Vol. 166, No. 2

Bimodal Pattern of Killing of DNA-Repair-Defective or AnoxicallyGrown Escherichia coli by Hydrogen Peroxide

JAMES A. IMLAY AND STUART LINN*

Department ofBiochemistry, University of California, Berkeley, California 94720

Received 25 November 1985/Accepted 19 February 1986

Two modes of killing of Escherichia coli K-12 by hydrogen peroxide can be distinguished. Mode-one killingwas maximal with hydrogen peroxide at a concentration of 1 to 2 mM. At higher concentrations the killing ratewas approximately half maximal and was independent of H202 concentration but first order with respect toexposure time. Mode-one killing required active metabolism during the H202 challenge, and it resulted insfiA-independent filamentation of both cells which survived and those which were killed by the challenge. Thismode of killing was enhanced in xth, polA, recA, and recB strains and was accelerated in all strains by an

unidentified, anoxia-induced cell function. A strain carrying both xth and recA mutations appeared to undergospontaneous mode-one killing only under aerobic conditions. Mode-one killing appeared to result from DNAdamage which normally occurs at a low, nonlethal level during aerobic growth. Mode-two killing occurred athigher doses of H202 and exhibited a multihit dependense on both H202 concentration and exposure time.Mode-two killing did not require active metabolism, and killed cells did not filament, although survivorsdemonstrated a dose-dependent growth lag. Strains with DNA-repair defects were not especially susceptible tomode-two killing.

Active oxygen species may be produced by incompletereduction of 02 during respiration; such species are potentoxidants of lipids, proteins, and nucleic acids (13). It hasbeen proposed that rates of age-related deterioration offunction in higher organisms correlate with rates of respira-tion (1). In bacteria, oxidative stresses mediated byexogenous hydrogen peroxide (20, 22), superoxide (15, 26),or hyperbaric oxygen (14) can be lethal or mutagenic.

Escherichia coli produces scavengers of activated oxygenspecies, including catalase, superoxide dismutase, and per-oxidase. Certain mutants which are defective in catalase andperoxidase are oxygen intolerant (H. M. Hassan, Fed. Proc.35:130, 1976), but they can become oxygen tolerant with theacquisition of additional mutations which diminish respira-tion rates (H. M. Hassan, Fed. Proc. 35:130, 1976). Muta-tions in recA (7), polA (2), or xth (12) confer hypersensitivityto H202; furthermore, some polA (3) and polA recB (27)strains are intolerant of aerobiosis. These observations sug-

gest that oxidative respiration might pose a threat to DNAand that DNA repair processes are critical in protectingagainst respiration-borne DNA-damaging agents.

Inducible responses to oxidative stress in enterobacteria(11, 33) result in a diminished sensitivity to oxidative chal-lenge. These responses are accompanied by an increasedtiter of scavenging enzymes; it is not known, however,whether an increase in repair capacity also ensues. In fact,the specific killing lesions in wild-type cells are unknown.This paper reports some of the phenomenology of killing ofE. coli K-12 by H202 and suggests that DNA is the site oflethal damage at lower concentrations of the agent.

MATERIALS AND METHODSBacterial strains are listed in Table 1. Hydrogen peroxide

was purchased as a 30% aqueous solution from Mallinckrodt;chloramphenicol and tetracycline were from Sigma Chemical

* Corresponding author.

Co.; beef-liver catalase was a 20-mg/ml, 65,000-U/mgsuspension from Boehringer-Mannheim Biochemicals.

Unless otherwise indicated, liquid cultures were in Kmedium (1% glucose, 1% Casamino Acids, 1 ptg of thiaminehydrochloride per ml, 1 mM MgSO4- 7H20, 0.1 mM CaCI2,M9 salts [24]). Anoxic growth was achieved by incubation of4 x 107 to 8 x 107 CFU/ml in 10-ml cultures in 30-ml testtubes with 150 rpm of shaking in a New Brunswick modelG76 shaking water bath. Aerated growth was by a similarincubation of dilute ('107 CFU/ml) cells or by incubation ofdenser 10-ml cultures in 125-ml Erlenmeyer flasks with 195rpm of shaking. Oxygen concentrations were measured withan Orion Research model 97-08 oxygen electrode.

Unless otherwise noted, cells were challenged with H202at a density of 1 x 107 to 4 x 107 CFU/ml in 1 ml of Kmedium for 15 min at 37°C with 150 rpm of shaking. To studykilling, the challenge was terminated either by dilution intoM9 salts or by the addition of 2 ,ug of catalase. Cells were

plated in top agar onto L agar plates (1% bactotryptone,0.5% yeast extract, 1% NaCl, 0.1% glucose, 2.5 mM CaCl2,1% agar), and colonies were counted after 24 to 48 h. Tostudy postchallenge cell growth or division, H202 was elim-inated by the addition of catalase. For pretreatment, H202was added (final concentration, 30 ,uM) to 2 x 107 CFU ofaerated cells per ml, and aeration was continued for 70 minbefore the challenge. Cell filamentation in liquid K mediumor on the surface of plates was visualized by light micros-copy.Anaerobic growth on plates was by incubation in a Gas-

Pak vessel (BBL Microbiology Systems), and P1-mediatedtransduction of the recA56 allele was done as describedpreviously (9).Phage inactivation assays were done as described previ-

ously (11); briefly, phage were treated with 7.5 mM H202 inthe presence of 10 ,uM CUSO4 and then used to infectlog-phase cultures at a multiplicity of infection of 10-5. Therates of inactivation were linear and were calculated by a

519

JOURNAL OF BACTERIOLOGY, May 1986, p. 519-5270021-9193/86/050519-09$02.00/0Copyright © 1986, American Society for Microbiology

on June 27, 2018 by guesthttp://jb.asm

.org/D

ownloaded from

520 IMLAY AND LINN

TABLE 1. Strains used in this study

Genotype

F- thr-l leuB6 proA2 his-4 thi-J argE2 lacYl galK2 rpsL supE44 ara-14 xyl-15 mtl-l tsx-33As AB1157 plus xth-lz.(xthA-pncA) relAl spoT-I thi-JF- gal his thi endAAs JC4583 plus recA56Su- recB21 trpE9829 leu his-318 str-321 lac Bl1As AB1157 plus recB21 recC22 sbcBJ5Hfr P045 srlC300::TnJO recA56 thr-30 ilv-318 rpsE300As BW9091 plus sr/C300::TnJO recA56As AB1157 plus sr/C300::TnlO recA56As BW9091 plus sr/C300::TnJOAs AB1157 plus sr/C300::TnJOF+ A(gal-bio) thi-J relAl spoTIAs CM4722 plus polAJ TnlOAs CM4722 plus ApolA KanrF- rha lacZ thyA deo polAJ (W3110 origin)Prototrophic K-12 strainAs TA4131 plus oxyR2btuB::TnJO araD139 (argF-/ac)205flbB5301 non-9 gyrA219 relAl rpsLO5 metE70As RK4936 plus Aoxy-3As AB1157 plus nth::Tn5F- endA thyA deoA or B gal xyl thiAs AB3049 plus uvrA6As AB3049 plus uvrB5As AB3049 plus uvrC34As AB1157 plus ut'rD::Tn5HfrH thi rep-5lac trp pho supC(Ts) str malAs SC122 plus htpRthr leu his ura str tif tsl sfiA::Tn5

Source or reference

23B. Weiss66A. J. ClarkA. J. Clark9This paperThis paperThis paperThis paperC. JoyceC. JoyceC. Joyce25888810P. Howard-FlandersP. Howard-FlandersP. Howard-FlandersP. Howard-Flanders295

2828S. Gottesman

linear regression analysis of data from at least four timepoints.

RESULTSTwo modes of killing by hydrogen peroxide are kinetically

distinguishable, one of which is enhanced in several DNA-repair mutants. The survival of aerated E. coli AB1157 afterexposure for 15 min to various concentrations of H202 isshown in Fig. 1A. The broad shoulder of resistance was a

function of total exposure to H202 (H202 concentration xtime) at concentrations up to at least 100 mM. The slight dipin survival below 5 mM H202 is reproducible and is dis-cussed in detail below.

In contrast to DNA-repair-proficient cells, strain BW9091(xth), which lacks exonuclease III, exhibited unusual sur-vival behavior under these challenge conditions (Fig. 1A).An extreme sensitivity to H202 at concentrations below 3mM was partly reversed at moderate concentrations, andthis intermediate survival level was ultimately exceeded athigh H202 concentrations. A simple description of thisresponse would define two regions of killing with an inter-vening zone of partial resistance; the exonuclease deficiencywould exacerbate only the first region of killing. Killing atthe lower concentrations of H202 will be designated mode-one killing, whereas killing occurring at the higher,postshoulder concentrations will be designated mode-twokilling. As noted below, these two modes of killing each havedistinct characteristics.

Survival curves of the recA strain, JC4588, closely resem-ble those of the xth strain (Fig. 1A and B). The heightenedsusceptibility of this strain to mode-one killing is likely dueto the deficiency in recombinational DNA repair rather thanto the absence of other recA-mediated functions, because a

recB strain was also rapidly killed (Fig. 1B). Consistent withthis conjecture, an sbcB mutation in a recB recC strain,which restores recombination by activation of the recFpathway (16), abolished the special sensitivity of the recBmutant (Fig. iB).polAl strains, which lack the polymerase and 3'--5'

exonuclease activities of DNA polymerase I, were ex-tremely sensitive to H202 under these challenge conditions(Fig. 1A). This sensitivity resulted from a greatly increasedrate of mode-one killing, the kinetics of which could only beobserved with very short exposures (Fig. 1C). Thus, the lossof exonuclease III, DNA polymerase I, or DNA recombina-tion ability confers vulnerability to mode-one killing.Other strains with abnormal DNA repair functions were

indistinguishable from repair-proficient strains with regard toboth modes of H202 killing. These include strains carryingmutations in uvrA, uvrB, or uvrC loci, which collectivelyencode the uvrABC endonuclease; uvrD and rep, which codefor DNA helicases; htpR, which codes for a regulator of theheat-shock response; and, surprisingly, nth, which codes forendonuclease III, a thymine glycol-urea glycosylase and APendonuclease.An obvious question is whether the reduced sensitivity at

intermediate H202 concentrations is dependent upon theconcentration of H202 or upon the total exposure (concen-tration of H202 times duration of exposure). To answer thisquestion, time courses of killing were monitored at H202concentrations between 0.15 and 25 mM H202 (Fig. 2). In allcases the rate of killing was initially first order with respectto time. At the lower concentrations the rate of killingeventually subsided, possibly due to detoxification of themedium by cellular catalase. Significantly, at longer timesthere was no increased survival which could be analogous to

Strain

AB1157BW9091BW9101JC4583JC4588JC4695JC7623JC10240JI100JI101J1102J1103CM4722CM5409CJ261JG138TA4131TA4110RK4936TA4112RPC37AB3049NH4905AB3062NH5132GW3703PM5SC122K165GC4540

J. BACTERIOL.

on June 27, 2018 by guesthttp://jb.asm

.org/D

ownloaded from

BIMODAL E. COLI KILLING BY H202 521

lo10o- 90-second.2 Challengeo) ABI 157co (x-xLi.

x

:310-2 x~BOO1\C4S

x

10-3- JC469* (r.cB)

I ~~~~JC4586(recA)

1CM5409

041(polA1)x I I0 10 20 30 40 50 0 10 20 30 40 0 10 20

H202 Concentration (mM)FIG., 1. Survival of aerated cells after challenge with H202. Appropriate repair-proficient strains isogenic to these mutants were tested and

confirmed to display profiles similar to that of AB1157. (A and B) Two separate experiments with a 15-min challenge. (C) A 90-s challengeof CM5409 (polAl).

the protective effect of intermediate H202 concentrations.Instead, at low concentrations the initial killing rate wasimmediately as much as twofold higher than that at thehigher concentrations (Fig. 2, inset). Thus, the protection atintermediate H202 concentrations is not related to the dura-tion of exposure; it is a function of H202 concehtration only.The fact that the initial rate of killing was independent ofH202 concentration at intermediate concentrations suggeststhat a step leading,to the production of DNA damage wassaturated. (Extension of any challenge to long exposuretimes eventually led to the appearance of mode-two killing,since the shoulder of resistance to that form of killing wasultimately exceeded [Fig. 2, 25 mM curve].)Time courses similar to those in Fig. 2 were observed for

mode-one killing of recombination-deficient cells and, de-spite a significant reduction of mode-one killing-rate con-stants, for the low-level mode-one killing of DNA-repair-proficient AB1157 cells (data not shown).Anoxia induces sensitivity to mode-one killing. Mode-one

killing of both repair-deficient and repair-proficient strainscan be markedly amplified by prechallenge cell growth inanoxic medium. Under conditions of minimal agitation, cellcultures abruptly exhausted measurable dissolved oxygen asthey reached densities of 1 x 107 to 5 x 107 CFU/ml. Within45 min of oxygen depletion, the sensitivity of the repair-proficient AB1157 bacteria to mode-one killing by H202increased by more than an order of magnitude (Fig. 3), whilevulnerability to mode-two killing increased only marginally ifat all. All cells were diluted into aerated medium for the

challenge in these experiments, ensuring that oxygen tensionand cell density were invariant among samples during thechallenge procedure. Similar sensitivity occurred, however,when the challenge or subsequent plating was performed inthe absence of oxygen.Anoxic enhancement of cell death was also observed with

xth cells (Fig. 3), as well as with recA and recB mutants (datanot shown). Detailed analysis of the killing rates shows thatthe two sensitizing factors, anoxia and DNA repair defi-ciency, had multiplicative (synergistic) rather than additiveeffects upon the rates of killing.A reversal of the anoxic sensitization to H202 could be

accomplished by cell dilution into oxygen-saturated mediumand subsequent aeration (Fig. 4A). Under these conditionsresistance to mode-one killing was largely achieved within 30min. Remnarkably, this reversal was not blocked by thepresence of the inhibitor of protein synthesis, chloramphen-icol (Fig. 4A). The sensitization which ensued upon shiftingwell-aerated cells to low-oxygen conditions was inhibited bychloramphenicol, however (Fig. 4B). Taken together, theseexperiments showed that anoxia resulted in the synthesis ofproteins that enhanced H202-mediated DNA damage.Only actively metabolizing cells are subject to mode-one

killing. Mode-one killing exhibited an apparent requirementfor metabolic activity during the challenge. xth, recA, recB,and polAl strains and anoxic repair-proficient cells wereexempted from mode-one killing when they were held in M9salts for 80 min before being challenged. Mode-two killingwas unaffected by such a starvation regime.

VOL. 166, 1986

on June 27, 2018 by guesthttp://jb.asm

.org/D

ownloaded from

522 IMLAY AND LINN

0.15mM

1.25mM

Initial Killing Rate.'\ vs [H202)

It% 0- .

I0 5 10 15 20 25

H202 Concentration (mM)

63

Duration of Challenge (minutes)FIG. 2. Kinetics of killing of strain BW9091 by H202. Aerated cells (107 CFU/ml) were challenged with H202 as indicated; at the indicated

times, samples were diluted in M9 salts containing 2 ,ug of catalase per ml before plating. (Inset) Initial killing rate as a function of H202concentration. The H202 concentrations which gave the maximal killing rate were 0.75 and 1.25 mM. Not all of the killing rates plotted in theinset are shown in detail as time courses in the main figure due to space limitations.

Mode-one killing of starved xth cells could be immediatelyrestored by the addition of glucose with the H202 (Fig. 5).Chloramphenicol did not block this reversal. Thus, mode-one killing, but not mode-two killing, is dependent uponactive metabolism during exposure to the H202.Growth delay and filantentation in response to H202. The

two modes of killing by H202 correlate with distinguishablebehavior subsequent to challenge. Mode-two-killed cellsremained permanently unit-sized on plates and in liquidculture: they were unable to elongate. By contrast, mode-one-killed cells, i.e., cells which were killed by low concen-trations of H202 while metabolically active, underwent ex-tensive filamentation but did not septate. This filamentousresponse was seen in repair-proficient, repair-defective, andanoxic cells; it was also observed in recA and sfiA mutants.

Survivors of H202 exposure displayed two phases ofrecovery. A growth lag was initially seen which was linearlydependent upon H202 concentration and time of exposure; itwas observed to extend to at least 7 h in cells exposed to 60mM H202 for 15 min. The growth lag was followed by aperiod of filamentation of about 90 min. As with the initial

rate of mode-one killing (Fig. 2, inset), the period offilamentation was independent of peroxide concentrationabove 5 mM but was somewhat greater at lower concentra-tions. An exception to the above pattern was noted with cellsthat were held in M9 salts before and during the challenge.These cells, which were not prone to mode-one killing,exhibited a dose-dependent growth lag after plating but didnot filament prior to septation. A growth delay response toamino acid starvation and near-UV irradiation is associatedwith the relA locus (18); However, relA mutants exhibitednormal growth delay and filamentation subsequent to H202exposure.These data collectively suggest that H202 produces dam-

age in a dose-dependent manner which is repaired during agrowth lag. Failure to complete this repair may lead tomode-two killing. Successful completion of this repair al-lowed reinitiation of cell growth but not of septation. Duringthis filamentous phase, repair of a separate damage whichblocked septation (presumably DNA damage) occurred.Failure to repair the latter damage resulted in mode-onekilling.

100

10-1

cu

m 0-cOC

._

L.

cn

co-

0.9 -

0.6 -

z_ 0.3-

0.0

0 9 12 15

J. BACTERIOL.

on June 27, 2018 by guesthttp://jb.asm

.org/D

ownloaded from

BIMODAL E. COLI KILLING BY H2O2 523

Mode-one killing lesions probably occur more frequently atlow hydrogen peroxide concentrations. The resistance tomode-one killing seen at intermediate H202 concentrationswas not blocked by the presence of chloramphenicol duringthe challenge and was observed even with a 90-s challenge.(Fig. 1C). Therefore, the induction of scavenger activities isan unlikely explanation for the resistance. An alternativeexplanation might be the induction of a DNA repair systemafter the challenge. To explore this possibility, xth cells wereexposed to 10 mM H202 for 6 min and then immediatelydiluted into media containing a range of H202 concentrationsbetween 0.75 and 15 mM (Fig. 6). The initial exposure to 10mM H202 produced less killing than did a concomitant 2.5mM exposure, indicating that the protection mechanism was

acting at 10 mM H202. Subsequent dilution of these cells intothe lower H202 concentration accelerated the killing rate tothat observed with unprotected cells. Hence, the protectiveeffect of higher concentrations of H202 appears to be rapidlyreversible. Such reversibility is incompatible with the induc-

100 r

Aerated

AB1157(xth+)

AnoxicABI 157 a

10-

0

1012H202 ConAeratedCD ~~BWOQO1

c ~~~~~(xth)

10-3

s Anoxic

10-4

io-50 10 20 80

H20 Concentration (mM)

FIG. 3. Effects of anoxia and exonuclease III deficiency on H202-sensitivity. Aerated and anoxic cells were cultured as described in

Materials and Methods, before being challenged.

tion ofDNA repair systems but is compatible with a reducedrate of damage at the higher H202 concentrations.

Sensitivity of an xth recA double mutant to aerobiosis. xthand recA mutations each confer sensitivity to mode-onekilling, yet these functions are considered to participate indifferent aspects of DNA metabolism. To test whether thesensitizations produced by these mutations are additive, therecA56 mutation was transduced into an xth strain. (Selec-tion was for a tetracycline-resistant activity which wasclosely linked to the recA allele in the JC10240 donor strain.)All recA-deficient transductants were slow growing relativeto recA-proficient counterparts, and their subsequent platingefficiency on L agar plates was usually less than 1%.Microscopic examination of the xth recA transductantsrevealed excessive filamentation in aerated liquid mediumand on rich or minimal plates. This filamentation could beentirely suppressed, however, by liquid culturing underanoxic conditions, and a 100% plating efficiency was re-stored when plates were incubated anaerobically. Killing byH202 could then be assayed if aerobically challenged cellswere plated anaerobically; in this way, it was found that lessthan 0.1% of the cells survived a 90-s challenge with 1.25mM H202. Thus, the sensitizing effects of xth and recAmutations are at least additive.

Effect of the oxyR regulon on the two killing modes. E. coli(11) and Salmonella typhimurium (33) demonstrate inducibleresistance toward H202-mediated killing. In E. coli, bothmodes of killing fell within the domain of this protectiveeffect; the sensitivity to mode-one killing was reduced, andthe shoulder of resistance to mode-two death was broadened(Fig. 7A). Resistance could be induced by H202 concentra-tions ranging from 30 to 500 ,uM; this protective effect couldbe seen within 15 min of and as long as 150 min after theinitial exposure. Optimal resistance was produced within 80to 100 min of the addition of 100 ,uM H202 to 5 x 107 CFUof AB1157 per ml. (Because cellular catalases detoxify themedium during pretreatment, net exposure and cell responseare functions of cell density.) As reported by Demple andHalbrook (11), induced resistance is completely blocked bythe presence of chloramphenicol during the pretreatment.

Christman et al. (8) have identified a regulon under thecontrol of the oxyR locus, which apparently encodes apositive effector of this response; gene products of theregulon are overproduced after exposure to inducing levelsof H202. Among the gene products overproduced are thescavengers of active oxygen species, catalase, superoxidedismutase, and peroxidase. It is not known whether DNArepair functions are also regulated by the oxyR function;however, we have not observed increased levels of endonu-clease III (10), endonuclease IV (21), or exonuclease III (12)after activation of the oxyR regulon (unpublished data).A deletion of the oxyR locus results in a strain that is

deficient in the enhanced synthesis of many H202-inducibleproteins (8). The deletion mutant retained the aeration-associated resistance to mode-one killing, as-well as thepeculiarly lowered sensitivity to mode-one killing at inter-mediate H202 concentrations (data not shown). These ob-servations, in agreement with those from the chloramphen-icol experiments cited above, confirm that oxyR isuninvolved in these resistance phenomena.

Christman et al. also observed that this deletion strain issensitive to peroxide in simple disk inhibition assays; wefound, however, that it was only slightly more sensitive thanits parent to aerobic liquid-culture challenge (Fig. 7B),implying that the oxyR gene product is not particularly activein aerobic liquid culturing conditions in the absence of

VOL. 166, 1986

on June 27, 2018 by guesthttp://jb.asm

.org/D

ownloaded from

524 IMLAY AND LINN

c0

cmci.L

._

0)

c

:3

0 10 20 30

H202 Concentration (mM) H202 Concentration (mM)

FIG. 4. Effect of chloramphenicol upon shift to or from anoxia-linked H202 sensitivity. (A) Anoxically grown AB1157 was challengedimmediately (0), after 45 min of anoxic growth in 100 ,ug of chloramphenicol per ml (A), after dilution and 45 min of subsequent aeration (0),or after dilution and 45 min of aeration in 100 ,ug of chloramphenicol per ml (x). (B) Aerated AB1157 was shifted to anoxic growth conditionsas described in Materials and Methods and incubated anoxically in the presence (x) or absence (0) of 100 ±g of chloramphenicol for 45 minbefore challenge. As a control, anoxically grown AB1157 was incubated anoxically in 100 p.g of chloramphenicol per ml for 45 min beforechallenge (A).

preexposure to exogenous H202. The disk inhibition proce-dure may produce a more notable phenotype because itconfronts bacteria with a temporal gradient of H202, whichmay in turn allow induced expression of the oxyR regulon innormal strains. In any case, deletion of the oxyR locus didfully prevent any induced resistance by pretreatment (Fig.7B).

Reactivation of H202-treated phage correlates with DNApolymerase I proficiency. The ability of a host cell to besuccessfully infected by damaged phage is one criterion usedto measure DNA repair capacity. When phage inactivationwith H202 was mediated by the cupric cation, inactivationwas first order with regard to exposure time, indicating thata single hit might suffice to inactivate the phage. Comparisonof inactivation rates of P1 and phages showed the propor-tionality of inactivation rate with genome size that is ex-pected from inactivation via DNA damage. However, we didnot observe any significant differences in host-cell reactiva-tion ability between cells grown with and without inducingpretreatment and between cells that are constitutive andnormal in oxyR expression. Such a result does not entirelyrule out the possibility that oxyR mediates an induction ofDNA repair functions, however, as phage DNA lesions

might not be identical to those generated intracellularly byH202, and, moreover, such lesions might not be accessibleto repair enzymes. Phage inactivation rates might also beinsensitive to quantitative adjustments in DNA repair capac-ity, especially since it is unlikely that a single phage genomequantitatively taxes the entire DNA repair capacity of thecell.

Strains JC4588 (recA) and BW9091 (xth), while sensitiveto killing by H202, exhibited normal phage reactivation.However, inactivation measured in the polAl hosts CM5409and JG138 occurred at a rate twice that in other hosts; thus,phage DNA damaged by H202 is subject to a type of DNArepair that is mediated by DNA polymerase I.

DISCUSSIONAnoxic and DNA-repair-deficient cells were particularly

susceptible to mode-one killing by H202. Mode-two killingwas prominent in aerated, repair-proficient cells and occursat higher H202 concentrations. These two modes of killingare distinguishable by their kinetics, requirements for activemetabolism, and postchallenge growth behavior.Mode-two killing has been observed frequently in re-

sponse to both H202 treatment and near-UV irradiation. The

010

C 10 10

C)

coLL0)c

>10-2

cnC 1_

i0-

J. BACTERIOL.

on June 27, 2018 by guesthttp://jb.asm

.org/D

ownloaded from

BIMODAL E. COLI KILLING BY H202 525

critical lesion has not been identified; however, many cellu-lar targets of chemical oxidants exist, including lipids, pro-teins, and nucleic acids, and all suffer some damage at doseswhich give moderate levels of mode-two killing (18). Thebroadly shouldered killing curve may indicate gradualtitration of a tolerance for damage, due either to inactivationof multiple targets or to saturation of some repair capacity.

Sensitivity to mode-two killing can be modified in severalways. Killing curves for slow-growing strains display abroadened shoulder, as does the killing curve of any strainexposed to extended prechallenge starvation or cyanide.Sensitivity to mode-two killing can also be lessened byH202-mediated induction of the oxyR regulon, due at least inpart to the induction of increased levels of scavengingenzymes (8). Indeed, coincubation of naive cells with in-duced cells during challenge allowed sharing of the resist-ance, presumably by medium detoxification (unpublisheddata).To our knowledge, mode-one death has not previously

been explicitly defined. The approximately zero-order de-pendence of cell death upon H202 concentration in mode-one killing is consistent with saturation by H202 of some stepin the damage-producing pathway. Such a step would not beinvolved in production of mode-two killing lesions, since therate of production of the latter lesions is proportional toH202 concentration to at least 100 mM. An obvious possi-bility is that mode-one damage is internal, mode-two damageis external, and transport is limiting for the former. Objec-tions to this model include the report (31) that H202 entry isdiffusion limited.

Alternatively, H202 may not be the ultimate oxidant inmode-one killing, and the conversion of H202 to the toxicspecies might be saturable at H202 concentrations near 1mM. Single-electron reduction of H202 generates hydroxyl

c0

cn

IL

0)C

10;io

0 5 10 20

Duration of 1.25mM Challenge (Minutes)FIG. 5. Effect of withholding a carbon source on mode-one

killing. BW9091 (xth) cells were grown as indicated in Material andMethods or were starved in M9 salts for 120 min. They were thenchallenged with 1.25 mM H202; where indicated, chloramphenicol,glucose, or both were added with the H202 to final concentrations of100 ,ug/ml and 1%, respectively. At the time shown, samples were

diluted into M9 salts containing 2 ,ug of catalase per ml and plated.

100

1

I- ~ - ~ -- - 6' Challenge

CY)

102~~~~~~~~~~~~~~~./ ~ ~~~10mM x 6'Survivore

1o-2I

0 3 6 9 12 15

H202 Concentration (mM)

FIG. 6. Exposure to high H202 concentrations does not blockincreased sensitivity to low H202 concentrations. BW9091 waschallenged by the H202 concentrations indicated; at 6 min, sampleswere diluted for plating. Also at 6 min, samples of the cellschallenged with 10 mM H202 were diluted 100-fold into a range ofH202 concentrations and incubated for another 6 min before plating.

radical, an unstable species which mediates DNA in vitrostrand breaking and thymine adduct formation (4, 19, 30).This hydroxyl radical may act directly or indirectly toproduce mode-one killing. Starvation would then deplete thecell of available reducing equivalents, thus preventing theproduction of toxic damage. The addition of glucose wouldreplenish these equivalents and restore mode-one killing.The enzymatic machinery responsible for this putative

univalent reduction of H202 is unknown. However, suchmachinery may be enhanced under conditions of anoxia,resulting in the chloramphenicol-sensitive increase in sensi-tivity to mode-one killing. If the responsible anoxia-inducedenzyme(s) can be identified, the validity of the activationmodel may be tested.

It is established that components for alternativerespirative pathways are synthesized in response to condi-tions of low oxygen concentration (17), including pathwayswhich utilize nitrate, fumarate, or nitrite as terminal electronacceptors; moreover, a secondary oxygen-reductive chain isalso induced (32). Low residual activities of these systemshave been reported under incubation conditions which ap-proximate our fully aerated cultures, so slight mode-onesensitivity under those conditions might even be ascribableto their low-level function. Experiments are underway totest these possibilities.At H202 concentrations above 1.25 mM, the mode-one

killing rate was lessened by a factor of two; this effect is animportant parameter in the determination of viability. Itappears that this protective effect occurred by adjustment ofthe production, rather than the resolution, of lesions. Per-

VOL. 166, 1986

on June 27, 2018 by guesthttp://jb.asm

.org/D

ownloaded from

526 IMLAY AND LINN

0 0PretrntondTA4112

LLco 1 0-2

*~10-2

CD~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~l

i0-3 Naive BW90911-

TA41I1(OxyR)

10o 10 20 30 40 50 60 70 0 10 20 30 40 50

H202 Concentration (mM) H202 Concentration (mM)

FIG. 7. oxyR-directed adaptation to H202. (A) A culture of BW9091 was pretreated with 30 ,uM H202 for 70 min immediately before beingchallenged as indicated. (B) Cells were grown with aeration before being challenged; pretreatment was as described above.

haps the amount of the toxic species was reduced at highconcentrations of H202, either because it was directlyquenched or because its production was suppressed. Forexample, hydroxyl (or other) radicals might be quencheddirectly by hydrogen peroxide with the concomitant forma-tion of superoxide anion. Superoxide anion could be a lesstoxic species either because of its greater chemical stabilityor because of its vulnerability to scavenging by superoxidedismutase. The rate of generation of superoxide anion bythis model would be limited by the original reduction ofH202, which would be saturated above 1 mM. Alternatively,the univalent reduction of H202 might be partially sup-pressed by diversion of electrons to another process, forexample, to the divalent reduction of H202 by a peroxidasefunction. These models can be genetically tested.The oxygen-intolerant phenotype of the recA xth double

mutant demonstrates the importance of the H202 survivalstudies. Culturing in atmospheric oxygen is sufficient toblock colony formation and induce filamentation, an indica-tion that mode-one killing damage can be generated underroutine conditions of aerobic growth. Similar filamentousbehavior has been reported under aerobic but not anaerobicconditions for several other DNA repair mutants, including apolA 5'-33' exo(Ts) mutant (3) and a polAl recB doublemutant (27). The toxic agents present during aerobic growthmay be partially reduced oxygen species escaping from therespiration pathway. We hope that our studies with H202toxicity will ultimately help to identify the particular dam-aging agents mediated by aerobiosis, the mechanism of theirformation, and their sites of action.

ACKNOWLEDGMENTS

This work was supported by Public Health Service grantsGM19020 and P30 ES01896 from the National Institutes of Health andby a National Science Foundation graduate fellowship to J.A.I.

LITERATURE CITED1. Ames, B. N. 1983. Dietary carcinogens and anticarcinogens.

Science 221:1256-1264.2. Ananthaswamy, H. N., and A. Eisenstark. 1977. Repair of

hydrogen peroxide-induced single-strand breaks in Escherichiacoli deoxyribonucleic acid. J. Bacteriol. 130:187-191.

3. Boling, M., H. Adler, and W. Masker. 1984. Restoration ofviability to an Escherichia coli mutant deficient in the 5'- 3'exonuclease of DNA polymerase I. J. Bacteriol. 160:706-710.

4. Breimer, L. H., and T. Lindahl. 1985. Enzymatic excision ofDNA bases damaged by exposure to ionizing radiation oroxidizing agents. Mut. Res. 150:85-89.

5. Calendar, R., B. Lindquist, G. Sironi, and A. J. Clark. 1970.Characterization of Rep- mutants and their interaction with P2phage. Virology 40:72-83.

6. Capaldo-Kimball, F., and S. D. Barbour. 1971. Involvement ofrecombination genes in growth and viability of Escherichia coliK-12. J. Bacteriol. 106:204-212.

7. Carlsson, J., and V. S. Carpenter. 1980. The recA+ gene productis more important than catalase and superoxide dismutase inprotecting Escherichia coli against hydrogen peroxide toxicity.J. Bacteriol. 142:319-321.

8. Christman, M. F., R. W. Morgan, F. S. Jacobson, and B. N.Ames. 1985. Positive control of a regulon for defenses againstoxidative stress and some heat-shock proteins in Salmonellatyphimurium. Cell 41:753-762.

9. Csonka, L. N., and A. J. Clark. 1980. Construction of an Hfrstrain useful for transferring recA mutations between Esche-richia coli strains. J. Bacteriol. 143:529-530.

10. Cunningham, R. P., and B. Weiss. 1985. Endonuclease III (nth)mutants of Escherichia coli. Proc. Natl. Acad. Sci. USA 82:474-478.

11. Demple, B., and J. Halbrook. 1983. Inducible repair of oxidativeDNA damage in Escherichia coli. Nature (London)304:466-468.

12. Demple, B., J. Halbrook, and S. Linn. 1983. Escherichia coli xthmutants are hypersensitive to hydrogen peroxide. J. Bacteriol.153:1079-1082.

13. Fridovich, I. 1978. The biology of oxygen radicals. Science201:875-880.

J. BACTERIOL.

on June 27, 2018 by guesthttp://jb.asm

.org/D

ownloaded from

BIMODAL E. COLI KILLING BY H202 527

14. Gregory, E. M., and I. Fridovich. 1973. Induction of superoxidedismutase by molecular oxygen. J. Bacteriol. 114:543-548.

15. Gregory, E. M., F. J. Yost, Jr., and I. Fridovich. 1973. Super-oxide dismutases of Escherichia coli: intracellular localizationand functions. J. Bacteriol. 115:987-991.

16. Horii, Z., and A. J. Clark. 1973. Genetic analysis of the RecFpathway to genetic recombination in Escherichia coli K12:isolation and characterization of mutants. J. Mol. Biol. 80:327-340.

17. Ingledew, W. J., and R. K. Poole. 1984. The respiratory chainsof Escherichia coli. Microbiol. Rev. 48:222-271.

18. Jagger, J. 1983. Physiological effects of near-ultraviolet radia-tion on bacteria. Photochem. Photobiol. Rev. 7:1-75.

19. Lesko, S. A., R. J. Lorentzen, and P. O. P. Ts'o. 1980. Role ofsuperoxide in deoxyribonucleic acid strand scission. Biochem-istry 19:3023-3028.

20. Levin, D. E., M. Holistein, M. F. Christman, E. A. Schwiers, andB. N. Ames. 1982. A new Salmonella tester strain (TA102) withAT base pairs at the site of mutation detects oxidativemutagens. Proc. Nati. Acad. Sci. USA 79:7445-7449.

21. Ljungquist, S. 1977. A new endonuclease from Escherichia coliacting at apurinic sites in DNA. J. Biol. Chem. 252:2808-2814.

22. McCormick, J. P., J. R. Fischer, J. P. Pachlatko, and A.Eisenstark. 1976. Characterization of a cell-lethal product fromthe photooxidation of tryptophan: hydrogen peroxide. Science191:468-469.

23. Milcarek, C., and B. Weiss. 1972. Mutants of Escherichia coliwith altered deoxyribonucleases. I. Isolation and characteriza-tion of mutants for exonuclease III. J. Mol. Biol. 68:303-318.

24. Miller, J. H. 1972. Experiments in molecular genetics. Cold

Spring Harbor Laboratory, Cold Spring Harbor, N.Y.25. Monk, M., M. Peacey, and J. D. Gross. 1971. Repair of damage

induced by ultraviolet light in DNA polymerase-defective Esch-erichia coli cells. J. Mol. Biol. 58:623-630.

26. Moody, C. S., and H. M. Hassan. 1982. Mutagenicity of oxygenfree radicals. Proc. Natl. Acad. Sci. USA 79:2855-2859.

27. Morimyo, M. 1982. Anaerobic incubation enhances the colonyformation of a polA recB strain of Escherichia coli K-12. J.Bacteriol. 152:208-214.

28. Neidhardt, F. C., R. A. VanBogelen, and E. T. Lau. 1983.Molecular cloning and expression of a gene that controls thehigh-temperature regulon of Escherichia coli. J. Bacteriol. 153:597-603.

29. Pang, P. P., and G. C. Walker. 1983. Identification of the uvrDgene product of Salmonella typhimurium LT2. J. Bacteriol.153:1172-1179.

30. Repine, J. E., 0. W. Pfenninger, D. W. Talmage, E. M. Berger,and D. E. Pettiohn. 1981. Dimethylsulfoxide prevents DNAnicking mediated by ionizing radiation of iron/hydrogen perox-ide-generated hydroxyl radical. Proc. Natl. Acad. Sci. USA78:1001-1003.

31. Schwartz, C. E., J. Krall, L. Norton, K. McKay, D. Kay, andR. E. Lynch. 1983. Catalase and superoxide dismutase in Esch-erichia coli. Roles in resistance to killing by neutrophils. J. Biol.Chem. 258:6277-6281.

32. Shipp, W. S. 1972. Cytochromes of Escherichia coli. Arch.Biochem. Biophys. 150:459-472.

33. Winquist, L., U. Rannug, A. Rannug, and C. Ramel. 1984.Protection from toxic and mutagenic effects of H202 by catalaseinduction in Salmonella typhimurium. Mut. Res. 141:145-147.

VOL. 166, 1986

on June 27, 2018 by guesthttp://jb.asm

.org/D

ownloaded from