Pseudomonas aeruginosa RoxR, a response regulator related to Rhodobacter sphaeroides PrrA, activates expression of the cyanide-insensitive terminal oxidase


  • James C. Comolli,

    1. Department of Bacteriology, University of Wisconsin – Madison, 1550 Linden Drive, Madison, WI 53706, USA.
    Search for more papers by this author
  • Timothy J. Donohue

    Corresponding author
    1. Department of Bacteriology, University of Wisconsin – Madison, 1550 Linden Drive, Madison, WI 53706, USA.
    • *For correspondence. E-mail; Tel. (+1) 608 262 4663; Fax (+1) 608 262 9865.

    Search for more papers by this author


The facultative anaerobe Pseudomonas aeruginosa has multiple aerobic electron transport pathways, one of which is terminated by a cyanide-insensitive oxidase (CIO). This study characterizes a P. aeruginosa two-component system that regulates CIO production. The response regulator of this system (RoxR) has significant amino acid sequence similarity to PrrA of Rhodobacter sphaeroides and related proteins in other α-proteobacteria. In heterologous complementation analysis, R. sphaeroides PrrA rescued the growth defect of a P. aeruginosa mutant lacking RoxR, and RoxR enabled photosynthetic growth of an R. sphaeroides PrrA mutant. Also, RoxR could substitute for PrrA in activating transcription in vitro, demonstrating that these proteins are functional homologues. P. aeruginosa strains lacking RoxR or the sensor kinase (RoxS) were more sensitive than wild type to the respiratory inhibitors cyanide and azide. The phenotypes of these mutant strains correlated with reduced cyanide-insensitive O2 utilization and less cyanide-dependent expression of the locus encoding the CIO (cioAB). The ability of purified RoxR to bind to the cioAB promoter region also suggests that this protein acts directly to regulate cioAB transcription. Therefore, RoxR appears to play a role in regulating the transcription of loci for P. aeruginosa energy-generating enzymes similar to that of its homologues in α-proteobacteria.


The adaptable energy-producing systems of bacteria often contribute to their remarkable metabolic diversity. For instance, facultative bacteria can alter their mode of energy generation by taking advantage of different respiratory pathways (Poole and Cook, 2000; Richardson, 2000). One obvious distinction between these pathways is their final electron acceptor; aerobic electron transport chains use oxygen, whereas anaerobic pathways use compounds such as nitrate or dimethyl sulphoxide (DMSO). In recent years, it has become evident that most of the terminal enzymes in bacterial aerobic respiratory chains have catalytic subunits that are homologous to that of the eukaryotic mitochondrial cytochrome oxidase and possess a similar haem-copper bimetallic active site (Garcia-Horsman et al., 1994). However, the subunit composition, cytochrome content, electron donor and oxygen affinity of the bacterial oxidases varies, leading to the hypothesis that each has a specialized metabolic function (Poole and Cook, 2000). Consistent with this, facultative bacteria often alter the expression of their respiratory oxidases in response to environmental changes (Cotter et al., 1990; Gennis and Stewart, 1996; Delgado et al., 1998; Richardson, 2000). The mechanisms in place to regulate this complement of aerobic terminal oxidase enzymes in order to generate energy and survive in different habitats vary and are still largely uncharacterized. To address this issue, we have been studying the regulation of cytochrome oxidase production in the facultative anaerobe Pseudomonas aeruginosa.

It has been known for some time that P. aeruginosa, known for its ability to endure in diverse habitats, has a complex respiratory system. P. aeruginosa can respire by denitrification in the absence of oxygen or can use a branched aerobic respiratory chain (Fig. 1), one branch of which was shown to be relatively insensitive to the respiratory inhibitor cyanide (Matsushita et al., 1980; 1983; Zannoni, 1989). Two types of P. aeruginosa haem-copper enzymes have been purified and biochemically char-acterized (Matsushita et al., 1980; Fujiwara et al., 1992; Okamoto et al., 1995), and the genes that probably encode these oxidases can be found within the genome of P. aeruginosa PAO1 (Stover et al., 2000). One of these oxidases, an aa3-type cytochrome oxidase, is common to many bacterial species and is the enzyme most closely related to the mitochondrial cytochrome oxidase. The other, identified as a co3-type cytochrome oxidase, is most likely to be a mixture of the products of two adjacent loci predicted to encode oxidases related to the microaerophilic cbb3 cytochrome oxidases of rhizobia and other α-proteobacteria (Croft et al., 2000; Myllykallio and Liebl, 2000; Stover et al., 2000). The PAO1 genome also contains genes for an uncharacterized haem-copper quinol oxidase related to the Escherichia coli bo3 enzyme, which is produced by the cyo operon (Gennis and Stewart, 1996). A fifth P. aeruginosa oxidase, characterized for its role in cyanide-tolerant growth and called the cyanide-insensitive oxidase (CIO), is similar to the only known class of bacterial oxidases not of the haem-copper superfamily (Cunningham and Williams, 1995; Ray and Williams, 1996; Cunningham et al., 1997). The predicted presence of these five P. aeruginosa oxidases in addition to a complete denitrification system creates a respiratory system well-suited for the survival of this γ-proteobacterium in a range of environmental conditions (Fig. 1).

Figure 1.

Model of the P. aeruginosa electron transport system derived from ORFs present in the PAO1 genome. This facultative γ-proteobacterium can respire anaerobically by denitrification (black boxes) or aerobically using two quinol oxidases or three cytochrome oxidases. The cyanide-insensitive respiratory branch terminated by the CIO is indicated. The aa 3 oxidase is encoded by ORFs PA0105–0108 of the PAO1 genome, the bo 3 -type oxidase by PA1317–1321, cbb 3 #1 by PA1551–1554, cbb 3 #2 by PA1554–1557 and the bd -type oxidase by PA3929–3930. Note that the assignment of cytochromes to the bo 3 -type and bd -type oxidases has not been determined experimentally.

Pseudomonas aeruginosa is also one of the few bacterial species known to be capable of producing cyanide, a potent inhibitor of haem-copper oxidases, at concentrations that would inhibit their function ( Pudek and Bragg, 1974 ; Castric, 1975 ; Solomonson, 1981 ; Blumer and Haas, 2000 ). It has been postulated that the cyanide-insensitive respiratory branch, which uses the CIO, allows this bacterium to survive when cyanide is produced under oxygen-limiting conditions ( Castric, 1983 ; Matsushita et al., 1983 ; Cunningham and Williams, 1995 ). The two subunits of the CIO, CioA and CioB have 27.5% and 24.4% amino acid identity with E. coli cytochrome bd quinol oxidase subunits CydA and CydB respectively. However, spectroscopic studies have indicated that the CIO does not contain haem d and therefore has a cytochrome content distinct from the E. coli cytochrome bd oxidase ( Cunningham et al., 1997 ). Similar to the production of cyanide, the expression of the CIO has been shown to increase under oxygen limitation ( Cunningham et al., 1997 ).

This work demonstrates that expression of the P. aeruginosa CIO is regulated by a two-component regulator (RoxR) closely related to Rhodobacter sphaeroides PrrA and its homologues. Response regulators related to PrrA that are present in α-proteobacteria have been shown to control the expression of numerous loci encoding electron-generating or -utilizing processes (Sganga and Bauer, 1992; Eraso and Kaplan, 1994; Joshi and Tabita, 1996; Oh et al., 2000; Swem et al., 2001). Our data demonstrate that RoxR can substitute for R. sphaeroides PrrA in vivo and in vitro. Thus, P. aeruginosa RoxR can be included in a highly conserved group of response regulators known to be present in α- and γ-proteobacteria.


Pseudomonas aeruginosa RoxR is an analogue of R. sphaeroides PrrA

Open reading frame (ORF) PA4493 of the P. aeruginosa PAO1 genome encodes a protein with significant amino acid identity to members of a highly conserved family of two-component response regulators that have been described to date in six α-proteobacterial species (Sganga and Bauer, 1992; Eraso and Kaplan, 1994; Tiwari et al., 1996; Bauer et al., 1998; Masuda et al., 1999). Analysis of additional bacterial genomes indicated that related regulators are widespread in other facultative proteobacteria (J. C. Comolli and T. J. Donohue, unpublished). In R. sphaeroides and Rhodobacter capsulatus, these regulators are involved in the transcriptional regulation of loci encoding proteins involved in energy generation. We thought that PA4493, referred to as roxR, may play a similar role in P. aeruginosa. This ORF encodes a 20.6 kDa protein with 50.7% amino acid identity to PrrA including a stretch of 12 out of 21 conserved amino acids in the putative DNA-binding region of the C-terminal domain (Fig. 2A). RoxR is downstream of and possibly co-transcribed with a gene encoding a sensor histidine kinase, PA4494 (roxS), as the start codon of roxR is overlapped by the stop codon of roxS (Fig. 2B). The RoxS sensor kinase is a 49.8 kDa protein that has 22.6% amino acid identity to the R. sphaeroides sensor kinase PrrB, with much of the homology within the C-terminal catalytic kinase domain. The gene structure of roxSR differs from that of related α-proteobacterial systems in which the genes encoding the response regulator and sensor kinase are divergently transcribed, and a third gene encoding a protein of unknown function that acts in signal transduction (prrC in R. sphaeroides) precedes the response regulator gene (Eraso and Kaplan, 1995). Interestingly, an ORF (PA0114) encoding a protein with 36.5% amino acid identity to PrrC, including a conserved CxxC metal-binding motif (Eraso and Kaplan, 2000), is located elsewhere on the PAO1 chromosome.

Figure 2.

Description of RoxR and the roxSR locus.

A. Amino acid alignment of the C-termini of R. sphaeroides PrrA and P. aeruginosa RoxR. Identical amino acids (hatch marks) are indicated as is the 21-amino-acid domain that is absolutely conserved in α-proteobacterial species (box). The putative helices of the helix–turn–helix DNA-binding motif are highlighted by black bars.

B. Diagram of the P. aeruginosa roxSR locus. The relative orientation of PA4493 and PA4494 ORFs in the PAO1 genome is indicated. Also displayed are the deletions made in construction of the RoxR1, RoxS1 and RoxSR1 mutant strains.

The significant homology between P. aeruginosa RoxR and R. sphaeroides PrrA led to the hypothesis that the RoxSR and PrrBA two-component systems have similar functions. We tested this by substituting RoxR for PrrA and RoxS for PrrB. A plasmid containing P. aeruginosa roxR, roxS or roxSR was conjugated into a R. sphaeroides strain lacking prrA or prrB (PrrA2 or PrrB1), which have growth defects under anaerobic photosynthetic conditions (Eraso and Kaplan, 1994; 1995). This phenotype was evident, as PrrA2 or PrrB1 containing a plasmid without an insert (pRK415) was unable to grow anaerobically in the light (Table 1). A plasmid containing R. sphaeroides prrA (pUI1621) restored the ability of PrrA2 to grow photosynthetically, consistent with what has been observed in previous studies (Eraso and Kaplan, 1994; 1995). A plasmid containing prrA also restored photosynthetic growth to cells lacking prrB, presumably because the increased expression of PrrA allows this protein to function in the absence of its cognate sensor kinase (Eraso and Kaplan, 1995). P. aeruginosa roxR (pJC472) enabled PrrB1 or PrrA2 to grow photosynthetically, as did a plasmid containing roxSR (pJC453). This showed that P. aeruginosa RoxR could effectively substitute for R. sphaeroides PrrA in controlling the production of factors required for anaerobic growth by photosynthesis. RoxS did not appear to be able to replace PrrB, however, as a roxS-bearing plasmid (pJC473) did not allow PrrB1 to grow under photosynthetic conditions. RoxS and PrrB share less than 20% amino acid identity in regions other than the conserved domains present in all histidine kinases, so the most likely explanation for this result is that RoxS cannot interact properly with factors of the R. sphaeroides PrrB signal transduction chain.

Table 1. Complementation of photosynthetic growth defects of the R. sphaeroides prrA and prrB mutants.
Complementing plasmidPlasmid-borne locusStrain
  1. Growth under photosynthetic conditions (+ +, similar to wild-type; +, less than wild type; −, no growth).

pRK415None+ +
pUI1621 prrA + ++ ++
pJC453 roxSR + +++
pJC472 roxR + +++
pJC473 roxS + +

Based on the ability of RoxR to replace PrrA in vivo, this protein was overexpressed in Escherichia coli and purified for in vitro studies. Initially, we tested whether RoxR could be phosphorylated by the heterologous R. sphaeroides sensor kinase PrrB. When purified RoxR was mixed with the purified histidine kinase domain of PrrB (cPrrB), which was 32P labelled, the majority of the phosphate was transferred from cPrrB to RoxR within 2 min (Fig. 3A). Under the conditions tested, the production of RoxR~P was indistinguishable from the production of PrrA~P, the cognate R. sphaeroides response regulator for PrrB (Fig. 3A). When [32P]-acetyl phosphate was used as a phosphodonor, the rates of RoxR and PrrA phosphorylation were also indistinguishable (data not shown).

Figure 3.

Analysis of recombinant RoxR in vitro.

A. Phosphorylation of RoxR by R. sphaeroides cPrrB. A phosphorylation reaction that produced [32P]-cPrrB (top species) was untreated or mixed with 4 or 8 μM RoxR or 8 μM PrrA for 2 min. The amount of phosphorylated response regulator (bottom species) produced was assessed by SDS–PAGE.

B. Transcriptional activation by RoxR. The ability of RoxR to generate a specific transcript from the R. sphaeroides PrrA-dependent cycA P2 promoter was assessed using an in vitro transcription assay. Each transcription reaction contained a plasmid template containing the cycA P2 promoter and an RNA1 control promoter, R. sphaeroides RNAP holoenzyme and RoxR or PrrA at the concentration indicated. The response regulators were either treated with acetyl phosphate for 60 min before the assay (+) or untreated (−).

We also examined whether RoxR could substitute for PrrA in activating transcription from a R. sphaeroides PrrA-dependent promoter, the cytochrome c2 gene P2 promoter (cycA P2). RoxR was added to a reaction containing R. sphaeroides RNA polymerase (RNAP) holoenzyme, and the production of cycA P2-specific transcript was monitored. In the absence of RoxR or PrrA, no detectable product was produced from the cycA P2 promoter (Fig. 3B). The addition of 0.5 μM unphosphorylated PrrA to the reaction caused cycA P2-specific transcript to accumulate, and treatment of PrrA with acetyl phosphate increased the amount of product generated ≈ 50-fold compared with that seen with an equivalent amount of unphosphorylated PrrA (Fig. 3B). This was similar to previous observations seen with this in vitro assay (Comolli et al., 2002). P. aeruginosa RoxR also stimulated the production of cycA P2 transcript, but to a lesser extent than PrrA. The production of cycA P2 transcript increased in a dose-dependent manner with 1, 3 and 5 μM unphosphorylated RoxR, and phosphorylation further augmented RoxR-dependent transcript production roughly 10-, 3- and 2.5-fold (Fig. 3B). Either activator stimulated transcription of the cycA P2 promoter specifically, as no significant change was observed in the amount of transcript produced at the control RNA1 promoter.

These experiments show that RoxR can substitute for PrrA to stimulate the expression of PrrA-dependent target genes in vivo, and can interact productively with the R. sphaeroides transcriptional machinery to activate transcription in vitro. Purified RoxR and PrrA also behave similarly in that the unphosphorylated proteins have the ability to activate transcription, and this activity is significantly enhanced by phosphorylation.

Loss of roxSR causes a defect in cyanide-insensitive aerobic respiration

To investigate the roles of P. aeruginosa RoxR and RoxS, each gene was inactivated in strain PAK. PA4493 (roxR) was inactivated by the deletion of 432 bp and the insertion of a ΩGmr cassette (Fig. 2B). The resulting strain was designated RoxR1. An in frame 1030 bp deletion of PA4494 (roxS) was constructed to avoid potential polar effects on roxR, and this mutant was designated RoxS1. An additional mutant, in which both genes were inactivated (RoxSR1), was constructed by a 1393 bp deletion encompassing a portion of both genes followed by the insertion of a ΩGmr cassette (Fig. 2B).

As we suspected that the RoxSR locus plays a role in P. aeruginosa energy generation, as does PrrBA in R. sphaeroides, we compared the growth phenotypes of the mutant and wild-type strains under different respiratory conditions. The three mutant strains exhibited a growth rate indistinguishable from PAK wild type under aerobic conditions in rich or minimal medium (Fig. 4A; data not shown). A similar growth rate to wild type was also seen when the strains were grown anaerobically in rich or minimal medium containing 20 mM nitrate as an alternative electron acceptor (data not shown). This suggested that the aerobic and anaerobic respiratory pathways remained fully functional in the mutant strains. As P. aeruginosa is known to have a distinct branch of aerobic respiration that is insensitive to cyanide (Matsushita et al., 1983), we also assessed the function of this branch by testing the sensitivity of the mutants to cyanide. As shown in Fig. 4A, the addition of 0.2 mM KCN to the growth medium caused only a slight decrease in the growth rate of the wild type, but completely inhibited the growth of the RoxR1 and RoxSR1 mutants. The RoxS1 strain grew slightly more than RoxR1 or RoxSR1, so it appeared that loss of RoxS also caused a significant but less severe defect in aerobic growth when cyanide is present (Fig. 4A). By varying the cyanide concentration in the media, we found that the sensitivities of RoxR1, RoxS1 and RoxSR1 to this compound were increased 10- to 15-fold relative to the wild type (Fig. 4B). Wild-type PAK tolerated well over 1 mM KCN (data not shown), but the mutant strains were unable to grow at concentrations at or above 0.15 mM. As seen in the previous experiment, the KCN sensitivity of RoxS1 was slightly less than that of RoxR1 or RoxSR1, especially at 0.1 mM KCN, again showing that inactivation of roxS produced a less severe phenotype. The growth phenotype of each strain was comparable when the respiratory inhibitor sodium azide was used but, for the sake of brevity, this data set is not presented.

Figure 4.

Figure 4.

Cyanide sensitivity of P. aeruginosa roxSR mutant strains.

A. Growth curves of PAK wild type (circles), RoxR1 (diamonds), RoxS1 (triangles), RoxSR1 (squares) or CioAB1 under aerobic conditions in LB at 37°C. The cultures were untreated (solid) or contained 100 μM KCN (hollow).

B. Growth rates, measured as cell divisions per hour, of the strains indicated in (A) measured with different concentrations of KCN in the growth medium.

Figure 4.

Figure 4.

Cyanide sensitivity of P. aeruginosa roxSR mutant strains.

A. Growth curves of PAK wild type (circles), RoxR1 (diamonds), RoxS1 (triangles), RoxSR1 (squares) or CioAB1 under aerobic conditions in LB at 37°C. The cultures were untreated (solid) or contained 100 μM KCN (hollow).

B. Growth rates, measured as cell divisions per hour, of the strains indicated in (A) measured with different concentrations of KCN in the growth medium.

To confirm that inactivation of roxR or roxS was responsible for the inhibitor sensitivity, mutant strains were complemented with plasmids containing no insert (pUCP26), roxS (pJC461), roxR (pJC462) or roxSR (pJC460). The growth of these strains in the presence of 0.5 mM azide was measured, and the results are presented in Table 2. As anticipated from the previous results, the doubling time of RoxR1 and RoxSR1 containing pUCP26 increased the most relative to wild type, with that of RoxS1 slightly less affected. Consistent with inactivation of a single gene in each mutant, the doubling time of RoxR1 was restored to near that of wild type by the plasmid bearing roxR and RoxS1 by the plasmid bearing roxS. In addition, the rescue of RoxS1 by pJC461 showed that the in frame deletion in roxS did not have significant polar effects on the downstream roxR gene. The doubling time of RoxS1 was restored to approximately that of wild type by a plasmid containing only roxR, demonstrating that multiple copies of this gene can overcome a defect in roxS. The growth of RoxSR1 was also rescued by either roxSR or roxR alone on a plasmid.

Table 2. Complementation of the azide sensitivity of P. aeruginosa RoxR1, RoxS1 and RoxSR1 mutants.
time (min)
Wild typepUCP26None37.9 ± 1.0
RoxR1pUCP26None 243 ± 15
 pJC462 roxR 41.3 ± 1.0
 pJC470 prrA 46.1 ± 2.1
RoxS1pUCP26None 124 ± 2.7
 pJC461 roxS 42.7 ± 0.15
 pJC462 roxR 49.0 ± 1.7
 pJC471 prrB 38.8 ± 1.7
 pJC470 prrA 41.7 ± 2.1
RoxSR1pUCP26None 207 ± 16
 pJC460 roxSR 39.2 ± 0.48
 pJC462 roxR 52.4 ± 0.77
 pJC470 prrA 41.1 ± 0.21

In a complementary experiment, a plasmid bearing R. sphaeroides prrA or prrB was placed in P. aeruginosa RoxR1, RoxS1 or RoxSR1, and the doubling time of each strain was determined in the presence of 0.5 mM azide. As shown in Table 2, a plasmid containing prrA (pJC470) restored the growth rate of RoxR1 to near that of the PAK wild type. The same plasmid was also able to rescue the azide sensitivity of the RoxS1 and RoxSR1 mutants, in these cases actually better than pJC462, a plasmid bearing the native P. aeruginosa roxR. RoxS1 containing a prrB-bearing plasmid had a growth rate similar to that of the wild-type strain, indicating that PrrB effectively substituted for RoxS in restoring azide resistance. This data set provides independent support for the notion that the amino acid identity between RoxR and PrrA reflects a functional similarity between these two-component systems.

RoxS and RoxR regulate cyanide-insensitive oxidase production

The cyanide-insensitive branch of P. aeruginosa electron transport has been shown to be terminated by a respiratory oxidase related to the E. coli cytochrome bd oxidase that is referred to as the cyanide-insensitive oxidase (CIO) (Fig. 1). At cyanide levels that inhibit the activity of haem-copper oxidases, this enzyme remains active by the nature of its bihaem reactive site (Cunningham and Williams, 1995; Cunningham et al., 1997). We speculated that the cyanide sensitivity of RoxR1 and RoxS1 resulted from a CIO deficiency. To test this notion, we compared the activity of this oxidase in wild-type and mutant strains lacking RoxR or RoxS.

To accomplish this, CIO activity was assayed by measuring the NADH-dependent oxygen consumption from purified cytoplasmic membranes in the presence of cyanide. In membranes from aerobically grown wild-type cells, roughly 16% of the total measured oxidase activity persisted after the addition of 0.1 mM cyanide (Table 3). In contrast, when 0.1 mM cyanide was present in the growth medium, 89% of the total oxidase activity in wild-type membranes was cyanide resistant. This showed that cyanide present during growth induced the cyanide-insensitive fraction of oxidase activity about 5.5-fold in the wild-type strain.

Table 3. NADH-dependent oxidase activity in the presence and absence of cyanide.
StrainKCNaActivity –KCNbActivity + KCNb% CIO
  • a.  

    0.1 mM KCN added to growth medium.

  • b.   NADH-dependent oxidase activity (μmol of O 2  min −1  mg −1 protein) ± 0.1 mM KCN.

PAK184 ± 1729.6 ± 0.016
RoxR1320 ± 2327.9 ± 3.2 8.7
RoxS1319 ± 1426.1 ± 2.7 8.2
CioAB1231 ± 4.814.8 ± 2.1 6.4
PAK+311 ± 31276 ± 2689
RoxR1+395 ± 17126 ± 9.232
RoxS1+390 ± 46 53 ± 1.614
CioAB1+309 ± 1225.2 ± 2.3 8.1

The fraction of cyanide-resistant oxidase activity in membranes from RoxR1 or RoxS1 was approximately 50% that of wild type, primarily because of an increase in the total activity in membranes from the mutant strains (Table 3). When cyanide was added to the growth medium of RoxR1 or RoxS1, only 32% or 14% of the total oxidase activity from purified membranes was cyanide resistant. Thus, the induction of CIO activity in response to cyanide was decreased 64–84% relative to the wild type. This deficiency in CIO activity caused by loss of RoxSR, named to denote its involvement in respiratory oxidase expression, could account for the increased cyanide sensitivity of the mutant strains.

To compare the cyanide sensitivity of RoxR1 and RoxS1 with that of an isogenic strain completely lacking CIO, we constructed a PAK strain with the cioAB locus, encoding for the two subunits of the cyanide-insensitive oxidase (Cunningham et al., 1997), inactivated. The resulting strain, designated CioAB1, displayed a reduced growth rate relative to the wild type in the presence of 0.2 mM cyanide (Fig. 4A). Interestingly, the CioAB1 strain was not as sensitive to cyanide as the RoxR1, RoxS1 or RoxSR1 mutants (Fig. 4B). This suggested that the cyanide-related phenotype of RoxR1 and RoxS1 was not solely the result of reduced expression of the CIO, but that the two-component system probably affected other cellular processes involved in resistance to respiratory inhibitors such as cyanide. The observation that membranes from RoxR1 and RoxS1, grown in the presence or absence of cyanide, contained more cyanide-resistant oxidase activity than membranes from CioAB1 (Table 3) supports this notion.

As R. sphaeroides PrrA and R. capsulatus RegA act by regulating the transcription of target genes, we also determined whether inactivation of roxR or roxS altered the expression of a fusion of the cioAB promoter to E. coli lacZ. Under aerobic conditions in the absence of exogenous cyanide, the β-galactosidase activity from the cioAB:lacZ reporter was only slightly lower (a 35–40% decrease) in the RoxR1 and RoxS1 mutant strains than in wild-type cells (Fig. 5), suggesting a limited role for RoxSR in cioAB expression under these conditions. However, the cyanide-dependent induction of cioAB transcription was greatly reduced by loss of RoxR or RoxS. In the wild-type strain, 50 μM or 100 μM cyanide in the growth medium stimulated cioAB:lacZ expression roughly three- and eightfold respectively. This increase in cioAB expression was attenuated to only a 1.5- to twofold increase when the RoxR1 or RoxS1 mutant strains were grown in 100 μM KCN. Thus, inactivation of roxR or roxS negatively affected cioAB expression, consistent with the reduction in CIO activity that was seen in these mutant strains. It should be noted that some stimulation of cioAB:lacZ expression, albeit reduced, was observed in the absence of RoxR or RoxS. This implies that additional factors are involved in controlling the induction of CIO expression (Cunningham et al., 1997).

Figure 5.

Activity from a β-galactosidase reporter plasmid containing a cioAB : lacZ fusion in PAK, RoxR1 or RoxS1. Each strain was grown aerobically in LB with no addition (black) or with 50 μM (stippled) or 100 μM KCN (white).

RoxR interacts with the cioAB promoter

To ascertain whether RoxR was directly involved in controlling cioAB expression, we tested whether this protein was able to interact with a cioAB promoter fragment that is known to contain the transcriptional initiation site of this operon (Cunningham et al., 1997). No significant transcription product was generated when RoxR/RoxR~P or PrrA/PrrA~P was used with R. sphaeroides RNAP to activate transcription from a template containing the cioAB promoter (data not shown). However, a direct interaction of RoxR with the cioAB promoter was evident when a gel shift mobility assay was used. Unphosphorylated RoxR at a concentration of 3 or 5 μM generated a significant mobility shift of a labelled cioAB promoter fragment that contained ≈ 230 bp upstream of the previously determined transcriptional start site (Fig. 6, left lanes). RoxR-dependent complex formation with this cioAB DNA appeared to be specific, as it was unaffected by the 100-fold excess of non-specific competitor DNA that was present in the assay. The ability of RoxR~P, generated by incubation of the protein with phosphorylated cPrrB, to bind to the cioAB promoter was enhanced relative to unphosphorylated RoxR. Not only could a mobility shift now be detected using 1 μM RoxR~P, but more of the labelled cioAB fragment was present in the presumed complex when 3 μM protein was used (Fig. 6, right lanes). This effect resulted from phosphorylation of RoxR by cPrrB, as the mobility shift generated by RoxR incubated with cPrrB but lacking ATP appeared to be the same as the mobility shift of RoxR only (Fig. 6, middle lanes). Our conclusion is that RoxR is a positive regulator of CIO expression with the ability to bind to the cioAB promoter. Interestingly, RoxR had DNA-binding activity in the absence of phosphorylation, and this activity was stimulated by phosphorylation. A similar phenomenon has been observed with the binding of the related R. sphaeroides PrrA protein to target promoters (J. C. Comolli and T. J. Donohue, unpublished observations).

Figure 6.

RoxR binding to the cioAB promoter. A 32 P-labelled fragment of the cioAB promoter (−229 to +55) was incubated with RoxR at the concentration indicated, then analysed in a gel mobility shift assay. Binding reactions contained no RoxR (outside lanes), RoxR only (left lanes), RoxR treated with cPrrB (centre lanes) or RoxR treated with cPrrB~P (right lanes).


Role for an uncharacterized P. aeruginosa two-component system

This work demonstrates a function for two previously uncharacterized P. aeruginosa genes (ORFs PA4493 and PA4494 of the PAO1 genome) that are predicted to encode a two-component regulatory pair. The sensor kinase and response regulator genes were designated roxS and roxR as their products regulated the expression of the cioAB locus that encodes the cyanide-insensitive respiratory oxidase (CIO). Inactivation of either roxS or roxR resulted in a greater than 10-fold increase in the sensitivity of P. aeruginosa cells to the respiratory inhibitors cyanide and azide. This correlated with a 64–90% reduction in cyanide-dependent induction of CIO activity and cioAB expression in the mutant strains compared with wild type. RoxR was also demonstrated to interact directly with the cioAB locus in a gel mobility shift assay. When these observations are considered together, it appears that RoxR is a direct positive transcriptional regulator of CIO production.

RoxR is related to a family of α-proteobacterial regulators

In the sense that P. aeruginosa RoxR controls the production of CIO, a component of an energy-generating pathway, the protein has a role similar to that of homologous response regulators that have been characterized in α-proteobacteria. Two members of this response regulator family from purple non-sulphur bacteria, R. sphaeroides PrrA and R. capsulatus RegA, are global regulators of gene expression under low oxygen conditions. PrrA has been shown to be required for the production of R. sphaeroides factors required for photosynthesis, DMSO-dependent respiration, carbon assimilation and nitrogen fixation (Eraso and Kaplan, 1994; Joshi and Tabita, 1996; Mouncey and Kaplan, 1998; Oh et al., 2000). Similarly, RegA of R. capsulatus has been demonstrated to affect the expression of more than 10 loci involved in energy generation, including those that encode the cbb3 and cytochrome bd-type terminal respiratory oxidases (Sganga and Bauer, 1992; Elsen et al., 2000; Vichivanives et al., 2000; Swem et al., 2001; Kappler et al., 2002). Recent studies have shown that R. capsulatus RegA or Sinorhizobium meliloti RegR (also called ActR) can substitute for the homologous Bradyrhizobium japonicum response regulator in promoting root nodule symbiosis, thus demonstrating that these proteins are functionally interchangeable (Emmerich et al., 2000). We have extended this analysis to include a response regulator in this family from a γ-proteobacterium. The observations that P. aeruginosa RoxR was able to restore photosynthetic growth to an R. sphaeroides prrA mutant and that PrrA could substitute for RoxR in maintaining cyanide resistance show that RoxR is a functional homologue of the PrrA/RegA response regulator family. ORFs related to RegA/PrrA are evident in a number of α-, β- and γ-proteobacterial genomes, so it is possible that this type of regulatory system plays a similar role in controlling energy generation in other facultative proteobacteria.

The ability of RoxR to substitute for PrrA in vivo, and vice versa, probably means that these proteins are appropriately regulated by the heterologous sensor kinase and that there are common features in their mechanisms of transcriptional activation. Our in vitro demonstration that the kinase domain of PrrB phosphorylated RoxR and that RoxR interacted with R. sphaeroides polymerase to activate transcription from cycA P2 corroborate this supposition. RoxR also behaved similarly to PrrA in that the unphosphorylated regulator was able to activate transcription in vitro (Comolli et al., 2002). This may mean that the unphosphorylated form of RoxR plays a role in the regulation of P. aeruginosa energy generation in vivo under non-activating conditions, as PrrA seems to do in R. sphaeroides (Eraso and Kaplan, 1995; 1996; Karls et al., 1999).

This work also demonstrated the capacity of RoxR to recognize PrrA-dependent target promoters despite the conservation of only 12 of 21 amino acids in their C-terminal putative DNA-binding domains (Fig. 2A). This is significant as, in the characterized α-proteobacterial members of this response regulator family, this domain shares 100% amino acid identity (Bauer et al., 1998; Masuda et al., 1999). This strict conservation of amino acid sequence between species suggests that any variation would be detrimental to the function of the protein, but the ability of RoxR to replace PrrA in vivo and in vitro indicates that this need not be the case. However, it is still possible that some differences in the DNA recognition capabilities of RoxR and PrrA do exist but went undetected as we have examined the interaction of these proteins with only a limited number of promoters.

Model for RoxR-dependent signal transduction

The similar effect of loss of RoxS or RoxR on cyanide-inducible CIO expression, in addition to the ability of multicopy roxR to rescue the growth defect of a roxS mutant, provides strong genetic evidence that RoxS and RoxR interact as members of the same two-component signalling pathway. However, inactivation of roxS produced a less severe defect in cyanide or azide sensitivity than inactivation of roxR. This is probably caused by the activity of unphosphorylated RoxR or by phosphorylation of RoxR by small molecule phosphodonors and/or non-cognate sensor kinases in the absence of RoxS, each of which has been demonstrated to occur in vitro.

Although the signal(s) influencing RoxS and resulting in RoxR activation remain to be identified, roxS or roxR inactivation had a greater effect on CIO activity and cioAB expression in the presence of cyanide. This suggests that the two-component system may be responsive to cyanide. Cyanide has long been known to affect respiratory function by inhibiting haem-copper respiratory oxidases, and P. aeruginosa is capable of producing cyanide at concentrations able to reduce or prevent the function of these enzymes (Pudek and Bragg, 1974; Castric, 1975; 1977; Solomonson, 1981; Blumer and Haas, 2000). As P. aeruginosa CIO, like other oxidases related to the E. coli bd oxidase, is less sensitive to cyanide because of the bihaem composition of its active site (Cunningham and Williams, 1995; Gennis and Stewart, 1996; Poole and Cook, 2000), increased production of this enzyme would enable the bacterium to respire even when cyanide has blocked the function of the other respiratory oxidases. So, RoxSR may provide a signalling mechanism to activate the production of CIO when significant amounts of cyanide are generated.

Recent findings have shown that the related R. sphaeroides PrrBA system receives an inhibitory, electron transport-dependent signal from the cbb3 oxidase (O’Gara et al., 1998; Oh and Kaplan, 1999 2000). Reducing the electron flow through this oxidase by mutation, the use of inhibitors or a decrease in oxygen tension stimulated the expression of PrrBA target genes in a PrrA-dependent fashion (O’Gara et al., 1998; Oh and Kaplan, 2000). As the P. aeruginosa genome encodes two distinct cbb3 oxidases, it is possible that a similar signalling system regulates RoxSR. Our working model based upon these observations is that cyanide decreases electron flow through one or both the P. aeruginosa cbb3 oxidases by inhibiting their ability to reduce oxygen. This in turn relieves an inhibitory oxidase-derived electron transport signal in a manner similar to R. sphaeroides PrrBA, and RoxSR is activated in order to stimulate CIO expression. Our observations are consistent with this model, and experiments are in progress to test this hypothesis in the hope of delineating a signalling pathway leading to RoxSR.

Expression of the cyanide-insensitive oxidase

Although RoxSR may provide a mechanism to induce CIO synthesis in the presence of cyanide, it does not appear to act alone at the cioAB promoter. In the absence of RoxS or RoxR, cyanide-dependent induction of CIO activity and cioAB expression was reduced but still evident. The P. aeruginosa anaerobic regulator ANR is thought to repress cioAB expression on account of two consensus ANR binding sites at inhibitory positions within the promoter region (Zimmermann et al., 1991; Cunningham et al., 1997) but, to date, no other potential regulators of cioAB expression have been identified. The co-regulation of P. aeruginosa cioAB by RoxR and ANR is analogous to the dual control of a number of R. sphaeroides photosynthesis genes by PrrA and FnrL, a protein related to ANR (Oh et al., 2000; Oh and Kaplan, 2000). However, it is possible that the control of cioAB transcription is influenced by more than these two factors, as is that of the E. coli cydAB operon that encodes the cytochrome bd oxidase. The E. coli cydAB operon has multiple promoters, some of which are co-ordinately regulated by the anaerobic regulator FNR, the two-component system ArcBA and the transcription factor H-NS (Cotter and Gunsalus, 1992; Cotter et al., 1997; Govantes et al., 2000a,b). The interplay between multiple regulatory systems could help to ensure the appropriate expression of the E. coli bd oxidase or P. aeruginosa CIO in response to changes in oxygen tension or the oxidation–reduction state of the aerobic respiratory chain.

Other loci influenced by RoxR

Our observations suggest that P. aeruginosa has RoxSR-dependent functions other than the CIO to cope with the presence of cyanide. For example, loss of CIO by cioAB inactivation had a lesser effect on cyanide sensitivity than did inactivation of roxS or roxR. One possibility is that cyanide resistance requires other factors that are positively regulated by RoxSR and, therefore, are not expressed in mutants lacking RoxR or RoxS. Along these lines, P. aeruginosa has been reported to have a detoxification mechanism that involves the excretion of α-ketoglutarate, which will react spontaneously with cyanide to produce cyanohydrin (Von Tigerstrom and Campbell, 1966; Blumer and Haas, 2000). More recent studies have demonstrated that another Pseudomonad, Pseudomonas fluorescens, can use this cyanohydrin product as a source of nitrogen (Kunz et al., 1992; 1998). An alternative possibility that does not invoke additional cyanide resistance mechanisms is that a defect in RoxSR causes the inappropriate expression of factors that increase respiratory inhibitor sensitivity. The identification of other loci regulated by RoxSR will help to address this issue.

In conclusion, P. aeruginosa RoxSR appears by both amino acid sequence and function to be a member of a class of two-component systems that includes R. sphaeroides PrrBA. Although the roles of these systems have not been completely elucidated, they appear to be important in adapting the flexible electron transport pathways of facultative bacteria in response to environmental stimuli. The presence of related two-component pairs in many proteobacteria may reflect their importance in the metabolic regulation of energy-generating pathways of these facultative bacteria.

Experimental procedures

Strains and growth conditions

Strains and plasmids used in this study are listed in Table 4. E. coli strains were propagated on Luria broth (LB) containing 100 μg ml−1 ampicillin, 15 μg ml−1 gentamicin or 10 μg ml−1 tetracycline. R. sphaeroides was grown as described previously (Comolli et al., 2002). P. aeruginosa strain PAK was grown on Vogel–Bonner medium (Vogel and Bonner, 1956) or LB supplemented with 100 μg ml−1 gentamicin, 200 μg ml−1 carbenicillin or 100 μg ml−1 tetracycline. For aerobic growth assays, overnight P. aeruginosa cultures were diluted into 8 ml of LB with or without KCN or NaN3 in a 125 ml flask maintained at 37°C with shaking at 250 r.p.m. Growth rates were calculated by linear regression of at least three separate turbidity readings taken during exponential growth. Error bars represent the standard deviation from the mean of three independent samples.

Table 4. Strains and plasmids used in this study.
StrainsRelevant characteristicsSource/reference
P. aeruginosa  
 PAKWild typeJ. Mattick
 RoxR1PAK roxRΔSmaI–NsiI::ΩGmr cassette; cyanide sensitiveThis study
 RoxS1PAK roxS in frame Δ1030 bp; cyanide sensitiveThis study
 RoxSR1PAK roxSRΔPstI:: ΩGmr cassette; cyanide sensitiveThis study
 CioAB1PAK cioABΔStuI:: ΩGmr cassetteThis study
E. coli   
 DH5α recA1 endA1 gyrA96 thi-1 hsdR17 supE44 relA1

deoR Δ( lacZYA-argF ) U169 l -(φ80dL acZ ΔM15)
Bethesda Research Laboratories
 S17-1Conjugal donor; C600::RP4 2-(Tc::Mu)(Km::Tn7) pro res mod+ (Tpr Smr) Simon et al. (1983 )
 ER2566fhuA2 [lon] ompT lacZ::T7 gene1 gal sulA11 Δ(mcrC-mrr)114::IS10
R(mcr-73::miniTn10–TetS)2 R(zgb-210::Tn10)(TetS) endA1 [dcm]
New England Biolabs
 M15 (pREP4) lac ara gal mtl F recA + uvr + ; Km r Villarejo and Zabin (1974 )
R. sphaeroides   
 2.4.1Wild typeLaboratory strain
 PrrA22.4.1 prrAΔBstBI–PstI::ΩSmrSpr PS Eraso and Kaplan (1995 )
 PrrB12.4.1 prrBΔNruI–RsrII::ΩSmrSpr PS Eraso and Kaplan (1995 )
 pBS II-KSApr cloning vectorStratagene
 pRK415Broad-host-range plasmid; mob+; Tcr Keen et al. (1988 )
 pX1918GTSource of insertion cassette containing ΩGmr/xylE; Apr, Gmr Schweizer and Hoang (1995 )
 pEX100TSuicide vector for allelic replacement; oriT, sacB; Apr Schweizer and Hoang (1995 )
 pNHG1pLO1::1.48 kb DdeI–VspI fragment from pBBR1MCS-3; Kmr, Tcr Jeffke et al. (1999 )
 pUCP26 E. coli / P. aeruginosa shuttle vector; ori 1600; Tc r West et al. (1994 )
 pLP170Transcriptional lacZ fusions reporter; contains 1.8 kb stabilization fragment; Apr Preston et al. (1997 )
 pUI1621pRK415 with ΩSmrSpr plus a 0.8 kb fragment containing prrA Eraso and Kaplan (1994 )
 pUI1643pBS/BamHI–HindIII, 4 kb BamHI–HindIII 2.4.1 chromosomal fragment Eraso and Kaplan (1994 )
 HP45ΩGmrGmr cassette cloned into BssHI–SphI of pHP45ΩThis study; Prentki and Krisch (1984)
 pRKK148 In vitro transcription template containing cycA P2 –228 to +22 fused to Ω spc ; Ap r Karls et al. (1999 )
 pTYB2C-terminal intein/chitin-binding domain expression vector; AprNew England Biolabs
 pJC451pBSII/EcoRI–XhoI, 2.4 kb PCR-generated roxSR; AprThis study
 pJC453pRK415/EcoRI–XbaI(blunt), 2.4 kb pJC451/EcoRI–EcoRV; AprThis study
 pJC457pBSII/BamHI–HindIII, 5.3 kb PCR-generated PAK roxSR; AprThis study
 pJC458pJC457/PstI, ΔroxSR::Gmr cassette from pX1918GT/PstI; AprThis study
 pJC459pEX100T/SmaI, 6.4 kb pJC457/SpeI(blunt)–HindIII(blunt); AprThis study
 pJC460pUCP26/XbaI–SmaI, 2.2 kb XbaI–EcoRV pJC451 fragment; TcrThis study
 pJC461pUCP26/XbaI–SmaI, 1.7 kb XbaI–SmaI pJC451 fragment; TcrThis study
 pJC462pUCP26/KpnI,1.7 kb KpnI pJC451 fragment; TcrThis study
 pJC464pBSII/XbaI–HindIII, 4.2 kb insert ΔroxS (in frame); AprThis study
 pJC470pUCP26/SmaI–BamHI, 0.9 kb pUI1643 StuI–BamHI fragment; TcrThis study
 pJC471pUCP26/SmaI–HindIII, 2.5 kb pUI1643 StuI–HindIII fragment; TcrThis study
 pJC472pRK415/KpnI–XbaI(blunt), 1.5 kb pJC457 KpnI–EcoRV fragment; TcrThis study
 pJC473pRK415/BamHI–EcoRI(blunt), 3.0 kb pJC457 BamHI–SmaI fragment; TcrThis study
 pJC475pJC457/SmaI–NsiI, ΔroxR::Gmr cassette from pX1918GT/ClaI–PstI(blunt); AprThis study
 pJC481pNHG1/XbaI(blunt), 4.2 kb pJC464 XbaI(blunt)–HindIII(blunt) fragment; Tcr, KmrThis study
 pJC482pNHG1/XbaI–AscI, 5.3 kb pJC475 SpeI–AscI; Tcr, KmrThis study
 pJC494pBSII/XbaI–SacI, 5.3 kb PCR generated PAK insert with cioAB; AprThis study
 pJC496pBSII/XbaI–BamHI, 0.68 bp PCR-generated PAK insert with cioAB upstream
region; Apr
This study
 pJC701pLP170/SalI–BamHI, 0.43 bp pJC496/SalI–BamHI; AprThis study
 pJC714pJC494/StuI ΔcioAB::ΩGmr, 2.1 kb pHP45ΩGmr/SmaI; AprThis study
 pJC723pNHG1/XbaI–SacI, 4.4 kb pJC714/XbaI–SacI; Tcr, KmrThis study
 pJC724pTYB2/NdeI–SmaI, 0.6 kb PCR-generated PAK roxR/NdeI; AprThis study

Inactivation of roxS, roxR, roxSR and cioAB

A 5.28 kb fragment containing the ORFs designated PA4493 and PA4494 in the PAO1 genome (Stover et al., 2000) was amplified from P. aeruginosa PAK genomic DNA using primers derived from the PAO1 genomic sequence (pReg3: 5′-ATATAGGATCCCCAGGCCCATGTCGGTGAAGTAGG-3′; pReg4: 5′-ATATATAAGCTTGTCGCTTTCGCCACCATCGT C-3′), digested with BamHI and HindIII at sites contained in the primers and cloned into pBSII-KS (pJC457). For allelic replacement of roxSR, a 1393 bp PstI fragment was replaced with the ΩGmr cassette from pX1918GT to generate pJC758 (Fig. 2). A similar strategy was used for inactivation of roxR except that a 434 bp SmaI–NsiI fragment of pJC457 was replaced by the ΩGmr cassette from pX1918GT to generate pJC475. In frame deletion of codons 41–379 of roxS was accomplished by joining PCR amplification products lacking 1030 bp of the ORF. The 3′ end of roxS was amplified from pJC457 using the T7 and pRegB-D1 (5′-GGTTAGATCTAC TGGCGGAGAGCAGGAAAACC-3′) primers, and the 5′ end was amplified using the T3 and pRegB-D2 (5′-CAGCAGAT CTTGCCTGAGTATCCGCGACCAC-3′) primers. These products were digested with XbaI–BglII and HindIII–BglII, respectively, then ligated into pBSII-KS digested to generate pJC464.

The 5.3 kb region of the P. aeruginosa chromosome containing cioAB was amplified using the Cio3 (5′-ATATATTCTA GAGCTGCCTTCGTTGGGAATGGC-3′) and Cio4 (5′-ATATA TGAGCTCGAAACGCCGACACCGCTAAGC-3′) primers and cloned into pBSII-KS using XbaI and SacI sites within the primers. A 2.2 kb fragment of the resulting plasmid (pJC494) was removed by StuI digestion and replaced with the ΩGmr cassette from pHP45ΩGmr (pJC714).

The mutated genes were cloned into the suicide vectors pEX100T (roxSRΩGmr) or pNHG1 (roxRΩGmr, roxSΔ, cioABΩGmr), which were conjugated into P. aeruginosa strain PAK using E. coli S17-1. Proper insertion of the ΩGmr cassette in exconjugates was confirmed by Southern blot (for RoxSR1) or by diagnostic restriction digestion of an amplified genomic region containing the rearrangement (for RoxR1, RoxS1 and CioAB1).

Complementation of P. aeruginosa and R. sphaeroides mutants

A 2.4 kb fragment containing roxSR was cloned into pBSII-KS after amplification from the PAK genome with the pReg-1 (5′-GCTGAATTCACACTGTAGAGCATC-3′) and pReg-2 primers (5′-GGCTGCTCGAGCAGGCCGGAATAG-3′). The roxSR, roxS or roxR genes were excised from the resulting plasmid (pJC451) and placed into corresponding restriction sites of pUCP26 (West et al., 1994). A 2.2 kb XbaI–EcoRV fragment that contained roxSR, a 1.7 kb XbaI–SmaI fragment that bore roxS or a 1.7 kb KpnI fragment that contained roxR were used to generate pJC460, pJC461 or pJC462 respectively. A comparable strategy was used for constructing complementing plasmids containing the R. sphaeroides prrA or prrB genes. The prrA gene was isolated on a 0.9 kb SmaI–BamHI fragment from pUI1643 and cloned into identical sites of pUCP26 (pJC470), and the prrB gene was removed from pUI1643 on a 2.5 kb StuI–HindIII fragment and placed into pUCP26 (pJC471). The pUCP26-based plasmids were transformed into the appropriate P. aeruginosa strain by electroporation (Smith and Iglewski, 1989).

The roxR, roxS or roxSR loci were cloned into pRK415 for complementation studies in R. sphaeroides. The roxSR locus was removed from pJC451 on an EcoRI–EcoRV fragment. The individual roxR or roxS genes were excised from pJC457 by KpnI–EcoRV or BamHI–SmaI digestion respectively. These plasmids were conjugated into the appropriate R. sphaeroides strain using S17-1.

Purification of RoxR

Recombinant RoxR was expressed in E. coli and purified using an intein fusion system that has been successful for the analysis of PrrA homologues (Du et al., 1998; Comolli et al., 2002). The ORF encoding RoxR was amplified from pJC457 using the RoxR-int1 (5′-ATATATCATATGCTGAAG AGCTCTTGC-3′) and the RoxR-int3 (5′-GCGCCGTACCGG CCGCTTCTG-3′) primers. The 600 bp product was digested with NdeI, cloned into pTYB2 to generate a C-terminal fusion of RoxR to an intein/chitin-binding domain (New England Biolabs) on plasmid pJC724 and sequenced to ensure that no errors in the coding region were introduced by PCR. Production of RoxR from E. coli ER2566 carrying this plasmid was induced with 0.3 mM IPTG for 4.5 h at 30°C, after which the cells were harvested by centrifugation at 6000 g for 10 min at 4°C. The cell pellets were processed, and RoxR was purified as described previously for PrrA (Comolli et al., 2002). Purified protein was dialysed into 50 mM Tris-Cl, pH 8.0, 0.2 M KCl, 10 mM MgCl2, 1 mM EDTA, 1 mM dithiothreitol (DTT) and 20% glycerol before storage.

Phosphorylation assay

6×His-cPrrB was purified and phosphorylated using 25 000 c.p.m. pmol−1 [γ-32P]-ATP for 2 h at 30°C (Comolli et al., 2002). Phosphorylation reactions that contained 16 μM cPrrB, 50 mM Tris-Cl, pH 7.5, 50 mM KCl, 5 mM MgCl2 and 50 μM ATP were diluted with 0.5 volumes of 8 μM RoxR, 16 μM RoxR or 16 μM PrrA. After 2 min at 30°C, the reactions were stopped by the addition of 3× sample buffer (188 mM Tris-Cl, pH 6.8, 6% SDS, 30% glycerol, 0.3 M DTT, 0.03% bromophenol blue) plus 50 mM EDTA. The samples were separated by SDS–PAGE and analysed using a phosphorimager (Molecular Dynamics).

In vitro transcription

The assay was performed as described previously (Comolli et al., 2002) using purified R. sphaeroides RNAP holoenzyme and plasmid pRKK149, which contained sequences from −228 to +22 relative to the cycA P2 transcription start site (Karls et al., 1999). Where indicated, RoxR or PrrA was incubated with 25 mM acetyl phosphate or buffer for 60 min at 30°C before its addition to the transcription reaction. Each transcription reaction was initiated with a nucleotide mix containing [α-32P]-UTP, incubated for 20 min at 30°C and terminated with 0.5 volumes of loading buffer (95% formamide, 20 mM EDTA, 0.05% xylene cyanol, 0.05% bromophenol blue). The reaction products were visualized by electrophoresis on a 6% polyacrylamide–urea gel and phosphorimager analysis. Transcripts were quantified by determining the number of pixels in each product and subtracting the background present in that lane. These values were then standardized to the amount of the control RNA1 transcript generated.

Measurement of oxidase activity

Membranes were prepared as described previously (Cunningham and Williams, 1995) from PAK, RoxS1, RoxR1 or CioAB1 grown aerobically in LB with or without 100 μg ml−1 KCN except that cells were lysed by sonication. Protein concentrations of the membrane samples were determined by Lowry assay (Lowry et al., 1951). NADH-dependent oxidase activity was measured using a Clark-type oxygen electrode (Yellow Springs Instrument). Membrane samples (≈ 600 μg of protein) were diluted into 5 mM KPO4, pH 7.1, and 5 mM MgCl2 and stirred rapidly. After the injection of NADH to 0.5 mM, the rate of O2 consumption (saturated solution estimated to contain 250 μM O2) was determined using linear regression of the O2 content over time. To measure oxidase activity in the presence of cyanide, 0.1 mM KCN was added before the injection of NADH. Values represent the average of three separate measurements with error being the standard deviation. The percentage of CIO activity is the portion of total NADH-dependent oxidase activity remaining in the presence of KCN.

Construction of reporter plasmid and β-galactosidase assays

A 679 bp fragment containing the cioA promoter region (Cunningham et al., 1997) was amplified from pJC494 using the Cio3 and Cioup-2 (5′-ATATATGGATCCGGAAGGGTCGC AAGCAGATTC-3′) primers and cloned into pBSII-KS to create pJC496. A portion of the cioAB promoter (−378 to +55 relative to the start site of cioA transcription; Cunningham et al., 1997) was removed with SalI and BamHI and fused to lacZ on the transcription reporter plasmid pLP170. The resulting plasmid, pJC701, was introduced into the appropriate P. aeruginosa strain by electroporation. β-Galactosidase assays were performed as described previously (Comolli et al., 2002) using cells grown aerobically to mid-exponential phase. Values are the average of assays of three separate cultures with error being the standard deviation. Strains containing pLP170 without an insert reproducibly contained less than 50 Miller units of activity.

Gel shift assay

A 230 bp fragment containing the region upstream (−229 to +55) of the start site of cioAB transcription (Cunningham et al., 1997) was generated by PCR from pJC494 using the Cioup-2 and CioABshift-1 (5′-CATTCTTCGATCCTTGT AGGGC-3′) primers. The product (5 pmol) was end-labelled with 10 μCi of [γ-32P]-ATP using T4 polynucleotide kinase. Before the binding reactions, 20 μM RoxR was incubated for 5 min at 30°C with 10 μM cPrrB that had been treated with 1 mM ATP or buffer only for 60 min at 30°C. Each 10 μl binding reaction contained RoxR, 2 fmol of 32P-labelled DNA in binding buffer (20 mM Tris-Cl, pH 7.9, 5 mM MgCl2, 10% glycerol), 1 mM DTT, 0.5 mg ml−1 BSA and 0.5 μg of poly-(dI–dC). The binding reactions were incubated for 30 min on ice, 2 μl of 0.05% bromophenol blue in binding buffer was added, and the samples were run on a 7% 1× Tris–glycine (25 mM Tris, 0.2 M glycine, 1 mM EDTA) polyacrylamide (37.5:1 bis) gel containing 5% glycerol and 5 mM MgCl2. The gel was then dried and analysed with a phosphorimager.


We appreciate the valuable laboratory assistance provided by D. Baker. Also, we would like to thank L. Passador for providing pLP170, H. Schweizer for pEX100T, pX1918GT and pUCP26, J.-I. Oh for pNHG1, and J. Eraso for providing R. sphaeroides PrrB1 and PrrA2, all of which were used in this work. We are also grateful to the P. aeruginosa genome website ( for providing access to the PAO1 sequence database. This study was funded in part by an NRSA GM 20344 from the National Institutes of Health (NIH) to J.C.C. and by NIH grant GM 37509 to T.J.D.