Two modes of killing of Escherichia coli by hydrogen peroxide can be distinguished. Mode-one killing is maximal at 1–2MM; at higher concentrations the killing rate is approximately half- maximal and is independent of H2O2 concentration but first order with respect to exposure time. Mutagenesis and induction of a phage lambda lysogen are similarly affected by H2O2 concentration, with reduced levels of response above 1–2 mM-H2O2. Mutagenesis is not affected by inactivation of umuC. Mode-one killing requires active metabolism during the H2O2 challenge and it results in sfiM-independent filamentation of both cells that survive and those that are killed by the challenge. This mode of killing is enhanced in xth, polA, recA and recB strains; however, it is unaffected by mutations in the nth, uvrA, uvrB, uvrC, uvrD, rep, gyrA, htpR and rel loci. Mode- one killing is normal in strains totally lacking catalase activity (katE, katG), glutathione reductase (gor) or glutathione synthetase (gshB), but enhanced in a strain lacking NADH dehydrogenase (ndh). Mode-one killing is accelerated by the presence of CN or by an unidentified function that is induced by anoxic growth and is under the control of the fnr locus. A strain carrying both xth and recA mutations and certain polA mutants appear to undergo spontaneous mode-one killing only under aerobic conditions. Taken together, these observations imply that mode-one killing results from DNA damage that normally occurs at a low, non-lethal level during aerobic growth. Models for the resistance to mode-one killing at doses above l–2mM-H2O2 will be discussed.

Mode-two killing occurs at high concentrations of H2O2 and longer times. It does not require active metabolism, and cells that are killed do not filament, although survivors demonstrate a dosedependent growth lag followed by a period of filamentation. Mode-two killing is accompanied by enhanced mutagenesis, but strains with DNA repair defects were not observed to be especially sensitive to this mode of killing.

Our laboratory has utilized H2O2 toxicity in Escherichia coli as a simple model system for studying DNA damage induced by free radical agents such as nearultraviolet light, gamma-irradiation, and chemical carcinogens (Demple & Linn, 1982; Demple et al. 1983; Demple & Halbrook, 1983; Imlay & Linn, 1986). In essence, we hope to learn by what mechanism and to what species H2O2 is activated to cause DNA damage, what the nature of the damage is, and how it is repaired. E. coli was selected because of the wealth of biochemical and genetic knowledge of its DNA repair functions, and it is expected that the knowledge gained will be applicable to eukaryotic microorganisms and finally to mammalian systems.

One important discovery in this program has been the ability of E. coli to ‘adapt’ to H2O2, i.e. after exposure to low levels of the agent the bacteria become resistant to normally toxic doses of H2O2 and gamma radiation (Demple & Halbrook, 1983). Christman et al. (1985) have recently shown that this response is at least partially under the regulation of a locus, oxyR, which is apparently responsible for a positive regulator that induces increased levels of scavenger enzymes, notably peroxidases, catalase and superoxide dismutase. In addition, the response was shown to overlap with the heat-shock response.

This paper summarizes some of our most recent findings concerning killing and mutagenesis of E. coli by hydrogen peroxide, and the possible involvement of at least some aspects of the SOS response in surviving exposure to the agent.

Challenges with H2O2 were for 15 min in K medium plus 1 % glucose, as described by Imlay & Linn (1986). E. coli strains utilized are either described therein or are as follows: C600(λ) was obtained from Dr H. Echols, University of California, Berkeley; GW2100 (as AB1157 plus umuCl22::Tn5) was from Dr G. Walker, MIT; CSH7 (lac7 rpsL thi-1) and UM1 (as CSH7 plus katEl katG14) were from Dr P. Lowen, University of Manitoba; RK4353 (ElacY169 araD139 non thi rpsLgyrA) and RK5279 (as RK4353 plus fnr250) were from Dr Valley Stewart, Stanford University; IY13 (Fthi his ilv trp rspL) and IY12 (as IY13 plus ndh) were from Dr Ian Young, Australian National University; JF511 (as JC7623 plus gshB:: Kan) was from Dr James Fuchs, University of Minnesota.

Some DNA repair mutants are sensitive to relatively low concentrations of H2O2

When DNA repair-proficient strains of E. coli are exposed to various concentrations of H2O2, a small amount of killing is generally observed near 2·5 mM, but drastic toxicity is only observed above roughly 25 mM during a 15-min exposure (Fig. 1; and Imlay & Linn, 1986). The killing at low doses is enhanced in several mutants lacking DNA repair enzymes (Fig. 1), most notably xth mutants, which lack exonuclease III, recB mutants, which lack recBC enzyme (exonuclease V), and particularly polA mutants, which lack DNA polymerase I. The sensitivity of recB mutants appears to be due to a defect in recombinational repair, because suppression of this defect with a second, recF’-activating sbcB mutation gives rise to cells that still lack the recBC enzyme, also lack exonuclease I, are recombinational repairproficient, and regain resistance to low concentrations of H2O2 (Fig. 1).

Fig. 1.

Survival of DNA repair-defective strains after challenge with H2O2.

Fig. 1.

Survival of DNA repair-defective strains after challenge with H2O2.

A property that distinguishes the killing curves of the DNA repair-defective mutants is their unique shape: killing is maximal at roughly 2·5 mM, then somewhat reduced and independent of H2O2 concentration up to 20 mM-H2O2, where the rate of killing begins to increase with concentration. This relative independence of killing from H2O2 concentration at low and moderate doses, the enhancement of killing at these doses in certain mutants, and other properties described below define a mode of killing at the low and moderate exposures, which we denote as mode-one, that is different from that at high doses, which we denote as mode-two.

We have also noted that extremely repair-defective mutants such as an xth recA strain (Imlay & Linn, 1986) or apolA deletion mutant (Morimyo, 1982) appear to undergo mode-one killing during aerobic growth, probably due to the generation of endogenous H2O2. Under anoxic conditions these strains grow well and without any indication of such killing.

A notably resistant strain (Fig. 1) is one carrying a mutation in nth, the structural gene for endonuclease III, a DNA glycosylase recognizing oxidized thymine residues such as thymine glycol (Demple & Linn, 1980; Cunningham & Weiss, 1985). Also showing normal levels of killing by H2O2 are strains carrying uvrA, uvrB, uvrC, uvrD, rep, gyrA and rel mutations (not shown).

Mutagenesis and stress responses induced by exposure to H2O2

The sensitivity of recA mutants to mode-one killing by H2O2 (Fig. 1) implies that the SOS responses might be normally induced by H2O2 to protect cells against this agent. Indeed, a lexA3 mutant also exhibits enhanced sensitivity to H2O2 and the peculiar dose-response seen for other DNA-repair mutants (Fig. 2). Moreover, when one scores for induction of phase λ in C600(λ), phage induction is seen at very low concentrations of H2O2, reaches a maximum below 0 ·5 mM, is constant until roughly 3 mM, then drops off with increasing concentration (Fig. 3). (In this experiment significant killing (<85% survival) occurred only above 15 mM- H2O2 Evidently the SOS response is induced by challenges of H2O2 that produce mode- one, but perhaps not mode-two, killing.

Fig. 2.

Survival of DM49 (lexA3) after challenge with H2O2.

Fig. 2.

Survival of DM49 (lexA3) after challenge with H2O2.

Fig. 3.

Induction of C600(λ) after challenge with H2O2. After 15 min of exposure to H2O2, catalase was added and cells were diluted and plated with E. coli AB 1157 as an indicator.

Fig. 3.

Induction of C600(λ) after challenge with H2O2. After 15 min of exposure to H2O2, catalase was added and cells were diluted and plated with E. coli AB 1157 as an indicator.

An important aspect of the SOS response in E. coli is the enhancement of mutagenesis due to the induction of the umuC and umuD genes (Walker, 1985). When mutagenesis was measured after exposure to H2O2 by induction of trimethoprim resistance (i.e. loss-of-function mutation of thyA, the structural gene for thymidylate synthetase), a dose-response similar to the killing dose-response was observed: a maximal level of mutagenesis near 2-5 mM, reduced mutagenesis at intermediate ranges, then enhanced mutagenesis at the high doses that give mode- two killing (Fig. 4). Most intriguing, induced mutagenesis was not altered in a strain carrying a uniuC::Tn5 allele, i.e. mutagenesis by H2O2 does not appear to be affected by the product of the umuC gene.

Fig. 4.

Mutagenesis of AB1157 (umuC+) and GW2100 (umuC::Tn5) to trimethoprim resistance after challenge with H2O2. After 15 min of exposure to H2O2, catalase was added and cells were plated with 2 ml of top agar containing 2 mg of thymine to score survival, or 2 mg thymine plus 0’2mg trimethoprim to score mutagenesis. Survival up to 20 mM-H2O2 for both strains was not significantly different than that of AB 1157 shown in Fig. 1. Survival at 25, 30 and 35mM-H2O2 was 86%, 20% and 4%, respectively, for AB 1157, and 82%, 21 % and 1 %, respectively, for GW2100.

Fig. 4.

Mutagenesis of AB1157 (umuC+) and GW2100 (umuC::Tn5) to trimethoprim resistance after challenge with H2O2. After 15 min of exposure to H2O2, catalase was added and cells were plated with 2 ml of top agar containing 2 mg of thymine to score survival, or 2 mg thymine plus 0’2mg trimethoprim to score mutagenesis. Survival up to 20 mM-H2O2 for both strains was not significantly different than that of AB 1157 shown in Fig. 1. Survival at 25, 30 and 35mM-H2O2 was 86%, 20% and 4%, respectively, for AB 1157, and 82%, 21 % and 1 %, respectively, for GW2100.

Though Christman et al. (1985) reported an overlap in the proteins induced by H2O2 exposure and by heat-shock, we have observed normal dose-responses for killing by H2O2 of an htpR mutant (Imlay & Linn, 1986), i.e. the htpR stress responses do not obviously protect against killing by H2O2.

As noted by Christman et al. (1985), the oxyR locus is reponsible for regulating, probably in a positive manner, a stress response to H2O2. Mutants defective in the oxyA-regulated response are not abnormally sensitive to mode-one or mode-two killing if the cells are not pretreated (Imlay & Linn, 1986). However, pretreatment of DNA repair-proficient cells with adapting doses of H2O2 results in protection against mode-two killing at high concentrations of H2O2 (Fig. 5), and this protection is not observed in the presence of chloramphenicol or in a mutant with a deletion of the oxyR locus (Imlay & Linn, 1986). Protection is also not seen in mutants that totally lack catalase (Fig. 5), so that the induction of catalase appears to be largely responsible for the induced protection of repair-proficient cells to mode-two killing. Remarkably, however, catalase mutants are not especially sensitive to H2O2 in the non-adapted state (Fig. 5).

Fig. 5.

Survival of CSH7 (kat+) and UM1 (katE katG) after challenge with H2O2 Cells in the left frame were not pretreated with adapting doses of H2O2; cells in the right frame were pretreated with 60 UM- H2O2 for 70 min before challenge.

Fig. 5.

Survival of CSH7 (kat+) and UM1 (katE katG) after challenge with H2O2 Cells in the left frame were not pretreated with adapting doses of H2O2; cells in the right frame were pretreated with 60 UM- H2O2 for 70 min before challenge.

When repair-deficient strains are adapted to H2O2, protection against mode-two killing at high H2O2 concentrations and against mode-one killing at intermediate concentrations is observed; however, the extent of mode-one killing at concentrations of H2O2 below SmM is not affected (Imlay & Linn, 1986). Therefore the oxyR-mediated adaptation appears not to be efficient against very low doses, i.e. those that themselves induce the adaptation.

In sum, it appears that the stress response by E. coli to H2O2 is quite complex and involves at least some functions of the /exA-regulated SOS response as well as the oxyR-regulated response. The former seems most effective at low H2O2 concentrations, whereas the latter appears most effective at high concentrations. The involvement of the htpR-regulated, heat-shock response, if any, is at present unclear.

Anoxic growth sensitizes cells to mode-one killing through function(s) regulated by the fnr locus

When DNA repair-proficient E. coli strains are grown anoxically, they become sensitive to mode-one killing by H2O2 with a dose-response similar to that seen for DNA repair-deficient strains (Imlay & Linn, 1986; and Fig. 6). This sensitivity is synergistic with that induced by DNA-repair deficiency. It is also dependent upon a functional fnr gene (Fig. 6), a gene that positively regulates the adaptation to low- oxygen or anoxic growth conditions (fnr =fumarate, nitrate reductase) (Shaw & Guest, 1982). Identification of the induced function that sensitizes the cells to mode-one killing is a major goal; we have already ruled out the involvement of menaquinone, cytochrome D and fumarate reductase.

Fig. 6.

Sensitization by anoxia to mode-one killing by H2O22 requires fnr function. RK43S3 (fnr+) and RK5279 9(fnr-250) were challenged after aerobic growth or after 60 min of anoxic growth. Similar resistance to mode-one killing was seen in a strain carrying the fnr allele, nirR22. The absence of a requirement for fnr function to enhance mode-two killing is a reproducible effect.

Fig. 6.

Sensitization by anoxia to mode-one killing by H2O22 requires fnr function. RK43S3 (fnr+) and RK5279 9(fnr-250) were challenged after aerobic growth or after 60 min of anoxic growth. Similar resistance to mode-one killing was seen in a strain carrying the fnr allele, nirR22. The absence of a requirement for fnr function to enhance mode-two killing is a reproducible effect.

Distinguishing features of the two modes of killing

Table 1 summarizes properties that distinguish mode-one from mode-two killing by H2O2. The requirement for active metabolism for mode-one killing is immediately observed upon addition of glucose to starved cells, even if chloramphenicol is present (Imlay & Linn, 1986). The linearity of mode-one killing with time occurs for as long as 15min at moderate doses; at very low doses killing eventually becomes slower because of medium detoxification, whereas at higher doses killing eventually increases as a result of the manifestation of mode-two killing. The phenotypic responses of mode-one killing challenges clearly differ from those of mode-two killing challenges. It appears that the former inhibit cell division, but not growth, whereas the latter inhibit growth as well. The ordered response noted for survival from mode-two killing challenges suggests, moreover, that mode- two killing lesions are repaired prior to mode-one killing lesions that might also be present.

Table 1.

Properties of the two modes of killing by H2O2

Properties of the two modes of killing by H2O2
Properties of the two modes of killing by H2O2

Models to explain the peculiar dose-responses observed for mode-one killing and mutagenesis

Two important unknowns concerning H2O2 toxicity are: (1) the mechanism(s) by which H2O2 is activated to induce DNA damage and other killing lesions; and (2) the origin of the peculiar dose-response noted for mode-one killing. It appears to us that the solution to (2) will help to provide answer(s) to (1).

The unique resistance to mode-one killing observed at moderate versus low concentrations of H2O2 presumably can be due either to a blockage of lesion production or an enhanced ability to tolerate lesions at the higher H2O2 concentrations. All normal induction phenomena are ruled out by various chloramphenicol experiments and by the fact that similar dose-responses are seen over 90-s exposures (Imlay & Linn, 1986). Activation effects (e.g. delayed DNA replication, DNA repair enzyme activation, etc.) or inactivation effects (e.g. destruction of agents that activate H2O2) are ruled out by the linearity of mode-one killing with time, the observation that once cells are exposed to low doses of H2O2 they cannot be rescued at higher doses, and, most notably, the observation that when cells are diluted out of 10mM-H2O2 into 0 ·5 M or lmM-H2O2, they immediately start being killed at the enhanced rate typical of the lower concentration (Imlay & Linn, 1986). These observations, coupled with the fact that import and export of H2O2 are diffusion limited, suggest that the resistance to higher doses results from suppression of activated intermediates by the higher H2O2 concentrations.

We should like to propose two very general types of models for this suppression. Both models assume that during electron flow from glucose to the ultimate electron acceptor, there is/are donor(s), D, which pass electrons (one or two at a time) to acceptors, A:

formula

H2O2 would be activated in this process when it accepts one electron from D in place of A so as to form the very toxic hydroxyl radical:

formula

This process is presumed to be saturated at some concentration of H2O2, however, due either to limiting D or to saturation by substrate of a reaction that follows standard enzyme kinetics.

To explain reduced toxicity as the H2O2 concentration is raised, the first model proposes a chemical quenching of the toxic OH species (or a related product) by the reaction:

formula

This reaction can be calculated from half-reaction potentials to have a favourable ΔG of − 9kcal (1 cal = 4 ·184J). The superoxide formed by this quenching reaction is presumably less toxic; however, it too can react with peroxide by the Haber-Weiss reaction:

formula

This situation can explain the dose-response as follows: at very low H2O2 concentrations killing increases as OH generation increases with increasing H2O2; however, killing begins to be suppressed at higher doses as the rate of OH’ generation is saturated, while OH’ is converted to HO2. by the quenching reaction. Finally, a kinetic steady-state is reached (Fig. 7) in which the quench and Haber—Weiss reactions serve to carry out the catalase reaction in a redox cycle that generates levels of OH and HO2 (and hence killing) that are independent of H2O2 concentration.

Fig. 7.

A hypothetical scheme for the production and redox cycling of free radicals.

Fig. 7.

A hypothetical scheme for the production and redox cycling of free radicals.

In favour of this model is the effect of starvation, which stops the cycle, and of adaptation to H2O2, which reduces the killing at moderate doses but not very low ones. The latter effect could be explained by the induction of superoxide dismutase, an enzyme that would perturb the steady-state levels of OH and H2O2 (hence killing) as shown in Fig. 7.

A second type of model to generate the dose-response would propose an electron diversion at higher H2O2 concentrations from one-electron H2O2 activation to a peroxidase function(s) that catalyses the two-electron reduction of H2O2 to H2O (Fig. 8). By this scheme, higher H2O2 concentrations draw electrons away from activation of H2O2 to OH towards one or more peroxidase sites that have lower affinities for H2O2 than the one-electron activation reaction. If the peroxidase also can be saturated with H2O2, a constant level of killing would be observed above some H2O2 concentrations (Fig. 8).

Fig. 8.

A hypothetical scheme for the diversion of electrons from one-electron H2O2 activation to two-electron peroxidation. On the graph on the right: (—) the observed rate of killing; (…..) the saturable H2O2 activation reaction; (– – –) the electrons available for activation in the presence of electron diversion to peroxidase(s). The combination of the dotted and broken line dependences could generate the observed killing response by adjustment of the H2O2 activation rates.

Fig. 8.

A hypothetical scheme for the diversion of electrons from one-electron H2O2 activation to two-electron peroxidation. On the graph on the right: (—) the observed rate of killing; (…..) the saturable H2O2 activation reaction; (– – –) the electrons available for activation in the presence of electron diversion to peroxidase(s). The combination of the dotted and broken line dependences could generate the observed killing response by adjustment of the H2O2 activation rates.

Consistent with the feasibility of electron diversion is the sensitization of cells to mode-one killing by CN (Fig. 9). In this instance, one might envisage that blockage of haem-dependent electron flow with CN- diverts electrons to activation, and, at higher H2O2 concentrations, to peroxidase(s) so as to generate the dose-response noted in Fig. 9.

Fig. 9.

Enhancement by KCN of mode-one killing by H2O2. Strain AB 1157 was grown aerobically and challenged in the presence of 3mM-KCN.

Fig. 9.

Enhancement by KCN of mode-one killing by H2O2. Strain AB 1157 was grown aerobically and challenged in the presence of 3mM-KCN.

A recent further observation in this regard is that, while cells lacking glutathione reductase (gor) or glutathione synthetase (gshB) are not abnormally sensitive to H2O2, a strain lacking NADH dehydrogenase (ndh) gives a dose-response for mode- one killing that looks very much like that of Fig. 9 regardless of whether CN is present or not. Such an effect would be expected if the activation of H2O2 competes with respiration functions for NADH. Blockage of NADH oxidation might then divert electrons to the activation reaction as well as to peroxidase functions.

We are hopeful that a genetic and biochemical analysis will serve ultimately to validate and/or modify the above general models for H2O2 activation and the mode- one killing dose-response. In the process, it is also hoped that we will begin to understand how other DNA-damaging agents that act through the generation of H2O2, OH- or other free radicals serve to damage DNA and induce protective responses. Such knowledge gained in bacteria such as E. coli cannot help but be applicable to higher organisms.

Another objective is to begin to dissect mode-two killing and the lesions involved. The inhibition of growth by mode-two killing doses suggests that perhaps lipid oxidation is involved; however, enhanced mutagenesis suggests the presence of DNA damage. Whatever the site(s) of damage, the observation of mode-two killing in the absence of a carbon source predicts the occurrence of a novel and potentially interesting activation reaction that does not require electron flow.

This research was supported by grant GM19020 from the National Institutes of Health, US Department of Health and Human Services.

Christman
,
M. F.
,
Morgan
,
R. W.
,
Jacobson
,
F. S.
&
Ames
,
B. N.
(
1985
).
Positive control of a regulon for defenses against oxidative stress and some heat-shock proteins in Salmonella typhimurium.
Cell
41
,
753
762
.
Cunningham
,
R. P.
&
Weiss
,
B.
(
1985
).
Endonuclease III (nth) mutants of Escherichia coli.
Proc, natn. Acad. Sci. U.S A.
82
,
474
478
.
Demple
,
B.
&
Halbrook
,
J.
(
1983
).
Inducible repair of oxidative DNA damage in Escherichia coli.
Nature, Lond.
304
,
466
468
.
Demple
,
B.
,
Halbrook
,
J.
&
Linn
,
S.
(
1983
).
E. colixth mutants are hypersensitive to hydrogen peroxide
.
J. Bad.
153
,
1079
1082
.
Demple
,
B.
&
Linn
,
S.
(
1980
).
DNA N-glycosylases and UV repair
.
Nature, Lond.
287
,
203
208
.
Demple
,
B.
&
Linn
,
S.
(
1982
).
5,6-Saturated thymine lesions in DNA: Production by ultraviolet light or hydrogen peroxide
.
Nucl. Acids Res.
10
,
3781
3789
.
Imlay
,
J. A.
&
Linn
,
S.
(
1986
).
A bimodal pattern of killing of DNA repair-defective or anoxic Escherichia coli by hydrogen peroxide
.
J. Bact.
166
,
519
527
.
Morimyo
,
M.
(
1982
).
Anaerobic incubation enhances the colony formation of apolArecB strain of Escherichia coli K-12
.
J. Bact.
152
,
208
214
.
Shaw
,
D. J.
&
Guest
,
J. R.
(
1982
).
Amplification and product identification of the fnr gene of Escherichia coli K-12. J’, gen. Microbiol.
128
,
2221
2228
.
Walker
,
G.
(
1985
).
Inducible DNA repair systems. A
.
Rev. Biochem.
54
,
425
457
.