Advertisement

Are Respiratory Enzymes the Primary Sources of Intracellular Hydrogen Peroxide?*

  • Author Footnotes
    ‡ Recipient of National Institutes of Health Training Grant GM07283.
    Lauren Costa Seaver
    Footnotes
    ‡ Recipient of National Institutes of Health Training Grant GM07283.
    Affiliations
    Department of Microbiology, University of Illinois, Urbana, Illinois 61801
    Search for articles by this author
  • James A. Imlay
    Correspondence
    To whom correspondence should be addressed. Tel.: 217-333-5812; Fax: 217-244-6697;
    Affiliations
    Department of Microbiology, University of Illinois, Urbana, Illinois 61801
    Search for articles by this author
  • Author Footnotes
    * This study was supported by National Institutes of Health Grant GM49640. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked “advertisement” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
    ‡ Recipient of National Institutes of Health Training Grant GM07283.
Open AccessPublished:September 10, 2004DOI:https://doi.org/10.1074/jbc.M408754200
      Endogenous H2O2 is believed to be a source of chronic damage in aerobic organisms. To quantify H2O2 formation, we have generated strains of Escherichia colithat lack intracellular scavenging enzymes. The H2O2 that is formed within these mutants diffuses out into the medium, where it can be measured. We sought to test the prevailing hypothesis that this H2O2 is primarily generated by the autoxidation of redox enzymes within the respiratory chain. The rate of H2O2 production increased when oxygen levels were raised, confirming that H2O2 is formed by an adventitious chemical process. However, mutants that lacked NADH dehydrogenase II and fumarate reductase, the most oxidizable components of the respiratory chain in vitro, continued to form H2O2 at normal rates. NADH dehydrogenase II did generate substantial H2O2 when it was when overproduced or quinones were absent, forcing electrons to accumulate on the enzyme. Mutants that lacked both NADH dehydrogenases respired very slowly, as expected; however, these mutants showed no diminution of H2O2 excretion, suggesting that H2O2 is primarily formed by a source outside the respiratory chain. That source has not yet been identified. In respiring cells the rate of H2O2 production was ∼0.5% the rate of total oxygen consumption, with only modest changes when cells used different carbon sources.
      Superoxide ((O2.¯)) and hydrogen peroxide (H2O2) are partially reduced oxygen species that are substantially more reactive than is oxygen itself. Treatments that artificially generate large amounts of these species inside living cells, including the administration of hyperoxia, redox-cycling drugs, or authentic hydrogen peroxide, can disrupt metabolism and generate mutagenic or even lethal doses of DNA damage (
      • Imlay J.A.
      • Linn S.
      ,
      • Fridovich I.
      ). Even absent such forcing conditions, some(O2.¯)and H2O2 are generated when molecular oxygen chemically oxidizes reduced metabolites or enzymes inside the cell. For this reason, it has been suggested that endogenous(O2.¯)and H2O2 may create the cell damage that underlies important human pathologies, including those that derive from carcinogenesis and the aging process. A key question, then, is whether(O2.¯)and H2O2 are produced inside cells in doses sufficient to account for these disabilities. To answer that question we need to know both the rates at which they are formed and the doses that can cause toxicity.
      Evidence for the toxicity of endogenous oxidants was first obtained in Escherichia coli. In the mid-1980s, Carlioz and Touati created mutant strains of E. coli that lack cytosolic superoxide dismutase (
      • Carlioz A.
      • Touati D.
      ). These mutants exhibited defects in amino acid biosynthesis and tricarboxylic acid cycle function that led to the discovery that superoxide can inactivate dehydratases containing iron-sulfur clusters (
      • Kuo C.F.
      • Mashino T.
      • Fridovich I.
      ,
      • Flint D.H.
      • Tuminello J.F.
      • Emptage M.H.
      ,
      • Gardner P.R.
      • Fridovich I.
      ,
      • Gardner P.R.
      • Fridovich I.
      ,
      • Liochev S.I.
      • Fridovich I.
      ). Analogous phenotypes were subsequently discovered in superoxide dismutase-deficient yeast (
      • Chang E.C.
      • Crawford B.F.
      • Hong Z.
      • Bilinski T.
      • Kosman D.J.
      ,
      • van Loon A.P.G.M.
      • Pesold-Hurt B.
      • Schatz G.
      ,
      • Wallace M.A.
      • Liou L.-L.
      • Martins J.
      • Clement M.H.S.
      • Bailey S.
      • Longo V.D.
      • Valentine J.S.
      • Gralla E.B.
      ). Mouse mutants that lacked mitochondrial superoxide dismutase died within 10 days, an occurrence that may also have reflected deficiencies in tricarboxylic acid function (
      • Li Y.
      • Huang T.T.
      • Carlson E.J.
      • Melov S.
      • Ursell P.C.
      • Olson J.L.
      • Noble L.J.
      • Yoshimura M.P.
      • Berger C.
      • Chan P.H.
      ). These phenotypes demonstrated that endogenous superoxide production is sufficiently rapid to poison a cell that lacks superoxide dismutase activity. Subsequent dosimetric analysis indicated that wild-type E. coli synthesizes just enough superoxide dismutase to keep its superoxide-sensitive enzymes predominantly active (
      • Gort A.S.
      • Imlay J.A.
      ).
      Similarly, mutants of E. coli that cannot scavenge endogenous H2O2 also exhibit severe growth defects (
      • Seaver L.C.
      • Imlay J.A.
      ). Although the underlying injuries have not yet been identified, the phenotype confirms that H2O2 is also generated inside the cells at rates that require the presence of scavenging enzymes.
      How are the intracellular(O2.¯)and H2O2 formed? Molecular oxygen is a triplet species, meaning that it cannot remove more than one electron at a time from organic molecules (
      • Naqui A.
      • Chance B.
      ). This spin restriction has an important consequence: because molecular oxygen has a weak affinity for that first electron (Em = –0.16 V), there are few biomolecules that can spontaneously transfer an electron to it. This feature greatly diminishes the potential toxicity of oxygen. It also suggests that the respiratory chain is among the few plausible sites of(O2.¯)and H2O2 formation, because its flavins, quinones, and metal centers are all univalent electron carriers of sufficiently low potential to react with oxygen.
      Indeed, both(O2.¯)and H2O2 have been detected as trace by-products when mitochondrial or bacterial membrane vesicles respire in vitro (
      • Chance B.
      • Sies H.
      • Boveris A.
      ,
      • Imlay J.A.
      • Fridovich I.
      ,
      • Messner K.R.
      • Imlay J.A.
      ). When mitochondrial complex III is inhibited with antimycin, electrons are rapidly transferred from its Qo site to oxygen (
      • Turrens J.F.
      • Alexandre A.
      • Lehninger A.L.
      ). Superoxide is produced stoichiometrically, although it subsequently dismutates to produce H2O2 as a secondary product. This(O2.¯)is evidently formed on the outer aspect of the mitochondrial membrane so that it would contribute to oxidative stress in the inner membrane space but not within the matrix (
      • Han D.
      • Williams E.
      • Cadenas E.
      ). Mitochondrial complex I (NADH dehydrogenase) generates H2O2in vitro either when downstream inhibitors favor reverse electron flow from reduced ubiquinone or when rotenone blocks turnover of the NADH-reduced enzyme (
      • Lambert A.J.
      • Brand M.D.
      ). In both cases, the specific site of electron transfer to oxygen is unclear. Furthermore, it is difficult to know whether the same mechanisms of oxidation pertain when inhibitors are not present.
      In contrast to the mitochondrial studies, in vitro analyses of the E. coli respiratory chain have indicated that(O2.¯)and H2O2 are generated primarily by the autoxidation of reduced dehydrogenases (
      • Messner K.R.
      • Imlay J.A.
      ). NADH dehydrogenase I, the bacterial homologue of complex I, was not a major contributor, but the much simpler NADH dehydrogenase II reacted with oxygen at a substantial rate. Succinate. dehydrogenase produced scanty but still detectable(O2.¯). Its anaerobically synthesized isozyme, fumarate reductase, autoxidized rapidly when exposed to air. With each of these enzymes, the(O2.¯)and H2O2 are formed when molecular oxygen adventitiously oxidizes a reduced, solvent-exposed flavin. Although the spin restriction dictates that(O2.¯)must be the initial product of these electron transfer reactions, in most cases a second electron was transferred to(O2.¯)before it diffused out of the active sites of these enzymes, thereby producing more H2O2 than(O2.¯)as a product. The flavins lie in cytoplasmic domains of these membrane-bound proteins so that the(O2.¯)and H2O2 they generate would be formed inside the intact cell.
      These mechanisms were identified in vitro. Because enzyme behavior in vitro does not always represent what happens in vivo, it is important to test whether these enzymes are actually responsible for most(O2.¯)and H2O2 formation inside cells. Small amounts of H2O2 are indeed released by intact state 4 mitochondria in which electrons are backed up on the respiratory chain due to the absence of ADP as a F1-F0-ATPase substrate. When ADP was provided or the membrane potential was otherwise dissipated, the amount of H2O2 fell to undetectable levels (
      • Boveris A.
      • Cadenas E.
      ,
      • Boveris A.
      • Chance B.
      ). The fact that H2O2 production depends upon the respiratory state confirms that the respiratory chain is the likely site of state 4 mitochondrial H2O2 production but makes it difficult to estimate the amount of H2O2 that would be generated in actively respiring cells in vivo. Furthermore, recent calculations indicate that most H2O2 is scavenged by glutathione peroxidase before it diffuses out of the matrix (
      • Antunes F.
      • Cadenas E.
      ). Thus, these efflux experiments may have detected only the H2O2 that was produced on the outer aspect of the mitochondrial membrane.
      Until recently, it was not possible to measure the H2O2 formed inside living E. coli, as this bacterium contains an NADH peroxidase and two catalases and does not release the H2O2 that is generated in its cytosol (
      • Seaver L.C.
      • Imlay J.A.
      ). However, this problem can now be circumvented by the use of mutants that lack these enzymes (
      • Seaver L.C.
      • Imlay J.A.
      ). In the present study, the stepwise elimination of respiratory enzymes allowed us to appraise directly their contribution to the overall rate of H2O2 formation in vivo. Surprisingly, the results indicate that most H2O2 is formed elsewhere.

      MATERIALS AND METHODS

      Chemicals and Enzymes—Catalase, cytochrome c, horseradish peroxidase (type II), 4-hydroxybenzonic acid, hydrogen peroxide (30% w/v), isopropyl-β-d-thiogalactopyranoside, deamino-NADH, NADH, NADPH, o-nitrophenyl-β-d-galactopyranoside, o-dianisidine, plumbagin, potassium cyanide, riboflavin, copper/zinc superoxide dismutase, and uracil were purchased from Sigma-Aldrich. Total protein was measured with Coomassie protein reagent (Pierce). β-Mercaptoethanol, EDTA, and dimethyl sulfoxide were purchased from Fisher, and Amplex Red was purchased from Molecular Probes. Water for the buffers was purified with a Labconco Water Pro PS system using house deionized water as a feedstock. Ampicillin, chloramphenicol, and tetracycline were used at 100, 20, and 14 μg/ml, respectively.
      Strain Construction—The strains used in this study were derived from E. coli K-12 and are listed in Table I. Mutant strains were constructed by P1 transduction (
      • Miller J.H.
      ). The tetracycline-sensitive version of JI377 (LC106) was created by transducing into JI372 a Δ(katG17::Tn10)1 allele linked to argE86::Tn10, selecting for tetracycline resistance, and then replacing the argE mutation by transduction of argE+, with selection for arginine prototrophy. The resultant strain was confirmed to be hydroperoxidase I-deficient (a katG mutant) by enzyme assay (
      • Seaver L.C.
      • Imlay J.A.
      ). The menA null allele was created using the λ Red recombinase method (
      • Datsenko K.A.
      • Wanner B.L.
      ). Transduction of the menA allele was done by chloramphenicol selection on Luria broth with glucose (LBg)
      The abbreviations used are: LBg, Luria Broth with glucose; Frd, fumarate reductase; Hpx, hydroperoxidase-deficient; MES, 4-morpholineethanesulfonic acid.
      medium. Isolates were then screened by PCR for anaerobic growth on glycerol/fumarate (40 mm each). All frd and menA mutants were supplemented with uracil (1 mm) to bypass the requirement for function of the respiration-linked dihydroorotate dehydrogenase. menA mutants were all screened for hydroperoxidase I activity, because these genes can be co-transduced. The loss of the chloramphenicol marker in the menA deletion was achieved by using pCP20, which encodes FLP recombinase (
      • Datsenko K.A.
      • Wanner B.L.
      ). The frd deletion was co-transduced with zjd::Tn10 and screened for the inability to grow on minimal medium containing glycerol/fumarate (40 mm each). The ubiA420 point mutation was co-transduced with malE52::Tn10 and then screened for growth aerobically without 4-hydroxybenzoic acid (
      • Wallace B.J.
      • Young I.G.
      ). The undefined nuo point mutation was co-transduced with a zej-223::Tn10 (
      • Calhoun M.W.
      • Gennis R.B.
      ). The ndh, nuo, frd, and fre mutants were all screened using the appropriate enzyme assays that are described below. Transformation of all plasmids was done using the transformation and storage buffer protocol (
      • Chung C.T.
      • Miller R.H.
      ). Transductions and transformations into a hydroperoxidase-deficient (Hpx) background were done anaerobically. Phenotypic comparisons were always between isogenic strains.
      Table IE. coli strains and plasmids
      StrainRelevant genotypeSource or reference
      BW6165argE86::Tn10
      • Wanner B.L.
      KM34As AN384 plus malE52::Tn10Lab collection
      SP41fre::kan... zih-102::Tn10Lab collection
      MG1655F- wild-typeHoward Steinman
      JI301nuo... zej-223::Tn10Lab collection
      JI222Δ(frdABCD)8... zjd::Tn10
      • Imlay J.A.
      JI364Δ(katG17::Tn10)1
      • Seaver L.C.
      • Imlay J.A.
      JI372ΔahpCF′ kan::′ahpF Δ(katE12::Tn10)1
      • Seaver L.C.
      • Imlay J.A.
      JI377ΔahpCF′ kan::′ahpF Δ(katG17::Tn10)1 katE12::Tn10
      • Seaver L.C.
      • Imlay J.A.
      LC100As JI364 plus argE86::Tn10P1(BW6165) × JI364
      LC104As JI372 plus argE86::Tn10... Δ(katG17::Tn10)1P1(LC100) × JI372
      LC106ΔahpCF′ kan::′ahpḞ Δ(katG17::Tn10)1 Δ(katE12::Tn10)1P1(JI372) × LC104
      LC109As JI377 plus pMW01This study
      LC110As JI377 plus pfn3This study
      LC114As JI377 plus pBR322This study
      LC118As LC106 plus fre::kan... zih-102::Tn10P1(SP41) × LC106
      LC126As LC106 plus Δfrd(frdABCD)18... zjd::Tn10P1(JI222) × LC106
      LC128As JI377 plus pH3This study
      LC132ΔmenA::cmThis study
      LC138As LC106 plus nuo... zej-223::Tn10P1(JI301) × LC106
      LC141As LC106 plus pHS1-4This study
      LC145As LC106 plus ΔmenA::cmP1(LC132) × LC106
      LC147As LC145 plus ubiA420... malE52::Tn10P1(KM34) × LC145
      LC149As LC106 plus ubiA420... malE52::Tn10P1(KM34) × LC106
      LC150As LC147 plus pMW01This study
      LC156As LC138 plus ndh::cmP1(MW03) × LC138
      LC160As LC147 made cmsThis study
      LC165As LC160 plus ndh::cmP1(MW03) × LC160
      MW03ndh::cm pMW01Lab collection
      MW11As JI377 plus ndh::cmP1(MW03) × JI377
      GS022araD139 Δ(argF-lac)169 λ-flhD5301 fruA25 relA1 rpsL150 rbsR22 deoC1 λRS45 ϕ(katG::lacZ)Gisela Storz
      LC70As GS022 plus ΔahpCF′ kan::′ahpF
      • Seaver L.C.
      • Imlay J.A.
      LC133As GS022 plus pMW01This study
      LC134As LC70 plus pMW01This study
      LC137As GS022 plus pBR322This study
      pMW01pBR322 plus ndh+ insert
      • Calhoun M.W.
      • Gennis R.B.
      pfn3pJF119EH plus fre+ insert
      • Fieschi F.
      • Niviere V.
      • Frier C.
      • Decout J.L.
      • Fontecave M.
      pH3pBR322 plus frdABCD+ insert
      • Blaut M.
      • Whittaker K.
      • Valdovinos A.
      • Ackrell B.A.C.
      • Gunsalus R.P.
      • Cecchini G.
      pHS1-4pHC79 plus sodB+ insert
      • Sakamoto H.
      • Touati D.
      Cell Growth and Media—LB (pH 7) contained (per liter) 10 g of bactotryptone (Difco), 5 g of yeast extract (Difco), and 10 g of NaCl (
      • Miller J.H.
      ). When glucose was added to LB, its final concentration was 0.2%. To prevent the photochemical formation of hydrogen peroxide, LB medium was shielded from light and used within 24 h of its preparation. Minimal growth medium consisted of minimal A salts (
      • Miller J.H.
      ), 1 mm MgSO4·7H2O, 5 mg thiamine per liter, and a specified carbon source. Minimal glucose-casamino acids medium contained both glucose and casamino acids (2 g/liter each). l-amino acid supplements were used at a final concentration of 0.5 mm. Histidine was routinely added to minimal media for the growth of any derivative in the MG1655 background. This strain requires the addition of histidine to grow anaerobically (data not shown). To minimize the chemical production of hydrogen peroxide (
      • Seaver L.C.
      • Imlay J.A.
      ), the minimal media used in experiments (as noted) were prepared immediately before use and sterilized by filtration.
      Aerobic cultures were routinely grown in flasks at 37 °C unless noted otherwise. Anaerobic cultures were grown in a Coy chamber (Coy Laboratory Products, Inc.) under 85% N2, 10% H2, and 5% CO2. Optical densities of cultures were measured at 600 nm. For studies of growth on various carbon sources, cells were grown in minimal growth media containing 0.5 mm histidine, phenylalanine, tryptophan, and tyrosine (each) in order to foster better aerobic growth of the Hpx strain without providing an additional carbon source.
      J. Sobota and J. A. Imlay, unpublished data.
      Carbon sources tested were glucose (0.2%), casamino acids (2.0%), gluconate (0.2%), glycerol (40 mm), pyruvate (0.4%), lactate (0.4%), succinate (40 mm), and acetate (0.25%). Cells were first grown in standing overnight cultures because the resulting microaerobic conditions disfavored the outgrowth of suppressors of the Hpx strain. To grow a standing culture overnight with glycerol, nitrate (40 mm) was added. All overnight cultures were then diluted to an OD of 0.005 into fresh aerobic media and grown aerobically with vigorous shaking to log phase (OD of 0.1–0.2). The glycerol/nitrate overnight culture was washed twice in minimal salts before dilution into fresh glycerol medium lacking nitrate. The rates of respiration and H2O2 formation were measured from the same cell cultures within 15 min of each other.
      For β-galactosidase measurements, anaerobic overnights were grown in LB. Overnights were then diluted into anaerobic LB to an OD of 0.01. Cultures were grown to ∼0.1 OD anaerobically and then shifted to aerobic conditions for one to two generations. β-galactosidase activity was then measured (see below).
      Measurement of Respiration Rate—Respiration by whole cells was measured using a Rank Brothers digital model 10 oxygen sensor. Oxygen consumption was determined using 5 ml of log-phase cells in growth medium at 37 °C.
      Enzyme Assays—Each data set presented in this study was derived from at least three independent experiments. All assays were performed on log-phase cells at an OD600 of 0.1–0.2.
      β-Galactosidase—Cultures were centrifuged, and pellets were washed with 50 mm cold potassium Pi buffer (pH 7). Cells were resuspended in 50 mm potassium Pi buffer (pH 7) at one-tenth the culture volume and lysed by sonication. Cell debris was removed from crude extracts by centrifugation at 13,000 × g for 20 min. β-Galactosidase activity was assayed in a 1.2-ml reaction consisting of 0.2 ml of o-nitrophenyl-β-d-galactopyranoside (4 mg/ml), extract, and Z buffer (
      • Miller J.H.
      ) at 28 °C. Absorbance was monitored at 420 nm.
      Flavin Reductase—Cells were centrifuged and washed in cold 50 mm Tris-HCl buffer (pH 7.8) twice. Final resuspension was in one-hundredth of the original culture volume in cold 50 mm Tris-HCl buffer (pH 7.8). Cells were lysed by passage through a French pressure cell. Cell debris was removed by centrifugation at 13,000 × g for 20 min. Membranes were then removed by ultra-centrifugation (100,000 × g for 2 h). The soluble fraction was assayed for flavin reductase activity at room temperature by monitoring A340 in the presence of 0.25 mm NADPH and 15 μm riboflavin (
      • Woodmansee A.N.
      • Imlay J.A.
      ).
      Fumarate Reductase—Anaeobically grown log-phase cultures were incubated on ice for 10 min with 150 μg/ml chloramphenicol. This prevented the induction of succinate dehydrogenase during subsequent aerobic processing. Cells were then centrifuged at 10,000 rpm for 10 min, and pellets were washed twice with cold 50 mm potassium Pi buffer (pH 7.8). Cells were lysed by passage through a French pressure cell, and cell debris was removed. Inverted membrane vesicles were then isolated from the supernatant by ultra-centrifugation at 100,000 × g for 2 h. The inverted vesicles were resuspended in cold 50 mm potassium Pi buffer (pH 7.8) at 0.5% the original culture volume. The inverted vesicles were then assayed for succinate:plumbagin oxidoreductase activity at room temperature at A550 in the presence of 0.4 mm plumbagin, 3.3 mm potassium cyanide (pH 7.8), 20 mm succinate, and 10 μm cytochrome c (
      • Imlay J.A.
      ).
      NADH Dehydrogenase I—Cells were centrifuged and washed twice with cold 50 mm MES buffer containing 10% glycerol (pH 6). Final resuspension was in one-fortieth the original culture volume. Cells were lysed by passage through a French pressure cell, and cell debris was removed. Inverted membrane vesicles were then isolated from the supernatant by ultra-centrifugation at 100,000 × g for 2 h. The inverted vesicles was resuspended in cold 50 mm MES buffer with 10% glycerol (pH 6) at one-fortieth the original culture volume. Vesicles were assayed immediately, because NADH dehydrogenase I activity declines when membranes are stored on ice. The inverted vesicles were assayed for NADH dehydrogenase activity at A340 with 200 μm either deamino-NADH or NADH as the substrate. Deamino-NADH is a substrate for NdhI but not for NdhII (
      • Matsushita K.
      • Ohnishi T.
      • Kaback H.R.
      ). NdhI utilizes deamino-NADH and NADH with equal efficiency (
      • Calhoun M.W.
      • Gennis R.B.
      ).
      NADH Dehydrogenase II—Inverted membrane vesicles were isolated as described above. The inverted vesicles were resuspended in cold 50 mm potassium Pi buffer (pH 7.8) at one-fortieth the original culture volume. Vesicles were held on ice overnight to eliminate the activity of NdhI activity, which is unstable at this pH. The inverted vesicles were then assayed for NADH dehydrogenase II activity at A340 in the presence of 100 μm plumbagin, 3 mm potassium cyanide (pH 7.8), and 200 μm NADH (
      • Imlay J.A.
      ).
      Superoxide Dismutase—Cells were pelleted at 10,000 rpm for 10 min. Cells were then washed in cold 50 mm potassium Pi buffer (pH 7.8) and resuspended in one-hundredth the original culture volume in 50 mm potassium Pi buffer (pH 7.8) with 0.1 mm EDTA (pH 8). Extracts were made by passage through a French pressure cell, and debris was removed by centrifugation at 13,000 × g for 20 min. Superoxide dismutase was assayed using the xanthine oxidase/cytochrome c method (
      • McCord J.M.
      • Fridovich I.
      ).
      H2O2 Measurements—Cells were grown for at least four generations (to an OD of 0.1–0.3). This preculture typically was aerated; however, when experiments included strains lacking quinone (which cannot grow aerobically), the preculture medium was anaerobic. Log-phase cells were immediately pelleted by room temperature centrifugation at 4,000 × g for 5 min. Because components of complex media interfere with H2O2 measurements, cells grown in LBg were washed and assayed in fresh, prewarmed minimal media containing 0.02% of both glucose and casamino acids. There is no loss of respiration due to this medium switch (data not shown). Cells grown in minimal media did not require washing and were resuspended after pelleting in the same minimal media containing one-tenth of the original concentration of carbon source. The lower concentration of carbon source was used in the assay because some carbon sources autoxidize and thereby contribute to H2O2 formation (
      • Seaver L.C.
      • Imlay J.A.
      ).
      Cells were finally resuspended at an OD of ∼ 0.1 with shaking at 37 °C. At selected time points the aliquots were removed, cells were removed by 1-min centrifugation in a microfuge, and H2O2 levels were determined by the Amplex Red/horseradish peroxidase method (
      • Seaver L.C.
      • Imlay J.A.
      ). Fluorescence was measured in a Shimadzu RF Mini-150 fluorometer and converted to H2O2 concentration using a curve obtained from standard samples in the same assay medium. Rates were normalized to the OD of cells at the 20-min time point. A small amount of H2O2 is generated by the dye/horseradish peroxidase detection system itself; this amount was accounted for by the standard curves. H2O2 formation rates were also corrected for any background H2O2 formed by medium alone. These backgrounds were ≤0.001 μm H2O2/min. Measurements in pyruvate medium were corrected for the ability of pyruvate to scavenge H2O2 using the equation d[H2O2]/dt = (cellular rate of H2O2 formation) – (k[pyruvate][H2O2]). k was determined by measuring H2O2 concentration over time after the addition of 1.5 μm H2O2 to the assay medium (k = 0.350). All rates were averaged from three separate measurements and presented with standard deviation. p values were calculated using Student's t test.
      To increase O2 concentration during H2O2 measurements, pure O2 was bubbled vigorously through the resuspended cells at 37 °C during the period of measurement. Under these conditions, H2O2 formation rates were also corrected for any H2O2 formation due to bubbled media alone.

      RESULTS

      Quantitation of H2O2 Formed inside Living Cells—E. coli mutants that lack catalase and alkylhydroperoxide reductase, an NADH peroxidase, cannot scavenge H2O2 (for brevity and because the E. coli catalases are denoted hydroperoxidases I and II, these triple mutants are designated in this study as hydroperoxidase-deficient, or Hpx). These strains excrete into the growth medium any H2O2 that is formed endogenously (
      • Seaver L.C.
      • Imlay J.A.
      ,
      • Seaver L.C.
      • Imlay J.A.
      ). We have modified an established horseradish peroxidase/Amplex Red assay in order to quantify the rate of H2O2 formation. Exponentially growing cultures were washed and resuspended into a 37 °C medium that had been freshly prepared in order to minimize the accumulation of H2O2 from the chemical oxidation of glucose. At selected time points the cells were removed, and the accumulated H2O2 was measured. This procedure circumvents the possibility that artifacts will arise from the direct interaction of cells with the detection system, an issue that has been raised in studies of mitochondrial H2O2 excretion (
      • Staniek K.
      • Nohl H.
      ,
      • Staniek K.
      • Nohl H.
      ).
      The Hpx strain grows progressively slower when it is cultured for an extended time in aerobic medium (
      • Seaver L.C.
      • Imlay J.A.
      ), in concert with the progressive accumulation of DNA damage.
      S. Park and J. A. Imlay, manuscript in preparation.
      We were concerned that the slowed growth might affect the rates of metabolism and H2O2 production. To minimize this possibility, the Hpx mutant was grown to log phase anaerobically and then subcultured into fresh aerobic medium containing amino acids. This regime allowed the mutant to grow almost as well as the wild-type for several subsequent generations (Fig. 1A). At the point when cultures were harvested for measurements of H2O2 production, the respiration rate was approximately that of the wild-type strain.
      Figure thumbnail gr1
      Fig. 1Measurement of H2O2production. A, aerobic growth of an Hpx strain. Anaerobic wild-type (MG1655, shown as circles) and Hpx (LC106, shown as squares) cultures in LBg were grown into log phase (white). Where indicated, some cells were diluted into fresh aerobic LBg and grown aerobically in shaking flasks (black). Hpx strains were collected for measurement of H2O2 formation at an OD of 0.10 (asterisks). At this point their respiration rates approximated that of wild-type cells. B, H2O2 excretion by log-phase Hpx cells suspension in fresh upon glucose/amino acids medium. No H2O2 is detected in analogous cultures of Hpx+ cells (not shown).
      Exponentially growing cultures generated ∼0.5 μm H2O2/min per OD600 in glucose medium. The vast majority of this H2O2 was formed by the bacteria, because the rate was >95% lower in the sterile medium (Fig. 1B). By normalizing the rate of H2O2 production to the cytoplasmic volume of the cells, we calculate a rate of 10–15 μm/s. This rate is about twice what was estimated through extrapolation of the rate at which H2O2 was formed by respiratory vesicles in vitro (
      • Messner K.R.
      • Imlay J.A.
      ). Therefore, either those experiments underestimated the rates at which respiratory enzymes autoxidize, or another substantial source of H2O2 exists in E. coli. Several avenues of H2O2 formation have been observed during in vitro studies: stoichiometric H2O2 production by oxidases (
      • Yamashita M.
      • Azakami H.
      • Yokoro N.
      • Roh J.-H.
      • Suzuki H.
      • Kumagai H.
      • Murooka Y.
      ), autoxidation of excreted metabolites (
      • Thornalley P.
      • Wolff S.
      • Crabbe J.
      • Stern A.
      ), dismutation of periplasmically generated superoxide (
      • Huycke M.M.
      • Moore D.
      • Joyce W.
      • Wise P.
      • Shepard L.
      • Kotake Y.
      • Gilmore M.S.
      ), superoxide-driven chain reactions (
      • Chan P.C.
      • Bielski B.H.J.
      ), and the autoxidation of redox enzymes (
      • Chance B.
      • Sies H.
      • Boveris A.
      ,
      • Imlay J.A.
      • Fridovich I.
      ,
      • Messner K.R.
      • Imlay J.A.
      ,
      • Messner K.R.
      • Imlay J.A.
      ,
      • Massey V.
      • Strickland S.
      • Mayhew S.G.
      • Howell L.G.
      • Engel P.C.
      • Matthews R.G.
      • Schuman M.
      • Sullivan P.A.
      ). We examined each in turn to evaluate their contribution to overall H2O2 production by E. coli.
      H2O2 Formation Is Due to an Adventitious Reaction with O2—The rates at which flavoenzymes react in vitro with O2 to produce(O2.¯)and H2O2 is proportional to the concentration of dissolved oxygen (
      • Messner K.R.
      • Imlay J.A.
      ,
      • Messner K.R.
      • Imlay J.A.
      ). If such adventitious reactions govern H2O2 formation in vivo, then increased O2 levels would significantly increase H2O2 production rates. In contrast, if H2O2 were generated as a stoichiometric product of an oxidase, then increases in O2 concentration would have little affect, because such enzymes are saturated by low concentrations of O2.
      The Hpx strain was grown to log phase aerobically. These cells were washed in fresh, prewarmed media and resuspended at an OD of ∼ 0.1. H2O2 concentration was measured over time in identical cultures that were vigorously aerated with air or pure oxygen. The culture with increased O2 produced ∼2.5-fold more H2O2 (Fig. 2). This result suggests that the most significant source(s) of H2O2 react with oxygen adventitiously rather than as an intended substrate. In fact, only one H2O2-generating oxidase is known to exist in the K12 strains of E. coli, the periplasmic monoamine oxidase (
      • Yamashita M.
      • Azakami H.
      • Yokoro N.
      • Roh J.-H.
      • Suzuki H.
      • Kumagai H.
      • Murooka Y.
      ), and none of its known substrates were present during these measurements.
      Figure thumbnail gr2
      Fig. 2H2O2 production persists in mutants lacking NADH dehydrogenase activity. LC106, MW11, LC138, LC156, LC109, and LC114 cultures were grown in LBg anaerobically to log phase, subcultured, and grown in aerobic LBg media to log phase. Cultures of cells with plasmids contained ampicillin throughout growth. Cells were then assayed for H2O2 formation in minimal glucose-casamino acids medium or harvested for NdhII activity. NdhII activity is normalized to that of the background strain and presented in parentheses above peroxide measurements. *, p < 0.0001; **, p < 0.004, and ***, p < 0.01 versus background strain (Hpx).
      H2O2 Is Not Formed by an Excreted Product—It was possible that the Hpx strain excreted a metabolite that subsequently autoxidized to form H2O2 outside the cell. To check this possibility, cells were rapidly removed, and H2O2 formation rates were measured in the spent medium. LC106 (Hpx) was grown to log phase in aerobic minimal casamino acids glucose medium and then removed by centrifugation. The spent medium of aerobic cultures generated <0.01 μm H2O2/min, compared with 0.4 μm H2O2/min·OD produced by LC106 cultures (in contrast, the spent medium produced by filtration of anaerobic cultures included an unknown substance that gradually oxidized during the first 10 min when the medium was aerated).
      We cannot exclude the possibility that in these control experiments some H2O2 was generated by an excreted product that autoxidized during the 60 s that was required to remove the cells. However, any metabolite that autoxidizes so efficiently could not have generated even more H2O2 when the oxygen level was raised (see above). We conclude that most or all of the H2O2 that we detected was generated inside the cells.
      Periplasmic(O2·¯)Does Not Contribute Substantially to Endogenous H2O2—Recent work in our lab has determined that some(O2.¯)is formed within the periplasm during aerobic respiration.
      S. Korshunov and J. A. Imlay, manuscript in preparation.
      Because(O2.¯)dismutation generates H2O2, we sought to determine whether this periplasmic(O2.¯)was responsible for much of the H2O2 that effluxes from the Hpx cells (
      • Seaver L.C.
      • Imlay J.A.
      ). Periplasmic(O2.¯)was measured and found to account for no more than 10% of the cellular H2O2.
      S. Korshunov and J. A. Imlay, manuscript in preparation.
      Menaquinone is necessary for periplasmic(O2.¯)production in these cells. For that reason H2O2 production was measured in an Hpx strain with an additional menA mutation (LC132). In this mutant, H2O2 production was not appreciably decreased (Fig. 3). Therefore, the periplasmic source of(O2.¯)did not contribute significantly to H2O2 formation.
      Figure thumbnail gr3
      Fig. 3NdhII is the major source of H2O2 in cells blocked in respiration. LC106, LC149, LC145, LC147, LC150, and LC165 cultures were grown in LBg plus uracil media anaerobically to log phase. Cultures of cells with the NdhII plasmid pNdhII (pMW01) contained ampicillin throughout growth. Cells were then assayed for H2O2 formation upon air exposure in minimal glucose (0.02%) and 20 amino acids (0.05 mm) plus uracil (0.1 mm). *, p < 0.00005 versus background (Hpx); **, p < 0.00005 versus background (Hpx Ubi Men).
      Superoxide-mediated Chain Reactions Are Not a H2O2 Source—(O2.¯)can univalently oxidize some small biomolecules, such as catechols, to form hydrogen peroxide and organic radicals, which subsequently transfer their unpaired electron to molecular oxygen. The latter reaction regenerates(O2.¯)and propagates a chain reaction. Because(O2.¯)dismutation ends the chain, such reactions are commonly used as assays for superoxide dismutase (
      • Marklund S.
      • Marklund G.
      ). Similarly, NADH that is bound within the active site of mammalian lactate dehydrogenase or glyceraldehyde-3-phosphate dehydrogenase can react with protonated(O2.¯)to generate H2O2 and an NAD· radical, which then regenerates(O2.¯)(
      • Bielski B.H.
      • Chan P.C.
      ,
      • Chan P.C.
      • Bielski B.H.
      ). These chain reactions may continue in vitro for many cycles, creating H2O2 at each cycle. To test the possibility that such reactions might be occurring in E. coli, H2O2 production was measured in concert with overproduction of iron-containing superoxide dismutase. In LC141 (Hpx pFe-SOD), a 26-fold overproduction of superoxide dismutase did not alter the H2O2 formation rate from that of LC106 (Hpx) (data not shown). We conclude that superoxide-mediated chain reactions do not contribute significantly to H2O2 stress in E. coli.
      The Majority of H2O2 Formed May Not Originate from the Respiratory Chain—The previous results led us to believe that most H2O2 is formed by the straightforward autoxidation of redox enzymes. Several flavoenzymes of E. coli have been identified that react with oxygen in vitro, forming H2O2 (
      • Messner K.R.
      • Imlay J.A.
      ,
      • Imlay J.A.
      ,
      • Gaudu P.
      • Touati D.
      • Niviere V.
      • Fontecave M.
      ). To evaluate the involvement of these enzymes in H2O2 formation, we eliminated or overexpressed the structural genes of each candidate flavoenzyme in Hpx cells.
      The most compelling candidate was NADH dehydrogenase II, the primary respiratory dehydrogenase when E. coli grows in glucose medium. Previous in vitro studies found that respiring vesicles prepared from ndh mutants generated much less H2O2 than did wild-type vesicles. Conversely, H2O2 production increased when NdhII was overproduced (
      • Messner K.R.
      • Imlay J.A.
      ).
      To test if NdhII was a significant source of H2O2in vivo, NdhII was deleted from the Hpx background. However, H2O2 production by this ndh mutant was not significantly less than that by its parent (MW11) (Fig. 2). Given the inherent error of this assay, we deduce that NdhII could contribute no more than 15% of the total H2O in Hpx–2 cells.
      NdhII is responsible for ∼90% of the total NADH dehydrogenase activity that is detected in membranes from E. coli grown in glucose medium (data not shown), with NdhI contributing the other 10%. However, inside cells the actual division of the NADH flux between NdhI and NdhII is likely to depend upon both substrate concentration and protonmotive force, and it has not been quantified. We tested whether H2O2 production could be elevated if the electron flux were forcibly directed through NdhII by eliminating NdhI. However, the addition of a nuo mutation (generating LC138) did not increase the rate of H2O2 formation (Fig. 2). Thus, the data indicated that another source obscures the contribution, if any, of NdhII to H2O2 production in vivo.
      When cells are grown in glucose medium, most respiration derives from NADH oxidation by these enzymes. Thus oxygen consumption of the nuo ndh double mutant was <20% of that of the parent (data not shown). Surprisingly, the combination of nuo and ndh mutations did not decrease the H2O2 production rate, as would be expected if flux through the respiratory chain were needed to make endogenous H2O2 (Fig. 2). This result suggests that most H2O2 is generated outside of the respiratory chain, in contradiction to the prevailing idea in this field.
      NADH Dehydrogenase II Forms Some H2O2 in Vivo—Because the ndh mutation had no discernible effect upon H2O2 production, we wondered whether the autoxidation of the enzyme that had been observed in vitro was entirely artifactual. Because the rate at which inverted respiratory vesicles generated H2O2 was elevated when NdhII was overproduced (
      • Messner K.R.
      • Imlay J.A.
      ), we conducted the same experiment in vivo, overproducing NdhII 16-fold in the Hpx background. Initial observations were consistent with increased H2O2 production, as the doubling time in the aerobic medium increased from an already slow 33 min in the Hpx strain to 44 min in the overproducer. In anaerobic media these strains grew at the same rate as the wild-type strain (data not shown). H2O2 production was then tested directly. The NdhII-overproducing strain formed ∼3.5-fold more H2O2 than the parental Hpx strain with or without the empty vector (Fig. 2).
      Earlier studies showed that NdhII generated H2O2 most rapidly when membrane vesicles were prepared from mutants lacking ubiquinone (
      • Messner K.R.
      • Imlay J.A.
      ), apparently because the absence of the downstream acceptor caused electrons to remain on the auto-oxidizable flavin of the enzyme. The same effect was observed in vivo; the quinoneless (menA ubiA) Hpx mutant (LC147) exhibited a substantial increase in H2O2 production (Fig. 3). Upon the addition of an ndh mutation, the rate of H2O2 formation was again diminished. Conversely, overproduction of NdhII in the quinoneless Hpx mutant (LC150) resulted in an even greater (9-fold) increase in H2O2 production. These data follow the pattern that had been observed in vitro and definitively show that NdhII can generate H2O2in vivo.
      Physiological Evidence of Significant H2O2 Production from NdhII—To support the conclusion that overproduced NdhII generates H2O2 in E. coli, we employed a katG::lacZ fusion, which serves as a reporter of OxyR activity (
      • Seaver L.C.
      • Imlay J.A.
      ). In E. coli, OxyR directly senses steady-state H2O2 concentration and, when it rises, positively activates transcription of a defensive regulon that includes katG (
      • Christman M.F.
      • Storz G.
      • Ames B.N.
      ). The katG::lacZ expression was elevated 10- to 15-fold in a strain lacking Ahp, the primary scavenger of endogenous H2O2 (Table II and Ref.
      • Seaver L.C.
      • Imlay J.A.
      ). The addition of the NdhII-overproducing plasmid raised β-galactosidase activity further. No effect was observed in Ahp-proficient strains.
      Table IIEndogenous H2O2 from NdhII enhances OxyR activation
      Strain
      All strains were isogenic and harbored a λRS45 (katG::lacZ). Cells were grown in LBg. The same conditions were used for NdhII activity and H2O2 formation measurements.
      β-Galactosidase activity
      units/mg
      Wild-type0.03 ± 0.01
      pBR3220.04 ± 0.01
      pNdhII
      pNdhII = pMW01.
      0.06 ± 0.01
      Ahp-0.57 ± 0.1
      Ahp- pNdhII
      pNdhII = pMW01.
      0.92 ± 0.1
      a All strains were isogenic and harbored a λRS45 (katG::lacZ). Cells were grown in LBg. The same conditions were used for NdhII activity and H2O2 formation measurements.
      b pNdhII = pMW01.
      In sum, the experiments with NdhII confirm that the enzyme reacts with oxygen in vivo as it does in vitro. However, the results indicate that it is evidently only a minor source of endogenous H2O2.
      Fumarate Reductase Forms Little H2O2 When Anaerobic Cells Are Aerated—The other respiratory enzyme that was identified as an H2O2 source in vitro was fumarate reductase (Frd). Frd is a member of the anaerobic respiratory chain and would not have been present in the preceding experiments; it is induced when oxygen is absent. Because this enzyme readily autoxidizes when it is exposed to air (
      • Imlay J.A.
      ), we tested the possibility that Frd would generate substantial H2O2 when anaerobic cells are abruptly aerated. We grew the Hpx mutant anaerobically to log phase, resuspended it in fresh glucose-fumarate medium, and then aerated the culture. The rate of H2O2 production was not substantially different from that of cultures grown continuously in aerobic medium. 8-fold overproduction of Frd increased H2O2 production only 1.8-fold (Fig. 4). This moderate acceleration of H2O2 formation presumably did not derive from an increased biosynthetic burden per se, because a far greater overproduction of β-lactamase had no effect upon H2O2 rates (Fig. 2).
      Figure thumbnail gr4
      Fig. 4Effect of fumarate reductase synthesis upon H2O2 formation. Aerobic overnight cultures of LC106, LC126, LC114, LC128, and LC141 were grown with an additional catalase. Cells were then diluted into anaerobic minimal glucose (0.2%) and fumarate (40 mm) plus histidine or LBg and grown to log phase. Cultures of cells with the Frd plasmid (pH3) contained ampicillin. Cells were then assayed for H2O2 formation and Frd activity upon air exposure. Frd activity is normalized to the background strain and presented in parentheses above peroxide measurements. *, p < 0.007 versus background (Hpx).
      Interestingly, an Hpx Frd mutant (LC126) also showed an increase in H2O2 production upon aeration. This result appeared contrary to the expected result, a decrease in H2O2 formation, if Frd were normally a substantial H2O2 source. Because Frd is a terminal electron acceptor in the respiratory chain, this increase may be due to a “back-up” of electrons onto a component upstream. This component could be NdhII, as was seen with the strain lacking quinone (Fig. 3). In sum, whereas Frd produced some H2O2in vivo, it did so at substantial rates only when it was overproduced.
      Overproduction of Flavin Reductase Results in H2O2 Formation in Vivo—The preceding experiments suggested that the primary H2O2 source might lie outside the respiratory chain. We looked, therefore, toward free flavins as a potential H2O2 source. Reduced free flavins, generated by flavin reductase, react with oxygen to produce H2O2 in vitro (
      • Gaudu P.
      • Touati D.
      • Niviere V.
      • Fontecave M.
      ), presumably by same mechanism as the autoxidation of flavoenzymes. However, Hpx cells lacking flavin reductase formed H2O2 at approximately the same rate as did the flavin reductase-proficient strain (Fig. 5). The Hpx strain that overproduced flavin reductase 80-fold formed only 1.9-fold more H2O2 than did the Hpx strain (Fig. 5). These data support the ability of free flavins to adventitiously make H2O2, but they indicate that flavin reductase is a relatively insignificant H2O2 source under these growth conditions.
      Figure thumbnail gr5
      Fig. 5Overexpression of flavin reductase increases H2O2 formation. Aerobic cultures of LC106, LC118, and LC110 were grown in LBg to log phase; at an OD600 of 0.02, 0.4 mm isopropyl-β-d-thiogalactopyranoside was added to the strain containing the flavin reductase plasmid pFre (pfn3) to induce expression. At an OD of ∼0.3, cells were assayed for flavin reductase (Fre) activity and a H2O2 formation rate. Flavin reductase activity is normalized to the background strain and presented in parentheses above the peroxide measurements. *, p < 0.007 versus background strain (Hpx).
      Thus the three candidate enzymes NdhII, Frd, and flavin reductase generated varying amounts of H2O2in vivo. However, they produced much less than did another unidentified source.
      Endogenous H2O2 Production during Growth on Different Carbon Sources—The amount of oxidative stress that a bacterium experiences depends on the rates at which intracellular(O2.¯)and H2O2 are produced. These rates should depend on the titers of auto-oxidizable enzymes in the cell, which, in turn, might change in response to growth on different carbon sources. To test this possibility we attempted to measure the rates of H2O2 production during log-phase growth on glucose, gluconate, glycerol, lactate, pyruvate, succinate, acetate, and casamino acids (2%). Respiration rates were measured in parallel with H2O2. H2O2 production rates varied modestly from one source to another (Table III). Unfortunately, these results were not very informative in suggesting the identity of the major H2O2 source.
      Table IIIH2O2 formation by an Hpx strain grown on various carbon sources
      Carbon source
      Carbon source concentration, amino acid supplementation, and growth conditions are described under “Materials and Methods.”
      tD
      tD = Doubling time in log phase.
      O2 consumption rate
      Measurements were done in duplicate and an average is presented.
      H2O2 formation rate
      Measurements were done in duplicate and an average is presented.
      H2O2/O2 consumed
      minμm O2 s-1 OD-1μm H2O2 s-1 OD-1
      Casamino acids482.48.3 × 10-33.5 × 10-3
      Glucose1041.47.2 × 10-35.1 × 10-3
      Gluconate1081.69.9 × 10-36.2 × 10-3
      Glycerol1551.38.3 × 10-36.4 × 10-3
      Pyruvate1121.54.8 × 10-33.2 × 10-3
      Lactate1322.211.0 × 10-35.0 × 10-3
      SuccinateNG
      NG = No growth.
      AcetateNG
      NG = No growth.
      a Carbon source concentration, amino acid supplementation, and growth conditions are described under “Materials and Methods.”
      b tD = Doubling time in log phase.
      c Measurements were done in duplicate and an average is presented.
      d NG = No growth.
      Interestingly, whereas this strain could grow aerobically in liquid culture on fermentable carbon sources, it was unable to accommodate a transfer from an anaerobic glucose medium to aerobic media containing either succinate or acetate as the sole carbon source. No increase in OD600 was seen in 20 h. In contrast, the Hpx+ parental strain (MG1655) grew well with both. Notably, these two carbon sources are catabolized via the trichloroacetic acid cycle; the others can be catabolized via other pathways. Thus, the low level of H2O2 that accumulates in the Hpx mutant is evidently sufficient to disrupt some catabolic activities. We note that this disruption could influence the rate of metabolic H2O2 production.

      DISCUSSION

      Autoxidation of Respiratory Enzymes—Aerobic cells generate enough internal H2O2 to cripple themselves. Mutant strains of E. coli that are stripped of antioxidant defenses, that is, of the ability to scavenge endogenous H2O2 or, alternatively, to repair oxidized DNA, grow poorly or not at all in aerobic environments (
      • Seaver L.C.
      • Imlay J.A.
      ,
      • Imlay J.A.
      • Linn S.
      ,
      • Boling M.
      • Adler H.
      • Masker W.
      ,
      • Morimyo M.
      ). In this study we attempted to identify the mechanisms by which H2O2 forms inside E. coli and to determine whether the rate of H2O2 formation depends upon the identity of the growth substrate.
      Our earlier work with respiratory vesicles in vitro had indicated, as shown in Reactions 1,2,3,
      FADH+O2FADH·~O2·¯REACTION1
      Eq.1


      FADH·~O2·¯FAD·+H++O2·¯REACTION2
      Eq.2


      FAD·+O2FAD++O2·¯REACTION3
      Eq.3


      or in Reactions 4 and 5,
      FADH+O2FADH·~O2·¯REACTION4
      Eq.4


      FADH·~O2·¯FAD++H2O2REACTION5
      Eq.5


      that several dehydrogenases transfer electrons singly or sequentially to oxygen when it collides with their solvent-exposed reduced flavins (
      • Messner K.R.
      • Imlay J.A.
      ). In those experiments, the greatest yield from the aerobic respiratory chain arose from NADH dehydrogenase II, and the greatest yield from the anaerobic chain arose from fumarate reductase. The autoxidation rate was greatest when the oxygen concentration was elevated or when oxidized quinone acceptors were unavailable so that the electrons were backed up onto the enzymes. Overproduction of either enzyme enhanced the yield of H2O2 or(O2.¯).
      The in vivo data of this study confirmed that NdhII and Frd can react with oxygen. There were several plausible reasons why it might have turned out otherwise. A fundamental concern was that, in our in vitro studies, rough handling during the preparation of either the vesicles or the purified enzymes might have caused some fraction of the enzymes to misfold and become artifactually vulnerable to oxidation. That was evidently not the case, because these enzymes turned out to autoxidize in vivo as well.
      However, a more subtle issue is that the redox status of the enzymes and the distribution of electrons among the redox moieties within the enzymes are dependent both upon substrate concentrations and membrane potential. In fact, whereas the in vitro experiments had suggested that NdhII might generate 5–10 μm/s H2O2in vivo, the present data shown that NdhII actually produced <1.5 μm/s. Why the difference? Right now we can only speculate. The steady-state redox status of the FAD moiety of NdhII reflects the dynamic balance between reduction by NADH, allosteric control by NAD+, and oxidation by ubiquinone. Efforts were made in the in vitro experiments to replicate the physiological dinucleotide pools; however, differences between in vitro and in vivo pH, counter-ion concentrations, and protonmotive force may have altered these interactions. Similarly, because fumarate can block the reaction of Frd with oxygen (
      • Messner K.R.
      • Imlay J.A.
      ,
      • Imlay J.A.
      ), the intracellular fumarate levels may have a big impact upon autoxidation rates.
      The Main Source of H2O2 Is Unknown—Ultimately, this study has not answered a key question: what is the primary source of endogenous H2O2? Our results indicate that it lies outside the respiratory chain. This is a surprise, as the amenability of respiratory components to univalent redox reactions has always prompted workers to look there for biomolecules that can react with triplet oxygen. Formally, whenever we eliminated one respiratory enzyme, the electron flow could have been redirected to another one that generated H2O2 at the same rate. However, this idea is quantitatively improbable, because the rates at which flavoproteins react adventitiously with oxygen vary by orders of magnitude (
      • Imlay J.A.
      ,
      • Massey V.
      • Strickland S.
      • Mayhew S.G.
      • Howell L.G.
      • Engel P.C.
      • Matthews R.G.
      • Schuman M.
      • Sullivan P.A.
      ). Thus, the replacement of a section of one redox chain with another is unlikely to precisely recreate the same flux of H2O2. Instead, we are inclined to believe that most H2O2 is formed when either a non-respiratory enzyme or a cellular metabolite reacts adventitiously with oxygen. The simplicity of this assay and the genetic tractability of E. coli give us hope that this experimental system will allow the question to be answered. This work is in progress.
      The Rate of H2O2 Formation Is Consistent—Uncertainty has arisen regarding the rate at which mitochondria generate H2O2, with published values ranging from <0.15 to 2% of oxygen consumption (
      • Chance B.
      • Sies H.
      • Boveris A.
      ,
      • St-Pierre J.
      • Buckingham J.A.
      • Roebuck S.J.
      • Brand M.D.
      ). The disparity has been variously attributed to the use of inhibitors (
      • Fridovich I.
      ), variability with different substrates (
      • St-Pierre J.
      • Buckingham J.A.
      • Roebuck S.J.
      • Brand M.D.
      ), interference from endogenous peroxidases (
      • Antunes F.
      • Cadenas E.
      ), and artifacts from the detection system (
      • Staniek K.
      • Nohl H.
      ). These problems have been circumvented with this E. coli system. Although we do not yet know the mechanism of its formation, with a variety of growth substrates H2O2 was produced in exponentially growing cells at rates between 0.3 and 0.7% of oxygen consumption and 9–22 μm/s inside the cell. Prior measurements of the intracellular scavenging activity had led to an estimate that these cells contain ∼20 nm steady-state H2O2 (
      • Seaver L.C.
      • Imlay J.A.
      ). This value, then, is a quantitative measure of the H2O2 stress that growing cells experience. It is probable that the value differs in nutrient-starved cells.

      Acknowledgments

      We are grateful to the colleagues cited in Table I who provided strains for this work.

      References

        • Imlay J.A.
        • Linn S.
        Science. 1988; 240: 1302-1309
        • Fridovich I.
        Annu. Rev. Biochem. 1995; 64: 92-112
        • Carlioz A.
        • Touati D.
        EMBO J. 1986; 5: 623-630
        • Kuo C.F.
        • Mashino T.
        • Fridovich I.
        J. Biol. Chem. 1987; 262: 4724-4727
        • Flint D.H.
        • Tuminello J.F.
        • Emptage M.H.
        J. Biol. Chem. 1993; 268: 22369-22376
        • Gardner P.R.
        • Fridovich I.
        J. Biol. Chem. 1991; 266: 1478-1483
        • Gardner P.R.
        • Fridovich I.
        J. Biol. Chem. 1991; 266: 19328-19333
        • Liochev S.I.
        • Fridovich I.
        Proc. Natl. Acad. Sci. U. S. A. 1992; 89: 5892-5896
        • Chang E.C.
        • Crawford B.F.
        • Hong Z.
        • Bilinski T.
        • Kosman D.J.
        J. Biol. Chem. 1991; 266: 4417-4424
        • van Loon A.P.G.M.
        • Pesold-Hurt B.
        • Schatz G.
        Proc. Natl. Acad. Sci. U. S. A. 1986; 83: 3820-3824
        • Wallace M.A.
        • Liou L.-L.
        • Martins J.
        • Clement M.H.S.
        • Bailey S.
        • Longo V.D.
        • Valentine J.S.
        • Gralla E.B.
        J. Biol. Chem. 2004; 279: 32055-32062
        • Li Y.
        • Huang T.T.
        • Carlson E.J.
        • Melov S.
        • Ursell P.C.
        • Olson J.L.
        • Noble L.J.
        • Yoshimura M.P.
        • Berger C.
        • Chan P.H.
        Nat. Genet. 1995; 11: 376-381
        • Gort A.S.
        • Imlay J.A.
        J. Bacteriol. 1998; 180: 1402-1410
        • Seaver L.C.
        • Imlay J.A.
        J. Bacteriol. 2001; 183: 7173-7181
        • Naqui A.
        • Chance B.
        Ann. Rev. Biochem. 1986; 55: 137-166
        • Chance B.
        • Sies H.
        • Boveris A.
        Physiol. Rev. 1979; 59: 527-605
        • Imlay J.A.
        • Fridovich I.
        J. Biol. Chem. 1991; 266: 6957-6965
        • Messner K.R.
        • Imlay J.A.
        J. Biol. Chem. 1999; 274: 10119-10128
        • Turrens J.F.
        • Alexandre A.
        • Lehninger A.L.
        Arch. Biochem. Biophys. 1985; 237: 408-414
        • Han D.
        • Williams E.
        • Cadenas E.
        Biochem. J. 2001; 353: 411-416
        • Lambert A.J.
        • Brand M.D.
        J. Biol. Chem. 2004; 279: 39414-39420
        • Messner K.R.
        • Imlay J.A.
        J. Biol. Chem. 2002; 277: 42563-42571
        • Boveris A.
        • Cadenas E.
        Oberley L.W. Superoxide Dismutase. 2. CRC Press, Boca Raton1982: 15-30
        • Boveris A.
        • Chance B.
        Biochem. J. 1973; 134: 707-716
        • Antunes F.
        • Cadenas E.
        FEBS Lett. 2000; 475: 121-126
        • Seaver L.C.
        • Imlay J.A.
        J. Bacteriol. 2001; 183: 7182-7189
        • Miller J.H.
        Experiments in Molecular Genetics. Cold Spring Harbor Laboratory Press, Cold Spring Harbor, NY1972: 201-205 (352–355, and 431–433)
        • Datsenko K.A.
        • Wanner B.L.
        Proc. Natl. Acad. Sci. U. S. A. 2000; 97: 6640-6645
        • Wallace B.J.
        • Young I.G.
        Biochim. Biophys. Acta. 1977; 461: 75-83
        • Calhoun M.W.
        • Gennis R.B.
        J. Bacteriol. 1993; 175: 3013-3019
        • Chung C.T.
        • Miller R.H.
        Nucleic Acids Res. 1988; 16: 3580
        • Woodmansee A.N.
        • Imlay J.A.
        J. Biol. Chem. 2002; 277: 34055-34066
        • Imlay J.A.
        J. Biol. Chem. 1995; 270: 19767-19777
        • Matsushita K.
        • Ohnishi T.
        • Kaback H.R.
        Biochemistry. 1987; 26: 7732-7737
        • McCord J.M.
        • Fridovich I.
        J. Biol. Chem. 1969; 244: 6049-6055
        • Staniek K.
        • Nohl H.
        Biochim. Biophys. Acta. 1999; 1413: 70-80
        • Staniek K.
        • Nohl H.
        Biochim. Biophys. Acta. 2000; 1460: 268-275
        • Yamashita M.
        • Azakami H.
        • Yokoro N.
        • Roh J.-H.
        • Suzuki H.
        • Kumagai H.
        • Murooka Y.
        J. Bacteriol. 1996; 178: 2941-2947
        • Thornalley P.
        • Wolff S.
        • Crabbe J.
        • Stern A.
        Biochim. Biophys. Acta. 1984; 797: 276-287
        • Huycke M.M.
        • Moore D.
        • Joyce W.
        • Wise P.
        • Shepard L.
        • Kotake Y.
        • Gilmore M.S.
        Mol. Microbiol. 2001; 42: 729-740
        • Chan P.C.
        • Bielski B.H.J.
        J. Biol. Chem. 1974; 249: 1317-1319
        • Massey V.
        • Strickland S.
        • Mayhew S.G.
        • Howell L.G.
        • Engel P.C.
        • Matthews R.G.
        • Schuman M.
        • Sullivan P.A.
        Biochem. Biophys. Res. Commun. 1969; 36: 891-897
        • Marklund S.
        • Marklund G.
        Eur. J. Biochem. 1974; 47: 469-474
        • Bielski B.H.
        • Chan P.C.
        J. Biol. Chem. 1976; 251: 3841-3844
        • Chan P.C.
        • Bielski B.H.
        J. Biol. Chem. 1980; 255: 874-876
        • Gaudu P.
        • Touati D.
        • Niviere V.
        • Fontecave M.
        J. Biol. Chem. 1994; 269: 8182-8188
        • Christman M.F.
        • Storz G.
        • Ames B.N.
        Proc. Natl. Acad. Sci. U. S. A. 1989; 86: 3484-3488
        • Imlay J.A.
        • Linn S.
        J. Bacteriol. 1986; 166: 519-527
        • Boling M.
        • Adler H.
        • Masker W.
        J. Bacteriol. 1984; 160: 706-710
        • Morimyo M.
        J. Bacteriol. 1982; 152: 208-214
        • St-Pierre J.
        • Buckingham J.A.
        • Roebuck S.J.
        • Brand M.D.
        J. Biol. Chem. 2002; 277: 44784-44790
        • Fridovich I.
        Aging Cell. 2004; 3: 13-16
        • Wanner B.L.
        J. Mol. Biol. 1986; 191: 39-58
        • Fieschi F.
        • Niviere V.
        • Frier C.
        • Decout J.L.
        • Fontecave M.
        J. Biol. Chem. 1995; 270: 30392-30400
        • Blaut M.
        • Whittaker K.
        • Valdovinos A.
        • Ackrell B.A.C.
        • Gunsalus R.P.
        • Cecchini G.
        J. Biol. Chem. 1989; 264: 13599-13604
        • Sakamoto H.
        • Touati D.
        J. Bacteriol. 1984; 159: 418-420