RegB/RegA, a Highly Conserved Redox-Responding Global Two-Component Regulatory System

In many species of purple photosynthetic bacteria, synthesis of the photosynthetic apparatus occurs predominantly under conditions of low oxygen tension (12). To define the mechanism behind redox regulation of photosystem synthesis, genetic screens were used to isolate R. capsulatus mutants that have reduced pigmentation after prolonged growth under dark anaerobic conditions. This screen led to the isolation of regA and regB mutants exhibiting a defect in anaerobic induction of the photosystem (55, 78). Null mutations in regB and regA were subsequently shown to be defective in high-level anaerobic expression of the puh, puf, and puc operons that encode apoproteins for the light-harvesting I, light-harvesting II, and reaction center complexes (55, 78). The similar phenotypes displayed by these mutants led to the hypothesis that they may be cognate trans-acting partners constituting a two-component regulatory system. This was confirmed by sequence analysis which demonstrated that RegB exhibits homology to histidine protein kinases (55, 70, 81, 82) and that RegA exhibits homology to DNA-binding response regulators (70, 78, 81, 82). Disruption of regB affected photosystem gene expression less dramatically than did disruption of regA, suggesting that another sensor kinase(s) may also phosphorylate RegA (55) at a low level or that RegA may obtain phosphates from small-molecule donors, such as acetyl phosphate or carbamyl phosphate (82).

Soon after the discovery of RegB/RegA from R. capsulatus, homologues were found in Rhodobacter sphaeroides that were called either RegB/RegA (71) or PrrB/PrrA (29, 30). Mutations in the R. sphaeroides RegB/RegA homologues exhibited defects in photosystem synthesis that were similar to those caused by RegB/RegA mutations in R. capsulatus (29, 30, 71). Homologous two-component regulatory systems have also been genetically characterized in many other species such as the RegS/RegR system from Bradyrhizobium japonicum (6), ActS/ActR from Sinorhizobium meliloti (89), RoxS/RoxR from Pseudomonas aeruginosa (15), and RegB/RegA from Rhodovulum sulfidophilum and Roseobacter denitrificans (53). Genome sequence studies have identified RegA and RegB homologues in many other photosynthetic as well as nonphotosynthetic α- and γ-proteobacterial species (Fig. 1 and 2, respectively).

Sequence analysis indicated that regB and regA genes from most species are localized to a similar region of the chromosome. In R. capsulatus, R. sphaeroides, R. sulfidophilum, and R. denitrificans, the regB gene is divergently transcribed from a putative senC-regA-hvrA operon (10, 29, 30, 31, 53, 55). A variation occurs in Caulobacter crescentus, where a putative regB-regA operon is transcribed divergently from senC. The other organisms listed in Fig. 1 and 2 have regB and regA genes linked in a single putative regB-regA operon without senC or hvrA genes being found within close proximity on the chromosome (details of SenC and HvrA are discussed below).

As shown in the sequence alignments of RegB (Fig. 2) and RegA (Fig. 1) homologues, RegB and RegA proteins are highly conserved among photosynthetic and nonphotosynthetic α- and γ-proteobacteria. The RegA homologues can be placed into two groups based on the sequence similarity of their output domain. Group 1 members have 100% conservation of the HTH DNA-binding motif, while group 2 homologues have a similar but nonidentical HTH region (Fig. 1). Overall, group 1 homologues exhibit a remarkable 78.7 to 84.2% sequence identity to RegA from R. capsulatus, with 93% overall identity in the C-terminal output domain. This level of conservation suggests that there are significant constraints on the variability of RegA to undergo mutational change in these species. In contrast to extensive analysis of group 1 homologues, there has been limited characterization of group 2 homologues. Mutational analysis of roxR, which is a group 2 RegA homologue in P. aeruginosa, indicates that RoxR controls respiratory gene expression in P. aeruginosa as does the RegB/RegA system in R. capsulatus (15, 83). Interestingly, several photosynthetic and nonphotosynthetic species, such as C. cresentus, R. capsulatus, and Mesorhizobium loti, contain both group 1 and group 2 homologues (Fig. 1).

Among RegB homologues, there are also two identifiable groups composed of similar members, as discussed for RegA homologues (Fig. 2). Differences between group 1 and group 2 RegB homologues occur in the regions between transmembrane 6 and the H box (amino acids 245 to 269), between the redox box and the N box (amino acids 335 to 355), and at the carboxyl terminus (amino acid 455 to end) (Fig. 2).

In addition to RegB and RegA proteins having a high level of sequence identity, they have functional similarity. For example, Emmerich et al. (25) demonstrated that RegB and RegA homologues from R. capsulatus, B. japonicum, and S. meliloti are interchangeable. More precisely, phosphotransfer was observed between the B. japonicum RegB homologue (RegS) and R. capsulatus RegA. The RegA homologue from B. japonicum (RegR) was shown to bind to the R. capsulatus puf and puc operon promoters. R. capsulatus RegA, and its homologue from S. meliloti (ActR), can activate transcription of the fixR nifA operon from B. japonicum in vivo (25). Furthermore, Comolli and Donohue (15) demonstrated that RegA homologues from R. sphaeroides (PrrA) and P. aeruginosa (RoxR) also succeeded in heterologous complementation. These studies demonstrate that homologous Reg proteins are interchangeable in vitro as well as in vivo.

Mutational analysis of several α-proteobacterial RegA homologues indicates that they control a similar set of target genes such as respiratory and nitrogenase genes, details of which are discussed below. Thus, it appears that the RegB/RegA system is involved in global regulation in many diverse species of bacteria.

Initial kinase assays demonstrated that a His-tagged cytosolic domain of R. capsulatus RegB was capable of autophosphorylation in vitro as well as phosphotransfer to its cognate response regulator, RegA. The rate of autophosphorylation was initially reported to be low, with half-maximal phosphorylation observed after 45 min of incubation with [γ-³²P]ATP (7, 41). Since the kinetics were not affected by the ATP concentration, it was suggested that the rate-limiting step of RegB autophosphorylation was phosphotransfer from bound ATP to the histidine residue or the dimerization of the protein, rather than binding of ATP (7). Recent results from Swem et al. (84) demonstrate that a non-His-tagged (truncated) version of RegB exhibits a significantly higher autophosphorylation rate, reaching half-maximal phosphorylation within 5 min. Similar results were seen with truncated cytosolic forms of RegB homologues from R. sphaeroides (14) and B. japonicum (26), as well as with a recently purified full-length version from R. sphaeroides (74). Phosphorylated full-length RegB exhibits decreased the stability of the phosphate compared to the truncated version of RegB (half-lives of about 34 min and 5.5 to 6 h, respectively), pointing to a role of the transmembrane domain in regulating the phosphorylation state of RegB (14, 74). Since mutations in the N-terminal transmembrane domain led to RegB proteins with constitutive kinase activity in vivo, the “unsignaled” state of the protein has been proposed to be “autophosphorylation dominant” (30, 67).

The first demonstration of phosphotransfer from R. capsulatus RegB∼P to RegA was reported by Inoue et al. (41). Phosphotransfer studies showed that the transfer of phosphate from the cytosolic domain of RegB to RegA in vitro is rapid (<1 min) (7). Similar phosphotransfer kinetics have been reported by investigators using truncated cytosolic and full-length versions of R. sphaeroides RegB homologues. No back-transfer of phosphate from RegA∼P to RegB has been observed (14, 74). A soluble truncated form of the RegB homologue (RegS_C) in B. japonicum was shown to autophosphorylate at residue His219 in the presence of [γ-³²P]ATP and to transfer the phosphate to the Asp63 residue of the RegA homologue (RegR) (26). Since phosphotransfer is very rapid, autophosphorylation of RegB appears to be the rate-limiting step in the phosphorylation of RegA.

Like many histidine kinases, RegB can modulate the level of phosphorylated RegA, not only through phosphorylation but also by exerting phosphatase activity on RegA∼P. Dephosphorylation of RegA∼P in vitro is dependent on the amount of unphosphorylated RegB, which is a good indication that RegB possesses phosphatase activity toward RegA (7). Similar results were observed for the B. japonicum RegB homologue (RegS) (26) and for the full-length version of RegB from R. sphaeroides (74). Phosphatase activity has also been characterized for a truncated soluble form of R. sphaeroides RegB by measuring the stability of the phosphate on RegA∼P in the presence and absence of RegB. Results showed that the presence of RegB resulted in a >16-fold reduction in the stability of the phosphate on RegA∼P (14).

Because both truncated and full-length RegB exhibit the same phosphatase activity, it is assumed that modulation of phosphatase activity does not require the N-terminal domain of RegB (74). However, further studies are required to determine if phosphatase activity is redox regulated.

In the E. coli sensor kinase EnvZ, it is apparent that a threonine residue positioned 4 residues downstream of the conserved phosphorylated histidine is important for phosphatase activity (42). This threonine residue is positioned on the same α-helical face directly below the phosphorylated histidine. It is thought that the histidine residue may deprotonate the threonine hydroxyl group to form a good nucleophile that attacks the phosphoryl group bound to the aspartate residue of the cognate response regulator. Interestingly, RegB homologues also contain a 100% conserved threonine residue located 4 residues downstream of the phosphorylated histidine (Fig. 2), suggesting that RegB may exhibit a similar phosphatase mechanism with RegA. Nonetheless, Potter et al. (74) point out that a significant difference between RegB and EnvZ phosphatase activity exists, given that EnvZ has a requirement for ATP or a nonhydrolyzable analogue as a cofactor for phosphatase activity while RegB lacks this nucleotide requirement.

The RegB/RegA system was originally described as a two-component regulatory system that was required for anaerobic activation of photosynthetic gene expression (55, 78). For some time, it has been presumed that RegB kinase activity is directly inhibited by oxygen. Although this may be true to some degree, it is also clear that RegB is capable of responding to additional redox signals beyond the simple presence or absence of oxygen. For example, R. capsulatus is fully capable of derepressing pigment biosynthesis under chemiautotrophic growth conditions involving growth in the presence of oxygen, hydrogen, and carbon dioxide (51). This clearly indicates that RegB is not directly affected by oxygen and, instead, may simply respond to the redox state of the cell.

DNA-binding activity of RegA was initially demonstrated by using a constitutively active variant of RegA called RegA* (18). RegA* was isolated in a RegB-null mutant strain by selecting for a suppressor mutation that exhibited elevated photosynthesis gene expression in the absence of RegB. Analysis of RegA* revealed that this form of RegA was capable of binding DNA and activating transcription independently of phosphorylation (7, 8, 18). Sequence analysis of RegA* revealed a mutation at amino acid 95 that substitutes serine for a conserved alanine. It is hypothesized that this mutation alters the conformation of RegA such that it mimics the phosphorylated state of the wild-type protein (7). Interestingly, Asp63 of RegA* is still capable of accepting a phosphate from RegB, with the phosphoryl group on RegA* being much more stable than the phosphoryl group on wild-type RegA (7). As a consequence of a shift in equilibrium between the phosphorylated and unphosphorylated states of the proteins, RegA* is capable of being phosphorylated to a greater extent in vitro than is wild-type RegA.

The inherent stability of the aspartyl-phosphate among the RegA homologues, as a measure of half-life, is high relative to that observed with other response regulators. The half-life was about 90 min for R. capsulatus RegA∼P (7), about 162 min for the RegA homologue (RegR∼P) from B. japonicum (26), and 330 min for R. sphaeroides RegA∼P (14). RegA from R. sphaeroides has also been shown to use the small phosphodonor acetyl∼P in vitro (14). Autophosphorylation of R. capsulatus RegA in the presence of acetyl∼P was also reported by Hemschemeir et al. (39).

DNase I footprint analysis of RegA binding to the puc promoter region revealed that phosphorylated and unphosphorylated wild-type RegA, and RegA* protect identical regions with different affinities. This observation indicates that phosphorylation does not affect the points of protein-DNA interaction but, rather, affects the affinity for the binding site (7). Indeed, phosphorylation is reported to increase the DNA-binding affinity of RegA by at least 16-fold. Interestingly, unphosphorylated RegA* exhibits DNA-binding activity comparable to that of phosphorylated wild type, supporting in vivo evidence that RegA* is able to promote transcription activation even in the absence of phosphotransfer from RegB. A further six-fold increase in the DNA-binding activity of RegA* is achieved upon phosphorylation (7). These results are consistent with what was reported for the B. japonicum RegA homologue (RegR), whose DNA-binding activity is increased by at least eightfold on phosphorylation (26). Phosphorylation-induced stimulation of DNA-binding activity was also reported for P. aeruginosa RoxR (15). However, Hemschemeier et al. (40) have reported the same affinity (K_D, 50 to 70 nM) for phosphorylated and unphosphorylated RegA when using the puc promoter as a probe in vitro. The reason for this discrepancy is not clear. In addition, they showed that deletion of the last 13 amino acids, which contain the DNA-binding motif of RegA, inhibited DNA-binding ability in vitro as well as the activation of puc and puf expression in vivo. This confirms the importance of DNA binding for the transcriptional activation activity of RegA.

Analysis of mutations at the site of phosphorylation (Asp63) in R. capsulatus RegA, as well as with RegA homologues from B. japonicum and R. sphaeroides, has been performed. A D63N mutation in RegR of B. japonicum rendered the protein unable to be phosphorylated and also unable to bind DNA, as demonstrated by gel retardation experiments with a fixR-nifA promoter probe (26). In contrast, experiments involving a D63K RegA mutant from R. capsulatus showed that the mutant protein was capable of binding DNA, despite an inability to be phosphorylated. This indicated that phosphorylation might not affect the DNA-binding ability but, rather, might facilitate a conformational change allowing appropriate interactions of RegA with RNA polymerase (38). This theory was further supported by in vitro transcription assays by Comolli et al. (14), involving wild-type PrrA and a D63A mutant of PrrA from R. sphaeroides. Their studies revealed that both unphosphorylated and phosphorylated wild-type PrrA are able to activate the transcription of cycA P2, with phosphorylated PrrA exhibiting greater activity than unphosphorylated PrrA. Interestingly, the D63A form of PrrA was unable to activate any detectable amounts of transcription. These data suggest that phosphorylation of Asp63 and the mere presence of Asp63 may affect several steps of activation such as the DNA binding of RegA and the interaction of RegA∼P with RNA polymerase.

Superimposed on the effect of phosphorylation are numerous in vivo observations that unphosphorylated RegA is also capable of affecting transcription. Specifically, mutational analysis indicates that phosphorylated RegA functions as an anaerobic repressor of cytochrome cbb₃ oxidase expression while unphosphorylated RegA functions as an aerobic activator of cytochrome cbb₃ oxidase (83, 85). In addition, both phosphorylated and unphosphorylated RegA are involved in activation of ubiquinol oxidase expression (85), activation of cbb operon expression (75), and repression of hupSLC expression (23). Additional DNA-binding studies with both phosphorylated and unphosphorylated RegA are clearly needed to obtain an understanding of the mechanism of activation or repression by unphosphorylated RegA.

DNase I footprint analysis initially demonstrated that purified wild-type RegA and RegA* bind to identical specific sites in the puf and puc promoters (18, 40). Subsequent DNase I footprint analysis demonstrated that RegA also binds to a plethora of other operons under the direct regulation of the RegB/RegA system, including nifA2, hupSLC (23), regB, senC-regA- hvrA (19), petABC, cycA, cycY, cydAB, ccoNOPQ (83), cbbI and cbbII (93), and cheOp2 (77). DNA-binding activity has also been reported for the phosphorylated forms of RegR in B. japonicum (26) and RoxR in P. aeruginosa (14) by using gel mobility retardation experiments.

The number of RegA DNA-binding sites on target promoters ranges between 1 and 6, based on DNase I protection assays (Fig. 3). Individual sites in the promoter regions have different affinities, as determined by the amount of RegA needed to obtain half-maximal protection of individual sites (83). This finding leads to a classification of RegA-regulated promoters into three types based on the location of the RegA-binding sites. The first type, exemplified by the R. capsulatus puf promoter, has RegA DNA-binding sites that extend from bp −22 to −80 relative to the start site of transcription (Fig. 3). The second type, comprising the R. capsulatus puc, nifA2, cydAB, ccoNOQP, and cbb_II promoters and the R. sphaeroides cbb_I, cbb_II, and cycA promoters, possesses several DNA-binding sites that extend from approximately bp −50 to −890 upstream from the transcriptional start site. The third type, represented by the R. capsulatus cbb_I, petABC, cycA, and cycY promoters, contains one site overlapping the start site of transcription (from bp −19 to +22) as well as DNA-binding sites from bp −50 to −225. Although it is possible that additional DNA-binding sites have been left unrevealed, these different binding locations may allow RegA to interact with RNA polymerase in more than one manner at these promoters. This is especially true in comparisons of promoters that have RegA DNA-binding sites at the region from bp +20 to −50 with promoters that have binding sites just in the region from bp −300 to −890. For example, RegA binding to the puc promoter is centered at around −60, which presumably favors the interaction of RegA with the α subunit of RNA polymerase (18). In contrast, RegA binds to the region of the puf promoter from bp −22 to −52 (18), which presumably favors interaction with the σ subunit of the RNA polymerase. There is also RegA binding to the region of the puf promoter from bp −68 to −80, which would favor an additional interaction with the α subunit of the RNA polymerase. These interactions should overcome the presence of poor σ⁷⁰ recognition sequences harbored by the puf promoter, which uses the σ⁷⁰ housekeeping RNA polymerase (8). Clearly, additional detailed analyses of the mechanism of RegA activation and repression at these different classes of promoters is warranted, and indeed needed, to better understand how RegA interacts with RNA polymerase and controls gene expression.

Initially it was hypothesized that RegA recognizes structural features of the DNA rather than a specific nucleotide sequence for DNA binding. Lack of sequence specificity was proposed as a result of the failure of early DNA-binding studies to recognize a consensus sequence that had a limited number of protection sites for analysis (18). However, the alignment of 21 RegA-binding sites from R. capsulatus and R. sphaeroides revealed that RegA indeed binds to a consensus sequence of 5′-G(C/T)G(G/C)(G/C)(G/A)NN(T/A)(T/A)NNC(G/A)C-3′ (83). An in vitro method was also used to determine a DNA-binding consensus sequence for RegR from B. japonicum (28). The selection of RegR-interacting oligonucleotides from a randomized oligonucleotide pool (SELEX), coupled with site-directed mutagenesis of known RegR DNA-binding sites on the fixR nifA promoter, revealed a “RegR box,” 5′-GNG(A/G)C(A/G)TTNNGNCGC-3′ that contains an imperfect inverted repeat with 4 nucleotides per half-site. The 5-bp spacing sequence was shown to also contain critical nucleotides for RegR DNA-binding activity. The RegR box is similar to the consensus sequence obtained from R. capsulatus RegA-protected promoter sequences, which is not surprising given that there is total identity of the putative HTH DNA-binding motifs between RegA and RegR (53).

A nuclear magnetic resonance (NMR) spectral structure was recently solved for the DNA-binding output domain of RegA bound to DNA (49). This structure demonstrated that the RegA DNA-binding domain is composed of a three-helix bundle encompassing an HTH motif (49). The three helices within the three-helix bundle are labeled in Fig. 1 as α6, α7, and α8, with α7 and α8 comprising the predicted DNA-binding HTH motif. This structure is similar to the three-helix bundle found in the E. coli Fis family of DNA-binding proteins (49). This DNA-binding effector domain of RegA is more structurally similar to Fis than to the effector domain of the response regulator, NtrC (49). The NMR structure has also provided new insights into the DNA recognition and DNA-binding ability of RegA. As demonstrated previously by DNase I protection assays and SELEX (systemic evolution of ligands by exponential enrichment) RegA has a distinct affinity for a GCG…CGC palindromic sequence (28, 83). The NMR structure of a RegA-DNA cocomplex has confirmed that the RegA consensus recognition sequence consists of YGCGRCR_xTATA_xGNCGC (with x denoting a variable number of bases) (49). Surprisingly, the GCG inverted repeat can be separated by a 3- to 9-nucleotide AT-rich spacer region (49). This AT-rich spacer region could presumably allow for DNA bending, which aids RegA recognition and binding (49). The NMR structure clearly identifies the α8 helix and the DNA recognition helix with this helix binding specifically to the GCG inverted repeat sequence within the major DNA groove. The α6 helix appears to undergo nonspecific interactions with the DNA phosphate backbone (49). These data provide a very revealing picture of how RegA recognizes and binds to the numerous promoters under RegB/RegA transcriptional control.

In vitro transcription assays demonstrated that housekeeping σ⁷⁰-RNA polymerase in the presence of purified RegA* can activate transcription from puc and puf promoters (8). Unphosphorylated RegA* was shown to stimulate puc expression fourfold over basal-level expression observed in the absence of RegA. Phosphorylation of RegA* by RegB∼P leads to a further 12- to 60-fold increase in transcription, depending on the RNA polymerase concentration (8). This result is surprising, given that RegA* is fully active in vivo in the absence of RegB. It was also observed that only the major RegA-binding site was required for RegA-dependent in vitro transcription, even though the puc promoter has both major and minor DNA-binding sites for RegA (7, 18).

As detailed in the following section, it is a common occurrence that other transcription factors regulate operons found within the RegB/RegA regulon. For instance, RegA regulates the transcription of photosystem promoters in concert with the histone-like proteins, integration host factor (IHF), and HvrA (10, 46, 58), as well as with CrtJ (8, 24, 72, 73) and AerR (17) (details of these additional transcription factors are covered in references 3 through 5). The occurrence of additional regulators of genes that are also controlled by RegA is common. Thus, RegA constitutes a global regulator that provides an overarching layer of redox control on operons that have additional transcription factors that respond to photosynthesis, respiratory electron transport, organic carbon, fixed nitrogen, hydrogen, etc.

The RegB/RegA system was discovered by selecting for mutations that exhibited reduced synthesis of the photosystem ini R. capsulatus (55, 78). An intact copy of regA was shown to be absolutely required for R. capsulatus to grow photosynthetically under dim light, which is a growth condition that requires maximum transcription of the puh, puf, and puc operons, which code for apoproteins of the light-harvesting and reaction center complexes. As is the case for RegA, RegB is also necessary for anaerobic induction of the puc, puf, and puh operons (55). Mutations in the RegB and RegA homologues from R. sphaeroides show similar effects with respect to the control of puh, puf, and puc expression, with the difference that RegA is indispensable for photosynthetic growth under all light intensities (29).

As is the case for many RegB/RegA-regulated promoters, numerous other transcription factors have been found to control puf, puh, and puc operon expression in R. capsulatus (3, 4). DNase I footprint assays showed that CrtJ and RegA compete for binding to overlapping sites in the puc promoter (8). An additional aerobic repressor called AerR also represses the puf operon; the site of AerR repression has not yet been determined (17). In addition to CrtJ and AerR, the H-NS-like protein HvrA regulates the puf and puh promoters, where it functions as an activator in R. capsulatus and as a repressor in R. sphaeroides (10, 59, 80). The location of HvrA binding in R. sphaeroides is unknown, but in R. capsulatus, the HvrA DNA-binding site is adjacent to RegA DNA-binding sites (48). In both R. capsulatus and R. sphaeroides, the histone-like protein IHF is required for maximal puc operon transcription (58, 96). FnrL in R. sphaeroides also anaerobically activates puc expression, whereas inactivation of fnrL in R. capsulatus has no noticeable affect on photosynthetic growth (reviewed in reference 96). Thus, there are numerous potential protein-protein interactions that might exist between RegA and a number of other transcription factors that control these photosystem promoters. Currently, little is known about the nature of these interactions or their importance in controlling photosystem gene expression.

In addition to controlling the synthesis of light-harvesting and reaction center apoproteins, RegA affects the expression of the bchE promoter in R. sphaeroides, which encodes an enzyme in the bacteriochlorophyll biosynthesis pathway (62). There are also reports that RegA from R. sphaeroides controls the expression of hemA, hemZ, and hemN (62, 65), which encode one δ-aminolevulinic acid (ALA) synthase isoenzyme and two coproporphyrinogen III oxidases, respectively: enzymes involved in the Mg- and Fe-tetrapyrrole biosynthetic pathways. Abada et al. (1) have also reported an involvement of the R. capsulatus RegB/RegA system in bacteriochlorophyll-dependent expression of the puc, puf, and bchC operons. Therefore, not only does RegB/RegA control the synthesis of the photosystem and cytochrome apoproteins (the role of RegB/RegA in cytochrome biogenesis is discussed below), but also they control synthesis of bacteriochlorophyll and heme that are bound by these respective apoproteins.

An important discovery was made in 1994 that hinted at the global nature of the RegB/RegA signal transduction cascade. The RegA homologue from R. sphaeroides (PrrA) was found to positively regulate the expression of cycA, which encodes cytochrome c₂ (29). Cytochrome c₂ is an important element of the electron transfer system shared by several energy-generating pathways including those of photosynthesis and respiration. Primer extension and in vitro transcription studies have indicated that PrrA positively controls the P2 promoter, which is one of three promoters driving the expression of cycA under both aerobic and anaerobic conditions. Expression from P2 is responsible for basal expression under aerobiosis as well as for induction under anaerobic conditions (45). Evidence that RegA directly controls cycA expression was also provided by DNase I protection assays, which showed that RegA* from R. capsulatus binds to a region of P2 centered at bp −50 from the start site of transcription (45). Furthermore, in vitro transcription assays confirmed that R. sphaeroides PrrA directly activates cycA transcription (14).

Swem et al. (83) demonstrated that RegB/RegA controls synthesis of cytochrome c₂ as well as cytochrome c_y and the cytochrome bc₁ complex in R. capsulatus. The expression patterns of these different cytochrome genes were compared in wild-type and regA-disrupted strains, which revealed that RegA activates the biosynthesis of cytochromes bc₁ and c₂ under anaerobic, semiaerobic, and aerobic growth conditions, whereas it activates cytochrome c_y only under semiaerobic and anaerobic conditions. DNase I protection assays also demonstrated that RegA binds to two sites on the promoter of the pet (bc₁) operon and to four sites on the promoters of the cycA and cycY genes encoding cytochrome c₂ and cytochrome c_y, respectively (83) (Fig. 3). These three promoters belong to the type III class of RegA-regulated promoters (discussed above), where there is one DNA-binding site that overlaps the start site of transcription and one site located just upstream from the −35 promoter region. The in vivo involvement of these different RegA-binding sites still needs to be determined.

Differences in expression patterns of these various cytochromes suggest that additional transcription factors may also regulate the expression of these genes. For example, evidence suggests the involvement of another protein besides RegA, that regulates cycA expression in the absence of O₂ in R. sphaeroides (45). There is also a response regulator and a putative repressor located just upstream of the pet operon, which is suspected to be involved in controlling the expression of cytochrome bc₁ apoproteins (91). Thus, as with other cellular processes, RegB/RegA appears to be just one component of a more complex regulatory network that controls the biosynthesis of cytochrome apoproteins.

Like many bacterial species, R. capsulatus possesses a branched respiratory chain involving two different terminal oxidases. In one branch, the ubiquinol (ubihydroquinone) oxidase takes electrons directly from the quinone pool to reduce O₂ to H₂O. The second branch, which is similar to the mitochondrial electron transfer chain, is composed of the cytochrome bc₁ complex, cytochromes c₂ or c_y, and a cbb₃-type cytochrome c oxidase (37). Cytochromes c_y, c₂, and bc₁ are also involved in photosynthetic electron transfer events, shuttling electrons from cytochrome bc₁ back to the photosynthetic reaction center either by the membrane-associated cytochrome c_y or by the soluble cytochrome c₂ (37).

Both of the terminal oxidases are maximally synthesized under semiaerobic growth conditions. However, the two oxidases are differentially regulated in regards to high and no (or very low) oxygen levels, with the ubiquinol oxidase exhibiting a low level of expression under aerobic conditions and a higher level under anaerobic conditions. The converse is true for cytochrome cbb₃ oxidase, which exhibits higher expression under aerobic than anaerobic growth conditions (83, 85). This expression pattern indicates that cytochrome cbb₃ oxidase from R. capsulatus may have a lower affinity for oxygen than does the ubiquinol oxidase (87, 95).

RegA has been observed to activate cytochrome cbb₃ oxidase expression semiaerobically and aerobically while repressing expression anaerobically. The mechanism and the significance of this observation are not yet well understood (83). DNase I footprint analysis revealed that RegA directly controls the synthesis of cytochrome cbb₃ oxidase by binding to a site on the ccoNOQP promoter located just upstream from the −35 sequence (Fig. 3) (83). Although there was an early report that FnrL does not affect cytochrome cbb₃ oxidase synthesis in R. capsulatus (96), more recent analysis indicates that expression of ccoNOPQ is indeed lower in an fnrL-disrupted mutant under semiaerobic and anaerobic conditions (85).

The expression pattern of the cydAB operon, encoding ubiquinol oxidase, suggests that the enzyme has a higher affinity for oxygen than does cytochrome cbb₃ oxidase (83, 85). As observed for the ccoNOQP operon, RegB/RegA is involved in the regulation of cydAB expression, with RegA being required for activation of cydAB transcription under all growth conditions tested. DNase footprint assays indicate that RegA binds to two sites upstream from the −35 region of the promoter (Fig. 3) (83).

Recently, RegB/RegA homologues from P. aeruginosa (RoxS/RoxR) were reported to be involved in the control of aerobic respiration (15). More precisely, RoxS/RoxR controls the induction of the cyanide-insensitive oxidase in the presence of cyanide. It is proposed that RoxR coregulates the cioAB promoter with another anaerobic regulator, ANR, thereby permitting the integration of different stimuli in the control of cyanide-insensitive oxidase expression.

R. capsulatus and R. sphaeroides are both capable of anaerobic respiration using dimethyl sulfoxide (DMSO) as a terminal electron acceptor (95). The reduction of DMSO is catalyzed by a membrane-bound DMSO reductase enzyme that is encoded by the dorCDA operon. The dor operon is under the transcriptional control of a two-component signal transduction system, DorS/DorR, that responds to the availability of DMSO (56, 57, 79). The sensor kinase, DorS, is known to autophosphorylate in the presence of DMSO, with the phosphate transferred to the response regulator, DorR, which then activates dorCDA expression. The dorCDA operon is known to also be under the control of the RegB/RegA system, with RegA acting as a repressor of the dorCDA operon during photoheterotrophic growth in the presence of malate as a carbon source (44). However, RegA seems to lose control of the dorCDA operon if the cells are grown on pyruvate rather than malate. This indicates that another, unidentified, regulator can suppress the regA mutant phenotype in cells grown on pyruvate but not in cells grown on malate. It is not yet known if the effect of RegA on the dorCDA operon is direct or indirect. However, since no obvious RegA DNA-binding sequence has been found in the dorCDA promoter, the influence of RegA on dorCDA expression may be indirect. Nonetheless, this appears to be another instance in which the RegB/RegA system exerts transcriptional control over a system that is responsible for energy generation.

The Calvin-Benson-Bassham reductive pentose phosphate pathway allows the production of organic carbon via the assimilation of CO₂. Carbon fixation also plays an important role under photoheterotrophic growth conditions, where it acts as an electron sink that is needed to balance the redox potential of the cell. Consequently, mutants of R. capsulatus and R. sphaeroides that are devoid of a functional Calvin cycle do not grow photoheterotrophically unless exogenous electron acceptors such as DMSO are provided (reviewed in references 86 and 89 and references therein).

Enzymes of the Calvin cycle are encoded by the cbb_I and cbb_II operons. Transcription of these operons is regulated in response to carbon by the transcriptional activator CbbR, which is a member of the LysR family of transcription factors. CbbR is absolutely required for expression of the cbb_I operon, since inactivation of the R. sphaeroides cbbR gene leads to the absence of transcription through the cbb₁ operon and to a strong reduction of cbb_II expression (34). CbbR directly regulates cbb_I expression by binding to two sites located in the promoter-proximal region, as demonstrated by in vitro DNase I footprint experiments (20, 21).

An involvement of the RegB/RegA system in the biosynthesis of Calvin cycle enzymes was first discovered in R. sphaeroides by screening for mutants that derepress an alternative CO₂ fixation pathway (75). From studies of these mutants, it was demonstrated that RegB (PrrB) of R. sphaeroides was required for positive regulation of the cbb operons, both anaerobically in the light and aerobically in the dark (75). Using purified R. capsulatus RegA*, Dubbs et al. (20, 22) demonstrated that RegA directly controls R. sphaeroides cbb expression by binding to four sites in the cbb_I promoter and to six sites on the cbb_II promoter (Fig. 3). The authors hypothesized that the locations of RegA binding could allow direct interactions with CbbR and/or with RNA polymerase. Furthermore, binding of RegA to the two sites located in the upstream activating sequence in the cbb_I promoter appears responsible for a RegA-mediated 41-fold enhancement in cbb_I expression (20).

Gibson et al. (33) demonstrated that chemoautotrophically grown regA (prrA) mutants of R. sphaeroides differentially express the two cbb operons with expression of the cbb_II promoter being severely reduced and expression of the cbb_I promoter being enhanced in the prrA mutant strain. This result indicates that PrrA functions as an activator of cbb_II and a repressor of cbb_I. Analysis of promoter mutants suggests that RegA may bind to distinct regions in cbb_II and in cbb_I during photoautotrophic and chemoautotrophic growth.

In R. capsulatus, the RegB/RegA system also controls the expression of the two cbb operons that are present in this species (93). The cbb_I and cbb_II operons are regulated by cognate CbbR proteins encoded by the cbbR_I and cbbR_II genes. The cbbR_I and cbbR_II genes are located upstream of, and are divergently transcribed from, the cbb_I and cbb_II operons, respectively. The CbbR_I protein is able to control its own expression (down regulation) as well as cbb_II expression under certain conditions (69, 93). On both the cbb_I and cbb_II operon promoters, CbbR binds to a site that overlaps the −35 region, which suggests that protein-protein interactions between CbbR and the RNA polymerase are required for transcriptional activation. Inactivation of regA and regB affects cbb_I and cbb_II expression, with only 14 and 10% of wild-type levels, respectively, found in a regA-disrupted strain under photoautotrophic growth conditions. RegA* was also shown to bind to two DNA-binding sites in both the cbb_I and cbb_II promoter regions. There is a major high-affinity RegA binding site located at bp −102 to −121, upstream of the cbb_I transcription start site, that is assumed to be involved in transcriptional activation in concert with CbbR_I. A low-affinity RegA binding site is located at positions −4 to −19, overlapping a CbbR_I DNA-binding site located at positions −18 to −79. Presumably the low-affinity RegA binding site that overlaps the CbbR site plays a negative role as a result of RegA-mediated occlusion of CbbR_I binding to this region. On the cbb_II promoter, there are two high-affinity RegA-binding sites, one located at positions −101 to −116 and the second located at positions −124 to −169. The upstream location of these binding sites suggests that they are involved in activation. The cbb_II promoter also contains a CbbR_II DNA-binding site located at positions −19 to −78.

It has been reported that RegA homologues from B. japonicum (RegR) and S. meliloti (ActR) also function as activators of cbb operons in concert with CbbR (27, 32). Thus, as was demonstrated for R. capsulatus, RegA homologues appear to control CO₂ fixation in a number of photosynthetic and nonphotosynthetic bacteria.

Conditions of nitrogen and oxygen limitation are known to activate the expression of nif genes, which are required for the biosynthesis of molybdenum nitrogenase (reviewed in reference 52). Joshi and Tabita (43) made the surprising discovery that RegA (PrrA) from R. sphaeroides was also involved in the control of nitrogen fixation, underscoring the global role of the RegB/RegA system. Specifically, they observed that nitrogenase synthesis is derepressed in the presence of excess ammonium in strains that lacked a functional CO₂ fixation pathway. They reasoned that CO₂ fixation normally functions as an electron sink to dissipate excess reducing equivalents generated by photoheterotrophic growth. In the absence of CO₂ fixation, they concluded that nitrogenase becomes derepressed to serve as an alternative secondary electron sink. Interestingly, a functional regB gene is required for derepression of nitrogenase in the absence of carbon fixation.

Elsen et al. (23) shed light on the mechanism of derepression of nitrogenase in R. capsulatus by showing that the RegB/RegA system indirectly controls expression of the nifHDK operon, which encodes the molybdenum-containing nitrogenase complex. In R. capsulatus and in many other species, nitrogenase expression is regulated by nitrogen limitation through the NtrB/NtrC two-component system. Under nitrogen-limiting conditions NtrB phosphorylates NtrC, which then activates nifA transcription by binding to two tandem sites centered >100 bp upstream of the transcriptional start site. NifA then activates the expression of numerous nif genes, including nifHKD (reviewed in reference 52). In R. capsulatus, there are two functional copies of nifA, nifA1 and nifA2, either of which can activate nifHDK expression. Elsen et al. (23) demonstrated that RegA binds to the nifA2 promoter between the tandem NtrC DNA-binding sites and the −35 and −10 promoter sequences. Interestingly, RegA-mediated activation of nifA2 transcription requires NtrC, indicating that RegA∼P alone is not sufficient to stimulate nifA2 expression (23). Thus, RegA appears to provide an overarching layer of redox control on top of the control of nitrogen availability provided by NtrC.

In B. japonicum, the RegB/RegA homologues (RegS/RegR) are required for the aerobic and anaerobic expression of the fixRnifA operon (6). Interestingly, a mutation that disrupts the response regulator RegR reduces fixR nifA expression and consequently nitrogen fixation activity. However, no related phenotype was observed on disruption of the sensor kinase, RegS. RegR mutants of B. japonicum form nodules, but the nodules are functionally incapable of fixing nitrogen (a Fix⁻ phenotype). Electron micrographic analysis indicates that nodules formed on infection by the RegR-disrupted strain failed to produce bacteroids, which are differentiated B. japonicum cells that undertake nitrogen fixation (6).

Recently, the R. sphaeroides RegB/RegA system (PrrB/PrrA) was shown to control the expression of nitrite reductase, which is a terminal electron acceptor involved in denitrification (50). Specifically, RegB- and RegA-disrupted strains reduced the expression of the nitrite reductase structural gene, nirK, which resulted in an inability to grow anaerobically on nitrite-containing medium. nir expression is also regulated by nitrite availability through the transcription factor NnrR, which is a member of the FNR/Crp family. Thus, RegA presumably acts in concert with NnrR to coordinate nirK expression.

R. capsulatus possesses the hupSLC operon, which codes for a membrane-bound uptake [NiFe]hydrogenase that catalyzes H₂ oxidation. This enzyme allows the bacterium to grow autotrophically with H₂ as the sole electron source (reviewed in reference 94). Hydrogenase can also recycle electrons from H₂ to nitrogenase under photoheterotrophic growth conditions. Biosynthesis of hydrogenase is regulated by growth conditions, with expression and activity being highest in the presence of its substrate, H₂ (13).

H₂ regulation is mediated by the two-component regulatory system HupT/HupR, with the response regulator HupR directly activating hupSLC transcription in the presence of H₂ (16). The nonphosphorylated form, HupR, binds to the hupSLC promoter at a palindromic sequence centered at bp −157 with respect to the transcription start site. Maximal expression of hupSLC also requires the binding of IHF between the HupR and RNA polymerase DNA-binding sites (reference 92 and references therein).

Elsen et al. (23) demonstrated that RegA is involved in repressing hupSLC expression under both aerobic and anaerobic heterotrophic growth conditions. A major DNA-binding site of RegA was shown to be located close to the −35 promoter recognition sequence, with a second, lower-affinity RegA-binding site overlapping the IHF DNA-binding region. At that location, it is possible that RegA could prevent either the RNA polymerase or the IHF protein, or both, from binding to the hupSLC promoter. Interestingly, a deletion in RegB can be suppressed by addition of multiple copies of the sensor kinase HupT (36). Presumably, increased amounts of HupT are capable of phosphorylating RegA in the absence of RegB.

Glutathione-dependent formaldehyde dehydrogenase plays an important role in the detoxification of formaldehyde by converting it to formate. Analysis of the expression of the glutathione-dependent formaldehyde dehydrogenase gene, adhI, demonstrated that adhI expression is under the control of several effectors that respond to formaldehyde, methanol, or other formaldehyde adducts (2). This enzyme is absolutely required for growth with carbon sources such as methanol, that generate formaldehyde. Formaldehyde oxidation creates reducing power in the form of NADH, thereby providing cellular energy as a product. Interestingly, RegA (PrrA) was shown to be essential for normal aerobic expression of the adhI gene in R. sphaeroides (2). Analysis of RegA binding to the adhI promoter has not been undertaken, so it is not yet certain whether RegA directly or indirectly affects the expression of formaldehyde dehydrogenase.

In S. meliloti, the RegB and RegA homologues, ActS and ActR, control the biosynthesis of three dehydrogenases, formaldehyde dehydrogenase, formate dehydrogenase, and methanol dehydrogenase, as well as CO₂ fixation (32). The ActS/ActR system is also involved in acid tolerance (90).

Romagnoli et al. (77) reported that the aerotactic motility response of R. sphaeroides is partially under the control of the RegB/RegA system. Their study indicated that aerotaxis in R. sphaeroides involves the second chemosensory operon cheOp2, which is one of three che clusters present in R. sphaeroides. Deletion of cheOp2 genes results in complete loss of aerotaxis, while deletion of regB (prrB) results in partial loss of aerotaxis. Deletions of the three linked regulatory genes regB, regA, and senC (prrBCA) restores the aerotactic ability of a cheOp2 deletion (77). It is not well understood why a deletion of the sensor kinase gene, regB, results in a nonaerotactic phenotype whereas deletion of the entire reg operon allows aerotaxis.

It will take further analysis to determine the exact role of RegB and RegA in controlling aerotaxis, such as the ability of RegB to directly affect phosphorylation of the chemotaxis cascade or simply modulate the abundance of the chemosensory apparatus. Nevertheless, aerotactic control going through the RegB/RegA cascade is certainly novel.