AcsF Catalyzes the ATP-dependent Insertion of Nickel into the Ni,Ni-[4Fe4S] Cluster of Acetyl-CoA Synthase*

Acetyl-CoA synthase (ACS) catalyzes the reversible condensation of CO, CoA, and a methyl-cation to form acetyl-CoA at a unique Ni,Ni-[4Fe4S] cluster (the A-cluster). However, it was unknown which proteins support the assembly of the A-cluster. We analyzed the product of a gene from the cluster containing the ACS gene, cooC2 from Carboxydothermus hydrogenoformans, named AcsFCh, and showed that it acts as a maturation factor of ACS. AcsFCh and inactive ACS form a stable 2:1 complex that binds two nickel ions with higher affinity than the individual components. The nickel-bound ACS-AcsFCh complex remains inactive until MgATP is added, thereby converting inactive to active ACS. AcsFCh is a MinD-type ATPase and belongs to the CooC protein family, which can be divided into homologous subgroups. We propose that proteins of one subgroup are responsible for assembling the Ni,Ni-[4Fe4S] cluster of ACS, whereas proteins of a second subgroup mature the [Ni4Fe4S] cluster of carbon monoxide dehydrogenases.

The cellular maturation of metalloenzymes, previously considered a spontaneous process in vivo, typically depends on a machinery of uptake, storage, processing, and delivery factors (1). How metalloenzymes mature has been investigated for some systems, revealing surprisingly complex maturation pathways (2)(3)(4)(5)(6)(7)(8). Enzymes containing nickel, although still relatively small in number, play critical roles in archaea, bacteria, and eukarya, through which they impact the global hydrogen (Ni,Fe-hydrogenase), nitrogen (urease), and carbon (acetyl-CoA synthase, carbon monoxide dehydrogenase, and methyl-CoM reductase) cycles (9). For most of these enzymes, we now have a good understanding of how nickel is incorporated into the active site (8,10).
The nickel-enzymes acetyl-CoA synthase (ACS) 2 and carbon monoxide dehydrogenase (CODH) are found in a variety of anaerobic microbes, including bacterial sulfate reducers, acetogens, and hydrogenogens, as well as archaeal methanogens and sulfate reducers, where they act as the prime CO 2 and CO converter (11)(12)(13)(14). ACS and CODH can be found as independent monofunctional enzymes in Carboxydothermus hydrogenoformans (15) but are typically found in other microorganisms as protein complexes: in acetogens ACS and CODH form a bifunctional (␣␤) 2 complex, whereas in methanogens they are part of a large (␣␤␥␦⑀) 8 multienzyme complex (11,13,15,16). CODHs catalyze the reversible reduction of CO 2 to CO at the C-cluster, in which a single nickel ion, embedded within a 3Fe-4S scaffold with an additional iron in exo, binds and activates CO and CO 2 for turnover (17)(18)(19). ACS catalyzes the reversible condensation of CO, CoA, and a methyl-cation donated by the methylated corrinoid iron-sulfur protein to form acetyl-CoA (13). ACS depends on a Ni,Ni-[4Fe4S] cluster (also called A-cluster) for activity, in which the two nickel ions have distinct coordinations: the nickel ion distal to the [4Fe4S] cluster (Ni d ) is coordinated by two amide nitrogen atoms and two cysteine thiolates within a Cys-X-Cys motif, holding nickel in a stable square-planar coordination environment (15,20,21). In contrast, the nickel ion proximal to the [4Fe4S] cluster (Ni p ) is weakly bound by three cysteine thiolates, is removable by 1,10-phenanthroline (22), and may be replaced by zinc and copper, thereby inactivating ACS (15,20,21,23,24). Ni p likely adopts different oxidation states during turnover and is the presumed place where substrates are activated and converted (13).
Despite their ubiquity and importance in different metabolisms, little is known about the maturation of CODH and ACS, which both limits our understanding of metalloenzyme maturation and hampers their adaptation for biotechnological processes. CODH activity in vivo depends on an ATPase termed CooC (25)(26)(27)(28). CooC belongs to the MinD-type ATPases, an enzyme family with diverse functions and a synapomorphic KGG signature in their Walker A motif (GKGGhGK(T/S)) containing a characteristic lysine residue (signature lysine) in addition to the P-loop lysine (29). Common to the MinD-type ATPases is an ATP-dependent dimerization, which is central to their function and a prerequisite to hydrolyze ATP (30). CooC binds Ni(II) through the Cys-X-Cys motifs of two monomers that combine to form a metal binding site within the dimer interface (31,32). Additional proteins support the maturation of the CODH in Rhodospirillum rubrum (25); however, these additional proteins are not conserved in bacteria outside the Rhodospirillaceae and may be functionally replaced by non-homologous isofunctional enzymes in other organisms. Although CooC was shown to mature monofunctional CODHs (25)(26)(27)(28), it was proposed that the homologous ATPase AcsF supports the assembly of the C-cluster in the bifunctional ACS-CODH complex in Moorella thermoacetica (33).
In contrast, how the Ni,Ni-[4Fe4S] cluster of ACS is formed in vivo is not known. The purpose of this paper is to provide insights into the role of an ATPase closely related to CooC in this process, which we named AcsF Ch . We provide evidence that AcsF Ch catalyzes the nickel and MgATP-dependent activation of ACS by forming a complex with apoACS that serves as a platform for nickel binding from which active ACS is generated at the expense of ATP.

Results
CooC Proteins Can Be Divided into Three Subgroups-The genome of Carboxydothermus hydrogenoformans contains three genes, annotated as cooC1, cooC2, and cooC3, which encode for proteins of the CooC family (34). The cooC1 and cooC2 genes are located in the same gene cluster as the genes encoding CODH-III and ACS (Fig. 1A). The cooC3 gene is located in proximity to the gene encoding CODH-I.
The InterPro database lists 881 protein sequences as members of the CooC family (35,36). A sequence similarity network analysis of the non-redundant sequences of the CooC proteins indicates that the sequences may be grouped into three different clusters (Fig. 1B). The two largest clusters, hereafter called "CooC-type" cluster and "AcsF-type" cluster, contain proteins typically associated with gene clusters of CODHases or ACS/ CODH. The third cluster contains 116 proteins, mostly found in methanogenic archaea (Methanomicrobia, Methanococci). The CooC-type cluster harbors 298 sequences and includes all CooC proteins shown to be involved in CODH maturation, such as CooC from R. rubrum (CooC Rr ) (26), CooC3 Ch (27), and CooC from Desulfovibrio vulgaris (CooC Dv ) (28). The CooC-type cluster also encompasses CooC1 from C. hydrogenoformans (CooC1 Ch ), which is a nickel-binding ATPase whose crystal structure has been determined (27,31).
The AcsF-type cluster encompasses 309 protein sequences, among them AcsF from M. thermoacetica (AcsF Mt ). AcsF Mt has been described as an ATPase of unknown function whose involvement in CODH maturation of bifunctional ACS/CODH was postulated but not shown (27,33). The AcsF-type cluster also encompasses the gene product of cooC2 from C. hydrogenoformans, which we therefore named AcsF Ch . All clusters include proteins from both bacteria and archaea. Most organisms (63%) containing a gene in the CooC-type cluster also harbor a paralogous gene of the AcsF-type cluster in their genome and only 13% of the organisms with an AcsF-type sequence have no CooC-type protein in the Interpro database.
AcsF Ch Dimerizes in the Presence of ATP, Does Not Bind Nickel and Has Low ATPase Activity-Typical size exclusion chromatograms of AcsF Ch showed peaks at 83.2 and 90.5 ml. The corresponding molecular masses, 57.7 and 28.8 kDa, were determined by comparison to a standard curve and agree with the predicted molecular masses of dimeric (55.4 kDa) and monomeric AcsF Ch (27.7 kDa). The monomer:dimer ratio depended on the cultivation batch rather than on individual purifications. Additionally, two variants of AcsF Ch were expressed and purified. The K10A-AcsF Ch variant, lacking the signature lysine Lys-10, showed similar monomer:dimer ratios for as isolated protein as wild type AcsF Ch , whereas the C108A-AcsF Ch variant was only present as monomer.
We analyzed by size exclusion chromatography (SEC) whether monomeric AcsF Ch dimerizes in the presence of MgADP and MgATP ( Fig. 2A). Although the presence of MgADP had no influence on the chromatogram, we found AcsF Ch in the presence of MgATP to be predominantly dimeric, as expected for a MinD-type ATPase. The monomeric fraction of the K10A-AcsF Ch variant remained in the monomeric state independent of the presence of MgADP or MgATP (Fig. 2B).  AcsF Ch shares with the other CooC proteins the CXC sequence motif. In CooC1 Ch the two Cys residues of this motif are able to bind nickel, which can be followed by an increased absorption in the region 310 -350 nm originating from a ligand to metal charge transfer between Cys-S Ϫ and Ni(II) (31). The UV/visible absorption spectrum of AcsF Ch had a maximum at 274 nm. When we added increasing amounts of NiCl 2 to AcsF Ch , the UV/visible spectral features remained unchanged. Furthermore, when we titrated NiCl 2 to AcsF Ch in an anoxic isothermal titration calorimetry (ITC) experiment, we only observed heats of dilution. Thus, in contrast to CooC1 Ch , we could not detect specific nickel binding by AcsF Ch .
The amino acid sequence of AcsF Ch contains the characteristic Walker A signature of the MinD-type ATPase. We determined the specific ATPase activity of AcsF Ch with a coupled assay and a colorimetric assay. Determined specific activities varied with each purification batch and averaged at 7.4 nmol min Ϫ1 mg Ϫ1 at 25°C and 25.5 nmol min Ϫ1 mg Ϫ1 at 45°C in the coupled assay. In the colorimetric assay the average activities measured were 3.9 nmol min Ϫ1 mg Ϫ1 at 25°C, 20.4 nmol min Ϫ1 mg Ϫ1 at 45°C, and 56.5 nmol min Ϫ1 mg Ϫ1 at 60°C. In contrast, we could not detect ATPase activity above background for the K10A-AcsF Ch variant. There is an intrinsic risk that low ATPase activities, such as those determined for AcsF Ch , could arise from minor protein impurities of high specific activities and not the protein investigated. However, all AcsF Ch preparations tested had substantially higher activities than the preparation of the K10A-AcsF Ch variant, which was purified using the same protocol. Furthermore, the ATPase activity increased until 60°C, which is expected for an ATPase of a hyperthermophilic organism such as C. hydrogenoformans. We thus conclude that AcsF Ch is an ATPase of low activity.
AcsF Ch Forms a Complex with ApoACS-A maturing metalloenzyme typically forms a specific complex to receive metal(s) from its accessory protein(s). We therefore tested by SEC whether AcsF Ch was able to form a stable complex with apoACS. When we loaded a preincubated 1:2 mixture of apoACS and AcsF Ch (1:1 of apoACS and dimeric AcsF Ch ) on the column, the resulting chromatogram revealed a new peak at 70.2 ml, which was clearly separated from the peaks of apoACS and AcsF Ch (Fig. 3A) and corresponded to a molecular mass of 199 kDa. Because apoACS and AcsF Ch were incubated in a 1:2 stoichiometry, and no further peaks appeared in the chromatogram, we assume that the complex is very stable and has a 1:2 stoichiometry of apoACS and AcsF Ch (141 kDa). SDS-PAGE of the peak fractions showed two bands corresponding to ACS (86.6 kDa) and AcsF Ch (27.7 kDa). This confirms the presence of an ACS-AcsF Ch complex (Fig. 3B). Densitometric analysis of the SDS gel indicated a 1:1.8 stoichiometry, confirming a complex of one molecule of ACS and two molecules of AcsF Ch . The complex was formed regardless of whether ACS was incubated with dimeric or monomeric AcsF Ch preparation.
CooC1 Ch and CooC3 Ch belong to the CooC-type proteins ( Fig. 1) and share 28.6 and 28.3% amino acid sequence identity with AcsF Ch , respectively. Despite the similarity, apoACS did not form a stable complex with CooC1 Ch or CooC3 Ch , indicating that its interaction with AcsF Ch is specific (Fig. 3C).
Furthermore, we determined the ability of the AcsF Ch variants K10A-AcsF Ch and C108A-AcsF Ch to form a complex with apoACS ( Fig. 3D). Although the K10A-AcsF Ch variant formed a complex with apoACS, we did not observe a complex between apoACS and the C108A-AcsF Ch , indicating that Cys-108 is involved in complex formation. ApoACS Can Be Activated by AcsF Ch , NiCl 2 , and MgATP-The UV/visible spectrum of ACS had the characteristic absorption shoulder of a [4Fe4S] 2ϩ cluster at 420 nm. Metal analysis by inductively coupled plasma optic emission spectrometry (ICP-OES) confirmed the presence of iron in as isolated ACS, whereas nickel could not be detected (Table 1). In agreement with the lack of nickel, the enzyme is inactive. Therefore, the as isolated ACS is designated as apoACS.
The lack of nickel and catalytic activity in apoACS served as a starting point to investigate the maturation of ACS. When we added stoichiometric amounts of nickel, MgATP, and AcsF Ch to ACS, it reached within 2 h a specific activity of 70 nmol min Ϫ1 mg Ϫ1 (Fig. 4A). In contrast, when we omitted AcsF Ch from the mixture, the activity of ACS increased only marginally. Activation of ACS by AcsF Ch strictly depended on MgATP addition. Thus, AcsF Ch catalyzes the Ni 2ϩ and MgATP-dependent maturation of ACS. The process is catalytic as also substoichiometric amounts of AcsF Ch (1:20 AcsF Ch :apoACS) sufficed to activate apoACS.
In a next step, we investigated how the amino acid exchanges K10A and C108A influence the capability of AcsF Ch to activate apoACS. Neither the ATPase-deficient variant K10A-AcsF Ch , nor the C108A-AcsF Ch variant was able to activate apoACS (Fig. 4B). This indicates that the ATPase activity of AcsF Ch , as well as the presence of the CXC motif, are essential for the activation process.
AcsF Ch is a specific maturation factor of ACS, as neither CooC1 Ch nor CooC3 Ch can replace AcsF Ch in the activation assay (Fig. 4B). Even though the data may indicate a slight increase of ACS activity over time (Fig. 4B, inset), the determined activities of ACS are within the error of the activity assay. They remained within the range detected in the absence of AcsF Ch and are more than 37 times lower than when incubated with AcsF Ch . Thus only AcsF Ch , not CooC1 Ch or CooC3 Ch , acts as a specific maturation factor of ACS.
The ACS-AcsF Ch Complex Binds Nickel Stoichiometrically and Can Be Subsequently Activated by MgATP-We determined whether there is a difference between the nickel binding properties of apoACS and the apoACS-AcsF Ch complex. Therefore, we titrated the proteins with nickel using ITC and analyzed the metal contents of these samples by ICP-OES after removal of excess nickel. When we titrated apoACS with NiCl 2 , we observed substoichiometric nickel binding; the ICP-OES metal analysis detected only 0.43 nickel atoms per 4 Fe atoms in nickel-titrated apoACS ( Table 1). The apoACS-AcsF Ch complex, by contrast, binds nickel stoichiometrically; the ICP-OES analysis detected 1.85 nickel atoms. Thus, whereas apoACS alone cannot bind Ni 2ϩ stoichiometrically, the apoACS-AcsF Ch complex is able to bind approximately two Ni 2ϩ ions ( Table 1).
As the apoACS-AcsF Ch complex has the full complement of metal ions after titration with nickel, we determined its specific activity. Despite the presence of two nickel ions, the nickel-bound apoACS-AcsF Ch complex lacked ACS activity. But when we removed excess Ni 2ϩ and incubated the nickel-bound apoACS-AcsF Ch complex with 10 mM MgATP (1 h, 60°C), it gained an activity of 67 nmol min Ϫ1 mg Ϫ1 , the same specific ACS activity as observed when all components were present in our activation assay. Thus, MgATP is needed to convert the active site in the ACS-AcsF Ch complex from an inactive nickelbound to an active nickel-bound state.

Role of AcsF Ch in Ni,Ni-[4Fe4S]
Cluster Formation-Our analysis allows first mechanistic conclusions about how AcsF Ch supports the assembly of the Ni,Ni-[4Fe4S] cluster of ACS (Fig.  5). In the first step, two AcsF Ch monomers or one AcsF Ch dimer form a complex with apoACS. Complex formation creates at least one additional high-affinity nickel binding site in the complex, as only the apoACS-AcsF Ch complex binds the full complement of nickel when presented with low physiological concentrations: we detected no affinity for AcsF Ch to bind nickel; apoACS binds less than one nickel ion when titrated with nickel; however, the apoACS-AcsF Ch complex binds two nickel ions under the same conditions. Thus, complex formation likely FIGURE 3. Complex formation of apoACS with AcsF Ch . A, complex formation monitored by SEC. The apoACS-AcsF Ch sample was prepared by incubating 13 nmol of apoACS with 13 nmol of dimeric AcsF Ch for 10 min. For comparison, apoACS and dimeric AcsF Ch were also loaded individually on the column. The elution profile of the apoACS-AcsF Ch incubation mixture is shown as dotted line and the elution profiles of the individual proteins apoACS and AcsF Ch are shown as dashed and solid lines, respectively. Gray arrows mark the elution volumes of the molecular weight standards. The apoACS-AcsF Ch complex elutes at 70.2 ml, whereas apoACS and AcsF Ch elute at 76.0 and 83.2 ml, respectively. Complex formation of apoACS with AcsF Ch was observed in more than five independent samples. B, SDS-PAGE analysis of SEC experiments. From each chromatography run, the main peak fractions were loaded on a gel (apoACS-AcsF Ch complex, 67 to 76 ml; apoACS, 72.5 to 80 ml; AcsF Ch , 80.5 to 85.5 ml). Lanes 1, 5, 9, and 12 show a marker with the indicated molecular masses. C, CooC1 Ch , AcsF Ch , or CooC3 Ch were individually incubated with apoACS and complex formation was analyzed by SEC. The elution profiles of the incubation mixtures of apoACS with CooC1 Ch , AcsF Ch , and CooC3 Ch are shown as solid, dashed, and dotted lines, respectively. Complex formation experiments of apoACS with CooC1 Ch and CooC3 Ch were carried out with two independent samples. D, complex formation of the AcsF Ch variants K10A-AcsF Ch and C108A-AcsF Ch with apoACS. Either monomeric K10A-AcsF Ch or monomeric C108A-AcsF Ch was incubated with apoACS and complex formation was followed by SEC. The elution profiles of apoACS incubated with K10A-AcsF Ch and C108A-AcsF Ch are shown as dashed and dotted lines, respectively. Complex formation experiments of apoACS with K10A-AcsF Ch and C108A-AcsF Ch were carried out with two independent samples. induces conformational changes, creating new binding sites for nickel, in apoACS, AcsF Ch , or at the interface between both.
When the Cys-X-Cys motif in AcsF Ch is altered, the complex does not form, suggesting that it is part of the ACS-AcsF Ch interface. It is tempting to assume that the presence of strictly conserved Cys-X-Cys motifs in both AcsF Ch and ACS is not purely coincidental (37), but serves to have binding sites with similar affinity for Ni 2ϩ in both proteins.
Once nickel is bound to the apoACS-AcsF Ch complex, MgATP is required to activate ACS. Although MgATP binding and its hydrolysis are key to the activation, their impact on the bound nickel ions remains to be defined. Furthermore, it should be noted that the observed rate constant of apoACS activation of 1.47 h Ϫ1 is substantially lower than expected from the determined rate of ATP hydrolysis by AcsF Ch alone.
Two Classes of CooC Proteins Are Responsible for the Maturation of ACS and CODH-Despite the sequence similarity between AcsF and the paralogous CooC1 and CooC3 proteins of C. hydrogenoformans, it is obvious that the CooC proteins, as defined by InterPro, can be grouped in at least three classes (Fig.  1). We suggest that enzymes clustered in the CooC-type group may act in the maturation of the [Ni4Fe4S] cluster of CODHs, whereas those that belong to the AcsF-type group, such as AcsF Ch and AcsF Mt , catalyze the MgATP and nickel-dependent maturation of the Ni,Ni-[4Fe4S] cluster of ACS. A possible division of selected CooC proteins in two classes has also been proposed based on a phylogenetic tree (27).
CODH and ACS differ not only in their function and overall fold, but their active sites differ in (i) composition: CODH has 1 nickel ion per 4 iron ions, whereas ACS has 2 nickel ions; (ii) architecture: in CODH nickel is integrated into an FeS scaffold, whereas in ACS the 2 nickel ions are bound next to a [4Fe4S] cluster; and (iii) accessibility: in CODH the Ni,Fe-cluster is connected to the surface via small channels, whereas the Ni,Fecluster of ACS is surface exposed in at least one conformation of the protein (21). It is surprising that despite all these differences the maturases involved in nickel-and MgATP-dependent activation of the enzymes belong to the same ATPase family, have remarkably similar amino acid sequences, and likely evolved from a common ancestor.
Comparison with Homologous NTPases-AcsF Ch is homologous and functionally related to other NTP-hydrolyzing maturation factors, which act, for example, in the assembly of ureases (UreG), Ni,Fe-hydrogenases (HypB), nitrogenases (NifH), and cytosolic Fe/S proteins (Cfd1/Nbp35). UreG and HypB are homologous GTPases that are essential for the maturation of ureases and hydrogenases, respectively. Both are like AcsF Ch   (1), which contains the [4Fe4S] cluster, but no nickel, can form a complex with either one dimer or two monomers of AcsF Ch . The apoACS-AcsF Ch complex (2) is then able to bind two nickel ions. The nickel ions may bind either sequentially (3) or simultaneously. However, neither the details of the nickel binding mechanism nor the location of the nickel binding site are known. ATP hydrolysis by AcsF Ch converts inactive (4) to active ACS (5).
SIMIBI NTPases, but belong to the G3E family and carry the signature motif (ESGG) and the guanine specificity loop (NKTD) characteristic for this family (29,38). In contrast to AcsF Ch , they act together with other maturation factors to transfer a metal to a target protein: UreG forms a complex with the additional factors UreD (UreH), UreE and UreF to facilitate apo-urease maturation (8,39,40), whereas HypB is believed to process and deliver Ni 2ϩ in cooperation with HypA and SlyD to the large subunit of Ni,Fe-hydrogenase (7, 10, 41, 42). In contrast, AcsF Ch is able to mature apo-ACS in vitro without additional proteins required.
The CooC/AcsF proteins are not the only MinD-type ATPases involved in metalloenzyme maturation: they are joined by Cfd1/Nbp35 and NifH. Cfd1/Nbp35 is a heterodimeric ATPase essential for cytosolic iron-sulfur cluster maturation (43), whereas NifH is involved in molybdenum-homocitrate transfer to the maturing Mo,Fe-cluster on NifEN (44). Mutagenesis studies indicate that Cys residues in the dimer interface of Cfd1/Nbp35 are involved in coordinating the [4Fe4S] cluster to be transferred (45). Although NifH and Cfd1/ Nbp35 are part of a larger maturation machinery, they seem capable, just like AcsF Ch , of directly transferring a metal or metal cluster to a maturing metalloenzyme. It is thus tempting to assume that the prototypical conformational changes in the dimer interface of MinD-type ATPases upon ATP hydrolysis are used by Cfd1/Nbp35, NifH, and AcsF Ch to transfer one or more metal ions from one protein to another.
ApoACS can also be activated by incubating it for several hours with an excess of nickel (46 -48). However, in common with many other metalloenzymes, under physiological conditions the maturation of apoACS requires an accessory enzyme, which can mature apoACS without the need for high nickel concentrations and long incubation times. In contrast to other metalloenzymes, the maturation of the A-cluster may require only one specific maturation factor, which gives us the unique opportunity to study the assembly of a complex metal cluster in a well defined system in vitro.

Experimental Procedures
Bioinformatic Methods-881 sequences were extracted from InterPro entry IPRO14433 (35,36). The sequence of CooC from D. vulgaris (CooC Dv ) was added, because it was missing in the InterPro entry. After removal of redundant (identical) sequences 723 sequences remained. These sequences were aligned using MAFFT (49) and a distance matrix was calculated from the multiple sequence alignment using protdist/PHYLIP (50) using the Kimura distance as a criterion for similarity. The distance cutoff used to define an interaction in the sequence similarity plot was 1.3. Cytoscape 3.3.0 was used for clustering and displaying the sequence similarity network using the yFiles organic layout provided in Cytoscape (51).
Chemicals-All chemicals were at least of analytical grade and purchased from Sigma, AppliChem, or Roth. Restriction enzymes, DNA polymerases, and T4 DNA ligase were purchased from New England Biolabs, Fermentas (Thermo Scientific), or Agilent Genomics. N 2 , N 2 /H 2 (95/5%), and CO were purchased from Air Liquide. Chromatography columns were purchased from GE Healthcare or IBA. Bottles for metal-free buffers were rinsed with dilute HCl and then washed with Mil-liQ/H 2 O. Chelex 100 resin (Bio-Rad) was added to all metalfree buffers (5 to 10 g liter Ϫ1 ). All anaerobic solutions were prepared at a Schlenk line in a bottle equipped with a butyl rubber septum by at least four cycles of evacuating and purging with N 2 . CO-saturated buffer was prepared by bubbling CO through the buffer for at least 10 min. Ti(III)citrate was prepared as previously described (52). Methylcobinamide was synthesized from methylcobalamin using the method from Zou et al. (53) except that it was purified with a SepPak C 18 reversed phase column (Waters).
Cloning, Mutagenesis, and Expression-The cooC2 gene was amplified by PCR from genomic DNA of C. hydrogenoformans Z-2901 using Pfu DNA polymerase with primer C2_fw with an NdeI restriction site and C2_rv with an Eco31I restriction site (primer sequences are listed in Table 2), which produces a BamHI compatible ligation site. The PCR product was digested with Eco31I and NdeI and then ligated into an NdeI/BamHIdigested pET11a vector (Novagen). The ligation product was named pET11a_AcsF. From this plasmid, the cooC2 gene was cloned into a modified pET28a vector with an N-terminal streptag and a tobacco etch virus (TEV) protease cleavage site, named pET28aTEVstrep, via polymerase incomplete primer extension (PIPE) cloning (54,55). For PIPE cloning the vector PCR contained 0.1 ng of template, 1 M of each primer (PIPE_ vec_fw and PIPE_vec_rv; Table 2), and 25 l of Phusion High Fidelity PCR master mix in a 50-l reaction volume. The DNA was amplified using the Mastercycler pro PCR machine (Eppendorf) with the following protocol: 2 min initial denaturation at 98°C, 35 cycles of 98°C for 10 s, 65°C for 10 s, and 72°C for 2 min, followed by immediate cooling to 4°C. The PCR product was purified and aliquots were stored at Ϫ20°C. Using appropriate primers and templates, the set up for the insert

Name Sequence
C2_fw 5Ј-GGA ATT CCA TAT GGC CTT TAA AAT TGC GGT TG-3Ј  C2_rv  5Ј-CAT GGT CTC GGA TCC TCA GAT AAT ACC CGC ATT TG-3Ј  C3_fw  5Ј-GGA ATT CCA TAT GCA GGT GTC TG-3Ј  C3_rv  5Ј-CCC AAG CTT CTA ACA ATT CTG CC-3Ј  PIPE_vec_fw  5Ј-GGA TCC GAA TTC GAG CTC CGT CG-3Ј  PIPE_vec_rv  5Ј-CAT ATG GCC CTG GAA ATA CAA GTT TTC GGT CG-3Ј  PIPE_C2_fw  5Ј-CTT GTA TTT CCA GGG CCA TAT GGC CTT TAA AAT TGC GGT TGC AG-3Ј  PIPE_C2_rv  5Ј-CGA CGG AGC TCG AAT TCG GAT CCC TAG ATA ATA CCC GCA TTT TGT AAA ATT TTC-3Ј  PIPE_ACS_fw  5Ј-CTT GTA TTT CCA GGG CCA TAT GAG CGA AGT TAT TAA TTT TGA TCA AAT TTT TG-3Ј  PIPE_ACS_rv  5Ј-CGA CGG AGC TCG AAT TCG GAT CCT TAG AGT AGT GGC TCC ATG GAA AGA GCT GG-3Ј PCR was the same as for the vector PCR. The reaction mixture was treated as follows: 2 min initial denaturation at 98°C, 35 cycles of 10 s at 98°C, 10 s at 62°C, and 12 s at 72°C, followed by immediate cooling to 4°C. After purification of the insert, 1 l of vector was annealed with 1 l of insert for 1 min on ice and then directly transformed into Escherichia coli DH5␣. The plasmid was named pET28a_AcsFstrep. K10A-AcsF Ch and C108A-AcsF Ch variants were prepared using the QuikChange method following standard procedures. The cooC3 gene was amplified by PCR from genomic DNA of C. hydrogenoformans Z-2901 using Herculase II DNA fusion polymerase with primer C3_fw with an NdeI restriction site and C3_rv with a HindIII restriction site ( Table 2). The PCR product was digested with the respective restriction enzymes and ligated into the original pET28a vector and the modified pET28aTEVstrep vector. The latter plasmid was named pET28a_C3strep. The acsB gene was cloned into the pET28aTEVstrep vector from genomic DNA of C. hydrogenoformans Z-2901 (10 ng) using PIPE cloning as described above with primers PIPE_ACS_fw and PIPE_ACS_rv ( Table 2). The plasmid was named pET28a_ACSstrep. The pET28a_ACSstrep plasmid was also digested with NdeI and BamHI and the digestion product was ligated into a pET28a_twinstrep vector, which has a twin-streptag and a TEV protease cleavage site. The plasmid was named pET28a_ ACStwinstrep. All plasmids were transformed into E. coli DH5␣ and verified by nucleotide sequencing (Eurofins Genomics). For expression, all plasmids were transformed into E. coli BL21(DE3). Cells containing the pET28a_AcsFstrep or pET28a_C3strep plasmid were grown aerobically in LB medium containing 50 g ml Ϫ1 of kanamycin at 37°C. After the optical density at 600 nm (A 600 ) reached 0.7 Ϯ 0.1, the culture was induced with 0.2 mM isopropyl ␤-D-thiogalactopyranoside and the temperature was decreased to 28°C. Cells were harvested 16 to 24 h after induction and washed with 50 mM Tris-HCl, pH 8.0, 100 mM NaCl, 2 mM tris(2-carboxyethyl) phosphine, and 2 mM EDTA. Cell pellets were frozen in liquid N 2 and stored at Ϫ80°C. Cells with the pET28a_ACSstrep or pET28a_ACStwinstrep plasmid were cultivated in modified TB medium containing 50 g ml Ϫ1 of kanamycin in 5-liter bottles in a water bath at 37°C. The culture was stirred aerobically and after 1 h the medium was supplemented with 0.1 mM FeSO 4 and 0.5 mM cysteine. When the A 600 reached 0.6 Ϯ 0.1 cultures were induced with 0.5 mM isopropyl ␤-D-thiogalactopyranoside and transitioned to anaerobic growth by closing the bottles with a butyl rubber septum. After 24 h cells were quickly harvested aerobically, frozen in liquid N 2 , and stored at Ϫ80°C. Purification-All purifications were carried out at room temperature in an anaerobic glove box (model B; COY Laboratory Products) using metal-free buffers. CooC1 Ch was prepared as previously described (31). For AcsF Ch , 5 to 10 g of frozen cells were resuspended in buffer A (50 mM Tris-HCl, pH 8.0, 100 mM NaCl, 2 mM tris(2-carboxyethyl)phosphine) with a small amount of avidin and lysozyme. The cell suspension was sonicated (Bandelin Sonoplus 2200) three times for 5 min each (5 ϫ cycle, 50% amplitude) in a rosette cooling cell on ice. It was then centrifuged in a polycarbonate bottle with cap assembly at 142,400 ϫ g for 1 h at 12°C. The supernatant was loaded on a StrepTactin-Sepharose High Performance column (10 -30 ml) equilibrated in buffer A. After washing the column with 3 column volumes of buffer A, AcsF Ch was eluted with a linear gradient of 0 -2.5 mM desthiobiotin in the same buffer. The streptag of AcsF Ch was cleaved by incubation with a streptagged TEV protease for ϳ16 h. The protein solution was applied to a PD-10 desalting column and was then reloaded on the StrepTactin column equilibrated with buffer A. The flowthrough was collected, concentrated, and loaded on a Superdex 200 Hiload 16/60 column equilibrated in buffer B (50 mM Tris-HCl, pH 8.0, 100 mM NaCl). Fractions corresponding to monomeric and dimeric AcsF Ch were pooled, concentrated, frozen in liquid N 2 (in glass vials equipped with a PTFE/silicone septum), and stored at Ϫ80°C. For K10A-AcsF Ch , C108A-AcsF Ch , and CooC3 Ch , the purification procedure was carried out as described above. ACS was prepared following the same basic strategy except that buffers did not contain tris(2-carboxyethyl)phosphine and the streptag or the twin-streptag was not cleaved. Cells containing the pET28a_ACSstrep or pET28a_ACStwinstrep plasmid were lysed, centrifuged, and captured on the streptactin column as described above. The protein was then eluted with desthiobiotin, concentrated, and directly loaded on the Superdex 200 size exclusion column. The fractions corresponding to monomeric ACS were collected, concentrated, frozen in liquid N 2 (in glass vials equipped with a PTFE/silicone septum), and stored at Ϫ80°C. UV/Visible Spectroscopy-UV/visible spectra were obtained inside the glove box, using an Agilent 8453 photodiode array spectrophotometer with a Peltier temperature controller at 25°C. Black walled quartz cuvettes with 1-cm path length were used to record spectra.
Measurements of ATPase Activity-To measure the ATPase activity of AcsF Ch a coupled assay using pyruvate kinase and lactate dehydrogenase was performed (56). In this assay the conversion of one molecule of NADH to NAD ϩ by lactate dehydrogenase equals the consumption of one molecule of ATP by AcsF Ch . The rate of ATP hydrolysis was determined by measuring the decrease of absorption at 340 nm for 10 min and calculating the reaction rates using the extinction coefficient of NADH ⑀ 340 nm ϭ 6220 M Ϫ1 cm Ϫ1 . The assay solution was prepared in 50 mM HEPES, pH 7.2, containing 150 M KCl, 10 mM MgCl 2 , 2.3 mM phosphoenolpyruvate, 140 M NADH, 9 units of lactate dehydrogenase, 6 units of pyruvate kinase, and 1 mM MgATP. The assay was started by addition of 5 M AcsF Ch . This assay was performed at two different temperatures, at 25 or 45°C. The velocities obtained by the coupled assay were confirmed with a modified malachite green ammonium molybdate assay (57). This assay was performed at 25, 45, and 60°C. In this assay the amount of phosphate released by the hydrolysis of ATP is detected by following the increase of absorption at 630 nm, which is due to the formation of a malachite green phosphomolybdate complex. The rate calculation was based on a standard curve, which was prepared with KH 2 PO 4 in the range from 1 to 9 M. The assay solution contained 5 M AcsF Ch , 10 mM MgCl 2 , and 1 mM MgATP in 0.1 mM HEPES, pH 7.2. Aliquots were taken at different time points, incubated with the malachite green ammonium molybdate mixture and 34% (w/v) sodium citrate for 15 min, and then the absorption was measured at 630 nm. Control experiments in the absence of AcsF Ch were performed to assess the rate of autohydrolysis of ATP at different temperatures.
Analysis of Complex Formation Behavior between ACS and AcsF and CooC Proteins-ApoACS and AcsF Ch were incubated for 10 min inside the glove box and loaded on a Superdex 200 HiLoad 16/60 column equilibrated in 50 mM Tris-HCl, pH 8.0, 100 mM NaCl. The flow rate was 1 ml min Ϫ1 . Elution profiles were recorded by following the absorption at 280 nm. As a control apoACS and AcsF Ch were also loaded individually on the column. From these three chromatography runs, the fractions were analyzed by SDS-polyacrylamide gel electrophoresis. The stoichiometry of the complex was analyzed with the software GelAnalyzer 2010a. Complex formation of apoACS with K10A-AcsF Ch , C108A-AcsF Ch , CooC1 Ch , and CooC3 Ch was also analyzed by size exclusion chromatography as described for wild type AcsF Ch . A calibration of the Superdex HiLoad 16/60 column was performed with the following molecular mass standards: aprotinin (6.5 kDa), RNase A (13.7 kDa), DNase I (31 kDa), ovalbumin (44 kDa), conalbumin (75 kDa), lactate dehydrogenase (140 kDa), aldolase (161 kDa), and catalase (250 kDa). The decadic logarithm of the molecular weight was plotted against the elution volume and the calibration curve was determined by linear regression.
Reconstitution of ApoACS-All incubation and reaction mixtures were set up inside an anaerobic glove box (model B; COY Laboratory Products). ApoACS was reconstituted by incubating 20 M apoACS with 40 M NiCl 2 , 40 M AcsF Ch , 10 mM MgCl 2 , and 10 mM MgATP in 0.1 M HEPES, pH 7.2, at 60°C. Four control experiments were set up; in each case, one component of the incubation mixture was missing. Reconstitution experiments were also performed with K10A-AcsF Ch , C108A-AcsF Ch , CooC1 Ch , and CooC3 Ch instead of AcsF Ch . Aliquots were taken at different time points and the rate of acetyl-CoA formation was analyzed by UV/visible spectroscopy.
Measurements of Acetyl-CoA Formation Activity-Acetyl-CoA synthesis activity of ACS from CO, methylcobinamide, and CoA was determined by following the conversion of methylcob(III)inamide (MeCbi) to cob(I)inamide (Cbi). The change of absorption was measured at 387 and 462 nm. The reaction rates were calculated using the differences of the extinction coefficients, ⌬⑀ 387 nm(MeCbi-Cbi) ϭ Ϫ17,410 M Ϫ1 cm Ϫ1 and ⌬⑀ 465 nm(MeCbi-Cbi) ϭ 7,840 M Ϫ1 cm Ϫ1 , which were calculated from spectra of Robertson et al. (58). The reaction was set up inside the glove box as follows: 0.1 M HEPES, pH 7.2, 50 M MeCbi, and 300 M Ti(III)citrate were placed in a screw cap cuvette; the ACS incubation mixture (see ACS reconstitution) was then added to yield a final concentration of 2 M ACS. After closing the cap, CO-saturated buffer (10% of the final volume) was injected with a syringe and the reaction was started by adding 200 M CoA. All measurements were performed in darkness.
Isothermal Titration Calorimetry-ITC experiments were performed inside a glove box (LABstar, MBRAUN) with a MicroCal VP-ITC system (GE Healthcare) at 25°C in 50 mM Tris, 100 mM NaCl. Three different experiments were performed: (a) 15 M AcsF Ch were titrated with 300 M NiCl 2 , (b) 24.7 M apoACS were titrated with 750 M NiCl 2 , and (c) 24.4 M ACS-AcsF Ch complex were titrated with 750 M NiCl 2 . The concentrations of apoACS and ACS-AcsF Ch complex were normalized to the amount of iron, which was determined after ITC measurements by ICP-OES. The enzyme was present in the reaction cell (1.4 ml) and NiCl 2 was in the syringe. The initial injection was 3 l (lasting 6 s) and the 47 subsequent injections were 6 l (lasting 12 s). There was an interval of 300 s between injections. The solution in the reaction cell was stirred at 307 rpm.
Removal of Excess Metal from ITC Samples-When the ITC experiment was finished, samples b and c were used for further experiments, i.e. activity measurements and metal analysis. The protein samples were concentrated to ϳ300 l and the excess metal was removed using a PD MiniTrap-G25 column. Only colored fractions were collected. To ensure that all excess metal was removed a buffer sample was prepared for metal analysis: 130 M NiCl 2 were added to the buffer and the sample was treated exactly the same as the protein samples.
Metal Analysis-Metal contents were analyzed using a PerkinElmer Optima 2100 DV ICP-OES spectrometer. Samples included two different purifications of apoACS and ITC, samples b and c, which were treated as described above. 500-l protein samples were wet-washed overnight in a 1:1 mixture with 65% nitric acid (Suprapur, Merck, Darmstadt, Germany) at 100°C. The samples were diluted with 4 ml of H 2 O prior to their injection into the ICP-OES. The multielement standard solution XVI (Merck) was used as a reference.