The periplasmic hydrogenase of Desulfovibrio vulgaris (Hildenbourough NCIB 8303) belongs to the category of [Fe] hydrogenase which contains only iron-sulfur clusters as its prosthetic groups. Amino acid analyses were performed on the purified D. vulgaris hydrogenase. The amino acid composition obtained compared very well with the result derived from the nucleotide sequence of the structural gene (Voordouw, G., Brenner, S. (1985) Eur. J. Biochem. 148, 515-520). Detailed EPR reductive titration studies on the D. vulgaris hydrogenase were performed to characterize the metal centers in this hydrogenase. In addition to the three previously observed EPR signals (namely, the "isotropic" 2.02 signal, the rhombic 2.10 signal, and the complex signal of the reduced enzyme), a rhombic signal with resonances at the g-values of 2.06, 1.96, and 1.89 (the rhombic 2.06 signal) was detected when the samples were poised at potentials between 0 and -250 mV (with respect to normal hydrogen electrode). The midpoint redox potentials for each of the four EPR-active species were determined, and the characteristics of each EPR signal are described. Both the rhombic 2.10 and 2.06 signals exhibit spectral properties that are distinct from a ferredoxin-type [4Fe-4S] cluster and are proposed to originate from the same H2-binding center but in two different conformations. The complex signal of the reduced hydrogenase has been shown to represent two spin-spin interacting ferredoxin-type [4Fe-4S]1+ clusters (Grande, H. J., Dunham, W. R., Averill, B., Van Dijk, C., and Sands, R. H. (1983) Eur. J. Biochem. 136, 201-207). The titration data indicated a strong cooperative effect between these two clusters during their reduction. In an effort to accurately estimate the number of iron atoms/molecule of hydrogenase, plasma emission and chemical methods were used to determine the iron contents in the samples; and four different methods, including amino acid analysis, were used for protein determination. The resulting iron stoichiometries were found to be method-dependent and vary over a wide range (+/- 20%). The uncertainties involved in the determination of iron stoichiometry are discussed.