Interaction of Potassium Cyanide with the [Ni-4Fe-5S] Active Site Cluster of CO Dehydrogenase from Carboxydothermus hydrogenoformans*

The Ni-Fe carbon monoxide (CO) dehydrogenase II (CODHIICh) from the anaerobic CO-utilizing hydrogenogenic bacterium Carboxydothermus hydrogenoformans catalyzes the oxidation of CO, presumably at the Ni-(μ2S)-Fe1 subsite of the [Ni-4S-5S] cluster in the active site. The CO oxidation mechanism proposed on the basis of several CODHIICh crystal structures involved the apical binding of CO at the nickel ion and the activation of water at the Fe1 ion of the cluster. To understand how CO interacts with the active site, we have studied the reactivity of the cluster with potassium cyanide and analyzed the resulting type of nickel coordination by x-ray absorption spectroscopy. Cyanide acts as a competitive inhibitor of reduced CODHIICh with respect to the substrate CO and is therefore expected to mimic the substrate. It inhibits the enzyme reversibly, forming a nickel cyanide. In this reaction, one of the four square-planar sulfur ligands of nickel is replaced by the carbon atom of cyanide, suggesting removal of the μ2S from the Ni-(μ2S)-Fe1 subsite. Upon reactivation of the inhibited enzyme, cyanide is released, and the square-planar coordination of nickel by 4S ligands is recovered, which includes the reformation of the Ni-(μ2S)-Fe1 bridge. The results are summarized in a model of the CO oxidation mechanism at the [Ni-4Fe-5S] active site cluster of CODHIICh from C. hydrogenoformans.

The hydrogenogenic thermophilic bacterium Carboxydothermus hydrogenoformans utilizes CO 3 as a sole source of energy and carbon under anaerobic chemolithoautotrophic conditions (1,2). The oxidation of CO is catalyzed by two Ni-Fe CO dehydrogenases, designated CODHI Ch and CODHII Ch , according to the equation CO ϩ H 2 O 3 CO 2 ϩ 2 H ϩ ϩ 2 e Ϫ . Crystal structures of CODHII Ch in different functional states have been solved to 1.1 Å resolution (3,4). The homodimeric enzyme contains five metal clusters, of which clusters B, BЈ, and a subunit-bridging cluster D are conventional cubane-type [4Fe-4S] clusters (3). The active site clusters C and CЈ in dithionite-reduced active CODHII Ch (CO oxidation activity ϳ14,000 mol of CO oxidized min Ϫ1 mg Ϫ1 at 70°C) have been modeled as [Ni-4Fe-5S] centers containing a Ni-( 2 S)-Fe1 subsite. Their integral nickel ion is coordinated by 4S ligands with squareplanar geometry (3,4). The nickel coordination by 4S ligands in CODHII Ch was also apparent from x-ray absorption spectroscopy (XAS) (5). A defined non-functional [Ni-4Fe-4S] form of cluster C can be produced by treatment of CODHII Ch with CO in the absence of low potential reductants, resulting in inactivation and the loss of the bridging 2 S (4).
It has been assumed that CODHII Ch catalyzes the oxidation of CO at the Ni-( 2 S)-Fe1 subsite of cluster C (3). The prime candidate for CO binding is the nickel ion because of its facile accessibility through the substrate channel and its empty apical coordination site (3). Fe1 is the presumed OH Ϫ donor ligand in CO 2 formation (3,6). The CODH Oc from the aerobic bacterium Oligotropha carboxidovorans oxidizes CO at the Mo-( 2 S)-Cu subsite of the [Cu-S-MoO 2 ] active site, in which copper and molybdenum are bridged by a cyanolyzable sulfane 2 S (7,8). The enzyme is inactivated when 2 S is removed and reactivated when 2 S is reinserted (9). The Mo-( 2 S)-Cu subsite resembles the Ni-( 2 S)-Fe1 bridge in cluster C of CODHII Ch . The mechanism of CO oxidation based on the x-ray structure of [Cu-S-MoO 2 ] CODH Oc with bound inhibitor n-butyl isocyanide involves a thiocarbonate-like intermediate state and proposes the binding of CO between the 2 S and copper (equivalent to nickel in CODHII Ch ) and the binding of an OH Ϫ group at molybdenum (equivalent to Fe1 in CODHII Ch ) (7).
Structures of cluster C of Ni-Fe CODHs from Rhodospirillum rubrum (CODH Rr ) (10) and Moorella thermoacetica (CODH Mt ) (11,12) also showed the positions of the five metal ions in cluster C of CODHII Ch but did not reveal the bridging 2 S. Since sodium sulfide was found to inhibit CODH Rr and CODH Mt , it has been concluded that cluster C with the bridging 2 S, as has been observed in CODHII Ch from C. hydrogenoformans (3,4), might represent an inhibited form (13). On the other hand, it has been shown that the [Ni-4Fe-4S] cluster miss-* This work was supported by the Deutsche Forschungsgemeinschaft (Grants SV10/1-1 and SV10/1-2). 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. □ S The on-line version of this article (available at http://www.jbc.org) contains a supplemental ing the bridging 2 S is an inactivated decomposition product originating from the [Ni-4Fe-5S] cluster (4). A catalytic mechanism suggesting the apical binding of CO at nickel and the coordination of OH Ϫ in the bridging position between the nickel and Fe1 ions was proposed for CODH Rr and CODH Mt (13)(14)(15).
To clarify some of these aspects and get further insights into the mechanism of CO oxidation, we were interested to study how CO interacts with cluster C of CODHII Ch and how the bridging 2 S might be involved in catalysis. Although some crystal structures have modeled apical CO at the nickel ion, the occupancies of CO were very low in CODH Mt (12), or the potential CO was observed in the non-functional [Ni-4Fe-4S] cluster C of CODHII Ch (4). We could not identify a CO ligand in the functional cluster C of CODHII Ch , which can be ascribed to the high turnover rate of CO oxidation (4,000 s Ϫ1 at 23°C) (2,4). Therefore, we have focused on the reactivity of CODHII Ch with potassium cyanide. CO and the cyanide ion show similar reactivities because they are isosteric and isoelectronic, display a similar -donor and -acceptor ligand character, and share the presence of a non-bonding pair of electrons in the sp-hybridized orbital of the terminal carbon atom. Cyanide has a similar size as CO, allowing its passage through the substrate channel of CODHII Ch (3), and inhibits the oxidation of CO by Ni-Fe CODHs and [Cu-S-MoO 2 ] CODH Oc (9, 16 -19). The inhibition of Ni-Fe CODHs has been attributed to the binding of cyanide to the nickel ion of cluster C in CODH Rr (17) or to the Fe1 ion (ferrous component II) of cluster C in CODH Rr and CODH Mt (20). It has been established that the inhibition of [Cu-S-MoO 2 ] CODH Oc is due to the removal of the copper ion as copper cyanide and the sulfane 2 S as thiocyanate, resulting in the formation of a [MoO 3 ] center, which is catalytically inactive (7,9).
We have studied the reactivity of CODHII Ch with potassium cyanide and the nickel coordination derived from XAS. We show that cyanide behaves as a competitive inhibitor of CODHII Ch with respect to the substrate CO and apparently interacts with cluster C in a similar fashion. It inhibits reduced CODHII Ch reversibly and forms a nickel cyanide in the equatorial plane. Thereby one of the four square-planar sulfur ligands of nickel, apparently the labile 2 S, is replaced but remains bound to Fe1. Reactivation cleaves the nickel cyanide and results in square-planar 4S-coordination of nickel and reformation of the Ni-( 2 S)-Fe1 bridge. The results are summarized in a model of CO oxidation at cluster C of CODHII Ch from C. hydrogenoformans.

EXPERIMENTAL PROCEDURES
All experiments were done under strictly anoxic conditions because of the oxygen sensitivity of CODHII Ch (2).
Purification and Assay of CODHII Ch -Growth of C. hydrogenoformans Z-2901 (DSM 6008) on CO and purification of CODHII Ch were carried out as detailed (2,21). CO oxidation activity of CODHII Ch was assayed at 70°C by following the CO-dependent reduction of oxidized methyl viologen employing an ⑀ 578 of 9.7 mM Ϫ1 cm Ϫ1 as described (2). For the assays, 1-ml volumes of reaction mixture composed of anoxic 50 mM HEPES/NaOH (pH 8.0) (buffer A) with 20 mM methyl viologen and 2 mM dithiothreitol (DTT) were flushed with CO in screwcapped cuvettes sealed with a rubber septum. Reactions were initiated by injecting 10 l of diluted enzyme solution. One unit of CO oxidation activity is defined as 1 mol of CO oxidized/min. Protein estimation employed conventional methods with bovine serum albumin as a standard (22). Purified as isolated CODHII Ch displayed a CO oxidation activity of 14,800 units mg Ϫ1 of protein and had a protein concentration of 8.8 mg ml Ϫ1 .
Inhibition of CODHII Ch by Potassium Cyanide under Turnover Conditions-Potassium cyanide (KCN) stock solutions were prepared in anoxic 10 mM aqueous NaOH under N 2 . KCN was added to the cuvettes for the assay of CO oxidation activity prior to the addition of CODHII Ch . The different CO concentrations for kinetic measurements were established by adding the appropriate amounts of CO-saturated reaction mixture to N 2 -saturated reaction mixture. At 70°C and 1 atm pressure, the CO concentration in CO-saturated reaction mixtures was taken to be 645 M (23).
Inhibition of CODHII Ch by Potassium Cyanide under Nonturnover Conditions-Assays (1 ml) were performed at 23°C and contained CODHII Ch in buffer A with 4 mM sodium dithionite plus 2 mM DTT, 4 mM Ti(III) citrate, 2 mM DTT or without reductants under an atmosphere of CO or N 2 . Inhibition was initiated by the addition of KCN. Aliquots were removed with time and analyzed for CO oxidation activity.
Reactivation of Cyanide-inhibited Reduced CODHII Ch -Cyanide-inhibited reduced CODHII Ch was prepared by treatment of the enzyme with KCN under non-turnover conditions. CODHII Ch (109 g) was incubated for 20 min at 23°C under N 2 in 1 ml of buffer A containing 75 M KCN, 4 mM dithionite, and 4 mM DTT. Such treatment resulted in complete loss of CO oxidation activity. To lower the concentration of KCN in reactivation assays to non-inhibitory 0.07 M, samples of inhibited CODHII Ch were diluted 11 times in buffer A containing 2 mM DTT under N 2 . For reactivation, 10 l of diluted samples were added to 1 ml of buffer A without reductants or with 4 mM dithionite plus 2 mM DTT, 4 mM Ti(III) citrate, or 2 mM DTT under CO or N 2 . Reactivations were performed at 23, 50, and 70°C.
Effect of Sodium Sulfide on CODHII Ch -Sodium sulfide (Na 2 S) was added to CODHII Ch activity assays under turnover conditions, to the reactivation assays of cyanide-inhibited reduced CODHII Ch , and to the as isolated CODHII Ch under non-turnover conditions.
Preparation of Samples for XAS-Dithionite-reduced CODHII Ch (CODH-DT, 15,400 units mg Ϫ1 ) was produced by treatment of as isolated enzyme under N 2 with 4 mM dithionite for 5 min at 23°C followed by concentration under N 2 to 99 mg ml Ϫ1 . CO-treated dithionite-reduced CODHII Ch (CODH-CO, 14,600 units mg Ϫ1 ) was produced by incubation of as isolated enzyme under CO in the presence of 4 mM dithionite for 1 h at 50°C followed by concentration under CO to 85 mg ml Ϫ1 . For dithionite-reduced CODHII Ch reversibly inhibited by cyanide (CODH-CN a and CODH-CN b ), the as isolated enzyme was diluted under N 2 with buffer A containing 4 mM dithionite and 2 mM DTT to 0.25 (CODH-CN a ) or 0.46 mg ml Ϫ1 (CODH-CN b ). KCN was added to the final concentrations of 1 mM (CODH-CN a ) and 200 M (CODH-CN b ). After a 40-min incubation at 23°C with gentle stirring, the activity was completely inhibited. The samples were concentrated under N 2 to 64 (CODH-CN a ) and 132 mg ml Ϫ1 (CODH-CN b ) and displayed activities of 700 and 44 units mg Ϫ1 , respectively. CODH-CN a and CODH-CN b were reversibly inhibited since they could be reactivated to 5,400 and 13,100 units mg Ϫ1 , respectively, upon incubation at 70°C under CO or N 2 in buffer A containing 4 mM Ti(III) citrate or 4 mM dithionite plus 2 mM DTT.
CODHII Ch reactivated after the inhibition by cyanide (CODHreact., 13,100 units mg Ϫ1 ) was prepared from the concentrated CODH-CN b . To remove the unbound cyanide, CODH-CN b (70 l) was dissolved in 0.5 ml of buffer A under N 2 containing 4 mM dithionite and 2 mM DTT and concentrated to 50 l by ultrafiltration under N 2 . The procedure was repeated two times. For the reactivation, the concentrated enzyme (50 l) was dissolved in 30 ml of buffer A with 4 mM dithionite and 2 mM DTT to a protein concentration of 0.3 mg ml Ϫ1 and incubated at 70°C under N 2 for 40 min. Reactivated CODHII Ch was concentrated under N 2 to 157 mg ml Ϫ1 , yielding sample CODH-react.
Ni-K Edge XAS Measurements-For XAS measurements, 25 l of each sample were filled into plastic cells covered with Kapton foil. Cells were sealed and kept in liquid N 2 . The Ni-K edge XAS data, 8.2-9.2 keV, were collected in fluorescence mode with a silicon (111) monochromator at the EMBL beamline D2 (DESY, Hamburg, Germany). Harmonic rejection was achieved by a focusing mirror (cut-off energy 20.5 KeV) and a monochromator detuning to 70% of its peak intensity. The sample cells were kept at ϳ20 K in a two-stage Displex cryostat. Automated data reduction, such as normalization and extraction of the fine structure, was performed with KEMP (24) assuming an energy threshold E 0,Ni ϭ 8,333 eV. Sample integrity during exposure to synchrotron radiation was checked by monitoring the position and shape of the absorption edge on sequential scans. No change in redox state or metal environment was detectable. X-ray Absorption near Edge Structure (XANES) Analysis-The Ni-K edge position was defined at the energy corresponding to a normalized absorbance of 0.5 (25). The extracted 1s 3 3d pre-edge feature, isolated by subtracting an arctangent function and a first order polynomial to the rising edge (25), was fitted to a single Gaussian function centered at ϳ8,332 eV. Its intensity corresponds to the integrated area of the Gaussian function.
Extended X-ray Absorption Fine Structure (EXAFS) Analysis-The extracted Ni-K edge (20 -700 eV) EXAFS were converted to photoelectron wave vector k-space and weighted by k 3 . The spectra were analyzed with EXCURV98 (26), refining the theoretical EXAFS for defined structural models based on the curved-wave theory. In addition to single scattering contributions, multiple scattering units were defined for linear Ni-C-N and square-planar Ni-4S. Parameters of each structural model, namely the atomic distances (R), the Debye-Waller factors (2 2 ), and a residual shift of the energy origin (EF), were optimized, minimizing the fit index (⌽) while keeping the number of free parameters below those of the independent points (26,27). Throughout the data analysis, the amplitude reduction fac-tor was kept at 1.0. The reduced 2 test verified the significance of an additional ligand in the models (26).

RESULTS AND DISCUSSION
Inhibition of CODHII Ch by Potassium Cyanide-Potassium cyanide inhibits CO oxidation activity of CODHII Ch under catalytic (turnover) conditions in the presence of CO and electron acceptor methyl viologen (Figs. 1, A and B, and 2A) as well as under non-turnover conditions in the absence of CO and acceptor (Fig. 1, C and D). The rate of inhibition depends on time (Fig. 1, A-D), the cyanide concentration (Fig. 1, A-C), and the incubation temperature (Fig. 1, A and B). Since CODHII Ch displays a high temperature optimum for activity (2), the temperature dependence of inhibition indicates a similar mode of interaction of cyanide and CO with the active site. This is supported by the double reciprocal plot of initial activity versus CO concentration as a function of cyanide concentration, revealing a pattern characteristic of competitive inhibition and a K i of 21.7 M cyanide ( Fig. 2A).
The inhibition by cyanide under non-turnover conditions greatly depends on the redox state of CODHII Ch (Fig. 1, C and D). Reduced CODHII Ch incubated with low potential reductants dithionite or Ti(III) citrate (redox potential ϳϪ500 mV) is inhibited more strongly than the more oxidized enzyme incubated with the weak reductant DTT (ϳϪ330 mV) or without reductants. This indicates an efficient interaction of cyanide with the highly reduced cluster C of CODHII Ch at redox potentials of ϳϪ500 mV. Obviously, CO also interacts with cluster C of CODHII Ch at very low potentials since the oxidation of CO in C. hydrogenoformans (Ϫ520 mV) is coupled to the reduction of protons to H 2 (Ϫ410 mV) (2). Therefore, cyanide interacts with the reduced cluster C of CODHII Ch in a similar fashion as the substrate CO.
CO protects reduced CODHII Ch against inhibition by potassium cyanide since there is no decrease of activity under CO in contrast to complete inhibition under N 2 (Fig. 1D), whereas the oxidized enzyme is not protected by CO (Fig. 2B). The protection by CO suggests that cyanide and CO share a common binding site at the reduced cluster C. The effects of temperature ( Fig. 1, A and B), redox dependence ( Fig. 1, C and D), protection by CO (Fig. 1D), and competitive character of inhibition with respect to CO ( Fig. 2A) suggest that the inhibition of reduced CODHII Ch by cyanide is due to the occupation of the CO binding site.
Reactivation of Cyanide-inhibited Reduced CODHII Ch -Inhibition of dithionite-or Ti(III) citrate-reduced CODHII Ch by potassium cyanide is fully reversible since the enzyme can be completely reactivated (Fig. 1, E and F). CO, high temperature (70°C), and the presence of dithionite or Ti(III) citrate accelerate the reactivation and increase the maximum level of regained activity as compared with reactivation under N 2 at lower incubation temperatures (23 or 50°C) and in the absence of low potential reductants (Fig. 1, E and F).
The significant acceleration of reactivation under CO (Fig. 1,  E and F), which apparently displaces cyanide at the CO binding site, indicates again that the interaction of reduced CODHII Ch with cyanide mimics its interaction with CO. The effect of CO is evident at 23 and 50°C (Fig. 1E). At 70°C, the effect of CO is negligible, and CODHII Ch can be completely reactivated after 25 min in the absence of CO (Fig. 1, E and F). Apparently, the dissociation of cyanide from the active site is accelerated by high temperatures, whereas at low temperatures, CO is required to displace the bound cyanide.
Low potential reductants are required for fast and complete reactivation (Fig. 1F). Inhibited CODHII Ch regains initial activity after a 15-25-min incubation at 70°C with dithionite or Ti(III) citrate under CO or N 2 . In contrast, slower and partial reactivation to 30 -50% of the initial activity occurs with DTT or without reductants (Fig. 1F).  A, competitive inhibition of CO oxidation by CODHII Ch under turnover conditions in the presence of CO, methyl viologen, and cyanide. KCN was added to serum-stoppered cuvettes for the assay of CO oxidation activity prior to the addition of CODHII Ch . Reactions were initiated by the addition of 10 l of stock enzyme solution (0.144 g ml Ϫ1 ). The different CO concentrations were established by adding the appropriate amounts of CO-saturated reaction mixture to assays containing the same reaction mixture saturated with N 2 . V 0 indicates the initial activity in units mg Ϫ1 . KCN concentrations in the cuvettes were 0 (F), 24.6 (), and 49.0 (f) M. B, effect of CO on the inhibition of oxidized CODHII Ch by cyanide under non-turnover conditions. CODHII Ch (0.19 g ml Ϫ1 ) was incubated with 2 mM KCN in the absence of reductants under an atmosphere of CO (F) or N 2 (E) at 23°C. C, effect of sodium sulfide on the activity of CODHII Ch under turnover conditions in the presence of CO and methyl viologen. Na 2 S was added to serum-stoppered cuvettes for the assay of CO oxidation activity containing 1.05 ng ml Ϫ1 CODHII Ch . 100% activity corresponds to 14,800 units mg Ϫ1 .
The reactivation patterns discussed above further substantiate that the inhibition of reduced CODHII Ch originates from the occupation of the CO binding site by cyanide. This inhibition is not due to any decomposition of cluster C since the activity can be completely recovered (Fig. 1, E and F).
Effect of Sodium Sulfide on CODHII Ch -Sodium sulfide has no effect on the reactivation of cyanide-inhibited CODHII Ch in the presence of dithionite (Fig. 1G). Partial reactivation with sulfide alone (Fig. 1G) is brought about by its function as a strong reductant and not as a sulfur source since the enzyme can be completely reactivated in the presence of Ti(III) citrate alone (Fig. 1F). Sulfide does not inhibit CODHII Ch under nonturnover conditions in the presence or absence of dithionite (Fig. 1H) as well as under turnover conditions (Fig. 2C). Apparently, sulfide does not affect CO oxidation by CODHII Ch , which is in contrast to the reported inhibition of CODH Rr and CODH Mt by sulfide and to the suggested inhibitory role of the bridging 2 S (13).
XAS of Dithionite-reduced CODHII Ch -Ni-K edge XAS on highly active dithionite-reduced CODHII Ch (CODH-DT, 15,400 units mg Ϫ1 ) reveals the nickel coordination in functional CODHII Ch in solution. The XANES spectrum (Fig. 3A) resembles that of the four-coordinate square-planar complexes of nickel (25). It shows a small shoulder near 8,337 eV, which has been observed in tetragonal geometries lacking one or more axial ligands and has been assigned to a 1s 3 4p z transition (with shakedown contributions). The spectrum exhibits a very weak 1s 3 3d pre-edge peak centered at ϳ8,332 eV. The normalized integrated area of this peak is 0.030 eV. The 1s 3 3d transition is dipole-forbidden; however, it can gain intensity due to p-d mixing in non-centrosymmetric geometries. Thus, planar complexes will feature weak transitions with areas of ϳ0.0 -0.029 eV, whereas the tetrahedral ones will display stronger transitions with areas of ϳ0.08 -0.114 eV (25). In CODH-DT, the combination of weak 1s 3 3d transition and a shoulder on the rising edge indicates that the nickel ion is fourcoordinate with a square-planar geometry. The edge energy of 8338.5 eV is consistent with a Ni 2ϩ oxidation state of the nickel ion (25).
The EXAFS spectrum provides further insight into the metal coordination (Fig. 3B, trace a). The amplitude envelope of the oscillations, e.g. its maximum at ϳ6.5 Å Ϫ1 , is indicative of the presence of elements heavier than oxygen and nitrogen in the vicinity of the absorber atom. The lack of the beat node-like change in the EXAFS amplitude marks a homogenous ligand sphere. This is further substantiated by the Fourier transform of the EXAFS data showing one dominant peak at ϳ2.2 Å and small contributions at ϳ2.8 and ϳ4.4 Å (Fig. 3C,  trace a). Both 2.2 and 4.4 Å peaks could be best fitted with four Ni-S interactions at 2.23 Å in the square-planar geometry ( Table 1). No further interactions with light atoms, e.g. Ni-O, at shorter bond lengths were required for a good fit.
The Ni-S bond lengths depend on the nickel geometry. In four-coordinate Ni 2ϩ complexes containing thiolate ligands, the Ni-S bond lengths range from 2.14 to 2.24 Å for approximately square-planar geometries and from 2.26 to 2.33 Å for tetrahedral geometries (28). The Ni-S distances in CODHII Ch (Table 1) are in the range for square-planar complexes. Inde-pendent evidence for this coordination arises from the multiple scattering via the central nickel atom within the Ni-4S system, visible at about 4.4 Å in the Fourier transforms. Such features only occur when the scattering vector is close to 180°.
To identify the potential contributors to the ϳ2.8 Å peak in the Fourier transform spectrum, two different scenarios have been considered. Based on the crystal structure of functional CODHII Ch , the most probable ligands to nickel at this distance are 2Fe ions at ϳ2.8 and ϳ2.9 Å (3). To test their presence, a single 2Fe shell was modeled first, but its refinement resulted in a relatively high Debye-Waller factor, and thus, had to be discarded. Followed by that, a single 1Fe shell at ϳ2.7 Å was fitted (supplemental Table 1). The inclusion of a second iron contri- (trace e). Calculated spectra in B and C are shown by black lines; experimental spectra based on the models given in Table 1 are represented by colored curves. The abbreviations are: , EXAFS signal; k, photoelectron wave number; R, interatomic distance; FT, Fourier transform (modulus). The Fourier transform is phase-corrected for the shortest metal ligand contribution, and therefore, the peaks do not appear at the refined metal neighbor distances. For sample labels, see "Experimental Procedures." bution at ϳ2.9 Å significantly improved the fit, as shown by the 5% drop of the fit index ( Table 1). The data published previously on the as isolated CODHII Ch did not show any evidence for Ni-Fe contribution(s), whereas for the CO-treated and Ti(III) citrate-reduced samples, only one Ni-Fe interaction at ϳ2.74 Å was detected (5). The lack of the Ni-Fe contributions or their weak signal was then attributed to the destructive interference between 2.7 and 2.9 Å Ni-Fe components within the ϳ5-10 Å Ϫ1 range (5). In the present studies, a partial cancellation of both Ni-Fe signals takes place as well (especially between 5 and 8 Å Ϫ1 ). However, a longer photoelectron wave number (k) range as compared with the previous XAS data on CODHII Ch (5) ensures that both of the components can be detected. The lack of any substantial numerical correlation during the EXAFS data refinement between the structural parameters of both iron shells further supports this statement.
The model comprising 4S atoms at 2.23 Å in a square-planar geometry and 2Fe atoms at 2.71 and 2.99 Å (Table 1) correlates well with the crystal structure of functional CODHII Ch (3,4). However, the 2.71 Å Ni-Fe distance is shorter than the shortest Ni-Fe bond (2.82 Å) found by x-ray crystallography (3,4). It may reflect a redox-dependent conformational change in cluster C due to the slightly different redox state of the protein in samples studied by XAS and crystallography as it was previously suggested (5).
XAS of CO-treated Dithionite-reduced CODHII Ch -XAS on highly active CO-treated dithionite-reduced CODHII Ch (CODH-CO, 14,600 units mg Ϫ1 ) was performed to determine the effect of CO on the nickel coordination and to identify the binding position of CO. XAS revealed no change in nickel geometry and ligand sphere composition upon CO treatment. The Ni-K edge spectrum is almost identical to that of the CODH-DT with a small shoulder at ϳ8,337 eV and a pre-edge peak at ϳ8,332 eV (Fig. 3A). The edge energy has not changed and is consistent with the Ni 2ϩ state. EXAFS demonstrated that the nickel coordination has not been altered by CO treatment (Fig. 3, B and C, traces b). The final structural model is consistent with the model for CODH-DT and comprises 4S atoms at 2.23 Å in a square-planar geometry and 2Fe atoms at 2.71 and 2.96 Å (Table 1). A single 1Fe shell at ϳ2.7 Å was tested as well, but the fit was significantly worse, as demonstrated by the 8% increase of the fit index value, as compared with 2Fe model (supplemental Table 1). The obtained results are similar to the data on CO-treated CODHII Ch published previously (5). However, the average Ni-S bond length found in this study is slightly shorter (2.231 (2) versus 2.252 (3) Å) and has a lower Debye-Waller factor (0.0080 (3) versus 0.0155 Å 2 ), which indicates the lower structural disorder of the 4S shell in the present COtreated CODHII Ch sample. As the crystal structure of functional CODHII Ch , briefly treated with CO (4), the model for CODH-CO indicates the presence of 2 S and the absence of bound CO. Therefore, after turnover of CO, cluster C remains in the functional state with 4S ligands at nickel. Since a carbon atom was not apparent in the vicinity of the nickel ion, the reaction product CO 2 obviously leaves the active site very quickly without the formation of a stable carboxyl intermediate.
XAS of Dithionite-reduced CODHII Ch Reversibly Inhibited by Cyanide-XAS on dithionite-reduced CODHII Ch reversibly inhibited by cyanide (CODH-CN a with 700 units mg Ϫ1 and CODH-CN b with 44 units mg Ϫ1 ) elucidates cyanide binding to cluster C. The XANES patterns of both samples almost line up with each other (Fig. 3A) but differ significantly from those of CODH-DT and CODH-CO, indicating significant changes of structure and/or ligand composition of the nickel site. Both XANES spectra also exhibit features of CODH-DT and CODH-CO, i.e. weak 1s 3 3d pre-edge peak and a small shoulder due to the 1s 3 4p z transition. However, the position of the shoulder is shifted slightly toward lower energies, and the edge energy The numbers (n) of ligand atoms (L), their distance to the nickel ion (R), the respective Debye-Waller factor (2 2 ), the C-N bond length (R CN ), the Fermi energy for all shells (EF), and the Fit Index (⌽), indicating the quality of the fit, are shown. For the Ni-4S square-planar (*) and Ni-C-N linear (**) units, multiple scattering up to the fifth or third order has been included, respectively. The presence of the first shell Ni-4S multiple scattering is visualized by the peak at 4 (2) increases by ϳ1.2 eV. This edge shift could indicate an increase in the nickel oxidation state. However, the preserved low intensities of the 1s 3 3d transition (0.031 and 0.036 eV for CODH-CN a and CODH-CN b , respectively) exclude such a possibility and are consistent with the Ni 2ϩ oxidation state of the nickel ion (25). Instead, the change in the hardness of some of the donor atoms is more likely. A general shift to lower edge energies has been observed for complexes with increasing numbers of sulfur-donor ligands (25). Thus, the observed changes indicate that cyanide substitutes for one of the sulfur ligands without affecting the square-planar geometry of the nickel site.
In the EXAFS spectra of CODH-CN a and CODH-CN b , the sharp oscillations at ϳ6.5 Å Ϫ1 , present in CODH-DT and CODH-CO, are diminished (Fig. 3B, traces c and d), and a beat node-like change in the regular sinusoidal pattern emerges. This indicates heterogeneous ligand sphere, most probably caused by cyanide binding, and is visualized in both Fourier transform spectra (Fig. 3C, traces c and d). As compared with CODH-DT and CODH-CO, a small peak at ϳ1.8 Å emerges, whereas the ϳ2.8 Å contribution is replaced by a broad peak centered at ϳ3.0 Å. The 4.4 Å peak marking a 4S square-planar geometry in CODH-DT and CODH-CO decreases to the noise level. The 2.2 Å peak intensity decreases by ϳ30%. These features are likely caused by a cyanide ligand replacing one of the sulfur ligands. Based on the reduced 2 test (26), the best fit among all considered models was obtained for a nickel ion coordinated by three sulfur atoms at ϳ2.20 or ϳ2.23 Å and one CN group with a Ni-C distance of 1.81 or 1.84 Å for CODH-CN a and CODH-CN b , respectively ( Table 1). The refined Ni-C and C-N bond lengths are consistent with Ni 2ϩ complexes with cyanide ligands (29). Assuming Ni-C distances of 1.81 or 1.84 Å, Ni-N distances of 3.00 Å, and C-N distance in cyanide of 1.15-1.18 Å (29,30), cyanide binds to the nickel ion of cluster C by its carbon atom in a linear fashion.
Therefore, in CODHII Ch reversibly inhibited by cyanide, one of the Ni-S bonds is cleaved and one CN ligand is bound to nickel in square-planar geometry. This suggests that cyanide cleaves the labile bond between nickel and the bridging 2 S (4) and binds to nickel at the coordination site previously occupied by the 2 S.
XAS of CODHII Ch Reversibly Inhibited by Cyanide and Then Reactivated-XAS on CODH-react. (13,100 units mg Ϫ1 ) determines the nickel coordination in highly active CODHII Ch formed after reactivation of enzyme reversibly inhibited by cyanide. The Ni-K edge shape is almost identical to that of CODH-DT and CODH-CO, indicating the same square-planar geometry and oxidation state of nickel in CODH-react. (Fig.  3A). However, the normalized integrated area of the 1s 3 3d increases (0.040 eV), suggesting a slightly disordered geometry of the nickel site. EXAFS confirms this observation (Fig. 3B, trace e). As compared with CODH-DT, the intensity of the Ni-S backscattering contribution is lowered by ϳ15% (Fig. 3C, trace e), but multiple scattering contributions within the square-planar Ni-4S unit significantly improve the fit. Thus, a slightly disordered square-planar geometry of the nickel site is likely, especially because the average Ni-S bond length has not changed as compared with CODH-DT and CODH-CO (Table  1). This is consistent with the CODHII Ch crystal structure where the partial occupancy of the 2 S has been observed (4) and refers to a minor component lacking the fourth sulfur ligand. Activities (15,400 units mg Ϫ1 in CODH-DT versus 13,100 units mg Ϫ1 in CODH-react.) indicate the presence of roughly 15% catalytically non-competent enzyme in CODHreact., which presumably contains a Ni-CN. Then, the reactivated component must be formed entirely as a NiS 4 site. Thus, the model for reactivated CODHII Ch is similar to that of CODH-DT, comprising four square-planar sulfur at 2.23 Å and 2Fe at 2.69 and 2.97 Å (Table 1).
CO Oxidation at the [Ni-4Fe-5S] Cluster of CODHII Ch -This study shows that cyanide is an inhibitor of CODHII Ch because it competes with CO at the reduced [Ni-4Fe-5S] cluster. XAS indicates that the reversible inhibition of CODHII Ch with 4S coordinated nickel (CODH-DT) results in a 3S and 1CN coordinated nickel (CODH-CN a and CODH-CN b ). The binding of cyanide to nickel cleaves the bond between the nickel ion and the bridging 2 S, which stays bound to Fe1 since after reactivation, the 4S coordination of nickel is reestablished (CODHreact.), and external sulfide is not required for reactivation (Fig.  1, F and G). The requirement of reduced conditions for reactivation (Fig. 1F) indicates that the Fe1-bound 2 S should be in its S 2Ϫ state to produce the bridge.
We feel that our data do not support a mechanism of CO oxidation at the [Ni-4Fe-5S] cluster of CODHII Ch from C. hydrogenoformans involving binding of oxygen in a bridging position between nickel and iron (13)(14)(15) since this position will be occupied by sulfur in as isolated state or by CO after the binding of the substrate, and XAS did not identify an oxygen ligand to nickel in any of the examined states of the enzyme. On the other hand, since we have not captured any of the intermediates described in Fig. 4, the possibility of an oxygen atom, which transiently bridges the two metals during catalysis, cannot finally be ruled out. Our results also argue against an inhibitory role of the bridging 2 S in cluster C of CODHII Ch (13). We did not observe an inhibition of CODHII Ch by sulfide. As we cultivate C. hydrogenoformans in the presence of 3.3 mM Na 2 S as a reductant of the growth medium, the compound is apparently not toxic to the bacteria.
In analogy to the dithionite-reduced, cyanide-inhibited, and reactivated states of CODHII Ch analyzed by XAS, as well as in accordance with the structure-based mechanism of CODHII Ch (3), we propose a mechanism of CO oxidation at the [Ni-4Fe-5S] cluster (Fig. 4). An incoming CO molecule reaches the cluster through the substrate channel ending at the nickel ion (Fig.  4A) where it binds, resulting in a square-pyramidal five-coordinate intermediate (Fig. 4B). A water molecule binds as OH Ϫ to the histidine-coordinated Fe1 as proposed previously (6). The resulting CO to OH Ϫ distance exceeds 4 Å, which does not allow interaction (3). The 2 S ligand dissociates from the nickel and remains bound to pentacoordinated Fe1 as S 2Ϫ . A simultaneous rearrangement leads to a square-planar nickel intermediate with three sulfur and one CO ligand. This moves CO toward the OH Ϫ , making a nucleophilic attack possible (Fig.  4C). The proposed rearrangement agrees with the models describing the conversion of CO by a nickel to sulfur rebound mechanism studied with model compounds containing fourcoordinate nickel (31,32). The nickel-bound CO undergoes a nucleophilic attack by the Fe1-bound OH Ϫ , forming a nickelbound carboxylic acid group. The carboxylic acid group is deprotonated by the Fe1-bound S 2Ϫ , which is ideally situated to act as a catalytic base, resulting in a thiol group and a preformed CO 2 (Fig. 4D). The proton of the thiol group is subsequently transmitted via His-261 to His-93 and Lys-563 of the assumed proton transfer chain (33), whereas the sulfide stays bound to Fe1 (Fig. 4E). Dissociation of the nickel carboxyl generates two electrons that are delocalized on the [3Fe-4S] subcluster and transferred further through clusters B, BЈ, and D to the external electron acceptor. Simultaneously, the Ni-( 2 S)-Fe1 bridge is being reformed (Fig. 4A), and a new reaction cycle can proceed. This mechanism resembles that of CODH Oc from O. carboxidovorans, containing a Cu-( 2 S)-Mo bridge in the active site. In that enzyme, the insertion of CO between the copper, the sulfido-ligand, and the hydroxo-group at molybdenum, thereby forming a thiocarbonate intermediate, has been proposed (7,8).