A Universal Scaffold for Synthesis of the Fe(CN)2(CO) Moiety of [NiFe] Hydrogenase*

Background: The active site iron of [NiFe] hydrogenases is equipped with a carbonyl ligand (CO) and two cyanides (CN−). Results: A complex of the hydrogenase accessory proteins HypC and HypD contains both CN− and CO ligands. Conclusion: The entire Fe(CN)2(CO) moiety is assembled on a scaffold and subsequently transferred to apo-hydrogenase. Significance: Scaffold-assisted cofactor assembly is a common trait of hydrogenases and other metalloenzymes. Hydrogen-cycling [NiFe] hydrogenases harbor a dinuclear catalytic center composed of nickel and iron ions, which are coordinated by four cysteine residues. Three unusual diatomic ligands in the form of two cyanides (CN−) and one carbon monoxide (CO) are bound to the iron and apparently account for the complexity of the cofactor assembly process, which involves the function of at least six auxiliary proteins, designated HypA, -B, -C, -D, -E, and -F. It has been demonstrated previously that the HypC, -D, -E, and -F proteins participate in cyanide synthesis and transfer. Here, we show by infrared spectroscopic analysis that the purified HypCD complexes from Ralstonia eutropha and Escherichia coli carry in addition to both cyanides the CO ligand. We present experimental evidence that in vivo the attachment of the CN− ligands is a prerequisite for subsequent CO binding. With the aid of genetic engineering and subsequent mutant analysis, the functional role of conserved cysteine residues in HypD from R. eutropha was investigated. Our results demonstrate that the HypCD complex serves as a scaffold for the assembly of the Fe(CN)2(CO) entity of [NiFe] hydrogenase.

[NiFe] hydrogenases are complex metallo-enzymes that catalyze the interconversion of H 2 and protons. Basically, these enzymes are composed of two subunits. The large subunit accommodates the active site, and the small subunit contains one to three iron-sulfur clusters mediating the electron transfer between the catalytic center and the primary electron donor/ acceptor. The catalytic center of [NiFe] hydrogenases displays a sophisticated architecture comprising one nickel and one iron ion that are jointly coordinated by four conserved cysteines. Of the four cysteine residues that ligate the nickel, two serve as bridging ligands to the iron. A 5-fold coordination of the iron is established by three additional nonproteinogenic ligands, namely two cyanides (CN Ϫ ) and one carbon monoxide (CO). The complex biosynthesis of the diatomic ligands and the assembly of the active site rely on a set of six conserved accessory proteins, designated HypA, -B, -C, -D, -E, and -F (1-3). The corresponding hyp genes have been found in all genomes encoding functional [NiFe] hydrogenases underlining their indispensable role in hydrogenase maturation. Hence, the minimal set of genes required for synthesis of an active [NiFe] hydrogenase includes the two structural genes encoding the two hydrogenase subunits and six hyp genes required for active site assembly.
The current model of [NiFe] hydrogenase active site assembly is shown in Fig. 1A. Previous studies uncovered a complex of the proteins HypC and HypD as a crucial intermediate in the formation of the iron moiety equipped with at least one cyanide ligand. Among the Hyp proteins, HypD is the only member containing an iron-sulfur cofactor in the form of a [4Fe-4S] cluster (4). It is presumed that HypD and HypC are connected through an additional iron ion that is ligated by one conserved cysteine residue of each of HypC and HypD (Fig. 1B) (4 -6). The HypCD-Fe complex receives cyanide groups that are synthesized from carbamoyl phosphate by the concerted action of the HypF and HypE proteins (7)(8)(9). According to the current model, the Fe(CN) 2 CO moiety is subsequently transferred to the apo-forms of the hydrogenase large subunits. In the case of the model systems employed in this study, these are the large subunits of the membrane-bound hydrogenase (MBH), 4 the soluble, NAD ϩ -reducing hydrogenase (SH), and the H 2 -sensing regulatory hydrogenase from Ralstonia eutropha (1,10) as well as the corresponding subunits of hydrogenase-1, -2, and -3 from Escherichia coli (Fig. 1A) (2,3). The MBH of R. eutropha represents an exception as the Fe(CN) 2 CO moiety is first transferred to the MBH-specific scaffold protein HoxV and subsequently to the apo-form of the large subunit ( Fig. 1A) (11).
Upon installation of the Fe(CN) 2 CO fragment, the nickel is being inserted in a process catalyzed by the HypA and HypB proteins (and SlyD in the case of E. coli) (12,13). Finally, the large subunits of most but not all hydrogenases undergo a specific proteolytic processing that triggers dimerization with the cognate electron-conducting small subunit (14,15).
Although the overall process of the [NiFe]-active site assembly is fairly established, major questions concerning the origin of carbon monoxide, the sequence of events, and the precise function of the central HypCD complex remained open. The transfer of at least one CN Ϫ ligand from HypE to the HypCD complex has been shown previously (9). However, information on ligand stoichiometry is still missing. It is known that the CO ligand is derived from a metabolic precursor (16). Unlike anticipated previously, this precursor is not carbamoyl phosphate (17,18). Only scarce information is available on the role of the HypCD complex in the assembly of the carbonyl group.
Infrared (IR) spectroscopy has been proven to be a powerful technique to monitor the CN Ϫ and CO ligands in hydrogenase biosynthesis. Specific absorptions of the diatomic molecules in infrared light allow conclusions on the nature and number of the diatomic iron ligands (11,17,18). In this study, we have used this technique to identify of the diatomic ligands in HypCD complexes. The HypCD protein complexes were purified from different genetic backgrounds of two bacterial species, R. eutropha and E. coli, for which extensive information on [NiFe] hydrogenase maturation is available (2,3,10,19). The results reveal that the HypCD complex serves as a scaffold for the assembly of the Fe(CN) 2 (CO) entity of the active site in [NiFe] hydrogenases and provide insights into the chronological order of ligand binding to the HypCD complex.

EXPERIMENTAL PROCEDURES
Bacterial Strains and Plasmids-The strains and plasmids used in this study are listed in supplemental Table S1. Strains with the initials HF were derived from R. eutropha H16. R. eutropha HF962 is derived from the megaplasmid-and hydrogenase-free derivative R. eutropha HF210 (20) and harbors a deletion within the promoter region of the acoX gene. This modification renders the strain incapable of catabolizing acetoin. For the isolation of Strep-tagged HypC and HypD proteins from R. eutropha, plasmids pIB12 and pIB15 were constructed as follows. Using primers 1 and 2 (supplemental Table S2) and pLO11 as the template, a 674-bp fragment was amplified by PCR. The fragment was digested with AflIII and SacI and inserted back into pLO11, which was previously linearized with AflIII and SacI. The resulting plasmid pIB01 carries a His codon (5Ј-CAT-3Ј) upstream of the SacI recognition site. Subsequently, the complete R. eutropha hyp1 region was cut from pCH783 as a 7.8-kbp SacI-HindIII fragment and inserted into pIB01 cut with the same restriction enzymes. This resulted in plasmid pIB02 carrying the hypA1, -B1, -F1, -C, -D, -E, and -X genes under control of the acoX promoter. For attachment of Strep-tag II encoding sequences to the hypC and hypD genes, a 750-bp AscI fragment derived from pCH301 was cloned into vector pNEB193 resulting in plasmid pIB08, which served as the template for the subsequent PCRs. Primers 3 and 4 were used to synthesize a 324-bp fragment encoding an N-terminally Strep-tagged HypD. The resulting PCR fragment was cut with NaeI and cloned into NaeI-linearized vector pIB08, resulting in pIB10. From this plasmid, a 750-bp AscI fragment was removed and inserted into AscI-digested plasmid pIB02. This procedure yielded the overexpression plasmid pIB12 (P acoX -hypA1B1F1C( Strep D)EX).
Plasmid pIB15 (P acoX -hypC( Strep D)) was synthesized by amplifying a 1.4-kb fragment by PCR using primers 5 and 6 and genomic DNA from R. eutropha H16 as the template. The resulting amplicon was digested with PciI and BglII and inserted into NcoI/BglII-linearized pLO11, yielding pIB15. All PCR-derived fragments were verified by sequencing. Plasmids for hyp gene overexpression in E. coli were described previously (6). Plasmids carrying R. eutropha hypD alleles encoding singlesite amino acid exchange variants of HypD were constructed as follows. Plasmid pCH1055, harboring the hypD gene equipped with a sequence encoding an N-terminal Strep-tag II, was used as the template. Therefore, all descendants of pCH1055 encode N-terminally Strep-tagged HypD variants. For amino acid replacements C323S, C325S, C338S, C345S, C354S, and C362S in the C-terminal part of HypD, a 536-bp KpnI fragment containing the 3Ј region of hypD derived from pCH1055 was inserted into the KpnI site of pCH1464, which is a pBlueScript KS(ϩ) derivative lacking the EcoRV restriction site (supplemental Table S1). This resulted in pCH1465 that was subsequently used as a template for the insertion of the point mutations by site-directed mutagenesis through overlap-extension PCR. The first PCR fragment containing the mutations was amplified using the universal primer 7 in combination with either of the mutagenic primers 14 -19. The second PCR product was generated with primers 8 and 9. Both overlapping PCR products served as templates for overlap extension PCR with primers 7 and 8. The resulting amplicons were cut with EcoRV and NcoI, and the resulting 310-bp fragments were re-inserted into pCH1465 yielding pCH1473-pCH1478. The correctness of the PCR-derived sequences was verified by DNA sequencing. From plasmids pCH1473-pCH1478, 536-bp KpnI fragments were transferred to pCH1055. The resulting plasmids were subsequently digested with HindIII-SpeI, and 1420-bp fragments were cloned into pEDY309 yielding the final expression plasmids pGE679-pGE684.
For the amino acid replacements C36S, C64S, and C67S in the N-terminal part of HypD, a 613-bp HindIII-AccI fragment containing the 5Ј region of hypD derived from pCH1055 was inserted into plasmid pBlueScript KS(ϩ). The resulting plasmid was named pCH1524 and served as a template for the subsequent PCR mutagenesis. The mutations were introduced by using primer 10 in combination with either of 11, 12, and 13. The PCR products were cut with NdeI-SphI (C36S) or NdeI-AfeI (C64S and C67S), and the resulting 125-and 231-bp fragments, respectively, were re-introduced into pCH1524. The PCR-derived sequences in the resulting plasmids pCH1525, pCH1528, and pCH1531 were verified by DNA sequencing. Subsequently, 613-bp HindIII-AccI fragments containing the individual mutations were cut from these three plasmids and then inserted into pCH1055, yielding pCH1526, pCH1529, and pCH1532 (supplemental Table S1). The resulting plasmids were digested with HindIII-SpeI, and 1420-bp fragments were cloned into pEDY309 yielding the expression plasmids pGE676, pGE677, and pGE678. All pEDY309 derivatives encoding hypD alleles under the control of the acoX promoter were transferred by conjugation to R. eutropha HF338 (⌬hypD).
Growth of R. eutropha and E. coli Derivatives and Induction of Gene Expression-For overproduction of HypCD in R. eutropha, the respective transconjugants were cultivated aerobically in a glass fermentor (Biostat MD; B. Braun, Melsungen, Germany) filled with 10 liters of fructose/glycerol minimal medium (21). The medium was inoculated to 1% with a pre-culture grown for 72 h in fructose minimal medium. At an A 436 of ϳ6.0, which was reached after ϳ24 h, expression of hyp genes was induced with 10 -15 mM acetoin. After further cultivation for 24 h, the cells were collected by centrifugation (4300 ϫ g, 4°C), washed with ice-cold 50 mM Tris-HCl, pH 8.0, and centrifuged again. The resulting cell pellets were frozen in liquid nitrogen and stored at Ϫ80°C. Overexpression of hypCD in E. coli was accomplished as described previously (6). Briefly, upon induction with 0.2-1.0 mM isopropyl 1-thio-␤-D-galactopyranoside, transformant strains were grown anaerobically in TGYEP medium for 12 h at 22°C.
Purification of Strep-tagged HypCD Protein Complexes-All purification steps were performed at 4°C. Strep-tagged protein complexes were purified from 20 to 80 g (wet weight) of cells. The cell pellet was resuspended in resuspension buffer (50 mM Tris-HCl, pH 7.5) containing 1 tablet of Complete EDTA-free per 50 ml, DNase I (Roche Applied Science), and either 0.1 mM or 1 mM DTT (1 ml of buffer ϫ g Ϫ1 of cells). Cells were disrupted by two passages through a French pressure cell (900 PSI, G. Heinemann, Schwäbisch Gmünd, Germany). Cell debris was removed by ultracentrifugation for 45 min at 90,000 ϫ g. Soluble extract was diluted 2:1 with resuspension buffer. Purification of tagged proteins from the soluble cell extract was performed with Strep-Tactin resin (IBA, Göttingen, Germany) following the protocol of the manufacturer. The columns were washed with 40 column volumes of washing buffer (50 mM Tris-HCl, pH 7.5, 50 mM NaCl, 0.1 mM, 1 mM DTT). Protein elution took place by addition of 6 column volumes of washing buffer containing 5 mM desthiobiotin. Fractions containing protein were pooled and concentrated in buffer S (10 mM Tris-HCl, pH 8.0, at 4°C, 5% (w/v) glycerol) to a final volume of ϳ20 l using Amicon Ultra-15 and Microcon filtration devices (Millipore). The final protein concentration ranged between 15 and 126 mg/ml as measured with a bicinchoninic acid-based assay. The purified protein samples were stored in liquid nitrogen.
Infrared Spectroscopy-Spectra were recorded on a Fourier transform spectrometer (Bruker Tensor 27) equipped with a liquid nitrogen-cooled mercury-cadmium-telluride detector at a spectral resolution of 2 cm Ϫ1 . The sample compartment was purged with dried air, and the sample was held in a temperature-controlled (10°C) gas-tight liquid cell (volume ϳ9 l, optical path length ϭ 50 m) with CaF 2 windows. Spectra were base line-corrected by using a spline function implemented within OPUS 6.5 software supplied by Bruker.
Western Immunoblot Analysis-Proteins were separated using denaturing PAGE (22). Western immunoblot analysis was performed according to a standard protocol (23). For immunological detection of hydrogenase maturation proteins, host-specific anti-HypC and anti-HypD antisera were applied in a 1:10,000 dilution.
Hydrogenase Assays-Activity of the MBH was routinely measured in a rubber-sealed cuvette under anaerobic conditions using a spectrophotometric assay at 30°C based on the H 2 -dependent reduction of methylene blue (24). The activity of the soluble, NAD-reducing hydrogenase (EC 1.12.1.2) was routinely measured in a whole cell assay (25).
Metal Determination-The HypD protein was expressed and purified as described by Blokesch et al. (6). The iron content NOVEMBER 9, 2012 • VOLUME 287 • NUMBER 46 was determined by inductively coupled plasma-optical emission spectrometry (ICP-OES; Optima 2100DV; PerkinElmer Life Sciences). Protein samples were incubated overnight in a 1:1 mixture with 65% nitric acid (Suprapur, Merck) at 100°C. Samples were filled to a 10-fold volume with water prior to ICP-OES analysis. Samples were analyzed for zinc, iron, and nickel in parallel. As reference, the multielement standard solution XVI (Merck) was used.

Construction of Expression Plasmids for Functional Characterization of HypCD Protein Complexes from R. eutropha-To
obtain sufficient amounts of HypCD protein for spectroscopic analysis, we constructed expression plasmids harboring the hypCD genes of R. eutropha in two different contexts (Table 1  and supplemental Table S1 and supplemental Fig. S1). The broad host range plasmid pIB01, a derivative o,ZSIf pLO11 (26), served as vector backbone. This plasmid carries the ribosomal binding site of the R. eutropha hoxF gene and the inducible acoX promoter (P acoX ) of the R. eutropha acoXABC operon (10,27).
Plasmid pIB12 contains the hypA1, -B1, -F1, -C, -D, -E, and -X genes (supplemental Fig. S1), which constitute the entire hyp1 operon of R. eutropha. It has been shown before that the hyp1 operon is necessary and sufficient for maturation of all three [NiFe] hydrogenases present in R. eutropha (28 -30). For purification of the HypCD complex, the 5Ј-end of hypD on pIB12 was equipped with a sequence encoding Strep-tag II (the corresponding gene product is henceforth denoted HypD Strep ). The second expression plasmid, pIB15, encoded solely the HypC and HypD Strep proteins (supplemental Fig. S1).
To analyze the functionality of the modified genes on pIB12 and pIB15, the plasmids were conjugationally transferred to R. eutropha derivatives lacking one or more hyp genes. The resulting transconjugants were grown in fructose/glycerol minimal medium (21) containing 10 mM acetoin for induction of the hyp gene expression. The cells were harvested in the glycerol-supported growth phase, and the activities of the MBH and SH were determined.
It has been shown previously that single hypC and hypD deletions result in hydrogenase-negative R. eutropha derivatives (29). This phenotype could be suppressed by the transfer of the modified hypCD genes cloned on plasmid pIB15 to the R. eutropha recipients HF340 (⌬hypC) and HF338 (⌬hypD). Both transconjugants showed almost wild-type activities for MBH and ϳ40% of SH activity (Table 1). However, plasmid pIB15 failed to restore hydrogenase activities (data not shown) in R. eutropha HF575, which harbors comprehensive deletions in all hyp genes required for SH and MBH maturation (29). Hence, it can be inferred that although the pIB15-encoded HypC and HypD Strep proteins are functional, they are not sufficient for hydrogenase maturation in a host lacking all hyp genes.
By contrast, transfer of plasmid pIB12 to the R. eutropha derivative HF575 led to the recovery of 95 and 67% of MBH and SH wild-type activities, respectively (Table 1). This result demonstrates that the hypA1B1F1CDEX operon on pIB12 is fully functional in MBH and SH maturation.
Purification of the HypCD Protein Complexes from R. eutropha and E. coli-For purification of HypCD Strep complexes from R. eutropha, plasmid pIB12 was transferred to R. eutropha HF962. This recipient is missing the megaplasmid pHG1 that carries all hydrogenase-related genes, including the hyp genes. Because this strain is devoid of the large subunits of the indigenous hydrogenases, which are considered the final targets receiving the iron cofactor from HypCD, we are anticipating accumulation of the HypCD complex in its "cofactor-loaded" configuration, i.e. equipped with the diatomic ligands. While the transconjugant R. eutropha HF962(pIB12) was used for the isolation of the loaded HypCD Strep complex, strain HF962(pIB15) served as control for purification of the ligandfree HypCD Strep complex.
Cells of the R. eutropha transconjugants were grown aerobically in minimal medium in a 10-liter fermenter as described under "Experimental Procedures." Soluble extracts were prepared, and the HypCD Strep protein was purified by Strep-Tactin affinity chromatography. In the case of E. coli, cultivation of the respective transformants (see below) and induction of gene expression were performed as described previously (6). Purification of the E. coli HypCD complex by affinity chromatography was based on the presence of a Strep-tag at the C terminus of the HypC protein.
Typical purification results for HypCD complexes from E. coli and R. eutropha are shown in Fig. 2, A and B, respectively. Despite the presence of contaminating protein bands, the attachment of a Strep-tag peptide on either HypC in the case of E. coli or HypD in the case of R. eutropha resulted in a substantial enrichment of the respective fusion proteins. Moreover, in both cases the cognate partner protein was co-purified demonstrating the formation of stable HypCD complexes in both bacterial species. This was verified immunologically with host-specific antibodies raised against HypC and HypD as shown in Fig.  2C. The band intensities in the pool fractions (Fig. 2, A and B) do not indicate a perfect stoichiometric appearance of the HypC and HypD proteins. As often observed, the bait protein was isolated in excess. The resulting HypCD Strep protein samples were concentrated to a final volume of 10 -20 l with protein concentrations ranging from 0.3 to 1.9 mM and subsequently subjected to IR spectroscopic analysis.
IR Spectroscopic Characterization of HypCD Protein Complexes from R. eutropha-The spectrum obtained for the Hyp-CD Strep complex prepared from R. eutropha HF962 cells harboring plasmid pIB12 revealed two prominent bands at 2099 and 2077 cm Ϫ1 indicative for (CN) stretching vibrations (Fig.  3A). This observation supports the notion that the complex contained two cyanide ligands and confirms previous assumptions based on radiochemical investigations of the E. coli HypCD complex (6). Furthermore, a third major absorption band was detected at 1969 cm Ϫ1 , which can be attributed to a (CO) stretching mode (Fig. 3A). This observation and the fact that the overall band pattern is similar to the arrangement found in the active site of [NiFe] hydrogenases (31, 32) and corresponding model complexes (33) suggests that the HypCD protein complex carries an iron equipped with two cyanides and one CO ligand.
A completely different result was obtained when plasmid pIB15 was used for the isolation of the HypCD Strep protein. Absorption bands specific for CN Ϫ or CO ligands were entirely absent (Fig. 3B). In conclusion, these data suggest that the HypCD complex plays a key role as a scaffold in the assembly of both the cyanide and carbon monoxide ligands. The data also show that co-expression of one or more of the hypA1B1F1EX genes together with hypCD is crucial for proper attachment of the diatomic ligands to the HypCD complex.

IR Spectroscopic Characterization of HypCD Protein Complexes from E. coli-Previous results revealed that HypE and
HypF are essential for endowing the HypCD complex with CN Ϫ ligands, whereas the HypA and HypB proteins are supposed to deliver nickel into the apo-forms of [NiFe] hydrogenase large subunits (2,12,13). To determine the minimal set of genes that needs to be co-expressed to obtain a loaded HypCD complex, we took advantage of pre-existing plasmids in E. coli (6). Plasmids pT-hypDEFCStrep, pT-hypDECStrep, and pT-hypDCStrep (plasmid maps are shown in supplemental Fig. S1) were transferred to the wild-type strain E. coli BL21(DE3). It is important to note that this E. coli strain harbors a chromosomal set of the hypA, -B, -C, -D, -E, and -F genes that is co-expressed with the plasmid-borne copies. The HypCD Strep complexes isolated from the individual transformants were examined by IR spectroscopy. The HypCD Strep complex purified from E. coli(pT-HypDEFCStrep) revealed three major absorption bands that can be attributed to cyanide ((CN) ϭ 2099, 2075 cm Ϫ1 ) and CO ligands ((CO) ϭ 1951 cm Ϫ1 ) (Fig. 4A).
Traces B and C of Fig. 4 illustrate the spectral properties of the HypCD Strep complexes originating from plasmids pT-hyp-DECStrep and pT-hypDCStrep, respectively. The absence of the characteristic ligand absorptions in Fig. 4C indicates that   NOVEMBER 9, 2012 • VOLUME 287 • NUMBER 46 co-overexpression of HypE is necessary to obtain a loaded Hyp-CD Strep complex. This result is consistent with the previous observation that the formation of a ternary, presumably stoichiometric complex of HypC, HypD, and HypE proteins is crucial for CN Ϫ ligand synthesis (6). Furthermore, our results reveal that the chromosomal expression level of hypF in E. coli is sufficient for the formation of a HypCD Strep complex equipped with two cyanides and one CO ligand. Although stoichiometric amounts of HypE are obviously required for the attachment of the CN Ϫ and CO ligands on the HypCD complex, it remains unclear whether HypE is directly involved in CO synthesis or attachment.

Scaffold Protein Complex in Maturation of [NiFe] Hydrogenase
A suitable tool to address this question is a strain incapable of producing carbamoyl phosphate, which serves as the precursor of the CN Ϫ group but not for the CO ligand (17,18). So far, it was impossible to isolate a carbamoyl-phosphate synthetasefree derivative of R. eutropha (17). Therefore, we used a carAB mutant of E. coli transformed with plasmid pT-hypDECStrep for further analysis. The yield of the corresponding HypCD Strep protein was similar to that for wild-type E. coli harboring the same plasmid. Moreover, the metal determination revealed that the HypD proteins isolated from both E. coli derivatives had a quite similar iron content (wild type, 3.6 Ϯ 0.8 iron per protein; carAB mutant, 3.9 Ϯ 1.8 iron per protein) indicating that the carAB mutation did not affect the formation of the [4Fe-4S] cluster in HypD (4,5). The IR spectrum of the Hyp-CD Strep complex isolated from the carAB strain, however, did not show any absorption band (Fig. 4D). Although only carbamoyl phosphate is missing, neither CN Ϫ nor CO-related stretchings could be detected in the complex. Recall that carbamoyl phosphate serves exclusively as the precursor in CN Ϫ ligand biosynthesis (17,18,34). Because both HypE and HypF are present in sufficient amounts for CN Ϫ ligand assembly, the result suggests that a functional pathway for cyanide biosynthesis providing the substrate for ligand assembly is a prerequisite for the coordinated attachment of the CO group to the HypCD complex.
Band Positions of the Diatomic Ligands Respond to Changes in the Redox Status of the HypCD Complex-The HypD protein contains a redox-active [4Fe-4S] cluster and cysteine disulfides, which are proposed to be crucially involved in the formation of the cyanide ligands (4 -6). As dithiothreitol (DTT) is commonly used to reduce disulfide bonds and the spectroscopic signatures of the iron-bound CN Ϫ and CO ligands are good probes for detecting redox changes. The incubation of the oxidized HypCD complex with sufficient amounts of DTT is expected to result in a shift of the stretching frequencies of the diatomic ligands.
The HypCD complex was purified aerobically from E. coli in the presence of either 0.1 mM DTT (Fig. 4A) or 1 mM DTT (Fig.  4E). At high DTT concentrations, the signal pattern changed significantly. An absorption band at 1963 cm Ϫ1 occurred at the expense of the band at 1951 cm Ϫ1 . The CN Ϫ -related absorptions remained basically unchanged. This result suggests that the formation of disulfide bonds affects the electronic properties of the Fe(SR) 2 (CN) 2 CO moiety and are in line with previous observations that redox-active cysteine residues are involved in loading the HypCD complex (4,5).
Cysteine Residues Involved in Function of the HypD Protein from R. eutropha-The crystal structure of the HypD protein from Thermococcus kodakarensis (5) revealed two disulfide bonds that are formed by two pairs of cysteines, Cys 66 -Cys 68 and Cys 325 -Cys 354 , which correspond to residues Cys 64 -Cys 67 and Cys 325 -Cys 354 , respectively, of the HypD protein of R. eutropha ( Fig. 1B and supplemental Fig. S3). Interestingly, only the Cys 64 -Cys 67 pair seems to be strictly conserved in HypD proteins. A prominent example of the latter HypD-type is the HypD protein from E. coli, which carries a leucine residue at the position of Cys 354 , suggesting that this type of HypD protein forms only a single disulfide bond, if at all. The HypD from R. eutropha, however, is more related to the T. kodakarensis counterpart because it also possesses the second pair of cysteine residues (Cys 325 and Cys 354 ).
To investigate the functional role of selected cysteines in HypD, we exchanged them by serine residues ( Table 2). The corresponding hypD alleles were expressed under the control of the P acoX promoter on a broad host range plasmid that was conjugated into the isogenic hypD deletion derivative HF338 of R. eutropha (29). The resulting transconjugant cells were grown under hydrogenase-derepressing conditions. Functionality of the HypD derivatives was tested on the basis of H 2 -dependent reduction of NAD ϩ mediated by the SH, which served as representative of the three well characterized R. eutropha [NiFe] hydrogenases. Furthermore, the stability of the HypD variants was tested immunologically with a polyclonal anti-HypD serum ( Table 2 and supplemental Fig. S2).
The hypD deletion strain R. eutropha HF338 carrying the control vector pEDY309 showed ϳ3% of the H16 wild-type activity (Table 2), which emphasizes the importance of HypD for biosynthesis of hydrogenase (29). Successful complementation of R. eutropha HF338 (⌬hypD) with pCH1062, encoding Strep-tagged wild-type HypD, was demonstrated by the recovery of about 70% of SH activity. As expected, the cysteine residue at position 36 was absolutely essential for HypD activity. This residue is proposed to be involved in the coordination of the iron atom in a concerted action with His 203 from HypD and Cys 2 from HypC (see Fig. 1B) (5). The corresponding exchange in E. coli HypD resulted, however, in a stable but inactive HypD variant ( Table 2) (4). Exchanges of the disulfide-forming Cys 64 and Cys 67 residues by serines abolished HypD function in R. eutropha. Although the C64S exchange resulted in a complete loss of protein, HypD stability was not affected in the C67S derivative. This observation supports previous results obtained with HypD variants from E. coli (4). Remarkably, individual exchanges of Cys 325 and Cys 354 , which are proposed to constitute the second disulfide, did not affect HypD function neither on protein stability nor on hydrogenase activity (Table 2 and supplemental Fig. S2). This observation clearly shows that these two residues, although conserved in 63% of HypD proteins, are dispensable for protein function. Residues Cys 323 , Cys 338 , Cys 345 , and Cys 362 are likely involved in coordination of the [4Fe-4S] cluster of HypD. Consequently, serine exchanges resulted in unstable HypD proteins, and almost no hydrogenase activity was observed in the respective mutant cells. The mutant carrying the C323S exchange in HypD represents an interesting exception because it showed residual hydrogenase activity (12% of the wild-type activity; Table 2). Because of the close proximity of Cys 323 and Cys 325 (5), it is conceivable that Cys 325 may partially replace Cys 323 . In fact, residue Cys 323 of E. coli HypD, which aligns at position Cys 325 of the R. eutropha HypD (supplemental Fig. S3

DISCUSSION
A complex of the proteins HypC and HypD has been previously identified to be crucial for the active site assembly of [NiFe] hydrogenase. A composite of genetic, biochemical, and structural studies revealed a model that suggests a stepwise binding of the two cyanide ligands and the CO group onto an iron ion that is jointly coordinated by HypC and HypD (4 -6, 9). So far, these studies have demonstrated that the HypCD complex is capable of receiving at least one cyanide group from HypE (9). In this study, we show that the HypCD complexes isolated from two nonrelated bacterial species, E. coli and R. eutropha, carry already the full complement of diatomic ligands, namely cyanide and carbon monoxide. This assignment is based on distinct IR absorption bands related to characteristic stretching vibrations of the diatomic ligands attached to iron. Thus, our results also provide indirect evidence for the postulated iron atom that serves as the target for ligand assembly.
The absorption band positions of the (CN) and (CO) stretching vibrations in both HypCD complexes are close to those observed in [NiFe] hydrogenases (32). The relatively large difference of about 20 wavenumbers between the (CO) stretching frequencies in the oxidized HypCD complexes of R. eutropha (1969 cm Ϫ1 ) and E. coli (1951 cm Ϫ1 ) might be related to differences in the direct protein environment as well as to a divergent content of cysteine residues, which tune the redox environment of the Fe(CN) 2 (CO) unit (see below). Notably, Rauchfuss et al. (35)   ase. This might be related to a higher electron density at the iron of the model complex, which in turn leads to a stronger back-donation in the anti-binding *-orbital of the ironbound CO. (32,36). While performing the synthesis of Fe(SR) 2 (CN) 2 (CO) mimics, Rauchfuss et al. (35) came to the conclusion that the high ligand-field strength of bound cyanide stabilizes low-spin Fe II , which subsequently enables binding of CO. The underlying mechanism behind this phenomenon of cooperative binding is supposed to be a synergetic effect arising from strong -donor properties of CN Ϫ ligands that stabilize the -back-bonding between Fe II and CO (37). This assumption is fully consistent with our results obtained for the biological system.
It has been shown before that carbamoyl phosphate serves as a metabolic precursor of the cyanide ligands synthesized by HypF and HypE (7). We have shown in this study that neither the CN Ϫ ligands nor the CO ligand is attached to HypCD in a strain that lacks carbamoyl-phosphate synthetase. In previous work, it has been demonstrated that the CO ligand is definitely not derived from carbamoyl phosphate (17,18). Therefore, we conclude that binding of the cyanide ligands is a prerequisite for subsequent attachment of CO to the iron.
Analysis of the crystal structures of HypC, HypD, and HypE from T. kodakarensis revealed a sophisticated model of thiolbased redox chemistry involved in the concerted assembly of the Fe(CN) 2 (CO) unit on HypCD. In addition to a [4Fe-4S] cluster coordinated by four cysteine residues, two pairs of cysteines seem to form disulfide bonds. It was postulated that the conversion of the four cysteine-derived thiols into two disulfide bonds provides sufficient electrons for two cyanide transfer reactions from HypE to an iron atom, which is trapped by Cys 36 of HypD and Cys 2 of HypC (Fig. 1B) (5). Watanabe et al. (5) had already recognized that the cysteine residues forming the proposed disulfide bridges are not conserved in all HypD proteins. The cysteine residues corresponding to Cys 64 and Cys 67 of R. eutropha HypD seem to be invariant, whereas Cys 325 and Cys 354 , which are proposed to form a disulfide bridge close to the [4Fe-4S] cluster, are less conserved.
In fact, the HypD proteins can be grouped into two subclasses. Among the 448 HypD isoforms, 63% contain the four cysteines required for two disulfide bonds (DDB-type, double disulfide bond), whereas the isoforms with only one putative disulfide bond (single disulfide bond, SDB-type) in HypD represent the minority. For example, HypD of E. coli belongs to the SDB-type, whereas the HypD proteins from R. eutropha and T. kodakarensis are members of the DDB-type. Although more than 60% of the HypD proteins contain cysteines sufficient for the formation of two disulfide bridges, mutant analysis of the DDB-type HypD from R. eutropha revealed that Cys 325 as well as Cys 354 are dispensable for HypD function. The other cysteine exchanges made in this study almost abolished HypD activity (Table 2), which clearly supports their crucial function in assembly of the Fe(CN) 2 (CO) unit.
In addition to the large subunits of [NiFe] hydrogenases, there was up to now only the HoxV protein known to carry cyanide-and carbonyl-ligated iron. HoxV is involved in maturation of the O 2 -tolerant membrane-bound [NiFe] hydrogenase of R. eutropha, and it has been proposed that HoxV receives the Fe(CN) 2 (CO) moiety from the HypCD complex and serves as a transient carrier until the Fe(CN) 2 (CO) group is eventually transferred to the hydrogenase apoprotein. IR spectroscopic analysis of the Fe(SR) 2 (CN) 2 (CO) moiety present in HoxV showed that even incubation with the strong reductant dithionite did not affect the band positions of the diatomic ligands (11). This observation is in contrast to the "loaded" HypCD complex that exhibits a shift of the CO stretching upon moderate reduction conditions, presumably due to a reversible formation of disulfide bridges within HypD.
In this study, we have shown that HypC and HypD form the core complex for the assembly of the Fe(CN) 2 (CO) moiety at the active site of the [NiFe] hydrogenase. The attachment of a carbonyl ligand to a low spin metal with d 6 -orbital configuration has been proposed to be crucial for heterolytic cleavage of dihydrogen (38). Thus, the elucidation of the mechanism of carbonylation of the iron in the HypCD complex is of paramount interest also for understanding the catalytic mechanism itself, as well as for tailoring synthetic catalytic compounds in the future. Furthermore, despite recent progress in search of the metabolic source of the CO ligand in [NiFe] hydrogenases (16), the detailed synthesis and the precise functions of the participating proteins still have to be unraveled.