Structural Analysis of the fds Operon Encoding the NAD+-linked Formate Dehydrogenase of Ralstonia eutropha *

The fdsGBACD operon encoding the four subunits of the NAD+-reducing formate dehydrogenase ofRalstonia eutropha H16 was cloned and sequenced. Sequence comparisons indicated a high resemblance of FdsA (α-subunit) to the catalytic subunits of formate dehydrogenases containing a molybdenum (or tungsten) cofactor. The NH2-terminal region (residues 1–240) of FdsA, lacking in formate dehydrogenases not linked to NAD(P)+, exhibited considerable similarity to that of NuoG of the NADH:ubiquinone oxidoreductase from Escherichia colias well as to HoxU and the NH2-terminal segment of HndD of NAD(P)+-reducing hydrogenases. FdsB (β-subunit) and FdsG (γ-subunit) are closely related to NuoF and NuoE, respectively, as well as to HoxF and HndA. It is proposed that the NH2-terminal domain of FdsA together with FdsB and FdsG constitute a functional entity corresponding to the NADH dehydrogenase (diaphorase) part of NADH:ubiquinone oxidoreductase and the hydrogenases. No significant similarity to any known protein was observed for FdsD (δ-subunit). The predicted product offdsC showed the highest resemblance to FdhD from E. coli, a protein required for the formation of active formate dehydrogenases in this organism. Transcription of the fdsoperon is subject to formate induction. A promoter structure resembling the consensus sequence of ς70-dependent promoters from E. coli was identified upstream of the transcriptional start site determined by primer extension analysis.

Apart from molecular hydrogen, formate can serve as an alternative energy source for autotrophic growth of the aerobic, facultatively chemoautotrophic bacterium Ralstonia eutropha (formerly Alcaligenes eutrophus) (1). The oxidation of formate in this organism is catalyzed by two distinct types of FDH 1 : a soluble, NAD ϩ -linked enzyme (S-FDH; EC 1.2.1.2) and a membrane-bound enzyme coupled directly to the respiratory chain via an unknown electron acceptor (2). S-FDH catalyzes the irreversible oxidation of formate to CO 2 with concomitant reduction of NAD ϩ to NADH. Assimilation of CO 2 proceeds via the reactions of the reductive pentose phosphate cycle (3). S-FDH exhibits diaphorase activity by reducing electron acceptors such as methyl viologen, benzyl viologen, or ferricyanide with NADH as electron donor (4). Therefore, the enzyme has two distinct activities, a FDH and a NADH dehydrogenase activity, which combine to perform the complete catalytic reaction. The enzyme is composed of four nonidentical subunits (␣␤␥␦) and contains one molecule of each MGD and FMN in addition to a number of redox-active [Fe-S] centers as cofactors (4,5).
Based on their general structure, FDHs can be divided into two groups. The first group of enzymes comprising heteromeric FDHs with various physiological functions is characterized by the possession of molybdenum or tungsten cofactors and [Fe-S] centers. Their catalytic subunits show significant sequence similarity. Depending on the physiological function of the individual enzymes, structures and cofactor contents of the remaining subunits are more diverse (6). The S-FDH from R. eutropha belongs to this group together with FDHs from various bacterial and archaeal organisms such as Escherichia coli, Wolinella succinogenes, Moorella thermoacetica (formerly Clostridium thermoaceticum) and Methanobacterium formicicum. The second group represents the homodimeric, NAD ϩ -reducing FDH from methylotrophic bacteria and yeasts and from plants. These enzymes contain neither cofactors nor metals. Their amino acid sequences resemble considerably and are also similar to those of NAD ϩ -dependent, D-specific 2-hydroxyacid dehydrogenases like lactate dehydrogenase (7). The enzymes of the two groups share very little similarity except for a short sequence region.
The present work reports on the first cloning and sequencing of genes encoding a heteromeric FDH from an aerobic, autotrophic organism. The fds genes from R. eutropha H16 apparently form an operon consisting of the four structural genes of S-FDH (fdsA, fdsB, fdsG, and fdsD) and an additional gene (fdsC) of unknown function not related to the enzyme. Analysis of the deduced amino acid sequences enabled a prediction of the cofactor sites within the S-FDH subunits and allowed us to hypothesize on a path of intramolecular electron transfer. Furthermore, structural relationships between the subunits of S-FDH and FDHs from other organisms are discussed, and similarities to subunits of NAD ϩ /NADP ϩ -reducing hydrogenases and NADH:ubiquinone oxidoreductases (complex I) are unveiled. Transcriptional studies are also presented, aimed at the regulation and promoter identification of the operon.

EXPERIMENTAL PROCEDURES
Organisms, Phages, and Plasmids-The bacterial strains, phages, and plasmids used in this study are listed in Table I. R. eutropha was grown under air at 30°C in a mineral salts medium supplemented with either 0.2% (w/v) formate (autotrophic growth), 0.2% (w/v) fructose (heterotrophic growth), or 0.2% (w/v) formate plus 0.1% (w/v) fructose * This work was supported in part by a grant from the Deutsche Forschungsgemeinschaft. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
The nucleotide sequence(s) reported in this paper has been submitted to the GenBank TM /EBI Data Bank with accession number(s) AJ223295.
‡ Recipient of a predoctoral fellowship from the Deutscher Akademischer Austauschdienst. (mixotrophic growth) as described previously (4). Alternatively, hydrogen and CO 2 (8:1, v/v) were used as energy and carbon sources, respectively, for aerobic lithoautotrophic growth of the organism. Cultures to be employed in RNA isolation were grown in low phosphate mineral medium (11) to an optical density of 2-3 (mid-exponential phase) measured at 436 nm. Strains of E. coli were cultivated aerobically at 37°C in Luria-Bertani medium. If required, ampicillin was added to the medium at a concentration of 50 g/ml. Phages were propagated in E. coli WL87 or WL95 as host strains according to Sambrook et al. (12).
Manipulation and Sequencing of DNA-Genomic DNA from R. eutropha H16 was isolated according to Ausubel et al. (13). Plasmid DNA from E. coli XL-1 Blue was prepared by alkaline lysis of cells as detailed by Sambrook et al. (12) and, if necessary, purified further using special chromatographic columns (Qiagen, Hilden, Germany). Recombinant DNA manipulations were performed by standard procedures (12). For double-stranded sequence analysis, nested deletions in pSOH1-1, pSOH1-2, pSOH2, and pSOH5 were generated by treatment with exonuclease III and mung bean nuclease according to the instructions of the manufacturer (Stratagene, Heidelberg, Germany). Nucleotide sequences were determined by the cycle sequencing method employing a reagent kit (SequiTherm Cycle Sequencing Kit; Biozym, Hessisch Oldendorf, Germany) together with either 35 S-or fluorescence-labeled oligonucleotide primers. The oligonucleotides were purchased from or synthesized by Pharmacia (Freiburg, Germany) or MWG-Biotech (Ebersberg, Germany), respectively. Nucleotide and deduced amino acid sequences were analyzed by the latest version of the GCG program package (14). Multiple alignments of sequences were constructed by means of the programs ClustalW (15) or MACAW, version 2.0.0 (16). For Southern hybridizations, restriction fragments of DNA were separated by agarose gel electrophoresis and transferred onto a nylon mem-brane (Biodyne B; Pall, Dreieich, Germany) using a vacuum blotting device (Vacu-Gene XL; Pharmacia). Labeling of DNA probes, hybridization, and signal detection were carried out using the ECL 3Ј-Oligolabeling and Detection System as instructed by the manufacturer (Amersham Buchler, Brunswick, Germany).
Construction of a Partial Genomic Library-Genomic DNA (300 g) from R. eutropha H16 was digested to completion with 700 units of restriction endonuclease BamHI. To isolate fragments of the 8 -20-kb size range, the digested DNA was electroeluted from the corresponding gel area after agarose (0.8%, w/v) gel electrophoresis in Tris acetate buffer, pH 8.1. The fractionated DNA was then ligated into the BamHI site of vector phage L47 and subjected to in vitro packaging using the Gigapack II Gold Packaging Extract (Stratagene). The resulting phage particles representing a partial genomic library of strain H16 were initially propagated in E. coli WL95 and subsequently screened by plaque hybridization after infection of E. coli WL87. Labeling of the oligonucleotide probes specific for fds genes and signal detections were performed employing the ECL 3Ј-Oligolabeling and Detection System.
RNA Isolation and Analysis-Total RNA was isolated from R. eutropha H16 as described by Oelmü ller et al. (17). For Northern hybridization experiments, denatured RNA (20 g/lane) was applied to a formaldehyde agarose gel, separated by electrophoresis, and transferred onto a nylon membrane (Biodyne B) by vacuum blotting. DNA probes used in RNA hybridizations were labeled radioactively with [␣-32 P]dCTP by means of a random primer labeling system (Life Technologies, Eggenstein, Germany).
Primer Extension Analysis-A 30-mer oligonucleotide primer complementary to nucleotide positions 30 -59 downstream of the translational start of the fdsG gene was radioactively labeled at its 5Јend using [␥- 33 (13). To determine the sizes of these products, the same oligonucleotide was used as primer in a sequencing reaction with pSOH1-1. Autoradiography was done using either Hyperfilm ␤-max (Amersham Buchler) or Cronex 10S film (NEN).

Cloning and Sequence Analysis of the fds Gene Locus
The approach to clone the S-FDH genes of R. eutropha was based on the known NH 2 -terminal amino acid sequences of the four subunits of the enzyme (5). Southern blotting revealed that a 12-kb BamHI fragment of the chromosomal DNA from R. eutropha H16 hybridized with degenerate oligonucleotide probes deduced from these sequences (data not shown). Phage clone AEC6 containing the fragment was isolated by screening a partial genomic library of strain H16 using the ␣and ␦-probes. After subcloning of the fragment into pUC18 in both orientations relative to lacZЈ (pOH1-1 and pOH1-2), the relative positions of the S-FDH structural genes (fds genes) within the fragment (Fig. 1) were determined by restriction analysis and Southern hybridization employing all four oligonucleotides as probes.
A genomic segment of 6,589 bp was sequenced and found to contain a cluster of five colinearly oriented open reading frames that were designated as genes fdsG (531 bp), fdsB (1,563 bp), fdsA (2,880 bp), fdsC (861 bp), and fdsD (384 bp) (Fig. 1). Start (ATG) and stop codons (TGA) of the genes either overlap (fdsGB, fdsCD) or are separated by 15-(fdsAC) or 35-bp (fdsBA) intergenic regions. The codon usage agrees well with that of known R. eutropha genes. The fds genes are thus likely to form a pentacistronic operon.
A comparison of the deduced amino acid sequences with the previously determined NH 2 -terminal amino acid sequences of the S-FDH subunits (5) revealed that fdsG, fdsB, fdsA, and fdsD represent the structural genes of the four subunits ␥, ␤, ␣, and ␦, respectively. The calculated isoelectric points of the putative gene products were in the weakly acidic range for FdsA (pI ϭ 6.77), FdsB (5.44), FdsC (6.99), and FdsG (6.32), whereas it was remarkably basic (pI ϭ 9.86) for FdsD. A hydropathy analysis (18) of the five proteins suggested that none of them contains potential membrane-spanning helices and hydrophilic and hydrophobic residues to be distributed evenly within the polypeptide chains (data not shown), confirming the cytoplasmic location of S-FDH.
A multiple sequence alignment of FdsA and related proteins ( Fig. 2) revealed that eight regions (C5, O1, F1, O2, M1, F2, O3, and M2) are conserved in the ␣-subunits of all MGD-containing FDHs a well as in the periplasmic nitrate reductase of R. eutropha (34). The F1 and F2 regions are well conserved only in the FDHs and in the periplasmic nitrate reductase, but not in other molybdopterin cofactor-containing oxidoreductases (biotin sulfoxide reductase, dimethyl sulfoxide reductase, polysulfide reductase, trimethylamine N-oxide reductase; [35][36][37][38]. The selenocysteine residue essential for catalytic activity of the FDH isoenzymes FDH-H and FDH-N from E. coli (Fig. 3A) and of the tungsten-containing enzyme of M. thermoacetica (not shown) is located in the F1 region. In the other FDHs and in the periplasmic nitrate reductase cysteine (Cys 378 in FdsA) replaces selenocysteine. Selenium and sulfur, respectively, are the proposed ligands of the molybdenum in MGD (42). The neighboring histidine residue (His 379 in FdsA) is conserved in all MGD-containing FDHs and plays a role in orienting the substrate molecule formate and in proton abstraction from formate during catalysis (43). It seems conceivable that the enzymes containing cysteine instead of selenocysteine have a different catalytic mechanism, a presumption supported by the fact that FDH from M. formicicum has a sulfido group as a ligand of molybdenum (MoϭS; 44), which does not occur in FDH-H (43). The sulfido group, instead of the selenol group, possibly serves as a proton acceptor during the transfer of a hydride ion from formate to the molybdenum cofactor (45). It has been shown that S-FDH of R. eutropha is inactivated irreversibly by cyanide (4) much in the same manner as FDH from M. formicicum and the members of the xanthine oxidase family containing the MoϭS group. Cyanide inactivates molybdoenzymes by replacing MoϭS with MoϭO to yield the inactive desulfo form of the enzymes (45 and references therein). Furthermore, mutated FDH-H from E. coli, in which cysteine replaces selenocysteine, showed a 300 times lower activity than the wild-type enzyme (46). The recent determination of the crystal structures of dimethyl sulfoxide reductase from Rhodobacter sphaeroides and FDH-H from E. coli revealed in both enzymes two MGD molecules to be involved in the coordination of one molybdenum atom by means of four dithiolene ligands (43,47). In discord with these findings, quantitation of MGD yielded 0.71 mol/mol of R. eutropha S-FDH, indicating that this enzyme contains only one molecule of molybdenum cofactor (4), although the catalytic subunits of all MGD-containing FDHs share high sequence similarity. Thus, the catalytic core and mechanism of S-FDH appear to be distinct from those of FDH-H. It is proposed that S-FDH, despite the high degree of sequence similarity to FDH-H, contains only one MGD and at least one sulfido group in the coordination sphere of molybdenum.
The COOH-terminal half of the F2 region in these FDHs exhibited significant local similarity to a corresponding stretch in the NAD ϩ -reducing, homodimeric FDH from Pseudomonas sp. 101, where a catalytically essential arginine (Arg 284 in FDH of Pseudomonas sp. 101, Arg 579 in FdsA) is located (Fig. 3B). It was demonstrated by x-ray crystallography with FDH-H (FdhF) of E. coli (43) and FDH of Pseudomonas sp. 101 (7) that this residue forms a hydrogen bond with formate in the active site, in FDH-H together with the above mentioned histidine. Both F1 and F2 thus seem to be parts of the active center of FDHs. The fact that the two regions are also present in periplasmic nitrate reductase suggests a structurally similar active center to occur in this protein. Nitrate, the substrate of periplasmic nitrate reductase, is regarded as a structural analog of formate in the transition state during catalysis by FDH (7).
The C5 region includes the sequence motif Cys-Xaa-Xaa-Cys-Xaa-Xaa-Xaa-Cys-Xaa 26 -34 -Cys (see Figs. 2 and 3A, Fig. 4 (49). This binuclear center has not been detected in complex I from other organisms. Because the Cys cluster of the C5 region is only present in NuoG of E. coli but not in its homologs from other organisms (Fig. 4), it was proposed to be involved in the coordination of the [2Fe-2S] center N1c. The rhombic EPR spectrum of N1c (g x, y, z ϭ 1.92, 1.95, 2.00; 49) is identical to that of the [2Fe-2S] center of formate-reduced S-FDH (5). It was reported that inactivation Interestinglsy, of the remaining regions (O1, O2, O3, M1, and M2) conserved in MGD-containing enzymes, three (O1, O2 and O3) are also present in NuoG of E. coli (40) and its homolog from the (mitochondrial) complex I from B. taurus (NuaM; 41), which do not contain a molybdenum cofactor (see Fig. 2). The three regions are thus unlikely to be directly required for coordination of MGD but may rather be related to the general oxidoreductase functions of these enzymes. The occurrence in a polypeptide of regions corresponding to M1 and M2 may be an indication of the presence of a molybdenum cofactor because these two regions seem to be conserved only in enzymes containing the cofactor.
The sequence motif in the C1 cluster of FdsA, Cys-Xaa 10 -Cys-Xaa-Xaa-Cys-Xaa 13 -Cys, is similar to that of [2Fe-2S] ferredoxins and ferredoxin-like proteins (Cys-Xaa 4 -Cys-Xaa-Xaa-Cys-Xaa n -Cys) from bacteria and plants (58), suggesting that this cluster probably coordinates a [2Fe-2S] center. Only three cysteine residues are conserved in cluster C2 (Fig. 4). Because four ligands are required to coordinate one tetranuclear [Fe-S] center, a histidine (His 112 in FdsA) lacking in HoxU might provide the fourth ligand for a [4Fe-4S] center through the nitrogen atom (N-3) of its imidazol ring as proposed for respiratory nitrate reductases from E. coli (59,60). However, it cannot be ruled out that this cluster is involved in the formation of a [3Fe-4S] center, or the fourth ligand might be provided by a cysteine residue from another subunit. Eight cysteines are conserved in the C3 and C4 clusters of FdsA, HndD, and HydI as well as in HoxU from A. variabilis, whereas only four cysteines occur in the corresponding regions of NuaM and its homologs and in HoxU from R. eutropha (Fig. 4). The residues presumably coordinate two [4Fe-4S] centers and are organized in the repeated sequence motif Cys-Xaa-Xaa-Cys-Xaa-Xaa-Cys-Xaa-Xaa-Xaa-Cys. In such cases the first three cysteines of each cluster coordinate the same center, and the remaining cysteine of each cluster serves as the fourth ligand for the other center (61) FdsA appears to consist of two functional domains. The NH 2terminal domain (residues 1-240) is closely related to HoxU and the NH 2 -terminal part of NuoG. This domain is lacking in the catalytic subunits of the other FDHs which do not catalyze the reduction of NAD(P) ϩ . Therefore, in S-FDH the NH 2 -terminal domain of FdsA is suggested to play an equivalent role in the diaphorese (NADH dehydrogenase) activity of the enzyme like HoxU in the hydrogenases, whereas the remaining part of the polypeptide is directly involved in the oxidation of formate. Such a modular structure of FdsA suggests a possible evolution of FdsA, HoxU, and NuoG from a common ancestor, during which FdsA acquired the additional function of formate oxidation. Alternatively, FdsA might be the fusion product of a FDH polypeptide and the NH 2 -terminal domain of a NuoG-related protein.
FdsB-The protein encoded by fdsB is composed of 520 amino acids and has a calculated molecular mass of 55.1 kDa, again in good agreement with earlier data (4). It revealed no significant sequence similarity to subunits of known FDHs, but a relatively high degree of identity (34 -45%) to NuoF from E. coli (40) and homologs of the latter from P. denitrificans (Nqo1; 62) and B. taurus (NubM; 63). The similarity extends to the major portion (residues 196 -602) of HoxF (␣-subunit of NAD ϩreducing hydrogenase) from R. eutropha (39) (Fig. 5). These related proteins are involved in the diaphorase activity of each corresponding enzyme and are assumed to contain one FMN and one [4Fe-4S] center as cofactors. Three conserved regions were recognized. The first includes a glycine-rich fingerprint motif (Gly-Xaa-Gly-Xaa-Xaa-Gly-Xaa 25-26 -Glu), which is thought to be involved in the formation of the ADP pocket within the NAD ϩ binding site (64). In FdsB the third glycine residue is replaced by alanine. The presence of this motif in the protein suggests that the reduction of NAD ϩ is catalyzed by the ␤-subunit of S-FDH. The second region is also a glycine-rich segment probably forming the FMN binding site (63,65). Four conserved cysteines in the third region are arranged into the sequence motif Cys-Xaa-Xaa-Cys-Xaa-Xaa-Cys-Xaa 39 -40 -Cys typical for [4Fe-4S] ferredoxins, suggesting that they serve as ligands of a [4Fe-4S] center.
FdsG-The fdsG gene encodes a protein of 176 amino acids with a calculated molecular mass of 18.7 kDa, which also corresponds to a previous determination (4). The protein showed a sequence identity of 27-34% to NuoE from E. coli (40) and homologs from P. denitrificans (Nqo2; 66) B. taurus (NuhM; 67). Significant resemblance was also detected to the NH 2 -terminal domains (residues 1-150) of HoxF from R. eutropha (39) (Fig. 6). Evidence provided by EPR studies, sequence analyses, and heterologous expression experiments allowed the assignment of one [2Fe-2S] center to this group of proteins (49,61,68). In FdsG such a center should be accommodated by the four conserved cysteines Cys 87 , Cys 92 , Cys 128 , and Cys 132 . Based on the sequence analyses of FdsA, FdsB, and FdsG, the R. eutropha S-FDH holoenzyme appears to contain three [2Fe-2S] and four [4Fe-4S] centers, which is in close agreement with the chemically determined contents of 21.5 g atoms of iron and 21 g atoms of acid-labile sulfur/mol of enzyme (4,5).
FdsD-The FdsD protein encoded by fdsD represents the ␦-subunit of S-FDH from R. eutropha. It has a calculated molecular mass of 13.9 kDa, being in reasonable agreement with the value reported earlier (4). The protein exhibited no significant sequence similarity to other proteins filed in the data bases. Because the calculated isoelectric points of the ␣-, ␤-, and ␥-subunits were 6.77, 5.44, and 6.32, respectively, and that of the purified holoenzyme was found to be 4.4 (4), the surface charge of these subunits should be negative under physiological conditions. The high pI of FdsD (9.86) suggests that this subunit may play a role in maintaining the quaternary structure of S-FDH by means of electrostatic interactions with the other subunits. FdsD does not appear to participate in the intramolecular electron transfer within the enzyme because it lacks redox components as concluded from the sequence analysis.
FdsC-Although the fdsC gene is located between fdsA and fdsD as part of the presumed fds operon, its product is not a constituent of S-FDH. FdsC is composed of 286 amino acids with a calculated molecular mass of 31 (71). Likewise, NarQ is required for activity of the phenazine methosulfatelinked FDH in B. subtilis. FdhD is not a structural component of a FDH isoenzyme, does not function in the biosynthesis of the molybdenum cofactor in E. coli, and does not participate in the transcriptional regulation of the fdnGHI operon encoding the FDH-N (72). A possible involvement of FdsC in the formate metabolism of R. eutropha remains to be investigated.

Intramolecular Electron Transfer within S-FDH
Judging from x-ray crystallographic data on FDH-H from E. coli (43), C5 in FdsA is assumed to be the primary electronaccepting [Fe-S] center in R. eutropha S-FDH which is reduced upon oxidation of formate. The further intramolecular electron transport may involve all or some of the remaining centers (C1, C2, C3, and C4) in the NH 2 -terminal domain of FdsA, probably depending on their redox potentials. FMN is known as a redox component in electron transport chains which mediates electron transfer between one-electron ([Fe-S] centers) and two-electron carriers such as NAD(P) ϩ . The flavin is thus thought to take up two electrons (in one-electron steps) from the [4Fe-4S] center (C6) in FdsB and to transfer them to NAD ϩ . The C7 center ([2Fe-2S]) in FdsG could have a bridging function in the transfer between FdsA and FdsB (Fig. 7). In accordance with this proposal, a [2Fe-2S] (g x,y,z ϭ 1.93, 2.00, 2.01) and a [4Fe-4S] center (g x,y,z ϭ 1.90, 1.95, 2.04) were observed by EPR after reduction of S-FDH with NADH (5), which may correspond to C7 and C6, respectively. More extensive EPR studies together with Mössbauer spectroscopy in addition to determination of the midpoint potentials and mutational inactivation of the centers will be required to elucidate the electron flow in S-FDH.

Relationship Between S-FDH and Complex I of E. coli
Complex I (NUO) of respiratory chains couples the electron transfer from NADH to ubiquinone with translocation of protons across the mitochondrial inner membrane in eukaryotes or the cytoplasmic membrane in prokaryotes. The complexes of B. taurus and N. crassa contain at least 41 and 35 nonidentical polypeptides, respectively, whereas NUO of E. coli is composed of only 14 subunits (40,49,61), representing a minimal form of a functional complex I. One of the three NUO subcomplexes is a NADH dehydrogenase fragment that consists of the three subunits, NuoE, NuoF, and NuoG, and contains one molecule of FMN, probably three [2Fe-2S] centers (N1a, N1b, and N1c) and at least two [4Fe-4S] centers (N3 and N4). The fragment is regarded as the electron-input module of NUO (49). NuoE, NuoF, and NuoG considerably resemble FdsG, FdsB, and FdsA, respectively (Fig. 8), although the overall sequence similarity between NuoG and FdsA is much lower than those of the The nuo operon of E. coli is comprised of 14 genes (nuoA-N), of which nuoE, nuoF, and nuoG encode the NADH dehydrogenase fragment of NUO (40,49). It is interesting that the relative arrangement of these three genes within the operon is same as that of their counterparts fdsG, fdsB, and fdsA in the fds operon of R. eutropha. The start codons of nuoF and fdsB overlap with the stop codons of their respective preceding genes, nuoE and fdsG. In contrast, nuoF and nuoG are separated by a 52-bp intergenic region, whereas the intergenic sequence between fdsB and fdsA has a length of only 35 bp. The conserved gene arrangement and conspicuous sequence similarities of the corresponding gene products suggest that the NADH dehydrogenase fragment of NUO is phylogenetically related to S-FDH, i.e. three subunits of S-FDH and the NADH dehydrogenase fragment have evolved from a common ancestor. This suggestion supports the modular evolution theory of complex I (61).

Relationship Between S-FDH and NAD(P) ϩ -reducing Hydrogenases
It is of considerable interest to note that the S-FDH subunits FdsB and FdsG, together with the NH 2 -terminal domain (residues 1-240) of FdsA, have functional and structural relations to the diaphorase part of the NAD ϩ /NADP ϩ -reducing hydrogenases from R. eutropha and other organisms (Fig. 8). The R. eutropha hydrogenase is a heterotetramer (␣␤␥␦) with the ␣␥dimer (HoxF-HoxU) constituting the diaphorase part, whereas the other two subunits (HoxH-HoxY) are homologous to dimeric [NiFe] hydrogenases (39,51,55,73). The NH 2 -terminal part (residues 1-150) of HoxF resembles FdsG, although it does not carry a [2Fe-2S] center, and the remaining part of HoxF is homologous to FdsB of R. eutropha S-FDH. Furthermore, HoxU is related to the NH 2 -terminal part of FdsA. It is thus assumed that FdsB, FdsG, and the NH 2 -terminal domain of FdsA constitute a functional unit within the S-FDH corresponding to the diaphorase dimer of the hydrogenases from R. eutropha and A. variabilis. The minimal structural prerequisites for a diaphorase appear to be subunits or domains corresponding to NuoE, NuoF, and the NH 2 -terminal domain of NuoG which carry an appropriate set of [Fe-S] centers. In these two hydrogenases these requirements are met by the ␣␥-dimer because HoxF is regarded as a fusion product of NuoE-and NuoF-related proteins (73), and HoxU corresponds to the NH 2 -terminal domain of NuoG. In contrast, three subunits of the hydrogenase (HndA, HndC, and NH 2 -terminal domain of HndD) from D. fructosovorans and of S-FDH (FdsG, FdsB, and FdsA) from R. eutropha seem to be required for the diaphorase function.

Transcriptional Studies
To gain initial insight into the transcriptional regulation of the putative fds operon, Northern hybridizations were performed. For this purpose total RNA was isolated from organoautotrophically (formate), lithoautotrophically (H 2 ϩ CO 2 ), or heterotrophically (fructose) grown R. eutropha H16 and probed with a DNA fragment specific for the proximal fdsG gene. Hybridization signals were detected only in the RNA from formate-grown cells (Fig. 9A), suggesting that the induction of the operon requires the presence of formate. No distinct fds mRNA species but a wide range of transcripts with a maximal size of about 6 kb was observed which probably represented the steady state of rapid synthesis and degradation/processing of the primary pentacistronic transcript. Another probe comprising the two distal genes fdsC and fdsD also hybridized only to RNA from formate-grown cells (Fig. 9B). These findings support the notion that the five fds genes form a transcriptional unit subject to formate induction.
The potential transcriptional start site of the fds operon was mapped by primer extension experiments. With RNA isolated from organoautotrophically or mixotrophically grown cells, a single extension product of the same length was observed (Fig.  10A). The 5Ј-end of this product corresponded to the 5Ј-end of the fds transcript and was located 100 bp upstream of the translational start of fdsG. Because no signal was observed with RNA from fructose-grown cells, the data corroborated the conclusion that formate is required for the induction of fds operon transcription. A sequence resembling 70 -dependent promoters of E. coli was detected immediately upstream of the determined transcriptional start site (Fig. 10B). The Ϫ35 region and particularly the Ϫ10 region are well conserved and separated by 17 bp, being consistent with optimal spacing between these regions of 70 -dependent promoters (74).
Formate induction of the genes was also observed in E. coli harboring plasmid pOH1-2, which carries the fds genes in an orientation divergent to lacZЈ (data not shown). These findings suggests that the 12-kb BamHI fragment contains cis-and trans-acting control elements involved in transducing the formate signal to the induction of the genes. Future studies will focus on the genetic basis of the regulation of the fds operon in R. eutropha.