Occurrence of ferredoxin:NAD+ oxidoreductase activity and its ion specificity in several Gram-positive and Gram-negative bacteria

A ferredoxin:NAD+ oxidoreductase was recently discovered as a redox-driven ion pump in the anaerobic, acetogenic bacterium Acetobacterium woodii. The enzyme is assumed to be encoded by the rnf genes. Since these genes are present in the genomes of many bacteria, we tested for ferredoxin:NAD+ oxidoreductase activity in cytoplasmic membranes from several different Gram-positive and Gram-negative bacteria that have annotated rnf genes. We found this activity in Clostridium tetanomorphum, Clostridium ljungdahlii, Bacteroides fragilis, and Vibrio cholerae but not in Escherichia coli and Rhodobacter capsulatus. As in A. woodii, the activity was Na+-dependent in C. tetanomorphum and B. fragilis but Na+-independent in C. ljungdahlii and V. cholerae. We deleted the rnf genes from B. fragilis and demonstrated that the mutant has greatly reduced ferredoxin:NAD+ oxidoreductase activity. This is the first genetic proof that the rnf genes indeed encode the reduced ferredoxin:NAD+ oxidoreductase activity.


INTRODUCTION
Like any other cell, microorganisms use two energy currencies to couple exergonic catabolic reactions to endergonic anabolic reactions, ATP and a transmembrane electrochemical ion gradient. Both are connected by the ATP synthase, a key element in cellular bioenergetics present in all domains of life. Under "respiratory" conditions, the ATP synthase is driven by the electrochemical ion gradient across the membrane to synthesize ATP (Boyer, Cross & Momsen, 1973). The coupling ion is a proton in most cases but some bacteria and archaea use Na + instead of H + as the coupling ion (Grüber et al., 2014;Müller et al., 2005). The reaction is freely reversible and under of the bacterium to fix nitrogen (Schmehl et al., 1993). The similarity of the Rnf proteins to subunits of the membrane-integral, ion translocating NADH:quinone-oxidoreductase (Biegel et al., 2011;Kumagai et al., 1997) led the authors to speculate that the Rnf proteins may constitute a membrane-integral protein complex involved in nitrogen fixation (Schmehl et al., 1993). The nitrogenase reaction requires low potential electrons that can only be generated from NADH at the expense of energy and it was speculated that the Rnf complex catalyzes this endergonic electron transfer with the energy derived from the transmembrane electrochemical ion gradient. Thus, in R. capsulatus, energy conservation is by photosynthetic electron transport phosphorylation and the Rnf complex is thought to be involved in anabolic reactions by providing low potential electrons for biosynthetic reactions (Schmehl et al., 1993).
rnf genes are widely distributed in bacteria and are also present in a few methanogenic archaea (Biegel et al., 2011). Rnf complexes thus far have not been purified from any source and thus, the final biochemical proof that they are indeed ion-translocating membrane proteins is still missing. Moreover, the anticipated reaction catalyzed by the complex, electron transfer from reduced ferredoxin to the acceptor NAD + has only been reported for A. woodii and C. tetanomorphum (Biegel et al., 2011;Boiangiu et al., 2005). Despite the obvious lack of experimental data with a purified enzyme, as well as the experimental proof that Rnf in the methanogen Methanosarcina acetivorans uses not NAD + but an alternative electron acceptor (Schlegel et al., 2012), the presence of rnf genes in bacteria is often taken as indication that these organisms have an ion-translocating ferredoxin:NAD + oxidoreductase. To shed more light on the function of the Rnf complexes in different bacteria, we have isolated the cytoplasmic membranes of several organisms and assayed them for ferredoxin:NAD + oxidoreductase activity. In addition, we determined the ion specificity of the reaction in the different organisms and will present genetic evidence that the ferredoxin:NAD + oxidoreductase activity in B. fragilis is indeed catalyzed by the Rnf complex.

Preparation of washed membranes
Membranes were prepared as described (Imkamp et al., 2007) under strictly anaerobic conditions with some modifications. Cells were grown to the end of the exponential growth phase, harvested by centrifugation (11,300 Â g at 4 C; Contifuge Strato; Heraeus, Osterode, Germany) and washed twice with buffer A (50 mM Tris-HCl (pH 7.0), 20 mM MgSO 4 , 20% glycerol, 2 mM DTE (dithioerythritol) and 4 mM resazurin). The cells were resuspended in 15 ml buffer A and disrupted by a single passage through a French press (16,000 psi) or for B. fragilis by sonication under anaerobic conditions. Cell debris and whole cells were removed by one centrifugation step (23,700 Â g, 20 min, 4 C). The cell free extract was separated into cytoplasmic and membrane fraction by ultracentrifugation (190,000 Â g, 1 h, 4 C). The resulting sediment was washed in buffer A and membranes were again sedimented by an ultracentrifugation step (190,000 Â g, 1 h, 4 C). Membranes were resuspended in 5 ml buffer A and used immediately for the experiments. Protein concentrations were determined as described previously (Bradford, 1976).

Preparation of inverted membrane vesicles of C. ljungdahlii
Cells were grown in a 20-l-scale as described above in MES-buffered complex medium containing 56 mM fructose. 20 g (wet weight) cells were resuspended in ∼150 ml buffer A (50 mM Tris, 20 mM MgSO 4 , 20% glycerol, 2 mM DTE, 4 mM resazurin, pH 7.5) and digested with 400 mg lysozyme at 37 C for 20 minutes. Afterwards, the cells were passed once through a French Press at 41 MPa. A low speed centrifugation step (4500Â g, 45 minutes, 4 C) was succeeded by ultracentrifugation at 120,000 Â g and 4 C for another 45 minutes. The vesicles were washed once in buffer A and vesicles were sedimented again via ultracentrifugation as described above. The resulting pellet was resuspended in the same buffer in a volume of 5 ml.

Measurement of Fd red :NAD + oxidoreductase activity
Measurement of electron transfer from reduced ferredoxin to NAD + by either membranes or inverted membrane vesicles was performed as described (Hess, Schuchmann & Müller, 2013) in anaerobic cuvettes filled with 1 ml 20 mM Tris-HCl buffer (pH 7.7) containing 2 mM DTE and 4 mM resazurin at a pressure of 0.5 Â 10 5 Pa CO. Ferredoxin (30 mM; purified from C. pasteurianum as described (Schönheit, Wäscher & Thauer, 1978)), Acs/CODH (30 mg/ml; purified from A. woodii as described (Hess, Schuchmann & Müller, 2013)), and washed membranes or inverted membrane vesicles (150 mg/ml) were added. If indicated, the ionophores ETH2120 or TCS were added at a concentration of 10 mM. The reaction was started by addition of NAD + (4 mM). Formation of NADH was measured at 340 nm. Enzyme activity was measured at the same temperature at which the corresponding cells were grown. Na + dependence of the electron transfer activity was measured as described (Hess, Schuchmann & Müller, 2013).

Generation of the deletion mutant B. fragilis Árnf
We used the two-step procedure as previously described (Brigham & Malamy, 2005) for the isolation of chromosomal deletions in broad specificity hexokinase genes of B. fragilis. Briefly, a 480 bp PCR fragment beginning 300 bp upstream of rnf B and ending 180 bp into rnf B, was ligated to a 590 bp PCR fragment containing 260 bp of the 3′ end rnfA and 330 bp of adjacent sequence, then ligated into the cloning vector pYT102 to create plasmid pRAG394 in E. coli. A tri-parental mating was performed to transfer the pRAG394 construct into the chromosome of B. fragilis ADB77. The resulting co-integrates were verified by PCR analysis and candidates were subjected to the resolution protocol as described (Brigham & Malamy, 2005). Resolvants that contained the rnf deletion were identified by PCR analysis.

Generation of the deletion mutant E. coli Árnf
The deletion of the rnf operon from the genome of E. coli was carried out using the Red-recombinase method as reported (Datsenko & Wanner, 2000). The parent strain was E. coli TOP-10 (Invitrogen). The rnf operon was replaced with a chloramphenicol cassette, which remained in the bacterial chromosome; thus the deletion strain is resistant to chloramphenicol up to 30 mg/ml. The deletion of the rnf operon was confirmed by PCR and by verifying the loss of fluorescent bands corresponding to RnfD and RnfG in SDS-PAGE membrane preparations.

RESULTS
For the survey of ferredoxin:NAD + oxidoreductase activity we used membranes of the strict anaerobes A. woodii, C. tetanomorphum, C. ljungdahlii, the nanoaerophile B. fragilis, and the facultative aerobes R. capsulatus, V. cholerae and E. coli. The genetic organization of the rnf genes in these species is shown in Fig. 1.

A. woodii
As mentioned above, the acetogenic bacterium uses the Rnf complex as a membrane potential generator during growth on H 2 + CO 2 or fructose; and in the reverse reaction to drive the endergonic reduction of ferredoxin with electrons derived from NADH during heterotrophic growth on, for example, 2,3-butanediol, ethanol or lactate. As can be seen from Table 1, NAD + reduction occurred at a rate of 50 ± 5 mU/mg protein and 53 ± 3 mU/mg protein, respectively, in membranes from cells grown on fructose or 2,3-butanediol, but the activity was highest in cells grown on ethanol 169 ± 8 mU/mg) or lactate (398 ± 21 mU/mg).
C. ljungdahlii C. ljungdahlii can grow autotrophically on H 2 + CO 2 or CO and produces acetate and ethanol. Inhibitor studies that had been done with whole cells were consistent with the presence of an enzyme that generates a primary electrochemical proton potential across the cytoplasmic membrane (Tremblay et al., 2012). Deletion of the rnf genes had resulted in a loss of the proton motive force which is consistent with the hypothesis of the presence of a proton-translocating Rnf complex in C. ljungdahlii. Indeed, membranes of C. ljungdahlii catalyzed electron transfer from reduced ferredoxin to NAD + with an activity of 256 ± 7 mU/mg (Table 1). This activity was not dependent on Na + (measured in a range of 68 mM to 10 mM NaCl) ( Fig. 2A). Electron transfer from reduced ferredoxin to NAD + in inverted membrane vesicles was stimulated by the protonophore TCS but not by the sodium ionophore ETH2120 (Fig. 2B). These data are consistent with the hypothesis that the Rnf complex of C. ljungdahlii uses H + as coupling ion, not Na + . F 1 F O (Rahlfs & Müller, 1997;Müller, 2003) and A 1 A O (Müller, Rupert & Lemker, 1999) ATP synthases have a characteristic Na + binding motif in their c subunits (Q/E in helix one, E/D and S/T in helix two). The absence of a Na + -binding motif in subunit c of the ATP synthase in C. ljungdahlii is in line with the hypothesis of a H + -Rnf.   Fno activity was measured in anoxic cuvettes filled with 1 ml 20 mM Tris (sodium free)-HCl buffer (pH 7.7) containing 2 mM DTE and 2 mM resazurin at a pressure of 0.5 Â 10 5 Pa CO. NaCl was added to the concentration indicated. Ferredoxin (30 mM), Acs/CODH (30 mg/ml), and washed membranes or inverted membrane vesicles (150 mg/ml) were added. If indicated, the ionophores ETH2120 or TCS were added at a concentration of 10 mM. The reaction was started by addition of NAD + (4 mM). Formation of NADH was measured at 340 nm.

C. tetanomorphum
This obligate anaerobic, glutamate fermenting bacterium was quoted in review articles to have an Rnf complex that could be partially purified (Boiangiu et al., 2005;Buckel & Thauer, 2013). The ion specificity was not discussed but speculated to be Na + . However, these data are not published. Membranes of C. tetanomorphum grown on glutamate catalyzed electron flow from reduced ferredoxin to NAD + with a rate of 900 ± 29 mU/mg (Table 1), 18-fold higher than the rate observed in fructose-grown cells of A. woodii. Without additional Na + the activity was only 431 ± 16 mU/mg, which is consistent with the hypothesis that the enzyme uses Na + as coupling ion.

E. coli
The rnf genes in E. coli are termed rsx (Koo et al., 2003) and have been shown to be upregulated during aerobic growth (Partridge et al., 2006). However, membranes from aerobically grown cells prepared under anaerobic conditions had no ferredoxin: NAD + -oxidoreductase activity. Furthermore, a deletion of the rnf operon from E. coli in the strain TOP10 showed no growth phenotype, neither under aerobic nor under anaerobic conditions.

R. capsulatus
As mentioned above, the rnf genes were first described in R. capsulatus and hypothesized to be involved in nitrogen fixation by providing low potential electron for nitrogenase (Schmehl et al., 1993). Membranes of cells grown diazotrophically had no ferredoxin: NAD + -oxidoreductase, neither if the cells were grown in the light nor under respiratory conditions with glucose or DMSO.

B. fragilis
The nanoaerophilic bacterium B. fragilis not only has rnf genes but also nqr and nuo (complex I) genes. Cytoplasmic membranes prepared from cells grown on complex medium catalyzed electron transfer from reduced ferredoxin to NAD + with a rate of 37 ± 5 U/mg (Table 1). In the absence of NaCl, the activity was only 5.3 ± 0.7 mU/mg (Table 1), arguing for a Na + -dependence of the reaction. Since a genetic system is available for this organism, we deleted the rnf genes to establish the gene-polypetide correlation of the observed activity. Indeed, the ferredoxin:NAD + oxidoreductase activity was reduced to 11% in the Árnf mutant in both the presence and absence of NaCl (Table 1), which is evidence that the majority of the ferredoxin:NAD + oxidoreductase measured is encoded by the rnf genes.

V. cholerae
The presence of rnf genes in V. cholerae was long known, however, a function of Rnf in this bacterium has not yet been described. Cytoplasmic membranes of V. cholerae catalyzed electron transfer from reduced ferredoxin to NAD + with a rate of 7.1 ± 0.7 mU/mg ( Table 1). As in C. ljungdahlii, this activity was independent of the Na + concentration, arguing for a H + -dependent enzyme.

DISCUSSION
The assay described before to detect ferredoxin:NAD + oxidoreductase activity in A. woodii (Hess, Schuchmann & Müller, 2013) was apparently suitable to detect the same activity in C. ljungdahlii, C. tetanomorphum, B. fragilis, and V. cholerae. This is actually all the more surprising since the assay requires that the ferredoxin:NAD + oxidoreductase accepts electrons from C. pasteurianum ferredoxin. Not only must the redox potential be suitable but also the protein-protein interaction of the ferredoxin and its cognate receptor, most likely RnfB (Suharti et al., 2014) has to allow electron flow from the donor protein to the acceptor protein. The lack of activity in membranes of E. coli or R. capsulatus may be due to the differences in ferredoxin or simply reflect the fact that ferredoxin is not the electron carrier used by these enzyme complexes. Bacteria such as A. woodii encode several different ferredoxins but the one used by the ferredoxin:NAD + oxidoreductase is not known. Therefore, phylogenetic analysis cannot be done. Thus far, the electron acceptor used by the Rnf complexes of E. coli and R. capsulatus remain to be identified. Moreover, the Árnf strain of E. coli has no phenotype and thus, the physiological role in E. coli also remains to be elucidated. Koo et al. (2003) has suggested that Rnf is the electron donor to the Fe-S protein SoxR that is part of a ROS sensor in the cell (Hidalgo, Ding & Demple, 1997). The role of Rnf would be to maintain SoxR in a reduced, and thus inactive state. In the presence of oxygen radicals, oxidized SoxR signals the cell to activate a response to ROS thus in the rnf deletion mutant, SoxR is expected to always be in the oxidized state and the cellular systems that protect the cell against ROS would always be activated. This would make it difficult to define phenotypic differences between the rnf deletion mutant and wild type E. coli. In our hands the growth curve for the deletion mutant was essentially the same as for wild type in both rich and minimal media, and inclusion of ROS producing substances such as paraquat and hydrogen peroxide did not reveal any clear differences. Preliminary expression data for V. cholerae suggest that Rnf is expressed at the same levels in several growth and infection conditions, suggesting that changes in the environment have little effect on the expression of rnf. However, Rnf seems to be essential for V. cholerae since we were not able to generate a Árnf deletion mutant. Of interest is the finding that the rnf deletion in B. fragilis did not alter the anaerobic growth properties of this strain on BHIS medium (which contains glucose) when compared to the wild type parental strain. There are other annotated oxidoreductases in the B. fragilis genome coded for by the nuo and nqr operons whose functions may be redundant with rnf; this might also explain why 11% of the ferredoxin:NAD + oxidoreductase activity could still be measured in the rnf deletion strain.
The data presented here show that the ferredoxin:NAD + oxidoreductase activity in A. woodii is regulated, since growth on different substrates resulted in different activities. There was no correlation between activity and direction of electron flow in context of the metabolism. However, there was a correlation between activity and the number of cycles the complex has to pass for the phosphorylation of 1 ATP in the metabolism: When converting lactate, the Rnf complex has to catalyze 2.5 cycles for every ATP that is gained in this pathway (Weghoff, Bertsch & Müller, 2015). When converting ethanol, the phosphorylation of 1 ADP requires the Rnf complex to catalyze 1.3 cycles . The conversion of 2,3-butanediol only needs 0.85 cycles for every ATP  and degradation of fructose is predicted to require 0.11 Rnf cycles per ATP. The correlation of these values is in good agreement with the specific Rnf activities we observed for each of these substrates (Table 1). Thus, the more Rnf activity is necessary for the metabolism, the higher is the experimentally determined activity.
The highest ferredoxin:NAD + oxidoreductase activity was found in C. tetanomorphum. Again, this would argue for a high electron transfer rate through Rnf during glutamate fermentation. The activity was stimulated by Na + indicating that the Rnf complex uses Na + as coupling ion. This is also consistent with the hypothesis of a Na + /glutamate symporter in C. tetanomorphum (Buckel & Thauer, 2013) as well as a V-type ATP synthase (encoded in the genome) with the c subunit having a Na + binding motif in the second hairpin. In contrast, the activity was not Na + -dependent in the acetogen C. ljungdahlii. The observed stimulation of NAD + reduction in inverted membrane vesicles by a protonophore clearly resembles the phenomenon of respiratory control, showing a strict coupling of electron transfer to the translocation of an ion across the membrane. The electrochemical potential established by the ion and charge translocation sets up a thermodynamic backup pressure and reduces the rate of electron flow (as well as the coupled ion flow). Dissipation of the ion gradient will release the thermodynamic backup pressure and stimulate electron flow. Since this was observed only with a protonophore but not a sodium ionophore it can be concluded that the coupling ion is a proton, not a sodium ion. As discussed before (Biegel et al., 2011), there is no reason to believe that the Rnf complex can only translocate Na + as it does in A. woodii. Actually, other membrane transporters are known that can translocate either Na + or H + , depending on the species. In the case of the F 1 F O ATP synthase, only the exchange of two out of roughly 4,000 residues will change the ion specificity of the enzyme from H + to Na + (Meier et al., 2009).
The final proof that the Rnf complex is a redox-driven ion (Na + /H + pump) still has to await purification of the complex and its reconstitution into liposomes. An equally important proof is presented here for the first time in bacteria: deletion of the rnf genes of B. fragilis reduced the ferredoxin:NAD + oxidoreductase activity by 89%. Although the physiological role of the Rnf complex and the origin of the remaining 11% activity remains to be identified, this is clear evidence that the rnf genes encode the ferredoxin:NAD + oxidoreductase activity and in line with the observation that a Árnf mutant of the archaeon Methanosarcina acetivorans is impaired in Na + transport coupled to electron flow from reduced ferredoxin to heterodisulfide (Schlegel et al., 2012).