CO Dehydrogenase from CZostridium thermoaceticum EPR AND ELECTROCHEMICAL STUDIES IN CO* AND ARGON ATMOSPHERES*

The EPR and redox properties of the metal com- plexes in CO dehydrogenase (CODH) from Clostridium thermoaceticum were studied. Controlled potential coulometric reductive titrations of CODH were per- formed under argon and CO, atmospheres. In the titrations performed under argon, five to eight electrons/ dimer were required for reduction, and four distinct EPR signals appeared. These included a signal with 4? aYe = 1.82 (E, = -220 mV), two signals with the same g values but different linewidths at g,,, = I.94 (E,,, = -440 mV), and a signal at g,,, = 1.86 (E, = -530 mV). All of the S = l/2 EPR signals had low spin concentra- tions; values between 0.2 and 0.3 spinsldimer were typically obtained for each signal. Features between g = 6 and 4, typical of S = 312 states, were also observed, and these may account, at least to some degree, for the low spin concentration values.

The EPR and redox properties of the metal complexes in CO dehydrogenase (CODH) from Clostridium thermoaceticum were studied. Controlled potential coulometric reductive titrations of CODH were performed under argon and CO, atmospheres.
In the titrations performed under argon, five to eight electrons/ dimer were required for reduction, and four distinct EPR signals appeared.
These included a signal with 4? aYe = 1.82 (E, = -220 mV), two signals with the same g values but different linewidths at g,,, = I.94 (E,,, = -440 mV), and a signal at g,,, = 1.86 (E, = -530 mV). All of the S = l/2 EPR signals had low spin concentrations; values between 0.2 and 0.3 spinsldimer were typically obtained for each signal. Features between g = 6 and 4, typical of S = 312 states, were also observed, and these may account, at least to some degree, for the low spin concentration values. Under COz, and at negative potentials, CODH served as an electrocatalyst in the reduction of COz to CO. The apparent half-maximal activity for this reduction at pH 6.3 occurred at -430 mV, a potential near the thermodynamic value. An EPR signal, arising from a complex containing Ni, Fe, and the carbon from CO/ COz developed along with this activity. The reduction of this complex is probably the last step to occur prior to the catalysis of CO2 reduction.
Carbon monoxide dehydrogenase (CODH)' from Clostridium ttzermoaceticum catalyzes the following two reactions (l-3): CO  In the first reaction, carbon monoxide is oxidized to carbon dioxide, while in the second, carbon monoxide, a methyl group, and coenzyme A are condensed by CODH to form acetyl-CoA.
The ability of CODH to catalyze the second reaction has only recently been discovered, and this has established the enzyme as a central participant in the synthe- sis of acetate from COZ and HP in acetogenic bacteria (4). We have proposed that CODH might be more appropriately called acetyl-CoA synthase (4). The pathway responsible for this reaction involves at least three additional enzymes, including methyltransferase, a corrinoid/Fe-S protein ([Co]), and disulfide reductase. Recently, it was found that the genes coding for CODH, methyltransferase, and the corrinoid/Fe-S protein are clustered within a IO-kilobase region of DNA (5).
In this pathway, CO, is reduced to 5-methyltetrahydrofolate via a series of reactions involving formate dehydrogenase and tetrahydrofolate-related enzymes (1,6). Methyltransferase catalyzes the transfer of this methyl group to the corrinoid/ Fe-S protein, forming enzyme-bound methylcobalamin (7). The corrinoid-bound methyl group is then transferred to a site, apparently a cysteine residue (8), on CODH. Finally, CODH binds CO and CoASH and acetyl-CoA is formed (4). At some point in the final steps, a disulfide reductase may be involved (9). A low potential electron carrier, such as an eightiron ferredoxin, is required for efficient catalysis of acetyl-CoA synthesis (4, 10).
CODH from the acetogenic bacterium, C. thermoaceticum, has been purified to apparent homogeneity (11). Polyacrylamide gel electrophoresis of the enzyme under denaturing conditions yields two protein bands of equal intensity, corresponding to (Y and fl subunits of M, = 71,000 and 78,000 (4, 11). Gel filtration studies indicate that CODH has a molecular weight of 440,000, suggesting an (a& hexameric protein subunit structure' (11). Pore-limit electrophoresis suggests that the cup dimer may be active as well (12). Metal and sulfide analyses yield about 11 irons, 1.8 nickels, 1 zinc, and 14 acidlabile sulfide ions/dimer (11). Upon prolonged incubation with disulfide reductase, -17 sulfhydryl groups/dimer are formed (9).
The metals are arranged into a variety of complexes with poorly understood structures. Upon incubation of CODH with CO, an axial EPR spectrum (the NiFeC signal) with effective g values of gi = 2.08 and g ,, = 2.02 is observed (13). Owing to its slow electronic spin relaxation rate, the signal is observable at temperatures as high as 170 K. This signal arises from a species containing Ni, Fe, and the carbon from CO, as demonstrated by the hyperfine broadening of the signal when samples are enriched in either 61Ni (14), 13C0 (14), or '?Fe (15). Therefore, the complex responsible for this EPR signal has been referred to as a NiFeC complex (15).
A rhombic signal with g values of 2.062, 2.047, and 2.028 is sometimes observed in spectra of . It is thought to arise from an altered form of the NiFeC complex (15). Addition of CoASH converts the rhombic signal into the ' Very recently, Ramer et al. (48) suggested that the subunit stoichiometry of the CODH holoenzyme is actually (a@~)~ and that the y subunit is the disulfide reductase polypeptide.

3874
EPR and Electrochemistry of CO Dehydrogenase axial form, which led to the suggestion (15) that CoASH binds near or at the complex. When CODH is treated with Nbromosuccinimide, an agent known to modify tryptophan residues, only the rhombic signal is observed (16). From fluorescent quenching studies, it has been suggested that a tryptophan residue may be involved in the binding of CoASH to CODH (17).
Both iron and nickel EXAFS of the enzyme have been reported (20,21). Analysis of the iron EXAFS spectrum indicates the presence of iron-sulfur clusters. The reported Fe-Fe coordination number of 2.6 is most consistent with the presence of [Fe&] clusters (21). The nickel ions hpve been proposed to be bound to either four sulfurs, at 2.16 A, (21) or to a combination of sulfur and oxygen/nitrogen ligands (20). The CODHs from other anaerobic bacteria contain nickel and iron complexes as well and exhibit related but not identical EPR signals. The enzyme from the acetogen Acetobacterium woodi has the same (@)s subunit structure (12) and physiological role (16) as the C. thermoaceticum enzyme. The reduced form of the enzyme from A. woodi yields EPR signals similar to the g,, = 1.94 and NiFeC signals of C. thermoaceticum (12). The CODH from Methanosarcinu thermophilu exhibits a NiFeC-type EPR signal similar to the acetogenic enzymes (22). The reduced CO dehydrogenase from Rhodospirillum rubrum exhibits two g,,, = 1.94 signals, a g,, = 1.82type signal (termed signal A), and a g,,, = 1.86-type signal (termed signal B) (23, 24). No NiFeC-type signal has been reported for the R. rubrum enzyme. A nickel-deficient inactive form of the R. rubrum CODH exhibits g = 1.94 signals identical to those in the holoenzyme (25). Unlike the enzyme from C. thermouceticum, the R. rubrum enzyme is not involved in acetate synthesis.
We have studied the magnetic and redox properties of the metal complexes in CODH by combining EPR spectroscopy with controlled potential coulometry (CPC). Besides providing estimates of the reduction potentials for each EPR-active complex, we identify low field EPR signals between g = 6 and 4 and report that the enzyme can serve as an efficient electrocatalyst in the reduction of COZ to CO. We also highlight some major problems facing future biophysical studies of this enzyme. In the following paper we report the results of a Mossbauer study of this enzyme.  In COZ, and at potentials < -350 mV, the cell current never returned to background levels, rendering the determination of the number of electron equivalents used impossible. (We will argue below that this steady-state current is due to the catalytic reduction of CO, to CO.) However, at potentials between 0 and -300 mV, currents did return to background levels, and quantitations were possible. Two electron equivalents reduced the oxidized enzyme in this potential region.
The electrochemically poised samples obtained from the titrations were analyzed by EPR spectroscopy. We observed a variety of signals, which are summarized in Fig. 1 and Table   II and are characterized in the following sections. &", = 1.94 Signal-In earlier studies, a g,,, = 1.94 signal was observed in spectra of reduced CODH (13). We have found that, depending on the protein preparation used, two gwe = 1.94 signals, with the same g values (2.04, 1.94, 1.90) but different linewidths may be observed. The narrow signal, gN = 1.94, is shown in Fig. L4, while the broad one, gB = 1.94, is shown (along with gN = 1.94) in Fig. 1B. The proportion of each signal observed depends somewhat on the sample and the EPR conditions. The gN = 1.94 signal saturates at lower powers (-10 microwatts uersus >5 mW, at 10 K) and shows relaxation broadening at higher temperatures (-60 uersus -20 K) than the gB = 1.94 signal. At higher temperatures and lower microwave powers, the gN = 1.94 Signal iS mOSt COnSpiCuous. We do not presently know why some preparations yield more of the broad form than others.
to that used in A. The   Procedures." The numbers on the right are the potentials (in mV) at which samples were poised. The dashed lines are simulations using the same (or similar) parameters as those described in Fig. 1. The intensities of each component are given in Table I. The spectral region -50 G to either side of g = 2.0 was removed as described in Fig. 1. EPR conditions were the same as in Fig. 1A except that the sample temperature was 11 K.
We have attempted to assess the reduction potentials of the g,,, = 1.94 species. From plots of signal intensities uersus potential, an average reduction potential of z-440 mV was obtained for the different titrations.
The midpoint potential associated with the gN = 1.94 species appears to be somewhat more positive than that yielding the ga = 1.94 signal, but further studies and more experience with enzyme preparations are required to quantify this difference.
The values obtained for the spin concentrations of these signals were low; the sum of both signals had an average of 0.64 + 0.14 spins/dimer (17 different samples). When the separate contributions from each signal was assessed, the narrow signal yielded an average of 0.26 & 0.05 spins/dimer, and the broad one 0.34 + 0.14 spins/dimer (11 samples). Low Field EPR Features-The EPR spectra of [Fe&]'+ clusters in S = 312 spin states exhibit low intensity features between g values of 6 and 4 (37-40). In spectra of reduced CODH, we noticed similar features. The features became more intense at negative potentials, but they did not develop in a simple Nernstian fashion. Some intensity remained in spectra which lacked the g.,, = 1.94 signals. It may be that the low field features reflect contributions from more than one complex.
go", = 1.82 Signal-The g,,,, = 1.82 signal, with g values of 2.01, 1.81, and 1.65, is shown in Fig. 1C (along with gN =  1.94). The spin concentrations of the signal, assumed to result from an S = l/2 system, were significantly below unity; we obtained an average of 0.18 f 0.06 spins/dimer (26 spectra). We found the redox behavior associated with this signal to be somewhat unusual. In argon-atmosphere titrations, it developed according to a midpoint potential of about -220 mV (Fig. 2), and then completely disappeared at more negative potentials.4 In the CO, titration (Fig. 4), and other experiments performed under COZ or CO, the signal did not develop at any potential tested. We have observed the g.,. = 1.82 signal in CO-reduced CODH samples to which oxidized corrinoid/Fe-S protein was added (18,19). &", = 1.86 Signal-The g.,. = 1.86 signal, with g values of 1.97, 1.87, and 1.75, is shown in Fig. 1D (along with the gN = 1.94 signal). Spin quantitations of this signal yielded an average of 0.32 f 0.10 spins/dimer (seven samples). The redox behavior of this signal was unusual as well. In the argon atmosphere titration illustrated in Fig. 2, the signal appeared at quite negative potentials, according to a midpoint potential E, = -530 mV, while in the titration performed under CO, (Fig. 4), the signal appeared when samples were poised at -350 mV and below, in accordance with E,,, = -360 mV.4 This potential is roughly 170 mV more positive than that obtained from the titration performed under argon.
NiFeC Sign&-The so-called NiFeC signal, with g values of gL = 2.08 and gll = 2.02 is known to arise from samples reduced by CO. An example of this signal is shown together with the g,, = 1.94 signal and the gaYe = 1.86 signal in Fig.  1F. We have observed that it is also present when samples are prepared under CO, at negative potentials (Fig. 1E). We have never seen the signal when samples were reduced electrochemically in an argon atmosphere or reduced by dithionite.
At 10 K, the NiFeC signal begins to saturate homogeneously at roughly 10 microwatts microwave power; at 50 mW, it is virtually absent. At temperatures above 77 K, saturation was not a problem. Quantitation of the signal at 100 K and 10 mW yielded an average of 0.3 f 0.2 spins/dimer.
The CPC titration performed under a CO, atmosphere yielded, not surprisingly, results quite different from those performed under argon. With argon, the current returned to background level after the reduction of the enzyme was com-4 In some titrations, the g., = 1.82 signal did not disappear at more negative potentials and the g., = 1.86 signal never developed. Subsequent titrations (C. M. Gorst and S. W. Ragsdale, unpublished results) suggests that the disappearance of the g.,, = 1.82 and the appearance of the g.,. = 1.86 signals at negative potentials are reproducible characteristics of CODH. plot of steady-state current in CO2 atmosphere titration uersus potential (lower). The dashed line in upper plot is of a reduction (-600 mV) performed under a CO, atmosphere; the solid line is of one performed under argon at the same potential. In both cases, 9.8 nmol of thionin-oxidized CODH were injected after -5 min. Background currents of 2.2 pA, for the titration under CO*, and 0.7 WA, for that under argon, have been subtracted. In the lower plot, the currents present 60 min after the injection of the protein are plotted against the solution potentials. The maximum steady-state currents are being limited by a factor other than the enzyme concentration and do not reflect the turnover number intrinsic to the enzyme. Details of this experiment were the same as described in Fig. 2. plete (Fig. 3, upper plot, solid line). Under COa, and at potentials sufficiently negative, currents never returned to background (dashed line).
To test whether this behavior was due to the electrocatalytic reduction of CO* to CO by CODH, we analyzed for CO in a repeat experiment by including deoxyhemoglobin in the buffer (poised at -500 mV). Prior to the addition of CODH, the current in the cell was at background levels. After a catalytic amount of CODH was added, the current increased immediately, and enough current was passed to allow the generation of sufficient CO to bind to all of the hemoglobin in the sample. The shift in the 430 nm absorption band (typical of deoxyhemoglobin), observed prior to the addition of CODH, to 419 nm (typical of carboxyhemoglobin (41)) after the addition, indicated that CO was being generated through the catalytic action of CODH.
The steady-state current thereby generated was plotted against the solution potential (Fig. 3, lower plot). The potential at half-maximal activity, Ecat, was z-430 mV. EPR spectra obtained from this titration are shown in Fig. 4. Three signals developed, the g.,, = 1.86 signal at E, = -360 mV, and the g.,, = 1.94 and NiFeC signals at E, = -410 mV. Spin quantitations for the signals are listed in Table I.

DISCUSSION
The EPR and redox properties of CODH are immensely complicated, and many problems must be overcome before they will be fully understood.
Three major difficulties are evident from our studies. First, many of the EPR signals of CODH exhibit similar properties; they develop when the enzyme is reduced and in the same spectral region. This makes their unobscured study difficult. Second, the values obtained in the spin quantitations (spin/protein) of every signal are significantly less than unity. Ideally, such quantitations yield one spin/complex, and these values can then be used to determine the number of complexes/protein. However, the low values obtained for the CODH signals indicate that such a straightforward analysis is not possible here. Third, the redox behavior of the enzyme seems to be dependent on subtle differences in protein preparations.
The two g,,, = 1.94 species observed here are probably the most readily understood. Such signals are known to arise from either [Fe4SJ1+ or [Fe2S211+ clusters (42). The two signals observed in CODH, with the same g values but different linewidths, were originally considered to be a single signal and were assigned to an [Fe&]" cluster (13). From our EPR study, we were unable to determine whether the signals arise from one cluster or two. However, there is precedence for both possibilities.
That two gave = 1.94 signals may result from a single cluster is exemplified by the corrinoid/Fe-S protein from C. thermoaceticum. This protein contains a single [Fe&]'+ cluster which yields two g.,, = 1.94 signals, also with different linewidths (18). The features observed between g = 6 and 4 probably arise from [Fe&J1+ clusters with S = 3/2 spin states (37-40). Similar features are observed in spectra of the nitrogenase Fe protein (37). About half the molecules of the Fe protein have an [Fe4SJ1+ cluster in an S = l/2 spin state, the other half have S = 3/2 states. The S = l/2 fraction gives rise to a g,,, = 1.94 signal, while the S = 3/2 fraction yields features between g = 4 and 6. Meyer and co-workers (43) have recently shown that the S = 3/2 form of the Fe protein arises from an interaction between the cluster and its environment at low temperatures.
A similar situation may apply to the clusters in CODH, but this is by no means established. Using M&sbauer spectroscopy, we have not been able to identify the cluster(s) associated with the low field EPR features (29).
The g.,.e = 1.86 and 1.82 signals have been mentioned only briefly in earlier studies (13,18,19), and the types of structures from which they arise have not been identified.
The redox properties of the complexes yielding these signals are difficult to understand.
Perhaps the g,,, = I.82 signal disappeared when samples were poised at potentials below about -400 mV because the complex yielding this signal became further reduced into an EPR silent state. Alternatively, the spin-state properties of the complex could have changed in a potential-dependent manner such that the g,,, = 1.82 signal was replaced by the g.,, = 1.86 signal. The disappearance of the g,,, = 1.82 signal could also be caused by spin-coupling to a paramagnetic species which is generated at low potential. The appearance of the g,,, = 1.82 and 1.86 signals was dependent on the presence or absence of CO&O and, in some of our earlier titrations," on the protein preparation used. That COr can modulate the redox properties associated with these signals, especially at potentials (e.g. -300 mV) more positive than that required for COs reduction, indicates that at these positive potentials CO1 might be binding at a site on the enzyme.
Determination of the midpoint potential of the complex yielding the NiFeC EPR signal from the CO, atmosphere titration data is difficult because the resulting EPR data reflect the redox states of the clusters during catalytic turnover (44). Consequently, the intensities do not only depend on E, but also on the rates of reduction and reoxidation of the complex, and perhaps other factors. These considerations are not relevant for samples poised at potentials > -350 mV, because under these conditions no turnover occurs. This potential provides an upper limit for the E, of the complex. Since some reduced signal developed in the turnover region (at -450 mV and below), the redox potential of the complex should be within about 70 mV of these values. With these considerations, we estimate -350 > E, > -520 mV for the complex in the presence of CO, at pH 6.3. Recent studies indicate that the redox potential is slightly more negative than this estimate." We have shown in this paper that purified CODH is able to catalyze the reduction of CO, to CO. The potential at halfmaximal activity, E,,, = -430 mV, is near the thermodynamic reduction potential for the CO&O couple at pH 6.3 (-517 mV) (calculation using data of reference 45). 6 We found that both the NiFeC and g,,, = 1.94 EPR signals developed concomitantly with CO, reduction. Most likely, the last electron transfer step required before catalysis can occur is the reduction of one of the clusters yielding these signals. Since the complex producing the NiFeC signal is known to bind CO*, it is probably the one which must be reduced before catalysis can occur.
Much remains to be learned about the metal complexes of CODH. The plethora of EPR signals suggests that some of them arise from different states of the same complex. However, it is not yet known which of the signals are so associated. It is likely that some of the complexes have unique structures, given their unusual EPR and redox properties.
Mijssbauer spectroscopy is well suited to address these problems (46,47), and we have undertaken such a study, utilizing the results ' C. M. Gorst and S. W. Ragsdale, unpublished results. "The thermodynamic value is almost 90 mV more positiue than the apparent E,,,. It may be that the currents observed at the more negative potentials are limited by some factor other than the enzyme concentration, and so the value of E,,, calculated from these data is more positive than the true value. Other factors may be involved as well, and further studies are needed to fully explain this situation. described here. In the following paper, the results of that study are reported, and our preliminary conclusions regarding the structures of the metal complexes in CODH are drawn.