Chapter 95

Responses to Molecular Oxygen

A. S. LYNCH and E. C. C. LIN

INTRODUCTION

During the course of the evolution of microbial bioenergetics, progenitors of Escherichia coli and its cousins probably acquired in a stepwise temporal fashion metabolic pathways for fermentation, different routes of anaerobic respiration, and eventually aerobic respiration for the process of energy transduction (68, 227). The acquisition of these different pathways, combined with the apparent development of successive and interactive layers of regulatory mechanisms to control them, enabled the bacteria to exploit adroitly energy sources to their greatest possible advantage. A key strategy of this metabolic modulation is to maximize ATP production under a given growth condition. Accordingly, electron transport processes are preferred over substrate-level phosphorylation reactions, and among the various routes of electron transport available, the one associated with the largest –Δ G is usually preferred. Conspicuous consequences are improved growth rate and/or yield for a given carbon and energy source. In the hierarchical regulation system for energy transduction, central roles are played by two global transcriptional regulators, Fnr and Arc, in concert with other global and specific regulatory proteins (for recent reviews, see references 82, 83, 84, 105, 108, 139, and 198; see also chapter 17 in this volume).

GLOBAL REGULATION BY Fnr IN ANAEROBIC METABOLISM

It has been known for many years that the activity levels of numerous proteins involved in anaerobic electron transport are significantly higher in anaerobically grown cells than in aerobically grown cells (70, 71, 91), observations which gave rise to speculation that a global regulator may be operative in E. coli. Subsequent genetic studies led to the identification of a key regulatory locus, designated fnr (for fumarate nitrate reduction) and located at min 29 on the E. coli chromosome, in which mutations were found to cause defects in the synthesis of nitrite, nitrate, and fumarate reductases (130). The same locus was also identified in studies of nitrite reduction and was designated nirA or nirR (31, 38). The corresponding gene in Salmonella typhimurium (official designation, Salmonella enterica serovar Typhimurium), designated oxrA, was identified by mutations that affect regulation of the pepT locus, encoding endopeptidase T (205).

The Global Regulator Fnr

Determination of the sequence of the fnr gene revealed it to encode a protein (of 250 amino acid residues) which shows significant homology to Cap (for catabolite activator protein); indeed, all of the major secondary structure elements present in the subsequently determined three-dimensional structure of Cap are predicted to be retained in Fnr (9, 226). However, a number of significant differences between the two proteins are apparent. First, as opposed to Cap, which is dimeric, Fnr is isolated from cells in monomeric form (28 kDa as judged by gel filtration), although the protein appears to be dimeric when bound to DNA (76). Second, the residues in Cap that interact with cyclic AMP, and a group of surface-exposed residues in Cap that are thought to interact directly with the α subunit of RNA polymerase, are not conserved in Fnr (9, 219, 226), although in the latter case, a different group of residues in Fnr may serve the same purpose (9); another potential activation region is conserved in both proteins (226). Finally, a feature of Fnr that is not found in Cap is the possession of a cysteine-rich amino-terminal extension, which contains three of the four essential cysteine residues (Cys-20, Cys-23, Cys-29, and Cys-122) thought to be involved in the binding of iron (75, 150, 185). The amino-terminal region probably serves as a redox sensor, which mediates a redox-sensitive conformational change that determines the activity of Fnr as a transcriptional regulator. However, of five Fnr mutants studied that contain single cysteine substitutions, only the variant with a change of Cys-122 to Ala (Cys122Ala variant) is substantially deficient in the binding of iron (75). Although there is no current direct evidence for an iron-sulfur cluster in the protein (76), recent studies of certain mutant Fnr variants by low-temperature electron paramagnetic resonance spectroscopy have led to the suggestion that the active form of the wild-type protein contains one 4[Fe-S] cluster per dimer (124).

The Fnr Modulon

Thirty-one transcriptional units (including over 70 genes) have been recognized as members of the Fnr modulon (Table 1). Their expression is either negatively or positively regulated. With one exception (cea, encoding colicin E1), all of the genes in the Fnr modulon encode proteins that are involved in cellular adaptations to growth in an anoxic environment. Transcription of many Fnr target operons is controlled cooperatively by other global regulators. For instance, maximal expression of ansB (encoding asparaginase II) requires both Fnr and Cap, implying that the gene product is most useful in cells growing anaerobically in a medium deficient in a highly rewarding carbon and energy source such as glucose (116). As is discussed in a later section, the apparent positive regulatory effect of Fnr on arcA expression (41) renders hazardous any definitive interpretation of studies of the dual Fnr/Arc control of certain genes. However, irrespective of whether Fnr effects are direct or indirect, the apparent dual regulatory mechanisms of cyoABCDE, cydAB, sodA, and focA-pfl are all examples in which the combined activities of ArcA and Fnr allow for fine-tuning of gene expression during transition from aerobic to anaerobic growth.

Table 1The Fnr and Arc modulonsa

Whereas it is apparent in some instances that preferential anaerobic utilization of certain exogenous electron acceptors requires the intervention of transcriptional regulatory proteins in addition to Fnr, it remains a formal possibility that hierarchical levels of Fnr affinity for different target promoter elements act in some instances as an alternate mechanism for sequential gene activation during anaerobic adaptation. However, the way by which nitrate (midpoint potential or E_m,7 = +433 mV) is used preferentially over fumarate (E_m,7 = +30 mV) as an exogenous electron acceptor is now clear. Although expression of both the frdABCD (encoding fumarate reductase) and narGHIJ (encoding the principal nitrate reductase) operons requires activation by Fnr, the presence of nitrate (and molybdate, which is required as a component of nitrate reductase) results in activation of the narGHIJ operon and simultaneous repression of frdABCD. This is effected by the activities of a two-component signal transduction system (NarX/NarL), through which the presence of nitrate is apparently detected by the sensor kinase (NarX), resulting in stimulation of its autophosphorylation activity and consequent activation as a kinase for the response regulator (NarL). The phosphorylated form of NarL is thought to act directly as a transcriptional activator of narGHIJ and repressor of frdABCD (183; see also chapter 17 in this volume).

DNA-Binding Sites.

A 22-bp Fnr-binding site consensus (or Fnr box), deduced from comparisons of target sequences, is a partial palindrome containing a 5'-TGAT-3' half-site motif, resembling the 5'-GTGA-3' motif present in the Cap binding site consensus (53a, 113, 193). Following delineation of the Fnr consensus, the close structural relationship between the Cap and Fnr proteins has been elegantly shown by demonstrating that switching of the noncongruent base pair in the half-site motifs of target promoters can result in their in vivo regulation by the corresponding noncognate regulator (192). DNase I footprinting studies of several positively regulated promoters (FFpmelR, pfl, fdn, narGHIJ, and nirB) and two negatively regulated promoters (fnr and ndh) confirm the consensus site for Fnr binding and have further demonstrated that two Fnr monomers are bound at each site (10, 76, 132, 186). In addition, such in vitro studies demonstrate the importance of the location of Fnr-binding sites at target promoters. At positively regulated promoters, which typically have weak –35 elements, the sites are centered at about –41.5 and are therefore analogous to class II type Cap-dependent promoters. At negatively regulated promoters, the location of Fnr sites is variable, and thus transcriptional repression mediated by a simple promoter occlusion mechanism does not appear to account for all cases (74, 209).

Further evidence for the close structural relatedness of Fnr and Cap has emerged from studies of a series of Cap-Fnr hybrid proteins that activate Cap-dependent promoters in response to anoxia or that activate Fnr-dependent promoters in response to glucose starvation (192). DNA bending mediated by Fnr has, however, yet to be experimentally demonstrated.

In Vitro Transcription.

Although holo-Fnr is not essential for DNA binding, a sufficiently high iron content is essential for transcriptional activation and repression activity in vitro. Apo-Fnr can be activated for in vitro transcription by incubation with Fe²⁺ and β-mercaptoethanol; this reaction is reversible by the addition of iron-chelating agents (72, 73, 76, 186).

Signaling

Demonstration of the functional importance of the three amino-terminal cysteine residues of Fnr, and the dependence of the protein on Fe²⁺ for in vitro transcription regulatory activity, does not solve all the overall mystery of how the protein senses the invivo signal for anoxia. There are, however, a number of experimental clues regarding the in vivo activation mechanism.

First, at least a portion of the holo-Fnr purified from aerobically grown cells retains specific transcriptional regulatory activity (186). Second, the addition of hexacyanoferrate III (E_m,7 = +520 mV) to growth media apparently diminishes Fnr function in vivo in the absence of O₂ (220). This O₂-like effect can be mimicked by the addition of iron chelators to growth media, an effect that can be overcome by Fe²⁺ (194, 215). Finally, the reversible in vivo activation of Fnr, related to the availability of O₂ and iron, occurs without de novo protein synthesis (56). Since the total iron content of E. coli is not sensitive to the presence or absence of O₂ (158), the combined data point to the availability of Fe²⁺ as the important cellular variable. The currently favored hypothesis is that Fe²⁺ induces the transcriptionally active conformation of Fnr, which is either free in the cytoplasm or already bound to DNA in dimeric form (J. R. Guest, personal communication). This notion is consistent with the successful isolation of Fnr* variants, which contain substitutions in the putative dimer interface that increase dimerization and which become partially (but not fully) active under aerobic conditions (125, 150). Further elaboration of the model will be necessary to explain some apparent aerobic transcriptional activity of wild-type Fnr, including autorepression of the fnr gene (171, 182).

GLOBAL REGULATION BY THE Arc SYSTEM DURING AEROBIC METABOLISM

The ArcAB Two-Component Signal Transduction System

The tricarboxylic acid cycle of E. coli is highly operative only in aerobically grown cells (5, 28, 70, 71, 90, 91), with the key regulatory control responsible for determining the levels of tricarboxylic acid cycle enzymes exerted at the transcriptional level. The principal elements of this regulation are encoded by the arcA (min 0) and arcB (min 69.5) genes (arc for aerobic respiration control) and are members of a two-component signal transduction system (98, 104, 109). Activation of the ArcB sensor probably occurs by detection of one or more specific stimuli and results in its autophosphorylation and consequent activation as the cognate kinase for the response regulator component (ArcA); the phosphorylated form of ArcA (ArcA-P) generally acts as a transcriptional regulator at target operons of aerobic respiratory function. In many cases, the expression of the genes involved is also coordinated with carbon, nitrogen, and phosphorus metabolism. This coordination probably depends largely on the cellular activities of other transcriptional regulators but may also in some cases be effected by a more direct mechanism, as implied by the apparent in vitro ability of ArcA to use acetyl phosphate and carbamoyl phosphate as substrates for autophosphorylation (A. S. Lynch and E. C. C. Lin, unpublished observations; G. Sawers, personal communication; see also references 146, 147, 165, and 225).

The arcA Gene.

The arcA gene was previously known as the dye gene on the basis that mutations at this locus resulted in growth sensitivity to the phenothiazine redox dyes methylene blue and toluidine blue (23, 24, 25, 104). Analysis of the deduced amino acid sequence of ArcA (238 amino acid residues) places the protein in the OmpR subfamily of two-component response regulators, which all contain a phosphoryl receiver domain at their amino termini and a helix-turn-helix-type DNA-binding domain at their C termini (2, 167, 170, 176, 207, 221). In the ArcA protein, Glu-10, Asp-11, and Asp-54 correspond to the three conserved amino acid residues which form an acidic pocket in the crystal structure of the homologous CheY protein. By analogy with studies of other response regulator proteins, Asp-54 of ArcA is expected to be the phosphoryl group acceptor residue, and Lys-103 corresponds to an invariant residue conserved in the receiver domains of all response regulator proteins (221).

Recent studies indicate that expression of the arcA gene, as monitored by Φ(arcA-lacZ) fusion constructs, increases about fourfold anaerobically. This anaerobic induction is observed to be entirely dependent on Fnr and to be maximal when arcA ⁺ is also present. It is proposed that the anaerobic repression of an arcA promoter (which is active in aerobically grown cells), and simultaneous activation of a stronger promoter (which is inactive in aerobically grown cells), is due to Fnr binding at a consensus site located within the sequence of the anaerobically repressed promoter (41). Some discrepancies between these results and those of previous studies (99, 168) can probably be accounted for by differences in the penetrance of the arcA and fnr alleles employed and/or differences in the structures of the gene fusions used. The significance of the inclusion of the arcA gene within the Fnr modulon presumably ensures that the effects of ArcA are accentuated during growth in highly anaerobic conditions.

As has been demonstrated in studies of the OmpR/EnvZ two-component system (191), overexpression of the response regulator (ArcA) is observed to be epistatic over null mutations in the cognate sensor kinase, ArcB (98).

The arcB gene.

Of 70 independent mutants isolated on the basis of elevated anaerobic activity levels of β-galactosidase in an sdh ⁺ (the operon encoding the succinate dehydrogenase complex) Φ(sdh-lac) merodiploid, five mapped at min 69.5 on the chromosome. Characterization of this locus, arcB, revealed an open reading frame of 778 amino acid residues (109). Analysis of the deduced sequence of ArcB suggested that the protein belongs to a subgroup of response sensors referred to as hybrid kinases, which possess a receiver domain in addition to the signal transmitter domain. The subgroup currently includes some seven members, including the RscC, BarA, and EvgS proteins of E. coli (2, 95, 153, 154, 170, 204); however, as with the other members in this subgroup, no helix-turn-helix DNA-binding motif is apparent in the C terminus of ArcB. Sequence analysis also suggested that ArcB is a membrane-associated protein with the two putative transmembrane segments (Phe-23 to Val-50 and Ser-58 to Val-77) separated by about seven amino acid residues, which are therefore predicted to be exposed to the periplasmic compartment (109). Cell fractionation studies of native ArcB protein (109) and of amino-terminally truncated variants (96, 106, 109) support the notion that ArcB is localized to the cytoplasmic membrane.

In the transmitter domain, His-292 is expected to be the autophosphorylation site, while in the receiver domain, Glu-532, Asp-533, and Asp-576 are expected to constitute the canonical acidic pocket, with Lys-628 as the invariant downstream residue. Site-directed mutagenesis studies indicate that His-292, Asp-576, and Asp-533 (but not Asp-624) are all important for normal ArcB function (107).

The key role of the receiver domain of ArcB was demonstrated in strains bearing a null arcB chromosomal mutation and harboring plasmid-borne arcB alleles encoding ArcB proteins truncated at residue Phe-516 or Asp-517. In these strains, the aerobic/anaerobic ratio of Φ(sdh-lac) expression was 13-fold lower than in an isogenic strain expressing an arcB ⁺ plasmid-borne allele (107). The significance of the receiver domain’s apparent ability to temper the activity of the transmitter domain to the in vivo regulation of the ArcB protein remains to be elucidated.

A study utilizing a Φ(arcB-lacZ) fusion indicated no respiratory regulation of ArcB at the transcriptional level (107). As a historical note, it should be mentioned that originally cpxA (encoding a two-component sensor protein) was paired with arcA (or dye) because mutations in either gene prevented the expression of an F⁺ phenotype (14, 15, 26, 133, 148, 159, 175, 187). However, it is now established that the cpxR gene encodes the cognate response regulator for the CpxA sensor protein, although the function of this two-component system remains obscure (51).

The Arc Modulon and Mutant Phenotypes

Target Genes.

Although arcA and arcB were first identified as trans-acting regulatory genes affecting the transcriptional control of the sdh operon, ensuing studies revealed that the expression of a considerable number of operons and regulons is modulated by the gene pair. Currently known control targets are listed in Table 1. A cursory inspection reveals that all of these target genes encode proteins whose synthesis would be expected to be coordinated during aerobic respiration. Increased anaerobic expression of the great majority of target operons in the arc mutants suggests that ArcA-P acts as a repressor. In a few exceptional cases, however, ArcA-P appears to function as an activator.

One example of positive regulation by ArcA-P is the control of the cydAB operon (encoding cytochrome d oxidase). The level of this enzyme is maximal under microaerobic conditions when the function of the enzyme becomes most useful to the cell, and initial studies indicated that Fnr was a coregulator of the operon (42, 44, 64, 99). However, it now seems most likely that Fnr exerts its effect indirectly by both elevating the ArcA level in anaerobic cells and increasing the ArcA-P/ArcA ratio. In part, at least, the latter effect is probably mediated by the influence of Fnr on the cellular levels of key anaerobic metabolites, such as d-lactate, acetate, and NADH. These compounds were shown to enhance the rate and level of ArcB autophosphorylation and thereby are expected to stimulate ArcA phosphorylation (96, 97, 140). This overall picture of cydAB regulation is supported by the observation that in strains bearing null arcA or arcB mutations, heat shock induction of the cyd operon is completely blocked (224). The inability of the arcA1 mutation to abolish expression of a Φ(cyd-lac) fusion under microaerobic and anaerobic conditions in a previous study (64) is apparently due to leakiness of the allele (S. Iuchi and E. C. C. Lin, unpublished observations).

Another positively regulated operon under dual Arc and Fnr control is pfl, encoding pyruvate-formate lyase, the gate for terminal fermentation reactions. In this case, the combined data suggest that both proteins probably act directly as transcriptional activators, with integration host factor serving in an architectural capacity by inducing a specific bend in the DNA located between the two major promoters activated by ArcA and Fnr (178, 180).

Less comprehensible at the current time is the positive control of expression of the F plasmid traY locus (101, 188, 189). An arcA1 mutant allele has been characterized and found to contain a 24-bp tandem duplication in the 5' end of the gene which results in insertion of eight amino acids between residues Ala-33 and Thr-34. Whereas strains bearing the arcA1 mutation are found to lack Arc function as judged by the apparent failure to anaerobically repress expression of a Φ(sdh-lac) fusion, aerobic activation of Φ(traY-lacZ) was largely unchanged (60%) in comparison with an isogenic arcA ⁺ strain (188). Similarly, the arcB1 mutation, presumed to be a null allele since only the amino-terminal 240 residues of ArcB are expressed (107), was found to derepress Φ(sdh-lac) expression but had little effect (<20%) on Φ(traY-lacZ) expression (188). In contrast, an ArcA^Val203Met variant (the product of the sfrA5 allele), which lowered the normal aerobic Φ(traY-lacZ) expression by approximately sevenfold, had no significant effects on anaerobic repression of Φ(sdh-lac) (188, 189). As expected, deletion of the arcA locus essentially abolishes both anaerobic repression of Φ(sdh-lac) and aerobic Φ(traY-lacZ) expression (104, 189). Whereas the molecular mechanism accounting for these apparent target-specific effects of ArcA awaits further experimental clarification, the data hint at the interesting possibility that the so-called Sfr (sex factor regulation) and Arc functions of the ArcA protein are physiologically and genetically separable (188).

Target Sequences for ArcA Regulation.

Comparisons of the transcriptional regulatory regions of target operons have thus far failed to reveal any obvious sequence that may correspond to an ArcA box, i.e., a consensus DNA-binding site presumably for ArcA-P. However, should such a sequence resemble that of NarL (a homologous response regulator protein), which comprises an imperfectly conserved heptameric repeat (135, 198, 217, 218), mere sequence inspection may not prove an adequate method to identify and define the putative ArcA box.

Specific binding of the ArcA protein to the regulatory region of the sodA gene (encoding the Mn-containing superoxide dismutase [MnSod]), whose transcriptional control depends on six regulatory proteins (with ArcA acting as an anaerobic repressor), has been demonstrated in experiments utilizing soluble cell extracts prepared from cells overexpressing the protein (40, 212). A DNase I footprint of approximately 65 bp overlapping the –35 to –10 region of the sodA promoter was obtained; however, definitive interpretation is rendered difficult by the observed failure of purified monomeric ArcA protein to bind to the sodA sequence (212). Two explanations of this apparent discrepancy seem most likely. First, it may be that only the phosphorylated form of ArcA is active in sodA DNA binding and that sufficient ArcA-P must therefore have been present in the crude extract for a DNase I footprint to be obtained. Alternatively, ArcA may bind to the sodA promoter only in a high-molecular-weight complex, possibly including other proteins (212).

For purified ArcA protein, site-specific DNA binding to the pfl promoter region has been demonstrated by both gel retardation and DNase I footprinting assays. In these studies, the DNA- binding activity of ArcA protein was significantly stimulated by prior incubation with either carbamoyl phosphate or acetyl phosphate (Sawers, personal communication); similar effects have been observed in studies of ArcA binding to the gltA-sdhCDAB intergenic region and to a promoter of the lctPRD operon (Lynch and Lin, unpublished observations).

Additional in vitro studies of ArcA and ArcA-P binding to target sequences, in combination with mutational analysis of putative target sites, are needed to define properly an ArcA box and to establish whether phosphorylation alters the DNA binding specificity of the protein or merely its intrinsic DNA binding affinity. By analogy with studies of other response regulators, several mechanistic scenarios are possible to account for changes in ArcA activity by phosphorylation.

Changes in Gene Expression in arc Mutants.

In arc mutants, the anaerobic activity levels of several enzymes required for aerobic respiration were unexpectedly found to exceed those of aerobically grown wild-type cells. Moreover, in some cases, null mutations in arcA gave stronger phenotypes than null mutations in arcB (98, 104). Two nonexclusive reasons might explain the observed phenomena: first, it is possible that there was an insufficient O₂ diffusion rate into the aerobic growth cultures; second, it is possible that there is always a basal level of ArcA-P in aerobic cells and that some target operons have relatively high-affinity binding sites for it. The latter hypothesis seems consistent with the results of recent studies in which the efficacy of ArcA acting as a transcriptional activator or repressor in aerobic cells was found to be further enhanced by anaerobiosis (46, 169); in these cases, Fnr-mediated stimulation of ArcA levels in anaerobic cells does not appear to account for the effects observed.

Unexplained Phenotypes.

Although arcA was previously characterized as dye on the basis of increased growth sensitivity to toluidine blue and methylene blue, the molecular basis of this mutant property ironically remains unknown. Mutations in arcB likewise cause dye sensitivity (98, 109). One possibility is that in aerobically grown arc mutant cells, the composition of the electron transport chain is perturbed such that the redox dyes can subvert electrons to futile flow. This model is compatible with the observation that the dye sensitivity phenotype of arc mutants is observed only in aerobically grown cells (Lynch and Lin, unpublished observations).

The arcB gene was recently identified as a multicopy suppressor of a defect in the activation of Φ(ompC-lacZ) and Φ(ompF-lacZ) fusions in a Δ(envZ) ompR ⁺ strain (153); the barA gene (also encoding a hybrid kinase) was similarly identified as a suppressor. Perplexingly, the multicopy suppression was subsequently found to require expression of only the carboxy-terminal 139 residues of ArcB or carboxy-terminal 118 residues of BarA (95). Comparison of the sequences of these regions of ArcB and BarA, along with the carboxy termini of five other hybrid kinases, revealed a limited but significant level of sequence similarity extending over some 150 residues which includes a single invariant histidine residue. In ArcB and BarA, these residues correspond to His-717 and His-861, respectively, and in both cases, the histidine residue was demonstrated to be essential in the multicopy suppression phenomenon (see also below, In Vitro Phosphorylation Studies).

Signaling

Characteristically, membrane-associated sensors of the two-component family are anchored by two membrane-spanning segments separated by an extensive periplasmic domain serving as the signal receptor. ArcB is unusual in that it apparently lacks any significant periplasmic domain, a property which suggests either that the transmembrane segments themselves serve for signal reception or that they simply anchor the protein to the cytoplasmic membrane to facilitate the reception of signals from components of the membrane. Available data indicate that signals to ArcB come from both the membrane and the cytosol.

The role of molecular oxygen as a direct signal is apparently excluded, since in a Δ cyo Δ cyd mutant lacking both the terminal oxidase with low O₂ affinity and the terminal oxidase with high O₂ affinity, strong repression of a Φ(lctD-lac) fusion was maintained during aerobic growth (99); in this edition, the lctD gene encoding l-lactate dehydrogenase has been renamed lld for consistency with dld, which encodes the d-lactate dehydrogenase (see chapter 17). It now appears that the small O₂ effect remaining in the strain doubly deleted in cyo and cyd is attributable to a third terminal oxidase (47). Nonetheless, the level of repression of Φ(lctD-lac) and Φ(sdh-lac) fusion does seem to correlate with the redox state of the cell. For instance, the highest expression level of Φ(sdh-lac) occurred with O₂(E_m,7 = +818 mV) as the electron acceptor, and fairly high expression of the fusion also occurred during anaerobic growth with NO₃ ^– (E_m,7 = 433 mV) as the electron acceptor. Significant expression occurred even during anaerobic growth with fumarate (E_m,7 = +30 mV) as the electron acceptor. Hence, the Arc system may respond to a metabolite that can exist physiologically in either an oxidized or a reduced form (104). Possible candidates include the lipid-soluble electron carriers ubiquinone (the adaptor for aerobic respiration and anaerobic nitrate respiration chains) and menaquinone (the adaptor for anaerobic respiration involving acceptors with relatively low midpoint potentials, such as fumarate). The possibility that either of the two cysteine residues in ArcB (Cys-180 or Cys-241) plays a role in redox sensing has been discounted, since glycine can be substituted for either residue without significant effects on the anaerobic repression of target operons (107).

In a recent study, it was reported that the level of aerobic induction of cytochrome d oxidase synthesis diminished as the proton potential across the plasma membrane was artificially reduced by treatment of cells with the protonphore pentachlorophenol or agents that oxidize elements of aerobic respiratory chains (18). As the response was found to be dependent on the integrity of the Arc system, it was suggested that the proton potential across the membrane is a source of signaling for ArcB, and it is proposed that ArcB acts as a protonmeter (18).

Additional studies are necessary to test the redox and the protonmeter models. Both models require that the ArcB protein be closely associated with the plasma membrane in order to receive the proposed signal(s). Indeed, by definition, a protein acting as true protonmeter would have to straddle the cytoplasmic membrane in order to sense proton levels on either side of the membrane. Finally, it is known that the autophosphorylating activity and/or the dephosphorylating activity of ArcB responds in vitro to fermentation product(s), such as d-lactate, acetate, and NADH, which signal transition to anoxia in the cell (see below). Although these metabolites may not serve as the primary stimuli, they probably do function physiologically as cytoplasmic effectors that modulate the ArcB kinase activity. Their cellular accumulation may lower the threshold sensitivity of ArcB to the true stimulus. It would seem that for any model to be complete, it would have to account for the intimate association of the ArcB protein with the cytoplasmic membrane. In conclusion, multiple different signals may exist for ArcB which, in combination, are likely to fine-tune its activity in response to numerous different environmental conditions.

In Vitro Phosphorylation Studies

Intact ArcB is difficult to purify because of its membrane association; however, purification of a catalytically active form is facilitated by genetically removing amino acid residues 1 to 128 of ArcB, thereby releasing the shortened protein (designated Arc^129–778 ) from the membrane. The truncated protein undergoes autophosphorylation at the expense of ATP, but not when His-292 is substituted with a glutamine residue (96, 106, 107).

When purified ArcB^129–778, labeled in vitro by [γ-³²P]ATP, is incubated at 43°C in a pH 12.5 buffer, a biphasic loss of the label occurs, indicative of dephosphorylation of both a histidyl and an aspartyl residue. Overall phosphorylation of a similar protein, with Asp-576 substituted by alanine, is greatly reduced. Finally, purified ArcB^129–778 catalyzes ATP-dependent phosphorylation of ArcB^His292Gln , but not ArcB^{His292Gln/Asp576Ala} , in everted membrane vesicle preparations (96, 106, 107). Apparently the terminal phosphoryl group of ATP is first transferred to His-292 and then transferred intermolecularly to Asp-576, as has previously been demonstrated in studies of other two-component sensor proteins (208, 235).

A surprising discovery is that in the presence of [γ-³²P]ATP, purified cytoplasmic membranes made from arcB ⁺ cells catalyze transphosphorylation of purified ArcB^639–778 , provided the latter possesses the wild-type His-717 residue. Although ArcB^639–778 is incapable of undergoing autophosphorylation itself, once phosphorylated, it acts as a kinase for purified OmpR, or intact ArcB, in cytoplasmic membrane preparations (95).

Strains expressing an arcB allele with a His717Leu mutation in single copy, however, exhibit no change in either ompC or ompF expression, retain normal anaerobic repressibility of a Φ(lctD-lac) fusion, and are resistant to toluidine blue (95). Hence, the physiological relevance of the phosphoryl acceptor and phosphoryl donor activities of His-717 remains to be established.

Purified ArcB^129–778 catalyzes the phosphorylation of ArcA in the presence of ATP (106). Additionally, studies of everted membrane vesicle preparations made from cells expressing various mutant arcB alleles indicate that ArcB variants lacking the receiver domain catalyze ArcA phosphorylation at a much slower rate than intact ArcB (107). In contrast, ArcB^Asp576Ala and ArcB^Asp533Ala are inactive as ArcA kinases despite the possession of a receiver domain (96, 107).

Studies of the autophosphorylation of intact ArcB in everted membrane vesicles, or of purified ArcB^129–778 , showed that d-lactate (even at 0.1 mM), acetate, pyruvate, or NADH increased both the rate and the maximal level of protein phosphorylation and decreased the rate of dephosphorylation. These results suggest that the site of action of the effectors is in the cytosolic domain of ArcB and that these effectors may inhibit dephosphorylation at Asp-576, thus promoting phosphorylation at His-292 (96, 107, 140).

Combined in vitro and in vivo studies have led to the following working model (shown in diagrammatic form in Fig. 1). In the quiescent state, the receiver domain of ArcB docks onto the transmitter domain. Upon stimulation, the His-292 undergoes autophosphorylation and subsequently intermolecularly phosphorylates Asp-576, which results in a conformational change in the protein that leaves the docking site of the receiver domain vacant. Further stimulation again results in autophosphorylation of His-292, creating a doubly phosphorylated ArcB species which acts as a kinase for ArcA by transferring the phosphoryl group from His-292 of ArcB to Asp-54 of ArcA. The presence of effectors, such as d-lactate, prolongs the half-life of Asp-576 and thereby stimulates the transphosphorylation of ArcA (96, 107, 140).

Fig. 1Model for the signaling process by ArcB under aerobic and anaerobic conditions. H, D, P, and m indicate the conserved His-292 and the conserved Asp-576 amino acid residues, the transferred phosphoryl group, and cellular metabolites such as D-lactate and NADH, respectively. Arrows indicate reaction steps. + and ? indicate positive and negative regulatory effects, respectively. ArcB may form a dimer or higher oligomer, and the phosphoryl group may be transferred between the subunits (96, 107, 140).

Source:

It has been suggested that Asp-576 in the receiver domain of ArcB may serve as the target for noncognate histidine kinases or phosphatases and further that such cross talk may play an integrative role in cellular metabolism (207). It has also been proposed that the conserved His-717 in ArcB may be the target of a noncognate sensor kinase protein, or that the His-717 domain phosphorylates a noncognate regulator protein. While interactions with other two-component systmes seem plausible, cross talk of the CpxA sensor protein with the ArcB (or ArcA) protein appears to be excluded (101).

OXIDATIVE STRESS

The advantages of aerobic metabolism are not without jeopardy. Some O₂ molecules unavoidably undergo reduction to toxic species such as O₂ ^{· –} and H₂O₂. O₂ ^{· –} can be converted to H₂O₂ either by superoxide dismutase (Sod) or spontaneously. H₂O₂, in turn, can be decomposed by catalase to yield H₂O. Surviving H₂O₂ can, however, also react with transition metal ions to form the highly destructive HO^· radical. Unfortunately, no enzymatic protection against HO^· is available, presumably because of its tendency to react with any target upon collision. Important vulnerable targets of reactive oxygen species are numerous, encompassing lipids, proteins, certain prosthetic groups of some enzymes, and DNA (62, 65, 66, 94, 123, 195).

E. coli is known to possess three Sods and two catalases. The first group comprises an MnSod (encoded by sodA), whose synthesis is controlled by at least six different regulators (40), an FeSod (encoded by sodB), whose level is influenced by intracellular iron concentration (157), and a periplasmic CuZnSod (12). The second group comprises (i) hydroperoxidase I (HPI) encoded by katG, whose levels respond to redox-cycling agents and H₂O₂ (61, 86), and (ii) HPII, encoded by katE, which is induced by starvation (222). Reinforcing these enzymes are metabolites such as NADPH, NADH, thioredoxin, and glutathione which remove harmful oxygen species by stoichiometric reactions (references 92 and 149 and references therein).

The SoxRS System

Although enzyme induction in response to oxidative stress (H₂O₂ or redox-cycling agents) was recognized almost two decades ago (61, 86), the wide scope of the adaptive defense mechanisms of the cell was not recognized and analyzed until recently. It is now known that two regulatory systems control families of operons responding to oxidative stress (35, 48, 49, 59, 78, 79, 126, 216, 223).

The soxRS Genes.

The two overlapping genes soxR and soxS (at min 92) are transcribed divergently, with transcription of soxS starting within the region between the two open reading frames and transcription of soxR beginning within the coding region of soxS (4, 79, 89, 216, 231). The deduced amino acid sequences of SoxR and SoxS indicate that both are DNA-binding proteins, SoxR being a member of the MerR-like family and SoxS being a member of the AraC-like family (4, 231). The combined data from numerous studies suggest that upon stimulation, SoxR is switched to a conformation capable of activating soxS and that the resulting SoxS protein directly regulates expression of target genes (4, 48, 163, 231, 232). Purified SoxR is an Fe-S protein which appears to be dimeric in form with or without bound Fe (88), and recent data support a dimer structure with two binuclear Fe₂S₂ clusters (E. Hidalgo and B. Demple, personal communication). Although apo-SoxR and Fe-SoxR bind with equal affinity to a soxS site encompassing the –35 and –10 regions of the promoter, only the Fe-SoxR form stimulates in vitro transcription (88).

Target Operons.

About 40 E. coli proteins are induced in response to superoxide-generating agents. Among these, the syntheses of about 10 appear to be under positive control of SoxR and SoxS (79, 89). The positively controlled operons include: (i) sodA, encoding MnSod, which might associate with DNA to provide special protection from superoxide damage (196); (ii) nfo, encoding endonuclease IV, which is involved in DNA repair (29, 45, 50, 134); (iii) zwf, encoding glucose-6-phosphate dehydrogenase, whose increase is expected to elevate the pool of NADPH (177); (iv) fumC, encoding the aerobically stable fumarase (141); (v) acnA, encoding the Fe₄S₄-dependent aconitase (65, 81); and (vi) fpr, encoding NADPH ferredoxin:oxidoreductase, which may serve by maintaining FeS groups in the reduced form (142). As an exception, SoxS negatively regulates its own gene expression (164). SoxS by itself appears to be sufficient to activate target operons. The purified protein binds tightly and specifically just upstream of the –35 elements of target promoters and recruits RNA polymerase to these promoters in vitro (60, 138).

Unexpected Functions.

SoxRS-mediated activation of the micF-encoded antisense RNA results in impairment of synthesis of the OmpF membrane porin (34). The effects on OmpF levels mediated by the antisense micF RNA appear to be more relevant to multidrug resistance than to oxygen stress, unless there are redox-cycling compounds in the natural environment. Possibly certain drugs can be damaging in more than one way, making the collective protection by several stress response systems essential. SoxRS activity is also known to influence modification of the carboxy terminus of ribosomal protein S6 (79, 122).

Signaling.

The SoxRS system was initially characterized as responding to structurally unrelated redox-cycling agents (160, 163, 216). However, the preponderance of evidence indicates that O₂ ^{· –} is required for activation of SoxR in vivo (69, 93, 123, 162, 163, 216). Recently, nitric oxide (NO^?) was discovered as another signal that can act externally on the cell (161, 162). The response to NO^? probably evolved to protect the cell when it is phagocytosed by macrophages which utilize NO^? as one of their cytotoxic weapons (155). The precise mechanism(s) by which SoxR is activated by different radicals is not yet clear.

The OxyR System

The synthesis of about 40 proteins is enhanced after exposure of cells to H₂O₂ in the range of 5 to 200 μM. Some of these are also elevated during other stress responses, including exposure to redox-cycling agents and heat shock (35, 78, 152, 223).

The oxyRS Genes.

The product of oxyR (min 89) usually functions as a transcriptional activator. OxyR has six cysteine residues and belongs to the LysR family of DNA-binding proteins (36); during purification, OxyR can be activated by exposure to O₂ (203). The Cys-199 residue is of critical importance, since substitution at this position with a serine residue results in a protein that is nonactivatable for transcription. The reversibility of the transition between active and inactive forms of OxyR led to the suggestion that Cys-199 might undergo interconversion to sulfenic or sulfinic acid (37, 203). Changing Ala-233 to valine activates OxyR; perhaps this change locks the protein in an active conformation (36, 203, 214).

The oxyS gene is divergently transcribed from oxyR. OxyR stimulates transcription of oxyS to generate a small untranslated RNA (203, 214), the function of which remains to be defined.

Target Operons.

Operons known to be regulated by OxyR include (i) katG encoding HPI catalase (61), (ii) ahpCF encoding an NADPH-dependent alkyl hydroperoxide reductase, (iii) gorA encoding glutathione reductase (35), and (iv) dps encoding a protein that nonspecifically binds to DNA to protect cells from H₂O₂ toxicity (3). In strains deleted for oxyR, 20 to 30 proteins are still inducible by H₂O₂. These mutants, however, are hypersensitive to peroxides and redox-cycling agents and have increased spontaneous mutation rates (77, 202). An intriguing unanswered question relates to whether cyd is controlled by OxyR, as well as by the Arc system, as the synthesis of cytochrome d is known to be induced by heat shock and loss of the enzyme renders cells sensitive to H₂O₂ (224). It is relevant to note that hypersynthesis of HPI catalase, HPII catalase, or alkyl hydroperoxidase suppresses H₂O₂ hypersensitivity in Δ oxyR strains (77).

Whereas the reduced and oxidized forms of OxyR and the OxyR^Cys199Ser variant all bind and repress the oxyR promoter, reduced forms of OxyR or the OxyR^Cys199Ser mutant do not seem to bind to the katG or ahpC promoter (214). An in vitro affinity selection using pools of DNA fragments led to the definition of a consensus OxyR-binding motif with elements of twofold symmetry (214). Four sets of 4-bp sequences are each separated by 7 bp of nonconserved sequence. The data suggest that transcription is stimulated by interaction of an activated OxyR tetramer with the promoter DNA spanning four consecutive helical turns in the –35 to –80 region and is mediated by contact with the α subunit of RNA polymerase (210, 213, 214). Other studies have demonstrated that OxyR autogenously represses expression of its own gene irrespective of the oxidation state of the protein (36, 214).

Signaling.

The levels of oxyR mRNA or OxyR protein do not change significantly following exposure of cells to H₂O₂ (36, 203). Hence, it seems most likely that posttranslational regulation by an H₂O₂-generated signal activates preexisting OxyR protein.

OVERVIEW

In bacterial gene regulation, overlaps in the membership of modulons presumably evolved to facilitate integration between different metabolic pathways in the overall control of cellular metabolism (104, 105, 156). This phenomenon is elegantly demonstrated in the case of modulons involved in coordinating physiological adaptations to anaerobiosis and oxidative stress and presumably enables the cell to fine-tune its metabolism to exploit maximally the energetically favorable aerobic respiration while minimizing oxidative damage of vital macromolecules. The sheer multiplicity of the genes involved in these responses is a further indication of the importance of oxygen as the preferred electron acceptor in the evolution of E. coli, S. typhimurium, and their cousins. It is clear that our knowledge of the membership of the modulons involved is still far from complete, as is a comprehensive understanding of the different intracellular and extracellular sensing mechanisms involved.

In the case of Fnr, clearly additional studies are needed to elucidate the molecular mechanism underlying in vivo activation of the protein during transition of cells to anoxia. Further, although studies to date indicate that Fnr plays only an anaerobic regulatory role, a potential regulatory role of the apoprotein in aerobic cells has not been rigorously investigated. At the metabolic level, it is somewhat surprising that null mutations in fnr do not prevent E. coli from growing anaerobically at a practically normal rate in minimal glucose medium (S. Iuchi and V. Stewart, personal communications), despite the fact that the main gateway leading to the fermentation products is catalyzed by pyruvate-formate lyase encoded by the pfl gene, which is thought to be strongly activated in anaerobic cells by Fnr. In this regard, it would be interesting to establish whether activation of the pfl promoter(s) by ArcA-P alone is sufficient for this particular growth condition.

In the Arc system, several key areas need to be addressed. First, the physiological significance of alternate (ArcB-independent) routes of ArcA phosphorylation in vivo needs to be investigated, in particular the possibility of aerobic control of the Arc modulon mediated by autophosphorylation of ArcA with use of low-molecular-weight phosphoryl donor substrates. Second, the physiological significance of the apparent regulatory function of the receiver domain of ArcB needs to be established. Third, identification of the true nature of specific chemical effectors and/or physical effectors of the ArcB kinase would appear to be the key in distinguishing whether the protein senses the redox or energy state (or both) of the cell. Finally, and of special interest, is the elucidation of the role of the carboxy-terminal kinase domain of ArcB.

The current intensive investigations of the oxygen stress responses should soon clarify the signal transduction processes and the molecular mechanisms underlying the switches between active and inactive forms of the key transcriptional regulators involved. Nature being so opportunistic, it would not be surprising to discover that the potentially damaging oxygen radicals can also be exploited in a beneficial manner. Evidence is already accumulating that the aerobic production of active oxygen selectively hastens the destruction and turnover of enzymes of anaerobic function and thus facilitates adaptation to aerobiosis (E. C. C. Lin and J. Aguilar, unpublished data). At the level of population genetics, it is conceivable that oxygen radical damage to DNA increases mutation rates to promote the survival of subclonal descendants.

Finally, a considerable expansion in quantitative biochemical and physiological data will be required before a holistic understanding of this area of metabolism can be realized.

ACKNOWLEDGMENTS

We thank J. R. Guest and B. Demple for providing us with unpublished reviews, J. R. Guest, J. Green, A. S. Irvine, and S. Shapiro for making available a list of Fnr targets prior to publication, and Alison Ulrich for help with preparation of the manuscript. Work in this laboratory is supported by Public Health Service grants R01 GM40993 and R01-GM11983 from the National Institute of General Medical Sciences.