Electron Paramagnetic Resonance Studies on the Mechanism of Activation and the Catalytic Cycle of the Nickel-containing Hydrogenase from Desulfovibrio gigas*

Desulfovibrio gigas hydrogenase (EC 1.12.2.1) is a complex enzyme containing one nickel, one 3Fe, and two [Fe4S4] clusters (Teixeira, M., Moura, I., Xavier, A. V., Der Vartanian, D. V., LeGall, J., Peck, H. D., Jr., Huynh, B. H., and Moura, J. J. G. (1983) Eur. J. Biochem. 130, 481-484). This hydrogenase belongs to a class of enzymes that are inactive "as isolated" (the so-called "oxygen-stable hydrogenases") and must go through an activation process in order to express full activity. The state of characterization of the active centers of the enzyme as isolated prompted us to do a detailed analysis of the redox patterns, activation profile, and catalytic redox cycle of the enzyme in the presence of either the natural substrate (H2) or chemical reductants. The effect of natural cofactors, as cytochrome C3, was also studied. Special focus was given to the intermediate redox species generated during the catalytic cycle of the enzyme and to the midpoint redox potentials associated. The available information is discussed in terms of a "working hypothesis" for the mechanism of the [NiFe] hydrogenases from sulfate reducing organisms in the context of activation process and catalytic cycle.

The D. gigas [NiFe] hydrogenase has a molecular mass of 89 kDa with two subunits of molecular mass 63 and 26 kDa, respectively. It contains approximately 1 g atom of nickel, 11 g atoms of iron, and 11-12 g atoms of sulfide/molecule of 89 kDa. Mossbauer and EPR spectroscopic studies established that in the purified enzyme the iron-sulfur clusters are organized into a [FeSSxIox cluster (EPR active) and two [FerS4I2+ clusters ("EPR silent").' The [Fe3Sxlox cluster is the origin of an almost isotropic EPR signal centered around g = 2.02 observable below 30 K. A rhombic EPR signal with g values at 2.31, 2.23, and 2.02 (Ni-signal A) was also observed (3). It was definitively assigned to nickel(II1) and accounts for 50-100% of the chemically detectable nickel, depending on preparation. This assignment was confirmed by the observation of hyperfine coupling in 61Ni isotopic labeled hydrogenase and by comparison with model nickel compounds (6,lO). A minor species can also be detected at g values 2.33, 2.16, and around 2.0 (Ni-signal B), but its intensity varies with the preparations (3). Redox titrations at pH 8.5 indicate redox transitions at -70 mV measured by the disappearance of the 2.02 signal and at -220 mV measured by the disappearance of the Ni-signal A (4, 5 ) . The midpoint redox potential associated with the disappearance of Ni-signal A was shown to be pH dependent with a slope of -60 mV/pH unit ( 5 ) .
Although x-ray diffraction data of 20-nm resolution has been obtained for the 3Fe cluster present in A. vinelandii ferredoxin I (41), the structure of these clusters remains controversial. X-ray data on A. vinelandii ferredoxin I indicates that the 3Fe cluster is an essentially planar core with Fe-Fe distances of 41 nm. However, EXAFS studies on the 3Fe cluster present in D. gigas (42) and in aconitase (43) indicate a shorter Fe-Fe distance (-27 nm), in closer agreement with those found for [Fe&] cubane-type structures. Also a careful determination of labile sulfide in aconitase indicates that the ratio Fe to S2is 3:4 (43). These results suggest the presence of a [Fe3S4] structure that can be built by removal of one iron from the [Fe4S4] cluster, which is quite attractive since it explains the facile interconversion between these type of centers (43,44). To address the discrepancy between the x-ray, the extended x-ray absorption fine structure, and the Szchemical determinations it was suggested that either two substantially different structures for [Fe3Sx] exist or some of the structural studies are in error (42). So, since formal charge of the core depends on the number of labile sulfide assumed to be present, throughout this text these clusters will be represented as [Fe3Sx], with a subscript denoting the oxidation state described (ox or red).

a943
Further reduction of hydrogenase results in the appearance of a new EPR signal with g values at 2.19, 2.16, and 2.02 (Nisignal C), which subsequently disappears upon a long exposure to hydrogen gas or in the presence of excess dithionite (3, 6, 11). This last signal was assigned to nickel since hyperfine structure is also induced by 61Ni isotope substitution (6).
The oxidation state of the nickel center giving origin to Nisignal C has been tentatively assigned to Ni(III), although the hypothesis that the signal originates from a Ni(1) species has not yet been eliminated (11).
In the presence of a chemical reductant (dithionite) the center associated with the g = 2.02 EPR signal is reduced prior to the disappearance of the Ni-signal A; however, in the presence of hydrogen both redox centers disappear almost simultaneously (4).
Most of the so-called "oxygen-stable" hydrogenases, such as D. gigas hydrogenase, appear to be reversibly inactivated by oxygen. They are fully active under anaerobic conditions after exposure to activators such as Hz, dithionite, reduced viologens, or tetraheme cytochrome c3 (13 kDa). The function of activation was thought to be associated with the removal of oxygen bound to an essential redox center; however, recent reports have shown that the process is more complex (7, 12).
Hydrogen evolution, hydrogen consumption, and hydrogendeuterium exchange experiments carried out on the native preparations, which are catalytically inactive enzymes, always show a lag phase and an activation phase before full expression of activity is achieved. The disappearance of the lag phase, with concomitant increase of specific activity of the enzyme, was observed under proper reducing conditions. However, oxygen scavengers (such as glucose oxidase) could only reduce the lag phase, the activation step still being required.
Berlier et al. (12) and Lissolo et al. (7) have proposed that a two-step process takes place: removal of oxygen bound to a catalytic site followed by reduction of the involved redox centers.
In this communication we describe: 1) the possible intermediate species and redox states occurring during the activation and catalytic cycle of the enzyme; 2) the activation phenomenon which is necessary in order to transform the enzyme into a fully active state; 3) the conditions for the appearance of the transient Ni-signal C and the estimation of the midpoint redox potential associated with the development of this signal; 4) the interaction between the redox centers; 5) the properties of the [Fe&] clusters; 6) the possible role of cytochrome c3 in the redox cycling of the enzyme.
Sufficient information was now accumulated in order to propose a "working hypothesis" for the mechanism of the [NiFe] hydrogenase from the sulfate-reducing bacteria, in the context of activation and catalytic cycles.

MATERIALS AND METHODS
Growth of the Organisms and Purification of Hydrogenase and Cytochrome c3-D. gigas was grown on a medium as described by Le Gall et al. (13). The periplasmic hydrogenase was purified using DEAE-Bio-Gel and hydroxylapatite chromatographic procedures as described in Ref. 3 with the difference that the crude cell extract, obtained by pressure disruption of the cells, was applied to the column instead of the cell washings. Precautions were taken against oxygen by flushing buffers with purified argon and maintaining all fractions under the same atmosphere. The purification method outlined gave a high protein yield, approximately 800 mg of pure enzyme from 4 kg of wet cells. The preparations used showed a single electrophoretic band and a specific activity between 400 and 440 pmol Hz min" mg-' of protein (unit of activity used throughout this text).
EPR Spectroscopy-EPR spectra were recorded on a Bruker ER-200 tt spectrometer equipped with an Oxford Instruments continuous flow cryostat interfaced with a Nicolett 1180 computer where mathematical manipulations were performed. Signal intensities were determined by double integration with base-line corrections. Cu-EDTA (1 mM) or D. gigas ferredoxin (11) were used as reference standards. Concentrations of ferredoxin (11) samples were calculated using the extinction coefficient c m = 16 mM" cm" (19).
Midpoint Redox Potential Determinations-Midpoint redox potentials were measured by poising the enzyme at different redox potentials in the presence of oxidation/reduction mediators, all at 80 PM (20). The potential (platinum calomel electrode) was adjusted by addition of dithionite (0.2 M in Tris-HC1, pH 9.0) or ferricyanide (0.2 M) solutions. The protein concentration in the titration vessel was 80 p~, as estimated by the extinction coefficient. After equilibration at a fixed redox potential a sample was transferred into an EPR tube under argon pressure and immediately frozen at 77 K for further quantification.
Generation of Intermediate Redox States of Hydrogenuse-All experiments were conducted anaerobically in EPR tubes sealed with serum caps. Additions of reductants and gases were performed through metal needles and gas-tight Hamilton syringes. Intermediate redox states of hydrogenase were obtained by incubation under a hydrogen atmosphere for different lengths of time or by addition of substoichiometric or excess amounts of dithionite. The effect of equimolecular amounts of cytochrome c3 in the redox pattern of hydrogenase was studied in different experimental conditions. Sample reoxidations were accomplished by replacing hydrogen atmosphere by argon.

RESULTS AND DISCUSSION
Activity Profile of the Enzyme (Hydrogen Evo1ution)"The profile for the hydrogen evolution activity as well as the correspondent changes of specific activities measured as a function of time are shown in Fig. 1 although the lag phase may vary. The maximal specific activity reported for the enzyme was taken from the linear part of the hydrogen evolution profile and is 440. Both hydrogen evolution and hydrogen uptake (7) require an activation step in order to express full activity.

Ni EPR Signals of Oxidized Hydrogenases-Fig.
2 shows the EPR spectra of different preparations of native D. gigas hydrogenase as well as a comparison with other bacterial [NiFe] hydrogenases. The data were recorded at 77 K, a temperature at which the g = 2.02 signal of the [Fe4Sx] cluster was broadened beyond detection, and only the Ni signals were observable. As purified, D. gigas hydrogenase exhibits a rhombic EPR signal, g = 2.31,2.23, and 2.02 (Ni-signal A, Fig. 2A). The quantitation of Ni-signal A varies from 0.5-1.0 spin/ molecule depending on the preparation. In correlation with the activation profile of the enzyme we notice that in all cases the Ni-signal A is associated with an inactive form of the enzyme. In addition to the Ni-signal A, a minor component with g values at 2.33, 2.16, and 2.02 (Ni-signal B) is observed in some preparations (Fig. 2B). The relative intensity of Nisignals A and B varies from preparation to preparation and can be altered by redox cycling of the enzyme. When D. gigas hydrogenase reduced under Hz is allowed to reoxidize slowly by replacing the hydrogen atmosphere with argon, an increase in the intensity of Ni-signal B relative to the Ni-signal A is 2.31 observed (Fig. 2C, also see below), indicating that there exists different Ni(II1) environments in the anaerobically oxidized enzyme. It should be noted that the nickel(II1) signal intensities (Ni-signals A and B) are recovered. The fact that Nisignal B relative intensity increases upon anaerobic reoxidation and that anaerobic reoxidation results in the drastic decrease of the lag phase suggest that this Ni-signal B represents an environment resulting from the first step of the activation process, namely the removal of oxygen from the enzyme (see below).
In Fig. 2, we have also shown the high temperature EPR spectra of [NiFe] hydrogenases isolated from various bacterial sources (Fig. 2, D-F). It is interesting to notice that native hydrogenases from different species yield different EPR signals. The periplasmic hydrogenase from Desulfouibrio desulfuricans (ATCC 27774) exhibits an EPR signal similar to the Ni-signal B (Fig. 2 0 ) (21) while Methanosarcina barkeri (DSM 800) (22) shows a nickel signal with g values at 2.24, 2.20, and 2.02 and a minor component at 2.30,2.12, and -2.0 (Fig. 2E). Yet another nickel signal with g values at 2.20,2.06, and 2.0 is observed for the periplasmic [NiFe] hydrogenase isolated from Desulfouibrio baculatus (ATCC 9974) (Fig. 2F).' Intermediate Oxidation States Generated by Hydrogenase Reduction- Fig. 3 shows EPR spectra representing a typical sequence of events during the anaerobic reduction of D. gigas hydrogenase with Hz. The spectra were recorded at 77 K. this temperature. The first step of events is the concomitant disappearance of the Ni-signal A, Ni-signal B, and the isotropic g = 2.02 signal of the [Fe&] cluster (as observed at 10 K). A radical-type signal with very low intensity and of unknown origin is observed next (Fig. 3B). The disappearance of this radical-type signal is accompanied by the development of a new rhombic EPR signal with g values at 2.19, 2.16, and 2.02 (Ni-signal C) (Fig. 3C). During the course of a few hours, this Ni-signal C develops through a maximum of intensity which quantitates to 40-60% of the chemically detectable nickel (Fig. 3 0 ) . After a long incubation (36-48 h) under hydrogen, an EPR silent state is obtained when measured at 77 K (Fig. 3E). At low temperature (below 15 K), EPR signals typical of [Fe4S411+ clusters are now observed. This sequence of events can be reproducibly reversed by anaerobically oxidizing the reduced sample under argon and completely repeated by exposing the reoxidized sample to hydrogen.
At redox states of the enzyme such that Ni-signal C has developed, low temperature studies reveal the presence of another EPR active species. Fig. 4 shows the EPR spectra of such a sample recorded at different temperatures. Below 15 K, the shape of the EPR spectra changes drastically, and a new set of signals at g = 2.21, 2.10, and broad components at Experimental conditions of the EPR spectra as follows. Temperature and gain: A , 4 K, 3.2 X lo'; B, 7 K, 5 X lo'; C, 9 K, 6.3 X lo'; D, 11  higher field are clearly discernible in the 4 K spectrum (Fig.  4A). This new set of EPR signals also exhibits power dependence different from that of Ni-signal C. Fig. 5 shows a power study performed at 10 K. The Ni-signal C is readily saturated at low microwave power (Fig. 5A), indicating slow electronic relaxation. The origin of the new set of signals is not yet understood. Since these signals are only observable at low temperature and show fast electronic relaxation, they may originate from an iron-sulfur cluster; however, the observed g values appear to be too high. Another possible explanation is that they originate from a Ni center that is weakly interacting with another paramagnetic center nearby, resulting in the fast relaxation behavior.
The Ni-signal C has been reported previously and was attributed to a nickel species since hyperfine structure was induced by 61Ni isotopic substitution ( 6 ) . Ni-signal C is different from Ni-signal A and Ni-signal B in both g values and in hyperfine coupling constants. These differences have been proposed to reflect a replacement of ligands in the nickel coordination sphere or a different coordination number (11).
It was also proposed that this transient signal might be due to a hydride-bound species by analogy with nickel catalysts involved in hydrogenation processes (23). Because of the midpoint redox potential studies reported below, we favor the later proposal. It is quite tempting to speculate that Ni-signal C may represent a Ni (1)

Studies of Midpoint Redox Potentials Associated with Ni-
Signal C-Titrimetric determinations of the oxidation-reduction potentials associated with the appearance and disappearance of the Ni-signal C were followed by EPR. Since Nisignal C is slow to equilibrate with the electrode in the presence of dye mediators and due to the transient nature of this signal, there is a large scattering of the experimental data. The redox titration was independently determined from three different experiments, and the data is superimposed and plotted in Fig. 6 . Ni-signal C develops at a potential below completely disappears below -400 mV.
These observations place the redox potential for the appearance of Ni-signal C between -300 mV and -350 mV. Hz. Their study indicated that the hydrogenase activation phenomenon is a 1-electron process with a midpoint redox potential at approximately -340 mV. This potential correlates very well with the potential at which Ni-signal C appears, strongly suggesting that this signal may represent an activated state of the enzyme.  Fig. 7 shows EPR spectra of dithionite-reduced hydrogenase. The spectra shown in Fig.   7, A to E, depict a temperature study of a hydrogenase sample reduced with substoichiometric amounts of dithionite. Nisignal C is observed at a high temperature (Fig. 7 E ) , and the g = 2.21 signal appears at a low temperature (Fig. 7 A ) .
The isotropic signal at g = 2.00 may be due to dithionite oxidation products; however, as we have already pointed out, a radical species is observed during the reductive pattern of the enzyme (Figs. 3 and 8).
In addition to the Ni-signal C and the g = 2.  (Fig. 7F).
The spectrum is complex and may have origin in interacting paramagnetic centers. Redox Cycling of Hydrogenase in the Presence of Ferricyto- signal. Spectral gain, 1 X 10'. Temperatures: A, 4 K; B, 6.7 K C, 10 K D, 12 K, and E, 40 K. F, reduced enzyme with a 3-fold excess of dithionite; temperature, 12 K. Other experimental conditions: modulation amplitude, 1 millitesla; microwave power, 2 milliwatts; microwave frequency, 9.41 GHz. chrome c3-The effects of cytochrome c3 on the redox pattern of hydrogenase were examined since cytochrome c3 is the natural electron donor and acceptor for this enzyme (1, 2). Cytochrome c3 contains four low-spin hemes, and the EPR spectrum of the oxidized state is quite complex. The g, region (not shown) exhibits several superimposed signals originating from the nonequivalent hemes ( g z = 3.3 -2.8)) and a large derivative peak is observed around g, = 2.28 (24).
EPR spectra representing the time course of redox cycling of native hydrogenase in the presence of ferricytochrome c3 were recorded at 77 K. In order to visualize the EPR signals originating from nickel, the contribution of cytochrome c3 was subtracted from the spectra, when necessary. The EPR spectrum of the native hydrogenase is shown in Fig. 8A. Ni-signals A and B are observed. A stoichiometric amount of ferricytochrome c3 was mixed into the native hydrogenase sample ( Proteins were mixed under argon in the native state (oxidized). A , native D. gigm hydrogenase; B, complex 1:l cytochrome c3:hydrogenase; C, same as B subtracted from cytochrome ca spectral contribution; D-G, reduction of the protein complex under hydrogen; E, F, and G show no contribution from cytochrome c3, already in the reduced state. Other experimental conditions: temperature, 77 K, microwave power, 2 milliwatts; modulation amplitude, 1 millitesla; microwave frequency, 9.41 GHz, variable gain.

8B), and the protein mixture was incubated under an H z
atmosphere, the reductive process being monitored by EPR (Fig. 8, C-G). Reoxidation of the proteins was achieved by replacing the Hz atmosphere by argon (Fig. 8, H-N). The sequence of events during this redox cycling followed closely the redox pattern reported previously in the absence of cytochrome q. However, the initial reduction of hydrogenase occurred more rapidly. In order to further investigate the effect of ferricytochrome c3 upon the redox cycle of hydrogenase reduced by HP the following experiment was also carried out. Native hydrogenase was reduced by Hz until the Ni-signal C was fully developed. Then an argon-saturated cytochrome c3 solution was added to the sample, and the Hz atmosphere was replaced by argon. The hydrogenase was reoxidized, and the sequence of redox events was similar to that indicated in EPR signal was recovered after the redox cycling of the enzyme.

CONCLUSIONS
A large effort has been put forward in order to understand hydrogen metabolism in sulfate- reducing bacteria (1,2). This includes the study of the mechanism of hydrogen activation, the type and structure of different hydrogenases, as well as their compartmentalization and role in the cellular bioenergetics.
A comprehension of the mechanism of action of hydrogenase can only be achieved by a full characterization of the structure and physicochemical properties of the redox centers as well as their interaction and behavior during the catalytic cycle. A reasonable understanding of the enzyme metal centers in the "as isolated" state has already emerged from spectroscopic studies (3,4).

Nickel Chemistry in the Context of Its Biological Role
One of the major unresolved questions concecning the [NiFe] hydrogenase is the role of the nickel during redox cycling, namely the oxidation states involved, its mode of ligation, as well as its interaction with other metal centers. Two possible schemes may be considered. One scheme involves redox cycling between Ni(II1) and Ni(I1) (hypothesis A), and the other requires the transition from Ni(II1) to Ni(0) (hypothesis B) (see Table I).
Nickel can exist in oxidation states ranging from Ni(0) to Ni(1V). Generally, the Ni(I1) state is favorably found, in agreement with the decrease in stability of higher oxidation states along the first series of transition metals (25, 26). However, Ni(II1) and Ni(1V) states can be stabilized by electronegative ligands and are found to coordinate N, 0, F, and anionic ligands. The lower oxidation states Ni(0) and Ni(1) are also not common, except with electron acceptor ligands (e.g. carboniles, phosphines, and thiolates). Restriction on the commonly available biological ligands favors a scheme involving fewer oxidation states (such as hypothesis A which involves only Ni(I1) and Ni(II1) states) rather than undergoing a redox transition (hypothesis B) from Ni(II1) to Ni(0).
Nickel chemistry indicates that nickel can occur with different coordination numbers 4, 5, and 6, using structures ranging from square planar to tetrahedral, trigonal bipyramidal, square pyramidal, and octahedral geometries. The octahedral coordination is commonly found for Ni(I1) ( S = l). A   (1) oxidation states are rarely found as pentacoordinated complexes, and the coordination number 4 is preferred. Ni(II1) and Ni(1V) oxidation states are frequently found in an octahedral arrangement, Ni(II1) also being found in pentacoordinatedcomplexes. Consequently, the Ni(III)/Ni(II) redox transition offers a wide range of opportunities: rearrangement of ligands through different preferred parameters, possibility for conformational and spin equilibrium, and a capability of altering the number and type of ligands in the vicinity of the center (29). The redox transitions involved are a reflex of (and/or controlled by) the peculiar chemistry of nickel. The very high and very low oxidation states are not stable, and a consequence is the very negative and very positive redox potentials at which model compounds undergo oxidationreduction transitions (Ni(I)/Ni(O) and Ni(III)/Ni(II)) as indicated schematically in Table 11. This table compiles data on nickel compounds thought relevant for this discussion. The examples were chosen in respect to their relevance to biological systems (i.e. macrocycles, peptides, Schiff bases, and dithiolenes). Also, only compounds where the oxidation states +3 and +1 were unequivocally assigned (namely by EPR) were included. Many of the nickel(1) compounds generate EPR radical-type spectra due to the delocalization of the electron density toward the ligand (30).
Nickel was shown to have a primary role in very different biological situations, being a constituent of several enzymes: urease (31), CO dehydrogenase (32)) coenzyme M methyl reductase (33), and some hydrogenases (11). The biological occurrence of nickel includes active sites identified as containing macrocycles (F430) and amino acids using N (urease) and S (hydrogenase) coordinating atoms. So far only the nickel redox transitions occurring in D. gigas hydrogenase have been studied in detail. The span of redox values associated with the nickel EPR active species is narrow and in the limiting region for the Ni(III)/Ni(II) transition (Table 111).
However, modulation of redox potential by the biological ligands involved in metal coordination is known to be a determining parameter for the redox potential (34).
The preliminary extended x-ray absorption fine structure results of the D. gigas hydrogenase ( EPR g values of hydrogenases purified from different bacterial sources and Ni(II1) complexes with S and N containing peptides (Table 111) showed that these complexes mimic the enzyme active center (36). Although these results are obtained by solution chemistry, the EPR spectra of the complexes show a high rhombicity, and an increase in sulfur coordination tends to cause a shift of the observable g values to higher field.

Working Hypothesis
As stated earlier, it is now well established that the so called "oxygen-stable " [NiFe] hydrogenases (e.g. D. gigas hydrogenase) are not fully active in the "as isolated" state. Studies of the hydrogenase activity (7, 12) indicate that the enzyme must go through a lag phase as well as an activation one in order to be able to exIjress full activity. This complex phenomenon seems to involve the removal of oxygen (lag phase) followed by a reduction step (activation phase). This complex process is illustrated in Fig. 1. When hydrogenase is incubated with dithionite-reduced methyl viologen, the amount of Hz evolved by the system is not linear during the first few minutes of the assay. The hydrogen evolution becomes linear with time after 25-30 min under the assay conditions. The exact time lag for the full activity to appear depends on the hydrogenase preparation as well as on the incubation conditions. Taking into consideration the hydrogenase activity studies, the preferred Ni(III)/Ni(II) redox cycling scheme, and the sequence of events observed by EPR, a working hypothesis is proposed for the mechanism of the [NiFe] hydrogenases from the sulfate-reducing bacteria in the context of both an activation and a catalytic scheme (see Diagram 1).
A can be increased drastically through anaerobic reoxidation (Fig. 2). EPR and Mossbauer studies in the enzyme "as isolated" (4) indicate that there is no magnetic interaction between these four redox centers. The active state of the enzyme (Form 3) is EPR silent. It can be attained either from Form 1 through a complex and slow activation process (removal of oxygen followed by a reduction step), or it can be reached directly from Form 2 (without a lag phase). When O2 is admitted a lag phase is required. During this activation process, both the isotropic g   19 EPR signal could be due to a transient Ni(II1) state in a different coordination, resulting from the breaking of the coupling, the g = 2.21 signal being due to the interacting Ni(II1) and [Fe4S4]+' centers bound, respectively, to hydride and proton.) The midpoint redox potential for the development of Nisignal C is below -330 mV, and this value is consistent with a catalytically active species (7). It is worth noticing that the midpoint redox potential (-220 mV) assigned here to the [Fe4S4] cluster interacting with the nickel center is pH dependent and that of the 3Fe cluster (-70 mV) is pH independent.
Forms 5 and 6 represent further reduced forms of the enzyme. The nickel center is reduced into a Ni(I1) state and becomes EPR silent. The proton is released from the [Fe4S4]l+ cluster, and a "g = 1.94" type EPR signal typical of [Fe4S4]'+ clusters is observed. During the catalytic cycle, all three forms of the enzyme (Forms 3,4, and 5) can be present simultaneously resulting in a complex EPR spectrum (e.g. see Fig. 7 A ) . When excess amounts of dithionite are added to the sample, a super-reduced state (Form 6) is attained. In this super-reduced state all metal centers are reduced. Both the Ni(I1) and the [Fe3Sx]d centers are EPR silent while the two [Fe4S4]'+ are EPR active. The fact that two sets of EPR signals are observed may indicate different conformations for these two [Fe4S4] clusters. The complexity of the EPR spectrum may also indicate weak interaction between the two This proposed mechanism offers a framework for interpreting the present available data; however, it must be viewed as tentative and as a guide for future experimental work.