Antioxidants are ineffective at quenching reactive oxygen species inside bacteria and should not be used to diagnose oxidative stress

Summary.

A wide variety of stresses have been proposed to exert killing effects upon bacteria by stimulating the intracellular formation of reactive oxygen species (ROS). A key part of the supporting evidence has often been the ability of antioxidant compounds to protect the cells. In this study some of the most-used antioxidants—thiourea, glutathione, N-acetylcysteine, and ascorbate—have been examined. Their ability to quench superoxide and hydrogen peroxide was verified in vitro, but the rate constants were orders of magnitude too slow for them to have an impact upon superoxide and peroxide concentrations in vivo, where these species are already scavenged by highly active enzymes. Indeed, the antioxidants were unable to protect the growth and ROS-sensitive enzymes of E. coli strains experiencing authentic oxidative stress. Similar logic posits that antioxidants cannot substantially quench hydroxyl radicals inside cells, which contain abundant biomolecules that react with them at diffusion-limited rates. Indeed, antioxidants were able to protect cells from DNA damage only if they were applied at concentrations that slow metabolism and growth. This protective effect was apparent even under anoxic conditions, when ROS could not possibly be involved, and it was replicated when growth was similarly slowed by other means. Experimenters should discard the use of antioxidants as a way of detecting intracellular oxidative stress and should revisit conclusions that had been based upon such experiments. The notable exception is that these compounds can effectively degrade hydrogen peroxide from environmental sources before it enters cells.

Introduction.

The production of intracellular reactive oxygen species is an unavoidable feature of life in oxic environments. Superoxide (O₂⁻) and hydrogen peroxide (H₂O₂) are formed when molecular oxygen adventitiously collides with the reduced flavins and metal centers of redox enzymes [1]. These oxygen species have the potential to inactivate enzymes that use solvent-exposed [4Fe-4S] clusters or Fe(II) cofactors to catalyze non-redox reactions. The O₂⁻ and H₂O₂ oxidize the iron moieties, leading to loss of the catalytic iron atom and, if the process is unchecked, failure of the pathways to which such enzymes belong [2–5]. Such injuries are disabling but non-lethal, as the cell can repair or rebuild the enzymes once the stress subsides.

A more irreversible event involves the oxidation of DNA. The cell maintains a pool of loose iron in order to metallate new apoproteins as they are made. Ferrous iron adheres both to anions and to aromatic systems, so this pool binds to the surfaces of lipids, proteins, and nucleic acids. A reaction between this iron and H₂O₂ forms hydroxyl radicals through the Fenton reaction [6]:

$$\text{Fe}\left(\text{II}\right)+{\text{H}}_2{\text{O}}_2\to \text{Fe}\left(\text{III}\right)+{\text{OH}}^{-}+\text{HO}$$

Hydroxyl radicals are extremely reactive, and so they tend to damage the biomolecules that bind iron. In the case of DNA, the resultant damage is potentially mutagenic and or lethal.

Organisms have evolved several layers of defense to avoid such injuries. Superoxide dismutases, peroxidases, and catalases minimize the accumulation of O₂⁻ and H₂O₂, and their effectiveness derives from the fact that they are among the most abundant and efficient enzymes inside cells. Although the rates of O₂⁻ and H₂O₂ formation are approximately 5 and 10 μM/sec inside aerobic E. coli, the steady-state concentrations of these species are estimated to be only 0.2 and 50 nM, respectively [1]. Cells go further to protect their DNA, carefully minimizing the amount of loose iron and employing multiple pathways to repair any lesions that might arise. Accordingly, committed aerobes and facultative organisms are able to thrive in air-saturated environments.

However, these defenses can be overwhelmed. Both plants and bacteria secrete redox-cycling antibiotics in order to poison competitors; these compounds greatly accelerate the formation of O₂⁻ and H₂O₂ inside target cells and, despite the presence of scavenging enzymes, can elevate ROS concentrations to toxic levels [7, 8]. Similarly, bacteria sometimes encounter environmental H₂O₂ that is produced by lactic acid bacteria, photochemistry, or cell-based immune responses [9]. This H₂O₂ can diffuse across membranes [10] and similarly override cell defenses. E. coli uses the SoxR and OxyR transcription factors to detect redox drugs and H₂O₂, respectively, and to initiate further defensive tactics [11–13]. These include the further induction of scavenging systems and the activation of enzyme and DNA repair systems. Collectively, these tactics seem to position bacteria to withstand most oxidative threats that they might encounter in natural habitats.

Nevertheless, biologists have often wondered whether various stresses might disrupt metabolism in a way that stimulates intracellular ROS formation, even to the point of killing cells. This idea has been proposed to explain the toxicity of a large number of conditions, ranging from thermal shocks to antibiotic treatments (e.g., [14, 15] [16] [17] [18, 19] [20] [21] [22, 23] [24]. In many of these cases, a key piece of evidence has been the observation that chemical antioxidants diminished the rate of cell death. These antioxidants are organic compounds that can directly reduce O₂⁻, H₂O₂, or hydroxyl radicals and thereby protect important biomolecules. Thiourea, N-acetylcysteine, glutathione, and ascorbic acid are the antioxidants that have been most commonly used in experiments; each of them exhibits detectable ROS-degrading activity in vitro.

To relieve oxidative stress inside a cell, antioxidants must provide scavenging activity that is at least comparable to that of the native scavenging enzymes. Only in that case could they substantially diminish the lifetime—and thus the steady-state concentration—of the ROS. In this study we have quantified the scavenging action of these compounds, and we find that it is much too low to make an impact in this way. Using strains that manifest injuries by each of the three ROS, we confirmed that these antioxidants were ineffectual in vivo. Instead, these compounds have the ability to slow cell growth—and often it is this effect that correlates most closely with their ability to rescue cells from stress.

Results.

The antioxidant actions of small molecules are regarded as arising from their abilities to scavenge reactive oxygen species. In this study we focused upon superoxide (O₂⁻), hydrogen peroxide (H₂O₂), and hydroxyl radical (HO), which are species that are formed in all aerobic organisms and that have been suspected of damaging cells under a range of stress conditions. Other oxidants of biological consequence include molecular oxygen, singlet oxygen, hypohalous acids, and peroxynitrite; these species are formed in more specialized conditions and were not considered here.

Our approach was to quantify the rates at which antioxidant compounds react with key oxidants in vitro, in order to predict the impact that they would have upon oxidant concentrations inside cells. We then tested their ability to suppress damage in E. coli strains suffering from excessive intracellular concentrations of prescribed oxidants. We chose a range of antioxidants that have been widely used by experimenters: thiourea, glutathione, N-acetylcysteine, ascorbic acid, Tiron, and endaravone.

Superoxide stress is not alleviated by these compounds.

The rate of superoxide scavenging by test compounds was tested as indicated in Figure 1A,B. Xanthine oxidase was used to generate O₂⁻ in vitro, and we monitored the ability of the O₂⁻ to react with cytochrome c. This is the classical set-up employed in assays of superoxide dismutase activity [25]. Because the concentration of cytochrome c was known, as is the rate constant with which it reacts with O₂⁻, one can use inhibition data to calculate the rate constant with which competitors degrade O₂⁻ (Materials & Methods).

Figure 1. Antioxidants are ineffective scavengers of intracellular superoxide.

(A) The efficiency of superoxide scavenging was quantified by the ability of the compound to competitively inhibit reduction of cytochrome c. (B) An example reaction is shown. Rate constants are listed in Table 1. (C) Thiourea (TU, 120 mM) does not suppress the aerobic growth defect of a superoxide dismutase-deficient (SOD⁻) strain. Exponentially growing cells (MG1655, KCI415) were diluted into aerobic minimal medium at 3 hrs. Superoxide stress blocks growth by inactivating enzymes needed to synthesize amino acids. Similar results were obtained when the medium contained ascorbate, glutathione, N-acetylcysteine, edaravone, or Tiron (Fig. S2). (D) Thiourea (150 mM) does not protect the activity of threonine dehydrogenase (TDH), a Fe(II) enzyme that is sensitive to O₂⁻. The enzyme was assayed after strains (MG1655, KCI415) were aerated for 45 min.

We observed that thiourea, glutathione, and endaravone each slowed cytochrome c reduction, whereas N-acetylcysteine did not. Control experiments tested whether any of the inhibition might be due to suppression of xanthine oxidase turnover and whether the residual cytochrome c reduction was sensitive to enzymic SOD. Thiourea indeed did slow xanthine oxidase, which required consideration when we calculated the rate constants.

Because ascorbate directly reduces cytochrome c, its ability to scavenge O₂⁻ was tested by its ability to block the reaction between O₂⁻ and nitroblue tetrazolium [26]. The oxidation of ascorbate radical by O₂⁻ produces an ascorbate radical that directly reduces NBT; therefore, we measured the concentration of SOD that was required to block the reaction of O₂⁻ with ascorbate and used this method to calculate a rate constant (Materials & Methods).

Table 1 presents the rate constants. The data verify that the majority of these compounds have some capacity to scavenge O₂⁻. However, the key question is whether they are sufficiently reactive that they will substantially reduce the O₂⁻ concentration inside cells that already possess substantial titers of SOD. The Table presents the amount of each compound that would provide O₂⁻ scavenging activity equivalent to the endogenous SOD; i.e., this amount would be necessary to reduce the steady-state O₂⁻ level in a wild-type cell by two-fold. The data show that grossly unrealistic concentrations would be needed.

Table 1.

Small-molecule antioxidants are poor intracellular scavengers of ROS.

	k (M−1 s−1)	k (M−1 s−1)	Conc’n in	Intracellular	Needed conc’ns (M)
Compound	with O 2 − a	with H 2 O 2 b	medium (M) c	conc’n (M) d	O 2 − e	H 2 O 2 f
TU	91 +/− 4	0.25 +/− 0.02	0.15	0.15	99.	5.6
GSH	820 +/− 50	0.20 +/− 0.05	0.010	0.0028	11.	7.0
NAC	ND	0.085 +/− 0.01	0.010	0.0032	--	16.
Asc	5000 +/− 40	0.44 +/− 0.02	0.005	--	4.5	3.8

^{a, b}Rate constants for reactions of thiourea (TU), reduced glutathione (GSH), N-acetylcysteine (NAC), and ascorbate (Asc) with O₂⁻ and H₂O₂. ND: below detection. Values approximate those available in the literature [84] [85].

^cConcentration of antioxidant provided in the medium.

^dResultant concentration in the cell. NAC value represents total low-molecular-weight thiols, including GSH.

^e,fThe hypothetical intracellular concentration of each antioxidant that would be needed to provide scavenging activity equal to that of the 1200 +/− 100 U/ml SOD or 750 +/− 76 U/ml catalase already inside E. coli.

The issue can also be considered from the perspective of the amounts of antioxidants that can enter cells. For example, thiourea is commonly employed as an antioxidant at concentrations up to 150 mM, and it rapidly equilibrates across biological membranes so that this concentration would also obtain inside the cell within seconds (Materials & Methods [27]). From the Table one can calculate that the scavenging action of 150 mM thiourea would reduce the intracellular O₂⁻ lifetime—and thus its steady-state concentration—by only 0.15%.

The other compounds are charged and do not easily enter cells. We determined that the glutathione concentration inside E. coli was approximately 2.5 mM in rich medium (and somewhat lower in minimal medium). Supplementation of the medium with 10 mM GSH or 10 mM N-acetylcysteine had very little, if any, effect on the intracellular concentration of low-molecular-weight thiols, which would include GSH, N-acetylcysteine, and cysteine (Fig. S1). Our measurements suggested that the thiol level might rise to 3 mM; at that concentration, these compounds would reduce O₂⁻ concentration by 0.03%.

Similar arguments can be made for ascorbate and endaravone. In sum, these in vitro experiments indicate that none of these compounds could plausibly reduce O₂⁻ levels at a rate that would have a physiological impact.

The antioxidant effect was then tested using E. coli strains that exhibit defects from O₂⁻ stress. Figure 1C depicts the growth of a sodA sodB mutant, which lacks all cytoplasmic SOD activity, upon shift from anaerobic to aerobic conditions. Upon aeration endogenous O₂⁻ rises to a level that inactivates [4Fe-4S] dehydratases, including isopropylmalate isomerase and dihydroxyacid dehydratase in the pathway of branched-chain biosynthesis [2]. Accordingly, growth stops in the minimal medium that was used here. The additions of thiourea (Fig. 1C) and of ascorbate, glutathione, N-acetylcysteine, and endaravone (Fig. S2) failed to improve growth. We have previously shown that as little as 10% of normal SOD activity is sufficient to enable growth in this medium [28], so it is clear that these antioxidants failed to provide even that much scavenging activity.

Mononuclear Fe²⁺ enzymes comprise the second class of enzymes that O₂⁻ and H₂O₂ disable [5, 29]. A member of this family, threonine dehydrogenase [30], exhibited reduced activity in the sodA sodB strain, and activity did not increase in the presence of antioxidants (Figure 1D).

Redox-cycling antibiotics are important natural sources of O₂⁻ with which E. coli must contend [7, 8]. Figure 2 shows that strains lacking sodA (but synthesizing SodB) are more sensitive to paraquat than are wild-type cells that contain both enzymes—indicating that under this condition O₂⁻ is the key toxic species. Again, the addition of antioxidants failed to relieve the growth defect. Therefore, the growth and biochemical data confirm the implications of the rate constants: These compounds cannot contribute enough scavenging activity to have any significant impact upon intracellular O₂⁻ stress.

Figure 2. Thiourea does not protect cells from the superoxide that is generated by a redox-cycling antibiotic.

Cells growing in aerobic glucose medium were treated with paraquat (PQ) starting at t=0. All cells (MG1655, KCI746) were SodB⁺; where indicated, strains lacked the inducible SodA superoxide dismutase, in order to establish that the growth defect arose from superoxide. (A) Thiourea (TU, 150 mM) was supplemented as indicated. (B) Parallel sodA cultures were supplemented with 10 mM glutathione (GSH), 10 mM N-acetylcysteine (NAC), 0.1 mM edaravone (Eda), 10 mM Tiron, or 5 mM ascorbate (Asc).

These antioxidant compounds can scavenge extracellular H₂O₂, but they are ineffective inside cells.

Rate constants were determined for the scavenging of H₂O₂ (Figure 3A, Table 1). All the tested compounds were able to degrade H₂O₂; however, the rate constants indicated once again that impossibly high concentrations of these reagents would be required for them to degrade intracellular H₂O₂ as quickly as native catalase. In fact, their actual effect inside E. coli would be even smaller, given that most scavenging of H₂O₂ is conducted by the NADH peroxidase AhpCF rather than catalase [10, 31].

Figure 3. Antioxidants are ineffective scavengers of intracellular H₂O₂.

The efficiency of H₂O₂ scavenging was determined by tracking the disappearance of H₂O₂. (A) Example reaction, containing 1 mM H₂O₂ and 22 mM thiourea, at 37^o C. Rate constants are shown in Table 1. (B) Thiourea (TU) failed to protect threonine dehydrogenase inside catalase/peroxidase-deficient mutants (Cat/Prx⁻, MK186). Intracellular H₂O₂ is approximately 1 μM.

The impact of scavengers upon cellular damage was tested in catalase/peroxidase-deficient cells, which lack both AhpCF and catalase. The endogenous H₂O₂ formed in these cells (~ 1 μM steady-state) is sufficient to damage threonine dehydrogenase. The addition of thiourea was unable to protect the enzyme (Figure 3B).

However, antioxidants can protect organisms by scavenging environmental H₂O₂ before it enters the cell. As shown in Figure 4A, an oxyR mutant cannot form isolated colonies on standard aerobic plates, which contain micromolar levels of H₂O₂ that are formed abiotically by glucose oxidation and/or photochemistry [32, 33]. The addition of thiourea rescued the cells. Catalase and pyruvate, a known chemical scavenger of H₂O₂, did the same. A similar effect was observed in zone-of-inhibition experiments, in which H₂O₂ diffuses from a disk towards bacteria (Figure 4B). The half-life of H₂O₂ in 150 mM thiourea is 20 s; in 10 mM GSH, it is 6 min. In these scenarios the extracellular scavenger does not compete with enzymes, and the substantial time required for H₂O₂ to encounter and enter the bacterium provides ample opportunity for even inefficient scavengers to have an impact.

Figure 4. Antioxidants can shield cells from environmental H₂O₂.

(A) The oxyR mutant (AL343) cannot form isolated colonies on aerobic LB medium, as it contains low-micromolar H₂O₂. Growth is restored by the addition of catalase, pyruvate, or thiourea. (B) Catalase and thiourea protect wild-type cells (MG1655) from exogenous H₂O₂ that diffused from a disk. (C) Extracellular antioxidants partially reduce the H₂O₂ stress experienced by catalase/peroxidase-deficient mutants (Cat/Prx⁻). (Left) As shown in the diagram, endogenous H₂O₂ equilibrates across the membrane of non-scavenging cells, and external scavengers block its re-entry. (Right) Cytoplasmic stress was detected by the expression of a katG-lacZ fusion whose expression responds to the H₂O₂-activated OxyR transcription factor (AL441, MK186). TU, thiourea; GSH, reduced glutathione.

Interestingly, the ability of scavengers to work extracellularly but not intracellularly can be captured using catalase/peroxidase-deficient cells. Their endogenous H₂O₂ equilibrates across the membrane, and extracellular scavengers act as a sink that partially diminishes the intracellular concentration by blocking reentry of H₂O₂ (Figure 4C). This effect was observed by tracking the expression of the OxyR regulon. Extracellular catalase, for example, diminished expression of an OxyR-inducible fusion (katG-lacZ) by approximately 50%. Thiourea and glutathione did the same. Neither scavenger could further eliminate the stress because H₂O₂ can substantially activate OxyR before leaving the cell.

Importantly, this capacity for extracellular scavenging cannot protect a wild-type cell from endogenous H₂O₂. Approximately 90% of internally formed H₂O₂ molecules are scavenged by catalase and/or NADH peroxidase (AhpCF) before they ever leave the cell [10]—and perhaps 99% if catalase and AhpCF have been induced—so the re-entry of H₂O₂ is not a significant component of the stress that wild-type cells experience. Therefore, the stimulation of stressed-cell growth or survival by extracellular catalase—or any other antioxidant—should be regarded as evidence that ROS are formed extracellularly, not internally.

The upshot is that antioxidant compounds do not exert significant scavenging activity inside a cell, whose scavenging activities are orders of magnitude more effective. However, antioxidants can clear external medium of H₂O₂ and therefore can shield cells from exogenous H₂O₂ stress.

Antioxidants are ineffective at scavenging intracellular hydroxyl radicals.

Hydroxyl radicals receive particular attention in studies of stresses that kill cells. This species oxidizes virtually all organic compounds. Its second-order rate constant can exceed 10¹⁰ M⁻¹ s⁻¹ for reaction with sulfur compounds such as cysteine; this rate is diffusion-limited, meaning that every collision complex enables a reaction. Other amino acids (His, Trp, Tyr) react as quickly, and the remainder are not far behind [34]. Consequently, intracellular hydroxyl radicals generally react within a few molecular diameters of the site at which they are formed, and extremely high concentrations of a competitor would be required in order to intercept the hydroxyl radical before an adjacent molecule is damaged.

Formally, from the abundance and rate constants of amino acids inside the cell (including in proteins), one can calculate that 350 mM of a diffusion-limited scavenger would be needed just to equal the hydroxyl-scavenging activity supplied by proteinaceous amino acids. Glutathione is one diffusion-limited scavenger (10¹⁰ M⁻¹ s⁻¹), and indeed studies with mutants confirmed that the millimolar glutathione inside cells does not exert any protective effect against the hydroxyl radicals that are generated by ionizing radiation [35].

Models of cell killing by oxidative stress focus upon DNA damage. All four bases, plus the deoxyribose moiety, react with hydroxyl radicals with rate constants close to 10¹⁰ M⁻¹ s⁻¹ [34]. Consequently, hydroxyl radicals formed upon the surface of nucleic acids are expected to react on site. Hydroxyl radicals that enter the bulk phase can return to damage nucleic acids only in dilute solutions that lack other targets, such as the amino acids referenced above.

When dilute DNA was exposed in vitro to a Fenton reaction, damage was substantially diminished by the presence of thiourea (Figure 5A). In such systems much of the Fenton chemistry occurs at a distance from the DNA, and diffusion of hydroxyl radicals to the target takes time, so scavenging is effective. However, the situation differs when bacteriophage are exposed to a Fenton system. Their inactivation is due to phage DNA damage [36]. In this system hydroxyl radicals that are produced in the bulk solution react with capsid protein before they can reach the DNA; only hydroxyl radicals produced on-site by DNA-bound iron can damage DNA [6] [37]. In this situation antioxidants were ineffective (Figure 5B–C, Figure S4). This situation is expected to mimic the circumstance and outcome inside living cells.

Figure 5. Antioxidants cannot shield phage from Fenton-generated oxidants.

(A) Thiourea (TU) diminished DNA nicking from a Fenton reaction in vitro. Plasmid DNA (400 ng) in HEPES buffer was exposed to 8 μM Fe(II), and 0.5 mM H₂O₂ was added for a single-turnover reaction. The reaction of lane 3 included 150 mM thiourea. (B) Bacteriophage P1 in buffer was exposed to the indicated concentration of Fe(II) and then 4 mM H₂O₂ for a single-turnover reaction. The H₂O₂ was then quenched with catalase, and P1 survival was determined by infection into an E. coli indicator strain. Where indicated, 100 mM thiourea was included in the reaction mix. (C) The reaction was conducted as in part B using 5 μM Fe(II) and 1 mM H₂O₂; where indicated, 10 mM N-acetylcysteine (NAC), 10 mM glutathione (GSH), or 100 mM thiourea (TU) was included.

The protective effect of antioxidants can arise from inhibitory effects upon growth.

To look further, recA mutants of E. coli were briefly exposed to H₂O₂. The strain lacks recombinational DNA repair, and so DNA damage is lethal [36, 38]. When workers have reported the use of thiourea as an antioxidant, it has typically been used at concentrations of 100–150 mM and has reduced death rates only partially. Our results were similar: Thiourea modestly increased survival, but it did so only at high concentrations (Figure 6A). The effect was not proportionate to thiourea dose, suggesting that it did not arise from hydroxyl scavenging. In fact, careful measurements showed that protection occurred concomitant with a slowing of growth. This inhibitory effect of thiourea has been noted before [39]. Most agents that kill bacteria are less effective when cell growth slows, and so we examined the possibility that growth inhibition, rather than oxidant scavenging, might be the route of protection.

Figure 6. Thiourea protects cells from oxidative DNA damage only at doses that slow growth.

(A) An exponentially growing recA mutant (LEM17) was exposed to 4.5 mM H₂O₂ for 30 s before the H₂O₂ was removed by catalase. Cell survival (solid bars) was then determined by plating. Thiourea at indicated doses was included in the medium before and during H₂O₂ exposure. Thiourea was observed to slow cell growth at the concentrations that protected cells (hatched bars). (B) The impact of 150 mM thiourea upon wild-type cell growth (MG1655) in LB medium. (C) Growth rate in LB medium as a function of thiourea concentration. The gray zone represents the concentration of thiourea (100–150 mM) commonly used as a putative antioxidant. (D) The basis of growth inhibition by thiourea is unknown, but its capacity to inhibit enzymes is demonstrated here using xanthine oxidase.

Figure 6B–C shows that 100–150 mM thiourea substantially slows the growth of E. coli. Indeed, the range of thiourea concentrations used in antioxidant experiments coincides with the doses that slow growth. Other antioxidants—including N-acetylcysteine, glutathione, and ascorbic acid—also impair cell growth modestly but consistently (Figure S3). The underlying mechanisms are unclear and may vary from compound to compound. Thiourea binds metals directly; we observed, for example, that it potently inhibits xanthine oxidase, potentially by associating with its molybdopterin cofactor (Figure 6D).

To test the linkage between growth rate and sensitivity to oxidative killing, we treated cells with norvaline, a non-conventional amino acid that is imported by cells, where it impairs protein synthesis [40]. This inhibition itself is non-lethal, and growth is rapidly restored when the norvaline is removed. Figure 7A shows that norvaline treatment diminished cell killing by H₂O₂ at the same doses that inhibited growth. DNA oxidation requires ongoing metabolism as a source of electrons [36], and it is likely that in this experiment metabolic slowing was the basis of protection.

Figure 7. Growth inhibition during DNA damage enables cells to tolerate threats to DNA.

(A) An exponentially growing recA mutant of E. coli (LEM17) in aerobic minimal glucose medium was exposed to 4.5 mM H₂O₂ for 30 s before the H₂O₂ was removed by catalase. Where indicated, the growth medium included norvaline, a valine isomer that slows cell growth. After exposure, cells were diluted and plated on LB medium to test viability. (B) In the same medium as in panel A, wild-type cells were exposed to norfloxacin (0.4 ug/ml) for 1 hr before dilution and plating. Growth and death rates were determined. (MG1655) (C) Wild-type cells in anoxic glucose/amino acids medium were exposed to norfloxacin for 1 hr, and survival was determined. Where indicated, thiourea was included. In parallel cultures the impact of thiourea upon growth rate was measured. Gray zone: concentration of thiourea typically used as an antioxidant (100–150 mM). ROS cannot be formed in anoxic medium.

Norfloxacin is a clinical quinolone antibiotic that introduces lethal double-strand breaks into DNA by inhibiting topoisomerase [41]. Quinolones are among the antibiotics whose lethality was ascribed in part to the generation of oxidative stress; key evidence included protection by thiourea and other scavengers [20, 21, 42]. Figure 7B shows that norvaline inhibited norfloxacin toxicity, too, showing that mild growth inhibition is sufficient to produce protection. Accordingly, the protective effect of thiourea was reproduced under anoxic conditions, indicating that it was the slowed growth rate, rather than any scavenging of ROS, through which thiourea protects cells from this agent (Figure 7C).

The upshot is that although a wide variety of molecules can scavenge oxidants in vitro, their impact in vivo is likely limited to extracellular H₂O₂ scavenging or to the slowing of cell physiology. Their protective effects cannot be reliably ascribed to the intracellular scavenging of oxygen species.

Discussion.

We have shown that an array of ostensible antioxidants are ineffectual inside a scavenging-proficient strain. They simply do not degrade ROS fast enough. Several of these compounds improved bacterial survival in the face of DNA-damaging stresses—but this effect seems to result simply from their ability to slow growth. Workers have gravitated to antioxidant doses that fall just short of those that obviously block growth, but we found that even a modest slowing is enough to improve survival.

It is hard to kill bacteria by amplifying ROS formation.

The sheer number of stresses that experimenters have tentatively linked to ROS is enough to give pause. They include cold, heat, salinity, osmotic stress, and hydrostatic pressure; aminoglycosides, B-lactams, fluoroquinolones, hydroxyurea, antifolates, and fungicides; organic solvents and nanoparticles; fatty acid catabolism, phosphate limitation, carbon starvation, and senescence; cytotoxic proteases, peptidoglycan recognition proteins, type-six secretion systems, and toxin-antitoxic systems; and so on [43]. Indeed, one commentary [44] that surveyed these proposals was entitled, “Are All Stresses Oxidative Stresses?” However, the direct targets of these stresses are diverse, which makes it hard to conceive of an underlying mechanism that could enable them all to increase ROS formation. Only a few obvious stresses—hyperbaric oxygen, redox-cycling drugs, and iron overload—drive O₂⁻, H₂O₂, or HO production in a way that is mechanistically understood. Many studies have speculated that disruption of the respiratory chain could trigger inappropriate univalent or divalent reduction of oxygen upstream of the cytochrome oxidase. However, although complete inhibition of electron flow does increase ROS production by the E. coli chain, the effect is mild—approximately a factor of three [45, 46]. This change would be inadequate to cause growth or enzymatic defects [28, 47]. There has not been a satisfying hypothesis for how any particular stress would accelerate ROS formation to the point of toxicity.

Moreover, even if ROS formation were accelerated, it is unlikely that it would rise to the level of killing cells. Studies have repeatedly documented that rapid O₂⁻ and H₂O₂ accumulation can damage enzymes and impair growth—which is biologically and ecologically important—but they still do not kill cells. DNA damage occurs, and mutagenesis happens, but the combinations of base-excision and recombinational repair are sufficient to avoid lethality. For example, when wild-type E. coli was exposed to a constant 9 micromolar concentration of H₂O₂ over a period of four hours, the cells transiently lagged and then resumed growth [33], with the adaptation being driven by both its OxyR-specific H₂O₂ response and the SOS DNA repair system [48]. Given the permeability coefficient of the membranes for H₂O₂ [10], during this period of exposure the rate of H₂O₂ influx was approximately 50 times the rate of usual endogenous production. Similarly, mutant strains of E. coli that lack peroxidases and catalases are able to tolerate 1 micromolar intracellular H₂O₂, which is twenty times the usual level [48]. Implicitly, even a stress that amplified cellular H₂O₂ production by these enormous factors would be non-lethal.

The reasons that cells are so resistant to supranormal ROS are well understood. They encounter external H₂O₂ in natural habitats and have acquired defenses against it. Bacteria typically wield a consortium of scavenging enzymes, and they also have devices that keep iron levels low and DNA repair capacity high. Further, each of these tactics is heightened during H₂O₂ stress. The OxyR transcription factor is activated whenever intracellular H₂O₂ levels exceed 0.3 μM [10, 33]. It then induces both catalase and AhpCF by more than an order of magnitude [49]. It turns on Dps, a miniferritin that protects the cell by sequestering the loose iron so it is not associated with DNA [50–52]. It turns on both Fur, a repressor that blocks further synthesis of iron importers, and YaaA, a protein that diminishes iron pools by a mechanism that is still unknown [53–55]. And it turns on exonuclease III, a DNA repair enzyme that completes the process of base-excision repair and also trims the lesions that arise from DNA ribose oxidation [48]. Mutants that lack the OxyR or SOS stress responses exhibit some loss of viability if intracellular H₂O₂ levels rise; it seems likely that any stressor that amplifies ROS production might also have to subvert these defenses to have any chance of killing the cell.

Is physiological ROS stress ever lethal to wild-type cells? Both plants and microbes are known to secrete redox-cycling compounds that accelerate ROS formation inside their bacterial targets [7, 8, 56, 57]. Juglone and pyocyanin are familiar examples, and paraquat is a man-made analogue. The ROS produced by these agents inactivates enzymes and paralyzes growth, but it is still non-lethal. Indeed, streptonigrin, an antibiotic that is released by a Streptomycete bent on killing its competitors, may be the exception that proves the rule. Once inside the target cell, streptonigrin binds iron, adheres to DNA, and reduces molecular oxygen all the way to a hydroxyl-like species without releasing either O₂⁻ or H₂O₂. Thus, the drug is able to generate a lethal oxidant on the DNA surface, but it does so in a way that sidesteps the scavenging defenses of the cell [47].

What are reliable markers of oxidative stress?

Our results do not refute prior claims that any particular stress does not increase ROS formation inside cells; they simply call into question a particular kind of evidence. But they do caution experimenters. The efficacy of any proposed antioxidant should be validated, ideally in control strains that display oxidative phenotypes. And it is crucial to show that the compound does not also slow growth—or else one cannot reliably attribute survival outcomes to its antioxidant activity.

Our data show that antioxidants can protect cells in the narrow circumstance in which H₂O₂ flows into a cell from the extracellular environment, where antioxidant scavenging can be effective. In contrast, most studies that infer toxic ROS production by a cell presume that O₂⁻ and H₂O₂ are generated within the cytoplasm, with the respiratory chain suggested to be the direct site. Studies of unstressed cells indicate that, in fact, most respiratory ROS are produced by autoxidation of the flavin moieties of respiratory enzymes—which indeed occurs on the cytoplasmic face of the membrane [45]. Still, E. coli releases some O₂⁻ and H₂O₂ into the periplasm via menaquinone oxidation [58], and in Gram-positive bacteria with extracellular electron transfer capabilities [59] this ROS yield is higher [60]. If a stress were to somehow amplify the latter events, then extracellular antioxidants could exert an effect. However, this model is fairly constrained, and it would be tested most clearly using catalase rather than cell-permeable chemicals.

How then should intracellular oxidative stress be diagnosed? Elevated levels of H₂O₂ are detected by the OxyR and PerR systems of most bacteria, so perhaps the easiest and most sensitive way to detect such a stress is by tracking the expression of a regulon member [49, 61]. A derivative of OxyR fused to YFP, called HyPer, is an alternative that allows visualization [62]. Both H₂O₂ and O₂⁻ damage [4Fe-4S] dehydratases, lowering their activities, and so their inactivation may be a marker of stress [63]. A stress that authentically kills cells by oxidizing its DNA will be especially lethal to DNA repair mutants, including recA strains devoid of recombinational repair [64]. And in specialized strains that lack H₂O₂ scavengers, the rate of intracellular H₂O₂ production can be directly measured by its efflux into the cell medium [65].

Some other approaches to ROS detection are inadvisable. Redox-active dyes have often been employed in studies, because, like antioxidants, they are easy to use. However, experts in oxidative stress have repeatedly warned workers that these dyes are not ROS-specific and that one must normalize the dye signal to the dye content of the cell [66] [67] [68] [69] [43] [70]. Activation of the SoxRS system has sometimes been interpreted as signaling elevated levels of O₂⁻; however, unlike the OxyR system, the SoxRS system is not responsive to ROS per se. Instead, its activity reflects the dynamic balance between its reduction by cellular systems and its direct oxidation by redox-active antibiotics [71]. Finally, because bacteria lack polyunsaturated fatty acids, lipid peroxidation is not believed to be a sequela of ROS stress and should not be used as an indicator [72].

Can antioxidants work at all in biological systems?

Despite the inadequacy of antioxidants at scavenging ROS inside the cell, these compounds can be effective at degrading H₂O₂ in extracellular environments, where slow action may suffice and they do not have to compete with scavenging enzymes. For example, they have proven useful in degrading the H₂O₂ that accumulates in standard lab medium under room lights, or the peroxides that are present in pre-made commercial media [32, 33]. Those peroxide levels are sufficient to impair the growth of some wild-type bacteria—and workers have found that H₂O₂ scavengers, including catalase or pyruvate or cysteine, improve plating efficiency of bacteria that are otherwise difficult to culture [73–75]. One would expect these agents to succeed whenever any bacterium is confronted with extracellular H₂O₂, whether it arise from phagocytes, the plant response, or amoebae. These antioxidants are likely to quench other environmental oxidants, too, including hypochlorous and hypothiocyanous acids.

The present study focused upon E. coli, the model organism in which the physiology of oxidative stress has been most thoroughly explored, but its underlying logic is likely to pertain equally well to other bacteria. Antioxidants have been even more widely used in studies of mammalian phenomena, and they have proved to be commercially viable in the health food market. The issues that confound their efficacy in E. coli—primarily, the presence of high-titer catalytic enzymes that dwarf the intracellular actions of antioxidants—potentially apply to mammalian cells as well. Indeed, regulatory agencies have concluded that clinical trials have failed to demonstrate any benefit of antioxidant-rich diets. The USDA no longer reports the antioxidant content of foods in the US [76], and the term “antioxidant” is now effectively forbidden from European advertising [77].

Experimental Procedures.

Chemicals.

Medium components and salts (except casamino acids), Tris, and calcium chloride were from Fisher Scientific. Acid-hydrolyzed casamino acids, catalase, thiourea, N-acetylcysteine, glutathione reduced, 5,5’-dithiobis(2-nitrobenzoic acid) (DTNB), ascorbic acid, evadarone, Tiron, norfloxacin, norvaline, L-threonine, hydrogen peroxide, NAD⁺, Amplex Red, bovine xanthine oxidase, xanthine, horseradish peroxidase, bovine copper-zinc superoxide dismutase, nitro blue tetrazolium, horse heart cytochrome c, paraquat, ortho-nitrophenyl-β-galactoside (ONPG), and bovine serum albumin (BSA) were purchased from Sigma. Coomassie Blue dye was from Pierce. Stock solutions of edaravone (50 mM) and norfloxacin (1 mg/ml) were made in 100% ethanol prior to use in aqueous solutions; Amplex Red (2 mg/ml) was dissolved in DMSO and then added to 40 ml of 100 mM potassium phosphate (KPi) buffer (pH 7.5).

Growth and media.

Cells were grown in LB or minimal A media [78]. LB medium contained (per liter) 10 g bactotryptone, 5 g yeast extract, and 10 g NaCl at pH 7. Defined medium contained minimal A salts and 0.2% glucose; where indicated, it was supplemented with 0.2% acid-hydrolyzed casamino acids. For anaerobic experiments, medium was moved from the autoclave directly into a Coy anaerobic chamber; media were held there for at least 24 hrs before use to ensure the full removal of oxygen. Reagents for anaerobic experiments were dissolved in anoxic water in the chamber.

The impacts of antioxidants and norvaline upon growth rates were determined at 37 C with shaking (220 rpm) for aerobic experiments and in an anaerobic chamber for anaerobic experiments. Overnight cultures were diluted to ~ 0.020 OD₆₀₀, grown in medium for 4 generations, and then diluted to 0.040 OD₆₀₀ into the same medium containing various concentrations of the potential inhibitor. Growth was tracked by optical density with time points every 30–60 min, and growth rate was calculated as the Ln(OD_t/OD_i)/hours, where OD_i and OD_t represent optical densities at initial and later time points. Because glutathione (GSH), N-acetylcysteine (NAC), and ascorbic acid can acidify media, their stock solutions were freshly made, and the pH was corrected to 7.2.

Strains.

All strains were direct derivatives of the wild-type E. coli strain MG1655 [79] and were generated in our lab. KCI415 is a SOD-deficient mutant: as MG1655 plus (sodA::Mud PR13)25 (sodB-Kan)1-Δ2. KCI746 is a sodA mutant: as MG1655 plus ΔsodA1::kan. LEM17 [55] is a recA mutant lacking both recombinational and SOS-inducing ability: as MG1655 plus recA56 srl300::Tn10. AL343 [33] is unable to induce the OxyR regulon: as MG1655 plus ΔoxyR::spec. AL441 [33] possesses a katG-lacZ transcriptional fusion: as MG1655 plus attλ::[pSJ501::katG’-lacZ⁺]~cat. MK186 lacks catalase and Ahp peroxidase activities and contains the same katG-lacZ transcriptional fusion: as MG1655 plus Δ(katG17::Tn10)1 (ahpC-ahpF’)del kan::’ahpF Δ(katE12::Tn10)1 attλ::[pSJ501::katG’-lacZ⁺]~cat.

Catalase and SOD assays.

Overnight cultures in LB medium were diluted to 0.005 OD₆₀₀ in 30 ml of oxic LB medium and grown to OD 0.4. Cultures were centrifuged at 10000 g for 7 min, washed with ice-cold KPi buffer, and resuspended in 2 ml of 100 mM KPi buffer, pH 7.2. The resulting suspension was lysed using French press, and the lysate was clarified by centrifugation.

Catalase activity was measured by monitoring the decrease in the absorbance of hydrogen peroxide at 240 nm [80]. Aliquots of cell lysate were added to 1 ml of 10 mM solution of hydrogen peroxide in 100 mM KPi buffer (pH 7.2, room temperature). The activity was normalized per protein content. One unit of catalase activity corresponded to decomposition of 1 micromole of H₂O₂ in 1 min.

Superoxide dismutase/ superoxide scavenging ability activity was determined by the ability of cell lysate or an antioxidant to decrease the rate of cytochrome c reduction by superoxide produced by xanthine-xanthine oxidase system [25]. The 1-ml reaction contained 10 μM cytochrome c in 100 mM KPi buffer (pH 7.2), xanthine oxidase, and a known amount of cell lysate or antioxidant. The amount of xanthine oxidase was chosen to reduce cytochrome c at a rate of approximately 1 μM per minute. The reaction was conducted at room temperature and was initiated by the addition of 100 μM xanthine. The absorbance was monitored at 550 nm. Units of scavenging activity were calculated by the formula: Units = [(Control rate/sample rate]-1)/3.

In studies of xanthine oxidase inhibition, the activity of xanthine oxidase was tracked by its production of urate, at 295 nm. This reaction was conducted in the absence of cytochrome c, which would otherwise obscure the absorbance of urate.

The protein content of cell extracts was determined using Coomassie Blue dye binding agent. Calibration was performed using BSA, and the absorbance of the protein-dye complex was measured at 595 nm.

Rate of thiourea equilibration across membranes.

The change in intracellular TU concentration per unit time is the product of the out-to-in TU gradient, the membrane permeability coefficient P, and the cell surface area, divided by cell volume:

$$\text{d}\left[{\text{TU}}_{\text{in}}\right]/\text{dt}=\left\{\left[{\text{TU}}_{\text{out}}\right]-\left[{\text{TU}}_{\text{in}}\right]\right\}\times \text{P}\times \text{A}\times {\text{V}}^{-1}$$

The value of $\text{P}=5.5\times {10}^{-7}\text{cm}/\text{s}\left[27\right]$ and $\text{A}=1.4\times {10}^{-7}{\text{cm}}^2\left[10\right].$ For a cell of volume $3\times {10}^{-15}\text{L}$:

$$\begin{gathered} \text{d}\left[{\text{TU}}_{\text{in}}\right]/\text{dt}=\left\{0.15\text{M}-\left[{\text{TU}}_{\text{in}}\right]\right\}\left(5.5\times {10}^{-\text{7}}\text{cm}/\text{s}\right)\left(1.4\times {\text{10}}^{-7}{\text{cm}}^{\text{2}}\right)\left({\text{10}}^{-3}{\text{L}/\text{cm}}^3\right){\left({3\times 10}^{\times 15}\text{L}\right)}^{-1} \\ \text{d}\left[{\text{TU}}_{\text{in}}\right]/\text{dt}=\left\{0.00385\text{M}.0257\left[{\text{TU}}_{\text{in}}\right]\right\} \\ \text{By}\text{integration},\left[{\text{TU}}_{\text{in}}\right]=-0{.150\text{e}}^{-0.0257\text{t}}+0.150. \end{gathered}$$

Then [TU_in] = 75 mM in 27 s. This is the halftime for equilibration of thiourea across the membranes, and it means that the intracellular thiourea concentration will reach 90% of the extracellular concentration within 90 s.

Rate constants for reactions of antioxidants with O₂⁻.

The reaction constants with which chemical antioxidants scavenge superoxide anion were determined at room temperature in a xanthine oxidase/cytochrome c reaction as described in the preceding section, with sufficient antioxidant added to suppress cytochrome c reduction. Rate constants were deduced from the rate constant with which O₂⁻ reduces cytochrome c (2.6 × 10⁵ M⁻¹ s⁻¹ [81]) and the working concentration of the cytochrome c (10 μM). For example, if 5 mM of an antioxidant is needed to suppress cytochrome c reduction by 50%, then that antioxidant scavenges O₂⁻ with a rate constant of 5 × 10² M⁻¹ s⁻¹. None of the thiol antioxidants directly reduces cytochrome c at a rate sufficient to interfere with this analysis.

Because ascorbate directly reduces cytochrome c, 40 μM nitroblue tetrazolium (NBT) was used instead of cytochrome c, and the scavenging activity was calculated from the ability of an inhibitor to inhibit NBT reduction by O₂⁻ [26]. The formation of formazan was monitored at 540 nm; the relation between inhibition and SOD activity was established by the addition of known amounts of CuZnSOD, whose activity had been previously determined by the cytochrome c-based assay.

Measurement of acid-soluble thiols.

Cells were grown in 30 ml aerobic LB medium supplemented with 50 mM KPi (pH 7.2) from 0.01 OD₆₀₀ to 0.1 OD₆₀₀, and pH-balanced NAC or GSH were then added to a final concentration of 10 mM. Cells were grown for an additional 30 min, centrifuged, washed twice with ice-cold KPi buffer. The final cell pellet was suspended in 0.1 ml of 100 mM of KPi buffer; 10 μl was used to determine the OD₆₀₀ of the suspension. Then 0.5 ml of 7% sulfosalicylic acid was added to the cells. The mixture was thoroughly mixed and held on ice for 10 min. The solid material was removed by 3 min centrifugation in a table-top centrifuge, and the supernatant was mixed 1:10 with 0.4 mM DTNB solution in 1 M KPi (pH 7.2). The resultant absorbance was measured at 412 nm.

Intracellular O₂⁻ stress.

The ability of externally added antioxidants to relieve intracellular superoxide stress was studied in a ΔsodA ΔsodB mutant that lacks cytoplasmic SOD. Such strains cannot grow in aerated minimal media in the absence of branched-chain, aromatic, and sulfur-containing amino acids [82]. Wild-type (MG1655) and SOD⁻ (KCI415) strains were inoculated to 0.01 OD₆₀₀ into anoxic minimal A medium supplemented only with 0.2% glucose. After reaching 0.2 OD₆₀₀, cells were diluted tenfold into aerated medium. Where indicated, antioxidant was included in the aerobic medium.

The ability of antioxidants to relieve superoxide stress was also studied in a ΔsodA single mutant under oxic conditions in the presence of 1 uM of paraquat. Cells were pre-cultured in aerobic medium to 0.2 OD₆₀₀ and diluted 1:10 into the same fresh medium; paraquat and antioxidants were then added. Growth was monitored at 600 nm.

Threonine dehydrogenase.

The threonine dehydrogenase activity of cell lysates was assayed anaerobically by monitoring the reduction of NAD after addition of threonine [30]. Aerobically grown cells (30 ml of LB medium) were centrifuged (10000 g) aerobically at 4 C for 7 min and washed once with ice-cold phosphate buffer. Collected cells were put on ice, moved to anaerobic chamber, and resuspended in 2 ml of ice-cold anaerobic Tris medium (50 mM, pH 8.0). Sonication (3 sec pulses separated by 3 sec time intervals) lasted for 90 sec of total sonication time. Lysates were briefly spun on a table-top centrifuge (1 min × 13000 rpm), and 50 μl of lysate was mixed with 1 ml of same anoxic Tris buffer containing 2 mM NAD⁺ and 50 mM of threonine. Cuvettes were sealed anaerobically and moved to a spectrophotometer, where absorbance at 340 nm was monitored. Anoxic conditions must be maintained during the assay to avoid NADH oxidation by the aerobic respiratory chain in inverted vesicles.

Rate constants for reactions of antioxidants with H₂O₂.

Antioxidants (20 mM of TU, 5 mM of GSH, or 5 mM NAC) were incubated with 0.5 mM of hydrogen peroxide in 100 mM KPi buffer (pH 7.6) at 37 C. At intervals, 10 μl aliquots were mixed with 1.5 ml of the 100 mM KPi buffer (pH 7.2) containing Amplex Red dye (1 mg/40 ml) and 0.5 ml of the same buffer with horseradish peroxidase (1 mg/ml) [46]. The resulting fluorescence of the oxidized dye was measured using 520 nm excitation and 620 nm emission filters. Second-order rate constants were calculated by fitting the data to an exponential decay (semilog) function.

The oxidation of ascorbate by H₂O₂ was monitored directly by the decrease in ascorbate absorbance at 270 nm. Ascorbate (1 mM) in KPi buffer was mixed with 1 mM of H₂O₂ and incubated at 37 C. At intervals 50 μL aliquots were mixed with 950 μl of the buffer for analysis. The reaction rate was calculated over the first minute.

The expression of katG.

Expression of OxyR-controlled katG, encoding catalase, was measured using a katG-lacZ fusion in both wild-type (MG1655) and the catalase/peroxidase-deficient background (ΔahpCF ΔkatG ΔkatE, strain MK186). Cells were grown in deaerated LB anaerobically from OD 0.01 to OD 0.1. They were then moved from the anaerobic chamber to a laboratory shaker and incubated aerobically for 45 min, in the presence of corresponding antioxidants where indicated. Cell cultures (30 ml) were then collected, washed with ice-cold KPi buffer, resuspended in 2 ml of KPi buffer, lysed by French press, and briefly centrifuged (1 min × 10000 g) to remove crude debris. The β-galactosidase activity of the lysate was measured using ONPG (1 mg/ml) in 100 mM KPi buffer as a substrate [78].

Extracellular scavenging of hydrogen peroxide by antioxidants on solid medium.

Several micromolar hydrogen peroxide is naturally formed in LB medium under room lighting by photochemical processes, and oxyR mutants cannot form isolated colonies upon LB plates due to their inability to induce defensive responses [32, 33]. The ability of antioxidants to scavenge this H₂O₂ was appraised by testing the ability of the oxyR mutant AL343 to grow when the antioxidants were added to these plates. LB plates were poured aerobically and solidified for 20 h at regular room illumination. Where indicated, the plates contained 5 mM of pyruvate or 75 mM of TU in the agar, or 300 U catalase (in 0.1 ml KPi buffer) was top-spread upon them. The oxyR mutant was grown on anaerobic LB plates anaerobically; a loopful was suspended in buffer and then re-streaked on aerobic plates. The plates were incubated at 37 C aerobically in the dark. Growth was assessed after 16 hours.

The ability of scavengers to protect against hydrogen peroxide was also visualized by measuring the zone of inhibition around a peroxide-containing paper disc. Where indicated, LB plates contained 100 mM of TU, or 10 U/ml of catalase was injected into autoclaved LB agar after cooling to 45–50 C shortly before pouring. Wild-type (MG1655) cells were grown in aerobic LB medium to 0.2 OD and diluted 1:100 in fresh LB, and then 75 ul of this cell suspension was spread on the surface of the plate. A 1 cm Whatman disk was placed in the middle of the plate, 10 μl of 30% H₂O₂ was added to its center, and the plate was placed in the dark at 37 C. The zone of inhibition assessed after 16 hours of aerobic growth.

Plasmid DNA nicking.

Bluescript plasmid DNA (400 ng) [47] in 10 mM HEPES buffer (pH 7) was mixed with 8 μM of ferrous ammonium sulfate in the presence or absence of 150 mM of thiourea at room temperature under oxic conditions (20 μl total). The reaction was initiated by the addition of 0.5 mM H₂O₂. After 45 s 0.5 mM of DTPA and 300 U/ml (final) catalase were added. DNA fragments were separated using agarose gel (1%) electrophoresis and visualized via ethidium bromide staining. Notably, the rate constants for the Fenton reaction (~ 2000 M⁻¹ s⁻¹ at RT for DNA-bound iron, and higher for dissociated iron [83]) are high enough that > 99% of the iron reacts in the first five seconds.

Phage inactivation.

Bacteriophage P1 obtained from MG1655 wild type cells were diluted into 1% NaCl, 10 mM HEPES (pH 7.2) at 37 C (2 × 10⁵ pfu/ml). Fresh made (5 min old) ferrous ammonium sulfate solution in water was added to a final concentration of 1–10 μM and mixed by pipetting. The desired concentration of H₂O₂ was added (see figure legends). After 45 seconds, 0.1 ml of the phage reaction mix was added to 0.9 ml catalase (300 U/ml final in the same buffer). Then 0.1 ml of the phage suspension was added to 0.5 ml of 0.01 OD₆₀₀ MG1655 cells in LB medium supplemented with 10 mM of calcium chloride, and the resulting mixture was plated in 5 ml of LB top agar and spread on LB plates. Plates were incubated at 37 C, and plaques were counted after 16 hours.

Cell killing by oxidative DNA damage.

The ability of antioxidants to prevent DNA damage in vivo was studied in a recA mutant (LM17) grown in minimal A medium with 0.2% casamino acids. When norvaline was added to suppress the growth rate, casamino acids were omitted. Overnight cultures were resuspended in fresh medium and grown for four generations aerobically. Cells were then diluted to 10⁶ cfu/ml (approximately 0.002 OD₆₀₀) into pre-warmed (37 C); where indicated, this medium was supplemented with filter-sterilized antioxidant or norvaline in 100 mM KPi buffer (pH 7.2). After 10 min of incubation, H₂O₂ was added to a final concentration of 4.5 mM. Cells were incubated for precisely 30 s and then diluted 1:10 into a sterile catalase solution (300 U/ml) in 100 mM of KPi (pH 7.2, room temperature). After 5 min, 0.1 ml of the cell suspension was mixed with 5 ml of top LB agar and plated on LB plates. Colonies were counted after 16 hours.

Cell killing by norfloxacin.

Norfloxacin toxicity was studied in wild type cells in LB or in minimal medium, both aerobically and anaerobically. For anaerobic experiments all plates and top agars were solidified in the anaerobic chamber. Liquid media were de-aerated for 24 hours, and all stock solutions were prepared by dissolving chemicals in anoxic KPi with subsequent syringe filtration. The top agar was melted and kept in a heating block in anaerobic chamber. Cells were grown in minimal A medium containing 0.2% glucose and 0.2% casamino acids. The effect of thiourea on the growth rate was tested as described previously. When exponentially growing cells reached 0.1 OD, cell cultures were diluted (1:10) into a series of flasks with pre-warmed medium containing different concentrations of thiourea. Norfloxacin (0.4 ug/ml) was added after 6 min of incubation. Cultures were incubated for 1 hour, and samples were diluted 1:100 in 100 mM KPi (pH 7.2) and plated in 5 ml of LB top agar.

Aerobic experiments with norfloxacin were conducted in minimal A medium containing 0.2% glucose. Overnight cells were diluted to OD 0.01, grown in aerobic medium to OD 0.2, and diluted 1:10 into flasks containing various concentrations of norvaline. Norfloxacin (0.2 ug/ml) was added, and cells were incubated for 1 hour at 37 C. They were mixed with 5 ml LB top agar and spread over LB plates. Colonies were counted after 16 hours of incubation. In norfloxacin experiments the death rate was determined as a negative natural logarithm of the ratio between the number of CFU at the starting point of the incubation with antibiotic and the final CFU number.

Supplementary Material

Figure S1. GSH and NAC supplements do not significantly elevate low-molecular-weight thiols inside the cell. GSH (10 mM) or NAC (10 mM) were added to wild-type (MG1655) cell cultures in buffered LB for 30 min. Cells were lysed, and total low-molecular-weight thiols were quantified. Thiol content may have increased slightly, but the changes were not statistically significant.

Figure S2. Ascorbate (5 mM), glutathione (10 mM), N-acetylcysteine (10 mM), edaravone (0.1 mM), and Tiron (10 mM) do not suppress the aerobic growth defect of a superoxide dismutase-deficient strain. Exponentially growing cells (MG1655, KCI415) in anaerobic minimal medium were diluted into aerobic medium containing the indicated antioxidants at time zero. Superoxide stress blocks growth by inactivating enzymes needed to synthesize amino acids.

Figure S3. Antioxidants consistently slow cell growth. Antioxidants were added to LB medium, and growth of wild-type MG1655 was monitored. (Left). Example growth curve. (Right). Normalized data from three independent cultures.

Figure S4. Thiourea fails to protect phage at any H₂O₂ concentration. Bacteriophage P1 was exposed to 1 μM iron and the indicated concentrations of H₂O₂, as described in the caption of Fig. 5. All concentrations of H₂O₂ suffice to fully oxidize the Fe(II), thereby generating an equivalent amount of damage. Where indicated, thiourea was added prior to the H₂O₂.