The crystal structure of human glutathione reductase has been established at 1.54 Å resolution using a restrained least-squares refinement method. Based on 77,690 independent reflections of better than 10 Å resolution, a final R-factor of 18.6% was obtained with a model obeying standard geometry within 0.025 Å in bond lengths and 2.4 ° in bond angles. The final 2 F o — F o electron density map allows for the distinction of carbon, nitrogen and oxygen atoms with temperature factors below about 25 Å 2. Apart from 461 amino acid residues and the prosthetic group FAD, the model contains 524 solvent molecules, about 118 of which can be considered an integral part of the enzyme. The largest solvent cluster is at the dimer interface and contains 104 interconnected solvent molecules, part of which are organized in a warped sheet-like structure. The main-chain dihedral angles are well-concentrated in the allowed regions of the Ramachandran plot. The spread of dihedral angles in β-pleated sheets is much larger than in α-helices and especially in α-helix cores, indicating the higher plasticity of β-structures. The analysis revealed a large amount of 3 10-helix. The side-chain conformations cluster at the staggered positions, and show well-defined preferences. Also, a mobility gradient is observed for side-chains. Non-polar and polar side-chains show average temperature factor increases per bond of 10% and 25%, respectively. A number of alternative conformations of internal side-chains, in particular serines and methionines, have been detected. The extended FAD molecule also shows a mobility gradient between the very rigid flavin ( 〈B〉 = 8.7 A ̊ 2 ) and the more mobile adenine ( 〈B〉 = 16.2 A ̊ 2 ). The entire active center is particularly well ordered, with temperature factors around 10 Å 2. The dimer interface consists of a rigid contact area, which is well conserved in the Escherichia coli enzyme, and a flexible area that is not. Altogether, the buried surfaces at the crystal contacts are half as large as at the dimer interface, but less specific. The refined structure shows clearly that there are no buried cations compensating the charge of the pyrophosphate moiety of FAD. The flavin deviates slightly from standard geometry, which is possibly caused by the polypeptide environment. In contrast to an earlier interpretation, atom N5 of the flavin can accommodate a proton, and it is conceivable that this proton proceeds to the redox-active disulfide. Cys58 of this disulfide may be activated in a manner similar to the “charge relay system” of serine proteases.