A New Type of NADH Dehydrogenase Specific for Nitrate Respiration in the Extreme Thermophile Thermus thermophilus *

A four-gene operon ( nrcDEFN ) was identified within a conjugative element that allows Thermus thermophilus to use nitrate as an electron acceptor. Three of them encode homologues to components of bacterial respiratory chains: NrcD to ferredoxins; NrcF to iron-sulfur-containing subunits of succinate-quinone oxidoreduc-tase (SQR); and NrcN to type-II NADH dehydrogenases (NDHs). The fourth gene, nrcE, encodes a membrane protein with no homologues in the protein data bank. Nitrate reduction with NADH was catalyzed by membrane fractions of the wild type strain, but was severely impaired in nrc :: kat insertion mutants. A fusion to a thermophilic reporter gene was used for the first time in Thermus spp. to show that expression of nrc required the presence of nitrate and anoxic conditions. Therefore, a role for the nrc products as a new type of membrane NDH specific for nitrate respiration was deduced. Consistent with this, nrc :: kat mutants grew more slowly than the wild type strain under anaerobic Fitness Assays and Kinetics of Anaerobic Growth— An insertion mu- tant in a gene coding for glutamate dehydrogenase ( gdh :: kat ) was used as a wild type strain in these experiments to avoid any putative influ- ence arising from the presence of the kat gene during growth. A spon-taneous chloramphenicol-resistant mutant (Cam r ) was isolated from this strain to distinguish it after the competition experiments by its ability to grow on plates containing this antibiotic. Complementary experiments were developed in which Cam r derivatives from each of the nrc mutants were used to ensure the absence of selection bias because of chloramphenicol resistance. Competition assays were developed by inoculating an identical num- ber of viable cells of the strains to be assayed (a Cam r wild type, and Cam s nrc :: kat mutant, or vice versa) into the same tubes containing 10 ml of TB with potassium nitrate (40 m M ). Tubes were filled to the top with mineral oil, screw-capped, and incubated at 70 °C. After 24 h, viable cells were plated in parallel on TB plus kanamycin (permissive medium) and on TB plus kanamycin and chloramphenicol (restrictive medium) and incubated aerobically at 70 °C. Individual growth was followed in a similar way, but with each strain inoculated in separate tubes. Anaerobic growth was followed by inoculating identical amounts of cells of a specific organism in several tubes, which were incubated at 70 °C, as above. At given times, a tube was removed from the oven, and the optical density was measured and the number of viable cells esti- mated, as before. Competition experiments under aerobic conditions were conduced in a similar manner, but using a 100-ml flask containing 20 ml of TB and shaken at 150 rpm. Electron Transport Assays— Electron transport toward the NR was measured as the amount of nitrite produced after 5 min at 80 °C (21) by membrane fractions containing 10 (cid:4) g of protein isolated from cultures of the wild type and each nrc :: kat mutant grown for 7 h with nitrate (40 m M ) under anoxic conditions. To avoid extensive membrane oxidation during cell disruption, the buffer Tris-HCl,

Many facultative anaerobes adapt their respiratory electrontransport chains to changing environmental conditions by synthesizing specific primary dehydrogenases and final reductases (or oxidases) (1). As a paradigmatic example, Escherichia coli encodes up to 15 primary dehydrogenases and 10 final reductases, many of which appear to be biochemically redundant (2). However, these enzymes actually show differences, such as substrate affinity, or the ability to couple electron transport to proton or sodium extrusion (3), which are adapted to specific growth conditions. Therefore, there is a sophisticated regulatory network of transcriptional controls that finely tunes their expression according to environmental signals and leads to the coexistence of a limited number of dehydrogenase-reductase pairs in a given cell. Regulatory elements responding to dioxygen and to diverse electron acceptors are the major environmental factors responsible for such transcriptional tuning, allowing a hierarchical expression of final reductases and, in some cases, appropriate dehydrogenase(s), even when different electron acceptors are simultaneously present (3).
Two types of NADH dehydrogenases (NDHs) 1 can function as the first component of respiratory chains in E. coli and other Enterobacteria. Type I NDH is a H ϩ -translocating NADH: quinone reductase, homologous to the 42-subunit mitochondrial respiratory complex I (4). In E. coli, this enzyme consists of 14 subunits and a large number of prosthetic groups and redox centers (5). These 14 subunits are cotranscribed in a single mRNA, which is preferentially expressed under "energylimited" growth conditions (anaerobic or microaerophilic) to provide a higher H ϩ /e Ϫ ratio. Under aerobic conditions with highly reduced substrates ("high energy"), an alternative NADH:ubiquinone oxidoreductase, or type II NDH, plays the main role in supplying electrons to the respiratory chain. With the exception of the mitochondria from certain yeast and fungi, type II NDH homologues are absent from eukaryotes. Type II NDHs are monomeric enzymes (47 kDa) that do not couple NADH oxidation to proton pumping, thus keeping a lower H ϩ /e Ϫ ratio under high energy growth conditions. Functionally, these enzymes are peripheral membrane proteins containing FAD as a cofactor, which can feed the electron transport chains by direct reduction of the quinone pool (5).
Most isolates of the ancient thermophilic bacteria Thermus spp. require oxygen to grow (6), despite the presence in their genome of genes encoding homologues to proteins involved in the anaerobic energy metabolism of other bacteria. In the recently sequenced HB27 strain of the aerobe Thermus thermophilus, a single type I NDH and a succinate dehydrogenase seem to be the major electron providers for respiratory chains toward either low (caa3) or high (baa3) affinity cytochrome c oxidases (7).
By contrast with the aerobic character of the HB27 strain, the closely related isolate T. thermophilus HB8 can grow anaerobically with nitrate as the electron acceptor (8). Such dissimilative nitrate reduction ends with nitrite as the final product, and depends on the presence of a seven-gene operon (narCGHJIK 1 K 2 ), which encodes a respiratory nitrate reductase (NarGHI), a dedicated chaperone (NarJ), two proteins involved in nitrate/nitrite transport (NarK 1 , NarK 2 ), and a di-heme cytochrome c (NarC), which is required for the synthesis of active enzyme (9,10). The nar operon is actually encoded as part of a genetic element that could be transferred by conjugation to the aerobic strain HB27, allowing the exconjugant (HB27::nar) to grow anaerobically (11). Pulse-field gel electrophoresis revealed that a DNA fragment with a minimum size of 30 kbp was transferred from the chromosome of the donor strain (HB8) to that of the receptor (HB27) during conjugation. As the size of the nar operon (ϳ12 kbp) was around a third of this, it seemed likely that genes related to regulation and/or the use of nitrate as an electron acceptor could also be encoded within this mobile DNA element. Therefore, we wondered whether anaerobically expressed genes could also form part of this element to render the receptor cells optimally suited for anaerobic growth.
In this work we describe the presence of a four-gene operon (nrcDEFN) within this mobile DNA element that confers on the receptor cells the ability to grow anaerobically. We demonstrate that the nrc operon encodes a dedicated new type of primary dehydrogenase that is synthesized in parallel with the nitrate reductase (NR), leading to a complete respiratory chain with nitrate as the final electron acceptor. We present evidence to support the hypothesis that the nrc operon makes the organism capable of a more efficient anaerobic growth by using nitrate as an electron acceptor. Table I sets out the details of the description and origin of the bacterial strains and plasmids used in this work. E. coli was grown in LB (12) at 37 or 30°C. Kanamycin (30 mg/liter), ampicillin (100 mg/liter), and/or chloramphenicol (20 mg/liter) were used as required. Aerobic growth of T. thermophilus was conducted at 70°C with shaking (150 rpm) in TB (8). For plasmid and mutant selection, kanamycin (30 mg/liter) was added. For anaerobic growth, cells grown under aerobic conditions were inoculated in 10-ml cultures of TB with KNO 3 (40 mM), and subsequently incubated in 15-ml tubes filled to the top with mineral oil and screw-capped.

Strains, Plasmids, and Growth Conditions-
For induction of nitrate/anoxia-dependent promoters, cells were grown aerobically in TB without nitrate up to an A 550 of 0.4, and transcription was activated by the addition of KNO 3 (40 mM) and the simultaneous cessation of shaking. Because of the low solubility of oxygen at high temperatures, and its consumption by growing bacteria, unshaken cultures rapidly become anaerobic. Cells were kept in such conditions for 4 -7 h before being processed.
T. thermophilus transformation was achieved on naturally competent cells (13). Transformation of E. coli was carried out as described (14). Plates containing the appropriate antibiotic(s) and concentrations were used for selection.
Nucleic Acid Techniques-DNA isolation, plasmid purification, restriction analysis, and plasmid construction were developed as described (15). Automatic methods (Applied Biosystems) were used for sequencing. RNA was purified with the Tri-Reagent-ls kit (Molecular Research Center, Inc., Cincinnati, OH). PCR and RT-PCR of the experiment shown in Fig. 3 were performed with the DNA polymerase from T. thermophilus, and the Retrotools kit as described by the manufac-  (Table II) and PCR conditions were used. Northern and Southern blots were done as described, and specific RNA or DNA was detected with 32 P-labeled DNA probes (15).
Cloning of the nrc Operon-The nrc operon was cloned from a T. thermophilus HB8 gene library constructed in EMBL3 (17). 32 P-Labeled probes from the 5Ј extreme of the nar operon (narC gene) were used for screening. Positive clones were analyzed by enzyme restriction, and adequate DNA restriction fragments were subcloned into pUC119 for sequencing purposes. The plasmid pUP3b (Table I), which contains the whole nrc operon, was selected.
Isolation of nrc Mutants-nrc::kat mutants were obtained by insertion of the kat gene, which encodes a thermostable resistance to kanamycin (18) at appropriate restriction sites of the target gene, always maintaining the same transcription direction of the operon to allow the expression of downstream genes. Previous work on the role of narC allowed us to ensure the transcription of downstream genes after insertion of the kat cassette (10). To inactivate nrcF, the kat gene was inserted into a XhoI site, which led to the truncation of the NrcF protein at amino acid position 151. nrcN was inactivated by inserting the kat gene at a PstI site, which caused truncation of the NrcN sequence at amino acid position 304. Inactivation of nrcE required the insertion of kat gene at a NcoI site, leading to a truncated NrcE protein of 310 amino acids that included the whole membrane domain. Plasmids con-taining the inserted kat genes were used to transform T. thermophilus HB8 to obtain the expected replacement after homologous recombination. Those colonies that grew after 48 h at 70°C on plates containing kanamycin were analyzed for the presence of the expected mutation by PCR and Southern blot.
The suicidal plasmid pKUPRBS␤gal was used for the insertion of the bgaA gene into nrcN. This plasmid is a pK18 (Table I) derivative that carries a DNA fragment for recombination that includes the whole nrcF gene and a region from ϩ1 to 492 of the nrcN coding sequence, followed by the bgaA gene. This gene encodes a thermostable ␤-galactosidase from Thermus sp. T2 (19), preceded by the Shine-Dalgarno sequence of the slpA gene, which codes for the S-layer protein of T. thermophilus HB8 (20). After transformation with this suicidal vector, those clones resistant to kanamycin were tested for the presence of the expected transcriptional fusion.
Overproduction of NrcD and Antiserum Preparation-To overproduce NrcD, its coding region was amplified by PCR (primers ferred.ndeI and ferred.rev, Table II) and cloned into the NdeI and EcoRI restriction sites of plasmid pET22b (Novagen) to render pET22nrcD. After induction of exponential cultures of transformed E. coli BL21/DE3 cells with isopropyl 1-thio-␤-D-galactopyranoside (1 mM), the soluble fraction was subjected to heat denaturation to yield a NrcD-enriched soluble fraction. Slices of 10% SDS-PAGE containing the overexpressed protein were sent to a private company (Charles River Laboratories, Chalaronne, France) for immunization of New Zealand rabbits. Antiserum was incubated with E. coli cells (DH5␣) for 1 h and centrifuged before use.

New Type of NDH Involved in Nitrate Respiration
Fitness Assays and Kinetics of Anaerobic Growth-An insertion mutant in a gene coding for glutamate dehydrogenase (gdh::kat) was used as a wild type strain in these experiments to avoid any putative influence arising from the presence of the kat gene during growth. A spontaneous chloramphenicol-resistant mutant (Cam r ) was isolated from this strain to distinguish it after the competition experiments by its ability to grow on plates containing this antibiotic. Complementary experiments were developed in which Cam r derivatives from each of the nrc mutants were used to ensure the absence of selection bias because of chloramphenicol resistance.
Competition assays were developed by inoculating an identical number of viable cells of the strains to be assayed (a Cam r wild type, and Cam s nrc::kat mutant, or vice versa) into the same tubes containing 10 ml of TB with potassium nitrate (40 mM). Tubes were filled to the top with mineral oil, screw-capped, and incubated at 70°C. After 24 h, viable cells were plated in parallel on TB plus kanamycin (permissive medium) and on TB plus kanamycin and chloramphenicol (restrictive medium) and incubated aerobically at 70°C. Individual growth was followed in a similar way, but with each strain inoculated in separate tubes. Anaerobic growth was followed by inoculating identical amounts of cells of a specific organism in several tubes, which were incubated at 70°C, as above. At given times, a tube was removed from the oven, and the optical density was measured and the number of viable cells estimated, as before. Competition experiments under aerobic conditions were conduced in a similar manner, but using a 100-ml flask containing 20 ml of TB and shaken at 150 rpm.
Electron Transport Assays-Electron transport toward the NR was measured as the amount of nitrite produced after 5 min at 80°C (21) by membrane fractions containing 10 g of protein isolated from cultures of the wild type and each nrc::kat mutant grown for 7 h with nitrate (40 mM) under anoxic conditions. To avoid extensive membrane oxidation during cell disruption, the buffer Tris-HCl, 20 mM, dithiothreitol, 1 mM, pH 8, was used, and cells were ruptured by sonication under N 2 flow. No other special conditions were employed during the NADH-oxidation and nitrite-production assays, which were carried out under atmospheric oxygen pressure. NADH was assayed as putative electron donor over a wide concentration range to check for saturating conditions in a buffer of glycine (125 mM, pH 10) containing 20 mM potassium nitrate.
Enzyme Activities-The NR activity was measured at 80°C on entire cells permeabilized with (0.1% w/v) tetradecyl-trimethylammonium bromide (8) with reduced methyl viologen as the electron donor, and potassium nitrate (20 mM) as the electron acceptor (21). The thermostable ␤-galactosidase was assayed at 70°C with ortho-nitrophenyl galactopyranoside on the soluble fraction from disrupted cells. Enzyme units were normalized as described and expressed in Miller units (22). NDH activity was tested spectrophotometrically at 80°C on soluble or membrane fractions (10 g of protein) by NADH consumption after 15 min, using buffer, Tris-HCl, 50 mM, dithiothreitol 0.2 mM, pH 7.5. Potassium nitrate (20 mM) was added for the assays with membrane fractions. Different concentrations of rotenone ((2R,6aS,12aS)-1,2,6.6a,12,12a-hexahydro-2-isopropenyl-8, 9- Bacterial Two-hybrid Assays-To analyze the putative interactions between the products of the nrc genes, a bacterial two-hybrid system was used based on functional reconstitution of the adenylate cyclase of Bordetella pertussis (23). The complete amino acid sequences of NrcD, NrcE, NrcF, and NrcN and the C-terminal domain of NrcE (positions 264 -363) were expressed in plasmids pT25 and pT18 as N-or C-terminal fusion to the T25 and T18 domains of the adenylate cyclase from B. pertussis, respectively (23). Each coding gene was cloned in both plasmids to check for any possible dependence of the interactions on the N-or C-terminal location of the protein within the fusions. Constructions were carried out with the primers described in Table II. Positive controls for interaction between mesophilic proteins were developed by using fusions of a leucine zipper domain to the T18 (pT18zip) and T25 (pT25zip) domains (23). Two additional positive controls for the interaction between thermophilic proteins were used: (i) the homo-oligomeric glutamate dehydrogenase Gdh (Gdh/Gdh), and (ii) NarG and its chaperone NarJ (NarG/ NarJ) from the nitrate reductase, both from T. thermophilus HB8. Positive interactions were detected by ␤-galactosidase activity of E. coli BTH101 cells cotransformed with pT18 and pT25 derivatives.
Sequence Comparison and Analysis-Sequence comparisons and multiple alignments were carried out using the BLAST and CLUS-TALW programs (24) available at http://us.expasy.org. Phylogenetic analyses were done at http://www.genebee.msu.su/services/phtree_ reduced.html.

Cloning the nrc Cluster
The use of a probe against the narC gene in a EMBL3 gene library of the facultative anaerobe T. thermophilus HB8 allowed the cloning of a large DNA region upstream of the nar operon. Its sequence revealed the presence of a type II NDH homologue and components of electron transport chains in the same DNA strand as the nar operon. A ϳ3500-bp DNA fragment from this region, for which the 3Ј extreme was located 4.6 kbp upstream from narC, was subcloned into pUC119 to yield pUP3b, and sequenced. Based on their physical proximity to the nar operon and their similarity to electron transport components (see below), the genes were named nrc (nitrate respiration chain) (Fig. 1). When the sequence of the nrc cluster (accession number AJ585200) was compared with the genome of the aerobic strain T. thermophilus HB27 (accession numbers AE017221 and AE017222), no positive hits were found, implying that this cluster was specific to facultative anaerobic strains of T. thermophilus.

The nrc Cluster Is Transferred by Conjugation Along the nar Operon
The putative genetic linkage between the nrc gene cluster and the nar operon was checked by PCR. As shown in Fig. 2, amplification products corresponding to the first (nrcD) and the last (nrcN) genes of the nrc cluster were identified in the facultative anaerobes HB27::nar, and HB8 (lanes 2 and 3), but not in the aerobe HB27 (lanes 1). Identical results were obtained when a gene from the nar operon (narH) was assayed. By contrast, the gdh (glutamate dehydrogenase) 2 and csaB (cell-surface anchoring) (25) genes, which are not involved in energy metabolism, were identified in the three strains. Therefore, the nrc cluster was part of the conjugative genetic element transferred by conjugation along the nar operon during the isolation of the facultative anaerobic strain HB27::nar.

The Four nrc Genes Are Cotranscribed
The sequence and organization of the nrc genes suggested that they were cotranscribed into a single mRNA. To check this, we isolated total RNA from aerobically grown and nitrate/ anoxia-induced cultures ("Experimental Procedures") of T. thermophilus HB8, and tested for the presence of nrc mRNA by Northern blot and RT-PCR.
As shown in Fig. 3A, an mRNA of around 3.2 kb was detected in cells subjected to anoxic conditions with nitrate, but not in cells grown aerobically. Data from RT-PCR (  (Table II). Promoters for the nrc and nar clusters are indicated by curved arrows. The positions of the following primers are indicated: 1) ferred.nde1; 2) ferred.rev; 3) orfF.rev; 4) orfN.rev; 5) or-fF.dir; and 6) orfN.dir. into a single mRNA that extends from the 5Ј sequences of nrcD to the 3Ј extreme of nrcN. Parallel assays with sense primers located 0.5 kbp upstream of nrcD did not detect any transcripts (not shown), implying that nrcD constitutes the first gene of the nrc operon. On the other hand, the absence of any significant open reading frame in the sense strand downstream of nrcN, and the presence of an open reading frame encoded in the complementary strand of this region (not shown), suggest that nrcN is the final gene of the transcript. Consequently, the four nrc genes constitute an operon that requires either anoxia, nitrate, or both to be expressed.

Nitrate and Anoxia Are Required for Transcription of nrc
To check if nitrate or anoxia were independently able to induce expression of the nrc operon, and to quantify the induction levels achieved, we developed a new suicidal vector that carries a reporter gene (bgaA) encoding a thermophilic ␤-galactosidase preceded by its own Shine-Dalgarno sequence. A fragment of nrcN was placed upstream of the reporter as a recombination target. Insertion of this construction into the chromosome was selected by the thermostable resistance to kanamycin conferred by the kat gene located in the suicidal plasmid ("Experimental Procedures"). In the mutant strain nrcN::bgaA, the reporter gene was located at position 483 of nrcN (Fig. 1), giving rise to a transcriptional fusion nrcDEFN::bgaA.
As shown in Fig. 4

Analysis of the Proteins Encoded by the nrc Cluster
The four open reading frames encoded by the nrc cluster ( Fig.  1) were preceded by respective Shine-Dalgarno sequences (GGA(G/A)(G/A)) located between positions Ϫ11 and Ϫ9 with respect to their ATG start codons. With the exception of NrcN, the distance between the stop codon of one gene and the ATG start codon of the next was quite small, even overlapping by one base in the case of nrcE-nrcF. However, in nrcN, this distance was apparently longer, but the presence of an in-frame GTG codon preceded by a putative Shine-Dalgarno sequence close to the end of NrcF suggests that this could be the actual translational start point of the protein.
Computer predictions suggest a cytoplasmic location for the translation products of all these genes, except for NrcE, for which an integral membrane protein containing seven putative transmembrane spanning helices is strongly suggested (see below). This putative membrane protein has no homologues in the protein data banks, unlike the significant sequence similarities of NrcD, NrcF, and NrcN and proteins involved in electron transport and NADH oxidation (Table III). Phylogenetic analysis situates these Nrc proteins at the root of the corresponding phylogenetic trees. The essential sequence information about these proteins is provided below.
NrcD-NrcD encodes an 87-amino acid-long protein (9917 Da) highly similar (86% identity) to a soluble seven-iron ferredoxin from T. thermophilus (FdTt, 79 amino acids long, accession number P03942), whose three-dimensional structure is known (26). Sequence alignment of both proteins revealed a complete conservation of the cysteine residues that coordinate the [3Fe-4S] and [4Fe-4S] clusters of FdTt. The major difference between them is the presence of a 10-amino acid C-terminal sequence in NrcD that is absent from FdTt.
NrcF-NrcF is predicted to be a 26,223-Da protein similar in sequence to the iron-sulfur subunit (B subunit) from succinatequinone oxidoreductases (SQRs) and quinol-fumarate reductases. The cysteine residues involved in the coordination of the NrcN-NrcN is homologous to type II NDH (EC 1.6.99.3) of different origins. This protein family contains a pyridine nucleotide-oxidoreductase motif consisting of a small NADH binding domain within a larger FAD-binding domain. As is frequently the case with other thermophilic proteins, NrcN (395 amino acids) is smaller than its mesophilic counterparts (430 to 465 amino acids).
NrcE-The nrcE gene encodes a 362-amino acid-long protein (40,388 Da) without homologues in the protein data banks. Topology predictions suggest the presence of seven-transmembrane ␣-helices, with a small N-terminal region facing the periplasm, and a larger (110 amino acids) C-terminal cytoplasmic domain, which contains four cysteine residues that could coordinate a redox center. Unexpectedly for a thermophilic protein, a fifth cysteine residue is present within the sequence of the first transmembrane ␣-helix.

Location of Nrc Proteins
Immunodetection of NrcD revealed lower electrophoretic mobility (ϳ30 kDa) than expected from its predicted size (ϳ10 kDa), independently of the bacterial host in which it was expressed (Fig. 5). As no signal was detected in the aerobic strain (HB27, not shown) the specificity of the antiserum for NrcD was guaranteed.
As shown in Fig. 5, NrcD was overproduced as a soluble protein in E. coli (Ec), as predicted by computer. Moreover, as it remained soluble after 30 min of incubation at 70°C, its thermostability was also assessed. By contrast, NrcD was associated with the membrane fraction of nitrate/anoxia-induced cultures of T. thermophilus (lanes M). Therefore, the existence of an interaction with a membrane component in the thermophile was concluded. As the membrane location of NrcD was not affected in nrcF::kat and nrcN::kat mutants, neither NrcF nor NrcN could be responsible for attachment of NrcD to the membrane. Conversely, NrcD was not detected in the membrane fractions of nrcE::kat mutants, and only faint signals of NrcD and of small proteins were detected in its soluble fraction. Therefore, we concluded that NrcE anchors NrcD to the membrane, and that solubilization of NrcD results in degradation of the protein.
Oxidation of NADH provided some clues concerning the location of NrcN and NrcF. Membrane fractions of nitrate/anoxia-induced cultures exhibited significant NADH oxidative activity in the presence of nitrate (Fig. 6B). In the absence of nitrate, NADH oxidation by these membrane fractions was negligible. By contrast, the presence of nitrate did not affect NADH oxidation by membrane fractions of aerobically grown cells (data not shown).
Interestingly, parallel assays revealed low NADH oxidation activity (ϳ1/10 of that of the membrane) in soluble fractions of nitrate/anoxia-induced wild type cells. This soluble activity was not nitrate-dependent and disappeared completely in nrcN::kat and nrcN::bgaA mutants. By contrast, there was a 3-fold in-crease over the wild type values in nrcE::kat and nrcF::kat mutants, concomitantly with a decrease in the membrane fraction (Fig. 6B). Therefore, we concluded that: 1) NrcN was responsible for the observed NDH activity; 2) NrcN was bound to the membrane through NrcE and NrcF; and 3) NrcN became a permissive electron donor upon solubilization.

Nitrate Reduction by Membrane Fractions of nrc Mutants
To test whether the Nrc proteins were directly involved in electron transfer from NADH to the nitrate reductase, membrane fractions from cultures of the wild type strain and the nrc mutants grown for 7 h under anoxia in TB medium containing nitrate were tested for the production of nitrite with NADH as the electron donor. Reduced methyl viologen, a direct electron donor to the ␣␤ complex of the NR (27), was used as control for the amount of active NR in the membrane fractions.
As shown in Fig. 6, NADH was a highly efficient electron donor for nitrate reduction (296 Ϯ 36 nmol of nitrite/min mg of protein). Reduction of nitrate with NADH was decreased 10fold in nrcE::kat, nrcN::kat, and nrcN::bgaA mutants. Nitrate reduction in nrcF::kat mutants decreased by a factor of five (Fig. 6A). A similar pattern was observed for the oxidation of NADH: membranes from nitrate/anoxia-induced cultures of the wild type exhibited NADH-oxidation activity (353 Ϯ 11 nmol of NADH consumed/min mg of protein) in accordance with the amount of nitrite produced (296 Ϯ 36 nmol of nitrite/min mg of protein), whereas NADH consumption was effectively undetectable in all but the nrcF::kat mutant, in which it decreased by a factor of 10. These findings imply that the Nrc proteins constitute a membrane-bound NDH required for nitrate reduction. It is also clear from these results that a quantitatively less important Nrc-independent pathway for nitrate reduction is also present on the membranes of T. thermophilus HB8 during anaerobic respiration. Interestingly, membranes isolated from cell cultures of each of the mutants grown for shorter times (2-4 h) under nitrate/anoxia presented a higher percentage of NDH and nitrate reduction residual activities (data not shown), thus showing that this Nrc-independent pathway decreases in relevance during anaerobic growth. This fact is related with the repression of the transcription of the nqo operon, encoding type I NDH, detected by RT-PCR (Fig. 6C).

Quinones Mediate Electron Transfer between the Nrc and NR Complexes
To establish whether electron transfer between NADH and nitrate was mediated by the respiratory quinones, different concentrations of inhibitors such as rotenone, diuron, and HQNO were used, and the corresponding inhibitory concentrations compared with that required to inhibit NADH oxidation by membrane fractions of cells grown aerobically (type I NDH). Similar concentrations of diuron and HQNO (0.2 M) were required for complete inhibition of the membrane-bound NDH activity from aerobically grown cells (type I NDH) and from cells grown for 7 h under nitrate/anoxia conditions (Nrc-NDH). In contrast, a small difference in rotenone sensitivity was found: at 0.1 M, NDH from aerobic membranes (type I) was completely inhibited, whereas 50% NADH oxidation was still detected in nitrate/anoxia-induced membranes (Nrc). At 0.2 M, rotenone completely inhibited NADH oxidation by aerobic membranes, but a 20% of activity was still detected with "anaerobic" membranes ( Fig. 6B). At 2 M, both type I and Nrc NDH were completely inhibited. On the other hand, nitrate reduction was insensitive to such inhibitors when reduced methyl viologen, which donates the electrons directly to the NarG subunit of the NR, was used as electron donor (data not shown).

Interactions between the Ncr Proteins
We used a bacterial two-hybrid system (23) to test for putative interactions between Nrc proteins. Fig. 7 shows the mean results of six measurements of ␤-galactosidase activity in cells transformed with pairs of plasmids derived from pT18 and pT25 that express the assayed Nrc proteins as fusions to the C-terminal (Nrc-T18) and N-terminal (T25-Nrc) domains of the adenylate cyclase from B. pertussis (23). All the nrc genes were cloned in both plasmids to avoid specific effects because of their positions at the N or C terminus of the fusion proteins. Negative controls (Ͻ100 units) were performed for each of the Nrc fusions and the complementary T18/T25 fragment fused to a leucine zipper domain (Zip). Three kinds of positive control were used: the leucine zipper fusions to the T18 and T25 fragments (ϳ1000 units), and two pairs of thermophilic proteins from T. thermophilus HB8, the NarG subunit of NR and its chaperone NarJ (ϳ280 units), and the glutamate dehydrogenase, an homo-oligomeric enzyme (ϳ400 units).
Interaction was also detected between NrcE and NrcD (ϳ600 units). Weaker but positive signals were detected between NrcE and NrcF (ϳ200 units) and between NrcE and NrcN (ϳ300 units). However, no interactions were detected between NrcD and NrcF or NrcN (Ͻ100 units). On the other hand, NrcN and NrcF (ϳ300 units) were also shown to interact to a similar extent as do NarG and NarJ. All these results were independent of the direction used in the assays (i.e. Nrc proteins expressed as N-or C-terminal parts of the fusion proteins).
As NrcE seemed to interact with all the other Nrc proteins, we decided to check whether such interactions were dependent on its C-terminal domain, which is predicted to face the cytoplasm (amino acids 264 to 362). For this, the C-terminal domain of NrcE (c-NrcE) was fused to plasmids pT25 and pT18 and used in the same assays. As shown in Fig. 7, the results for c-NrcE were almost identical to those obtained for the entire protein, indicating that the interactions between NrcE and the other Nrc proteins take place through its cytoplasmic C-terminal domain.

The Role of the nrc Operon in Nitrate Respiration
To analyze the role of nrc in vivo, the growth of mutants in nrcE (nrcE::kat), nrcF (nrcF::kat), and nrcN (nrcN::kat) under anaerobic or aerobic conditions was assayed. As illustrated in Fig. 8B, all mutants grew anaerobically with nitrate after 24 h, implying that the nrc cluster was not an absolute requirement for anaerobic growth. However, when each mutant was cocultivated with the wild type strain, the latter outgrew each of them (ϳ15% of mutants remaining in the culture) after 24 h of anaerobic growth. This growth bias was the consequence of a slower growth rate of all mutants compared with the wild type under such anaerobic conditions (Fig. 8C).
Conversely, when the competition experiments were developed under aerobiosis, no significant population bias was detected (Fig. 8A). Therefore, the nrc operon confers a selective advantage only during anaerobic growth with nitrate. units that can couple quinone reduction to proton extrusion during respiration. In the genome of the aerobic strain T. thermophilus HB27, a single type I NDH is encoded, and is currently used as a model for the structural analysis of this type of enzyme (7,28). In many other bacteria, but not in aerobic strains of T. thermophilus, the so-called alternative or type II NDH can also reduce the membrane quinones under specific growth conditions (3). These are simple, monomeric enzymes, which interact directly with the membranes and do not extrude protons (3,29). What we describe in this article is a new type of respiratory NDH that is expressed concomitantly with nitrate reductase, leading to the production of a complete respiratory chain. The presence of their respective encoding operons in a conjugative plasmid incorporated into the chromosome (11) implies that they constitute the first description of a mobile "respiratory island." Specific points reinforcing these conclusions are discussed below.
NrcN Is the Catalytic Subunit of a Multimeric NDH-Although it was not possible to produce NrcN in an active form in E. coli or to assay the activity of the recombinant protein directly (not shown), the role of NrcN as the catalytic subunit of a NDH was first inferred from the presence of a pyridine nucleotide-oxidoreductase motif in its sequence, and from its similarity to monomeric respiratory type II NDH sequences. Further confirmation of this role was provided by the analysis of the NADH-oxidative capability of the membrane and soluble fractions of cells grown with nitrate under anoxic conditions, where our data showed a clear relationship between NADH oxidation and NrcN integrity.
It is interesting to note that NADH oxidation by membrane fractions of nitrate/anoxia-induced cells of the wild type strain was completely dependent on the presence of nitrate and proportional to the amount of nitrite produced. Given that mutations in nrcN, the final gene of the operon, made NADH oxidation undetectable and severely decreased nitrite production, it was concluded that NrcN is responsible for this membraneassociated NDH activity.
The nature of NrcN as a subunit of a multimeric enzyme was deduced from the effect of nrcF and nrcE mutations. In both cases, NADH oxidation by membrane fractions was severely decreased, leading to the conclusion that NrcE and NrcF are proteins associated to NrcN that are required for its membrane-bound NDH activity.
Interestingly, an increase in nitrate-independent NADH oxidation was found in the soluble fraction of nrcE and nrcF mutants. As there was no such activity in nrcN mutants and was very weak in the wild type strain, we concluded that it was associated with a solubilized NrcN protein. Therefore, NrcN has its own NADH oxidative capability that is specific for nitrate as the electron acceptor when bound to the membrane through NrcE and NrcF, but which becomes nitrate-independent upon mechanical (in the wild type) or genetic (nrcE and nrcF mutants) solubilization.
A Tentative Model for Nrc-NDH-A tentative model for the oligomeric structure of Nrc-NDH emerges from our data (Fig.  9), by which NrcE functions as a membrane anchor for the three other proteins. This is based on the computer-predicted presence of seven-membrane spanning helices within its sequence, and on the observed solubilization of NrcD and the NrcN-associated NDH activity in the absence of NrcE. The implication of the C-terminal cytoplasmic domain of NrcE in such subunit anchoring was first deduced from the position of the kat gene cassette close to the 3Ј terminus of the gene in the nrcE::kat mutant, which should allow the synthesis of a truncated, 310-amino acid long NrcE protein. Further evidence was obtained from the results of the two-hybrid assays, which revealed specific interactions between this C-terminal domain of NrcE and NrcD, NrcF, and NrcN.
Concomitantly, several lines of evidence support the computer-predicted soluble nature of the other components of the Nrc complex. The soluble nature of NrcN was deduced from the increase in the aforementioned solubilization of NDH activity, and that of NrcD, through their detection by Western blots in nrcE mutants. The soluble nature of NrcF can be deduced indirectly. First, NrcN is solubilized in nrcF mutants, suggesting an NrcN-NrcF interaction, which was subsequently confirmed by the two-hybrid assays. Second, NrcN is also solubilized in nrcE mutants. Finally, a complex between the A and B subunits of the respiratory SQR, homologues of NrcF and NrcN, respectively, can be solubilized in E. coli (Ref. 30, see below). Consequently, the most probable explanation of our results is that NrcF forms a soluble complex with NrcN in nrcE mutants. The presence of NrcD in this soluble complex cannot be ruled out by our data, but the absence of detectable interactions with NrcD, NrcF, and NrcN, and its rapid degradation in nrcE mutants, makes this unlikely. This Nrc enzyme model (Fig. 9) is distinct from that of the multimeric (13-14 subunits) respiratory type I NDH (28) and, also from that of the monomeric alternative type II NDH (29). By contrast, it shares some sequence and structural similarities FIG. 7. Interaction between components of the nrc operon. Two-hybrid bacterial assays were developed for the proteins encoded by the nrc operon. Each Nrc component and a C-terminal domain of NrcE (c-NrcE) were expressed in E. coli BTH101 as fusion proteins to the T18 or T25 domains of the adenylate cyclase from B. pertussis from plasmids derived from pET18 and pET25, respectively. ␤-Galactosidase activity was used as a measure of reconstituted adenylate cyclase. Positive controls with mesophilic (zip/zip) and thermophilic proteins (Gdh/Gdh, and NarJ/NarG) were used. Putative interactions were sought with protein domains fused both as C-or N-terminal domains. Gray and white bars correspond to assays with genes cloned as indicated in plasmids labeled on gray or white backgrounds, respectively. Controls for negative (Ϫ/Ϫ) and positive (zip/zip, Gdh/Gdh, NarJ/NarG) interactions were also performed. The values are the means of six measurements in two independent experiments. with another respiratory enzyme, the SQR (30,31). The catalytic domain of SQR is a soluble complex of the iron-sulfur B subunit (NrcF homologue) and a large (ϳ600 amino acids) A subunit, a protein with a pyridine nucleotide-oxidoreductase motif (30). In many bacteria, including E. coli, the A-B complex is attached to the membrane by the C and D subunits, each of which has three membrane-spanning helices. In other bacteria, like Bacillus subtilis and Wollinella succinogenes, a single membrane protein containing five membrane-spanning helices substitutes for the C-D subunits. In T. thermophilus two membrane proteins encoded downstream of the A and B subunit genes probably constitute the two-component membrane anchor and quinone reduction subunit of its SQR (7). NrcE plays a similar role in our model. In this comparative context, electrons from succinate (SQR) or NADH (Nrc complex) are transferred to the flavine at the A (SQR) or NrcN subunits, and then to the four iron-sulfur clusters of the B (SQR) or NrcF subunits, to reach the NrcE membrane subunit either directly, through the C-D complex (SQR), or indirectly, through a ferredoxin (NrcD). In our model, NrcE catalyzes the reduction of a membrane quinone at the cytoplasmic face of the membrane. Reduced quinones will then travel toward the NR, where they will be oxidized at the periplasmic face of the membrane. The absence of conserved sequence motifs for quinone binding sites among quinol reductases prevents us proposing a protein domain in NrcE as responsible for such quinone reduction. However, the observed inhibition of nitrate reduction and NADH oxidation by rotenone, diuron, and HQNO in our in vitro experiments strongly supports this model, according to which quinones shuttle electrons between the Nrc and NR complexes (4).
The quinone most likely to be involved in this process is menaquinone-8, which is by far the most important respiratory quinone in T. thermophilus (32). In fact, no ubiquinone has been found in T. thermophilus (32), and homologues of UbiC and other enzymes implicated in the synthesis of ubiquinone are not encoded among the sequence of T. thermophilus HB27 (7), being also not found among the 22 kbp of the nar element so far sequenced. Therefore, the Nrc and the NDH-I complexes likely use the same quinones. In support of this conclusion is the fact that the amount of residual nitrate reduction with NADH as electron donor in nrcN::kat mutant, where only NDH-I is functional, decreases along the time when cells are grown under nitrate/anoxia conditions. Thus, it follows that the quinone reduced by NDH-I, essentially menaquinone-8 (28,32), is oxidized by the NR. Consequently, it is unlikely that the specificity of nitrate reduction by Nrc-mediated NADH oxidation was based on the use of different quinones, but more probably on the repression of NDH-I expression, as revealed by the semiquantitative RT-PCR assays of Fig. 6C.
Nrc Is Specific for Nitrate Respiration-The model discussed above clearly shows that the Nrc complex constitutes a membrane-bound NDH that is structurally different from both type I and type II NADH dehydrogenases. Its involvement in nitrate respiration is based on three arguments: 1) its dependence on nitrate and anoxia for expression; 2) the NADH-dependent in vitro assays of nitrate reduction; and 3) the lower growth rates of nrc mutants under anaerobic conditions.
The insertion of a thermostable reporter gene in nrcN allowed us to define the requirements for transcription of the Pnrc promoter. In fact, as the NR can be easily detected by the production of nitrite (8), we were able to confirm that the expression of the nrc operon in nrcN::bgaA mutants was respectively. The number of cells of each of the strains was deduced from the comparison between the number of viable cells on TB plates with kanamycin (both strains grew) and that obtained on TB plates with kanamycin plus chloramphenicol (only the wild type strain grew). Panel C, independent anaerobic growth of the wild type strain (black circles) and each of the nrc::kat mutants (white symbols) was followed for the indicated time by measuring cell viability. FIG. 9. Model for the Nrc-NDH. A model for the complete respiratory chain formed by the Nrc-NDH and the nitrate reductase is proposed. NADH reduces the FAD cofactor predicted for NrcN. Electrons are then transferred through five iron-sulfur centers (black diamonds) of NrcF (three) and NrcD (two) to the membrane subunit NrcE, where they are used to reduce menaquinone (MQ). Reduced menaquinone (MQH 2 ) is subsequently oxidized by the NR, first at its cytochrome c (NarC) component. Electrons are canalized through hemes c (NarC) and b (NarI) (white hexagons) to five consecutive iron-sulfur centers of NarH (four) and NarG (one) to reach the molybdenum cofactor (Moco), where nitrate is finally reduced. Putative proton extrusion during NADH oxidation is indicated (H ϩ ?). The interactions detected by the twohybrid assays are indicated (white double arrowheads).
parallel to the expression of the NR. Such coordinated synthesis probably depends on signal-transduction systems that respond to nitrate and anoxia, similarly to the FNR and NarL/NarX two-component systems described for E. coli (3), although neither FNR nor NarL binding sequences are found around the Pnrc and Pnar promoters. However, the genes coding for transcription regulators involved in the expression of the Pnar promoter are transferred by conjugation along the nar operon (33). Accordingly, we have identified a homologue to DNR transcription factors encoded between the nar and nrc operons, but no NarL/NarX homologues could be identified neither in the conjugative element nor in the genome of T. thermophilus HB27 (7). The DNR transcription factors constitute a subgroup of the CRP/FNR protein family that were initially identified in the context of investigation of the anaerobic respiratory system of nitrate denitrification in Pseudomomas aeruginosa (34). Proteins of this subgroup have been implicated in the detection of specific N-oxides for the induction of the corresponding terminal reductases (35). One of them, DnrE, to which the protein encoded between the nar and nrc operons shows high similarity (35% of identity), seems to be implicated in a nitrate sensory pathway (35). Future work will reveal if its thermophilic homologue also plays such a role in T. thermophilus.
Our in vitro assays demonstrated that the Nrc complex is the major NDH activity that feeds electrons to the nitrate reductase in cultures grown for 7 h under nitrate/anoxia conditions, accounting for 80 to 90% of the nitrite produced in these assays. The low relevance of aerobic type I NDH can be explained on a genetic basis, as the transcription of its encoding operon decreases dramatically under these conditions. Therefore, in nrc mutants, NADH oxidation has to be carried out by a small number of type I NDH complexes, which are not enough to provide electrons to the NR at the same rate as in the wild type where a high number of Nrc-NDH complexes does. As a consequence, the nrc mutants grow at a lower rate than the wild type. Interestingly, the nrcF mutants are less affected than nrcE or nrcN mutants in electron transference from NADH to NR, a fact that could be related to the presence of a homologue, the B subunit of its SQR (65% identity) in the genome of this bacteria that could partially substitute for NrcF (7). In conclusion, all our data support the hypothesis that the nrc operon has evolved as an efficient primary electron donor system that is specifically tuned for nitrate respiration.
The Origin of the Nrc-NDH-A final point concerns the putative origin of this Nrc-NDH. As the genus Thermus spp. belongs to one of the oldest phylogenetic branches of the Bacteria kingdom, we may speculate about an "ancient" origin for the Nrc-NDH complex. This view is supported by the comparison of NrcN and NrcF with type II NDH, and B-subunits of SQRs from different origins, respectively, which place both Nrc proteins close to the root of the corresponding phylogenetic trees (not shown). However, as a type I NDH similar to that found in "modern" bacteria is the only respiratory NDH encoded in the genome of aerobic strain HB27 of T. thermophilus (7), and considering that Nrc-NDH is actually encoded by a conjugative plasmid (8), the origin of this cluster could even be earlier than that of the genus.
In any case, the presence of the nar and nrc operons encoding a complete respiratory chain and their transcriptional regulators within a transferable genetic element provides an excel-lent example of gene functional recruitment in a respiratory island and explains why adaptation to a completely different environment does not necessarily require a lengthy evolutionary process, but simply an appropriate mating partner.