的偏差更大。合成配合物计算和观测的四极分裂及异构体位移参数之间存在的高度相关性使我们得出结论，蛋白质活性位点产生误差的主要来源是与训练集中的合成配合物相比，蛋白质结构数据在原子水平分辨率上存在差异。与前者相关的X射线分辨率低得多，这在几个键长的准确性上引入了不确定性。这最终反映在蛋白质中Fe位点的计算异构体位移和四极分裂参数上。对于蛋白质，预测和观测的穆斯堡尔异构体位移之间最接近的对应顺序为MMOH(red)、RNR(red)、MMOH(ox)和RNR(ox)，与实验的平均偏差分别为0.17、0.17、0.17 - 0.20和0.32 mm/s，但这需要对铁 - 氧二聚体配合物进行DFT几何优化。
