Evidence for nickel and a three-iron center in the hydrogenase of Desulfovibrio desulfuricans.

Hydrogenase from Desulfovibrio desulfuricans (ATCC No. 27774) grown in unenriched and in enriched 61Ni and 57Fe media has been purified to apparent homogeneity. Two fractions of enzymes with hydrogenase activity were separated and were termed hydrogenase I and hydrogenase II. they were shown to have similar molecular weights (77,600 for hydrogenase I and 75,500 for hydrogenase II), to be composed of two polypeptide chains, and to contain Ni and non-heme iron. Because of its higher specific activity (152 versus 97) hydrogenase II was selected for EPR and Mössbauer studies. As isolated, hydrogenase II exhibits an "isotropic" EPR signal at g = 2.02 and a rhombic EPR signal at g = 2.3, 2.2, and 2.0. Isotopic substitution of 61Ni proves that the rhombic signal is due to Ni. Combining the Mössbauer and EPR data, the isotropic g = 2.02 EPR signal was shown to originate from a 3Fe cluster which may have oxygenous or nitrogenous ligands. In addition, the Mössbauer data also revealed two [4Fe-4S]2+ clusters iun each molecule of hydrogenase II. The EPR and Mössbauer data of hydrogenase I were found to be identical to those of hydrogenase II, indicating that both enzymes have common metallic centers.

Hydrogenases have been purified to apparent homogeneity from a number of microbial species and found to exhibit a diversity (1) which is unexpected in view of the simplicity of the reaction involved in the activation of hydrogen. Generally, hydrogenases contain 4 to 12 atoms of non-heme iron/molecule. EPR studies suggest that the types of non-heme iron clusterfs) show considerable variability (1)(2)(3). Most recently, nickel has been shown to be required for the biosynthesis of hydrogenase (1, 4-7) and reported to be a structural compo-* The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. nent of the hydrogenases from Desulfovibrio gigas (2,8) ' and Methanobacterium thermoautotrophicum (5). Both hydrogenases were demonstrated to contain 1 atom of nickel/molecule; however, only the hydrogenase from D. gigas was demonstrated to contain EPR-active non-heme iron. In the oxidized form (ie. as isolated), both enzymes exhibit EPR signals with g values of 2.3, 2.2, and 2.0 (2, 8). Using isotopic substitution of "Ni, the EPR signal of Methanobacterium thermoautotrophicum hydrogenase was proven to have arisen from Ni (9). Based on the EPR spectrum originating from Ni(II1) in the membranes of M. bryantii (10, ll), the EPR signal from D. gigas hydrogenase was also proposed to reflect the presence of Ni(III), but in contrast to the hydrogenase fromM. thermoautotrophicum, reduced D. gigas hydrogenase exhibited Ni signals. Similar g = 2.3 EPR signals have been detected in a high molecular complex containing hydrogenase from M. thermoautotrophicum (12).
In this communication, we describe the purification and characterization of a hydrogenase from cells of Desulfouibrio desulfuricans grown in unenriched and enriched 61Ni and "Fe media. EPR signals with g values at 2.32, 2.21 and 2.01 have been unequivocally shown to be due to Ni, and Mossbauer studies indicate the presence of two [4Fe-4S] clusters plus a 3Fe cluster.

MATERIALS AND METHODS
Growth ofMicroorganisn and Preparation of Extract-D. desulfuricans (ATCC No. 27774) was grown in the medium described by Liu and Peck (13) containing nitrate rather than sulfate as a terminal electron acceptor thus avoiding precipitation of metal sulfides. For the growth of isotopically labeled cells, 40 mg of 6'Ni (enrichment 86.4%) and 400 mg of 57Fe (enrichment 95%) were first dissolved in H,SO, then in HC1, neutralized, and added to 400 ml of media. The "Ni was obtained from Oak Ridge National Laboratory and the "Fe from New England Nuclear. In a typical preparation of the crude extract, 350 g of cells were suspended in 1 liter of 10 m~ Tris-HC1, pH 7.6, and ruptured in a French press at 9, OOO p s i . under a N, atmosphere. The extract was centrifuged at 19, OOO X g for 30 min and then at 180,000 X g for 75 min.
Assays-Hydrogenase activity was determined by the HZ evolution assay (14). Hydrogen was determined by means of a Varian 4600 gas chromotograph (2); iron by the method of Van De Bogart and Beinert (15); and sulfide by the method of Siege1 (16). Nickel was determined by plasma emission spectroscopy using the Jarrell-Ash model 750 Atomcomp. Protein was determined by t,he Bradford method (17) using bovine serum albumin as standard. The Bradford method was chosen because it gave values which were close to the ones obtained from calculations based on the extinction coefficient for D. gigas hydrogenase which has similar chromophore content.
Electrophoresis-Purity of the hydrogenase was established by polyacrylamide disc electrophoresis (18) as well as by sodium dodecyl sulfate-polyacrylamide electrophoresis (19).
Purification of Hydrogenase-For all purification procedures, the temperature was maintained at 5 "C and the pH of the buffers was 7.6 (measured at 5 "C for Tris-HCI). Precautions were taken against oxygen by flushing buffers with purified argon.
First DEAE-Bio-Gel Column-A DEAE-Bio-Gel column (8 X 22 cm) was prepared with 1100 ml of gel and successively washed with 300 ml of 1 M Tris-HCI, 300-400 ml of 10 mM Tris-HC1,300 ml of 0.5 M Tris-HC1 containing 10 mM Na&04, and 400 ml of 10 mM Tris-HCI. After the crude extract was loaded on the column, it was washed with 500 ml of 10 mM Tris-HC1 and the proteins were eluted with two Tris-HCI linear gradients (  The hydrogenase was collected in approximately 10-ml fractions between 2050 and 2800 ml. First Hydroxylapatite Column-A hydroxylapatite column (4.5 X with 300 ml of 1 M potassium phosphate buffer (KPB), 100 ml of 1 30.5 cm) was prepared with 485 ml of gel and washed successively M Tris-HCI, 200 ml of 0.2 M Tris-HC1, 300 ml of 0.5 M Tris-HC1 containing 10 m M Na2SzO4, and 300 ml of 0.2 M Tris-HC1. After the hydrogenase-containing fractions from the first column were applied, the column was washed with 50 ml of 10 m M KPB and the proteins were eluted with two phosphate linear gradients (I liter of 10 m M KPB and 1 liter of 0.15 M KPB; 1 liter of 0.14 M KPB and 1 liter of 0.3 M KPB). The hydrogenase was eluted between 1600 and 2250 ml and concentrated to 90 ml in a Diaflow apparatus using a PM-30 membrane. Second DEAE-Bio-Gel Column.-A DEAE-Bio-Gel column (6 X 21 cm) was prepared with 600 ml of gel and washed as previously described. The hydrogenase from the previous column was diluted with 150 ml of anaerobic water and applied to the column. After washing with 50 ml of 75 n m Tris-HC1, the column was developed with a Tris-HC1 linear gradient (1 liter of 75 mM Tris-HC1 and 1 liter of 0.25 M Tris-HC1). The hydrogenase eluted between 1310 and 1550 ml.
Second Hydroxylapatite Column-A hydroxylapatite column (4.5 X 21 cm) was prepared with 334 ml of gel and washed as described for the first hydroxylapatite column. After the hydrogenase from the previous step was absorbed on the column, it was washed with 50 ml of 10 mM KPB and the column was developed with a linear phosphate gradient (1 liter of 0.15 M KPB) at a rate of 40 ml/h. The hydrogenase was eluted in two bands: hydrogenase I, between 1270 and 1390 ml and hydrogenase 11, between 1710 and 1880 ml.

RESULTS AND DISCUSSION
The purification of the hydrogenases from D. desulfuricans is summarized in Table I. The overall recovery of hydrogenase activity was 13% and was divided about equally between two fractions containing hydrogenase activity, termed hydrogenase I and hydrogenase 11. Hydrogenase I1 was found to have higher specific activity than hydrogenase I. Hydrogenase I1 was demonstrated to have a molecular weight of 75,500 by ultracentrifugation and was judged to be homogeneous by polyacrylamide disc electrophoresis. On the basis of sodium dodecyl sulfate-polyacrylamide electrophoresis, both hydrogenases were demonstrated to be composed of two polypeptide chains. Hydrogenase I1 contains 10.9 atoms of iron, 11.4 atoms of acid-labile sulfur and 0.6 atom of Ni per molecule. Hydrogenase I has a molecular weight of 77,600 by ultracentrifugation and appeared to be homogeneous by polyacrylamide disc electrophoresis. In addition to chromatographic behavior and specific activity, hydrogenase I differs from hydrogenase I1 in containing 7.8 atoms of iron, 6.8 atoms of acid-labile sulfur, and 0.6 atom of Ni per molecule. Since hydrogenase I1 exhibited the highest specific activity, it was subjected to detailed EPR and Mossbauer studies discussed below. The EPR and Mossbauer spectra of hydrogenase I are similar to those of hydrogenase 11, indicating similar prosthetic groups in both enzymes. The observed differences in activity and metal content may suggest the presence of apoprotein in hydrogenase I.
The EPR spectrum of purified hydrogenase from D. desulfuricans exhibits an "isotropic" signal at g = 2.02 and a rhombic signal with g values at 2.32, 2.21, and 2.01. These values are slightly different but similar to those recorded for the purified hydrogenase from D. gigas (2). At low temperatures (2' < 20 K), the EPR spectrum of hydrogenase I1 is dominated by an intense isotropic signal at g = 2.02. Additional rhombic type signals are also observed. At temperatures higher than 77 K, the g = 2.02 signal becomes very broad and is no longer observable. Only the rhombic type signals remain. The high temperature EPR spectrum is relatively complex.
Resonances are observed at g = 2.32, 2.21, 2.16, and 2.01 regions, indicating the existence of multiple EPR species. Due to its complexity, a definite assignment for the EPR resonances is difficult. However, comparing our results with the EPR data reported for purified hydrogenases from D. gigus (2), and from M. thermoautotrophicum ( 6 ) , as well as for a Ni(II1) species in oxidized membranes from M. bryantii, we have assigned the observed resonances at g = 2.32, 2.21, and 2.01 to one species and the additional resonance to other species.
In Fig. 1 the EPR spectra of hydrogenase I1 as isolated from cells of D. desulfuricans grown on both unenriched, and "Niand "Fe-enriched media are compared at two temperatures, 11 and 90 K. Fig. 1A shows the spectra recorded at 90 K. The signals at 2.32 and 2.16 show significant broadening for the Ni-enriched hydrogenase when compared to the naturally occurring purified hydrogenase. In the g = 2.01 region, hyperfine structure is observed for the 'lNi-enriched protein. Lancaster (11) reported four distinct hyperfine lines in this region in the M. Bryantii "Ni-enriched system due to the nuclear spin (I = 'Ti) of "Ni. It is interesting to note that the separation between the two outermost lines for the "Ni-enriched hydrogenase is 79 G while that measured from Lancaster's data is 80 G. Thus, the hyperfine patterns for "Ni show both broadening and partial resolution of hyperfine lines indicating unequivocally that the rhombic signal observed in the oxidized state of hydrogenase is due to nickel.
Since the inner two hyperfine lines are not well resolved, and the enzyme is also enriched in 57Fe, the observed hyperfine pattern could be due to the nuclear spin ( I = %) of 57Fe. However, a hyperfine splitting of 80 G observed in an EPR spectrum for an electronic spin ?h system corresponds to an internal field of 800 kG at the "Fe nucleus. The following Mossbauer data (Table 11) completely rule out this possibility. Fig. 1B shows the EPR spectrum of purified hydrogenase recorded a t 11 K in the g = 2.02 region. Hyperfine broadening due to the 57Fe nuclear spin is very definitely observed in the 57Fe-enriched hydrogenase. The half-width signal of the "Fehydrogenase signal is 21 G while that of the "Fe-enriched hydrogenase signal is 32 G, indicating a broadening of 11 G.
There also appears to be hyperfine structure in the "Feenriched hydrogenase spectrum with an apparent splitting constant of approximately 13 G. These values of splitting constants are within the range of those found in iron-sulfur clusters, and agree very well with the magnetic hyperfine coupling constants obtained for a paramagnetic component observed in the 57Fe Mossbauer spectra (Table 11).
Upon reduction by hydrogenase, the isotropic g = 2.02 signal and the rhombic Ni signal disappear and are replaced by an intense "g = 1.94 type reduced [4Fe-4S]" signal and additional weak signals presumably due to nickel. The new Gain (natural abundant "'Fe hydrogenase) = 2.5 X 10' and gain ("Feenriched hydrogenase) = 3.2 X 10".

TABLE I1
Mossbauer parameters for the paramagnetic iron center in purified hydrogenase from D. desulfuricans The quadrupole splittings AEQ and isomer shifts 6 (relative to iron metal at room temperature) are values obtained at 85 K. 1) is the asymmetry parameter. g.p. is the nuclear magnetic moment. With the limited resolution, the A-tensor is assumed to be isotropic for site 2 and is undetermined for site 3. The values in parentheses are equivalent gauss at the electron. These values are consistent with the hvoerfine seDarations observed in the isotrooic e = 2.02 sienal. nickel signal has maximum g value at 2.28 and its intensity is about one-third of the g = 2.32 resonance. These observations will be the subject of a future communication. Fig. 2 shows Mossbauer spectra of "Fe-enriched hydrogenase I1 from D. desulfuricans. The data are recorded at 4.2 K with a magnetic field of 500 G applied parallel (Fig. 2 A ) and perpendicular (Fig. 2B) to the y-radiation. The spectrum -ogenase from D. desulfuricans consists of at least two subspectral components: an intense quadrupole doublet at the center and a paramagnetic component extended from -2 to +3 mm/s. The doublet accounts for approximately 70-80% of the total iron absorption. The observed quadrupole splitting (AEQ = 1.17 mm/s), the isomer shift (6 = 0.43 mm/s), and the general shape of the doublet are typical of a [4Fe-4SI2' cluster (20, 21). These observations, together with the iron quantitation, indicate that hydrogenase 11, as prepared, contains two [4Fe-4S] clusters in the 2+ oxidation state.
A close examination of Fig. 2, A and B, reveals that the absorption pattern of the magnetic component is field direction-dependent. To study the magnetic component in more detail, spectrum 2B is subtracted from spectrum 2A, and the difference spectrum is shown in Fig. 2C. The absorption of the intense doublet is cancelled in the difference spectrum because the [4Fe-4SI2+ cluster is diamagnetic and its spectrum will not be affected by changing direction of the applied field. Two important observations are evident in Fig. 2C. First, the magnetic component is strongly field direction-dependent, a phenomenon which suggests that an intense EPR signal should be associated with the magnetic component. This EPR signal is identified as the isotropic g = 2.02 signal. Second, at least two pairs of Am = 0 nuclear transitions are observed (indicated by the brackets). Since a single iron site can only have one pair of Am = 0 transitions, the above observation demonstrates that at least two iron sites are associated with the magnetic components.
We have analyzed the spectra of the two discernible magnetic iron sites using an S = % spin Hamiltonian, All symbols in Equation 1 have their conventional meanings (22). The solid lines in Fig. 2 are theoretical simulations. The parameters used for calculations are listed in Table 11. By comparing the absorption intensities of the peaks at -3 and -2 mm/s with those of the simulated spectra, the two iron sites are estimated to contribute 16-20% of the total iron absorption.
The EPR studies showed that at temperatures higher than 77 K, the g = 2.02 signal disappeared, indicating fast electronic relaxation rate. Consistent with this EPR observation, the Mossbauer magnetic component collapses into a sharp quadrupole doublet at 85 K. The measured parameters (AEQ = 0.63 mm/s and 6 = 0.36 mm/s) are typical for high spin ferric ion (S = Yz). However, two S = 5/2 irons cannot be coupled to form a total spin of %. A third half-integral spin is required.
Also the absorption intensity of the collapsed quadrupole doublet was found to be 25-30% of the total absorption, which is about % of that of the two discernible magnetic iron sites. We therefore conclude that the g = 2.02 center must contain a third iron site which at low temperature spectrum is masked by the intense central doublet.
Three-iron clusters have recently been found in various proteins (23-25). The physical properties of the g = 2.02 center in hydrogenase are very similar to those of a 3Fe cluster found in Azotobacter vinelandii ferredoxin (23). However, all the oxidized 3Fe clusters reported so far have an isomer shift less than 0.3 mm/s, a typical value for high spin ferric ion in a tetrahedral sulfur environment. In D. desulfuricans hydrogenase a relatively larger isomer shift, 0.36 mm/s, was found, suggesting that the iron coordination of this trinuclear iron cluster may contain nitrogenous or oxygenous ligands.
The hydrogenases isolated from D. gigas (2,26,27), D. desulfuricans, and the hydrogenase of D. vulgaris (28,29) exhibit significant differences in their molecular organization and specific activity. However, evidence collected from chemical analyses, EPR, and Mossbauer studies indicate that all three enzymes contain some common prosthetic groups. All three hydrogenases were reported to contain approximately 11-12 non-heme irons and comparable amounts of acid-labile sulfurs per molecule. Approximately 1 atom of nickel was found in hydrogenase from D. gigas (2) and from D. desulfuricans but there has been no report of Ni in hydrogenase from D. vulgaris. The EPR and Mossbauer spectra of isolated hydrogenases from D. gigas and from D. desulfuricans are practically the same, suggesting similar organization for the iron and sulfur atoms in the enzymes f i x . two [4Fe-4S] clusters and one 3Fe cluster per molecule). As isolated, the hydrogenase from D. vulgaris exhibits a g = 2.02 EPR signal which is lost upon reduction and replaced by a low intensity g = 1.94 type signal (3). These signals are consistent, respectively, with oxidized 3Fe cluster and reduced ferredoxin type [4Fe-4S] cluster.
Recently, it has been shown that [4Fe-4S] clusters can be converted into 3Fe clusters by ferricyanide oxidation (30-32). The activation of beef heart aconitase was shown to involve the transformation of 3Fe clusters into [4Fe-4S] clusters (33). Also, the conversion of 3Fe to [4Fe-4S] clusters was achieved in D. gigas

30.
dures, and it is therefore rather unlikely that the presence of 3Fe clusters in this enzyme is due to oxidation damage of [4Fe-4S] clusters.
Although nickel is not yet established as an essential component of all hydrogenases, it is found in hydrogenases purified in various species. Evidence gathered so far indicates that the nickel center is sensitive to the redox state of the enzyme (2), and it can undergo a one-electron reduction from Ni(II1) to Ni(I1) with pH-dependent midpoint redox potential (2, 8). These results suggest that nickel must play an important role in the activation of hydrogen in these hydrogenases.