In the recently determined crystal structure of human thioredoxin (hTRX), the protein is a dimer, covalently linked via an intermolecular disulfide bond involving Cys73 from each monomer (Weichsel et al., 1996). The active site loop, comprising Trp31-Cys-Gly-Pro-Cys35, forms part of the dimer interface and is blocked. Surprisingly, even a mutant protein in which Cys73 is replaced by Ser, is dimeric in the crystal, which led to the suggestion that the dimer of hTRX may be of physiological importance (Weichsel et al., 1996). In addition, the role of Asp60 in dimer formation was probed both crystallographically and with a dimerization assay using diamide as the oxidant (Andersen et al., 1997). Since no evidence for the presence of a dimer was found in the solution NMR structure of hTRX (Qin et al., 1994), it was suggested that the threonine for methionine substitution at position 74 in the hTRX used for the NMR studies might be responsible for this difference between crystal and solution states (Weichsel et al., 1996). To resolve this issue, we have examined the molecular weight and rotational correlation time of hTRX by analytical ultracentrifugation and NMR spectroscopy, respectively. Two variants were investigated, namely hTRX identical in amino acid sequence to the one whose NMR structure we previously determined (C62A, C69A, C73A, M74T) and hTRX (C62A, C69A, C73A, M74) containing the wild-type amino acid methionine at position 74.