Reduction of Flavodoxin by Electron Bifurcation and Sodium Ion-dependent Reoxidation by NAD+ Catalyzed by Ferredoxin-NAD+ Reductase (Rnf)*

Electron-transferring flavoprotein (Etf) and butyryl-CoA dehydrogenase (Bcd) from Acidaminococcus fermentans catalyze the endergonic reduction of ferredoxin by NADH, which is also driven by the concomitant reduction of crotonyl-CoA by NADH, a process called electron bifurcation. Here we show that recombinant flavodoxin from A. fermentans produced in Escherichia coli can replace ferredoxin with almost equal efficiency. After complete reduction of the yellow quinone to the blue semiquinone, a second 1.4 times faster electron transfer affords the colorless hydroquinone. Mediated by a hydrogenase, protons reoxidize the fully reduced flavodoxin or ferredoxin to the semi-reduced species. In this hydrogen-generating system, both electron carriers act catalytically with apparent Km = 0.26 μm ferredoxin or 0.42 μm flavodoxin. Membrane preparations of A. fermentans contain a highly active ferredoxin/flavodoxin-NAD+ reductase (Rnf) that catalyzes the irreversible reduction of flavodoxin by NADH to the blue semiquinone. Using flavodoxin hydroquinone or reduced ferredoxin obtained by electron bifurcation, Rnf can be measured in the forward direction, whereby one NADH is recycled, resulting in the simple equation: crotonyl-CoA + NADH + H+ = butyryl-CoA + NAD+ with Km = 1.4 μm ferredoxin or 2.0 μm flavodoxin. This reaction requires Na+ (Km = 0.12 mm) or Li+ (Km = 0.25 mm) for activity, indicating that Rnf acts as a Na+ pump. The redox potential of the quinone/semiquinone couple of flavodoxin (Fld) is much higher than that of the semiquinone/hydroquinone couple. With free riboflavin, the opposite is the case. Based on this behavior, we refine our previous mechanism of electron bifurcation.

at 45 M Fe 2ϩ , the flavodoxin content decreases to 0.2 mol g Ϫ1 and that of ferredoxin increases to 1.0 Ϯ 0.1 mol g Ϫ1 of protein (0.4 mM) (14). Flavodoxin from A. fermentans is a protein (molecular mass ϭ 14.5 kDa) with FMN as tightly bound cofactor. Ferredoxin from the same organism is a smaller protein (molecular mass ϭ 5.6 kDa) that contains two [4Fe-4S] clusters. The redox potentials of both electron carriers from A. fermentans have been measured. In the case of ferredoxin, the first electron encounters a potential of E 1°Ј ϭ Ϫ340 mV (Fd/Fd Ϫ ), and the second electron encounters a potential of E 2°Ј ϭ Ϫ405 mV (Fd Ϫ /Fd 2Ϫ ) (14); with flavodoxin, the corresponding potentials are E 1°Ј ϭ Ϫ60 mV and E 2°Ј ϭ Ϫ430 mV (15). Hence, from a thermodynamic point of view, flavodoxin can easily replace ferredoxin, if only the half/fully reduced carriers with the low redox potentials (E 2°Ј ) are involved. In a previous study, evidence was obtained that ferredoxin cycles between half/fully reduced states (Fd Ϫ /Fd 2Ϫ ) during electron transfer from Etf to hydrogenase and back (10). This study will show that flavodoxin behaves in a similar manner; it cycles between the neutral blue semiquinone (FldH ⅐ ) and the colorless hydroquinone (FldH Ϫ ) states, which are much easier to detect than the brownish/colorless Fd/Fd Ϫ /Fd 2Ϫ states. Furthermore, the generation of two-electron reduced ferredoxin and flavodoxin by electron bifurcation without the use of artificial electron donors allowed the confirmation of the postulated Na ϩ dependence of the Rnf from A. fermentans.

Materials and Methods
Growth of Microorganisms-A. fermentans strain VR4 (DSM 20731) was grown anaerobically on the glutamate/yeast extract/biotin medium as described earlier (16). Clostridium tetanomorphum DSM 526 was grown on the same glutamate medium but without added biotin. Clostridium pasteurianum DSM 525 was grown on a medium containing 100 mM glucose, 70 mM NaHCO 3 , and yeast extract (2 g/liter). E. coli strains DH5␣ and BL21-DE3 were grown aerobically at 37°C in Standard I nutrient broth (Merck, Darmstadt, Germany) using an orbital shaker. For growth of E. coli with inserted plasmids, the medium was supplemented with carbenicillin (50 g/ml). For overproduction of flavodoxin, 2 liters of Standard I broth, supplemented with carbenicillin, was inoculated with 25 ml of an overnight culture of BL21-DE3 that has been transformed with the pASG-IBA33 expression vector (IBA, Göttingen, Germany) containing the flavodoxin gene of A. fermentans. The culture was grown until it reached OD 600 ϭ 0.6, and then anhydrotetracycline (200 g/liter, IBA) was added to induce the expression of the recombinant gene during the further growth for the next 12 h at room temperature. The harvested cell paste was stored at Ϫ80°C until use.
Synthesis of CoA Esters-Crotonyl-CoA and butyryl-CoA were synthesized by acylation of CoASH in aqueous 1 M KHCO 3 using 1 M crotonic anhydride or butyric anhydride in acetonitrile with a slight molar excess. After acidification, the CoA thioesters were purified over C18 columns and stored as lyophilized powders at Ϫ80°C (17). The concentration of crotonyl-CoA was calibrated by the NAD ϩ -dependent ␤-oxidation to acetyl-CoA and acetyl phosphate as described for the assay of glutaconyl-CoA decarboxylase (16). HPLC of crotonyl-CoA and butyryl-CoA was performed on a C18 Kinetex column (5-m particle size, 100-Å pore size, 250 ϫ 4.6 mm, Phenomenex, Aschaffenburg, Germany) at a flow rate of 1 ml/min in 50 mM KH 2 PO 4 , pH 5.3, and 5% acetonitrile. During 20 min, a linear gradient up to 60% acetonitrile was applied. For detection of respective CoA esters from the assay mixtures, samples were acidified with concentrated HCl to pH 2.0, centrifuged, and filtered to remove the denatured protein. Samples of 10 l were loaded on the HPLC column.
Preparation of Ti(III) Citrate-Ti(III) citrate was prepared as described (18). TiCl 3 in HCl (12%, 0.5 ml) was added to 5 ml of 0.2 M sodium citrate. After 10 min, the mixture was neutralized with a saturated sodium carbonate solution to pH 7.0. The prepared Ti(III) citrate was stored under anaerobic conditions in air-tight bottles.
Recombinant DNA Methods-Genomic DNA was isolated from A. fermentans cells using a genomic DNA preparation kit (Fermentas, Thermo Fisher) according to the manufacturer's instructions. The flavodoxin protein from A. fermentans was identified earlier by N-terminal sequencing (15). The gene corresponding to the protein, Acfer_0269 (19), was amplified by PCR using genomic DNA as the template with the forward primer 5Ј-AAGCTCTTCAATGAGCAAAATCGCAGTGGT-GTTCT and the reverse primer 5Ј-AAGCTCTTCACCCTGC-CAGCGCCTCTCCCAG. The PCR product was cloned into the p-ENTRY vector and introduced in E. coli DH5␣ by chemical transformation. The cloned gene sequence was confirmed and then transferred to the expression vector pASG IBA 33 and finally transferred to E. coli BL21-DE3. The colonies harboring the plasmid were used for overproduction of the protein.
Purification of Flavodoxin-E. coli cells overproducing the recombinant flavodoxin (10 g of wet mass) were suspended in 50 mM potassium phosphate/150 mM NaCl, pH 7.0 (buffer A) and disrupted by three passages through a French press at 140 MPa. Cell debris and membranes were separated by centrifugation at 150,000 ϫ g for 45 min at 4°C to obtain a membranefree extract. The extract was passed through a 45-m filter and loaded on a 5-ml Ni-Sepharose column (HiTrap, GE Healthcare, München, Germany), which was pre-equilibrated with buffer A. All purification steps were performed at 4°C, unless otherwise mentioned. After loading the membrane-free extract, the column was washed with 10 column volumes of buffer A to remove contaminating proteins. The protein was eluted with 250 mM imidazole in buffer A. The purified protein was desalted and concentrated; FMN in molar excess was added and incubated overnight in the dark at 4°C. Chromatography on a Superdex 75 pg (GE Healthcare) yielded homogenous protein and removed excess FMN. The protein was stored at Ϫ80°C until use.
Purification of Ferredoxin-The purification was performed under strict anoxic conditions under an atmosphere of 95% N 2 and 5% H 2 (Coy Anaerobic Chamber) (14). Frozen C. tetanomorphum cells (10 g) were suspended in 50 mM potassium phosphate, pH 6.8 (buffer B), and passed three times through a French press at 140 MPa. The supernatant obtained by centrifugation at 150,000 ϫ g for 1 h at 4°C was loaded on a DEAE column, which was pre-equilibrated by buffer B. The column was washed with 2 column volumes of buffer B, and the protein was eluted with a gradient of 0 -100% 1 M NaCl achieved over 10 column volumes (600 ml). Active fractions of ferredoxin, identified by the bifurcation assay (see below), were pooled and concentrated by ultrafiltration with a 3-kDa membrane (Centricon, Merck Millipore, Darmstadt, Germany). The concentrated protein sample was chromatographed on a Superdex 75 column pre-equilibrated with 150 mM NaCl in buffer B. The fractions containing ferredoxin were identified by their dark brown color, concentrated, and stored at Ϫ80°C under anaerobic conditions. The ferredoxin concentration was measured by its iron content (14), and a molecular mass of 6 kDa was used. The concentrations of flavodoxin and ferredoxin in A. fermentans cells listed as mol/g of protein (14) were converted to mM by assuming a cell volume of 2.5 ml/g of protein (20).
Preparation of Hydrogenase-Frozen C. pasteurianum cells (5 g) were suspended in 10 ml of buffer B and broken by three passages through a French press at 140 MPa (20,000 p.s.i.) under strict anoxic conditions. The broken cells were centrifuged at 7,000 ϫ g for 20 min. The supernatant was heated to 55-60°C for 10 -15 min under a hydrogen atmosphere. Afterward the extract was cooled in ice for 30 min. The precipitated proteins were removed by centrifugation at 20,000 ϫ g for 30 min. The supernatant containing hydrogenase was stored at Ϫ80°C until use (21).
Electron-transferring Flavoprotein and Butyryl-CoA Dehydrogenase-Recombinant Etf from A. fermentans with a C-terminal His tag at the ␣-subunit produced in E. coli and native Bcd from A. fermentans were prepared as reported earlier (10). Both proteins were incubated with excess FAD overnight in dark at 4°C. Excess FAD was removed by using a desalting PD 10 column (GE Healthcare). The FAD content of the proteins was calculated using ⑀ 450 ϭ 11.3 mM Ϫ1 cm Ϫ1 .
Preparation of Membrane Extracts-Wet-packed cells of A. fermentans (20 g) were suspended in buffer B and broken by four passages through a French press at 140 MPa under strict anoxic conditions. Cell debris was removed by centrifugation 20,000 ϫ g for 20 min at 4°C. The crude extract was centrifuged at 150,000 ϫ g for 60 min. The supernatant was stored for further purification of soluble proteins, and the membrane pellet was homogenized and washed twice with buffer B by centrifuging at 150,000 ϫ g for 30 min. The obtained inverted vesicles were used for Na ϩ /Li ϩ dependence assays. For other kinetic measurements, the vesicles were solubilized with 2% n-dodecyl-␤-D-maltoside in buffer B supplemented with 0.5 M NaCl and then well homogenized. The homogenate was kept on ice for 30 min and centrifuged at 150,000 ϫ g for 30 min. The supernatant was used for the experiments (16).
Analytical Methods-Protein concentrations were estimated with the Bradford assay (22) (Bio-Rad-Microassay reagent, Bio-Rad Laboratories, Munich, Germany). Bovine serum albumin (Sigma, Darmstadt, Germany) was used as standard. SDS-PAGE was performed as described (23). All enzyme assays were performed under an atmosphere of 95% N 2 and 5% H 2 . The bifurcation assay, i.e. the measurement of the reduction of crotonyl-CoA to butyryl-CoA, was done in a quartz cuvette (total volume 500 l, d ϭ 1 cm) using, unless otherwise indicated, 0.5 M Etf Af , 1 M Bcd Af , 1 M ferredoxin, 30 g of crude hydrogenase, and 250 M NADH in buffer B. The reaction was started with 100 M crotonyl-CoA. The decrease in NADH concentration was monitored at 340 nm, ⑀ ϭ 6.3 mM Ϫ1 cm Ϫ1 (24). Stoichiometric amounts of flavodoxin (50 M) were used in place of ferredoxin/hydrogenase and monitored at 340 nm and also at 450 nm (oxidized flavin) and 578 (blue flavin semiquinone). Usually all measurements were repeated three times and the error bars (average with S.D.) were inserted into the figures when appropriate.
Determination of Na and Other Elements-After running the bifurcation/Rnf reaction, the whole assay was analyzed by inductively coupled plasma mass spectrometry (ICP-MS). The protein samples and buffer blanks were diluted 100-fold with ultrapure water and spiked with 10 g yttrium ( 89 Y)/kg as internal standard. The calibration of the ICP-MS was performed in the concentration range 0.1-100 g/kg using dilutions of a Merck ICP multi-element standard solution IV (Merck product number 111355). The elemental contents of Li, Na, Mg, K, Mn, Fe, Co, Ni, Cu, and Zn in the protein sample were determined by ICP-MS using an Agilent 7900 ICP-MS with HEN nebulizer and cooled Scott spray chamber under standard operating conditions. The instrument was auto-tuned to robust plasma conditions to avoid matrix effects. The isotopes 7 Li, 23 Na, 24 Mg, 26 Mg, 39 K, 55 Mn, 56 Fe, 57 Fe, 59 Co, 60 Ni, 62 Ni, 63 Cu, 65 Cu, 64 Zn, 66 Zn and 67 Zn were measured under NoGas, He collision, and H 2 reaction mode conditions. Some isotopes were strongly interfered from the K-containing matrix (mainly 55 Mn) in the NoGas mode and therefore rejected. The assay used to determine the dependence of Na contained 43 Ϯ 1 M Na, which was mainly introduced by the 50 mM potassium phosphate buffer, pH 6.8. In addition, about 8 M Mg, 2.5 M Fe, 0.8 M Ni, and 2 M Zn were detected.

Characterization of Recombinant Flavodoxin from A.
fermentans-The flavodoxin with a C-terminal His tag was affinity-purified from a cell lysate of E. coli BL21 (DE3) harboring the pASG IBA33 vector with the flavodoxin gene as described under "Materials and Methods." To assess its homogeneity, the protein was subjected to gel filtration at pH 7.0 whereby several peaks were observed. Each of the two most prominent peaks, when analyzed by SDS-PAGE, revealed a single band around the 15-kDa protein marker. Both proteins were identified to be flavodoxin by peptide mapping with MALDI-TOF mass spectrometry, showing that the flavodoxin tends to make a mixture of oligomeric and monomeric forms. The molecular mass of the monomeric form (15.37 kDa) fits to the calculated value based on the amino acid sequence from the protein (14.54 kDa) and the His tag (0.84 kDa). The UV-visible spectrum of the purified flavodoxin ( Fig. 1) exhibited the two characteristic peaks at 375 and 448 nm with a shoulder at 464 nm. The flavodoxin was heated at 80°C for 5 min, and the yellow compound in the supernatant of the precipitated protein was identified as FMN by TLC and reverse phase HPLC. The FMN content of the protein was quantified by using the known absorbance coefficients at ϭ 450 nm (⑀ ϭ 11.3 mM Ϫ1 cm Ϫ1 ) and ϭ 375 nm (⑀ ϭ 10.7 mM Ϫ1 cm Ϫ1 ) (25). Flavodoxin as isolated contained only 0.2 mol of FMN per monomer, which was raised to 0.5-0.6 upon overnight incubation with a molar excess of FMN in the dark at 4°C followed by gel filtration to remove the unbound FMN. In contrast, the FMN content of the flavodoxin purified from A. fermentans was reported to be 1.0 mol/mol of protein (15), which suggests that upon overproduction in E. coli, about half of the recombinant flavodoxin folded into a structure unable to bind FMN, which could be due to the presence of the His tag. The production of flavodoxin from Anabaena without a tag in E. coli yielded pure holo-flavodoxin with a ratio of A 465 /A 274 ϭ 0.17 (26), whereas the ratio of A 465 /A 274 obtained with the recombinant protein in this study was 0.146. The addition of 0.1 mM Ti(III)citrate to 35 M flavodoxin (22 M FMN) at pH 6.8 reduced the absorbance at 375 and 446 nm, and two new stable peaks at 500 and 578 nm appeared (Fig. 1), which are characteristic for the blue neutral semiquinone (FldH ⅐ ) (27). Upon increasing the concentration of Ti(III)citrate to 5 mM, both peaks disappeared and the spectrum changed to that of completely reduced flavodoxin (FldH Ϫ ).
Etf Exhibits Low Diaphorase Activity-Upon incubation of flavodoxin with NADH, the absorbance at 450 nm remained unchanged. In the presence of Etf, however, flavodoxin was slowly reduced by NADH to the blue semiquinone. As shown in  FMN) were mixed under strict anaerobic conditions, and the UV-visible spectrum of the oxidized Fld was obtained. Upon the addition of 30 M crotonyl-CoA, the peak at 450 nm decreased and that of the blue semiquinone (FldH ⅐ ) at 578 nm, which remained stable for more than 3 min, became visible. Upon further addition of 50 M crotonyl-CoA, the semiquinone was reduced to the hydroquinone (FldH Ϫ ). The complete reduction of flavodoxin indicated that an effective redox potential of E 2 Ј Յ Ϫ500 mV was reached. The same behavior was observed in the bifurcation process with ferredoxin (6,29). To get more insight into the kinetics of this reduction, we simultaneously measured the absorbance changes with time at three different wavelengths, NADH at 340 nm, quinone at 450 nm, and semiquinone at 578 nm (Fig. 3). After starting the incubation of Etf, Bcd, flavodoxin, and NADH with crotonyl-CoA, we observed NADH oxidation, quinone reduction, and semiquinone formation. Only after the quinone was quantitatively reduced to the semiquinone did an ϳ1.4 times faster reduction to the hydroquinone take place. At this point, the apparent rate of NADH consumption increased accordingly. This result showed that the thermodynamically more facile reduction of the quinone to the semiquinone (E 1°Ј ϭ Ϫ60 mV) occurred first and was followed by the more difficult reduction to the hydroquinone (E 2°Ј ϭ Ϫ430 mV) (15). In contrast, the rates of the two reductions did not follow the thermodynamics, indicating that at the Etf, the semiquinone is a better electron acceptor than the quinone.
Hydrogenase Catalyzes the Reoxidation of Flavodoxin Hydroquinone-In our previous bifurcation assays, the hydrogenase from C. pasteurianum recycled oxidized ferredoxin concomitant with hydrogen formation (6,29). With 100 M NADH, 50 M crotonyl-CoA, 0.5 M Etf, and 1.0 M Bcd (optimal Etf/Bcd ratio) from A. fermentans, specific activities up to V max ϭ 3.2 units (mg of Etf) Ϫ1 were achieved. The apparent K m  for ferredoxin was determined as 0.26 Ϯ 0.03 M (10). This system worked equally well with Fld, exhibiting a similar V max ϭ 1.2 units (mg of Etf) Ϫ1 and an apparent K m ϭ 0.42 Ϯ 0.05 M, based on a content of 0.5 FMN/Fld (Fig. 4). Initially, Fld is reduced to the hydroquinone (FldH Ϫ , Reaction 1). In the next step, the hydrogenase catalyzes the oxidation of FldH Ϫ to the semiquinone (FldH ⅐ , Reaction 2), because further oxidation to the quinone (Fld) is not possible due to the much more positive redox potential of E°Ј ϭ Ϫ60 mV than that of hydrogen, E°Ј ϭ Ϫ414 mV. FldH ⅐ is reduced again to FldH Ϫ by electron bifurcation (Reaction 3). In summary, crotonyl-CoA and 2 protons are reduced with 2 NADH to butyryl-CoA and hydrogen (Reaction 4), whereby flavodoxin cycles between the semiquinone and hydroquinone states. The same stoichiometry (2 NADH/ crotonyl-CoA) was observed earlier with ferredoxin (29). At the end of the reaction when crotonyl-CoA and/or NADH are consumed, the semiquinone cannot be reduced anymore (Reaction 3).

Reaction 4
Rnf Mediates the Reoxidation of Either Flavodoxin Hydroquinone or Reduced Ferredoxin-Membranes of A. fermentans solubilized with dodecylmaltoside contain a very active Rnf that catalyzes the exergonic oxidation of NADH with ferricyanide, specific activity ϭ 20 units (mg of protein) Ϫ1 (8,30,31). Now electron bifurcation allowed the complete reduction of ferredoxin or flavodoxin, with which the reaction could be measured in the physiologic direction with the natural electron donor (Reaction 5 with flavodoxin) (Fig. 5). Together with Reaction 3, the simple Reaction 6 for the reduction of crotonyl-CoA to butyryl-CoA by NADH was obtained. This stoichiometry could be verified experimentally (1.09 NADH/crotonyl-CoA, Fig. 6).   Sodium Dependence of Rnf-It was postulated that Rnfs from A. fermentans and many clostridia use the energy difference of about 200 mV between reduced ferredoxin or flavodoxin and NADH to generate an electrochemical Na ϩ gradient for ATP synthesis (7,8) as shown recently for Rnf from Acetobacterium woodii (9,32). Using inverted membrane vesicles from C. tetanomorphum and A. fermentans in two assays, with either NADH or Ti(III) citrate ϩ ferredoxin as donor and ferricyanide or NAD ϩ as acceptor, respectively, a scalar Na ϩ dependence could be not be detected. Furthermore, when the pH was raised above 7.0, Ti(III) citrate reduced NAD ϩ directly in the absence of ferredoxin and Rnf (30). In the reaction with inverted membrane vesicles from A. fermentans, Etf, Bcd, flavodoxin, NADH, and crotonyl-CoA (Reaction 6), a sodium or lithium ion dependence could indeed be observed (Fig. 7). For the determination of the apparent K m ϭ 120 Ϯ 20 M Na ϩ or 275 Ϯ 50 M Li ϩ , the Na ϩ concentrations of the assays before the addition of defined amounts of NaCl or LiCl were measured with ICP-MS. The lowest Na ϩ concentration in the assay mixture was 43 M, whereas that of Li ϩ was close to zero.

Discussion
Our results clearly demonstrate that flavodoxin can replace ferredoxin as acceptor in the Etf ϩ Bcd bifurcating system of A. fermentans, as donor of the [FeFe]hydrogenase of C. pasteurianum, and as donor/acceptor of the Rnf of A. fermentans. Thus flavodoxin or its semiquinone is reduced by NADH to the hydroquinone via electron bifurcation and reoxidized back to the semiquinone catalyzed either by the hydrogenase or by RnfABCDGE (Acfer_0108-0113) (19). The advantage of studying these reactions with flavodoxin is the facile identification of the blue neutral semiquinone, whereas with ferredoxin, the Fd, Fd Ϫ , and Fd 2Ϫ states are not easily differentiated. Thus it could be shown that the quinone was almost quantitatively reduced to the semiquinone before it was further reduced to the hydroquinone as expected from thermodynamics (Fig. 3). In contrast, the reduction of the flavodoxin semiquinone to the hydroquinone proceeded ϳ1.4 times faster than the reduction of the quinone to the semiquinone. Hence a one-electron transfer to a compound with an unpaired electron most likely proceeds over , where X signifies the concentrations of Na ϩ or Li ϩ added and X 0 signifies the concentration of Na ϩ ϭ 0.043 mM already present in the assay mixes. Error bars indicate Ϯ S.D. a lower barrier than the addition of one electron to a closed shell. Alternatively, the easier one-electron reduction of the semiquinone could be an adaption to the physiologic conditions under which flavodoxin cycles between semiquinone and hydroquinone, whereas the quinone hardly exists.
In this work, we showed for the first time that Rnf from A. fermentans indeed required Na ϩ for activity. The failure of earlier attempts to verify this effect was certainly due to the application of artificial electron acceptors and donors. The heterohexameric Rnf from C. tetanomorphum contains riboflavin, non-covalently bound riboflavin-5Ј-phosphate, two riboflavin-5Јphosphates covalently bound to RnfG and RnfD, about five iron-sulfur clusters (30), and probably one iron atom coordinated by RnfA and RnfE as in the homologous NADH-quinone oxidoreductase (Nqr) of Vibrio cholerae (33) (Fig. 8). Thus one of this plethora of cofactors could transfer an electron directly to ferricyanide or accept one electron from Ti(III)citrate without passing the Na ϩ -translocating site. Here we used electron bifurcation for the generation of flavodoxin hydroquinone (Fld 2Ϫ ) in the absence of Ti(III)citrate and were able to demonstrate the Na ϩ and the Li ϩ dependences of Rnf from A. fermentans. The apparent K m value of 120 Ϯ 20 M Na ϩ at pH 6.8 is similar to those of Rnf from A. woodii with K m values of 201 Ϯ 30 M Na ϩ at pH 6 and 155 Ϯ 39 M at pH 7.7 (34). The K m values for Li ϩ also lie in the same range for both Rnfs. Surprisingly, these values are much lower than those of the biotincontaining Na ϩ pumps glutaconyl-CoA decarboxylase from A. fermentans (35) or oxaloacetate decarboxylase from Klebsiella pneumoniae (36), which exhibit apparent K m values of 1.0 -1.5 mM Na ϩ and 25-100 mM Li ϩ .
The now confirmed Na ϩ dependence of Rnf fits well to the Na ϩ bioenergetics of A. fermentans (Fig. 8). This organism ferments glutamate via the radical dehydration of (R)-2-hydroxyglutaryl-CoA (37) followed by the Na ϩ -dependent decarboxylation of the product glutaconyl-CoA to crotonyl-CoA (35,38). Disproportionation of crotonyl-CoA yields acetate, butyrate, and hydrogen; the latter stems from reduced flavodoxin or ferredoxin generated by electron bifurcation with crotonyl-CoA and NADH. The Na ϩ -pumping enzymes of this organism are glutaconyl-CoA decarboxylase and Rnf, both of which most likely translocate two Na ϩ per turnover. According to the proposed pathway of glutamate fermentation, the decarboxylase contributes 2 Na ϩ /glutamate and Rnf adds 2 Na ϩ /5 glutamate (7). Because 1 Na ϩ / glutamate is most likely required for the import, from 2 ϩ 0.4 Ϫ 1 ϭ 1.4 Na ϩ /glutamate are left for the Na ϩ -dependent ATP synthase,leadingto1.4/4 ϭ0.35ATP/glutamateviaiongradientphosphorylation. The contribution of substrate level phosphorylation amounts to 3 ATP/5 glutamate, which sums up to 0.95 ATP/glutamate, the maximum theoretical yield (7).
The electron-bifurcating Etf ϩ Bcd system has been shown to provide a stable steady state concentration of almost completely reduced ferredoxin or flavodoxin (EЈ Յ Ϫ500 mV). Thus in addition to hydrogenase and Rnf, this system also allows us to study other enzymatic reactions in which these reduced electron carriers are substrates. The reaction rates are easily followed via the oxidation of NADH. Examples are the synthases that form 2-oxo acids from acyl-CoA and CO 2 , such as pyruvate synthase, or the tungsten-containing aldehyde dehydrogenases, which are also proposed to catalyze the reverse reactions, the reduction of free carboxylic acids to aldehydes. Further- FIGURE 8. Proposed sodium ion gradient phosphorylation driven by the reduction of crotonyl-CoA in A. fermentans. The whole picture is a modification of that taken from Ref. 9. The presence of an F 1 F o ATP synthase was deduced from the genome of A. fermentans. The model of Rnf was constructed from data obtained from Rnf of C. tetanomorphum (30) and NADH-quinone oxidoreductase (Nqr) of V. cholerae (33). For the structure of the Etf-Bcd complex see Ref. 10. more, the Etf ϩ Bcd system enables us to monitor even very difficult conversions such as nitrogen fixation or benzoyl-CoA reduction just by NADH oxidation at 340 nm.
Stimulated by the inverted redox potentials of flavodoxin as compared with those of free FAD or riboflavin (39), we propose a mechanism of electron bifurcation with Etf ϩ Bcd (Fig. 9). Crystallographic studies showed that ␣-FAD of Etf from A. fermentans is bound to domain II of the ␣-subunit with a flavodoxin-like fold (10). The similarity ␣-FAD with flavodoxin is manifested by the ability of both proteins to stabilize a semiquinone, a red anionic semiquinone in the case of ␣-FAD and the protonated blue form in the case of flavodoxin. The formation of stable semiquinones is due to the inverse redox potentials of flavodoxin (see above) and ␣-FAD. The redox potentials of ␣-FAD in the Etf from M. elsdenii (Etf Me ), a bacterium closely related to A. fermentans, have been reported as E°Ј 1 ϭ ϩ 81 mV for ␣-FAD/FAD ⅐ Ϫ and E°Ј 2 ϭ Ϫ136 mV for ␣-FAD ⅐ Ϫ /FADH (28). Although both potentials are much higher than the corresponding potentials of flavodoxin (Ϫ60 and Ϫ430 mV, respectively), there are indications that they decrease considerably when Etf forms a complex with Bcd as assumed in Fig. 9 (10, 40). Otherwise the reduction of crotonyl-CoA to butyryl-CoA (E°Ј ϭ Ϫ10 mV) by ␣-FAD ⅐ Ϫ would be difficult. Upon titration of Etf with 1 ⁄ 2 NADH ( 1 ⁄ 4 NADH/FAD), the stable red semiquinone anion ( max ϭ 370 nm), whose concentration is significantly increased in the presence of Bcd (10), is observed (see Fig. 2 and Ref. 28). A special feature of ␣-FAD is an internal hydrogen bond between the 4Ј-OH of its ribityl side chain to either N1 or to the C2ϭO group of the isoalloxazine ring. This bond, as well as several other interactions with the surrounding amino acid residues, perturbs the UVvisible spectrum of ␣-FAD in such a way that an extra absorption band around 400 nm between the two prominent peaks of the FAD quinone at 375 and 446 nm is observed. This band at 400 nm is quenched by reduction of Etf with NADH in a 1:1 ratio, whereby the remaining still oxidized ␤-FAD exhibits a normal flavin spec-trum. This spectrum is also obtained when the ␣-FAD is removed by treatment of the Etf with KBr (28,29).
␤-FAD is tightly bound to Etf between domain III, the ␤-subunit, and domain I of the ␣-subunit. It exhibits a more closed conformation, different from that of ␣-FAD and without the internal hydrogen bond. Therefore we conclude that the redox potentials of ␤-FAD are not reversed and similar to those of free riboflavin. Unfortunately, only the two-electron redox potential of the related Etf from M. elsdenii could be measured as E°Ј ϭ Ϫ279 mV for ␤-FAD/FADH Ϫ (28). Together with the one-electron redox potentials of riboflavin (Rf), E°Ј Rf1 ϭ Ϫ314 mV for Rf/RfH ⅐ and E°Ј Rf2 ϭ Ϫ124 mV for RfH ⅐ /RfH Ϫ (39), the real redox potentials of ␤-FAD could be E°Ј ␤ ϭ Ϫ279 Ϯ 1 ⁄ 2(E°Ј Rf1 Ϫ E°Ј Rf2 ) ϭ Ϫ374 mV and Ϫ184 mV. Due to interaction with the protein, the separation between E°Ј ␤1 and E°Ј ␤2 could be even larger, which is assumed in Fig. 9.
Using these estimated redox potentials for ␣and ␤-FAD, one can construct a possible mechanism for electron bifurcation. As shown by crystallographic studies, NADH binds close to ␤-FAD and reduces it to ␤-FADH Ϫ (29). Due to the ability of ␣-FAD to stabilize a semiquinone, only one electron is shifted from the low ␤-FADH ⅐ /FADH Ϫ level (about Ϫ90 mV, Fig. 9) to ␣-FAD/FAD ⅐ Ϫ (about Ϫ60 mV), which is 14 Å apart. According to the law of energy conservation, the other electron at the ␤-FAD/FADH ⅐ level becomes "red hot" (about Ϫ520 mV) (41) and reduces the much more closely located ferredoxin that is only 6 Å apart (29). Because the isoalloxazine ring of FMN sits on the surface of flavodoxin, an electron from ␤-FADH ⅐ to this acceptor probably has to pass the same distance. Now the flexible domain II carrying ␣-FAD ⅐ Ϫ must undergo a large conformational change to interact with ␦-FAD of Bcd, which, in the complex of human Etf and medium chain acyl-CoA dehydrogenase, has been shown to be 30 Å apart (42). After reduction of ␦-FAD to the semiquinone (␦-FADH ⅐ ), a second round of bifurcation generates ␦-FADH Ϫ that transfers a hydride to crotonyl- CoA, yielding butyryl-CoA. Alternatively, ␦-FADH ⅐ reduces crotonyl-CoA to an allylic ketyl radical (43,44), and the second ␦-FADH ⅐ completes the reduction to butyryl-CoA.
This mechanism is actually similar to that of the "redox seesaw" originally proposed by Peter Mitchell (41,55), although the redox potentials of the bifurcating ␤-FAD are normal and not inverted. The inverted redox potentials of ␣-FAD are necessary to generate the high potential one-electron acceptor able to form a stable semiquinone, which has to be transported over 30 Å to ␦-FAD of Bcd. Other bifurcating systems use iron sulfur clusters as high potential electron acceptors; established cases are the Rieske [2Fe-2S] cluster in the Q-cycle of complex III of the respiratory chain (45) and the [4Fe-4S] cluster N3 close to FMN of NADH-quinone oxidoreductase (Nuo, complex I) (46). The Rieske [2Fe-2S] cluster has a function very similar to that of ␣-FAD. It accepts the high potential electron from ubihydroquinone (UQH 2 ) and swings over to cytochrome c 1 , which becomes reduced. It has been proposed that in complex I, the two electrons of FMNH Ϫ bifurcate; one goes to the high potential [4Fe-4S] cluster N3, and the other goes to the lower potential [2Fe-2S] cluster N1a. Although the electron on N3 moves forward via six additional clusters to the ubiquinone, the other electron is stored on N1a until the way is free to follow the first electron. Thus the formation of a reactive oxygen species-forming FMNH ⅐ is avoided (46). With the exception of Etf ϩ Bcd, all other bifurcating systems also contain iron-sulfur clusters, which possibly serve as high potential electron acceptors (47)(48)(49)(50)(51)(52)(53)(54). Probably these clusters can be more easily tuned to the required redox potentials than flavins. Hence the Etf ϩ Bcd system appears to be a special case among the bifurcating flavoprotein complexes, in which the flavodoxin-like ␣-FAD adopts the function of an iron-sulfur cluster.
Author Contributions-N. P. C. and W. B. designed the study and wrote the manuscript. K. K. performed initial experiments. N. P. C. produced and purified recombinant flavodoxin and studied its reactions with Etf ϩ Bcd and Rnf. A. S. performed the metal analysis. All authors analyzed the results and approved the final version of the manuscript.