Structural insight into metallocofactor maturation in carbon monoxide dehydrogenase

The nickel-dependent carbon monoxide dehydrogenase (CODH) employs a unique heterometallic nickel–iron–sulfur cluster, termed the C-cluster, to catalyze the interconversion of CO and CO2. Like other complex metalloenzymes, CODH requires dedicated assembly machinery to form the fully intact and functional C-cluster. In particular, nickel incorporation into the C-cluster depends on the maturation factor CooC; however, the mechanism of nickel insertion remains poorly understood. Here, we compare X-ray structures (1.50–2.48 Å resolution) of CODH from Desulfovibrio vulgaris (DvCODH) heterologously expressed in either the absence (DvCODH−CooC) or presence (DvCODH+CooC) of co-expressed CooC. We find that the C-cluster of DvCODH−CooC is fully loaded with iron but does not contain any nickel. Interestingly, the so-called unique iron ion (Feu) occupies both its canonical site (80% occupancy) and the nickel site (20% occupancy), with addition of reductant causing further mismetallation of the nickel site (60% iron occupancy). We also demonstrate that a DvCODH variant that lacks a surface-accessible iron–sulfur cluster (the D-cluster) has a C-cluster that is also replete in iron but lacks nickel, despite co-expression with CooC. In this variant, all Feu is in its canonical location, and the nickel site is empty. This D-cluster–deficient CODH is inactive despite attempts to reconstitute it with nickel. Taken together, these results suggest that an empty nickel site is not sufficient for nickel incorporation. Based on our findings, we propose a model for C-cluster assembly that requires both CooC and a functioning D-cluster, involves precise redox-state control, and includes a two-step nickel-binding process.

Anerobic carbon monoxide dehydrogenases (CODHs) 4 catalyze the reversible oxidation of CO to CO 2 , enabling certain microbes, such as Rhodospirillum rubrum and Carboxydothermus hydrogenoformans, to live on CO as a sole source of carbon and energy (1,2). This microbial activity accounts for the removal of an estimated 10 8 tons of CO from the lower atmosphere each year, making CODHs an important part of the global carbon cycle (3). Given that CO is a toxic pollutant and a component of fossil fuel emissions, CODH has attracted attention as a possible bioremediation catalyst. Similarly, CODHs also have potential applications in the capture and removal of CO 2 via the Wood-Ljungdahl pathway of carbon fixation, in which the CO that is generated is incorporated into the acetyl group of acetyl-CoA (4). Use of CODH in such capacities would benefit from an ability to produce large amounts of active enzyme. In particular, CODH activity requires a complex heterometallic nickel-iron-sulfur cofactor (termed the C-cluster), the biogenesis of which is poorly understood.
The CODH C-cluster is housed within a homodimeric protein scaffold that contains two additional iron-sulfur clusters, termed the B-and D-clusters, that are used for electron transfer during catalysis (Fig. 1A). The D-cluster, depending on the bacterial species, is either a [4Fe-4S] or [2Fe-2S] cluster that resides at the CODH dimer interface and serves as an electron conduit to external redox partners, such as ferredoxins, whereas the B-cluster is a [4Fe-4S] cluster that mediates electron transfer between the C-and D-clusters (5)(6)(7)(8)(9). The C-cluster is a structurally unique metallocluster composed of a distorted [Ni-3Fe-4S] cubane linked through a sulfide ion (S L ) to a mononuclear iron site (Fe u ) (Fig. 1B) (5,6). This canonical C-cluster architecture is essential for catalysis as it allows for binding of CO at the nickel ion of the cubane, activating it for nucleophilic attack by a water molecule ligated in immediate proximity at Fe u (10 -13). To access this chemistry, organisms require dedicated cellular machinery for C-cluster assembly, similar to the requirements of other complex metalloclusters (14). Our understanding of the C-cluster assembly process, however, remains incomplete. It is still unknown what the biosynthetic origin of the iron-sulfur scaffold of the cluster is, how the nickel-iron-sulfur cluster is assembled, and what roles individual accessory proteins play in this assembly process.
Limited insight into the process of C-cluster assembly has been gleaned from the co-operonic expression of accessory proteins that appear to play roles in cluster maturation, in particular incorporation of nickel. Previous studies have shown that integration of nickel into the C-cluster depends on the accessory protein CooC (15)(16)(17). Certain organisms express additional proteins, CooJ and CooT, that have been implicated in C-cluster maturation; however, CooC appears to be the only dedicated and essential maturation factor expressed by all CODH-containing organisms (15, 18 -22). CooC is a P-loop ATPase with sequence similarity to UreG and HypB, maturation factors involved in nickel transfer to the active sites of urease and nickel-iron hydrogenase, respectively (15,23). In analogy to UreG and HypB, CooC has been proposed to use ATP hydrolysis to facilitate nickel insertion into CODH (15,16,23,24). Alternatively, CooC has been proposed to fold or otherwise mediate formation of the proper nickel-binding site in CODH (16,17,25).
To gain a further understanding of C-cluster maturation and the role of CooC, we have recently developed a means to heterologously express Desulfovibrio vulgaris CODH (DvCODH) in either the presence (DvCODH ϩCooC ) or the absence (DvCODH ϪCooC ) of the D. vulgaris CooC maturase (DvCooC) using Desulfovibrio fructosovorans as an expression host (17). This differential expression results in substantially different enzymatic phenotypes (see Table 1). As-isolated DvCODH ϩCooC binds about half of the expected nickel content and exhibits a lag phase in activity followed by a relatively low rate of CO oxidation (160 mol⅐min Ϫ1 ⅐mg Ϫ1 ) as compared with the previously published activities of monofunctional CODHs from other species, which range from ϳ4400 to 16,000 mol⅐min Ϫ1 ⅐mg Ϫ1 (17,26,27). Incubation of the as-isolated DvCODH ϩCooC with both NiCl 2 and the reductant sodium dithionite results in elimination of the lag phase and a 10-fold increase in CO oxidation activity; however, activation does not occur in the presence of NiCl 2 or sodium dithionite alone (17). In contrast, as-isolated DvCODH ϪCooC contains low amounts of nickel (0 -0.2 nickel/monomer), has nearly no activity (4% of as-isolated DvCODH ϩCooC ), and undergoes limited activation with NiCl 2 and sodium dithionite (17), suggesting that DvCooC is involved in constructing the appropriate nickel-binding site in DvCODH (25).
Interestingly, our previously published crystal structures of DvCODH ϩCooC revealed that the C-cluster adopts an alternative conformation upon exposure to oxygen in which the Ni, Fe u , and S L ions shift by as much as 3 Å, and the Ni and Fe u ions adopt new coordination environments (Fig. 1C) (9). Notably, this oxidized conformation of the C-cluster can be converted back to the canonical, reduced conformation by incubation with reducing agent (9). The oxidized conformation involves ligation by a cysteine residue that is strictly conserved in CODHs but that does not serve as a ligand to the active, reduced conformation of the cluster (9). Mutation Note that the B-cluster of one monomer completes the electron transfer pathway of the opposing monomer. B, the C-cluster in its canonical, reduced state (PDB code 6B6V). C, the oxidized C-cluster (PDB code 6B6W). A lysine residue that completes a distorted tetrahedral coordination geometry around the nickel ion has been omitted for simplicity. D, the C-cluster of DvCODH(C301S) ϩCooC (PDB code 6DC2). Residue numbers correspond to the sequence of DvCODH. Protein is shown in ribbon representation in pink with metalloclusters shown as spheres and sticks with nickel in green, iron in orange, and sulfur in yellow; in B-D, ligating amino acid residue side chains are shown as sticks with sulfur in yellow, nitrogen in blue, and oxygen in red. The structures shown in this figure are described in Ref. 9. Crystallographic studies of C-cluster assembly of this cysteine residue (Cys-301) in DvCODH ϩCooC to serine (C301S) was shown to result in an inactive enzyme that does not bind nickel (see Table 1) (9), similar to previous results on the CODH from Moorella thermoacetica (28). The crystal structure of DvCODH(C301S) ϩCooC revealed a partially assembled C-cluster in which Fe u adopted a split conformation: Fe u was in its canonical binding site at 70% occupancy, and Fe u was incorporated into the cubane portion of the cluster at 30% occupancy, taking up the canonical nickelbinding site (Fig. 1D) (9). This split C-cluster conformation combined with the inability of the DvCODH(C301S) ϩCooC variant to incorporate nickel led us to propose that the oxidized conformation of the cluster could be an intermediate in C-cluster maturation, although how this conformation may participate in the assembly process remained unclear (9).
To further interrogate the process of C-cluster assembly, we have now determined crystal structures of DvCODH produced in the absence of CooC (DvCODH ϪCooC ) and of a DvCODH variant produced in the presence of CooC and engineered to not contain the surface-accessible D-cluster (DvCODH(⌬D) ϩCooC ). Comparison of the DvCODH ϪCooC structure to that of DvCODH(C301S) ϩCooC (9) suggests a possible link between CooC-dependent cluster assembly and the ability to adopt the alternative, oxidized cluster arrangement. Furthermore, removal of the D-cluster leads to formation of an incomplete C-cluster, highlighting the importance of this redox-active iron-sulfur cluster for C-cluster maturation. Combined, these results expand our understanding of C-cluster biogenesis, with an emphasis on the importance of accessing different cluster conformations and redox states.

The C-cluster expressed in the absence of CooC is a [3Fe-4S] cluster with a mobile fourth iron
The CODH from D. vulgaris was expressed heterologously in D. fructosovorans in the absence of the C-cluster maturation factor CooC (DvCODH ϪCooC ), as described previously (17). The preparation of protein that was used for crystallization displayed no detectible CO oxidation activity and contained 0 nickel atoms/monomer and 10 iron atoms/monomer ( Table  1). The crystal structure of DvCODH ϪCooC was determined to 1.50 Å resolution ( Table 2). The structure aligns well (C␣ root-mean-square deviation of 0.18 Å for 1250 C␣ atoms within the CODH dimer) with our previously determined structure of DvCODH ϩCooC (9), and the B-and D-clusters of the enzyme are both present and fully intact. Thus, the over-all structure of DvCODH is retained when expressed in the absence of CooC.
At the C-cluster of DvCODH ϪCooC , the [3Fe-4S] partial cubane portion of the canonical C-cluster is intact and present at full occupancy (Fig. 2), indicating that CooC is not necessary for formation of this part of the C-cluster. Modeling of the Fe u ion, however, was more complicated. When the C-cluster was modeled as a [3Fe-4S]-Fe u cluster at full occupancy, residual positive difference electron density was observed in the open cubane position, indicating the presence of an additional atom ( Fig. 2A). Further, iron anomalous difference maps (Table 2) reveal a shoulder extending from the canonical Fe u -binding site into the cubane position, suggesting the presence of iron at partial occupancy (Fig. 2B). Given the lack of nickel in the sample that was crystallized, the positive difference density in the electron density maps, and the shoulder in the iron anomalous maps, we rationalized that Fe u could be present in a split conformation. Refinement of Fe u with a split conformation revealed that at 80% occupancy, Fe u is in its canonical binding site, ligated by His-266, Cys-302, and a water molecule, whereas at 20% occupancy, Fe u is incorporated into the cubane portion of the cluster and ligated by Cys-519, forming a distorted [4Fe-4S] cluster (Fig. 2C). Similar split Fe u conformations were observed across multiple crystal structures of DvCODH ϪCooC samples that lacked nickel. Interestingly, this split Fe u conformation is similar to what was observed previously in our structure of DvCODH(C301S) ϩCooC (9) (Fig. 1D), a CODH variant that is unable to adopt the alternative, oxidized C-cluster conformation because of the absence of the Cys-301 thiol for coordination to Fe u . Together, the structural similarity between the C-clusters in DvCODH ϪCooC and DvCODH(C301S) ϩCooC (9) suggests a link between Cys-301 and the role of CooC in cluster assembly, perhaps due to a CooC-induced conformational change in which Fe u becomes ligated by Cys-301 (see "Discussion"). Additionally, these data suggest that it is not the lack of an open coordination site for nickel that prevents nickel incorporation into the C-cluster. Although there is some Fe u in the nickel-binding site, there is not enough to explain the inability to reconstitute the C-cluster with nickel.

Reduction of DvCODH ؊CooC induces movement of Fe u into the cubane position
The presence of Fe u at partial occupancy in the nickel-binding site of the C-cluster in both our new structure, DvCODH ϪCooC , and the previous structure, DvCODH(C301S) ϩCooC (9), is intriguing. Notably, the [3Fe-4S] clusters of aconitase and ferredoxins, as well as synthetic model compounds, are wellknown to incorporate exogenous metal into their open This work Crystallographic studies of C-cluster assembly cubane site upon reduction due to the increased nucleophilicity of the open sulfide ions in the reduced state (29 -33). Therefore, we hypothesized that reduction of the immature pre-C-cluster, which in part resembles a [3Fe-4S] cluster, could have led to movement of Fe u from its canonical binding site into the cubane position in some CODH molecules.
To test this hypothesis, crystals of as-isolated DvCODH-ϪCooC were soaked in the reductant sodium dithionite prior to cryo-cooling and X-ray data collection. The structure of reduced DvCODH ϪCooC was determined to 1.72 Å resolution (Table 2), revealing greater incorporation of Fe u into the cubane position relative to the structure of the as-isolated enzyme (Fig. 3, A and B). Here, the Fe u ion resides in its canonical position at 40% occupancy and in the cubane portion at 60% occupancy (Fig. 3C). Together, these data suggest that reduction of the pre-C-cluster before nickel is inserted can lead to mismetallation of the nickel site, and therefore careful control of cluster redox state is likely essential during the C-cluster maturation process in vivo.

Crystallographic studies of C-cluster assembly The D-cluster is necessary for proper C-cluster assembly in the presence of CooC
To test the hypothesis that control of redox state is essential to C-cluster maturation, we sought to disrupt electron transfer between the C-cluster and external redox partners by removal of the solvent-exposed D-cluster, which serves as an electron conduit during CO/CO 2 interconversion. Toward this goal, we designed a DvCODH double-mutant variant in which the D-cluster-ligating cysteine residues (Cys-42 and Cys-45) were replaced with alanine resides to abolish binding of the D-cluster (DvCODH(⌬D)). This variant was expressed in the presence of CooC (DvCODH(⌬D) ϩCooC ) and purified to homogeneity. Similar to DvCODH ϪCooC and DvCODH(C301S) ϩCooC , DvCODH(⌬D) ϩCooC is inactive as-isolated and does not contain appreciable amounts of nickel (Table 1). No increase in activity is observed upon incubation with nickel (Table 1). These observations indicate that the D-cluster is essential for C-cluster maturation. To characterize the impact of a D-cluster deletion on C-cluster architecture, the crystal structure of DvCODH(⌬D) ϩCooC was determined to 2.48 Å resolution ( Table 2). The overall structure aligns well (C␣ root-meansquare deviation of 0.29 Å for 1242 C␣ atoms within the CODH dimer) with that of DvCODH ϩCooC . The structure contains both the B-and C-clusters and confirms that the D-cluster is not present in this protein variant (Fig. 4A). The absence of the D-cluster leads to local disorder, and residues 41-44 could not be modeled (Fig. 4A, inset). At the C-cluster of DvCODH(⌬D) ϩCooC , we observe an intact [3Fe-4S]-Fe u scaffold with Fe u present at 100% occupancy in its canonical binding site (Fig. 4, B and C). This result is consistent with the above-mentioned idea that movement of Fe u into the cubane is induced by reduction and that the D-cluster mediates that reduction. Additionally, it is notable that the C-cluster of DvCODH(⌬D) ϩCooC , in which 100% of Fe u is in the canonical location, cannot be activated by incubation with nickel. Further, the fact that the structure of DvCODH(⌬D) ϩCooC is largely unchanged by D-cluster loss suggests that it is the D-cluster's redox role, rather than a structural role, that is required for nickel insertion.

Discussion
Here we present a series of crystal structures of DvCODH to provide insight into the process of C-cluster assembly and maturation, the mechanisms of which remained largely elusive. Our structures suggest that the C-cluster maturase CooC is primarily involved in nickel insertion rather than in formation of the [3Fe-4S]-Fe u scaffold and reveal that nickel insertion is Ϫ F o (as-isolated)) reveals increased electron density at the canonical nickel-binding site of the cubane (green mesh, contoured to ϩ5) and decreased electron density at the canonical Fe u binding site (red mesh, contoured to Ϫ5) in the structure of reduced DvCODH ϪCooC relative to as-isolated DvCODH ϪCooC . B, iron anomalous difference map (orange mesh, contoured to 6) reveals a strong peak of iron anomalous signal in the canonical nickel-binding site of the cubane (compare with Fig. 2B). C, the C-cluster of DvCODH ϪCooC refined with an alternative conformation of Fe u . At 60% occupancy, Fe u is incorporated into the cubane portion of the cluster and ligated by Cys-519. At 40% occupancy, Fe u is ligated by His-266 and Cys-302 in its canonical binding site. Cys-302 adopts alternative conformations. 2F o Ϫ F c electron density (blue mesh) contoured to 1. Protein is shown in ribbon representation in teal with ligating amino acid residue side chains in sticks; cluster ions shown as spheres and sticks; iron in orange, sulfur in yellow, and nitrogen in blue. Proteins are shown as the C␣ trace of each structure. B-and C-clusters of DvCODH(⌬D) ϩCooC are shown as spheres. B, iron anomalous difference map (orange mesh, contoured to 5) suggests the presence of Fe u at full occupancy in its canonical binding site. C, refinement of DvCODH(⌬D) ϩCooC confirms the location and occupancy of Fe u . 2F o Ϫ F c electron density (blue mesh) contoured to 1. In B and C, protein is shown in ribbon representation in maroon with ligating amino acid residue side chains in sticks; cluster ions shown as spheres and sticks; iron in orange, sulfur in yellow, nitrogen in blue, and oxygen in red.

Crystallographic studies of C-cluster assembly
additionally dependent on the D-cluster, likely because of its role in mediating electron transfer. Together, these findings allow us to propose a model for C-cluster assembly and maturation involving multiple cluster conformations and redox states.
In our structure of as-isolated DvCODH ϪCooC , we observe a largely (80%) intact iron-sulfur scaffold that contains four iron ions and four sulfur ions arranged as a [3Fe-4S]-Fe u pre-Ccluster that lacks nickel. The presence of this prearranged ironsulfur scaffold in the absence of dedicated C-cluster assembly machinery suggests that the [3Fe-4S]-Fe u cluster arrangement can be formed using general iron-sulfur cluster biogenesis pathways, such as the sulfur utilization factor (SUF) nitrogen fixation (NIF) systems, both of which are present in the D. fructosovorans expression host, as well as D. vulgaris itself. Two possibilities for the formation of the iron-sulfur scaffold can be envisioned (Fig. 5A). First, the pre-C-cluster could be inserted in two pieces: a single iron ion inserted into the unique His-266/ Cys-302 site and a [3Fe-4S] cluster inserted into the cubane site. Linkage of Fe u and the [3Fe-4S] cluster via the cubane sulfide (S L ) could occur subsequently (Fig. 5A, upper pathway). Alternatively, the C-cluster-binding site of CODH could become loaded with a [4Fe-4S] cluster that is distorted by CODH concomitant with the insertion step or is acted upon by an unknown maturation factor to remove an iron ion from the cubane, forming Fe u (Fig. 5A, lower pathway). Regardless of the exact assembly mechanism, our data indicate that CooC is not necessary for formation of a four-iron-containing iron-sulfur scaffold and that its primary role is likely in facilitating nickel insertion.
Once the iron-sulfur framework of the C-cluster has been assembled in CODH, nickel insertion can occur to form the fully mature and active cluster. Here we consider two possibilities for nickel insertion. In the first, C-cluster maturation in vivo involves the CooC-dependent insertion of nickel into a preformed [3Fe-4S]-Fe u scaffold that resembles our DvCOD-H ϪCooC structures with Fe u in its canonical site coordinated by His-266 and Cys-302 (Fig. 5B). In the second, nickel is inserted into a [3Fe-4S]-Fe u scaffold in which Cys-301 coordinates Fe u (Fig. 5C), a state that is reminiscent of the metal positions observed in our previous structure of fully oxidized DvCODH ϩCooC (Fig. 1C) (9).
For scenario I (Fig. 5B), the key role of CooC, in addition to nickel insertion, may be to control the redox state of the pre-C-cluster, allowing for nickel insertion without mismetallation of the nickel site. In analogy to metal capture by [3Fe-4S] clus- Crystallographic studies of C-cluster assembly ters in other systems, nickel insertion into the scaffold as shown in Fig. 5B would likely require that the [3Fe-4S] framework be in a reduced state to increase the nucleophilicity of the open cubane site, allowing for binding of exogenous metal (29 -33). In the case of the C-cluster, however, addition of exogenous nickel is likely complicated by the presence of Fe u , which we have shown can migrate into the open coordination site of the reduced cubane (Figs. 2 and 3). In the context of C-cluster assembly, this observation indicates that the redox state of the pre-C-cluster must be tightly regulated to avoid mismetallation. One strategy for ensuring nickel incorporation in vivo could be to couple binding of CooC with cluster reduction, such that cluster reduction occurs just prior to nickel insertion. For example, binding of CooC could in some way facilitate interaction of CODH with a low-potential electron transfer protein, such as a reduced ferredoxin.
Although control of cluster reduction provides one route to prevent mismetallation, the previously reported structure of a fully oxidized C-cluster (Fig. 1C) (9) suggests another possible strategy for avoiding incorporation of Fe u into the cubane in vivo (Fig. 5C). In particular, the position of Fe u in the oxidized cluster, ligated by Cys-301, could represent an alternative binding site in which Fe u is positioned prior to nickel insertion, such that Fe u is not ligated in immediate proximity to the remainder of the [3Fe-4S] scaffold (Fig. 5C, state II). In this model, CooC could be involved in inducing a conformational change in the C-cluster prior to nickel insertion such that Fe u becomes ligated by Cys-301 (Fig. 5C, state I to state II). Given the inability of the C-cluster to incorporate nickel in the absence of the D-cluster, this conformational change could additionally be redox-dependent. In any case, nickel could then be inserted into the His-266/Cys-302 binding site that is normally occupied by Fe u , resulting in formation of the oxidized C-cluster conformation (Fig. 5C, state IIIa). Subsequent reduction, possibly facilitated by a change in reduction potential as a result of nickel binding, would then trigger formation of the active C-cluster via the three-atom migration of Fe u , S L , and nickel that we have described previously and that occurs upon reduction of the oxidized cluster conformation (Fig. 5C, state IIIa to state IV) (9).
With these two proposals in mind (Fig. 5, B and C), we revisited the CODH literature. In addition to our previous characterization of DvCODH(C301S) ϩCooC (9), several additional mutagenesis studies on the CODHs from M. thermoacetica (MtCODH) (28), R. rubrum (RrCODH) (34 -36), and C. hydrogenoformans (ChCODH-II) (37) are better explained by the mechanism shown in Fig. 5C than that in Fig. 5B. First, the proposal in Fig. 5B does not explain why substitution of the noncanonical C-cluster ligand Cys-301 in DvCODH and MtCODH results in inactive CODH variants that lack nickel (9,28), whereas the mechanism in Fig. 5C provides a role for Cys-301 in nickel insertion. Second, substitution of the C-cluster ligating histidine residue with valine (in RrCODH) or alanine (in ChCODH-II) resulted in CODHs with iron contents that were indistinguishable from WT but that were impaired in their ability to incorporate nickel in vivo (34,37). Additional mutagenesis experiments in which each of the canonical C-cluster ligating cysteine residues were mutated to alanine or serine revealed that His-266 and Cys-302 (D. vulgaris number-ing) are in fact the only protein-based ligands to the canonical C-cluster that are necessary for nickel incorporation (34 -37). Together, these data support the hypothesis that the His-266/ Cys-302 site serves as the binding site for nickel during nickel incorporation (Fig. 5C). Second, the kinetics of nickel activation in nickel-deficient RrCODH (produced in the presence of RrCooC) suggest a two-step mechanism in which nickel first binds to the enzyme reversibly and then is seated into its active and stable site (26,36).
Collectively, these findings support a model in which Cys-301 binds Fe u while nickel is first inserted into the His-266/Cys-302 site, followed by rearrangement to form the active C-cluster (Fig. 5C, upper pathway), consistent with structures of the C-cluster that we have observed experimentally (9). We note that an alternative assembly pathway could also involve coordination of Fe u by Cys-301 while nickel is inserted into its canonical site in the iron-sulfur cubane, although such a state has not been observed crystallographically (Fig. 5C, state IIIb) and does not explain the RrCODH mutational data mentioned above. One caveat of the model presented in Fig. 5C is that we have only observed Fe u coordinated to Cys-301 in the oxidized state of the C-cluster (9), whereas the presence of reducing agent is known to be essential for nickel-dependent activation in vitro (17,26). That being said, it has not yet been possible to visualize a nickel-deficient form of DvCODH ϩCooC in either an oxidized or reduced state to know whether Fe u movement occurs and/or is redox-dependent in the absence of nickel.
Overall, our data begin to reveal the requirements for assembly of a fully intact and activatable C-cluster: 1) the C-cluster maturase CooC (this work and Refs. [15][16][17], 2) the D-cluster (this work), and 3) the noncanonical Fe u ligand Cys-301 (9,28). Together, these observations begin to expand our understanding of the complex and tightly regulated process of C-cluster biogenesis. In particular, given the varied metal-binding sites that we have observed within the C-cluster scaffold, the insertion of nickel is not a straightforward process and appears to be more complicated than originally thought.

Cloning and purification of DvCODH ؊CooC and DvCODH(⌬D) ؉CooC
Protein was expressed and purified as described previously (17). Briefly, the D. vulgaris gene encoding CODH (cooS) was cloned into a modified pBGF4 shuttle vector under the control of the promoter of the D. fructosovorans nickel-iron hydrogenase operon and included an N-terminal Strep-tag. For DvCODH(⌬D) ϩCooC , the expression vector also contained the gene for the CooC maturase (cooC), and mutations encoding C42A and C45A were introduced into the cooS gene by sitedirected mutagenesis. To perform mutagenesis by PCR, the HindIII-SacI fragment of the modified pBGF4 plasmid, containing the 5Ј end of cooS, was subcloned into pUC19 to serve as a DNA template. The primers GAACAGACGCCGGCCGCC-AAATTCGCCGAATTGGGCACCACC (forward, mutations underlined) and GGTGGTGCCCAATTCGGCGAATTTGG-CGGCCGGCGTCTGTTC (reverse, mutations underlined) were used to generate the C42A/C45A variant. The mutated

Crystallographic studies of C-cluster assembly
HindIII-SacI fragment was then reintroduced into the HindIII-SacI-digested expression vector. The final mutated plasmid was verified by DNA sequencing. Protein was expressed in D. fructosovorans and purified under strictly anaerobic conditions in a Jacomex anaerobic chamber (100% N 2 atmosphere) by affinity chromatography on Strep-Tactin Superflow resin. Protein concentrations were determined by amino acid analysis at the Centre for Integrated Structural Biology (Grenoble, France). Metal content was analyzed by inductively coupled plasma optical emission spectroscopy. The asisolated samples contained nickel and iron as follows: DvCODH ϪCooC : 0 nickel atoms/monomer, 10 iron atoms/ monomer; DvCODH(⌬D) ϩCooC : 0.02 nickel atoms/monomer, 8.5 iron atoms/monomer. CO oxidation activity was assayed at 37°C by monitoring the reduction of methyl viologen at 604 nm (⑀ ϭ 13.6 mM Ϫ1 ⅐cm Ϫ1 ), as described previously (17). Neither DvCODH ϪCooC nor DvCODH(⌬D) ϩCooC exhibited detectable CO oxidation activity. Reconstitution of either sample with NiCl 2 under reducing conditions did not lead to an increase in activity.

Crystallization of DvCODH variants
DvCODH ϪCooC and DvCODH(⌬D) ϩCooC were crystallized in an N 2 atmosphere at 21°C by hanging-drop vapor diffusion in an MBraun anaerobic chamber. A 1 l aliquot of protein (10 mg/ml in 100 mM Tris-HCl, pH 8) was combined with 1 l of a precipitant solution (200 -275 mM MgCl 2 , 14 -20% PEG 3350) on a glass coverslip and sealed over a reservoir containing 500 -700 l of precipitant solution. Diffraction quality crystals grew in 4 -7 days. The crystals were soaked in a cryo-protectant solution containing 200 mM MgCl 2 , 20 -30% PEG 3350, and 10 -16% glycerol and cryo-cooled in liquid nitrogen. For structures of reduced DvCODH ϪCooC , the crystals were soaked in 250 mM MgCl 2 , 18% (w/v) PEG 3350, 5 mM sodium dithionite for 30 min prior to cryo-protecting and cryo-cooling in liquid nitrogen.

Data collection, model building, and refinement
Diffraction data were collected at the Advanced Photon Source (Argonne, IL) on Beamline 24-ID-C using a Pilatus 6M pixel detector and at a temperature of 100 K. Native and iron peak data were collected on the same crystal for each sample, where applicable. The DvCODH(⌬D) ϩCooC structure was determined and refined using data collected at the iron peak wavelength. The data for DvCODH ϪCooC (as-isolated and reduced) were integrated in XDS and scaled in XSCALE (38). The data for DvCODH(⌬D) ϩCooC were integrated and scaled in HKL2000 (39). All data collection statistics are summarized in Table 2.
Structures were determined by molecular replacement in the program Phaser (40) using our previously published structure of DvCODH (PDB code 6B6V) as a search model. Following molecular replacement, 10 cycles of simulated annealing refinement were performed in Phenix (41) to eliminate existing model bias. Refinement of atomic coordinates and atomic displacement parameters (B-factors) was performed in Phenix, and models were completed by iterative rounds of model building in Coot (42) and refinement in Phenix. Metal cluster geom-etries were restrained during refinement using custom parameter files. In advanced stages of refinement, water molecules were added automatically in Phenix (41) and modified in Coot (42) with placement of additional water molecules until their number was stable. For the DvCODH ϪCooC structures, final stages of refinement included translation-libration-screw (TLS) parameterization with one TLS group per monomer (43). Side chains without visible electron density were truncated to the last atom with electron density, and amino acids without visible electron density were not included in the model. Final models contain the following residues (of 629 total): as-isolated Final refinement yielded models with low free R-factors, excellent stereochemistry, and small root-mean-square deviations from ideal values for bond lengths and angles. The models were validated using simulated annealing composite omit maps calculated in Phenix (41). Model geometry was analyzed using MolProbity (44). Analysis of Ramachandran statistics indicated that each structure contained the following percentages of residues in the favored, allowed, and disallowed regions, respectively: as-isolated DvCODH ϪCooC : 96.7%, 3.0%, 0.3%; reduced DvCODH ϪCooC : 96.9%, 2.8%, 0.3%; and DvCODH(⌬D) ϩCooC : 95.8%, 4.0%, 0.2%. Refinement and geometry statistics are summarized in Table 2. The figures were generated in PyMOL (45). Crystallography packages were compiled by SBGrid (46).