The oxygen-tolerant and NAD+-dependent formate dehydrogenase from Rhodobacter capsulatus is able to catalyze the reduction of CO2 to formate


  • Tobias Hartmann,

    1. Department of Molecular Enzymology, Institute of Biochemistry and Biology, University of Potsdam, Germany
    Search for more papers by this author
  • Silke Leimkühler

    Corresponding author
    1. Department of Molecular Enzymology, Institute of Biochemistry and Biology, University of Potsdam, Germany
    • Correspondence

      S. Leimkühler, Department of Molecular Enzymology, Institute of Biochemistry and Biology, University of Potsdam, Karl-Liebknecht-Str. 24-25, 14476 Potsdam, Germany

      Fax: +49 331 977 5128

      Tel: +49 331 977 5603


    Search for more papers by this author


The formate dehydrogenase from Rhodobacter capsulatus (RcFDH) is an oxygen-tolerant protein with an (αβγ)2 subunit composition that is localized in the cytoplasm. It belongs to the group of metal and NAD+-dependent FDHs with the coordination of a molybdenum cofactor, four [Fe4S4] clusters and one [Fe2S2] cluster associated with the α-subunit, one [Fe4S4] cluster and one FMN bound to the β-subunit, and one [Fe2S2] cluster bound to the γ-subunit. RcFDH was heterologously expressed in Escherichia coli and characterized. Cofactor analysis showed that the bis-molybdopterin guanine dinucleotide cofactor is bound to the FdsA subunit containing a cysteine ligand at the active site. A turnover rate of 2189 min−1 with formate as substrate was determined. The back reaction for the reduction of CO2 was catalyzed with a kcat of 89 min−1. The preference for formate oxidation shows an energy barrier for CO2 reduction of the enzyme. Furthermore, the FMN-containing and [Fe4S4]-containing β-subunit together with the [Fe2S2]-containing γ-subunit forms a diaphorase unit with activities for both NAD+ reduction and NADH oxidation. In addition to the structural genes fdsG, fdsB, and fdsA, the fds operon in R. capsulatus contains the fdsC and fdsD genes. Expression studies showed that RcFDH is only active when both FdsC and FdsD are present. Both proteins are proposed to be involved in bis-molybdopterin guanine dinucleotide modification and insertion into RcFDH.


formate dehydrogenase


inductive coupled plasma–optical emission spectrometry


molybdopterin guanine dinucleotide


molybdenum cofactor




sodium dithionite


Rhodobacter capsulatus formate dehydrogenase


surface plasmon resonance


Formate dehydrogenases (FDHs) comprise a heterogeneous group of enzymes that catalyze the oxidation of formate to CO2 and H+ [HCOO = CO2 + H+ + 2e; E° = – 461 mV (pH 8)] [1, 2]. FDHs are found in anaerobic bacteria and archaea, in addition to facultative anaerobic or aerobic bacteria and yeast [2]. These enzymes are structurally diverse, and catalyze the oxidation of formate by the use of different cofactors, such as the molybdenum cofactor (Moco), FeS clusters, and/or FMN, or without any cofactor in case of the yeast enzymes [3]. In addition to the molybdenum-containing FDHs, some enzymes contain tungsten in place of molybdenum at the active site [4, 5]. Additionally, certain FDHs were reported to act as a CO2 reductases, which preferentially catalyze the reverse reaction [6-8]. FDHs that catalyze the CO2 reduction reaction are of interest for the sequestration of CO2 and for the production of formate as a stabilized form of hydrogen fuel. Formate or formic acid as solid fuel has the advantage of safe portability as compared with liquid fuels such as methanol, which is toxic and highly flammable. In organisms that catalyze the oxidation of formate, this reaction is a key process for obtaining energy and reducing equivalents. The CO2 formed is usually incorporated predominantly into the ribulose 1,5-bisphosphate cycle, as in autotrophic organisms [9, 10]. Metal-dependent FDHs from prokaryotic organisms are members of the dimethylsulfoxide family of mononuclear molybdenum-containing or tungsten-containing enzymes [1]. In these enzymes, the active site in the oxidized state comprises a hexacoordinated molybdenum or tungsten atom in a distorted trigonal prismatic geometry, which is coordinated by two dithiolene groups from two molybdopterin (MPT) moieties, an additional sulfido-ligand, and a protein-derived selenocysteine or cysteine ligand [1]. The other two enzyme families of molybdenum enzymes are the xanthine oxidase family and the sulfite oxidase family [11]. Enzymes of the xanthine oxidase family contain one MPT moiety, one oxo-ligand, one sulfido-ligand and one hydroxo-group coordinated to the molybdenum atom [12]. In contrast, enzymes of the sulfite oxidase family contain two oxo-ligands and one amino acid ligand from the protein, which is usually a cysteine [11]. The crystal structures of three metal-containing FDHs have been reported: two are the molybdenum-containing enzymes FDH-H and FDH-N from Escherichia coli, and one is the tungsten-containing FDH from Desulfovibrio gigas [13-15]. The large subunit of these enzymes contains the bis-MGD cofactor and one [Fe4S4] cluster. Two conserved arginine and histidine residues close to the molybdenum or tungsten atom were suggested to have a key role in catalysis, in addition to a lysine, which is proposed to be involved in electron transfer. Kinetic and spectroscopic data of E. coli FDH-H showed that the oxidation of formate starts with the cleavage of the C–H bond, resulting in CO2, a proton, and an electron. The proton of the substrate was proposed to be transferred to an acceptor group located near the molybdenum ion, which can be the selenocysteine or a conserved histidine at the active site [14, 16]. However, two different mechanisms for the reaction of formate oxidation have been proposed in the literature, so that other studies in addition to site-directed mutagenesis are necessary to investigate the catalytic mechanism further [14, 17]. The major problems in studying the selenium-containing enzymes are their extreme oxygen lability and the lack of suitable expression systems for the expression of large amounts of enzyme. Oxygen-tolerant FDHs are usually cytoplasmic, are composed of different subunits, and contain a cysteine instead of a selenocysteine at the active site. So far, enzymes from Pseudomonas oxalaticus, Ralstonia eutropha H16, Clostridium carboxidovorans and Methylosinus trichosporium OB3b have been purified and characterized [2, 3, 7, 18]. These enzymes are composed of different domains, and are either heterotetramers or dimers of tetramers. In general, the additional subunits contain several [Fe4S4] and [Fe2S2] clusters and FMN as cofactor. In these enzymes, NAD+ is used as the terminal electron acceptor. A crystal structure of this class of FDHs is not available so far.

Rhodobacter capsulatus is a facultative anaerobic and phototrophic purple nonsulfur bacterium [19]. The molybdoenzymes dimethylsulfoxide reductase and xanthine dehydrogenase have been characterized in detail from R. capsulatus [20-22]. Here, we identified a third molybdoenzyme in R. capsulatus, which was predicted to be a cytoplasmic FDH. This enzyme is encoded by the fds operon comprising five genes – fdsG, fdsB, fdsA, fdsC, and fdsD – which are located downstream of the moaD2 and moaE genes in the R. capsulatus genome. We expressed the operon in a heterologous expression system in E. coli, and characterized the purified protein. FDH from R. capsulatus (RcFDH) consists of an (αβγ)2 heterotrimer in which the large α-subunit FdsA (105 kDa) harbors the bis-molybdopterin–guanine dinucleotide (MGD) cofactor, and a set of four [Fe4S4] clusters and one [Fe2S2] cluster. FdsA is linked to the β-subunit FdsB (52 kDa), which binds one additional [Fe4S4] cluster and the FMN cofactor. The γ-subunit FdsG (15 kDa) binds an [Fe2S2] cluster. We characterized the kinetic constants and analyzed the back reaction for CO2 reduction. Additionally, we investigated the properties of the separately purified FdsGB diaphorase subunit for NADH regeneration. Furthermore, we propose a function for the so far uncharacterized proteins FdsC and FdsD in cofactor maturation and insertion into FDH.


Expression and purification of R. capsulatus fdsGBACD in E. coli

An operon containing the genes fdsG, fdsB, fdsA, fdsC and fdsD coding for a cytoplasmic FDH was identified downstream of the moaD2 and moaE genes in the R. capsulatus genome (Fig. 1A). The operon structure is identical to the one described for Ra. eutropha [23]. Amino acid sequence comparisons showed homologies with complex I, NiFe-hydrogenase, nitrate reductase, and related FDHs from organisms such as Ra. eutropha [23]. The 105-kDa FdsA subunit showed amino acid sequence homologies with Ra. eutropha FdsA, M. trichosporium FdsA, E. coli FdhF, and E. coli NuoG, containing conserved motifs for the binding of the bis-MGD cofactor in addition to four [Fe4S4] clusters and one [Fe2S2] cluster (Fig. S1A). The 52-kDa FdsB showed amino acid sequence homologies with Ra. eutropha FdsB, E. coli NuoF, and Ra. eutropha HoxF, suggesting the presence of FMN and an [Fe4S4] cluster (Fig. S1B). The 15-kDa FdsG subunit showed amino acid sequence homologies with Ra. eutropha FdsG, E. coli NuoE, and Ra. eutropha HoxF, comprising the binding site for an [Fe2S2] cluster (Fig. S1C). FdsC, with a size of 26 kDa, showed amino acid sequence homologies with Ra. eutropha FdsC and E. coli FdhD, a chaperone involved in the formation of the sulfurated bis-MGD cofactor in E. coli. The 7-kDa FdsD subunit showed amino acid sequence homologies only with other counterparts from organisms such as Ra. eutropha and M. trichosporium FdsD, and is so far uncharacterized. To characterize RcFDH, the operon fdsGBACD was cloned into the vector pTrcHis for heterologous expression in E. coli MC1061 cells. This construct expresses RcFDH as an N-terminal fusion of FdsG to a His6-tag. After expression, the protein was purified by Ni2+–nitrilotriacetic acid and size-exclusion chromatography. With this procedure, RcFDH was obtained with 90–95% purity and a yield of 6 mg per liter of E. coli cells (Fig. 1B). After size-exclusion chromatography, three major peaks were detected on an SDS/PAGE gel, corresponding to the calculated molecular masses of approximately 100 kDa, 50 kDa, and 15 kDa, representing FdsA, FdsB, and FdsG, respectively (Fig. 1B). Bands corresponding to masses of 26 kDa for FdsC or 7 kDa for FdsD were not identified after SDS/PAGE. Furthermore, analytical size-exclusion chromatography showed that RcFDH forms an (αβγ)2 dimer of heterotrimers in solution, with a molecular mass of ~ 340 kDa (Fig. 1B,C). However, two major peaks were eluted from the analytical size-exclusion column, with a second peak corresponding to a molecular mass of 60 kDa. SDS/PAGE revealed that the second peak contained only FdsG and FdsB. This represents only the βγ dimer of the diaphorase unit of FDH, lacking the molybdenum-containing FdsA. In total, the purified (αβγ)2 RcFDH was stable in phosphate buffer, with 50% loss of activity after 4 days at 4 °C. The addition of sodium nitrate to the buffer resulted in an increase in the stability of RcFDH, as reported before for Ra. eutropha [2]. FDH activity assays showed that RcFDH was not inactivated by molecular oxygen, and anaerobic purification did not result in an increase in stability or activity (data not shown). Thus, all further purifications and characterizations were performed under aerobic conditions.

Figure 1.

(A) Operon structure for subunit composition of FDH from R. capsulatus. The fds operon downstream of moaD1 and moaE contains five genes: fdsG, fdsB, fdsA, fdsC, and fdsD. (B) Size-exclusion chromatography of RcFDH. Size-exclusion chromatography was performed with a Superdex 200 column equilibrated in 75 mm potassium phosphate buffer (pH 7.5) containing 10 mm KNO3. Eluted protein peaks were recorded at 280 nm and 444 nm. Shown is the elution profile of 5 μm RcFDH and the corresponding 15% SDS/PAGE gels separating proteins from collected fractions of the elution maxima (arrows). Inset: plot of the standard proteins (Bio-Rad): thyroglobulin (670 kDa), gamma-globulin (158 kDa), ovalbumin (44 kDa), myoglobin (17 kDa), and vitamin B12 (1.3 kDa). (C) Model of the proposed subunit composition and cofactor content of RcFDH. RcFDH is a hexamer with an (αβγ)2 composition. FdsA harbors the bis-MGD cofactor, four [Fe4S4] clusters, and one [Fe2S2] cluster. FdsB binds the FMN cofactor and one [Fe4S4] cluster. FdsC binds one additional [Fe2S2] cluster.

Analysis of the cofactor composition of RcFDH

To determine the cofactor composition of the purified RcFDH, we quantified the amount of protein-bound molybdenum and iron by inductive coupled plasma–optical emission spectrometry (ICP-OES). Taking into account the predicted presence of five [Fe4S4] and two [Fe2S2] clusters from the amino acid sequence alignment, RcFDH showed an overall iron saturation of 48%. The molybdenum content was calculated to be 39%, and the flavin content was calculated to be 68% (data not shown). Thus, the enzyme was shown to be not completely saturated with its cofactors. Changing the expression conditions or the expression strain did not result in an increase in cofactor occupancy (data not shown). To verify the bis-MGD structure, the protein-bound Moco was oxidized to its stable degradation product Form A-GMP, with the method described by Johnson et al. [24] (Fig. 2A). The detection of Form A-GMP and the detection of GMP released from Moco (data not shown) confirmed that RcFDH binds the MGD cofactor. The Form A peak shown in Fig. 2A, which was also separated on the HPLC column, is probably the result of the hydrolysis of GMP during the release of the bis-MGD cofactor from the protein by acid treatment, as, so far, no free Mo-MPT has been detected in enzymes of the dimethylsulfoxide reductase family.

Figure 2.

(A) Analysis of the Moco present in RcFDH. The Moco of RcFDH was converted to Form A and Form A-GMP after treatment with acidic I2/potassium iodide, with the method of Johnson et al. [24]. Shown is the HPLC profile on a reversed-phase C18 octadecylsilane hypersil column, with relative fluorescence detected with excitation at 383 nm and emission at 450 nm. (B) UV–visible absorption spectra of RcFDH. Shown are the spectrum of the oxidized enzyme (solid line) and the spectrum of the enzyme reduced with 10 mm sodium formate (dashed line). Total reduction of the sample was achieved by addition of 10 mm NDT (dotted line). Spectra were recorded in 75 mm potassium phosphate buffer (pH 7.5) containing 10 mm KNO3.

The UV–visible spectrum of RcFDH in its oxidized form is similar to the one identified for purified Ra. eutropha FDH [2]. Characteristic are the typical shoulders resulting from [Fe4S4] and [Fe2S2] clusters, and the absorption maximum at 444 nm for the FMN cofactor (Fig. 2B). To calculate the amount of catalytically active FDH in solution, reduction spectra were recorded after the addition of 10 mm sodium formate and 10 mm sodium dithionite (NDT) (Fig. 2B). The results showed that 41% of RcFDH was reduced with sodium formate, in comparison with the fully reduced enzyme with NDT. This is consistent with the determined molybdenum saturation of 39%. In addition, this shows that the enzyme was purified in a form with full saturation of the terminal sulfido-ligand at the molybdenum site. Unfortunately, the sulfido-ligand could not be confirmed by cyanide treatment and quantification of the formed SCN. During the reaction, cyanide inactivated the enzyme, and resulted in the release of inorganic sulfide from the FeS clusters, so that the origin of the sulfide detected as SCN could not be verified (data not shown). However, the reduction spectrum implies that RcFDH was purified with saturation of 39% of the cyanolyzable sulfur, consistent with the molybdenum content.

Steady-state kinetics of RcFDH

The kinetic constants of purified RcFDH were determined following the reduction of NAD+ as the terminal electron acceptor. The pH dependence was determined with a range of buffers from pH 4.5 to pH 9.5 (Fig. 3A) with overlapping buffer systems, and showed a pH optimum at pH 9.0 in 100 mm Tris/HCl buffer. The temperature optimum was determined to be 40 °C (Fig. 3B). However, as R. capsulatus has a growth optimum at 30 °C, all kinetic constants were determined at 30 °C. As shown in Table 1A, the Km values for NAD+ and formate were calculated to be 173 ± 19 μm and 281 ± 10 μm, respectively, with a kcat of 2189 ± 38 min−1. Taking into account that only 41% of the purified protein was catalytically active, the kcat of a fully active enzyme can be calculated as 5339 min−1. The double reciprocal plots with varying formate and NAD+ concentrations revealed the formation of a ternary complex, in which both substrates are bound before product release (Fig. 3C,D). When 2,6-dichlorophenolindophenol or ferricyanide was used as the acceptor, kcat values of 1918 ± 82 min−1 (2,6-dichlorophenolindophenol) and 2380 ± 64 min−1 (ferricyanide) were determined, showing that electron transfer to the acceptor is probably not the rate-limiting step of the reaction.

Figure 3.

(A) pH dependency of RcFDH activity. The pH optimum was determined by measuring activities of RcFDH with formate and NAD+ in the pH range of 4.5–9.5 in 100 mm citrate/acetate buffer (round dots), 100 mm potassium phosphate buffer (squares), and 100 mm Tris/HCl (triangles). (B) Temperature optimum of RcFDH activity. The temperature optimum was determined by measuring RcFDH activities with formate and NAD+ in 75 mm potassium phosphate buffer (pH 7.7) in the range of 15–60 °C. (C) Lineweaver–Burk plot of RcFDH activities with constant formate concentrations and different NAD+ concentrations. (D) Lineweaver–Burk plot of RcFDH activities with constant NAD+ concentrations in and different formate concentrations.

Table 1. Steady-state kinetics of RcFDH. (A) FDH activity with formate as substrate. (B) FDH activity with NADH as substrate. Kinetic parameters were determined with a range of 0.05–5 mm of sodium formate and NAD+; 0.1 mm 2,6-dichlorophenolindophenol or 1 mm ferricyanide was used as an alternative electron acceptor; 0.2 mm NADH was used as the electron donor for CO2 reduction
ProteinElectron acceptorkcat (min−1)Km Formatem)Km NAD+m)
RcFDHNAD+2189 ± 38281 ± 10173 ± 19
Ferricyanide2380 ± 64278 ± 29
Dichlorophenolindophenol1918 ± 82277 ± 43
ProteinElectron acceptorkcat app. (min−1)Km app. NADHm)
  1. ND, not detectable.

RcFDHCO289 ± 1.0ND

Some FDHs have been shown to coordinate tungsten instead of molybdenum at the active site. In order to replace the molybdenum atom in the bis-MGD cofactor with tungsten, cells were grown in the presence of 10 mm tungsten without supplementation with molybdenum. After expression, the purified enzyme was shown to be completely inactive and devoid of Moco (data not shown), showing that molybdenum is specifically bound to RcFDH and can not be replaced by tungsten.

To analyze the back reaction for the reduction of CO2 to formate, NADH was used as the electron donor in a bicarbonate saturated solution. Under these conditions a kcat of 89 ± 1 min−1 was determined (Table 1B). The colorimetric assay described by Lang and Lang [25] confirmed the presence of formate formed during the reaction (data not shown). However, substrate inhibition at an NADH concentration of 300 μm was determined. This is consistent with a previous report on the soluble hydrogenase from Ra. eutropha H16, where substrate inhibition with concentrations of 200 μm NADH and higher was also observed [26].

FdsG and FdsB form an active diaphorase unit

Amino acid sequence comparisons of RcFDH with the soluble hydrogenase from Ra. eutropha shows high identities between the HoxF and HoxU subunits and R. capsulatus FdsG and FdsB. HoxF and HoxU were recently shown to form an efficient diaphorase module [26]. To determine the kinetic constants of FdsGB for NAD+ reduction and NADH oxidation, we purified the FdsGB heteromer after separate expression of fdsG and fdsB in E. coli (Fig. 4). The purified enzyme showed full saturation with both FeS clusters, and UV–visible spectroscopy confirmed the presence of the FMN cofactor in stoichiometric amounts. Determination of the steady-state kinetic parameters showed that FdsGB expressed activity for both NAD+ reduction and NADH oxidation, with ferricyanide as electron acceptor or methylviologen as electron donor, respectively (Table 2). The Km and kcat values were determined to be 1050 ± 148 μm and 1916 ± 110 min−1, respectively, for NAD+ reduction, and 129 ± 15 μm and 413 ± 10 min−1, respectively, for NADH oxidation. However, much higher turnover rates were obtained for the RcFDH (αβγ)2 heterohexamer, with a kcat of 9781 ± 310 min−1 for NAD+ reduction, and a kcat of 3566 ± 133 min−1 for NADH oxidation. In general, the affinity for NAD+ was high, with a Km value of 349 μm. Determination of the Km value for NADH was not possible, owing to substrate inhibition.

Table 2. Steady-state kinetic parameters of the FdsGB diaphorase subunit in comparison with RcFDH. (A) Kinetic parameters with NAD+ as electron acceptor. (B) Kinetic parameters with NADH as electron donor. Methylviologen was used as electron donor and ferricyanide as electron acceptor
ProteinElectron donorkcat app. (min−1)Km app. NAD+m)
R. capsulatus FdsGBMethylviologen1916 ± 1101050 ± 148
RcFDHMethylviologen9781 ± 310349 ± 41
ProteinElectron acceptorkcat app. (min−1)Km app. NADHm)
  1. ND, not detectable.

R. capsulatus FdsGBFerricyanide413 ± 10129 ± 15
RcFDHFerricyanide3566 ± 133ND
Figure 4.

UV–visible absorption spectrum of separately purified FdsGB. Shown is the UV–visible spectrum of oxidized FdsGB in 75 mm potassium phosphate buffer in addition to 10 μL of the protein separated on a 15% SDS/PAGE gel. The subunits are indicated by B (β-subunit, FdsB, m = 52 kDa) and G (γ-subunit, FdsG, m = 15 kDa + 2.5-kDa His6-tag). m, molecular mass marker.

FdsC and FdsD are essential for FDH maturation

In addition to the structural genes fdsG, fdsB and fdsA for RcFDH, the fds operon encodes two other proteins, FdsC and FdsD, which are not subunits of mature RcFDH. Studies characterizing FDH from Ra eutropha H16 and M. trichosporium OB3b showed FdsD to be part of the active FDH [2, 3]. Amino acid sequence comparisons categorized FdsD as a member of the FdsD superfamily predicted to have a role as a stabilizing protein in maintaining the quaternary structure of its target enzyme. However, after expression of the fdsGBACD operon, purified RcFDH did not show a band corresponding to the molecular mass of FdsD on Coomassie-stained or silver-stained SDS/PAGE gels (Fig. S2). In contrast, FdsC was classified as a member of the FdhD/NarQ superfamily, which constitutes a family of accessory proteins that are necessary for the activities of their respective target enzymes. Sequence alignments with FdsC revealed an amino acid sequence identity of 34% with FdhD from E. coli.

To analyze the role of FdsC and FdsD in the maturation of RcFDH, we expressed fdsG, fdsB and fdsA in the presence or absence of either fdsC or fdsD in E. coli cells. When fdsG, fdsB, fdsA and fdsD were expressed in E. coli MC1061 cells, which contain a copy of E. coli fdhD, an active RcFDH was obtained (data not shown). However, the expression of fdsG, fdsB, fdsA and fdsD in an E. coli ΔfdhD strain resulted in an inactive RcFDH, showing that E. coli FdhD and R. capsulatus FdsC can perform each other's role in FDH maturation. The fdsG, fdsB and fdsA genes were expressed in the E. coli ΔfdhD strain in the absence of both fdsC and fdsD, or only in the presence of either fdsD or fdsC, and FdsGBA was purified and characterized. The purified protein variants were analyzed for their total activity, metal saturation, and Form A-GMP content (Fig. 5). The data were normalized to 100% in the BW25113 ΔfdhD strain in which the proteins were expressed. However, the specific activity of the purified FDH was decreased in the BW25113 ΔfdhD strain (kcat app. = 1730 min−1) in comparison with the enzyme expressed in the MC1061 strain (kcat = 2189 min−1), owing to a lower level of bis-MGD insertion in this E. coli strain. The results show that purified FdsGBA is only active when expressed in the presence of both FdsC and FdsD. Analysis of the molybdenum content and the Form A-GMP content showed that FdsD influences Moco insertion, as the molybdenum content was reduced by ~ 50% in the strains lacking FdsD. In the absence of FdsC, the molybdenum content remained the same. However, as the protein was inactive, we suspect that FdsC has a role in the sulfuration of bis-MGD, as reported before for its counterpart E. coli FdhD [27]. The absence of the sulfido-ligand at Moco subsequently results in an active enzyme. To underline the role of both FdsC and FdsD in the maturation of RcFDH, protein–protein interaction studies were performed with surface plasmon resonance (SPR) (Table 3). Purified RcFDH, FdsC and FdsD were immobilized on a sensor chip. The interaction was analyzed with purified FdsC, FdsD, and RcFDH, in addition to Moco-free apo-RcFDH. The results showed that FdsC interacted with RcFDH (apo-form and holo-form) with KD values of 0.41 ± 0.03 μm and 1.54 ± 0.23 μm for the apo-form and holo-form, respectively, when FdsC was immobilized, and a KD value of 0.80 ± 0.45 μm when holo-RcFDH was immobilized (Table 3). Interestingly FdsD only showed binding with FdsC with KD values of 0.72 ± 0.31 μm when FdsD was immobilized and 6.39 ± 4.79 μm when FdsC was immobilized. No interaction of FdsD with apo-RcFDH or holo-RcFDH was identified.

Table 3. Analysis of protein–protein interactions between RcFDH, FdsD and FdsC by SPR. Mean values from four independent measurements are shown
AnalytesKD values (μm) obtained with immobilized proteins
RcFDHR. capsulatus FdsCR. capsulatus FdsDBSA
  1. ND, no interaction was determined.

Apo-RcFDH0.20 ± 0.100.41 ± 0.03NDND
RcFDH1.39 ± 0.981.54 ± 0.23NDND
R. capsulatus FdsC0.80 ± 0.451.80 ± 0.660.72 ± 0.31ND
R. capsulatus FdsDND6.39 ± 4.79NDND
Figure 5.

Detection of the cofactor content of purified RcFDH in relation to the presence of FdsD or FdsC during expression. Protein expression from plasmids pTHfds03, pTHfds04, pTHfds07 and pTHfds12 was carried out in a E. coli ΔfdhD mutant strain to exclude the influence of E. coli FdhD on RcFDH maturation. Purified FDH was analyzed for formate:NAD activity, and Mo, Fe and MGD (as Form A-GMP) content. Results from the expression of FdsGBA in the presence of FdsC and FdsD were set to 100%.


There is increasing interest in the biotechnological application of FDHs, as formate is suitable as an alternative fuel. In addition, the electrochemical reduction of CO2 to formate has been described [8], suggesting a suitable model not only for producing formate for biofuel applications, but also for using the accumulating greenhouse gas CO2 in the atmosphere. The main problems in working with most FDHs published so far are protein instability, sensitivity to oxygen, and low turnover rates [28-30]. In addition, most FDHs, like those studied in E. coli, require toxic selenium in their active sites, hindering large-scale protein expression, and the oxygen sensitivity of the proteins necessitates purification under anaerobic conditions [14, 31]. In order to avoid these problems, we purified and characterized a stable oxygen-tolerant and so far uncharacterized FDH from R. capsulatus. We have shown that the enzyme is able to perform the reaction of CO2 reduction in the presence of NADH. Whereas the number of cofactors and catalytic properties of RcFDH are similar to those of cytoplasmic FDHs from other organisms, our studies have shown that RcFDH is an (αβγ)2 heterotrimer, binding the bis-MGD cofactor in addition to four [Fe4S4] clusters and one [Fe2S2] cluster in the FdsA subunit. FdsB was predicted to bind one [Fe4S4] cluster and one FMN, and FdsG was predicted to bind one [Fe2S2] cluster. In our study, FdsD was not shown to be stably associated with the mature protein complex. In other studies, such as for the FDH from Ra. eutropha, the subunit composition was interpreted as an αβγδ heterotetramer [2]. Instead of selenocysteine, RcFDH harbors a cysteine ligand at the active site. Investigation of the catalytic mechanism of RcFDH showed that NAD+ is the physiological electron acceptor during formate oxidation. The double reciprocal plots with varying formate and NAD+ concentrations revealed a ternary complex mechanism, showing that both substrates bind before product release. RcFDH shows kcat values of 2189 min−1 for formate oxidation and 89 min−1 for NADH+ reduction. Reduction spectra showed that 41% of purified RcFDH was in a catalytically active state in solution. Calculated for a fully active protein (5339 min−1, 31 U·mg−1), these kinetic constants are consistent with the one determined for P. oxalaticus FDH (32 U·mg−1), but lower than those for Ra. eutropha H16 FDH (98 U·mg−1) and M. trichosporium OB3b FDH (80 U·mg−1).

In addition to the structural genes fdsG, fdsB and fdsA for FDH, two so far uncharacterized genes are present in the fds operon, named fdsC and fdsD. Our studies show that both FdsC and FdsD are involved in the production of an active RcFDH, but they are not subunits of the mature enzyme. FdsC shows high sequence similarities with FdhD from E. coli. E. coli FdhD was previously shown to be essential for the formation of active respiratory FDH in E. coli [32]. A recently published study on E. coli FdhD showed a direct role as sulfurtransferase, the enzyme being involved in sulfuration of the bis-MGD cofactor for E. coli FDH-H [27]. Our study showed that, when FdsC was absent, an active RcFDH was only obtained from the heterologous expression system when E. coli FdhD was also present. This shows that E. coli FdhD is capable of replacing FdsC in its role in RcFDH maturation. This implies that FdsC has a similar role to E. coli FdhD, and is involved in the addition of a sulfido-ligand to bis-MGD in RcFDH. The obtained interaction between FdsC and RcFDH in the SPR experiment supports this hypothesis. We showed that RcFDH expressed in the absence of FdsC in a E. coli ΔfdhD strain was inactive but contained bis-MGD (and there was consequently no impairment of cofactor insertion); therefore, it is likely that FdsC modifies the bis-MGD cofactor by adding a sulfur compound, which is essential for the catalytic activity. Whereas the role of FdsC seems to be similar to that of E. coli FdhD, a homolog of RcFdhD does not exist in E. coli. FdsD is present in other homologs of RcFDH comprising a group of cytoplasmic FDHs that are NAD+-dependent. Our studies show that the absence of FdsD led to an inactive RcFDH with a subtantially reduced bis-MGD composition. As the SPR studies revealed that FdsD only interacts with FdsC and not with RcFDH (apo-form or holo form), FdsD might be involved in insertion and/or stabilization of the bis-MGD cofactor bound to FdsC rather than to the apo-FDH. The role of FdsD might thus be restricted only for soluble FDHs, which are oxygen-tolerant. Thus, the role of FdsD might also be to protect the bis-MGD cofactor from oxidation under aerobic conditions when bound to FdsC. Proteins with a similar role have been described, for example, for R. capsulatus XDH, for which the molecular chaperone XdhC is involved in Moco insertion [33]. For E. coli FDH, the mechanism of bis-MGD insertion might be controlled in a different manner, as the protein is located in the membrane (FdhF) or in the periplasm (FdoG or FdnG), and here the maturation is controlled by the TAT pathway [34]. In conclusion, both FdsC and FdsD are essential for the maturation of active RcFDH, with a role in bis-MGD incorporation and modification by insertion of a terminal sulfido-ligand in bis-MGD.

RcFDH exclusively binds the bis-MGD cofactor with molybdenum as the active site metal, which could not be substituted for by tungsten. Molybdenum and tungsten have similar chemical properties, and some molybdenum-containing enzymes have been shown to be active when molybdenum is replaced by tungsten, such as R. capsulatus dimethylsulfoxide reductase [35-37]. Tungsten-containing enzymes from the family of molybdoenzymes were shown to catalyze low-potential reactions, so tungsten-containing FDHs are predicted to catalyze the reaction of CO2 reduction more easily [38]. A study by Reda et al. [8] characterized the reversible reaction of CO2 reduction by using the tungsten-containing FDH from Synthrobacter fumaroxidans. Here, the enzyme was adsorbed on an electrode surface, and was able to catalyze the efficient electrochemical reduction of CO2 to formate after application of small overpotentials. However, formate oxidation was more than five times faster than CO2 reduction with this enzyme [8]. Our studies showed that molybdenum-containing RcFDH is able to perform both reactions in solution. However, the forward reaction of formate oxidation was 24 times faster than the reaction of CO2 reduction, which was achieved by using high concentrations of NADH as substrate. Thus, electron potentials of the FeS clusters close to the molybdenum atom might hinder electron flow from the flavin to the molybdenum center. Thus, applying an overpotential with an electrochemical approach after coupling of RcFDH to an electrode might facilitate the back reaction. However, our results show that the back reaction is possible with RcFDH without overpotentials. As the enzyme is oxygen-tolerant and stable at room temperature, it might be a suitable enzyme for fuel cell applications in the future.

Besides its role in formate oxidation, RcFDH also shows diaphorase activity at the FMN cofactor site. FdsG coordinates the flavin and one additional [Fe4S4] cluster. FdsG and FdsB, which binds one [Fe2S2] cluster, form an active diaphorase unit, which could be separately purified and showed similarity to the HoxF and HoxU subunits characterized for the soluble hydrogenase of Ra. eutropha H16 [26]. R. capsulatus FdsGB showed activities in catalyzing both NAD+ reduction (413 ± 10 min−1) and NADH oxidation (1916 ± 110 min−1) reactions. However, much lower turnover rates were obtained than with the full-length FDH (NAD+ reduction, 3566 ± 133 min−1; NADH oxidation, 9781 ± 310 min−1). In general, both RcFDH and FdsGB showed higher turnover rates for NAD+ reduction, which shows that this direction is the preferred electron transfer path. Thus, the FdsGB diaphorase unit could also be used in cofactor regeneration systems for NAD+ or NADH production in the future.

RcFDH contains a cysteine instead of a selenocysteine in the active site. Replacement of the selenocysteine by cysteine in E. coli FdhF resulted in a reduction in kcat from 2800 s−1 to 9 s−1, a factor of 300 [39]. Here, it was argued that, because of the difference in the pKa values of cysteine (8.2) and selenocysteine (5.2), cysteine is not able to perform the reaction, as, for the reaction mechanism, the residue has to be deprotonated [1]. The selenocysteine ligand was proposed to act as acceptor of protons from molybdenum-bound formate. RcFDH with a cysteine at the active site is approximately 10 times faster than the cysteine-containing E. coli FdhF variant. The reaction mechanism of RcFDH will be investigated in future studies in more detail; however, it is expected to be similar to that of E. coli FdhF, as the crucial amino acids involved in catalysis (Lys295, His387, and Arg587) are conserved in RcFDH. We support the recent model of Mota et al., showing the influence of the terminal sulfido-ligand of bis-MGD in FDH. Our studies suggest that, in RcFDH, this sulfido-ligand is also present and essential for catalytic activity, as revealed by the role of FdsC.

Experimental procedures

Cloning of fds genes

The construct for expression of the R. capsulatus fdsGBACD operon was cloned after PCR amplification of the gene with respective primers, with chromosomal R. capsulatus DNA as template. The fdsGBACD DNA fragment was ligated into the pTrcHis vector with NheI–SalI restriction sites, and the resulting plasmid was designated pTHfds05. The fdsG, fdsB and fdsA genes were introduced into the pTrcHis vector after PCR amplification with the NheI–SalI restriction sites, and the resulting plasmid was designated pTHfds04. For coexpression, fdsC, fdsD and fdsCD were cloned into the NdeI–XhoI sites of the pACYCduet-1 vector, and the resulting plasmids were designated pTHfds03, pTHfds12, and pTHfds07, respectively. The fdsG and fdsB genes encoding the diaphorase subunit of RcFDH were cloned into the pTrcHis vector by use of the NheI–SalI sites.

Expression of RcFDH and FdsGB

RcFDH was expressed in E. coli MC1061 cells transformed with pTHfds05. Cells were grown at 30 °C and 130 r.p.m. in LB medium containing 1 mm molybdate, 20 μm isopropyl thio-β-d-galactoside, and 150 μg·mL−1 ampicillin, beginning with a 1 : 500 dilution of a preculture (same supplements, 12 h, 37 °C, no shaking), for 24 h. For the tungsten-containing RcFDH, cells were expressed in the presence of 10 mm tungstate instead of molybdate. The apo-form of RcFDH lacking the bis-MGD cofactor was obtained after expression in E. coli RK5200 ΔmoaA cells [40] without supplementation with molybdate. FdsG and FdsB were produced under the same conditions as used for RcFDH, but without addition of molybdate. For all coexpressions, the E. coli BW25113 ΔfdhD strain was used [41]. After growth, cells were harvested, resuspended in 50 mm NaH2PO4 and 300 mm NaCl (pH 8.0), and used directly for purification or stored at – 20 °C.

Purification of proteins

All purification steps were performed at 4 °C on ice or inside a cold chamber. E. coli cells were lysed either by sonication (HTU Soni130; G. Heinemann Ultraschall und Labortechnik, Schwaebisch Gmuend, Germany) (total time, 5 min; amplitude, 70%; bursts every 3 s for 2 s) or with a cell disruptor system (Constant Systems LTD, Northants, UK) (two loadings, 1.35-kbar pressure), depending on the amount of cells. Cell debris was spun down at 21 000 g for 1 h, and loaded twice onto a self-packed Ni2+–nitrilotriacetic acid column (0.3 mL of matrix per liter of culture). After washing with 20 column volumes of 10 mm and 20 mm imidazole in 50 mm NaH2PO4 (pH 8.0) and 300 mm NaCl buffer, RcFDH or FdsGB were eluted with the same buffer containing 250 mm imidazole. With PD-10 columns (Sephadex G-25 M; Amersham Biosciences, Uppsala, Sweden), the buffer was changed to 75 mm potassium phosphate (pH 7.5) and 10 mm KNO3, and samples were concentrated to 2 mL with ultracentrifugation devices (VIVASPIN 20, 50-kDa cut-off; Sartorius AG, Goettingen, Germany). Size-exclusion chromatography was carried out at 4 °C on a Superose 12 column (GE Healthcare, Uppsala, Sweden) equilibrated in 75 mm potassium phosphate (pH 7.5) and 10 mm KNO3. Fractions containing RcFDH or FdsGB were identified by SDS/PAGE, combined, frozen in liquid nitrogen, and stored at – 80 °C until use. Protein concentrations were determined by measuring the absorption at 280 nm, with ε = 169 500 m−1·cm−1 for RcFDH/RcFDH-W and ε = 48 100 m−1·cm−1 for FdsGB samples.

Quantitation of the FMN cofactor

Five hundred microliters of RcFDH (6 μm) was incubated with 80 μL of trichloroacetic acid (50% w/v) for 10 min on ice. The denatured protein was spun down at 14 000 g for 20 min, and the pellet was resuspended in 100 μL of trichloroacetic acid (5% w/v) and spun down again for 10 min. Both supernatants (550 μL and 100 μL) were combined, and 150 μL of 2 m K2HPO4 was added. The released FMN was detected photometrically on a Shimadzu UV-2401 PC photometer (Shimadzu Europa, Duisburg, Germany) at 444 nm. The protein-bound FMN saturation was calculated from the specific extinction coefficient of free FMN (12 200 m−1·cm−1 at 444 nm) and from a calibration curve derived from different concentrations of free FMN in solution.


The metal content (Mo, W, Fe) was measured with ICP-OES. Samples of 500 μL (10–30 μm) were mixed with the same volume of 65% HNO3, and wet ashed at 100 °C overnight. Samples were diluted with 4 mL of Millipore water, and applied to an Optima 2100 DV instrument (PerkinElmer Life and Analytical Sciences, Waltham, MA, USA). The multielement Standard XVI (Merck, Darmstadt, Germany) (Mo, Fe) and Standard XII (Merck, Darmstadt, Germany) (W) were used for calibration and quantification. The resulting mass concentrations were related to percentage saturation of molybdenum or tungsten according to complete saturation of FeS clusters (five [Fe4S4] clusters; two [Fe2S2] clusters).

Analysis of the bis-MGD cofactor

For detection of the bis-MGD cofactor, the cofactor was converted to its fluorescent degradation product Form A, as described by Johnson et al. [24]. Samples were oxidized in the presence of iodine at pH 2.5 overnight to convert the bis-MGD to the stable fluorescent product Form A-GMP. Analysis of samples was performed with HPLC on a C18 reversed-phase column (Thermo Fisher Scientific, Waltham, MA, USA) with 5 mm ammonium acetate and 15% (v/v) methanol. Fluorescence was monitored with a 1100 series detector (Agilent Technologies, Santa Clara, CA, USA), with excitation at 383 nm and emission at 450 nm.

For detection of 5′-GMP, RcFDH samples (20 μm) were incubated at 95 °C with sulfuric acid for 30 min. Released 5′-GMP was separated on a C18 reversed-phase column (Thermo Fisher Scientific, Waltham, MA, USA) with 50 mm diammonium phosphate buffer (pH 2.5) and 1% (v/v) methanol. Absorption was monitored at 280 nm and 260 nm with a 1100 series DAD detector (Agilent Technologies, Santa Clara, CA, USA).

SDS/PAGE analysis

SDS/PAGE was performed with 15% polyacrylamide gels as described by Laemmli [42]. Protein samples were separated at 15 mA per gel for 80 min. In-gel staining was achieved with a solution of Coomassie Blue R250 in 40% (v/v) methanol and 10% (v/v) acetic acid.

Enzyme assays

RcFDH activity was measured with a UV-2401PC spectrophotometer (Shimadzu Europa, Duisburg, Germany), with the change in NADH absorption being recorded at 340 nm (εNADH = 6220 m−1·cm−1). RcFDH or RcFDH-W (100 nm) was used in a standard assay containing 6 mm sodium formate and 2 mm NAD+ in 100 mm Tris/HCl (pH 9.0) at 30 °C. Reactions were started by the addition of enzyme, and followed for a period of 60 s. For determination of the pH optimum of enzyme activity, overlapping buffer systems with 100 mm citrate/acetate buffer, 100 mm potassium phosphate or 100 mm Tris/HCl in a total range of pH of 4.5–9.5 were used. The temperature optimum was detected by measuring activities in a range of 15 °C to 60 °C. Steady-state kinetics for RcFDH were determined by varying the concentrations of sodium formate and NAD+ (0.05–5 mm each). Steady-state kinetics with ferricyanide (1 mm) or 2,6-dichlorphenolindophenol (0.1 mm) as electron acceptor were determined by varying the sodium formate concentrations (0.05–5 mm), with detection at 420 nm (εferricyanide = 1040 m−1·cm−1) or 600 nm (εdichlorphenolindophenol = 16 100 m−1·cm−1), respectively. Turnover rates of CO2 reduction were calculated by following the oxidation of NADH (0.2 mm) at 340 nm in the presence of 100 mm NaHCO3 in 100 mm potassium phosphate (pH 6.8). For diaphorase activity measurements, 600 nm FdsGB or 200 nm RcFDH was used in an assay containing 1 mm ferricyanide, and a concentration range of 0.05–5 mm NADH was used for determination of steady-state kinetics, with detection at 340 mm for 60 s in 75 mm potassium phosphate (pH 8.0) with monitoring of NADH oxidation. NAD+ reduction assays were performed in a glove box (anaerobic conditions) with 1 mm methylviologen (A600 nm = 1.0 with NDT, εmethylviologen = 12 000 m−1·cm−1) and a concentration range of 0.05–3.5 mm NAD+. Reactions were started by the addition of 40 nm FdsGB or 10 nm RcFDH to the assay, and the oxidation of methylviologen at 600 nm in 100 mm Tris/HCl (pH 9.0) was monitored. Data obtained from three individual measurements were fitted with the Hill function ([y = Vmax × Xn/(Kmn + Xn)], n = 1) and origin software (OriginPro 8.1G SR1; OriginLab Corporation, Northampton, MA, USA). All activities were calculated with respect to one catalytically active protomer, αβγ (m = 172 kDa).

Detection of formate

The product formate from CO2 reduction catalyzed by RcFDH was detected with a colorimetric assay described by Lang and Lang [25]. Fifty microliters of 50 μm RcFDH in 100 mm potassium phosphate buffer (pH 6.8) was incubated in the presence of 2 mm NADH and 100 mm sodium bicarbonate for 30 min at 30 °C. The reaction was stopped by addition of 100 μL of reactant solution [0.5% (w/v) citric acid, and 10% (w/v) acetamide, in 50 mL isopropanol], and samples were spun down at 14 000 g for 10 min. Four microliters of aqueous sodium acetate solution (30%, w/v) and 350 μL of acetic anhydride were added to 150 μL of supernatant, and samples were further incubated at 50 °C for 2 h. The reaction product of formate and the citric acid-derived compound was detected at 515 nm on a plate reader photometer (Vario Scan Flash; Thermo Fisher Scientific, Waltham, MA, USA). Samples either without RcFDH or without NADH were used as negative controls, and a range of 20–0.2 μm sodium formate dissolved in 100 mm potassium phosphate buffer (pH 6.8) was used for standard calibration.

Reduction spectra

UV–visible spectra of the oxidized and reduced RcFDH were recorded in a range of 250–800 nm on a UV-2401 PC photometer (Shimadzu Europa, Duisburg, Germany) in 75 mm potassium phosphate buffer (pH 7.5) containing 10 mm KNO3. Reduction of the catalytically active portion was achieved by the addition of 10 mm sodium formate as the final concentration, and total reduction of the protein sample was achieved by the addition of 10 mm NDT as the final concentration.

Protein–protein interaction studies with SPR

Purified RcFDH, R. capsulatus FdsC, and R. capsulatus FdsD, as well as BSA (control), were immobilized on CMD 500m sensor chips (XanTec Bioanalytics, Duesseldorf, Germany), with response units (RU) between 400 and 1300. The binding experiments were performed on an SPR-based Biacore T200 instrument (GE, Uppsala, Sweden) at 25 °C with a flow rate of 30 μL·min−1. The running buffer comprised 9.9 mm Na2HPO4.2H2O, 1.8 mm KH2PO4, 138 mm NaCl, 2.7 mm KCl, 3.4 mm EDTA, and 0.005% (v/v) Tween-20 (pH 7.4). BSA, Apo-RcFDH, RcFDH, R. capsulatus FdsC and R. capsulatus FdsD at concentrations in the range 10 μm to 0.3 μm were injected with an autosampler rack cooled to 8 °C, and this was followed by 15 min of dissociation. Evaluation of the obtained data was performed with T200 software (GE, Uppsala, Sweden). Constants derived from four independent experiments were used to calculate average KD values.


This work was supported by Deutsche Forschungsgemeinschaft Grant LE1171/6-1 and the Cluster of Excellence ‘Unifying Concepts in Catalysis’.