Crystallographic studies of the hydrogenases (Hases) from Desulfovibrio gigas (Dg) and Desulfovibrio vulgaris Miyazaki (DvM) have revealed heterodinuclear nickel-iron active centers in both enzymes. The structures, which represent the as-isolated (unready) Ni-A (S = 1/2) enzyme state, disclose a nonprotein ligand (labeled as X) bridging the two metals. The bridging atom was suggested to be an oxygenic (O2- or OH-) species in Dg Hase and an inorganic sulfide in DvM Hase. To determine the nature and chemical characteristics of the Ni-X-Fe bridging ligand in Dg Hase, we have performed 35 GHz CW O-17 ENDOR measurements on the Ni-A form of the enzyme, exchanged into (H2O)-O-17, on the active Ni-C (S = 1/2) form prepared by H-2-reduction of Ni-A in (H2O)-O-17, and also on Ni-A formed by reoxidation of Ni-C in (H2O)-O-17. In the native state of the protein (Ni-A), the bridging ligand does not exchange with the (H2O)-O-17 solvent. However, after a reduction/reoxidation cycle (Ni-A up arrow Ni-C --> Ni-A), an O-17 label is introduced at the active site, as seen by ENDOR, Detailed analysis of a 2-D field-frequency plot of ENDOR spectra taken across the EPR envelope of Ni-A(O-17) shows that the incorporated O-17 has a roughly axial hyperfine tensor, A(O-17) approximate to [5, 7, 20] MHz, discloses its orientation relative to the g tensor, and also yields an estimate of the quadrupole tensor, The substantial isotropic component (a(iso)(O-17) approximate to 11 MHz) of the hyperfine interaction indicates that a solvent-derived O-17 is indeed a ligand to Ni and thus that the bridging ligand X in the Ni-A state of Dg Hase is indeed an oxygenic (O2- or OH-) species; comparison with earlier EPR results by others indicates that the same holds for Ni-B. The small Fe-57 hyperfine coupling seen previously for Ni-A (A(Fe-57) approximate to 0.9 MHz) is now shown to persist in Ni-C, A(Fe-57) similar to 0.8 MHz. However, the O-17 signal is lost upon reductive activation to the Ni-C state; reoxidation to Ni-A leads to the reappearance of the signal. Consideration of the electronic structure of the EPR-active states of the dinuclear center leads us to suggest that the oxygenic bridge in NI-A(B) is lost in Ni-C and is re-formed from solvent upon reoxidation to Ni-A. This implies that the reductive activation to Ni-C opens Ni/Fe coordination sites which may play a central role in the enzyme's activity.