Escherichia coli cytochrome c peroxidase is a respiratory oxidase that enables the use of hydrogen peroxide as a terminal electron acceptor

Significance

Hydrogen peroxide has been regarded exclusively as a hazard for bacteria; its sources and concentrations in natural habitats are uncertain. The cytochrome c peroxidase of Escherichia coli exhibits an expression pattern and flux rate that provides surprising insights into these issues. This periplasmic enzyme is induced only when H₂O₂ is present and molecular oxygen is absent. Intriguingly, it was ineffective as a defensive enzyme, but through its linkage to the quinone pool it did enable E. coli to respire using H₂O₂ as an anaerobic electron acceptor. We suggest that both chemical and biotic processes generate micromolar H₂O₂ at oxic/anoxic interfaces and that this scenario is common enough that microbes have evolved strategies to productively exploit the H₂O₂.

Abstract

Microbial cytochrome c peroxidases (Ccp) have been studied for 75 years, but their physiological roles are unclear. Ccps are located in the periplasms of bacteria and the mitochondrial intermembrane spaces of fungi. In this study, Ccp is demonstrated to be a significant degrader of hydrogen peroxide in anoxic Escherichia coli. Intriguingly, ccp transcription requires both the presence of H₂O₂ and the absence of O₂. Experiments show that Ccp lacks enough activity to shield the cytoplasm from exogenous H₂O₂. However, it receives electrons from the quinone pool, and its flux rate approximates flow to other anaerobic electron acceptors. Indeed, Ccp enabled E. coli to grow on a nonfermentable carbon source when H₂O₂ was supplied. Salmonella behaved similarly. This role rationalizes ccp repression in oxic environments. We speculate that micromolar H₂O₂ is created both biologically and abiotically at natural oxic/anoxic interfaces. The OxyR response appears to exploit this H₂O₂ as a terminal oxidant while simultaneously defending the cell against its toxicity.

The facultative anaerobe Escherichia coli lives adjacent to the epithelial layer of the mammalian gut, where it can respire by scavenging trace oxygen that diffuses into the lumen from the epithelial cells. When oxygen levels decline, E. coli can ferment, although it is less successful than the coresident obligate anaerobes. However, upon excretion, E. coli can thrive, whereas the obligate anaerobes enter a period of stasis.

In fully oxic habitats, E. coli must cope with reactive oxygen species that are generated internally through the adventitious oxidation of redox enzymes (1). Superoxide (O₂⁻) and hydrogen peroxide (H₂O₂) are potentially toxic, because they oxidatively inactivate cytoplasmic enzymes that use exposed [4Fe–4S] clusters or ferrous iron atoms as prosthetic cofactors (2–6). H₂O₂ also oxidizes the intracellular pool of unincorporated iron through the Fenton reaction, thereby generating hydroxyl radicals that can damage DNA (7). To protect itself from these oxidants, E. coli—like virtually all organisms—routinely synthesizes superoxide dismutases to keep O₂⁻ levels low and catalase and NADH peroxidase to minimize H₂O₂. As a consequence, wild-type E. coli grows well in oxic environments.

In addition to these basal protections, most bacteria have the capacity to ramp up their defenses when they encounter extracellular H₂O₂ (8, 9). Exogenous H₂O₂ moves freely through porins into the periplasm, and because it is small and uncharged, it can gradually diffuse across the cytoplasmic membrane into the cell interior, where it poses a hazard (10). When H₂O₂ levels rise in E. coli, the H₂O₂ directly oxidizes the sensory cysteine residue of the OxyR transcription factor, triggering the formation of a disulfide bond that locks the protein into an activated conformation (11, 12). OxyR then activates the expression of defensive genes (13). Catalase and NADH peroxidase activities are boosted more than 10-fold. Suf proteins improve the repair of iron–sulfur clusters (14, 15), the manganese importer MntH allows Mn to replace ferrous iron in mononuclear enzymes (16, 17), and the miniferritin Dps protects DNA from Fenton chemistry by sequestering loose iron (18–20). When H₂O₂ levels later fall, glutaredoxins deactivate OxyR by reducing its disulfide bond (21).

Data indicate that ∼200 nM steady-state intracellular H₂O₂ is sufficient to activate this response (10, 21). Because the cytoplasmic membrane is only semipermeable to H₂O₂ and the cytoplasm contains robust peroxidase and catalase activities, the internal H₂O₂ concentration is 5- to 10-fold lower than the external H₂O₂ concentration (10). Accordingly, measurements indicate that external H₂O₂ levels must exceed 2 μM to induce the intracellular OxyR response (21). Thus, one infers that despite dwelling in mostly hypoxic environments, E. coli must encounter micromolar levels of external H₂O₂ at some stage in its normal lifestyle. Strikingly, even obligate anaerobes manifest H₂O₂-inducible defenses, mediated either by OxyR or by PerR (22–24). The implication is that microbes may confront toxic doses of H₂O₂ even in low-oxygen environments.

Such a conclusion might resolve the conundrum of cytochrome c peroxidases. The biochemical activity of these enzymes was first described in 1940 (25), and although their reaction mechanism has been elaborated in detail (26), their physiological role continues to perplex. Notably, the bacterial enzymes are expressed only under anoxic conditions (27, 28). We now report that through YhjA, a cytochrome c peroxidase homolog, E. coli can exploit H₂O₂ as a terminal oxidant in a form of anaerobic respiration. This represents a surprising positive use of H₂O₂ in metabolism. YhjA is induced only when oxygen is scarce and H₂O₂ is present, implying that micromolar H₂O₂ can be found at oxic–anoxic interfaces. We suggest that YhjA be renamed Ccp, and we continue to use Ccp throughout this paper.

Results

Cytochrome c Peroxidase Is a Fourth Enzyme That Scavenges H₂O₂.

E. coli has three enzymes that are known to degrade H₂O₂ in vivo: the KatG and KatE catalases, plus the NADH peroxidase AhpCF. For convenience, mutants that lack all three enzymes (katG katE ahpCF) are termed hydroperoxidase-deficient (Hpx⁻) (29). When Hpx⁻ mutants are grown aerobically, they do not clear micromolar H₂O₂ from medium at a significant rate (Fig. 1). However, we were surprised to find that when the same strain was grown and exposed to H₂O₂ under anoxic conditions, it displayed substantial ability to degrade H₂O₂ (Fig. 1, rightmost bar).

Fig. 1.

Under anoxic conditions, a new H₂O₂-degrading activity appears. Wild-type (MG1655) and Hpx⁻ (LC106) cells were grown and assayed aerobically or anaerobically. Rates of H₂O₂ scavenging were measured as described. Error bars in this and subsequent figures represent SEM of three independent experiments. Asterisks represent statistical significance [*P ≤ 0.05; **P ≤ 0.01; ***P ≤ 0.001; not significant (ns), P > 0.05].

The culture filtrate did not degrade H₂O₂ (Fig. S1), indicating that the scavenging activity resided within the cells. We considered the possibility that H₂O₂ might be eliminated through Fenton reactions with the cytoplasmic pool of unincorporated iron. However, our calculations indicated that the cell was unlikely to contain enough iron to support the observed rate of H₂O₂ clearance (SI Materials and Methods), and indeed the scavenging activity persisted when cells were perfused with dipyridyl, a chelator that fully blocks intracellular Fenton chemistry (7) (Fig. S1).

Fig. S1.

H₂O₂ scavenging in Hpx⁻ strains is not due to reactions with secreted metabolites or intracellular Fenton chemistry. The rate of H₂O₂ scavenging was measured with Hpx⁻ cells, with the filtrate of anaerobically grown Hpx⁻ (LC106) cells, or with cells treated with 1 mM dipyridyl (a cell-permeable iron chelator). ND, the rate was below the detection limit (0.1 μM H₂O₂/min-OD). ns, not significant compared with Hpx⁻.

Therefore, we focused upon E. coli enzymes that have been reported to display peroxidase activities in vitro: thiol peroxidase (Tpx) (30), bacterioferritin comigratory protein (BCP) (31), a glutathione peroxidase homolog (BtuE) (32), a predicted cytochrome c peroxidase (YhjA, or Ccp) (28), and osmotically inducible peroxiredoxin (OsmC) (33). The physiological roles of these enzymes are uncertain, and they have not been reported to degrade H₂O₂ in vivo. We also considered the cytochrome bd-I, bd-II (AppBC), and bo terminal oxidases. These enzymes terminate aerobic respiration by transferring electrons from respiratory quinones to oxygen, but they also exhibit some peroxidase activity, because H₂O₂ is a formal intermediate in the oxygen-reduction process (34, 35). Each of the genes was deleted from the Hpx⁻ parent strain, and the rate of H₂O₂ scavenging was then remeasured under anoxic conditions. The rate was substantially diminished only in the Δccp mutant, and this effect was reversed when ccp was overexpressed from a plasmid (Fig. 2). The other mutations had no further impact even when they were introduced into the Hpx⁻ Δccp strain. We conclude that Ccp is an authentic scavenger, whereas the other proteins play little role in H₂O₂ clearance under these conditions. We do not know the source of the slight scavenging activity that persists in the Hpx⁻ Δccp strain, and it is not considered further here.

Fig. 2.

Ccp scavenges H₂O₂ in anoxic Hpx⁻ cells. The Hpx⁻ parent strain and mutant derivatives were grown anaerobically, and the rate of H₂O₂ scavenging was measured. Cross-hatched bars: The cytochrome c maturation machinery is required for Ccp activity. Spotted bar, far right: the Hpx⁻ Δccp mutant was genetically complemented using pACYC184-ccp under its own promoter. Asterisks represent statistical significance (*P ≤ 0.05; **P ≤ 0.01; ***P ≤ 0.001; ns, P > 0.05). The single mutants other than Hpx⁻ Δccp and Hpx⁻ Δccm were not significantly different from Hpx⁻. None of the double mutants were significantly different from Hpx⁻ Δccp. Strains used: Hpx⁻ (LC106), Hpx⁻ Δbcp (MK150), Hpx⁻ Δtpx (MK154), Hpx⁻ ΔbtuE (MK158), Hpx⁻ Δccp (MK146), Hpx⁻ Δcyd Δcyo (SSK53), Hpx⁻ ΔappBC (MK180), Hpx⁻ ΔosmC (MK208), Hpx⁻ Δccp Δbcp (MK172), Hpx⁻ Δccp Δtpx (MK174), Hpx⁻ Δccp ΔbtuE (MK176), Hpx⁻ Δccp ΔappBC (MK182), Hpx⁻ Δccp Δcyd (MK164), Hpx⁻ Δccp Δcyo (MK166), Hpx⁻ Δccp Δcyd Δcyo (MK170), Hpx⁻ Δccp ΔosmC (MK210), Hpx⁻ Δccm (MK198), Hpx⁻ Δccp Δccm (MK418), and Hpx⁻ Δccp pAcyc184-ccp (MK430).

Ccp Is an Authentic Peroxidase That Derives Electrons from the Quinone Pool.

The yhjA (here, ccp) gene is annotated as a potential cytochrome c peroxidase. Well-studied Ccp enzymes contain two c-type hemes and localize in the mitochondrial intermembrane space or bacterial periplasm. In vitro, these enzymes can receive electrons from reduced respiratory cytochrome c protein; the Ccp can then transfer electrons to H₂O₂, reducing it to water (25, 36–38).

The E. coli respiratory chain does not use cytochrome c, and its Ccp belongs to a second class that has a third c-type heme, which serves as the electron entry port when artificial reductants are provided (26). E. coli has a single known pathway for synthesizing and exporting c-type heme to periplasmic proteins; this ccmABCDEFGH operon has been shown to provide c-type hemes to periplasmic nitrite reductase, which is encoded by the adjacent nir operon. We found that the Hpx⁻ ΔccmABCDEFGH and the Hpx⁻ Δccp ΔccmABCDEFGH mutants phenocopied the Hpx⁻ Δccp mutant, confirming that the Ccm pathway is also the source of the c-type heme that activates Ccp (Fig. 2).

Because the Fenton reaction is chemically simple, it can be catalyzed at modest rates by many enzymes that contain solvent-exposed iron cofactors. However, such adventitious chemistry often requires high concentrations of H₂O₂. For example, although the bo and bd-I terminal oxidases can reduce H₂O₂ to water (35, 39), they do so effectively only at millimolar H₂O₂ concentrations that vastly exceed the low-micromolar concentrations of H₂O₂ that are likely to occur in nature (Fig. S2). In contrast, by measuring H₂O₂ clearance in an Hpx⁻ background, we determined that the apparent K_m of Ccp in intact cells was 5.2 ± 0.6 μM (Fig. 3A). We conclude that H₂O₂ is likely to be its physiological substrate.

Fig. S2.

Cytochrome bo and bd oxidases require millimolar doses of H₂O₂ for effective turnover. The rates of NADH oxidation by the bo and bd oxidases were measured using inverted membrane vesicles under anoxic conditions. The vesicles were prepared from Δcyd (AL454) or Δcyo (JEM1513) strains. The figure shows a representative curve from three independent experiments. The cytochrome bd-1 oxidase also exhibits some catalase activity, but this action similarly requires millimolar levels of H₂O₂ for rapid turnover (82).

Fig. 3.

(A) Ccp has a low effective K_m for H₂O₂. Rates of anoxic H₂O₂ scavenging were measured with Hpx⁻ (LC106) and Hpx⁻ Δccp (MK146) cells. Triplicate measurements determined K_m(app) to be 5.2 ± 0.6 μM. A single trial is shown here. (B) Respiratory quinones are required for Ccp function. The Hpx⁻ strain (LC106) and its derivatives lacking ubiquinone (LC148); menaquinone (SSK6); both (LC160); ccp (MK146); or ubiquinone, menaquinone, and ccp (MK184) were grown anaerobically in LB medium, and the rates of H₂O₂ scavenging were measured. Asterisks represent statistical significance compared with the Hpx⁻ strain (*P ≤ 0.05; **P ≤ 0.01; ***P ≤ 0.001; ns, P > 0.05).

Ccp has been predicted to be a single-pass enzyme that is tethered to the cytoplasmic membrane, with the bulk of its polypeptide in the periplasm (40). The electron transport chain in the cytoplasmic membrane is an available source of electrons for periplasmic redox enzymes. In vitro, some quinone analogs were found to be capable of reducing some bacterial Ccp enzymes (41–43), although the activity was not detected with the Ccp of E. coli (42). We observed that Hpx⁻ ubi mutants retained Ccp scavenging activity, but the activity was diminished in men mutants and was essentially absent from ubi men double mutants (Fig. 3B). Menaquinone is the primary quinone in membranes under the anoxic conditions of these experiments (44).

Periplasmic nitrite and nitrate reductases also acquire electrons from the quinone pool. In those cases, the structural genes for the catalytic enzymes sit alongside genes that encode bridging proteins that deliver electrons from the quinones to the catalytic enzymes. No analogous gene accompanies ccp. We verified that the Ccp scavenging activity persisted in mutants lacking the Nap and Nrf bridge proteins (Fig. S3); we infer that its association with the cytoplasmic membrane allows Ccp to receive electrons directly from the quinone pool.

Fig. S3.

Nap and Nrf are not required for H₂O₂ scavenging. The rate of H₂O₂ scavenging was measured using anoxic Hpx⁻ Δnrf (MK218) and Hpx⁻ Δnap (MK212) cells. The slight decrement in the Δnap mutant may be due to a polar effect upon downstream cytochrome c biosynthetic genes. ns, not significant compared with Hpx⁻.

Ccp enables anaerobic respiration with H₂O₂ as the terminal oxidant.

Partridge et al. (28) demonstrated previously that E. coli expresses ccp (called yhjA in their study) only under anoxic conditions, and mutants lacking either the Fnr or OxyR transcription factors failed to do so. They identified Fnr and OxyR binding sites upstream of the transcriptional start site. In their experiments, the involvement of OxyR appeared not to require the presence of H₂O₂. We found that the addition of H₂O₂ stimulated Ccp activity by threefold (Fig. 4A). We generated a ccp–lacZ transcriptional fusion and reexamined this pattern. Anoxia alone was insufficient to promote full expression; H₂O₂ was also required (Fig. 4B). This pattern was most easily observed with Hpx⁻ strains that do not rapidly clear H₂O₂ from the medium, but H₂O₂ stimulated expression in wild-type cells as well (Fig. S4). Neither fnr nor oxyR mutants showed induction (Fig. 4B). The requirement for H₂O₂ was relieved by expression of OxyR2, a mutant form of OxyR that remains activated even in the absence of H₂O₂ (Fig. 4C). We suspect that the contamination of LB by H₂O₂ (45) may previously (28) have given the misleading impression that H₂O₂ was unnecessary for ccp induction. Collectively, the dual regulation by the OxyR and Fnr transcription factors ensures that E. coli expresses ccp only if molecular oxygen is absent and H₂O₂ is present. The requirement for H₂O₂ confirmed our view that it is the natural substrate of the enzyme.

Fig. 4.

Strong ccp expression requires both the absence of oxygen and the presence of H₂O₂. (A) The rates of H₂O₂ scavenging were measured in Hpx⁻ cells grown in oxic or anoxic conditions. Where indicated, cells were incubated with 40 μM H₂O₂ for 1 h before assay. (B) β-galactosidase activity from the transcriptional ccp′–lacZ⁺ reporter fusion was measured in Hpx⁻ (MK250), Hpx⁻ Δfnr (MK274), and Hpx⁻ ΔoxyR (MK278) mutants grown in oxic or anoxic media. Where indicated, cells were incubated with 40 μM H₂O₂ for 1 h before harvesting. (C) β-galactosidase activity from the fusion was measured in wild-type cells containing either a pACYC184-oxyR2 plasmid (expressing a constitutively active form of OxyR) (MK346) or empty vector (MK344), grown anaerobically. Asterisks represent statistical significance (*P ≤ 0.05; **P ≤ 0.01; ***P ≤ 0.001; ns, P > 0.05).

Fig. S4.

ccp expression requires both the absence of oxygen and the presence of H₂O₂. β-galactosidase activity from the transcriptional ccp′–lacZ⁺ reporter fusion was measured in WT (MK246) cells grown in oxic or anoxic media. Where indicated, cells were incubated with 40 μM H₂O₂ for 1 h before harvesting. Asterisks represent statistical significance (*P ≤ 0.05; **P ≤ 0.01; ***P ≤ 0.001; ns, P > 0.05).

Molecular oxygen is an obligatory precursor in all known routes of H₂O₂ formation, so the fact that Ccp is synthesized only in its absence was perplexing. However, we and others have observed that both facultative and obligately anaerobic bacteria continue to express oxidative defenses even when oxygen is absent, suggesting that H₂O₂ can find its way into anoxic habitats (Discussion). Therefore, we considered the possibility that in anoxic environments Ccp might replace AhpCF as the primary scavenger of H₂O₂. However, an ahpC–lacZ fusion demonstrated that AhpCF continues to be synthesized at full force in anaerobic cells (Fig. S5A). In fact, AhpCF and catalase remained the major enzymes involved in clearing H₂O₂ from the bulk medium, because Hpx⁻ strains were defective at this activity. Ccp was a minor contributor (Fig. S5B).

Fig. S5.

Under anoxic conditions, cytoplasmic enzymes remain the primary scavengers of H₂O₂. (A) β-galactosidase activity of the transcriptional ahpCF′–lacZ⁺ reporter fusion was measured in Hpx⁻ (MK188) cells grown anaerobically or aerobically. (B) The rate of H₂O₂ scavenging was measured in anoxic wild-type (MG1655) cells, the Hpx⁻ (LC106) mutant, and the Δccp (MK416) mutant. Asterisks represent statistical significance (*P ≤ 0.05; **P ≤ 0.01; ***P ≤ 0.001; ns, P > 0.05).

The established role of scavenging enzymes is to shield cells from the toxic effects of H₂O₂. It seemed plausible that Ccp might act as a new layer of defense that would degrade exogenous H₂O₂ before it could diffuse into the cytoplasm, where H₂O₂-sensitive enzymes are located. When wild-type cells and ccp mutants were challenged with millimolar H₂O₂, the rates of killing were indistinguishable (Fig. S6). However, that protocol might fail to elicit the protective effect of a peroxidase because the H₂O₂ dose grossly exceeds the effective K_m of the enzyme, such that it can scavenge only a minute fraction of the H₂O₂. An alternative is to challenge cells with low doses of exogenous H₂O₂ that just barely poison metabolism. To do so, we tested the vulnerability of ccp mutants to a gradient of exogenous H₂O₂ through zone-of-inhibition experiments using a minimal glucose medium (Fig. 5 A and B). In this situation, the poisoning of amino acid biosynthetic enzymes sets the H₂O₂ sensitivity of the cell at 0.5–1 μM intracellular H₂O₂ (46); we thought this arrangement offered the best chance to detect any defensive impact of Ccp. However, the ccp mutants were again no more sensitive than wild-type cells. Only in the Hpx⁻ background did Ccp exert an impact, which we attribute to its providing the sole route of clearance of H₂O₂ from the bulk medium. These data resemble zone-of-inhibition results that were previously obtained with other bacteria (43, 47–50) and or fungi (51, 52), in which mutants showed either only a marginal increase in sensitivity, or none at all. In contrast, the Hpx⁻ strain was extremely sensitive, showing that in E. coli the cytoplasmic peroxidase and catalase activities play a much greater role in protecting the cell from H₂O₂.

Fig. S6.

A Δccp mutant is not more sensitive to millimolar doses of H₂O₂ than is the wild-type strain. Wild-type (MG1655) and Δccp mutant (MK416) cells were grown to OD₆₀₀ ∼0.3 and challenged with 2.5 mM H₂O₂. Percent survival was calculated based on cfus. In contrast to an earlier report (28), we did not observe any difference in the sensitivities of these strains.

Fig. 5.

Ccp cannot protect the cytoplasm from exogenous H₂O₂. (A) Wild-type (MG1655), Δccp (MK416), Hpx⁻ (LC106), and Hpx⁻ Δccp (MK146) mutant cells were grown to OD₆₀₀ ∼0.1 and spread on a plate. Disks soaked in H₂O₂ were put on the plate, and the diameter of zone of inhibition was measured after 24 h. (B) Data from A are shown as bar graphs. (C) Modeling (SI Materials and Methods) shows that H₂O₂ exchange between the external environment and the periplasm is too fast for Ccp to significantly diminish the periplasmic H₂O₂ level. Therefore, Ccp has minimal effect on H₂O₂ entry into the cytoplasm. Steady-state fluxes (in %) are calculated relative to the rate of H₂O₂ entry into the periplasm. Asterisks represent statistical significance (*P ≤ 0.05; **P ≤ 0.01; ***P ≤ 0.001; ns, P > 0.05).

To understand why Ccp fails to shield the cytoplasm, we modeled H₂O₂ fluxes as H₂O₂ moves from the external environment into the periplasm. Outer membrane porins allow very rapid H₂O₂ movement into the periplasm (53), whereas the cytoplasmic membrane exhibits a membrane permeability coefficient that is lower by at least two orders of magnitude (10, 54). Calculations showed that the activity of Ccp could diminish the periplasmic concentration of H₂O₂ by only 0.1% in the face of its rapid exchange with the external environment (SI Materials and Methods). Accordingly, the rate of subsequent influx into the cytoplasm was effectively unchanged by Ccp (Fig. 5C). Thus, it seems implausible that the main role of Ccp is to protect the cell from exogenous H₂O₂.

Because Ccp mediates electron flow from respiratory quinones to H₂O₂, its physiological role may be to exploit H₂O₂ as a terminal electron acceptor, an idea that was broached by Atack and Kelly (26). The regulatory data fit this model, as Ccp expression was blocked when oxygen, a superior acceptor, is available. Further, the presence of oxygen blocks Ccp turnover by competing for the electrons carried by the quinone pool (Fig. 6A). To probe further, we tested the effects upon ccp expression of added nitrate and fumarate, which are alternative anaerobic electron acceptors. Nitrate is a better respiratory substrate than fumarate, because nitrate reductase provides an additional coupling site in generating proton-motive force. Like oxygen, nitrate inhibited Ccp-mediated scavenging and ccp–lacZ expression (Fig. 6 B and C). E. coli has three transcription factors that respond to the presence of nitrate—NarL, NarP, and ArcA—but experiments failed to clarify which of these transcription factors might be involved in the repression (55) (Fig. S7).

Fig. 6.

Oxygen and nitrate block both Ccp activity and synthesis. (A) Hpx⁻ (LC106) cells were grown anaerobically, and the rate of H₂O₂ scavenging was measured in absence or presence of oxygen. (B) Hpx⁻ (LC106) cells were both grown and assayed for H₂O₂ scavenging either without an electron acceptor or in the presence of fumarate, nitrate, or oxygen. (C) Expression of the transcriptional ccp′–lacZ⁺ reporter fusion in Hpx⁻ (MK250) cells grown either without an electron acceptor or in the presence of fumarate, nitrate, or oxygen. Cells were incubated with 40 μM H₂O₂ for 1 h before harvesting. Asterisks represent statistical significance compared with anoxic assay in A and no terminal acceptor in B and C (*P ≤ 0.05; **P ≤ 0.01; ***P ≤ 0.001; ns, P > 0.05).

Fig. S7.

NarL, NarP, and ArcA do not mediate the inhibition of ccp transcription by nitrate. β-galactosidase activity of the transcriptional ccp′–lacZ⁺ reporter fusion was measured in Hpx⁻ (MK250), Hpx⁻ ΔnarL (MK404), Hpx⁻ ΔnarP (MK412), and Hpx⁻ ΔarcA (MK408) strains cultured with or without nitrate. If indicated, cells were incubated with 40 μM H₂O₂ for 1 h before harvesting. Asterisks represent statistical significance (*P ≤ 0.05; **P ≤ 0.01; ***P ≤ 0.001; ns, P > 0.05).

Fluxes through various E. coli metabolic pathways vary over five orders of magnitude. If the role of Ccp were to enable anaerobic respiration, the flux through Ccp should approximate fluxes to other electron acceptors. The rate of respiration through Ccp to H₂O₂ was measured under fully induced conditions; it was elevated threefold compared with the noninduced conditions that had been sampled in Fig. 1 (Fig. 7). This rate was found to be somewhat less than the respiratory rates with molecular oxygen and nitrate but comparable to the respiratory rate with fumarate. Thus, although the Ccp flux is insufficient to shield the cell from H₂O₂, it is sufficient to provide a respiratory benefit.

Fig. 7.

Rate of respiration of H₂O₂ is comparable to that of fumarate. Comparison of respiration rates by wild-type cells with different electron acceptors in glycerol medium. The grid pattern extension of the H₂O₂ bar represents calculated V_max, because unlike the other acceptors, H₂O₂ was provided at a subsaturating (10 μM) concentration. See Materials and Methods for details. The rate of respiration by fumarate was calculated in SI Materials and Methods.

To test this idea more directly, we monitored anaerobic growth with glycerol as sole carbon source. Glycerol catabolism depends upon the oxidation of α-glycerolphosphate by the respiratory GlpABC dehydrogenase complex, which transfers electrons to the menaquinone pool. Thus, respiration can proceed only if an exogenous terminal oxidant is available. We tested whether H₂O₂ could support growth. Unlike other respiratory substrates, which can be added in millimolar concentrations, H₂O₂ must be maintained at micromolar levels to avoid toxicity; therefore, 5 μM H₂O₂ was supplied, and the H₂O₂ levels in the medium were monitored and replenished periodically. Dilute cells were used so they would not exhaust the H₂O₂ too quickly, and growth was tracked by viable cell counts rather than optical density. After the preculture electron acceptor (fumarate) was removed, cells divided twice more and then stopped. Fig. 8 shows that H₂O₂ enabled continued growth. Growth did not occur if ccp was deleted, and it was restored in the complemented strain.

Fig. 8.

Ccp allows respiratory growth using H₂O₂ as the final electron acceptor. (A) Wild-type cells were grown in anoxic glycerol medium without any electron acceptor (○) or in the presence of 40 mM nitrate (▲), 25 mM fumarate (▪), or 5 μM H₂O₂ (●). Viable cells (cfus) were determined at different time points. Residual growth in the absence of H₂O₂ appears to be due to trace oxygen in the anaerobic chamber (Fig. S8). (B) WT and Δccp (MK416) strains were grown in anoxic glycerol medium in the presence or absence of 5 μM H₂O₂. (C) Δccp mutants were complemented with the ccp gene under its own promoter in a plasmid. Strains with pACYC184-ccp (MK436) and the empty vector (MK432) were grown in presence or absence of 5 μM H₂O₂. (D) The initial (t = 0) and the final (14-h) time points from three biological replicates of A were used to calculate the growth ratio for each growth condition. (E) The initial (t = 0) and the final (14-h) time points from three biological replicates of B were used to calculate the growth ratio for each growth condition. (F) The initial (t = 0) and the final (14-h) time points from three biological replicates of C were used to calculate the growth ratio for each growth condition. Asterisks represent statistical significance (*P ≤ 0.05; **P ≤ 0.01; ***P ≤ 0.001; ns, P > 0.05).

To rule out the possibility that the observed growth phenotype is specific to K12 strains, we confirmed that H₂O₂ also enabled the anaerobic respiration of an E. coli B strain and of Salmonella typhimurium (Fig. 9). In sum, the regulatory, flux, and growth data support the conclusion that the role of Ccp may be to exploit H₂O₂ as a respiratory substrate when better substrates are not available. Its location in the periplasm allows turnover even as AhpCF protects internal enzymes by keeping cytoplasmic H₂O₂ levels low.

Fig. 9.

E. coli B and S. typhimurium LT2 can grow using H₂O₂ as the final electron acceptor. (A) Wild-type E. coli B cells were grown in anoxic glycerol medium without any electron acceptor (○) or in the presence of 5 μM H₂O₂ (●). Viable cells (cfus) were determined at different time points. (B) The initial (t = 0) and the final (14-h) time points from three biological replicates of A were used to calculate the growth ratio for each growth condition. (C) Wild-type S. typhimurium LT2 was grown in anoxic glycerol medium without any electron acceptor (○) or in the presence of 5 μM H₂O₂ (●). Viable cells (cfus) were determined at different time points. (D) The initial (t = 0) and the final (14-h) time points from three biological replicates of A were used to calculate the growth ratio for each growth condition. Asterisks represent statistical significance compared with growth with no terminal acceptor (*P ≤ 0.05; **P ≤ 0.01; ***P ≤ 0.001; ns, P > 0.05).

SI Materials and Methods

Calculating an Upper Limit for the Rate of Fenton Reaction in E. coli.

We wish to calculate the maximum contribution that intracellular Fenton reactions might make to the degradation of H₂O₂ inside E. coli. The Fenton reaction degrades H₂O₂:

$${\text{H}}_2{\text{O}}_2+{\text{Fe}}^{2+}\to \text{HO}\cdot +{\text{OH}}^{-}+{\text{Fe}}^{3+}.$$

The hydroxyl radical that is formed will subsequently oxidize organic molecules. The ferric iron can be reduced back to the ferrous form by various reductants, including cysteine and reduced flavins (76, 77). During steady exposure to 10 μM H₂O₂, as used in measurements of H₂O₂ clearance (Fig. 1), the iron will be partitioned between ferric and ferrous forms in a ratio that has not been determined. To maximize the potential rate of Fenton chemistry, we will overestimate the amount of ferrous iron by assuming that all of the iron is in its reduced form. EPR measurements indicate that when E. coli is cultured in the defined media used in this study, its cytoplasm contains ∼30 μM of unincorporated iron that can participate in Fenton reactions (78). Therefore, the rate of H₂O₂ degradation is

$$\text{d}\left[{\text{H}}_2{\text{O}}_2\right]/\text{dt}=\text{k}\left[{\text{H}}_2{\text{O}}_2\right]\left[{\text{Fe}}^{2+}\right].$$

Previous studies reported the rate constants of the Fenton reaction for DNA-bound iron at 13 °C (k = 890 M⁻¹⋅s⁻¹) and 37 °C (k = 4,400 M⁻¹⋅s⁻¹). The reaction rate is faster for fully aqueous iron (k = 6,500 M⁻¹⋅s⁻¹ at 13 °C, and too fast to measure at 37 °C). If the activation energies for bound and aqueous iron as similar, we can estimate a rate of 32,000 M⁻¹⋅s⁻¹ for aqueous iron at 37 °C. Then, inside the cell,

$$\begin{array}{c}\text{d}\left[{\text{H}}_2{\text{O}}_2\right]/\text{dt}=\text{k}\left[{\text{H}}_2{\text{O}}_2\right]\left[{\text{Fe}}^{2+}\right]=\left(\mathrm{32,000}\hspace{0.17em}{\text{M}}^{-1}·{\text{s}}^{-1}\right)\times \left({10}^{-5}\hspace{0.17em}\text{M}\right)\times \left(30\times {10}^{-6}\hspace{0.17em}\text{M}\right)\\ =9.6\hspace{0.17em}\mathrm{\mu }\text{M}/\text{s}\text{,}\hspace{0.17em}\text{or}\hspace{0.17em}580\hspace{0.17em}\mathrm{\mu }\text{M}/\min .\end{array}$$

One liter of E. coli cells contains 0.5 mL cytoplasm per OD₆₀₀ (79). Therefore, the culture volume is 2,000× the cytoplasmic volume, and the rate of H₂O₂ clearance normalized to the full culture volume will be 0.29 μM⋅min⁻¹⋅OD₆₀₀. The actual measured rate of H₂O₂ clearance by Hpx⁻ cells (Fig. 1) was 6 μM⋅min⁻¹⋅OD₆₀₀. Therefore, even when generously overestimated, the Fenton reaction could not be more than a very minor contributor.

Calculating the Impact of Ccm Upon Periplasmic H₂O₂ Concentrations.

It has been shown previously that the cytoplasmic volume of E. coli cells growing exponentially in LB medium is 3.23 × 10⁻¹⁸ m³ (79). The volume of periplasm (Vol_peri) is ∼25% of the cytoplasm, or 8.08 × 10⁻¹⁹ m³. If E. coli cells are modeled as cylinders with two half-spheres at both ends, and stipulating that the cell length is 6× the radius, the surface area of the inner (A_in) and outer (A_out) membrane would be 1.35 × 10⁻⁷ cm² and 1.57 × 10⁻⁷ cm², respectively. These statements are approximately true, and deviations can be shown to have little impact upon final conclusions.

The H₂O₂ concentration in the periplasm is determined by the balance between its rate of entry through the outer membrane (V_in) and the three processes that clear it from the periplasm: its rate of efflux back to the environment through the outer membrane (V_efflux), the rate at which it passes through the inner membrane into the cytoplasm (V_cyto), and the rate of its degradation by Ccp (V_ccp). Reentry from the cytoplasm is essentially zero, due to efficient scavenging by cytoplasmic Ahp (10). Thus, in units of molarity/time,

$$\text{d}{\left[{\text{H}}_2{\text{O}}_2\right]}_{\text{peri}}/\text{dt}={\text{V}}_{\text{in}}–{\text{V}}_{\text{efflux}}–{\text{V}}_{\text{cyto}}–{\text{V}}_{\text{ccp}}.$$

The rate at which a solute crosses a membrane is proportional to solute concentration, membrane surface area (A), and the membrane permeability coefficient for that solute (P). Representing fluxes (J) in units of moles/time:

$$\begin{array}{c}\text{J}\left(\mathrm{mol}/\text{s}\right)=\left[{\text{H}}_2{\text{O}}_2\right]\times \text{P}\left(\text{cm}/\text{s}\right)\times \text{A}\left({\text{cm}}^2\right)\times {10}^{-3}\left(\text{L}/{\text{cm}}^3\right)\\ {\text{J}}_{\text{in}}={\left[{\text{H}}_2{\text{O}}_2\right]}_{\text{out}}\times {\text{P}}_{\text{out}}\left(\text{cm}/\text{s}\right)\times {\text{A}}_{\text{out}}\left({\text{cm}}^2\right)\times {10}^{-3}\left(\text{L}/{\text{cm}}^3\right)\end{array}$$

The permeability coefficient of the cytoplasmic membrane has been measured to be 1.6 × 10⁻³ cm/s (10, 54). Permeability across the outer membrane is primarily through porins; a lower limit of their permeability [determined for glycerol (53)] is 50 × 10⁻³ cm/s. The true value will be even higher for the smaller molecule H₂O₂, but the glycerol value will suffice to make the point. In this study, Ccp action upon H₂O₂ was measured with 10 μM H₂O₂.

$$\begin{array}{c}{\text{J}}_{\text{in}}=10\times {10}^{-6}\left(\text{M}\right)\times 50\times {10}^{-3}\left(\text{cm}/\text{s}\right)\times 1.57\times {10}^{-7}\left({\text{cm}}^2\right)\times {10}^{-3}\left(\text{L}/{\text{cm}}^3\right)\\ =7.85\times {10}^{-17}\hspace{0.17em}\mathrm{mol}/\text{s}\end{array}$$

$$\begin{array}{c}{\text{J}}_{\text{efflux}}={\left[{\text{H}}_2{\text{O}}_2\right]}_{\text{peri}}\times 50\times {10}^{-3}\left(\text{cm}/\text{s}\right)\times 1.57\times {10}^{-7}\left({\text{cm}}^2\right)\times {10}^{-3}\left(\text{L}/{\text{cm}}^3\right)\\ ={\left[{\text{H}}_2{\text{O}}_2\right]}_{\text{peri}}\times 7.85\times {10}^{-12}\hspace{0.17em}\text{L}/\text{s}\end{array}$$

$$\begin{array}{c}{\text{J}}_{\text{cyto}}={\left[{\text{H}}_2{\text{O}}_2\right]}_{\text{peri}}\times 1.60\times {10}^{-3}\left(\text{cm}/\text{s}\right)\times 1.35\times {10}^{-7}\left({\text{cm}}^2\right)\times {10}^{-3}\left(\text{L}/{\text{cm}}^3\right)\\ ={\left[{\text{H}}_2{\text{O}}_2\right]}_{\text{peri}}\times 2.16\times {10}^{-3}\hspace{0.17em}\text{L}/\text{s}\end{array}$$

The rate of peroxide clearance from media by Ccp when it is fully induced is 20 μM H₂O₂/min-OD or 0.33 μM H₂O₂/s-OD. One liter of E. coli cells at OD₆₀₀ = 1 contains 0.125 mL periplasm (80); that is, at OD₆₀₀ = 1, the ratio of medium volume to total periplasmic volume to media is 1,000/0.125. Therefore, the rate of H₂O₂ degradation within the periplasm will be 8,000× that in the total culture:

$${\text{V}}_{\text{ccp}}=0.33\left(\mathrm{\mu }\text{M}{\text{H}}_2{\text{O}}_2/\text{s}\right)\times \mathrm{8,000}=\mathrm{2,640}\hspace{0.17em}\mathrm{\mu }\text{M}/\text{s}$$

$$\text{The}\hspace{0.17em}\text{Ccp}\hspace{0.17em}\text{flux}\hspace{0.17em}{\text{J}}_{\text{ccp}}={\text{V}}_{\text{ccp}}\times \text{volume}\hspace{0.17em}\text{of}\hspace{0.17em}\text{the}\hspace{0.17em}\text{periplasm},\hspace{0.17em}\text{so}$$

$$\begin{array}{c}{\text{J}}_{\text{ccp}}=\mathrm{2,640}\left(\mathrm{\mu }\text{M}/\text{s}\right)\times 8.08\times {10}^{-19}\hspace{0.17em}{\text{m}}^3\times {10}^3\hspace{0.17em}\left(\text{L}/{\text{m}}^3\right)\times {10}^{-6}\hspace{0.17em}\mathrm{mol}/\mathrm{\mu }\mathrm{mol}\\ =1.99\times {10}^{-18}\hspace{0.17em}\mathrm{mol}/\text{s}\end{array}$$

At the steady state:

$${\text{J}}_{\text{in}}={\text{J}}_{\text{efflux}}\hspace{0.17em}–\hspace{0.17em}{\text{J}}_{\text{cyto}}\hspace{0.17em}–\hspace{0.17em}{\text{J}}_{\text{ccp}}$$

$$\begin{array}{c}7.85\times {10}^{-17}\mathrm{mol}/\text{s}={\left[{\text{H}}_2{\text{O}}_2\right]}_{\text{peri}}\times \left(7.85\times {10}^{-12}\hspace{0.17em}\text{L}/\text{s}+2.16\times {10}^{-3}\hspace{0.17em}\text{L}/\text{s}\right)+1.99\times {10}^{-18}\hspace{0.17em}\mathrm{mol}/\text{s}\\ \therefore {\left[{\text{H}}_2{\text{O}}_2\right]}_{\text{peri}}=9.48\hspace{0.17em}\mathrm{\mu }\text{M}\end{array}$$

By substitution the relative fluxes can be compared (% in parentheses):

$$\begin{array}{c}{\text{J}}_{\text{efflux}}=7.44\times {10}^{-17}\hspace{0.17em}\mathrm{mol}/\text{s}\hspace{0.17em}\left(94.9\%\right)\\ {\text{J}}_{\text{cyto}}=2.04\times {10}^{-18}\hspace{0.17em}\mathrm{mol}/\text{s}\hspace{0.17em}\left(2.6\%\right)\\ {\text{J}}_{\text{ccp}}=1.99\times {10}^{-18}\hspace{0.17em}\mathrm{mol}/\text{s}\hspace{0.17em}\left(2.5\%\right)\end{array}$$

Thus, the steady-state H₂O₂ concentration in the periplasm is primarily determined by the balance of H₂O₂ fluxes into and out of the outer membrane, with flux into the cytoplasm and consumption by Ccp having very minor effects. The impact of Ccp can be determined by reprising the calculations with the omission of Ccp flux:

$$\begin{array}{c}\text{d}{\left({\text{H}}_2{\text{O}}_2\right)}_{\text{peri}}/\text{dt}={\text{V}}_{\text{in}}\hspace{0.17em}–\hspace{0.17em}{\text{V}}_{\text{efflux}}\hspace{0.17em}–\hspace{0.17em}{\text{V}}_{\text{cyto}}\\ \text{d}{\left({\text{H}}_2{\text{O}}_2\right)}_{\text{peri}}/\text{dt}\times \text{cell}\hspace{0.17em}\text{volume}={\text{J}}_{\text{in}}\hspace{0.17em}–\hspace{0.17em}{\text{J}}_{\text{efflux}}\hspace{0.17em}–\hspace{0.17em}{\text{J}}_{\text{cyto}}\end{array}$$

The outcome is

$$\begin{array}{c}{\left[{\text{H}}_2{\text{O}}_2\right]}_{\text{peri}}=9.73\hspace{0.17em}\mathrm{\mu }\text{M}\\ {\text{J}}_{\text{efflux}}=7.64\times {10}^{-17}\hspace{0.17em}\mathrm{mol}/\text{s}\\ {\text{J}}_{\text{cyto}}\hspace{0.17em}2.10\times {10}^{-18}\hspace{0.17em}\mathrm{mol}/\text{s}.\end{array}$$

Thus, Ccp reduces the periplasmic H₂O₂ concentration by only 2.5%, and it therefore lessens the flux into the cytoplasm by an equivalent amount. Recalculations with any plausible change in cell dimensions do not change this outcome; if a higher OM permeability coefficient is used, consistent with the small size of H₂O₂, then the impact of Ccp will only be further diminished. The upshot is that Ccp action cannot shield the cytoplasm of an isolated cell from environmental H₂O₂.

Rate of Anaerobic Respiration in Glucose Medium with Fumarate as Terminal Electron Acceptor.

Fumarate is generated endogenously during the anaerobic catabolism of glucose, and it is then used as an electron acceptor for the respiratory chain. Tran et al. (81) reported that E. coli cells generate 0.07 mol succinate per mol of glucose catabolized in a minimal medium lacking amino acids. Using the same medium we determined that the doubling times of cells was 70 min and the cell yield was 0.5 OD₆₀₀ per 0.2% (11.1 mM) glucose provided. Therefore, the yield of succinate = 0.07 × 11.1 mM = 0.78 mM per 0.5 OD₆₀₀, or 1.6 mM/OD₆₀₀. The doubling time indicates a growth rate constant of 0.0099 min⁻¹, which applies equally to the production of end products. Thus, the rate of succinate formation in this medium = 1.6 mM/OD₆₀₀ × 0.0099 min⁻¹ = 15 mM/OD₆₀₀·min.

Discussion

Cytochrome c peroxidases are widely distributed among microbes. Previous studies demonstrated their activity in vitro, using artificial electron donors; here, experiments confirm that they degrade micromolar H₂O₂ at substantial rates in vivo. Ccps have generally been assumed to be defensive enzymes, but our analysis does not support this view, at least for E. coli Ccp. The enzyme cannot provide enough activity for isolated cells to lower the periplasmic H₂O₂ level below that of the surrounding medium, so Ccp does not shield the cytoplasm from H₂O₂. Further, because E. coli is a minor member of the intestinal flora, it is unlikely to assume the responsibility for clearing H₂O₂ from its environment. We do recognize that, in principle, if E. coli grew in a clonal biofilm with only a slow H₂O₂ influx from the surrounding environment, the scavenging activity that Ccp provides might help lower the H₂O₂ level within the biofilm itself. However, the scavenging activity provided by Ccp does not significantly exceed that which is already provided by cytoplasmic enzymes. Further, for Ccp to deplete H₂O₂ within a biofilm, the biofilm would have to be dense enough to limit H₂O₂ entry, but without simultaneously blocking the influx of the carbon sources that provide the respiratory electrons. Thus, Ccp could conceivably contribute as a defensive enzyme only in a very constrained set of circumstances.

However, our data do show that Ccp can allow E. coli to use H₂O₂ as a respiratory substrate. The regulatory controls upon ccp are more completely resolved for E. coli than for other bacterial or fungal Ccps, and they fit this model (Fig. 10). The transcriptional dependence upon OxyR dictates that the gene be transcribed only when H₂O₂ is present; Fnr collaborates to ensure that transcription only occurs when a better electron acceptor is not available. Apparent OxyR and Fnr binding sites have been identified (28). The details of promoter architecture and transcription factor interaction will be interesting, particularly because with other genes OxyR seems to operate independently of other transcription factors. This regulatory scheme is a nice demonstration that the two transcription factors are truly specific for their effector oxidants: in controlling ccp, Fnr is nonresponsive to H₂O₂, and OxyR is not activated by O₂ per se.

Fig. 10.

Proposed model for Ccp expression and function. H₂O₂ rapidly enters the periplasm through porins but penetrates the inner membrane at a lower rate. Activation of OxyR stimulates synthesis of catalase and NADH peroxidase, which keep the cytoplasmic H₂O₂ concentration low. In anoxic conditions, FNR is also activated, and together it and OxyR induce ccp. Ccp enters the periplasm and allows the respiratory chain to reduce H₂O₂ to H₂O. In isolated cells, Ccp enables anaerobic respiration, but it does not degrade H₂O₂ quickly enough to reduce the periplasmic or cytoplasmic H₂O₂ concentrations.

But the bigger picture presents an obvious question: In what circumstances might E. coli encounter H₂O₂ but not O₂? We suggest that at oxic–anoxic interfaces both abiotic and biotic processes may generate H₂O₂. Such interfaces occur along the margins of intestines and in static ground sediments, and their defining characteristic is that highly reducing environments containing reduced sulfur and metal species collide with an oxygenated one. In the intestinal lumen, sulfate-reducing bacteria generate millimolar amounts of sulfide (56, 57), which presumably diffuses to the margins where it will mix with oxygen. E. coli lives in those margins, as evidenced by the fact that to colonize a mouse intestine, it relies upon a specialized cytochrome oxidase with high oxygen affinity and the ability to function in the presence of sulfide (58–60). When reductants encounter oxygen, metal-catalyzed sulfide oxidation generates H₂O₂. Further, lactic acid bacteria commonly dwell in hypoxic environments. Although they are classic fermentative organisms, when oxygen becomes available many of them redirect substrate through pyruvate, lactate, or NADH oxidases. These enzymes produce H₂O₂ as a stoichiometric product (61, 62). Finally, Bacteroides species that predominate in the intestine generate substantial H₂O₂ when they encounter oxygen, presumably due to the adventitious autoxidation of their low-potential redox systems (63). Thus, at these interfaces, both chemical and biological processes are likely to diminish O₂ levels and send H₂O₂ diffusing into anoxic zones. We suspect that this is the source of the H₂O₂ that E. coli exploits. Indeed, Campylobacter ccp mutants exhibit colonization defects that are consistent with the notion that H₂O₂ is an important component of the intestinal environment (64).

Other bacteria also express Ccp only in hypoxic environments, suggesting that it serves a common purpose in most organisms. Seib et al. (65) noted that Neisseria gonorrhea encodes a Ccp enzyme, whereas its close relative N. meningitidis does not. The authors speculated that the key difference is that N. meningitidis dwells in the oxygen-rich nasopharynx, whereas N. gonorrhea lives in the urogenital tract where oxygen is scant and lactic-acid bacteria may excrete H₂O₂. Recently, Stacy et al. (66) noted a shift in the gene set used by Aggregatibacter actinomycetemcomitans when it was coinfected with Streptococcus gordonii in a murine thigh abscess model; they inferred that Aggregatibacter respiration was stimulated by a terminal oxidant that S. gordonii released. The data would fit the idea that Aggregatibacter Ccp was exploiting H₂O₂ that S. gordonii generates as a fermentation product.

The apparent K_m of E. coli Ccp is 5 μM, and the electron flux through Ccp would be enough to sustain growth only if H₂O₂ approached this level. We and others have modeled H₂O₂ flow through a porous outer membrane and a semipermeable inner membrane into a cytoplasm containing high titers of peroxidases and catalases (10, 54). The upshot is that micromolar H₂O₂ outside cells will generate an equivalent level in the periplasm but perhaps a 10-fold lower level in the cytoplasm. The set point of OxyR is submicromolar (10, 21), so this flux is enough to activate OxyR and to induce Ccp. This situation allows Ccp to use micromolar external H₂O₂ as a periplasmic oxidant while at the same time the threat to cytoplasmic biomolecules is minimal. The other activities comprising the OxyR response would presumably help to keep the cell fit even while it respired using H₂O₂.

Although catalase generates molecular oxygen when it degrades H₂O₂, the primary scavengers of H₂O₂ in enteric bacteria are NADH peroxidases (29, 67). Therefore, oxygen production by H₂O₂-fed cells was insufficient to enable cytochrome oxidase-dependent growth. In any case, in natural habitats any oxygen generated by cytoplasmic catalases would immediately escape the source bacterium before it could be employed for respiration, because oxygen crosses membranes at diffusion-limited rates.

The existence of H₂O₂ in microoxic environments may explain why many bacteria are programmed for robust anti-H₂O₂ responses—including obligate anaerobes, whose oxidative defenses otherwise seem counterintuitive. One might suppose that these systems are useful during the occasional entry of anaerobes into oxic environments—but some of their peroxidatic scavenging systems, including rubrerythrins, are ineffectual when oxygen levels are high enough to poison central metabolism (24). Implicitly, these peroxidases, like Ccp, must serve to degrade H₂O₂ in habitats that contain little oxygen. Interestingly, low-micromolar H₂O₂ levels are also predicted to obtain to the phagosomal vesicles. By chance, Salmonellae, which naturally move between intestinal and macrophage habitats, might be equipped by its OxyR regulon to handle both environments.

The employment of a toxin such as H₂O₂ as a respiratory substrate strikes an ironic note, but this evolutionary step recapitulates the adoption of O₂. Life emerged in an anoxic world; when O₂ accumulated 2 billion years later (68), cells were threatened by its propensity to deactivate enzymes that have radical or low-potential metal centers. Defensive enzymes arose. But the toxicity of oxygen did not preclude the simultaneous appearance of respiratory enzymes that exploit it as an electron acceptor. We infer that the story with H₂O₂ may be analogous.

Materials and Methods

Reagents.

All antibiotics (ampicillin, chloramphenicol, kanamycin, tetracycline, and spectinomycin), ortho-nitrophenyl-β-galactoside, 2,2′-bipyridyl, horseradish peroxidase, 30% hydrogen peroxide, casein acid hydrolysate, sulfanilamide, N-(1-naphthyl)ethylenediamine dihydrochloride, and NADH were purchased from Sigma-Aldrich. Amplex UltraRed reagent was obtained from Life Technologies.

Strains and Bacterial Growth.

Strains, primers, and plasmids used in this study are listed in Tables S1–S3. Deletion mutations were made using the λ-red recombinase (69) or ordered from the Keio collection (70). Mutant strains were assembled by P1 transduction (71), and the inheritance of mutant alleles was confirmed by PCR analysis. A malE::Tn10 mutation was used to cotransduce the ubiA420 allele; the malE mutation itself was inconsequential in these experiments because the media lacked maltose. Transcriptional lacZ fusions were created as described (72), and the fusions were recombined into the lambda attachment site so that the native alleles were retained.

Table S1.

Strains

Strain	Genotype	Source
MG1655	F− wild type	E. coli Genetic Stock Center
LC106 (Hpx−)	ΔahpF::kan Δ(katG::Tn10)1 Δ(katE12::Tn10)	2
BW25113	lacI rrnB ΔlacZ hsdK ΔaraBAD ΔrhaBAD	69
E. coli B	F− wild type	E. coli Genetic Stock Center
Salmonella typhimurium LT2	Wild type	Lab stock
AL427	Δ(katG17::Tn10)1 Δ(ahpCF1::cat)1 Δ(katE12::Tn10)1	Lab stock
Al443	F- λ- lacIq rrnBT14 ΔlacZWJ16 hsdR514 ΔaraBADAH33 ΔrhaBADLD78 ΔcydAB1::cat	Lab stock
AL454	As MG1655 ΔcydAB1::cat	Lab stock
JEM1513	As MG1655 with Δ(cyoABCDE)456::kan	Lab stock
JW1212-1	F− Δ(araD-araB)567 ΔlacZ4787(::rrnB-3) ΔnarL732::kan rph-1 Δ(rhaD-rhaB)568 hsdR514	E. coli Genetic Stock Center
JW2181-2	F− Δ(araD-araB)567 ΔlacZ4787(::rrnB-3) ΔnarP746::kan rph-1, Δ(rhaD-rhaB)568 hsdR514	E. coli Genetic Stock Center
JW4364-1	F- Δ(araD-araB)567 ΔlacZ4787(::rrnB-3) ΔarcA726::kan rph-1 Δ(rhaD-rhaB)568 hsdR514	E. coli Genetic Stock Center
KK220	F− Δ(pncA-xthA) thr-1 leuB6 proA2 his-4 thi-1 argE2 lacY1 galK2 rpsL supE44 ara-14 xyl-15 mtl-1 tsx-33 fnr(Bsm-Mlu)::Ω (SpR) zcj::Tn10	Lab stock
LC148	As LC106 with ubiA420∼malE::Tn10*	Lab stock
LC160	As LC148 with Δ(menA::cam)	Lab stock
MK144	As BW25113 with Δccp::cam	This study
MK146	As LC106 with Δccp::cam	P1(MK144) × LC106
MK148	As BW25113 with Δbcp::cam	This study
MK150	As LC106 with Δbcp::cam	P1(MK148) × LC106
MK152	As BW25113 with Δtpx::cam	This study
MK154	As LC106 with Δtpx::cam	P1(MK152) × LC106
MK156	As BW25113 with ΔbtuE::cam	This study
MK158	As LC106 with ΔbtuE::cam	P1(MK156) × LC106
MK160	As MK146 with camR gene flipped out	This study
MK162	As BW25113 with ΔcyoABCDE::cat	This study
MK164	As MK160 with ΔcydAB1::cat	P1(Al443) × MK160
MK166	As MK160 with ΔcyoABCDE::cam	P1(MK162) × Mk160
MK168	As MK164 with camR gene flipped out	This study
MK170	As MK168 with ΔcyoABCDE::cam	P1(MK162) × Mk168
MK172	As MK160 with Δbcp::cam	P1(MK148) × MK160
MK174	As MK160 with Δtpx::cam	P1(MK152) × MK160
MK176	As MK160 with ΔbtuE::cam	P1(MK156) × MK160
MK178	As BW25113 with ΔappBC::cam	This study
MK180	As LC106 with ΔappBC::cam	P1(MK178) × LC106
MK182	As MK160 with ΔappBC::cam	P1(MK178) × MK160
MK184	As LC160 with Δccp::cam	P1(MK144) × LC160
MK188	As LC106 with attλ::[pSJ501::pahpC’-lacZ+]∼cat	P1(SRK57) × LC106
Mk196	As BW25113 with Δ(ccmABCDEFGH)::cam	This study
MK198	As LC106 with Δ(ccmABCDEFGH)::cam	P1(MK196) × LC106
MK204	As BW25113 with ΔosmC::cam	This study
MK206	As BW25113 with Δ(napABCDFGH)::cam	This study
Mk208	As LC106 with ΔosmC::cam	P1(MK204) × LC106
MK210	As MK160 with ΔosmC::cam	P1(MK204) × MK160
MK212	As LC106 with Δ(napABCDFGH)::cam	P1(MK206) × LC106
MK216	As BW25113 with Δ(nrfABCDEFG)::cam	This study
MK218	As LC106 with Δ(nrfABCDEFG)::cam	P1(MK216) × LC106
MK240	As SJ130 with attλ::[pSJ501::yhjA’-lacZ+] ∼camR	This study
MK246	As MG1655 with attλ::[pSJ501::yhjA’-lacZ+] ∼camR	P1(MK204) × MG1655
MK250	As LC106 with attλ::[pSJ501::yhjA’-lacZ+] ∼camR	This study
MK274	As MK250 with Δfnr(Bsm-Mlu)::Ω (SpR) zcj::Tn10	P1(KK220) × MK250
MK278	As MK250 with ΔoxyR::spec	P1(XY20) × MK250
MK336	As MK246 with camR gene flipped out	This study
MK344	As MK336 with pACYC184	This study
MK346	As MK336 with pGSO58 (poxyR2)	This study
MK400	As AL427 with attλ::[pSJ501::yhjA’-lacZ+] ∼camR	This study
MK404	As MK400 with ΔnarL732::kan	P1(JW1212-2) × MK400
MK408	As MK400 with ΔarcA726::kan	P1(JW4364-1) × MK400
MK412	As MK400 with ΔnarP746::kan	P1(JW2181-2) × MK400
MK416	As MG1655 with Δccp::cam	P1(MK144) × MG1655
MK418	As MK160 with Δ(ccmABCDEFGH)::cam	P1(MK196) × MK160
MK420	As MK416 with camR gene flipped out	This study
MK430	As MK160 with pACYC184-ccp	This study
MK432	As MK420 with pACYC184	This study
MK436	As MK420 with pACYC184-ccp	This study
SJ130	As MG1655 with Δ(lacZ1::cat)1	Lab stock
SRK57	attλ::[pSJ501::PahpC’-lacZ+]∼cat	Lab stock
SSK53	As LC106 with cyd::cam Δ(cyoABCDE)456::kan linked to zaj::Tn10	Lab stock
SSK6	As LC106 with ΔmenA::cam	Lab stock
XY20	as MG1655 with ΔoxyR::spec	Lab stock

^*The ubiA420 mutation in LC148 is linked to malE::Tn10. Because none of the growth or assay conditions in this paper contained maltose, the malE mutation has no effect on the observed phenotype.

Table S3.

Plasmids

Plasmids	Characteristics	Source
pSJ501	pAH125 derivative with cat flanked by flp sites	15
pKD3	bla FRT cat FRT PS1 PS2 oriR6K	69
pKD46	bla PBAD gam bet exo pSC101 oriTS	69
pCP20	bla cat cI857 λPR flp pSC101 oriTS	74
pACYC184	TetR CmR	75
pGS058	pACYC184 containing oxyR2 [A233V]	75
pACYC184-ccp	pACYC184 containing ccp gene and promoter	This study

Table S2.

Primers

Primer	Sequence
ccp-deletion	5′-GAAAATGGTCTCACGTATTACCGCGATCGGCCTGGCTGGCGTCGCGTGTAGGCTGGAGCTGCTTCG-3′
	5′-CTGTGCAGGAAAGCTACGATATCATCCACATCCTCCTGCGGCAGCTCCATATGAATATCCTCCTTAG-3′
bcp-deletion	5′-GCCCACAGCCCCGCATCTGCGGACGCAGCAAATATTGAGCAAGCGTGTAGGCTGGAGCTGCTTCG-3′
	5′-GAAGTGAGGTATGCTGGCCGCAAGCGCAGCCAGCACGGAATGGAGCATATGAATATCCTCCTTAG-3′
btuE-deletion	5′-CGACGGTGAAGTGACCACGCTGGAGAAGTTCGCCGGTAATGTGCTGTGTAGGCTGGAGCTGCTTCG-3′
	5′-CGGCGTCATATCCGGGGAAAAACGCTGGATGACTTTTCCGTCCCTGCCCATATGAATATCCTCCTTAG-3′
tpx-deletion	5′-CCCCGGGCAATCACAAACGGGGTATGTAAGGGCCAGGCTTCCTCGTGTAGGCTGGAGCTGCTTCG-3′
	5′-GAGCTGCTTCGTAATCCGGCTCGGTGGTGATTTCATCCACCAGCTGGCCATATGAATATCCTCCTTAG-3′
appBC-deletion	5′-GTGGGATGTCATTGATTTATCGCGCTGGCAGTTTGCTCTGACCGCGCTGTAGGCTGGAGCTGCTTCG-3′
	5′-CTCGTTTTCGTTACGGCGGAGAGTTTCTGTTGTCATGCGCCCCCACCATATGAATATCCTCCTTAG-3′
osmC-deletion	5′-GGTCAGGCACACTGGGAAGGCGATATCAAACGCGGGAAGGGAACAGTTGTAGGCTGGAGCTGCTTCG-3′
	5′-GCATCAGGCATTCACGCGCCTGACGCGTCATCCGGCAATGCTTTACGCATATGAATATCCTCCTTAG-3′
cyoABCDE-deletion	5′-GCCCGCCCAGACTCCTTCAGCACTCCCCTTTTGTTATAACGCCCTTGTAGGCTGGAGCTGCTTCG-3′
	5′-GGTTTTGTTACCACACAGCAGCCAGCAGCGTATGCGAGTCCGGTACCATATGAATATCCTCCTTAG-3′
ccmABCDEFGH-deletion	5′-CTGAACGCAGGAGAGTGGGTACAAATCACCGGTAGCAACGGCGCGGTGTAGGCTGGAGCTGCTTCG-3′
	5′-ATCTCCCACGCGGCAACGGCTTCGCCAAATCGCTGCTGCTCAAAGGCATATGAATATCCTCCTTAG-3′
napABCDFGH-deletion	5′-GATTGATGCATCCCGTCGGGGCATACTCACTGGTCGCTGGCGCAAATGTAGGCTGGAGCTGCTTCG-3′
	5′-CCTGGCTCGACTTCACGCATATCGGGCAGCTTGTGCGCTATCCCTTTCATATGAATATCCTCCTTAG-3′
nrfABCDEFG-deletion	5′-TTGCCCCGCAGCATCCCGATCAATATCTCTCCTGGAAAGCCACCTTGTAGGCTGGAGCTGCTTCG-3′
	5′-CGACTCAACCAGCTGTGTTCGGTTAACTCGCGGTGAGTTGAGATCCACATATGAATATCCTCCTTAG-3′
ccp-promoter for lacZ fusion	5′-TTTCTGCAGCCGCAATCATGATATGCC-3′
	5′-TTTGAATTCCGGTGGTGAGTTATTGTG-3′
ccp gene for complementation	5′-TTTTCTAGACCGCAATCATGATATGCC-3′
	5′-TTTGGATCCGCCCAAACATTGCTGTTC-3′

Luria broth and base minimal A salts were of standard composition (71). Glucose–amino acid medium was minimal A salts supplemented with 0.2% casein acid hydrolysate, 0.2% glucose, 5 μg/mL thiamine, 0.02% MgSO₄, and 0.5 mM tryptophan. Where indicated, 40 mM nitrate or 25 mM fumarate was added. Uracil (1 mM) was added to the media in experiments that involved quinone biosynthetic mutants, because dihydroorotate dehydrogenase requires a quinone substrate.

β-galactosidase activity was measured as described (71), using cells that were grown to OD₆₀₀ of 0.1 and then challenged with H₂O₂ for 1 h if indicated. Protein concentrations were measured using Bradford assay (Coomassie protein assay reagent; Thermo Scientific) using BSA as the standard. All anoxic growth and assays were done using anoxic buffers, media, and reaction components in a Coy anaerobic chamber (Coy Laboratory Products, Inc.) under 85% N₂, 10% H₂, and 5% CO₂.

Hydrogen Peroxide Scavenging Assay.

Overnight cultures were diluted to OD₆₀₀ of 0.01 into fresh glucose–amino acids medium, and cells were grown to OD₆₀₀ of 0.2. Cells were then diluted to OD₆₀₀ of 0.02 in fresh medium containing 10 μM H₂O₂. The culture was maintained at 37 °C, and 1-mL samples were removed every 4 min for 20 min. The samples were centrifuged (1 min at 15,300 × g, 4 °C) and the supernatant was frozen on dry ice and stored at −80 °C freezer. The samples were subsequently thawed on ice, and the concentration of peroxide was measured using Amplex UltraRed/horseradish peroxidase (29) in a Shimadzu RF Mini-150 fluorometer.

Glycerol Growth Assays.

Cells were grown anaerobically overnight in minimal A medium supplemented with 40 mM glycerol, 25 mM fumarate, amino acids (histidine, tyrosine, isoleucine, leucine, valine, and phenylalanine, each at 0.5 mM), 0.02% magnesium sulfate, and 5 μg/mL thiamine. The branched-chain and aromatic amino acids were provided to avoid growth disruptions due to the potential inactivation of iron enzymes in their biosynthetic pathways. E. coli cannot catabolize these amino acids to use them as energy sources. Fumarate was included to enable anaerobic respiration. Cells were then precultured in the same medium to OD₆₀₀ of 0.1 and then washed four times with minimal A salts to remove fumarate. Bacteria were diluted to OD₆₀₀ of 0.0001 in the fresh 37 °C medium without fumarate. Respiratory oxidants (40 mM nitrate, 25 mM fumarate, or 5 μM H₂O₂) were added as indicated. The residual H₂O₂ levels were periodically determined by Amplex UltraRed analysis, and H₂O₂ was replenished as needed to restore the concentration to 5 μM. Samples were periodically removed, diluted, and plated on LB agar. Colonies were counted the next day, and the colony-forming units in the original culture were calculated. Slight residual growth in the absence of electron acceptors was apparent only at very low cell densities and probably indicates trace oxygen in the media (Fig. S8).

Fig. S8.

Trace oxygen is the probable source of slight residual anaerobic growth in the absence of added electron acceptors. Growth occurred only at very low cell densities (○); at higher cell densities that would quickly scavenge oxygen, no increase in biomass was perceptible (□). One sample culture is shown from two replicates.

NADH Oxidation Assay.

Inverted membrane vesicles were prepared as described (73) from ∆cyo and ∆cyd mutants that had been grown aerobically in LB medium to 0.25 OD₆₀₀. The NADH oxidation reaction was performed using anoxic reagents in the Coy chamber with inverted membrane vesicles, 50 μM NADH, and different concentrations of H₂O₂. NADH oxidation was measured using the extinction coefficient for NADH of 6,220 cm⁻¹⋅M⁻¹ at 340 nm.

Nitrate Reduction and Oxygen Consumption Assays.

Wild-type cells were grown to 0.25 OD₆₀₀ in glucose–amino acid medium supplemented with 40 mM nitrate. Cells were washed three times with minimal A salts to remove the residual nitrite and inoculated to 0.020 OD₆₀₀ in glucose-amino acid medium. The concentration of nitrite was measured every 30 min using a modified version of the Griess assay. Sulfanilamide (250 μL of 2% wt/vol in 5% HCl), N-(1-naphthyl)ethylenediamine dihydrochloride (250 μL of 0.1% wt/vol in water) and 500 μL sample were mixed, and the absorbance at 540 nm was measured after 20 min of incubation in room temperature.

The rate of oxygen consumption was determined during aerobic growth in glycerol medium, using a Clark electrode (Micrometrix). Wild-type cells were grown anaerobically from 0.01 to 0.1 OD₆₀₀. Cells were moved to the electrode chamber, and the level of oxygen was recorded every 10 s for 5 min. The rate of respiration with oxygen was then normalized to OD₆₀₀.

The rate of anaerobic respiration using fumarate as an acceptor is calculated in SI Materials and Methods.

H₂O₂ Killing Assay.

Cells were grown anaerobically to OD₆₀₀ = 0.3 in LB, and 2.5 mM H₂O₂ was added. At different time points, samples were diluted and plated on LB agar. Colonies were counted after a day, and percent survival was calculated based on colony-forming units.

Measuring Zones of Inhibition.

All steps were performed in the anaerobic chamber with anoxic materials. Cells were grown in minimal glucose medium to OD₆₀₀ = 0.1; they were then mixed with 4 mL top agar (0.8%, 50 °C) and spread on plates of the same composition. Sterile disks (6 mm) were soaked in 15 μL 100 mM H₂O₂ and placed in the middle of the plate. The diameter of the inhibition zone was measured after 24 h.