Regulation of cytochrome bd expression in the obligate aerobe Azotobacter vinelandii by CydR (Fnr). Sensitivity to oxygen, reactive oxygen species, and nitric oxide.

Azotobacter vinelandii is an obligately aerobic bacterium in which aerotolerant nitrogen fixation requires cytochrome bd. Regulation of cytochrome bd expression is achieved by CydR (an Fnr homologue), which represses transcription of the oxidase genes cydAB. cydAB mRNA was mapped by primer extension; the transcriptional start site was determined, and putative -10 and -35 regions were deduced. Two "CydR boxes," one at the +1 site and one upstream of the -35 region, were identified. Transcriptionally inactive, purified CydR was converted, by adding NifS, cysteine, and Fe(2+), into an active form possessing acid-labile sulfide and spectra suggesting a [4Fe-4S](2+) cluster. Reconstituted CydR specifically bound both CydR boxes cooperatively, with higher affinity for the nearer consensus +1 site. Low concentrations of O(2) or NO ([O(2)]/[[CydR] or [NO]/[CydR] = 0.1-0. 6) elicited loss of the 420 nm absorbance attributed to the [4Fe-4S](2+) cluster, formation of a 315 nm species, and loss of ability to retard DNA migration. Retardation by reconstituted CydR was enhanced by superoxide dismutase and/or catalase, suggesting a role for reactive oxygen species in CydR inactivation. The role of CydR in regulating cydAB expression in the supposedly anoxic cytoplasm of A. vinelandii and similarities to cydAB regulation by Fnr in Escherichia coli are discussed.

Azotobacter vinelandii is an obligate aerobe that fixes nitrogen under a wide range of oxygen concentrations, even at air saturation (1). Nitrogen fixation is an energy-demanding process that consumes 16 mol of ATP to convert 1 mol of N 2 to 2 mol of NH 3 (2). This energy requirement can be met only by aerobic respiration, yet paradoxically, nitrogenase is notoriously sensitive to oxygen damage (3). One way to avoid this damage is "respiratory protection," i.e. the rapid utilization of oxygen to achieve subinhibitory levels of oxygen, thus allowing the coexistence in a cell of aerobic respiration and nitrogenase activity (1,4). To scavenge traces of oxygen yet consume excess oxygen, A. vinelandii has a branched respiratory chain with at least two routes of electron transport to oxygen (1). One branch is terminated by an oxidase closely resembling the cytochrome bd-type oxidase of Escherichia coli and certain other bacteria (5,6). The oxidase comprises a low-spin cytochrome b 558 and two ligand-binding hemes, cytochromes d and b 595 (previously called cytochrome a 1 (8)) (7). Another respiratory branch is terminated by an oxidase of the heme-copper superfamily (9), which is probably the oxidase referred to previously as cytochrome o (8,10). However, a gene fragment from A. vinelandii has been independently sequenced and shown to resemble fixN or ccoN encoding a cb-type cytochrome c oxidase (11). It is not clear whether cytochrome c oxidase and cytochrome o are distinct oxidases.
Direct evidence for the essential role of cytochrome bd in respiratory protection of nitrogenase and the first molecular genetic analysis of respiratory metabolism in A. vinelandii were provided by Kelly et al. (12), who obtained mutants with transposon insertions in and around that region of the genome homologous to the E. coli cydAB genes. One class of mutants had insertions within the cydAB operon, had no spectroscopically detectable cytochrome bd, and significantly, could not fix nitrogen in air. Sequencing of the entire cydAB operon (13) revealed striking similarities to the E. coli cytochrome bd-type oxidase, in accord with spectral studies (5)(6)(7)(8)(9)14). Purification of A. vinelandii cytochrome bd (15,16) confirmed similarities in subunit composition, complement of redox centers, and reaction mechanisms in these bacteria. However, a remarkable difference between the E. coli and A. vinelandii oxidases is that the former is synthesized maximally microaerobically (17), whereas synthesis of the A. vinelandii oxidase increases with oxygen supply (1,18). Furthermore, and consistent with the patterns of regulation, cytochrome bd in A. vinelandii has a surprisingly low affinity for oxygen (apparent K m ϳ 4.5 M) (14), unlike the oxidase in E. coli, which has the highest affinity yet recorded (K m as low as 5 nM) for a terminal oxidase (19).
An explanation of the different responses in these bacteria of oxidase expression to oxygen supply is now beginning to emerge. In E. coli, regulation of cytochrome bd expression is complex and coordinated by the ArcAB two-component system and by Fnr, major global regulators of the aerobic/anaerobic switch (20,21). ArcA activates cydAB gene expression at lowoxygen tensions (22,23). As oxygen tension falls farther, Fnr is activated and represses cydAB expression (24). Recent work has identified two cydAB promoters, but the roles played by Fnr and ArcA have not been fully elucidated. Lynch and Lin (25) found three sites for ArcA, one of which (site III) was located downstream of the previously identified cydAB promoter P1 (referred to hereafter as P1). A second promoter was found downstream of this site, but could not be detected by analysis of RNA extracted from aerobically grown cells, suggesting that cydAB P1 is used preferentially under such con-ditions. It was suggested that ArcA-P (i.e. the active phosphorylated form) bound at site III activates cydAB anoxically when Fnr prevents transcription from P1 (25). Subsequently, Cotter et al. (26) demonstrated that a single site for ArcA-P upstream of P1 was sufficient for activation of cydAB expression. Significantly, two sites for Fnr were found, one at the start of cydAB transcription at P1 and another centered 53.5 bp 1 upstream of the ϩ1 site of P1. Thus, collectively, ArcA and Fnr afford maximal cydAB expression in E. coli growing in microaerobic environments, consistent with the finding that this quinol oxidase has a remarkably high affinity for oxygen (19).
A simpler pattern of cydAB expression in response to oxygen availability is observed at the physiological level in A. vinelandii. In this strict aerobe, cydAB transcription is up-regulated as the oxygen tension, and thus danger of nitrogenase damage, increases (1,3,4,14). Mutagenesis in the region of the A. vinelandii genome upstream of the cydAB genes revealed a region in which insertions resulted in marked overproduction of the cytochrome bd complex (12). These mutants failed to grow in a microaerobic atmosphere (1.5% O 2 ) on defined medium either containing (BSN medium) or lacking (BS medium) a supply of fixed nitrogen in the form of ammonium ions (12). This region, separated from the downstream cydAB operon by ϳ1 kilobase pair, was sequenced by Wu et al. (27) and revealed a gene whose deduced product is highly similar to Fnr. The gene was named cydR (a) to indicate clearly its role in the regulation of the cyd operon (this being the only operon in A. vinelandii so far demonstrated to be CydR-regulated) and (b) because the term fnr appears inappropriate since neither fumarate nor nitrate respiration occurs in A. vinelandii (27). It was postulated that CydR is a repressor of cydAB transcription (27).
Fnr senses anaerobiosis via an oxygen-labile [4Fe-4S] 2ϩ cluster that promotes dimerization of the protein and enhances site-specific DNA binding (20, 21, 28 -31). Homologues of Fnr control a variety of physiological functions in a diverse range of phylogenetically distinct prokaryotes (21). They are characterized by the presence of four essential cysteine residues that act as ligands for the [4Fe-4S] 2ϩ cluster and the amino acid sequence EXXSR in the DNA-binding region, which confers specificity for the Fnr "box" or -binding site with the consensus sequence TT-GAT . . . . ATCAA. Both of these features are conserved in CydR.
In this paper, purified CydR is shown to be an oxygenresponsive, DNA-binding, Fnr-like protein that is exceptionally oxygen-labile. Furthermore, it is shown for the first time that this member of the Fnr family is also responsive to nitric oxide, with wide implications for regulation of gene expression by Fnr-like proteins and probably several other Fe-S proteins.

EXPERIMENTAL PROCEDURES
Strains and Plasmids-The bacterial strains and plasmids used in this study are listed in Table I. E. coli and A. vinelandii cultures were grown at 37 and 30°C, respectively. The growth medium used for E. coli was 2ϫ YT broth (32), and that for A. vinelandii was BSN (12).
General DNA Manipulations-Standard procedures were used for DNA isolation and manipulation (32). Enzymes for DNA manipulation were obtained from Promega, Stratagene and MBI Fermentas. DNA end labeling was done with [␣-32 P]dCTP (Amersham Pharmacia Biotech) and Klenow fragment of DNA polymerase for use in the DNase I footprinting and gel retardation assays (see below). DNA sequencing was performed using an ABI 373A sequencer (Applied Biosystems).
The insert in pRKP1028 is in the opposite orientation from that in the other three. DNA fragments containing wild-type and mutated CydRbinding sites were isolated from plasmids pRKP1025, pRKP1026, and pRKP1029 after cutting with BamHI-ApaI and from pRKP1028 after cutting with BamHI-SacI.
Mapping of the cydAB Transcript-The cydAB transcript was mapped by extracting total RNA by the hot phenol method (33) from a culture of A. vinelandii strain MK8. This strain carries a Tn5-B20 mutation in cydR and consequently overproduces cytochrome bd (12,27); it was grown under conditions of high aeration (50 ml of culture in a 1-liter conical flask shaken at 200 rpm) for 7 h before harvest. Primer extension was performed according to Yagü e et al. (34). The primer used was purified by running a 15% sequencing gel, and the band was visualized using a "shadowing technique" against the background of a thin-layer chromatographic plate under UV light and isolated from the gel using standard procedures (32). Ten ng of 5Ј-end-labeled 34-mer oligonucleotide RP141 (5Ј-AAGATTGGGATGGTTTAGACATGTGGGC- Cloned promoter of cydAB 27 pMK41 ϩ1 CydR box mutated This work pMK435 Ϫ50.5 CydR box mutated This work pMK4351 Both CydR boxes mutated This work pRKP1025 PCR a fragment generated by primers RP38 and RP39 using pMK4 as template and cloned into pGEM-T-Easy This work pRKP1026 As above, but pMK41 used as template This work pRKP1028 As above, but pMK435 used as template and insert reversed This work pRKP1029 As pRKP1025, but pMK4351 used as template This work pRKP1082 NcoI-HindIII fragment containing cydR cloned in pGEX-KG This work a PCR, polymerase chain reaction. AGCTCCAA-3Ј) were mixed with 5 g of total RNA in 30 l of primer extension buffer (50 mM Tris-HCl (pH 8.3), 50 mM KCl, 10 mM MgCl 2 , 10 mM dithiothreitol, 1 mM each dNTP, and 0.5 mM spermidine), heated at 75°C for 3 min, and then cooled at room temperature for 10 min. Twenty units of Moloney murine leukemia virus reverse transcriptase (Promega, RNase H Ϫ ) containing sodium pyrophosphate (1.8 l of a 125 mM solution) were added to the annealed primer/RNA. The mixture was incubated at 42°C for 30 min, and the nucleic acids were precipitated with ethanol. The DNA was resuspended in formamide loading dye and resolved on a denaturing 6% polyacrylamide gel. The sizes of the primer-extended products were determined by running a known sequence ladder (M13mp18 DNA sequenced with the M13 Ϫ40 primer; U. S. Biochemical Corp.) in parallel.
Subcloning of cydR in the Gene Fusion Vector pGEX-KG-To clone cydR into the pGEX-KG vector, a NcoI site was introduced at the ATG initiation codon of cydR on plasmid pGW33 (Table I) using site-directed mutagenesis directed by oligonucleotides RP34 (5Ј-CAGAGAAAGTT-GCCCATGGCGGATAAATCC-3Ј) and RP35 (complementary to RP34). These primers additionally result in changes in the first two amino acid residues of CydR from Met-Ser to Met-Ala. The NcoI-HindIII fragment from pGW33 was then cloned into pGEX-KG to give plasmid pRKP1082.
Purification and Reconstitution of CydR-CydR was purified from French press extracts of aerobic cultures of E. coli strain RKP3363 (DH5␣ containing plasmid pRKP1082) grown in 2ϫ YT broth (32) plus ampicillin (200 mg/liter). Three h after inoculation with an overnight culture (1.6% of the final volume), the expression of GST-CydR was induced with isopropyl-␤-D-thiogalactopyranoside (0.1 mM final concentration) during the exponential phase of growth. Cells were harvested after a further 4 h of growth. Glutathione-Sepharose 4B (Amersham Pharmacia Biotech) was used to adsorb the GST-CydR fusion protein according to the manufacturer's instructions. The column was washed with Tris-buffered saline (25 mM Tris-HCl and 138 mM NaCl (pH 7.5)) containing 0.1% Triton X-100 (10ϫ the resin bed volume), and CydR was cleaved from GST by thrombin protease by incubation for 30 min at room temperature. For reconstitution of CydR (30), NifS (purified from A. vinelandii) and dithiothreitol (2.5 mM final concentration) were added to the purified CydR protein, and the mixture was flushed with O 2 -free nitrogen gas for 1 h. Cysteine (1 mM final concentration) and (NH 4 ) 2 SO 4 ⅐FeSO 4 ⅐6H 2 O (10 mol of iron/mol of CydR monomer) were added to the mixture in an Mk3 Anaerobic Work station (Don Whitley Scientific Ltd, Shipley, England). The reconstitution was monitored spectrophotometrically. Upon completion, CydR was separated from excess cysteine, iron, dithiothreitol, and residual nucleic acid (see below) by passing through a column packed with Toyopearl ether-650M hydrophobic interaction resin, which was first washed with 20 mM Tris-HCl (pH 7.0) (low-salt buffer; 5-10ϫ the resin bed volume) and then equilibrated with 20 mM Tris-HCl (pH 7.0) plus 1.7 M (NH 4 ) 2 SO 4 (high-salt buffer). The reconstituted CydR protein containing 1.7 M (NH 4 ) 2 SO 4 was loaded on the column, which was washed with high-salt buffer (5ϫ resin bed volume), and finally eluted with low-salt buffer (200 l at a time). Acid-labile sulfur in denatured CydR samples was determined by the method of Beinert (35) in the anaerobic cabinet. Protein concentrations were estimated by the Bio-Rad dye binding procedure with bovine serum albumin as the standard.
Spectroscopy and Inactivation of CydR-A Beckman DU Series 600 spectrophotometer or a Perkin-Elmer spectrophotometer with Winlab software was used for optical spectroscopy. CydR samples were transferred from the anaerobic cabinet in sealed cuvettes. Air-saturated H 2 O (ϳ230 M O 2 at 23°C) and freshly prepared NO solution (ϳ1.9 mM in H 2 O at room temperature) prepared as described by Poole et al. (36) were added to the sample by injection in the anaerobic cabinet.
Gel Retardation Assays-For determining CydR binding to the cydAB promoter, reconstituted CydR protein was diluted in 10 mM Tris-HCl (pH 7.0) and incubated anaerobically for 10 min with labeled promoter DNA (20 -50 counts/min/l), 150 ng of salmon sperm DNA, 2.5 g of bovine serum albumin, 5 mM dithiothreitol, and band-shift buffer (20 mM Tris-HCl (pH 7.5), 5% glycerol, and 100 mM KCl) in a final volume of 5 l. Sensitivity of CydR to O 2 was studied by preincubation of 10 M protein (monomer) for 20 min in 10 mM Tris-HCl (pH 7.0) in the presence of mixtures of anoxic water and air-saturated water to give final concentrations of up to 60 M O 2 . Sensitivity of CydR to NO was studied similarly, but in anoxic aqueous solutions of NO or NaNO 2 to give final concentrations of up to 60 M NO or NaNO 2 . Superoxide dismutase (5 units; Sigma) and/or catalase (5 units; Sigma) was added to these reactions where indicated. O 2 concentrations quoted refer to those in the preincubation mixture, not to the incubation with DNA. To allow CydR⅐DNA complex formation, 1-l samples of the reactions were then incubated for 10 min as described above in a final volume of 5 l.
The loading of samples was performed at a voltage of 20 V after pre-running at 120 V for 5 min. Promoter DNA and purified CydRprotein interactions were visualized on 5% polyacrylamide gels (19:1 acrylamide/bisacrylamide) buffered with 25 mM Tris and 250 mM glycine (pH 8.3). The gel was run for ϳ1 h at 60 mA (for two gels). Quantitation of shifted and non-shifted bands was performed by determining density expressed as counts/mm 2 using a Bio-Rad Model GS-S25 Molecular Imager® system.
DNase I Footprinting-DNase I footprinting using purified CydR and DNA restriction fragments containing the cydAB promoter was performed according to Green et al. (28) and Rhodes (37), except that reactions were not extracted with phenol/chloroform. The details of the binding conditions are presented under "Results." The C lane was included as a sequence reference to determine the location of the binding sites.

RESULTS
Mapping the Transcriptional Start Site by Primer Extension-It has been reported (38) that the cydA and cydB genes are cotranscribed and that the transcriptional start site is ϳ275 bp upstream of the ATG initiation codon of cydA. An oligonucleotide (RP141) starting 188 bp upstream of the cydAB ATG initiation codon was therefore used for primer extension experiments. Fig. 1 shows that the transcriptional start site is actually 268 -269 bp upstream of the ATG codon of cydA.
Putative Ϫ10 and Ϫ35 regions were sought. The sequence TTTATT (labeled "Ϫ10" in Fig. 4; see later) has four matches with the Ϫ10 motif (TATAAT) characteristic of E. coli promoters transcribed by the major 70 complex of the RNA polymerase. On the basis of ribonuclease protection analysis, Moshiri et al. (38) identified the sequence GTAAAT as the probable Ϫ10 site, having three matches with the consensus sequence. These workers also tentatively identified a Ϫ35 sequence (TGGTCA) that also has four matches with the E. coli consensus sequence TTGACA. However, as shown in Fig. 4, the gap between our Ϫ10 sequence and the putative Ϫ35 sequence is 24 bp, considerably exceeding the usual distance in E. coli, for example, of 17 bp.
Two sequences similar to the E. coli Fnr boxes (TTGAT . . . . ATCAA) were located in the promoter region of cydAB (see Fig.  4 below), which we designate here as the ϩ1 CydR box (TT-GAC . . . . ATCAA) and the Ϫ50.5 CydR box (TTGAC. . . . GT-CAA), the latter being centered 50.5 bp away from the more upstream G marked in Fig. 4. The ϩ1 site has one mismatch compared with the Fnr consensus sequence (39), whereas the Ϫ50.5 site has two mismatches, one in each half-site. From the beginning of one CydR box to that of the other, the distance is exactly five turns of the helix, which suggests that CydR molecules that bind to these two CydR boxes are located on the same face of the DNA and may therefore interact with each other. We obtained no evidence for a second promoter for the cydAB operon in A. vinelandii.
Overexpression and Purification of CydR-Using a GST fusion vector, CydR was first expressed as a GST fusion protein (GST-CydR) at levels up to 15% of soluble cell protein (Fig. 2,  lane 1). Purification was then achieved by glutathione-Sepharose 4B affinity chromatography. Most of the fusion protein in the supernatant was adsorbed by this column in 30 min (Fig. 2,  lane 2). After washing with Tris-buffered saline, GST-CydR was purified to near-homogeneity (Fig. 2, lane 3). CydR was then completely cleaved from GST by thrombin protease in ϳ30 min at room temperature (Fig. 2, lane 4). Typically, growth in the 30-liter airlift fermentor produced ϳ120 g of cells (wet weight), from which ϳ100 mg of CydR could be obtained. The cleaved CydR protein contains 15 extra amino acids at its N terminus; these do not prevent demonstration of the aerobic/anaerobic transcription switch in vitro in the case of E. coli Fnr (30).
Reconstitution of the Fe-S Cluster in CydR-Since apo-Fnr can be reconstituted to form an active protein containing two [4Fe-4S] 2ϩ clusters/dimer (29,30), the same procedure was used in an attempt to reconstitute CydR. Reconstitution of CydR could be achieved in ϳ3 h at room temperature, but since we found that CydR is very sensitive to high temperature and denatured instantly at 37°C, CydR was routinely reconstituted at 4°C overnight. Purified Fnr protein is reported to be contaminated with some nucleic acid (30), and the high absorbance of the CydR preparation at 260 nm (data not shown) suggested the same. However, after the reconstituted CydR protein was purified through a column packed with Toyopearl ether-650M hydrophobic interaction resin, it was a straw-brown color, and the absorbance at 260 nm was greatly reduced, indicating that contaminating nucleic acids that might inhibit promoter binding were largely eliminated.
Spectra recorded during a typical reconstitution experiment are shown in Fig. 3. The signal at 420 nm attributed to reconstitution of an Fe-S cluster increased in intensity with time. The model compound [Fe 4 S 4 (S-Et) 4 ] 2Ϫ has an ⑀ 280 of 17,200 M Ϫ1 cm Ϫ1 and an A 420 /A 280 ratio of ϳ0.7 in methylformamide, which shows only small variations with changes in solvent or thiol ligand (30). Assuming the presence of one [4Fe-4S] 2ϩ cluster/monomer in the reconstituted CydR protein, as in Fnr, the absorbance of the 420 nm species corresponds to a concentration of [4Fe-4S] 2ϩ clusters of ϳ60 M, i.e. ϳ40% of the anticipated concentration of protein after reconstitution. However, the final spectrum in Fig. 3, taken at 2.5 h, does not reveal the full intensity of the signal, but analysis was frustrated after longer incubations by the formation of a fine black precipitate.
The ratio A 420 /A 280 can also serve as a useful index of the iron-sulfur cluster content of a protein. The CydR protein contains 14 phenylalanines (⑀ 257 ϭ 220 M Ϫ1 cm Ϫ1 ) and one tyrosine (⑀ 274 ϭ 1440 M Ϫ1 cm Ϫ1 ), which contribute to the absorbance at 280 nm, and no tryptophan. Based on studies with Fnr (11 phenylalanines and five tyrosines) (30), we estimate that ⑀ 280 for CydR, with a much lower tyrosine content than Fnr, is on the order of 3000. Hence, it can be calculated that the A 420 /A 280 ratio for CydR containing one [4Fe-4S] 2ϩ cluster/monomer should be ϳ0.62. This value is close to that calculated for Fnr (0.56) because the Fe-S cluster makes a 4 -8-fold larger contribution to the absorbance at 280 nm than does the protein. The highest ratio determined experimentally for CydR was ϳ0.5, although as for Fnr, determination was frustrated by the persistence of absorbance at 260 nm due to nucleic acids, which artifactually raises the protein assay, and by slow precipitation of material after reconstitution (see below), which contributes a turbidity "base line" to the uncorrected absorbance spectrum. Nevertheless, these values are 1.6-fold higher than the ratio measured for Fnr (30).
Interaction of CydR with Wild-type and Mutant Target Sequences-In Fnr, the paradigm for such protein-DNA binding site studies, Glu-209, Ser-212, and Arg-213 in the second helix of the helix-turn-helix motif have a significant role in the recognition of an Fnr box (39). These amino acid residues are conserved in CydR, suggesting that CydR and Fnr will recognize very similar DNA-binding sequences. To determine whether the putative CydR boxes are actually recognized by CydR, mutations were made (Fig. 4) in which the central G, which interacts with Glu-209, was mutated to A, and the corresponding C, in the second half of the box, was changed to T. The interaction of CydR with its target sequences was then analyzed by gel retardation. Fig. 5 shows that the specific retardation was seen only with the wild-type sequence at CydR concentrations as low as 0.5 ϫ 10 Ϫ7 and 0.5 ϫ 10 Ϫ6 M (monomer). Little retardation could be detected when the ϩ1 CydR box, the Ϫ50.5 box, or both were mutated even at a CydR monomer concentration of 0.5 ϫ 10 Ϫ5 M. The unreconstituted CydR protein (up to 0.5 ϫ 10 Ϫ5 M) did not retard the same DNA fragment (data not shown).
DNase I footprint analysis confirmed that CydR protects the CydR boxes in the ϩ1 and Ϫ50.5 regions (Fig. 6). The concentration necessary for protecting both the CydR boxes (pMK4) was ϳ2 ϫ 10 Ϫ7 M (monomer). DNA protection was lost when the G or C nucleotides were changed in the two half-sites of the CydR boxes (pMK4351). Interestingly, mutation of the ϩ1 CydR box modified CydR protection at the Ϫ50.5 region. In pMK41, a higher CydR concentration (2 ϫ 10 Ϫ6 M (monomer)) was required to protect the Ϫ50.5 CydR box. In contrast, mutation of the Ϫ50.5 site (pMK435) did not change the CydR concentration required for protecting the ϩ1 site. These results allow us to distinguish between a primary and a secondary binding site, with CydR showing a higher affinity for the ϩ1 CydR box.
Interaction of CydR with Oxygen-In Fnr, reaction with oxygen of the [4Fe-4S] 2ϩ cluster in the reconstituted protein produces a non-DNA-binding, transcriptionally inactive form (31). Since A. vinelandii is an obligate aerobe that must maintain an effectively anoxic cytoplasm for nitrogenase function, it was of interest to compare the sensitivity of CydR and Fnr to oxygen. Sequential additions of O 2 (as O 2 -saturated buffer) were made to the reconstituted CydR protein in a volume of 0.4 ml in a sealed cuvette and monitored by UV-visible electronic spectroscopy (Fig. 7A). The same samples were subsequently used for gel retardation experiments after suitable dilution. Slight precipitation of material during the titration caused absorbance at all wavelengths to rise, but particularly at lower wavelengths, consistent with the increased turbidity. Nevertheless, an obvious change superimposed upon the turbidity signals was the increase of a peak centered at ϳ315 nm and a decrease of the peak at 420 nm tentatively attributed to a [4Fe-4S] 2ϩ cluster. To minimize the effect of the base-line shift, changes in absorbance at 315 nm were expressed as the absorbance difference between 315 and 295 nm, and the changes at 420 nm were expressed as the difference between 420 and 500 nm. Fig. 7A shows that nearly half of the absorbance change at 420 nm was lost on adding oxygen to give an [O 2 ]/ [CydR] ratio of ϳ0.03. Adding oxygen to give an [O 2 ]/[CydR] ratio of ϳ0.1 almost eliminated the A 420 signal. The absorbance at 315 nm increased as the preparation was exposed to small amounts of air, but was not seen when reconstituted CydR was fully exposed to air. The maximal absorbance was reached when the [O 2 ]/[CydR] ratio was ϳ0.1.
Interaction of CydR with NO and Reactive Oxygen Intermediates-Redox-sensitive metal clusters of proteins are sensitive not only to O 2 , but may also be sensitive to degradation by radicals of biological significance such as NO (40). In light of this and considering that A. vinelandii, a soil bacterium, is likely to encounter NO produced by denitrifying bacteria, we studied the effects of NO on CydR. Successive additions of NO (as a strictly anoxic gas-saturated solution) and monitoring of spectral changes (Fig. 7B) showed that the effects of NO were very similar to those of oxygen, both with respect to the nature of the spectral changes and the [NO]/[CydR] ratio required for abolition of the 420 nm signal (Fig. 7B).

DISCUSSION
An absolute requirement for aerotolerant nitrogen fixation in A. vinelandii appears to be synthesis of the quinol oxidase cytochrome bd (1,12). We have previously shown genetically that transcription of the cydAB operon, encoding the two subunits of cytochrome bd, is repressed by CydR (27) and that mutation of CydR causes elevation of oxidase synthesis (12,27). Cytochrome bd levels, as well as cytochrome bd-specific mRNA, increase in a wild-type strain when the oxygen concentration increases under non-nitrogen-fixing conditions (18,27). The converse is true in cydR mutants, i.e. the cytochrome bd concentration increases sharply when the oxygen concentration decreases. The trends are the same in cells grown under nitrogen-fixing conditions. 2 The transcriptional start site of cydAB was mapped in this work to 268 -269 bp upstream of the cydAB ATG translational initiation codon by primer extension. This method is probably more precise than the ribonuclease protection assay (38), which placed the transcriptional start site at 275-277 bp upstream of the cydAB translational initiation site. We and Moshiri et al. (38) have not obtained any evidence for a second promoter for 2 S. E. Edwards, S. Hill, and R. K. Poole, unpublished results. the cydAB operon in A. vinelandii. The transcriptional start site is the same under both nitrogen-fixing and non-nitrogenfixing conditions (38). Putative Ϫ10 and Ϫ35 regions were identified based on similarity to the promoters recognized by E. coli 70 . However, A. vinelandii RNA polymerase may recognize different promoter sequences, and further promoter analysis is required. Moshiri et al. (13) showed that the cloned A. vinelandii cydAB genes in E. coli could reconstitute cyanideinsensitive respiratory chain activity from NADH to O 2 , but not succinate-or lactate-dependent respiration, and that cytochrome d was detectable spectroscopically. In another study, however, the cloned cydAB genes did not complement an E. coli cydAB mutant for growth on Zn 2ϩ -and azide-containing medium, and no cytochrome d was detected spectroscopically (41). Moshiri et al. (38) demonstrated that the cydAB genes are up-regulated under nitrogen-fixing conditions in a 54 -dependent manner. However, the promoter of cydAB does not resemble the typical 54 -dependent promoter, but is more similar to the E. coli 70 -dependent promoter. It is possible that 54 regulates the expression of cytochrome bd indirectly.
The CydR protein has now been purified by glutathione-Sepharose 4B affinity column chromatography to near-homogeneity. The aerobically purified CydR protein can be reconstituted into an active form, as can the aerobically purified E. coli Fnr protein (29,30), and only in this state binds to target sequences in the cydAB promoter identifiable by footprinting studies and similarity to Fnr boxes. Two CydR-protected regions were seen at this promoter, as was also revealed in E. coli (26) by DNase I footprinting studies of Fnr binding to the cydAB regulatory region. One region of 25 bp extends from positions Ϫ17 to ϩ7 and thus overlaps the transcriptional start site (ϩ1). Another region of 24 bp extends from positions Ϫ61 to Ϫ38 and is centered at position Ϫ50.5 relative to the same start site. This arrangement of CydR (Fnr)-binding sites is very similar to that of the cydAB promoter in E. coli, in which the Fnr sites extend from positions Ϫ13 to ϩ10 and from positions Ϫ67 to Ϫ45, respectively, with reference to the start of P1 transcription. Only one putative Fnr site was considered by Lynch and Lin (25). Thus, in both A. vinelandii and E. coli, an Fnr-like protein acts directly to repress cydAB gene expression. In E. coli, however, both sites are bound by Fnr with similar affinity, and the sites become occupied simultaneously as the protein concentration is increased (26). In E. coli, the upstream site centered at position Ϫ53.5 is identical to the consensus sequence, whereas the site at the start of cydAB transcription has a single mismatch. In A. vinelandii, however, CydR binds both sites, but with higher affinity for the ϩ1 CydR box. This may reflect the closer match of the ϩ1 site to the Fnr consensus sequence. "Anaerobic" cydAB repression in A. vinelandii may involve the binding of two pairs of CydR monomers over the ϩ1 and Ϫ50.5 regions, which prevents essential RNA polymerase-DNA contacts. CydR binding to the primary (high-affinity) ϩ1 site could cooperatively help CydR binding to the secondary (low-affinity) Ϫ50.5 site.
Thus, even in this obligately aerobic bacterium, cydAB transcription is still regulated in response to O 2 . However, whereas the absorbance loss at 420 nm of the Fnr protein in E. coli requires an [O 2 ]/[Fnr] ratio of ϳ1 (31), a ratio of only ϳ0.1 is sufficient to cause loss of the distinctive 420 nm band of the Fe-S cluster in CydR. Furthermore, an [O 2 ]/[CydR] ratio of ϳ0.6 significantly prevents the retardation by CydR of its target DNA, whereas a ratio of 3 is needed to abolish the retardation by Fnr of its target DNA (31). This observation is perhaps not surprising given the requirement in this organism that the cytoplasmic oxygen tension should be maintained at very low levels, despite ready penetration of oxygen through the cytoplasmic membrane from an external growth environment that may be air-saturated. Although intracellular oxygen levels have not been measured in A. vinelandii (or any other bacterium), CydR appears to be a highly sensitive monitor of cytoplasmic oxygen, as anticipated for continued operation of nitrogenase under highly aerobic growth conditions. We envisage that, during growth under microaerobic conditions, intracellular oxygen concentrations are sufficiently low to allow nitrogenase function and that CydR would be active, repressing cydAB expression. Under conditions of stress imposed by high oxygen, the repressed levels of cytochrome bd may not maintain the essentially anoxic state of the cytoplasm that is required for nitrogenase and CydR will be inactivated; this in turn derepresses cytochrome bd synthesis, which provides respiratory protection.
Several lines of evidence indicate that cysteine-rich motifs of metal-binding proteins and redox-sensitive metal clusters of metalloproteins are natural biosensors not only of O 2 and Fe (40), but also of NO (42). Fe-S-containing proteins like dehydratases have long been known as targets of O 2 Ϫ and H 2 O 2 (43) and are also targets of NO. NO forms complexes with Fe-S clusters in model compounds (44), and mitochondrial Fe-S enzymes are inhibited by NO (45). Aconitase is especially sensitive to NO, but it has recently been proposed that the peroxynitrite anion (ONOO Ϫ ), formed in the reaction of NO with O 2 Ϫ , is the inactivating species (46,47). Regulatory Fe-S-containing proteins like SoxR and mammalian iron-responsive element-  1 (lanes 13 and 14), 0.6:1 (lanes 15 and 16), and 6:1 (lanes 17 and 18). Concentrations of CydR are given for the monomer.
binding protein 1 have also been shown to be NO-sensitive (40,48,49), but it was not known whether Fnr is sensitive to NO or insensitive, as is the molybdenum metalloenzyme xanthine oxidase (50). We now show for the first time that a member of the Fnr family is inactivated by NO as well as by oxygen. The mechanism of this inactivation needs further study. The physiological function of the effects of NO on CydR, if any, are unclear. However, although A. vinelandii is not itself a denitrifying bacterium, it inhabits environments where other bacteria produce NO as an intermediate in this pathway. NO may derepress cytochrome bd so that nitrogenase is protected by respiration and able to exploit the end product of denitrification, namely dinitrogen.
The peak observed in absorbance spectra at 315 nm formed after adding O 2 or NO may be a breakdown product of the [4Fe-4S] 2ϩ cluster or reflect the presence of substoichiometric iron levels in the protein (51). Several studies of Fnr including recent Mössbauer spectroscopy (52) show that the loss of the 420 nm-absorbing form of Fnr is due to conversion of the [4Fe-4S] 2ϩ cluster to a [2Fe-2S] 2ϩ cluster. Only stoichiometric amounts of O 2 are needed for this inactivation, whereas ferricyanide is required in considerable excess (31), suggesting that Fnr is a true O 2 sensor. The greater sensitivity of CydR to oxygen than of Fnr suggests that even substoichiometric amounts of O 2 are adequate for cluster inactivation. A plausible mechanism that accounts for the oxygen sensitivity of the [4Fe-4S] 2ϩ clusters of Fnr-like proteins and the biphasic nature of the response (40)  The low levels of cytochrome bd at low-oxygen tensions in wild-type cells are presumably due to repression by CydR, but levels of cytochrome bd under low aeration in cydR mutants are significantly higher than those in both wild-type (cydR ϩ ) and the cydR mutants under high aeration (18,27). This suggests that there may be another regulator that represses the expression of cytochrome bd under high aeration. There is no evidence for or against an ArcA/ArcB system in A. vinelandii. Therefore, at present, the possible involvement of ArcA-P in CydR (Fnr)dependent repression of cydAB is unknown, but unlike E. coli, mechanisms for maximizing cydAB expression microaerobically are not evident and may be unnecessary.
So far, only the cydAB operon has been unequivocally shown to be regulated by CydR. However, increased activity of NADH: ubiquinone oxidoreductase that is insensitive to capsaicin (i.e. NADH dehydrogenase II) is co-induced with cytochrome bd in a cydR mutant (53), and the O 2 -sensitive phenotype of a nifU mutant is corrected by the introduction of a cydR mutation (54). Furthermore, an unexplained phenotype of cydR mutants is their inability to grow under conditions of low aeration (12,27). A likely explanation is that one or more genes required for microaerobic growth are CydR-regulated. Comparative analysis of the proteome of wild-type A. vinelandii and a cydR mutant is now under way to determine the extent of gene regulation accomplished by CydR.