Rubredoxin-related Maturation Factor Guarantees Metal Cofactor Integrity during Aerobic Biosynthesis of Membrane-bound [NiFe] Hydrogenase*

Background: Biosynthesis of complex metal cofactors in [NiFe] hydrogenase is sensitive toward molecular oxygen. Results: A rubredoxin-like protein is required for hydrogenase maturation under aerobic conditions. Conclusion: The rubredoxin-like protein prevents oxidative damage of metallocenters, including the recently discovered [4Fe3S] center. Significance: Dedicated protection mechanisms enable biosynthesis of sophisticated metal centers in the presence of dioxygen. The membrane-bound [NiFe] hydrogenase (MBH) supports growth of Ralstonia eutropha H16 with H2 as the sole energy source. The enzyme undergoes a complex biosynthesis process that proceeds during cell growth even at ambient O2 levels and involves 14 specific maturation proteins. One of these is a rubredoxin-like protein, which is essential for biosynthesis of active MBH at high oxygen concentrations but dispensable under microaerobic growth conditions. To obtain insights into the function of HoxR, we investigated the MBH protein purified from the cytoplasmic membrane of hoxR mutant cells. Compared with wild-type MBH, the mutant enzyme displayed severely decreased hydrogenase activity. Electron paramagnetic resonance and infrared spectroscopic analyses revealed features resembling those of O2-sensitive [NiFe] hydrogenases and/or oxidatively damaged protein. The catalytic center resided partially in an inactive Niu-A-like state, and the electron transfer chain consisting of three different Fe-S clusters showed marked alterations compared with wild-type enzyme. Purification of HoxR protein from its original host, R. eutropha, revealed only low protein amounts. Therefore, recombinant HoxR protein was isolated from Escherichia coli. Unlike common rubredoxins, the HoxR protein was colorless, rather unstable, and essentially metal-free. Conversion of the atypical iron-binding motif into a canonical one through genetic engineering led to a stable reddish rubredoxin. Remarkably, the modified HoxR protein did not support MBH-dependent growth at high O2. Analysis of MBH-associated protein complexes points toward a specific interaction of HoxR with the Fe-S cluster-bearing small subunit. This supports the previously made notion that HoxR avoids oxidative damage of the metal centers of the MBH, in particular the unprecedented Cys6[4Fe-3S] cluster.

The membrane-bound [NiFe] hydrogenase (MBH) supports growth of Ralstonia eutropha H16 with H 2 as the sole energy source. The enzyme undergoes a complex biosynthesis process that proceeds during cell growth even at ambient O 2 levels and involves 14 specific maturation proteins. One of these is a rubredoxin-like protein, which is essential for biosynthesis of active MBH at high oxygen concentrations but dispensable under microaerobic growth conditions. To obtain insights into the function of HoxR, we investigated the MBH protein purified from the cytoplasmic membrane of hoxR mutant cells. Compared with wild-type MBH, the mutant enzyme displayed severely decreased hydrogenase activity. Electron paramagnetic resonance and infrared spectroscopic analyses revealed features resembling those of O 2 -sensitive [NiFe] hydrogenases and/or oxidatively damaged protein. The catalytic center resided partially in an inactive Ni u -A-like state, and the electron transfer chain consisting of three different Fe-S clusters showed marked alterations compared with wild-type enzyme. Purification of HoxR protein from its original host, R. eutropha, revealed only low protein amounts. Therefore, recombinant HoxR protein was isolated from Escherichia coli. Unlike common rubredoxins, the HoxR protein was colorless, rather unstable, and essentially metal-free. Conversion of the atypical iron-binding motif into a canonical one through genetic engineering led to a stable reddish rubredoxin. Remarkably, the modified HoxR protein did not support MBH-dependent growth at high O 2 . Analysis of MBH-associated protein complexes points toward a specific interaction of HoxR with the Fe-S cluster-bearing small subunit. This supports the previously made notion that HoxR avoids oxidative damage of the metal centers of the MBH, in particular the unprecedented Cys 6 [4Fe-3S] cluster.
The reversible oxidation of molecular hydrogen (H 2 ) into protons and electrons is catalyzed by hydrogenases and constitutes a key process in the metabolism of many bacteria, archaea, and unicellular eukaryotes. Aerobic H 2 -oxidizing Knallgas bacteria are characterized by their ability to oxidize H 2 using O 2 as terminal electron acceptor, thereby conserving the energy released from the highly exogenous "Knallgas" reaction (1). This lifestyle relies on O 2 -tolerant [NiFe] hydrogenases that transfer electrons from hydrogen oxidation to NAD ϩ or the quinone pool of the respiratory chain (2)(3)(4). These robust enzymes are exceptional among hydrogenases, which are usually inhibited by traces of O 2 . Ralstonia eutropha H16 is a model organism for chemolithautotrophic growth on H 2 , O 2 , and CO 2 (5,6). This metabolically versatile ␤-proteobacterium harbors four [NiFe] hydrogenases, which share the ability of H 2 conversion at ambient O 2 but serve different physiological functions (7,8). The membrane-bound hydrogenase of R. eutropha (ReMBH), 2 (Fig. 1A) consists of the large subunit that accommodates the heterobimetallic active site as follows: a cysteinebound [NiFe] cofactor, in which the iron ion is further coordinated by two cyanides and one carbonyl ligand (Fig. 1B). The large subunit is intimately associated with a small subunit harboring an electron transfer relay of three different Fe-S centers (9) that connect the catalytic center with the primary electron acceptor, a membrane integral b-type cytochrome (10).
In contrast to the MBH from R. eutropha, structurally related O 2 -sensitive "standard" [NiFe] hydrogenases, which have been initially characterized in sulfate-reducing bacteria, are rapidly inactivated in the presence of O 2 giving rise to a mixture of inactive oxidized states of the catalytic site (11,12). Two of these states, denoted Ni u -A and Ni r -B, can be detected by electron paramagnetic resonance (EPR) spectroscopy. Enzymes in the "ready" Ni r -B state contain a bridging hydroxo species between nickel and iron ( Fig. 1B) (13,14), which undergoes fast reactivation under relatively mild reducing conditions. In con-trast, Ni-Fe sites residing in the Ni u -A state are suggested to contain a hydroperoxo ligand (15,16) or a related oxidative modification (17,18). Hydrogenases in the Ni u -A state underlie an extremely slow reactivation process that requires strongly reducing conditions and presumably occurs only in vitro (19). Hence, the strict avoidance of Ni u -A-related states and a rapid reactivation of [NiFe] hydrogenase in the Ni r -B state are mandatory for sustained H 2 conversion under aerobic conditions (20,21). Accordingly, Ni u -A has never been observed in wildtype ReMBH (19,22), and because of their remarkable O 2 tolerance, MBH-like proteins are promising tools for application in H 2 -based technologies such as enzymatic fuel cells (23,24) and light-driven biohydrogen production (25,26).
Multidisciplinary studies revealed that the Fe-S center proximal to the catalytic site of the ReMBH differs in its electronic and molecular structure from the conventional [4Fe-4S] cubanes that are usually coordinated to the protein by four cysteine-derived thiolates at the corresponding position of standard [NiFe] hydrogenases (9,19,22). The Fe-S cluster proximal to the [NiFe] site of the ReMBH and homologous hydro-genases of Escherichia coli and Hydrogenovibrio marinus is coordinated by two additional cysteine residues leading to a Cys 6 [4Fe-3S] configuration (9,27,28) (Fig. 1C). This unique architecture of the Fe-S cofactor facilitates two concerted redox transitions at physiologically relevant potentials, a performance that cannot be achieved by regular [4Fe-4S] centers (19,29). The redox changes at the Cys 6 [4Fe-3S] center involve ligand exchanges of at least one of the iron atoms ( Fig. 1C) (27,28), which in turn tune the cluster's redox potential and stabilize three different redox states under physiological conditions. The two electrons stored in the fully reduced Cys 6 [4Fe-3S] 3ϩ cluster provide an electron-rich environment to the active site and are proposed to facilitate that attacking O 2 is efficiently reduced to harmless H 2 O without evoking oxidative damage (19,20).
Saggu and co-workers (22) compared the isolated MBH heterodimer, which was rather unstable and exhibited a relatively low, rapidly decreasing hydrogenase activity, with the heterotrimeric membrane-attached form of MBH, which reacts fully reversibly with H 2 and O 2 . As a consequence of this study, an improved purification protocol was developed that involves the full chemical oxidation of the membrane proteins prior to purification of the MBH heterodimer. This procedure resulted in rather stable homogenous MBH protein that resides up to ϳ80% in the Ni r -B state (19,30). These results indicate that only the fully oxidized purified MBH remains functional in the presence of O 2 , whereas the simultaneous presence of free electrons (in partially reduced enzyme) and O 2 results in protein damage. Presumably, both the Ni-Fe site and the proximal Fe-S center of the MBH are major targets for oxidative attack (22).
Apart from the catalytic reaction, hydrogenase biosynthesis under aerobic conditions requires biosynthetic devices that protect the hydrogenase subunit precursors against harmful effects of O 2 (31)(32)(33)(34). The MBH undergoes a particularly complex maturation process, which takes place in the cytoplasm (4). The gene cluster responsible for ReMBH biosynthesis encompasses a set of 21 genes (35). The Ni-Fe cofactor is incorporated into the large MBH subunit (HoxG) by at least six hyp gene products (36,37). In addition, a specific chaperone HoxL and the transfer protein HoxV were shown to be necessary for proper MBH maturation in cells grown under aerobic conditions (38,39). Similarly, the HoxL homolog HupF of Rhizobium leguminosarum was reported to stabilize the premature large hydrogenase subunit in the presence of O 2 (34). Moreover, it was shown that the chaperones HoxO and HoxQ interact with the small ReMBH subunit precursor (preHoxK), probably shielding the Fe-S centers in preHoxK against O 2 (7,32). Two additional proteins, HoxR and HoxT, were found to be required for efficient MBH maturation at high O 2 levels (33). HoxR Ϫ strains of R. eutropha showed an extremely O 2 -sensitive phenotype regarding MBH-dependent growth on H 2 , significantly reduced MBH levels in the membrane, and severely decreased hydrogenase activity. Furthermore, preliminary co-purification studies uncovered a large protein complex consisting of the two MBH subunits and several maturases, including HoxR (33).
In this study, we analyzed the reactivation properties and metal cofactor composition of MBH isolated from aerobically and microaerobically cultivated hoxR mutant cells using biochemical and spectroscopic techniques. Detailed co-purifica-FIGURE 1. Redox cofactors in the MBH of R. eutropha. A, schematic view of the physiologically active MBH, which is attached to the periplasmic side of the cytoplasmic membrane via a C-terminal hydrophobic extension of the small subunit (HoxK) and a membrane-integral cytochrome b. Electrons from H 2 oxidation at the Ni-Fe site in the large subunit (HoxG) are transferred via the chain of Fe-S centers in HoxK to the cytochrome and are finally fed into the quinone pool of the respiratory chain. B, oxidized Ni-Fe site is shown along with the Fe-S centers and ligating amino acid side chains (yellow, sulfur; gray, carbon; red, oxygen; blue, nitrogen; brown, iron). The six cysteine residues that coordinate the proximal [4Fe-3S] center in the small subunit are labeled, and key distances between metal atoms in the electron transfer relay are shown. The physiological redox states of the metal centers are indicated. C, proximal Fe-S center in MBH passes through three stable redox states, [4Fe-3S] 3ϩ/4ϩ/5ϩ , within an extraordinary narrow midpoint potential range (E m ). This transfer of two consecutive electrons is accomplished via redox-dependent structural rearrangements, i.e. upon "superoxidation," the bond between one iron and a sulfide is broken, and the deprotonated peptide amide-N of Cys-20 becomes a ligand to this iron.
tion studies with HoxR as bait were conducted to obtain further insights into the specific interaction of HoxR with the maturation complex. A recombinant HoxR variant was constructed to explore the role of the potential iron-binding motif in the function of this rubredoxin-like protein.

EXPERIMENTAL PROCEDURES
Bacterial Strains and Plasmids-E. coli JM109 was used as the host in standard cloning procedures, and PCR-generated fragments were confirmed by sequencing. For heterologous production of HoxR in E. coli, a PCR-generated 0.234-kbp NcoI-BamHI-digested fragment containing hoxR (33) was transferred to NcoI-BglII-cut pASK-IBA-3-NcoI-BglII 3 and NcoI-BglII-cut pQE60 (Qiagen). This resulted in the plasmids pCH1368 (hoxR-StrepTag II under the control of the tetA promoter) and pCH1370 (hoxR-His 6 under the control of the T5 promoter). The MBH-overproducing strains used in this study are derivatives of the megaplasmid-free strain R. eutropha HF631 carrying variants of the plasmid pLO6 that contains the complete MBH gene cluster from the wild-type strain R. eutropha H16 and a single copy ⌽(hoxKЈ-lacZ) translational fusion in the chromosome (40). For construction of the hoxR W32G-D65G allele, the 0.129-kbp PCR product obtained with the primers J30 (AATGCAAGATCTGCGGGTGGGAG-TACGATCC) and J31 (GCTCGGCTTCACCGCCGCAATT-CGGACACCGC) and pCH1370 as template was used as primer for a second PCR with pCH1370 as template. The resulting 7.672-kpb PCR product, containing the recombinant hoxR W32G-D65G -His 6 gene under the control of an isopropyl 1-thio-␤-D-galactopyranoside-inducible T5 promoter was digested with DpnI and transferred into E. coli JM109 resulting in the plasmid pCH1689. For genetic complementation experiments, expression plasmids containing hoxR-His 6 and hoxR W32G-D65G -His 6 , respectively, under the control of P hoxF were constructed by transferring a 0.261-kbp NcoI-HindIII fragment to NcoI-HindIII-cut pLO12, resulting in pCH1690 (hoxR-His 6 ) and pCH1691 (hoxR W32G-D65G -His 6 ).
Media and Growth Conditions-Media and growth conditions for R. eutropha strains have been described previously (33). Antibiotics for R. eutropha were used at the following concentrations: kanamycin, 400 g/ml; tetracycline, 10 g/ml. For E. coli, the following concentrations were employed: kanamycin, 50 g/ml; tetracycline, 10 g/ml; ampicillin, 100 g/ml. For purification of MBH variants, R. eutropha strains were cultivated heterotrophically under hydrogenase-derepressing conditions at 30°C in mineral medium containing 0.05% w/v fructose and 0.4% w/v glycerol (FGN) in baffled 5000-ml Erlenmeyer flasks filled with 2000 ml of culture ("well aerated") or 4000 ml ("O 2 -limited") and shaken at 120 rpm. Cells were harvested at an absorbance at 436 nm (A 436 ) of 8 -10 by centrifugation at 6000 ϫ g for 15 min at 4°C. For copurification experiments, R. eutropha strains were cultivated in mineral medium containing 0.2% w/v fructose and 0.2% w/v glycerol in baffled 5000-ml Erlenmeyer flasks filled with 2000 ml of culture and harvested as described above. For heterologous production of HoxR variants, JM109 strains carrying the plasmids pCH1368 and pCH1370 and pCH1689, respectively, were cultivated at 30°C in LB media containing ampicillin, shaken at 120 rpm, and gene expression was induced at an A 600 of about 0.6 with 200 ng/ml anhydrotetracycline (pCH1368) or 1 mM isopropyl 1-thio-␤-D-galactopyranoside (pCH1370). Cells were harvested at an A 600 of ϳ2 by centrifugation at 6000 ϫ g for 15 min at 4°C. Lithoautrophic growth of R. eutropha strains was tested on mineral agar plates at 30°C under an atmosphere of 3% v/v H 2 , 10% v/v CO 2 , and varying O 2 concentrations balanced by N 2 .
Isolation of Membranes and Strep-Tactin Affinity Chromatography-Purification of MBH heterodimer from the membrane fraction was carried out as described previously (33). This includes the oxidation of cell extracts by addition of K 3 (Fe(CN 6 )) to a final concentration of 50 mM prior to solubilization of membrane proteins using Triton X-114. Protein concentrations were determined with the BCA TM -kit (Pierce) with bovine serum albumin as standard. Purity of the samples was examined by visual inspection after separation of the proteins via SDS-PAGE and subsequent staining with Coomassie Brilliant Blue G-250.
Copurification Experiments with HoxR Strep as Bait-Purification of protein complexes from soluble extracts of MBHoverproducing R. eutropha strains was conducted via Strep-Tactin affinity chromatography, and proteins were identified via Western blot analysis. Both methods were carried out according to Ref. 33. For immunological examination, the amount of protein isolated from 1 g of cells (wet weight) was loaded on each lane of the SDS-polyacrylamide gels. Crude cell extracts were obtained by disrupting cell suspensions in a French pressure cell (SLM Aminco) via two passages at 18,000 lb/in 2 .
Purification of Heterologously Produced HoxR Variants-For purification of His 6 -tagged HoxR variants, cells were resuspended in buffer W (2 ml of 300 mM NaCl, 50 mM NaH 2 PO 4 , 35 mM imidazole, pH 7.1, per 1 g wet weight) containing Complete EDTA-free protease inhibitor mixture (Roche Applied Science) and DNase I, and the suspension was disrupted in a French pressure cell via two passages at 18,000 lb/in 2 . After ultracentrifugation (100,000 ϫ g, 60 min), the supernatant containing the soluble proteins was loaded onto nickel-nitrilotriacetic acid-agarose columns (Qiagen, Germany; 1-ml bed volume for up to 20 ml of soluble extract), which were run by gravity flow. The columns were washed with 10 bed volume equivalents of buffer W, and purified proteins were eluted with buffer W con-taining 350 mM imidazole. The pooled eluates were buffer-exchanged with buffer A (150 mM NaCl, 50 mM Tris, pH 7.4) and further purified by concentrating the flow-through from a 30-kDa cutoff ultracentrifugation unit in a 10-kDa cutoff ultracentrifugation unit (Amicon Ultra-15 30K NMWL and 10K NMWL, Millipore). Furthermore, all steps were performed with the same buffers containing 2.5 mM dithiothreitol as reducing agent. The StrepTag II-tagged variant of HoxR was purified by the same protocol using buffer A instead of buffer W and Strep-Tactin Superflow columns (IBA, Germany) for affinity chromatography. Proteins were eluted with buffer A containing 3 mM desthiobiotin and further purified and concentrated as described above.
Hydrogenase Activity Assays-MBH activity was measured in a spectrophotometric assay in a H 2 -flushed cuvette sealed with a rubber septum with methylene blue as the electron acceptor. Activity measurements of isolated membrane fractions were conducted at 30°C in H 2 -saturated 50 mM K-PO 4 buffer at pH 7.0, whereas measurements on purified MBH were carried out at pH 5.5. For reductive reactivation, purified protein (5-10 M MBH in 10% glycerol, 150 mM NaCl, 50 mM K-PO 4 buffer, pH 5.5) was incubated under a 1-bar moisturized H 2 atmosphere at RT, and samples were withdrawn at different time points to determine hydrogenase activity.
UV-visible Absorption Spectroscopy-UV-visible measurements were carried out on a CARY 5000 UV-visible NIR spectrophotometer (Varian). The protein solution (14 -50 M HoxR) was filled in a 100-l quartz cuvette with an optical path length of 1 cm and measured at 15°C against buffer. For reduction of the HoxR samples, sodium dithionite was added to a final concentration of 1 mM.
EPR Spectroscopy-9.5 GHz X-band EPR spectroscopy has been carried out using a Bruker ESP300E spectrometer equipped with a rectangular microwave cavity in the TE102 mode. For low temperature measurements, the sample was kept in an Oxford ESR 900 helium flow cryostat that allows for temperature control between 6 and 100 K (Oxford ITC4). The microwave frequency was detected with an EIP frequency counter (Microwave Inc.). For determination of g values, the magnetic field was calibrated with an external Li/LiF standard with a known g value of 2.002293. Spin quantifications have been performed by comparing the double integrated signal with the signal of a CuSO 4 standard of known concentration. Baseline corrections, if required, were performed by subtracting a background spectrum, obtained under the same experimental conditions from a sample containing only a buffer solution.
Fourier Transform Infrared Spectroscopy-FTIR spectra were recorded on a Bruker Tensor 27 spectrometer, equipped with a liquid nitrogen-cooled MCT detector, with a spectral resolution of 2 cm Ϫ1 . The sample compartment was purged with dried air, and protein samples (400 -500 M) were studied in a temperature-controlled (10°C) gas-tight IR cell (volume 7 l, path length 50 m) with CaF 2 windows. For one spectrum, 200 scans were averaged. The base-line correction was done with a spline function using the OPUS 6.5 software. For reduction, protein samples were incubated under a moisturized H 2 atmosphere for 30 -90 min at room temperature.
Metal Determination-Iron, nickel, and zinc contents of isolated HoxR preparations were quantified by inductively coupled plasma optical emission spectrometry analysis with an Optima 2100 DV from PerkinElmer Life Sciences. The multiple element standard solution XVI (Merck) was used as reference.  Fig. 2A), and only a moderate increase of 25% was observed for the MBH WT protein isolated from the well aerated cells. Remarkably, the activity of the corresponding MBH ⌬hoxR protein increased almost 200% within a time period of 30 min. The wild-type activity was, however, not reached  (Fig. 3A). At 20 K, the spectra were dominated by the typical broad complex signal resulting from the magnetic coupling of the proximal Cys 6 [4Fe-3S] 5ϩ center and the medial [3Fe-4S] ϩ cluster. The spectrum of the MBH ⌬hoxR contained a minor narrow component at 3395 G (g ϭ 2.01), which is characteristic of uncoupled [3Fe-4S] ϩ centers (Fig. 3A, trace b). Relative spin quantification revealed that the Fe-S signals of the mutant protein contain 10 -15% of the uncoupled [3Fe-4S] ϩ species. EPR signals attributed to Ni 3ϩ at the active site of MBH revealed no major difference between the wild-type and mutant proteins (Fig. 3A, traces a and b). The broadened components assigned to the ready Ni r -B state (g x ϭ 2.30, 2973 G; g y ϭ 2.17, 3162 G) result from the magnetic coupling of the Ni 3ϩ to the superoxidized Cys 6 [4Fe-3S] ϩ5 center. At a higher temperature (80 K), the broad nickel signals resulting from the magnetic coupling and all signals derived from the Fe-S centers disap-peared due to fast spin relaxation (Fig. 3A, traces c and d). As a consequence, the proportion of uncoupled narrow Ni r -B-related signals were well resolved (22).

Catalytic Features of MBH Protein Purified from hoxR
Considerably different paramagnetic fingerprints were observed with MBH WT and MBH ⌬hoxR samples isolated from well aerated cells (Fig. 3B). Spin quantification of the Fe-S signals of MBH ⌬hoxR revealed about 50% uncoupled [3Fe-4S] ϩ species (Fig. 3B, trace f) indicating that at least half of the oxidized MBH ⌬hoxR proteins contained an EPR-silent, proximal Fe-S center, although the medial cluster still resided in the EPRactive oxidized state. Moreover, in addition to the Ni r -B-related signals, the MBH ⌬hoxR protein exhibited another component at 3061 G (g ϭ 2.23), which can be assigned to the inactive Ni u -A state (Fig. 3B, trace f) and is commonly observed in O 2 -sensitive [NiFe] hydrogenases. When purified aerobically, these hydrogenases typically exhibit an uncoupled [3Fe-4S] ϩ center and a mixture of Ni u -A and Ni r -B states (both are distinguished only by the g y component (1, 22, 43)).
Quantitative assessment of the Ni u -A:Ni r -B ratio based on the g y components in the MBH ⌬hoxR spectra taken at 20 K revealed a ratio of ϳ1:4 (Fig. 3B, trace f), whereas at 80 K a ratio of 2:3 was observed (Fig. 3B, trace i). This change can be explained on the basis that a significant proportion (ϳ60%) of the Ni r -B species is coupled to the proximal Cys 6 [4Fe3S] 5ϩ center, indicated by a broadening of the g x and g y peaks, which is not detectable at higher temperature. The Ni u -A state in MBH ⌬hoxR , however, appeared to be mainly uncoupled because its g y peak has the same width at 20 and 80 K. This is consistent with the assumption that Ni u -A mainly occurs in damaged MBH proteins lacking the EPR-active superoxidized form of the Cys 6 [4Fe-3S] center. In summary, ϳ50% of the EPR-active Ni-Fe sites in MBH ⌬hoxR protein isolated from well aerated cells resided in the Ni r -B state coupled to the oxidized proximal Cys 6 [4Fe-3S] 5ϩ cluster. About 30 and 20% can be assigned to uncoupled Ni r -B and Ni u -A states, respectively. This is in line with the observation that the superoxidized Cys 6 [4Fe-3S] 5ϩ cluster is detectable in only 50% of the MBH ⌬hoxR preparation, indicating oxidative damage of this cofactor.
To examine whether similar oxidative damage can be induced in MBH WT , the isolated protein was reduced with H 2 and subsequently reoxidized under air. As a result of this treatment, the H 2 oxidation activity of the sample decreased to 10 -30% of the initial activity, and the overall EPR signal intensity was diminished by more than 40%. Furthermore, the Fe-S cluster-related signals in the reoxidized MBH WT indicate the presence of about 50% of the uncoupled [3Fe-4S] ϩ species, which is similar to the observations of as-isolated MBH ⌬hoxR (Fig. 3B, compare traces f and g). Interestingly, Ni u -A was not detected in the reoxidized MBH WT . However, the g y component of Ni r -B showed a splitting that was also detectable in the as isolated MBH ⌬hoxR (Fig. 3B, triangles). The emergence of a similar split signal upon aerobic reoxidation was previously observed in E. coli Hyd-1 variants (44). The results indicate structural alterations of the Ni-Fe site and the proximal cluster in MBH ⌬hoxR , which are similar to those in oxidatively damaged (reoxidized) wild-type MBH.
EPR spectra of H 2 -reduced MBH samples displayed the typical pattern of reduced Fe-S centers in MBH (Fig. 3C, traces k  and l). Notably, neither Ni r -B nor Ni u -A were clearly visible in the reduced MBH ⌬hoxR protein, indicating that both the Fe-S centers and the Ni-Fe site were redox-active. Nevertheless, despite similar protein concentrations, the overall Fe-S cluster spin count in reduced MBH ⌬hoxR amounted to only 20 -25% of the wild-type level. Prolonged incubation under H 2 did not change the signal composition and intensity significantly (Fig.  3C, trace m). According to a recent EPR study on the H 2 -reduced Hyd-1 protein from E. coli, which is quite similar to the ReMBH, signals related to Fe-S clusters originate exclusively from the proximal Cys 6 [4Fe-3S] 3ϩ center (45). Therefore, the spectroscopic observations for H 2 -reduced MBH ⌬hoxR are in agreement with the hypothesis that a major fraction of the hydrogenase molecules contained a damaged and/or EPR-silent proximal cluster.
To investigate if the alterations in MBH ⌬hoxR resulted from the purification procedure, EPR spectroscopy was conducted with an oxidized membrane fraction isolated from the hoxR mutant cells. Again, increased amounts of uncoupled Ni r -B and [3Fe-4S] ϩ species were detected in the membrane samples (data not shown), which coincides with the results obtained for the purified protein. However, a dominant signal at 3400 G (g ϭ 2), probably caused by interfering membrane components, made the assignment and quantification of paramagnetic species difficult. Moreover, the low MBH levels in membranes isolated from well aerated hoxR mutant cells impeded a clear iden-tification of Ni 3ϩ signals that could be assigned to Ni u -A and Ni r -B states.
FTIR Spectroscopy Provides Insights into the Altered Ni-Fe Site in MBH ⌬hoxR -The three diatomic ligands ligated to the iron at the active site can be specifically monitored by FTIR spectroscopy. The stretching vibrations of cyanide and carbon monoxide ligands are highly sensitive toward redox changes at and close to the active site. An overview of redox states in MBH that were assigned on the basis of FTIR is given in Table 1.
The MBH ⌬hoxR and MBH WT proteins isolated from O 2 -limited cells exhibited almost identical infrared absorption spectra (Fig. 4, traces b, c, and d) with major bands at 2097 and 2080 cm Ϫ1 (CN Ϫ stretching bands) and 1948 cm Ϫ1 (CO stretching band). These are characteristic for the ready Ni r -B state (ϳ80% of the overall CO absorption intensity). Also, the general IR profile of the less active MBH ⌬hoxR samples isolated from well aerated cells (Fig. 4, trace a) was rather similar to that of the corresponding wild-type protein. However, integration of the CO absorption area ranging from 1908 to 1964 cm Ϫ1 revealed that the overall signal intensity was decreased by ϳ40% in MBH ⌬hoxR . In addition, a considerable amount of spectral heterogeneity was detected. The maximum of the CO absorption was only slightly shifted (⌬ Ϸ ϩ1 cm Ϫ1 ) compared with Ni r -B, which can be explained by an overlap of absorptions characteristic for Ni u -A and Ni r -B species (11,46). Furthermore, prominent shoulders attributable to the EPR-silent inactive states Ni u -S and Ni ia -S (22) were detected in the mutant protein.
Compared with the MBH WT , the CN Ϫ -related bands in MBH ⌬hoxR were shifted by ϩ1 and ϩ3 cm Ϫ1 (Fig. 4, trace a,  indicated by arrows), respectively, which can be interpreted by an overlap of Ni r -B-and Ni u -A-related absorptions at 2080/ 2097 and 2081/2100 cm Ϫ1 , respectively (46).
In the H 2 -reduced samples, mainly the EPR-silent fully reduced SR states of the active site (CO stretchings at 1944(CO stretchings at , 1926, and 1919 cm Ϫ1 (22)) were detected (Fig. 4, traces e-h), which confirms the results gained from EPR analysis. However, in the reduced MBH ⌬hoxR protein that was isolated from well aerated cells, a minor fraction can potentially be related to Ni u -A or Ni ia -S states (Fig. 4, trace e, marked by the arrows), whereby the latter showed a CO stretching at 1930 cm Ϫ1 (22). This observation indicates that Ni-Fe sites in the MBH ⌬hoxR sample are not entirely redox-active.
To further examine oxidative damage, the H 2 -reduced samples from well aerated cells were reoxidized under air. The IR absorption pattern of the resulting samples (Fig. 4, traces i and  MARCH 14, 2014 • VOLUME 289 • NUMBER 11

Biosynthesis of [NiFe] Hydrogenase under Oxic Conditions
JOURNAL OF BIOLOGICAL CHEMISTRY 7987 j) revealed the accumulation of inactive species such as Ni ia -S in both the MBH ⌬hoxR and MBH WT proteins. Furthermore, another oxidized species was observed and tentatively assigned to Ni ia -SЈ (CO absorption at 1954 cm Ϫ1 , Fig. 4, traces i and j, arrows). Because these inactive species were in part already present in the as-isolated MBH ⌬hoxR from well aerated cells, this result confirmed that MBH in the well aerated hoxR mutant cells is subject to increased oxidative damage. Search for Partners Interacting with HoxR-Previous attempts isolating HoxR from R. eutropha were unsuccessful because only trace amounts of this protein could be purified from cells overexpressing the complete MBH gene cluster (33). Sensitive immunological analysis indicated that HoxR interacts with a transient maturation complex containing the premature small and large MBH subunits as well as several MBH-associated maturases (33).
To explore the interaction of HoxR with MBH auxiliary proteins in more detail, the HoxR-StrepTag II fusion protein was purified via affinity chromatography from MBH-overproducing R. eutropha cells carrying in-frame deletions in either the hoxK or the hoxG gene. The corresponding immunological analysis (Fig. 5A) unambiguously showed that even in the absence of the large subunit HoxG, HoxR copurified with the premature small subunit preHoxK and the HoxK-specific chaperone HoxQ (Fig. 5A, blots 5, 7, and 8). The HoxO protein was not detectable in samples purified from both the hoxG and hoxK deletion mutants (Fig. 5A, blot 9). Interestingly, HoxR was hardly detectable in purifications from the hoxK mutant (Fig.  5A, blot 5). To exclude any effect of the hoxK deletion on the expression of the downstream genes, protein levels of HoxG, HoxQ, and HoxV were also analyzed immunologically in crude cell extracts. Only the HoxQ level was significantly diminished in hoxK mutant cells (Fig. 5A, blots 1, 3, and 4). These results are in line with a HoxG-independent formation of a preHoxK-HoxQ-HoxR complex. Association with HoxO obviously requires the presence of HoxG, whereas the lack of HoxK destabilizes both HoxR and HoxQ.
Is HoxR a Rubredoxin?-To obtain sufficient quantities of HoxR protein for biochemical characterization, HoxR was heterologously overproduced in E. coli cells both as a translational fusion with a C-terminal His tag (HoxR His ) under the control of the T5 promoter and as a fusion with a C-terminal StrepTag II (HoxR Strep ) controlled by the tetA promoter. Even upon expression under the strong T5 promoter, only low HoxR protein levels were observed in the soluble extract as indicated by the lack of a dominant band in Coomassie-stained SDS-polyacrylamide gels (Fig. 5B, lanes SE). The two HoxR protein variants were purified by affinity chromatography and subsequent ultrafiltration (Fig. 5B). Regardless of whether the HoxR samples were prepared in the presence of 3 mM dithiothreitol or without reducing agents, they were completely colorless. Accordingly, UV-visible spectroscopy revealed the lack of absorptions between 300 and 800 nm characteristic for rubredoxins (Fig. 6,  solid line). Furthermore, the isolated proteins precipitated rapidly when incubated at room temperature (data not shown). Metal analysis by inductively coupled plasma optical emission spectrometry yielded a very low iron content of 0.04 to 0.02 mol of iron per mol of purified HoxR.
These observations are in marked contrast to common rubredoxins that have a red color, bind iron at high affinity, and are usually rather stable (47,48). When provided in excess, nickel, zinc, or cadmium can substitute for the iron in rubredoxins (48). Therefore, we investigated HoxR for the presence of these metals via ICP-OES. Zinc and cadmium were not detectable in any of the samples. Not surprisingly, 0.5 mol of nickel per mol was found in His 6 -tagged HoxR but was absent at relevant levels in HoxR Strep samples indicating that nickel binds rather to the His 6 tag than to the HoxR core protein.
High instability and low iron content indicate that HoxR differs significantly from canonical rubredoxins, in which usually two conserved CXXCG motifs constitute the iron-binding site. HoxR and homologous proteins exhibit highly conserved CXXCW and CXXCD signatures (33). To see whether these alterations are structurally and functionally important, we exchanged residues Trp-32 and Asp-65 in HoxR for glycine residues using site-directed mutagenesis. The resulting HoxR W32G-D65G protein was purified (Fig. 7) and turned out to be much more stable than its native counterpart, which was reflected by an almost 10-fold higher yield and increased solu- bility. Furthermore, in contrast to colorless wild-type HoxR, the reddish HoxR W32G-D65G variant revealed the typical absorption maxima of class I rubredoxins (Fig. 6, dotted lines).
To test whether the recombinant HoxR W32G-D65G variant is still functional in MBH maturation under aerobic conditions, plasmids carrying the genes hoxR His-tag or hoxR His tag W32G-D65G under the control of the P hoxF promoter were constructed for complementation experiments. The plasmids were transferred to the strain R. eutropha HF539, which carries an hoxR in-frame deletion. Only the transconjugant strain expressing wild-type hoxR fully recovered MBH-mediated growth at high O 2 (Fig. 8) clearly showing that the signatures in HoxR, which differ from   the canonical motifs of rubredoxins, are essential in preventing detrimental effects caused by O 2 during MBH maturation.

DISCUSSION
The availability of O 2 in the atmosphere in the course of the emergence of oxygenic photosynthesis enabled bacteria to exploit the energy released from enzymatically controlled combustion of H 2 with O 2 (1). As O 2 is a great challenge for enzymatic H 2 cycling and biosynthesis of metalloenzymes in general, ancient O 2 -sensitive [NiFe] hydrogenases had to acquire multiple adaptations to function under aerobic conditions by (4,49). Sophisticated modifications of MBH-like proteins include modified cofactors such as the unique Cys 6 [4Fe-3S] cluster and concomitantly a designated, highly complex MBH maturation machinery. The accessory protein HoxR is one of the crucial factors in MBH biosynthesis at ambient O 2 , because the maturation process is severely impaired in aerobically grown hoxR mutant cells, which exhibit low content and activity of MBH and therefore fail to grow with H 2 as the energy source (33). Biochemical and spectroscopic evidence presented in this study document that in the absence of HoxR, the metal cofactors in MBH are subject to increased oxidative damage during MBH biogenesis.
HoxR Is Not a Common Rubredoxin-The amino acid sequence of HoxR predicts similarity with rubredoxins. Rubredoxins are relatively small (ϳ6 kDa) redox-active proteins, in which the central iron atom is coordinated by four highly conserved cysteines in a tetrahedral arrangement. Although the exact physiological role of most rubredoxins is still elusive, they are anticipated to mediate electron transfer between a number of redox enzymes (50 -54). HoxR proteins show a sequence identity of ϳ50% to class I rubredoxins, which are more abun-dant in strictly anaerobic prokaryotes. The two canonical CXXCG motifs, however, which constitute the coordination sphere of the redox-active site in rubredoxins, are replaced by conserved CXXCW and CXXC(D/E) motifs in HoxR proteins (33). To examine the significance of this modification, we constructed a HoxR W32G-D65G variant containing two CXXCG motifs. Indeed, the mutant proteins proved to be more stable than wild-type HoxR and exhibited a rubredoxin-like UV-visible absorption pattern. Importantly, the altered protein no longer supported MBH-dependent growth at high O 2 , pointing to an essential role of these particular signatures of HoxR proteins in hydrogenase maturation. These observations suggest that HoxR protein separated from the MBH maturation complex is prone to degradation, which overall leads to low HoxR levels in wild-type cells (33). Obviously, HoxR enters an early stage of maturation represented by a complex consisting of the small subunit precursor preHoxK and the chaperone HoxQ. The HoxQ protein is proposed to protect preHoxK in concert with the chaperone HoxO against reactive oxygen species until the cofactor-containing large subunit is delivered for oligomerization (7,32). A similar role has been proposed for the HoxO, HoxQ, HoxR, and HoxT homologs in the legume nodule symbiont R. leguminosarum (31). A concerted action of the HoxR and HoxQ proteins gains support by the presence of hoxR-hoxQ gene fusions denoted hydG (formerly hyaF2) in hydrogenase gene clusters of Salmonella enterica serovar typhimurium and other Salmonella species that encode aerobically synthesized MBH-like proteins (55,56).
HoxR Protein Stabilizes the Cys 6 [4Fe-3S] Cluster-Previously, we have reported clear differences in the UV-visible spectra of MBH proteins isolated from wild-type and hoxR mutant cells (33). Accumulating evidence now leads to the conclusion that HoxR is instrumental particularly in the aerobic maturation of the unique proximal Cys 6 [4Fe-3S] cluster in MBH. (i) EPR spectroscopy of oxidized and reduced MBH isolated from aerobically cultivated hoxR mutant cells revealed severely diminished signals attributed to the proximal cluster. (ii) Genes encoding HoxR proteins are found exclusively in hydrogenase gene clusters encoding MBH-like proteins that are predicted to bear a Cys 6 [4Fe-3S] cluster. (iii) preHoxK and HoxQ copurify with HoxR, even in the absence of the large MBH subunit. (iv) Both HoxR and HoxQ are destabilized in the cells carrying a hoxK deletion.
To date, no detailed information is available on whether the formation of the Cys 6 [4Fe-3S] cofactor requires specific auxiliary factors. It is likely that first a [4Fe-4S] cubane cluster is incorporated into the proximal position of HoxK by the general Fe-S cluster assembly machinery (57). Subsequently, one sulfide has to be eliminated to convert the cubane to the [4Fe-3S] moiety. The asymmetrical and highly flexible conformation of the resulting Cys 6 [4Fe-3S] cluster (Fig. 1C) is anticipated to contribute to increased instability of the cofactor toward oxygen and might explain the need for additional protective proteins such as HoxR, particularly during the maturation process. Indeed, recent quantum chemical investigations revealed that the fully reduced Cys 6  most of the protein backbone and therefore may partially represent the solvent-exposed [4Fe-3S] cluster in the isolated small subunit.
Saggu et al. (22) presented the first evidence for severe damage of the proximal Fe-S cluster and the Ni-Fe site of the MBH when the partially reduced heterodimer reacts with O 2 (22). Because as-isolated MBH that had been chemically oxidized prior to purification remains active and stable under aerobic conditions (19), it is conceivable that HoxR maintains the metal cofactors, especially the proximal cluster, in an oxidized state in the course of the early MBH maturation process. This could be achieved if HoxR adopts the properties of common rubredoxins with their relatively high midpoint potential of around 0 mV (59) only within the maturation complex. The need for association of HoxR with an MBH maturation complex would ensure that the protein does not receive electrons randomly from other cytoplasmic redox compounds. It is noteworthy, that the deletion of a gene encoding a rubredoxin in the cyanobacterium Synechococcus sp. PCC 7002 resulted in the loss of the Fe-S cluster F x of photosystem I. Hence, it has been proposed that the rubredoxin acts as an electron shunt to prevent over-reduction of the labile cluster during photosystem I biogenesis (60).
Upon attachment to the membrane-integral cytochrome b, the MBH heterotrimer reacts fully reversibly with O 2 (22,61). As the cytochrome b and the quinone pool have also rather high midpoint potentials ranging between ϩ10 and ϩ166 mV (10), the mature, resting MBH is most likely kept in the stable Ni r -B state (22). If the mature MBH converts H 2 in the presence of O 2 , the designated electron transfer relay ensures that the attacking oxygen is efficiently reduced to one water molecule and a hydroxyl ligand in the bridging position of the active site (Ni r -B state), thereby preventing the accumulation of detrimental reactive oxygen species (4,7,19).
Interestingly, the EPR and IR spectroscopic data on MBH ⌬hoxR isolated from aerobic cells also revealed alterations of the Ni-Fe site that resembles the oxidized Ni u -A state in a minor fraction of the preparation. Importantly, the respective Ni u -A signals were not magnetically coupled. Therefore, this state is formed only in those proteins lacking an EPR-active superoxidized Cys 6 [4Fe3S] 5ϩ center. According to the model presented by Cracknell and co-workers (20), it is reasonable to conclude that this Ni u -A fraction is the result of altered proximal Fe-S centers no longer being able to provide an electronrich environment to the active site upon O 2 attack. A correlation between the occurrence of Ni u -A and a damaged proximal cluster of the MBH has also been discussed on the basis of spectroscopic features of variants carrying amino acid exchanges close to the Ni-Fe site (46).
Genes encoding HoxR homologs are not necessarily present in gene clusters encoding MBH-like proteins, e.g. the gene clusters of Hyd-1 in Aquifex aeolicus and E. coli are devoid of hoxR. The hydrogenases of these organisms are expressed under anaerobic or microaerobic conditions (62,63), which makes HoxR obviously dispensable (33). Apart from HoxR of R. eutropha, the thioredoxin-like HyaE protein of E. coli, a homolog of R. eutropha HoxO, was suggested to play a role in the biogenesis of the Cys 6 [4Fe-3S] center (49). Structure analysis of HyaE proteins from E. coli and S. enterica revealed thioredoxin-like pro-tein folds (64), which might point to a function in Fe-S cluster assembly. However, the conserved redox-active cysteines of common thioredoxins are substituted for the acidic residues aspartic acid and glutamic acid in the HyaE/HoxO proteins. These residues were suggested to bind transiently to the [4Fe-3S] precursor (49). Our database searches revealed that strict anaerobes such as Clostridia and green sulfur bacteria, which have the coding capacity for MBH-like proteins containing [4Fe-3S] clusters, even lack homologs of all of the MBH-specific accessory genes (hoxLOQRTV). This observation supports their designated function in protecting hydrogenase maturation against the detrimental effects of O 2 (4) raising the question whether two functionally divergent groups of MBH-like proteins exist as follows: (i) aerobically produced respiratory enzymes, such as the MBH from R. eutropha, which are involved in energy conservation, and (ii) anaerobically synthesized [4Fe-3S] cluster-containing hydrogenases that possibly play a role during transient oxygen exposure.