Where in the world do bacteria experience oxidative stress?

Summary

Reactive oxygen species—superoxide, hydrogen peroxide, and hydroxyl radicals—have long been suspected of constraining bacterial growth in important microbial habitats, and indeed of shaping microbial communities. Over recent decades, studies of paradigmatic organisms such as E. coli, Salmonella typhimurium, Bacillus subtilis, and Saccharomyces cerevisiae have pinpointed the biomolecules that oxidants can damage and the strategies by which microbes minimize their injuries. What is lacking is a good sense of the circumstances under which oxidative stress actually occurs. In this MiniReview several potential natural sources of oxidative stress are considered: endogenous ROS formation, chemical oxidation of reduced species at oxic-anoxic interfaces, H₂O₂ production by lactic acid bacteria, the oxidative burst of phagocytes, and the redox-cycling of secreted small molecules. While all of these phenomena can be reproduced and verified in the lab, the actual quantification of stress in natural habitats remains lacking—and, therefore, we have a fundamental hole in our understanding of the role that oxidative stress actually plays in the biosphere.

Introduction

Enzymes that scavenge superoxide (O₂⁻) and hydrogen peroxide (H₂O₂) were discovered by serendipity. Their existence, in virtually all organisms, implies that these reactive oxygen species (ROS) comprise threats to fitness. In this Minireview I pose two key questions: What are the natural sources of these oxidants? In what circumstances do their levels rise high enough to impinge upon cell fate? I cannot provide definitive answers, but I summarize ideas that have arisen in the literature. The main point is that this issue is a key unknown in our understanding of oxidative stress.

To tolerate oxygen bacteria must cope with endogenous ROS formation

Mutants of E. coli that lack either superoxide dismutase or catalase and peroxidase exhibit distinctive growth defects, which led investigators to pinpoint the specific injuries that O₂⁻ and H₂O₂ can produce (Carlioz and Touati, 1986; Seaver and Imlay, 2001a) (Fig. 1). These species inappropriately oxidize and disable [4Fe-4S] dehydratases of the aconitase class (Kuo et al., 1987; Gardner and Fridovich, 1991; Flint et al., 1993; Jang and Imlay, 2007) as well as mononuclear Fe(II) enzymes such as ribulose-5-phosphase 3-epimerase (Sobota and Imlay, 2011; Gu and Imlay, 2013; Sobota et al., 2014). Both enzyme families use solvent-exposed iron atoms to directly bind metabolites; therefore, their metal centers are accessible to direct oxidation by either H₂O₂ or O₂⁻. The oxidized metals dissociate from the enzymes, rendering them inactive and disabling the pathways to which they belong. Enzymes of these two families are requisite for core cellular processes: the TCA cycle, the pentose-phosphate pathway, the biosynthesis of branched-chain and aromatic amino acids, and others.

Figure 1. Endogenous oxidative stress.

The adventitious transfer of electrons from redox enzymes to oxygen generates a mixture of O₂⁻ and H₂O₂. These species oxidize the solvent-exposed iron centers of mononuclear Fe²⁺ enzymes and [4Fe-4S] dehydratases, provoking iron dissociation and activity loss. The H₂O₂ also reacts with the pool of loose iron, most notably leading to DNA damage from hydroxyl-radical production.

The existence of mutant phenotypes implied that these oxidants are formed continuously in the aerobic cell. Indeed, measurements of H₂O₂ efflux from catalase/peroxidase mutants have fixed the H₂O₂ production rate inside air-saturated cells at 10–15 μM/sec; O₂⁻ rates are estimated to be 5–10 μM/sec (Seaver and Imlay, 2001b). In wild-type (scavenger-proficient) cells, the balance between production and scavenging sets the steady-state level of H₂O₂ at an estimated 20–50 nM, while O₂⁻ is predicted to be about 0.2 nM (Imlay, 2013). The reaction rates of H₂O₂ and O₂⁻ with the metal centers of dehydratases and mononuclear enzymes (ca. 10⁴ and 10⁶ s⁻¹ (Imlay, 2013)) therefore predict that the enzymes will be damaged every 30 min or so. The damaging events are reversible, however: clusters are reassembled with half-times of ca. 5 min (Gardner and Fridovich, 1992), and remetallation of mononuclear enzymes may be faster yet. This means that at any moment in wild-type cells, the bulk of these enzyme populations are in their active form, with a small minority being inactive because of a recent damaging event. On balance, their net activities are sufficiently high that their pathways are functional.

Inside cells H₂O₂ also reacts with the pool of loose iron to generate hydroxyl radicals (Keyer and Imlay, 1996). These powerful oxidants can damage most biomolecules, including DNA. The contribution of oxidants to DNA damage—and to mutagenesis—has been proposed to be significant: mutation rates are lower, and the viability of repair mutants is greater, when E. coli is cultured in the absence of oxygen (Morimyo, 1982; Boling et al., 1984; Imlay and Linn, 1986; Sakai et al., 2006). However, this conclusion may be premature. Lab media routinely contain micromolar H₂O₂ that is formed by chemical and photochemical processes, and under many experimental conditions the flow of this H₂O₂ into the cell greatly outstrips the pace of H₂O₂ formation within it (Li and Imlay, 2018). Therefore exogenous H₂O₂ can be the predominant source of the intracellular H₂O₂ and DNA oxidation. The upshot is that DNA-oxidation phenotypes in lab media can be artifactual. The actual impact of endogenous oxidants is currently unclear and in principle may be quite small.

When defensive systems are not enough: endogenous ROS contributes to the phenomenon of obligate anaerobiosis

E. coli—and presumably all oxygen-tolerant bacteria—have acquired a mixture of scavenging and repair systems that adequately sustain the activities of oxidant-sensitive enzymes in the face of endogenous ROS production. At one time textbooks asserted that obligate anaerobes cannot tolerate oxygen because they lack superoxide dismutases and catalases. This idea is incorrect. Many obligate anaerobes actually have these scavenging enzymes, and the others rely upon superoxide reductases and peroxidases (Sheng et al., 2014). It is not for lack of evolution that anaerobes cannot thrive around oxygen.

However, it may still be true that ROS contribute to oxygen poisoning in these bacteria. The carbohydrate fermenting Bacteroides thetaiotaomicorn is a classic oxygen-sensitive bacterium: when fully aerated, growth stops, and it resumes only when anoxia is restored. Analysis showed that both branches of central metabolism are poisoned, due to the inactivation of the metalloenzymes fumarase and pyruvate:ferredoxin oxidoreductase (Pan and Imlay, 2001). The inactivation of fumarase, an aconitase-class [4Fe-4S] dehydratase, is striking because the same enzyme remains active inside air-tolerant bacteria (Flint et al., 1993; Liochev and Fridovich, 1993). The key difference? Aerated B. thetaiotaomicron generates intracellular ROS at 15 times the rate of aerated E. coli (Mishra and Imlay, 2013; Lu and Imlay, 2017; Lu et al., 2018). The reason is not clear. Biochemical studies have shown that redox enzymes can accidentally transfer electrons to oxygen instead of to their proper substrates, and it is supposed that the collective autoxidations of such enzymes are responsible for the O₂⁻ and H₂O₂ that are formed endogenously in aerated cells (Massey et al., 1969; Messner and Imlay, 1999; Imlay, 2013). The rate of ROS formation in aerated Bacteroides might therefore be especially high because it possesses redox enzymes that are especially reactive with molecular oxygen. The low-potential redox chains (Eo’ ~ −.4 V) in such anaerobes are plausible sources of these ROS, since they are thermodynamically competent to drive electrons onto oxygen (Eo’ = −0.16 V). To resolve this issue completely, it will be necessary to pinpoint the specific enzymes that are the predominant sources of internal ROS—a goal that has not yet been achieved in any organism.

It is important to note that ROS are not exclusively responsible for the oxic problems of obligate anaerobes: molecular oxygen itself can directly poison key enzymes that are exclusive to anaerobes. Oxygen forms adducts to glycyl-radical enzymes, leading to polypeptide cleavage (Wagner et al., 1992), and it seems likely that molecular oxygen itself can over-oxidize the low-potential metal centers of enzymes such as pyruvate:ferredoxin oxidoreductase, hydrogenase, and nitrogenase (Vita et al., 2008; Stiebritz and Reiher, 2012; Schlesier et al., 2015). These enzymes are rapidly damaged by the introduction of oxygen in vitro and in vivo.

Oxidative stress from the environment

In 1982 Demple and Halbrook reported that when E. coli was pre-exposed to 50 μM H₂O₂ it could survive a subsequent exposure to 5 mM H₂O₂, whereas naïve cells could not (Demple and Halbrook, 1983). The phenomenon was subsequently linked to the induction of a regulon under the control of an H₂O₂-activated transcription factor, named OxyR (Christman et al., 1985; Aslund et al., 1999; Choi et al., 2001). Hydrogen peroxide rapidly oxidizes its sensory cysteine residue, and in this form OxyR triggers synthesis of defensive proteins. Catalase and NADH peroxidase (AhpCF) are induced 10-fold, thereby helping to drive down the intracellular level of H₂O₂ (Zheng et al., 2001). Dps is a mini-ferritin that suppresses damage to DNA by sequestering the cellular pool of loose iron (Altuvia et al., 1994; Ilari et al., 2002; Park et al., 2005). SufABCDES comprises a secondary iron-sulfur-cluster assembly system that counteracts damage to the dehydratase clusters (Lee et al., 2004; Jang and Imlay, 2010). MntH imports manganese(II), which supplants iron(II) in mononuclear enzymes, providing nearly as much activity without the tendency toward oxidation by peroxide (Kehres et al., 2002; Anjem and Imlay, 2012). YaaA and Fur help suppress loose-iron levels, and HemF and HemH sustain heme synthesis to enable catalase induction (Varghese et al., 2007; Liu et al., 2011b; Mancini and Imlay, 2015). The roles of other members of the regulon are not yet completely clear. Thus the induced activities are well-matched to the types of injuries that H₂O₂ creates. Dosimetric studies indicate that OxyR is activated when cytoplasmic H₂O₂ reaches about 0.1–0.2 μM—appropriately close to the concentrations (0.3–0.5 μM) that begin to inactivate enzymes and disrupt growth (Seaver and Imlay, 2001b; Zheng et al., 2001; Sobota and Imlay, 2011).

OxyR turns out to be wide-spread among bacteria. Interestingly, still other bacteria use PerR, an iron-binding transcription factor of the Fur family (Lee and Helmann, 2006b). This transcription factor is activated when H₂O₂ reacts with its Fe(II) cofactor, creating a hydroxyl radical that irreversibly oxidizes an adjacent histidine residue. PerR represses many of the same genes that OxyR controls in E. coli; when H₂O₂ oxidizes PerR, its DNA-binding activity is lost, and the genes that it controls are expressed. The rate constant with which H₂O₂ oxidizes PerR is similar to that of OxyR, suggesting that these disparate bacteria have evolved to respond to similar amounts of H₂O₂ (Lee and Helmann, 2006a). This arrangement would make sense if OxyR and PerR serve to defend similar classes of enzymes against oxidation.

Aeration alone does not activate these transcription factors, because the basal levels of scavenging enzymes inside cells are sufficient to keep endogenous H₂O₂ well below the OxyR and PerR trigger points. Instead, it seems likely that they respond to larger influxes of H₂O₂ from the local environment. The cytoplasmic membrane is semipermeable to H₂O₂, and external concentrations > 2 μM are sufficient to drive internal concentrations up to the 0.2 μM dose that activates the responses (Li and Imlay, 2018). The key question is: When do these bacteria encounter these doses of H₂O₂?

Photochemically generated H₂O₂ in surface waters

Somewhat surprisingly, H₂O₂ is readily detected in both fresh and oceanic waters (Lesser, 2006; Mesle et al., 2017). The levels can rise to the low-micromolar range—enough to plausibly threaten bacteria and induce their stress responses. The diurnal cycling of H₂O₂ levels tipped off investigators that ultraviolet photochemistry lie at the root of the phenomenon (Wilson et al., 2000b; Wilson et al., 2000a). Dissolved organic compounds in these waters include chromophores that, when excited by solar radiation, transfer electrons from local reductants to molecular oxygen. The events recapitulate the flavin-dependent production of O₂⁻ that is used to visualize superoxide dismutase in activity gels (Beauchamp and Fridovich, 1971)—and also the steady accrual of micromolar H₂O₂ that occurs in complex growth media under room lights (Li and Imlay, 2018). Workers are exploring the involvement of ROS in solar disinfection protocols. It remains unclear whether the ROS generated in natural waters comprises enough of a stress to add structure to the microbial community.

H₂O₂ at oxic-anoxic interfaces

One of the genes regulated by E. coli OxyR is a periplasmic cytochrome c peroxidase (Ccp) (Partridge et al., 2007). The enzyme receives electrons from the respiratory chain and transfers them directly to H₂O₂ in the periplasm. Apparently the purpose of Ccp is not to clear H₂O₂, since H₂O₂ equilibration across the outer membrane is so fast that the enzyme cannot lower the periplasmic level below the environmental level. Instead, Ccp allows the cell to employ H₂O₂ as a terminal oxidant to support respiration (Khademian and Imlay, 2017). This function would be beneficial only when molecular oxygen, a superior acceptor, is unavailable—and accordingly Ccp is synthesized only when molecular oxygen is absent (Partridge et al., 2007).

The peculiar aspect of this arrangement is its implication that H₂O₂ is present in some hypoxic environments. E. coli inhabits the margins of the large intestine, where it is ensconced in the mucoid periphery of the lumen (Fig. 2). Some oxygen diffuses into the intestine from the surrounding epithelial cells, but the levels drop by orders of magnitude just a few microns towards the interior (Espey, 2013). Deeper in the lumen oxygen is essential absent, and sulfate-reducing bacteria generate copious amounts of hydrogen sulfide (Macfarlane et al., 1992; Pitcher et al., 2000). One possibility is that as this sulfide diffuses towards the microoxic margins, it reacts chemically with the oxygen, producing H₂O₂. A similar situation might occur at oxic-anoxic interfaces in soil and at hydrothermal vents (Ogino et al., 2018), where low- and high-potential fluids run into one another. The apparent K_M of Ccp is 5 μM, which is a good match to the extracellular dose that activates OxyR (Khademian and Imlay, 2017; Li and Imlay, 2018); this agreement lends further weight to the idea that low-micromolar concentrations of H₂O₂ are biologically relevant.

Figure 2. Plausible sources of oxidative stress at the oxic-anoxic interface near the intestinal epithelium.

Oxygen influx from the epithelium collides with sulfide generated by luminal bacteria, potentially generating H₂O₂ through direct reaction. Lactic acid bacteria near the epithelium are also likely to excrete H₂O₂ as a direct metabolic product, threatening by-stander bacteria.

Alternative possibilities are that intestinal H₂O₂ is generated by lactic acid bacteria, which can also thrive at such interfaces, or that H₂O₂ is released by phagocytic cells along the epithelial layer of the intestine. These sources are considered in more detail below. The most striking take-home message is that the intuitive idea, that H₂O₂ is most abundant in the most oxic environments, may be wrong.

Lactic acid bacteria spew H₂O₂ into their environment

Lactic acid bacteria (LAB) are primarily carbohydrate fermenters who have made metabolic compromises that enable them to thrive in low-iron habitats. They lack full respiratory chains, but many can nevertheless exploit oxygen as a direct electron acceptor by using soluble lactate and pyruvate oxidases (Spellerberg et al., 1996; Seki et al., 2004). When LAB do so, they produce additional ATP through acetate kinase (Pericone et al., 2003). The product of these oxidases is H₂O₂. In laboratory cultures of LAB, the H₂O₂ concentration in the cells and the medium can rise to the millimolar range (Tong et al., 2007; Liu et al., 2011a). The oxidases are expresses in human plaque samples (Zhu et al., 2014), for example, raising the prospect that the excreted H₂O₂ might shape the oral microbial community.

How do LAB avoid poisoning themselves? They lack [4Fe-4S] dehydratases; aconitase, for example, is absent by virtue of the lack of a TCA cycle. LAB famously accumulate millimolar concentrations of intracellular manganese (Archibald, 1986; Daly et al., 2004), and it is likely that they metallate their mononuclear enzymes with Mn rather than Fe. Thus they constitutively employ the same defensive tactic that E. coli activates only under H₂O₂ duress (Anjem et al., 2009; Anjem and Imlay, 2012). A Dps homolog keeps free-iron levels low enough that DNA damage is suppressed (Xu et al., 2014). Collectively these devices avoid damage to iron-sulfur dehydratases, mononuclear enzymes, and DNA, so that lactic acid bacteria are largely immune to the H₂O₂ that they generate.

It has long been suspected that the H₂O₂ that LAB excrete may suppress the growth of competing bacteria. The effect has been documented in mixed lab cultures (Pericone et al., 2000; Tong et al., 2007; Bao et al., 2017). However, these experiments represent richly fed bacteria in closed environments; in natural habitats H2O2 may be convected away and or degraded by scavenging-proficient cells. Much less H₂O₂ may accumulate (Margolis, 2009). The accurate measurement of H₂O₂ in natural LAB environments will be fundamental to understanding its impact upon flora.

A final interesting contrast between LAB and E. coli can be considered. E. coli has only one high-flux enzyme that generates H₂O₂ stoichiometrically: a monoamine oxidase that catalyzes the first step in phenylethylamine catabolism (Kumar and Imlay, 2013). The key distinction is that, unlike the oxidases in LAB, the E. coli oxidase is localized in the periplasm. Consequently the H₂O₂ that it produces rapidly leaves the cell through the outer membrane porins and thereby avoids damaging the cytoplasmic iron enzymes.

ROS as an antimicrobial weapon wielded by phagocytes

Mammalian phagocytes and plants respond to bacteria by engulfing them and then dousing them with superoxide that the generate with a dedicated NADPH oxidase (Bedard et al., 2007; Kawahara et al., 2007) (Fig. 3). Its role in suppressing microbial growth is clear, as humans and mice that lack the enzyme are vulnerable to infection (Shiloh et al., 1999; Thomas, 2017). However, it has been surprisingly hard to figure out exactly how phagocytic ROS production suppresses microbial growth (Slauch, 2011).

Figure 3. Oxidant formation in phagosomes.

The O₂⁻ produced by the host NADPH oxidase cannot penetrate the cytoplasmic membranes of captive bacteria; either it, or its more-reactive protonated form HO₂^., are believed to injure the extracytoplasmic surface of bacteria. Dismutation produces H₂O₂ that can penetrate membranes. Tentative calculations suggest that O₂⁻, HO₂^., and H₂O₂ levels may be in the ranges of 10–50 μM, 0.1–4 μM, and 1–4 μM, respectively, depending upon phagosomal pH (Imlay, 2009). Modeling of fluxes in neutrophils predicts similar levels (Winterbourn et al., 2006).

The oxidase directly generates superoxide. Because superoxide is anionic (O₂⁻) at neutral pH, it cannot cross bacterial membranes (Lynch and Fridovich, 1978; Korshunov and Imlay, 2002). Phagosomal acidification can partially protonate the superoxide to neutral HO₂^., which might in principle penetrate into the cytoplasm, but genetic studies generally support the view that superoxide acts upon an as-yet unidentified target either on the cell surface or in the periplasm (Craig and Slauch, 2009). Metalloenzymes of the aconitase or Rpe classes are not present in those regions, raising the possibility that HO₂^.—a much more reactive oxidant than O₂⁻—might attack a novel target. This idea has proven difficult to pursue, since it is hard to recreate the predicted micromolar levels of HO₂^. (Imlay, 2009) in experimental systems.

Superoxide spontaneously dismutates to H₂O₂, and one might wonder whether H₂O₂ formed in the phagosome exerts toxic effects upon bacteria. However, modeling suggests that H₂O₂ efflux from the macrophage phagosome is so rapid that, despite the impressive pace of its formation, the steady-state concentration of H₂O₂2 H will rise no higher than a few micromolar (Imlay, 2009). A similar result has been predicted for neutrophil phagosomes (Winterbourn et al., 2006). As was discussed, bacteria can cope with this level of H₂O₂, as long as they can activate their OxyR response (Li and Imlay, 2018).

This analysis considers bacteria trapped inside isolated macrophages; the extruded H₂O₂ is treated as disappearing from the system. However, it is conceivable that H₂O₂ might accumulate on the macroscopic level in inflamed tissue, potentially rising to higher concentrations. An abscess is a plausible example. Actual measurements of H₂O₂ will be needed to resolve this point. The surprising bottom line is that we still do not know the mechanism by which phagocytic O₂⁻ and/or H₂O₂ suppress microbial growth, nor whether their presence affects by-stander bacteria.

Redox-cycling antibiotics impose intracellular O₂⁻ stress

In their initial investigations into the function of SOD, Hassan and Fridovich used redox-cycling compounds to elevate the rate of O₂⁻ production inside E. coli (Hassan and Fridovich, 1979). The synthetic viologen paraquat is still commonly used for this purpose: it penetrates into bacteria, oxidizes their redox enzymes, and transfers the electrons to oxygen. The resultant O₂⁻ flux is enough to disable ROS-sensitive enzymes and block growth even of wild-type cells (Kuo et al., 1987).

We now realize that such experiments recapitulate natural phenomena. Although paraquat is a synthetic compound, both plants and bacteria synthesize and secrete phenazines and quinones that do the same thing (Turner and Messenger, 1986; Paiva et al., 2003; Inbaraj and Chignell, 2004). Their significance is reflected in the fact that enteric bacteria manifest a response regulon, governed by SoxRS (Greenberg et al., 1990; Tsaneva and Weiss, 1990), that has evolved to defend the cell against such compounds. Although this response was initially believed to be triggered by O₂⁻ itself, more recent data indicate that the system is activated when the redox-cycling compounds directly oxidize the [2Fe-2S] cluster of the SoxR regulator (Gu and Imlay, 2011). The activated protein then induces a regulon (Pomposiello et al., 2001) that includes drug-efflux pumps, proteins that diminish cell permeability, and defensive enzymes including SOD (Aiba et al., 1987; Ma et al., 1995; Lee et al., 2009). Interestingly, the system appears to have been laterally inherited and modified from SoxR regulons in the bacteria that produce these compounds, where SoxR controls their secretion and does not involve defensive enzymes (Dietrich et al., 2008).

Thus the new view is that such compounds are natural sources of ROS stress. Their benefit to the producing organisms, however, is a matter of ongoing discussion. On the one hand, they can suppress the growth of competitors. A persuasive example is that walnut trees provide bare ground for their nuts by poisoning undergrowth with the juglone in their fallen leaves. However, workers have proposed alternative functions. The secreted compounds can rescue iron-poor bacteria by reducing and solubilizing iron in mineral deposits (Wang et al., 2011). They can also carry electrons from bacteria embedded deep in anoxic biofilms to outer, oxic regions, thereby enabling respiration at a distance (Glasser et al., 2017). In these views, the toxicity that redox-cycling compounds impose upon by-stander bacteria such as E. coli is largely an accident—albeit one against which the bacterium must defend itself.

Do heat, metals, solvents, and clinical antibiotics create oxidative stress?

In contemporary toxicology it seems that oxidative stress is easy to suspect but hard to prove. Microbiologists have proposed that ROS underlie the toxic effects of many stresses that have no obvious connection to oxidants—from toxin/antitoxin systems (Kolodkin-Gal et al., 2008)and nanoparticle surfaces (Applerot et al., 2012), to high salinity (Mishra et al., 2009) and hydrostatic pressure (Aertsen et al., 2005). The list is so long as to invite caution (Imlay, 2015). The problem is two-fold: at the experimental level, ROS themselves are evanescent and difficult to detect directly; at the conceptual level, it might seem plausible that any stress that physically impairs redox enzymes could cause them to inadvertently transfer electrons to oxygen. How can ROS suspicions be tested?

A popular approach is the use of redox-sensitive dyes. By 2007 skeptical reviewers were able to cite > 2000 papers in which such dyes had been used to diagnose oxidative stress inside living cells (Wardman, 2007). The number is surely far higher now. Various dyes have been shown to be oxidized by hydroxyl radicals—or O₂⁻, or H₂O₂—in vitro, and the hope is that their oxidation in vivo can be regarded as evidence of the same oxidant (Kalyanaraman et al., 2012). However, problems have arisen. When coupled to cell-sorting, these analyses have sometimes been confounded by changes in bacterial morphology (Renggli et al., 2013; Paulander et al., 2014). Most of the experiments do not carefully quantify dye loading in stressed cells versus their unstressed controls; since stress may alter membrane packing, or influx and efflux energetics, he amount of dye inside sample cells may differ from that in control cells. It is worth considering that a ratiometric sensor may be more appropriate (Bilan et al., 2013; Lin et al., 2013). And rarely have control experiments rigorously established that the dyes are not vulnerable to the oxidation by high-valent metal centers, or countervailing reduction by cellular thiols (Imlay, 2015). These concerns are not intended to refute claims but to invite care when considering the evidence.

Other common tests of whether a phenomenon is driven by oxidants include the use of “antioxidants”—usually thiols or ascorbate—in suppressing it, or measurements of TBARS as evidence of lipid peroxidation. Thiols may be effective scavengers of extracellular H₂O₂, but they are not much more reactive with intracellular hydroxyl radicals than is everything else inside the cell, and they barely react with superoxide at all (Winterbourn and Metodiewa, 1999). Tests for TBARS (thiobarbituric-acid reactive species) are easy to perform but are not specific to peroxidized lipids; the latter issue is especially pertinent to bacteria, as they have no polyunsaturated fatty acids, which are requisite according to standard models of lipid peroxidation (Bielski et al., 1983).

So what are the most reliable markers of oxidative stress? Induction of the OxyR regulon is a good one, as it trades upon the cell’s own detection system for H₂O₂. Damage to [4Fe-4S] dehydratases and mononuclear enzymes are known effects. And to approach the problem head-on, the rate of H₂O₂ production can be directly determined with catalase/peroxidase mutants, if they are available. In the view of the author, these techniques provide the approaches that are currently the most specific and the least prone to artifacts.

Concluding remarks

Over the past forty years, much has been learned about the mechanisms by which O₂⁻ and H₂O₂ can poison bacteria and about the tactics by which bacteria strive to defend themselves. The point of this MiniReview is that we know much less about the natural circumstances in which oxidative stress occurs. In real environments H₂O₂ stress is likely to be low-micromolar in dose but protracted in time, and laboratory experiments should seek to replicate this regime to obtain the most pertinent outcomes. Meanwhile, we will need to determine H₂O₂ levels in natural samples if we are to know whether oxidative stress is a real occurrence there. And multiple, direct assays are needed to resolve uncertainties as to whether non-oxidative stresses trigger ROS formation.

Originality-significance statement.

The mechanisms by which oxidants can damage cells are increasingly understood, but the natural circumstances under which they do so remain poorly defined. This manuscript reviews what is known about the level of threat imposed by oxidants that are generated by endogenous, chemical, and biological-warfare processes. The point is to emphasize a key gap in our understanding.