Extracellular superoxide production by Enterococcus faecalis requires demethylmenaquinone and is attenuated by functional terminal quinol oxidases


  • Mark M. Huycke,

    Corresponding author
    1. The Muchmore Laboratories for Infectious Diseases Research, Research Service, Department of Veterans Affairs Medical Center, Oklahoma City, OK 73104, USA.
    2. Department of Medicine, University of Oklahoma Health Sciences Center, Oklahoma City, OK 73190, USA.
    • *For correspondence. Infectious Diseases (111C), 921 N.E. 13th Street, Oklahoma City, OK, USA 73104. E-mail mark-huycke@ouhsc.edu; Tel. (+1) 405 270 0501, ext. 3285; Fax (+1) 405 297 5934.

    Search for more papers by this author
  • Danny Moore,

    1. Department of Medicine, University of Oklahoma Health Sciences Center, Oklahoma City, OK 73190, USA.
    Search for more papers by this author
  • Wendy Joyce,

    1. The Muchmore Laboratories for Infectious Diseases Research, Research Service, Department of Veterans Affairs Medical Center, Oklahoma City, OK 73104, USA.
    Search for more papers by this author
  • Phillip Wise,

    1. The Muchmore Laboratories for Infectious Diseases Research, Research Service, Department of Veterans Affairs Medical Center, Oklahoma City, OK 73104, USA.
    Search for more papers by this author
  • Laura Shepard,

    1. The Muchmore Laboratories for Infectious Diseases Research, Research Service, Department of Veterans Affairs Medical Center, Oklahoma City, OK 73104, USA.
    Search for more papers by this author
  • Yashige Kotake,

    1. Free Radical Biology and Aging Program, Oklahoma Medical Research Foundation, Oklahoma City, OK 73104, USA.
    Search for more papers by this author
  • Michael S. Gilmore

    1. Departments of Ophthalmology, and Microbiology and Immunology, University of Oklahoma Health Sciences Center, Oklahoma City, OK 73190, USA.
    Search for more papers by this author


The intestinal commensal bacterium, Enterococcus faecalis, is unusual among prokaryotic organisms in its ability to produce substantial extracellular superoxide. Transposon mutagenesis, allelic replacement, and electron spin resonance (ESR)-spin trapping showed that superoxide production and generation of derivative hydroxyl radical were dependent on membrane-associated demethylmenaquinone. Extracellular superoxide was generated through univalent reduction of oxygen by reduced demethylmenaquinone. Moreover, extracellular superoxide production was inhibited by exogenous haematin, an essential cofactor for cytochrome bd, and by fumarate, a substrate for fumarate reductase. As integral membrane quinol oxidases, cytochrome bd and fumarate reductase redox cycle demethylmenaquinone, and are necessary for aerobic and anaerobic respiration respectively. A rat model of intestinal colonization demonstrated that conditions exist in the mammalian intestinal tract that permit a mode of respiration for E. faecalis that results in the formation of hydroxyl radical. These results identify and characterize the mechanism by which E. faecalis generates extracellular free radicals.


Superoxide is an anionic free radical produced by the univalent reduction of molecular oxygen. Because of its negative charge, superoxide is poorly diffusable through cell membranes and effectively constrained to intra- or extracellular locations, depending on its site of origin. Extracellular production is limited primarily to specialized eukaryotic cells expressing multimolecular NADPH-oxidase complexes that play important roles in cell signalling, oxygen sensing and inflammatory processes (Babior, 1999). In contrast, intracellular production of superoxide is ubiquitous among aerotolerant eukaryotes and prokaryotes, in which it is often considered deleterious (Fridovich, 1999), but may, on occasion, serve as an intracellular signal (Irani et al., 1997). Cytosolic concentrations of superoxide are normally quite small as this free radical is usually only a minor by-product of respiratory reactions. For example, intracellular superoxide is generated by the Q cycle of cytochrome bc1 in mitochondria, and by fumarate reductase or NADH dehydrogenase in bacteria (Barja, 1999; Imlay, 1995). Superoxide can be toxic under physiological conditions despite its limited production, short half-life, and limited reactivity, because it can readily dismute into hydrogen peroxide (k ≈ 2 × 105 M−1 s−1) and, in the presence of appropriate metal catalysts, produce hydroxyl radical via the Haber–Weiss reaction (Fridovich, 1997; Miller and Britigan, 1997). As a result, intracellular defences against superoxide-mediated damage are robust (Fridovich, 1997; Gort and Imlay, 1998).

Unlike specialized eukaryotic cells, production of extracellular superoxide by prokaryotic organisms is not well recognized. Recently, we observed that Enterococcus faecalis, an intestinal commensal of humans and animals, produces large amounts of extracellular superoxide (Huycke et al., 1996). Whereas nearly all E. faecalis strains produce extracellular superoxide, surveys of Gram-positive and Gram-negative bacteria have identified only a few other strains of enterococci among a limited number of species, and Lactococcus lactis, as expressing this unusual phenotype (Huycke et al., 1996; Winters et al., 1998). The mechanism of superoxide generation, however, was unknown. This investigation describes how inhibition of bacterial respiration is essential for generation of extracellular superoxide by E. faecalis. Moreover, growth conditions that favour superoxide production appear to be present in the mammalian intestinal tract, suggesting that this microorganism may be an important source of oxidative stress on the epithelium and other components of intestinal ecology.


E. faecalis mutants with attenuated extracellular superoxide production

Initially, we identified pathways involved in extracellular free radical generation using E. faecalis strain OG1RF, which produces 17.2 ± 0.3 nmol (mean ± SD) of extracellular superoxide per min per 109 colony-forming units (CFU) as measured by the ferricytochrome c reduction assay (Huycke et al., 1996). OG1RF was transformed with a temperature sensitive vector, pTV1(Ts)::Tn917, and two out of 3400 OG1RF colonies harbouring Tn917 insertions were found to be defective in superoxide production using a ferricytochrome c assay (Huycke et al., 1996). Southern blot of NcoI-digested DNA revealed a single insertion of Tn917 into the chromosomes of both mutants at apparently identical sites. One mutant, designated TM1, produced 0.18 ± 0.01 nmol extracellular superoxide per min per 109 CFU (1.0% compared with OG1RF) and was selected for further study.

The loss of superoxide production by mutant TM1 was confirmed by electron spin resonance (ESR)-spin trapping. An ESR spectrum generated by wild-type OG1RF cells, using 5-(diethoxyphosphoryl)-5-methyl-1-pyrroline-N-oxide (DEPMPO) as a spin trap, was observed to be similar to that produced by the control superoxide-generating system of xanthine and xanthine oxidase (Fig. 1). The spectrum is consistent with signals for superoxide and hydroxyl radical adducts calculated by computer simulation (Rouband et al., 1998). Addition of manganese superoxide dismutase eliminated both superoxide and hydroxyl radical signals, indicating that the hydroxyl-DEPMPO adduct arose from superoxide. Mutant TM1 cells yielded an attenuated spectrum (Fig. 1). Purified wild-type OG1RF membranes generated DEPMPO adducts for both superoxide and hydroxyl radical, but only after 100 µM NADH or l-lactate had been added as an electron donor. This localized the site of extracellular superoxide production to the enterococcal cell membrane. Superoxide and hydroxyl radical ESR-spin trapping signals were detected only if fermentable (e.g. d(+)-glucose), but not non-fermentable (e.g. d(+)-xylose), sugars were added to the buffer (Fig. 2).

Figure 1.

ESR-spin trapping spectra generated by E. faecalis strains OG1RF and TM1, and purified cell membranes.

A. superoxide generating control, 50 mM xanthine and 100 Units ml−1 xanthine oxidase with 20 mM DEPMPO.

B. OG1RF (1 × 109 CFU ml−1 of washed bacteria in 25 mM Tris buffer at pH 7.4 with 7 mM glucose).

C. Computer-simulated spectrum of (B) as a composite of DEPMPO-hydroxyl radical adduct with hyperfine coupling constants of aN = 13.9 G; aP = 46 G; aHβ = 13.36 G; aHγ = 0.6 G; and DEPMPO-superoxide adducts, corresponding to an exchange between two conformers with the following parameters: 1. (50%) aN = 13.05 G; aP = 50.0 G; aHβ = 12.9 G; aHγ = 0.87 G; 2. (50%) aN = 13.09 G; aP = 49.5 G; aHβ = 11.5 G; aHγ = 0.92 G. Unique peaks for superoxide (*) and hydroxyl radical (+) adducts of DEPMPO are indicated.

D. Loss of signals for OG1RF upon addition of manganese superoxide dismutase (100 units ml−1).

E. TM1 as in (B) showing no detectable superoxide or hydroxyl radical adduct signals.

F. Membranes from OG1RF in 25 mM Tris at pH 7.4 (1.0 mg protein ml−1) showing no signals.

G. As in (F), demonstrating signals similar to those from intact bacteria (B) after addition of 100 µM NADH. Arrow points up-field and length indicates 10 Gauss.

Figure 2.

Fermentable sugars necessary for extracellular superoxide by E. faecalis strain OG1RF. ESR-spin trapping of bacteria performed in 25 mM Tris buffer (pH 7.4) with a sugar (3 mM) using 20 mM DEPMPO as described in Fig. 1.

A. d(+)-glucose.

B. β-gentiobiose.

C. d(+)-cellobiose.

D. d(+)-mannose.

E. d(+)-xylose.

F. No sugar. Arrow points up-field and length indicates 10 Gauss.

Superoxide-deficient E. faecalis mutants blocked in demethylmenaquinone synthesis

To determine the nature of the mutation in TM1 that attenuated superoxide production, DNA flanking Tn917 was amplified by inverse polymerase chain reaction (PCR) and sequenced. Nucleotide sequencing for 8.3 kb of contiguous chromosomal DNA through the site of transposon insertion revealed seven genes coding for proteins with similarity to enzymes in the Escherichia coli pathway for aromatic amino acid biosynthesis (Fig. 3) (Pittard, 1996). To determine whether a block in the biosynthesis of aromatic amino acids had occurred, both wild-type OG1RF and mutant TM1 strains were grown in a defined medium. OG1RF is naturally auxotrophic for tryptophan (Murray et al., 1993) and, as expected, of the three aromatic amino acids only tryptophan was required for its growth. For TM1, however, phenylalanine and tyrosine, in addition to tryptophan, were required for growth. To confirm the involvement of aromatic amino acid biosynthesis in superoxide production, genes tentatively designated aroE, aroC and aroA were independently disrupted in OG1RF by allelic replacement. As anticipated, these mutants were also auxotrophic for tryptophan, phenylalanine and tyrosine. In addition, and as with TM1, each knockout showed ≥ 10-fold attenuation in the rate of superoxide production (data not shown).

Figure 3.

Map of chromosomal region for E. faecalis strain TM1 flanking Tn917 insertion. Seven deduced ORFs code for proteins with significant similarity to enzymes in the aromatic amino acid biosynthetic pathway of other bacteria: shikimate 5-dehydrogenase (aroE), phospho-2-dehydro-3-deoxyheptonate aldolase (aroF), 3-dehydroquinate synthase (aroB), chorismate synthase (aroC), prephenate dehydrogenase (tyrA), 3-phosphoshikimate 1-carboxyvinyltransferase (aroA), and prephenate dehydratase (pheA). This sequence has been deposited in GenBank under accession number AF318277.

Menaquinones are 1,4-dihydroxynaphthoquinone derivatives with isoprenoid side chains of varying length. These molecules serve as redox carriers for electron transport and are synthesized from intermediates in the aromatic amino acid biosynthetic pathway (Søballe and Poole, 1999). Most streptococci and other enterococcal species do not express menaquinone or produce superoxide (Collins and Jones, 1979). Enterococcus faecalis, however, produces demethylmenaquinones (DMKs) with eight (DMK-8) and nine (DMK-9) isoprene units (Baum and Dolin, 1965; Collins and Jones, 1979). Because chorismate is a common intermediate for aromatic amino acid and menaquinone biosynthesis, we hypothesized that attenuation of superoxide production in TM1 resulted from a block in the synthesis of chorismate and hence DMK. Thin-layer chromatography (TLC) and fast atom bombardment mass spectroscopy of lipids from mutant TM1 confirmed specific loss of DMK-8 and DMK-9 (Fig. 4).

Figure 4.

Thin-layer chromatogram of lipid extracts after development with iodine crystals for (A) OG1RF; (B) TM1 and (C) menB knockout of OG1RF. Materials analysed by fast atom bombardment mass spectroscopy are indicated by arrows (D). Peaks at 705 and 774 m z−1 are consistent with DMK-8 and DMK-9 respectively.

We explored the E. faecalis genome database at the Institute for Genomic Research for genes with similarity to those known to be involved in menaquinone biosynthesis in other systems (Meganathan, 1996). Five kilobases of DNA were identified containing four contiguous open reading frames (ORFs) similar to those encoding 1,4-dihydroxy-2-naphthoic acid synthase (menB), o-succinylbenzoate synthase (menE), menaquinone-specific isochorismate synthase (menD) and 2-succinyl-6-hydroxy-2,4-cyclohexadiene-1-carboxylic acid synthase (menF) from E. coli and Bacillus subtilis. To determine the involvement of these genes in menaquinone biosynthesis and superoxide production by E. faecalis, menB, menD and menE were independently disrupted in OG1RF by allelic replacement. Like the parenteral OG1RF strain, these mutants were only auxotrophic for tryptophan. However, as predicted, DMK-8 and DMK-9 were absent from lipid extracts of these mutants (Fig. 4). Knockouts resulted in 4.2–22.2% reductions in rates of superoxide production compared with OG1RF (menB, 0.73 ±0.12 nmol extracellular superoxide per min per 109 CFU; menD, 1.19 ± 0.04, and menE, 3.82 ± 0.06).

A direct relationship between the aroEFBCA and menBEDF biosynthetic pathways (Fig. 5) and superoxide production was established through complementation by growing mutants in 1,4-dihydroxy-2-naphthoic acid. This DMK precursor is synthesized via aromatic amino acid and menaquinone biosynthetic pathways, and serves as the substrate for 1,4-dihydroxy-2-naphthoate prenyltransferase (menA) that adds prenyl side chains to form demethylmenaquinone (Shineberg and Young, 1976). Extracellular superoxide production was restored to TM1 and menB mutant (Fig. 6), along with aroACE and menDE mutants (data not shown), after incubation with ≥ 1 µM 1,4-dihydroxy-2-naphthoic acid for as little as 15 min. Incubation of TM1 and the menB mutant with 1,4-dihydroxy-2-naphthoic acid also resulted in restoration of DMK-8 and DMK-9 in lipid extracts. In aggregate, these data demonstrated an essential role for membrane-associated DMK in the production of extracellular superoxide by E. faecalis, and confirmed phenotypic changes resulting from targeted gene knockouts were a result of allelic replacement or transposon insertion, and not polar effects.

Figure 5.

Pathway for biosynthesis of aromatic compounds showing chorismate as the branch point for aromatic amino acids and demethylmenaquinone. Boxed genes represent those inactivated by transposon insertion or allelic replacement using p3erm. E. faecalis OG1RF is auxotrophic for tryptophan, and the hashed arrow represents an absence or loss of activity of these biosynthetic genes.

Figure 6.

Attenuated extracellular superoxide production by E. faecalis strain TM1 and menB knockout of OG1RF is complemented by 1,4-dihydroxy-2-naphthoic acid. Spectra were obtained by ESR using DEPMPO as a spin trap (see Fig. 1).

A. 1,4-dihydroxy-2-naphthoic acid at 5 µM in 25 mM Tris at pH 7.4.

B. TM1 grown in BHI broth with 5 µM 1,4-dihydroxy-2-naphthoic acid (scale X 0.25 that in [A], [C], [D], [E] and [F]).

C. TM1 grown in BHI broth, washed, resuspended in 25 mM Tris at pH 7.4 with 1,4-dihydroxy-2-naphthoic acid added just before scan.

D. As in (C), with scan performed after incubating bacteria at 37°C for 15 min.

E. As in (C), except using menB knockout of OG1RF.

F. As in (E) with scan performed after incubating bacteria at 37°C for 15 min.

Terminal quinol oxidases attenuate extracellular superoxide production

Electron transport in mitochondria, chloroplasts and bacteria utilize quinone pools that undergo rapid redox cycling. These systems, however, typically produce little superoxide (Barja, 1999). Because quinol oxidases maintain oxidized quinone for electron transport, we hypothesized that E. faecalis counterparts may regulate extracellular superoxide production if superoxide was arising from DMK. E. faecalis contains two quinol oxidases: cytochrome bd (Pritchard and Wimpenny, 1978; Winstedt et al., 2000) and fumarate reductase (Aue and Deibel, 1967). Cytochrome bd is widely distributed among bacteria and contains two subunits (cydA and cydB) and three distinct cytochromes (b558, b595, and d) (Jünemann, 1997). E. faecalis cannot synthesize porphyrins, and so cytochrome bd remains an apo-cytochrome unless exogenous haematin is available (Pritchard and Wimpenny, 1978). When grown aerobically with added haematin, functional cytochrome bd forms, which initiates respiration and increases growth yields (Pritchard and Wimpenny, 1978).

To assess the role of cytochrome bd in extracellular superoxide production, brain–heart infusion (BHI) was supplemented with 8 µM haematin and OG1RF cultures aerated to maximize cytochrome bd expression. In the presence of exogenous haematin, the rate of extracellular superoxide production by wild-type OG1RF decreased by 92% (1.39 ± 0.14 nmol extracellular superoxide per min per 109 CFU), compared to growth without haematin (Fig. 7). Zinc protoporphyrin IX and protoporphyrin IX had no significant effect on the rate of extracellular superoxide production at 10 µM, indicating specificity of the effect on haematin that contains iron. To establish involvement of cytochrome bd in attenuating extracellular superoxide production, genes likely to encode cytochrome bd were identified within the E. faecalis genomic database. Putative cydA and cydB genes were identified and disrupted independently by allelic replacement. Loss of cytochrome bd after growth in the presence of exogenous haematin was confirmed by difference spectroscopy of purified cell membranes (Fig. 8). cydA and cydB knockouts produced extracellular superoxide at rates similar to wild-type OG1RF (13.3 ± 0.7 and 15.5 ± 1.0 nmol extracellular superoxide per min per 109 CFU for cydA and cydB mutants, respectively, compared with 17.2 ± 0.3 nmol extracellular superoxide per min per 109 CFU for OG1RF). Addition of 8 µM haematin to BHI did not alter significantly the rate of extracellular superoxide production for cydAB knockouts (16.6 ± 1.7 and 17.6 ± 1.3 nmol extracellular superoxide per min per 109 CFU for cydA and cydB mutants respectively). These findings indicated that functional cytochrome bd, when fully constituted by nutritional sources of iron-containing haematin, suppressed extracellular superoxide production.

Figure 7.

Superoxide production by E. faecalis strain OG1RF is attenuated by exogenous porphyrins.

A. OG1RF cells grown in BHI, with and without porphyrins, were resuspended in Hanks balanced salt solution to an O.D. of 0.2. AU and ferricytochrome c (20 µM) added. Reduction was measured over 5 min at 550 nm. Rates were calculated using linear regression and percent reduction found by dividing rates for bacteria grown in BHI with porphyrins by rates for bacteria grown in BHI alone. ESR-spin trapping of OG1RF in BHI supplemented with 10 µM (B), 1.0 µM (C), and 0.1 µM (D) haematin. Bacteria were washed and resuspended in Tris buffer (25 mM, pH 7.4) with glucose (3 mM) and DEPMPO (20 mM). Scans were performed immediately at ambient temperature as described in Fig. 1. Arrow points up-field and length indicates 10 Gauss.

Figure 8.

Light absorption difference spectra (dithionite-reduced minus air-oxidized) for E. faecalis membranes isolated from aerobically grown cells.

A. OG1RF grown in 8 µM haematin (0.67 mg protein ml−1) showing characteristic cytochrome b558 and d peaks; the b595 peak was not resolved.

B. OG1RF grown without haematin (1.7 mg protein ml−1).

C. OG1RF-derived cydA knockout grown with 8 µM haematin (1.2 mg protein ml−1).

D. OG1RF-derived cydB knockout grown with 8 µM haematin (1.5 mg protein ml−1).

Absence of peaks in scans in (B), (C) and (D) indicate lack of functional cytochrome bd. Vertical bar is absorption scale and peak wavelengths are in nanometres.

A second E. faecalis quinol oxidase is fumarate reductase (Aue and Deibel, 1967). This enzyme is constitutively expressed, consists of four subunits including a flavoprotein containing the catalytic site for fumarate reduction (frdA) and facilitates anaerobic respiration by catalysing the conversion of fumarate to succinate using menaquinol as an electron donor (van Hellemond and Tielens, 1994). Fumarate, like haematin, had a suppressive effect on the in vitro production of extracellular superoxide, reducing it by 75.0% (4.3 ± 0.85 nmol extracellular superoxide per min per 109 CFU) when added at 1 mM to OG1RF during ferricytochrome c assays.

Using the E. faecalis genomic database, we identified a gene with significant similarity to frdA in E. coli. This gene was disrupted in OG1RF by allelic replacement, and the knockout proved unable to convert greater than 5% of fumarate to succinate in overnight cultures supplemented with 10 mM fumarate. In comparison, wild-type OG1RF converted > 95% of exogenous fumarate to succinate. The frdA knockout still produced extracellular superoxide at rates similar to OG1RF (16.6 ± 0.9 nmol extracellular superoxide per min per 109 CFU for the frdA mutant compared with 17.2 ± 0.9 for OG1RF) indicating that this enzyme, by itself, was not a source of extracellular superoxide. However, in contrast with OG1RF, no attenuation of superoxide production was observed after addition of 1 mM fumarate to the frdA knockout during ferricytochrome c assays (15.6 ± 1.5 nmol extracellular superoxide per min per 109 CFU).

The relationship between terminal quinol oxidase activity and extracellular superoxide expression was further investigated using cydA, cydB and frdA mutants. Extracellular superoxide rates for cydA and cydB mutants decreased by 94.1% and 98% respectively, when assayed following addition of 0.4 mM fumarate (from 11.6 ± 0.5 to 1.1 ± 0.8 and from 11.4 ± 0.1 to 0.30 ± 0.35 nmol superoxide per min per 109 CFU respectively). Similarly, growth of the frdA mutant in 8 µM haematin decreased the rate of extracellular superoxide formation by 95.1% (from 12.1 ± 0.7 to 0.6 ± 0.1 nmol superoxide per min per 109 CFU) compared to growth of the mutant without haematin.

Ex vivo hydroxyl radical production by E. faecalis colonizing the intestine

To determine whether nutritive conditions in the intestinal tract were appropriate for production of extracellular superoxide from available oxygen, as might occur in proximity to the oxygenated epithelium, rats were inoculated via gastric intubation with streptomycin- and spectinomycin-resistant strains of E. faecalis OG1RF and the menB knockout (designated OG1RF-SS and PW18-SS respectively). These antibiotic-resistant derivatives showed no change in in vitro doubling times and had rates of extracellular superoxide production similar to parental strains. Spectinomycin and streptomycin antibiotics were included in the drinking water to eliminate competing intestinal flora and promote colonization. Five days post inoculation, the concentration of OG1RF-SS and PW18-SS in colon contents was similar, and colon contents were assayed by ESR using 5,5-dimethyl-1-pyrroline N-oxide (DMPO) as a spin trap (Fig. 9). A typical spectrum of colon contents harbouring OG1RF-SS showed DMPO-hydroxyl radical adduct arising from superoxide via metal-catalysed reactions and/or decomposition of a DMPO-superoxide adduct (Finkelstein et al., 1979). Substantially attenuated or no signals were observed for PW18-SS and sham-treated rats (Fig. 9). These results provide evidence that conditions exist in the intestine that favour the mode of E. faecalis respiration that results in production of the derivative hydroxyl radical.

Figure 9.

Production of hydroxyl radical by E. faecalis colonizing the rat intestine. ESR spectra were obtained as described in Fig. 1 using 40 mM DMPO as a spin trap.


B. OG1RF-SS upon addition of manganese superoxide dismutase (100 units ml−1).

C. PW18-SS.

D. PW18-SS with manganese superoxide dismutase (100 units ml−1).

E. Computer-simulation spectrum of (A) as a composite of the DMPO-superoxide adduct (47%) with hyperfine coupling constants of aN = 14.2 G; aHβ = 11.6 G; aHγ = 1.25 G; and DMPO-hydroxyl radical adduct (53%) aN = 14.9 G; aH = 14.9 G. Weanling rats on water containing streptomycin and spectinomycin at 500 mg l−1 for three days had gastric intubation with 1 × 108 CFU of E. faecalis strain OG1RF-SS or, as controls, 1 × 108 CFU of PW18-SS or an uninoculated sham. Five days later, 50–100 mg of colon contents were surgically retrieved, suspended in 25 mM Tris buffer at pH 7.4 with 20 mM DMPO, and immediately scanned as described in the Experimental procedures. Intestinal colonization with:

F. OG1RF-SS at 4 × 109 CFU gm−1 stool, showing hydroxyl radical adduct of DMPO.

G. PW18-SS at 7 × 109 CFU gm−1 stool with a markedly attenuated signal.

H. Uninoculated sham with no detectable signals. Arrow points upfield and length indicates 10 Gauss.


The univalent oxidation of reduced quinone, or quinol, produces semiquinone radicals that can react rapidly with molecular oxygen to form superoxide (Barja, 1999). One such reaction occurs during normal mitochondrial respiration, when a small percentage of electrons flowing through NADH:ubiquinone oxidoreductase and cytochrome bc1 leak out to form superoxide (Barja, 1999). In contrast, as shown schematically in Fig. 10, E. faecalis appears to have evolved a respiratory pathway that produces substantial amounts of superoxide depending on the availability of haematin as a cofactor for cytochrome bd, or on fumarate as a terminal electron acceptor. When the E. faecalis respiratory chain is impaired by lack of access to these nutrients, extracellular superoxide is generated by partially reduced DMK upon exposure to oxygen. As for mitochondria (Liu, 1999), these data support a proposal that extracellular superoxide is produced by E. faecalis through the respiratory chain due to rapid nonenzymatic reaction of semiquinone radicals of demethylmenaquinone with oxygen.

Figure 10.

Proposed model for extracellular superoxide production by E. faecalis. Reducing equivalents produced by sugar fermentation transfer from cytosol to cell membrane through oxidoreductases (for simplicity only NADH:quinone oxidoreductase is shown). Reduced demethylmenaquinone (DMKH2) can bind the terminal quinol oxidase fumarate reductase (Frd) through hydrophobic polypeptide anchors (FrdCD). Reducing equivalents pass through iron–sulphur clusters (FrdB) to an FAD-containing domain that harbours a dicarboxylate catalytic site (FrdA) where succinate is formed from fumarate. Alternatively, DMKH2 can be oxidized by a second terminal quinol oxidase in the form of cytochrome bd that can only be rendered catalytically active if exogenous haematin has been translocated to the cytosol, through an unidentified transporter. Haeme b558 in CydA efficiently oxidizes DMKH2 with reducing equivalents transferred to haemes b595 and d in CydB, where oxygen is reduced to water. DMKH2 oxidation releases protons into the extracellular compartment and helps establish a protonmotive force. In the absence of fumarate and haematin (A), extracellular superoxide is generated through the univalent reduction of oxygen by stabilized semiquinone radicals (thick arrows) at quinol binding sites on the two terminal quinol oxidases, or, potentially, at sites on other membrane-associated enzymes. Under neutral or acidic conditions, superoxide is readily dismuted into hydrogen peroxide and, in the presence of transition metals like iron, will be catalytically converted into hydroxyl radical.

These in vitro observations were extended in the intestinal environment using a rat colonization model. Ex vivo ESR measurements of wild-type E. faecalis strain OG1RF, colonizing the colon, produced hydroxyl radical whereas a menB mutant did not. These findings suggest insufficient fumarate or haematin is present in faeces of these animals which are fed a conventional diet that can suppress extracellular superoxide production by E. faecalis exposed to oxygen. What concentrations of fumarate or haematin might be required to modify superoxide production by E. faecalis in an intestinal environment remains to be determined.

In vivo measurement of free radical production in stool has not been reported. It remains likely, though, that free radicals are produced by E. faecalis in vivo as dissolved oxygen is ordinarily present in colon contents. For humans on a typical Western diet, the average concentration of dissolved oxygen in colon gas is 29.9 ± 7.2 (SEM) torr (Kirk, 1949). Similar valves have been reported for colon gas from dogs and germ-free, gnotobiotic and conventional rats (Steggerda, 1968; Bornside et al., 1976). This oxygen arises by passive diffusion from the colonic epithelium in which concentrations are similar to that of venous blood (approximately 40 torr). Potentially, E. faecalis in apposition to the colonic epithelium could generate extracellular superoxide through exposure to this source of oxygen. Transformation of superoxide to hydrogen peroxide and hydroxyl radical would then occur through spontaneous dismutation at the mildly acidic environment in the colon and by Fenton-type chemistry in the presence of appropriate metal catalysts (Fridovich, 1997; Henle and Linn, 1997; Miller and Britigan, 1997). Previous studies have shown that rat and human faeces produce hydroxyl radical (Babbs, 1990; Erhardt et al., 1997; Lund et al., 1999) but, until these findings, a potential microbiological source has not been identified.

Enterococci are the only intestinal bacteria known to produce extracellular superoxide (Huycke et al., 1996; Winters et al., 1998). Although the majority of aerotolerant prokaryotic organisms ordinarily produce small amounts of superoxide, this free radical production is located in the cytoplasm or periplasm. For E. coli the predominant source of intracellular superoxide is fumarate reductase (Imlay, 1995). Fumarate reductase is oriented in the cell membrane such that the flavoprotein and iron-sulphur protein subunits are located entirely within the cytoplasm (Iverson et al., 1999). The flavin redox centre is the site of electron leakage to oxygen (Imlay, 1995), so production of superoxide is intracellular. Using inverted membrane preparations, and under conditions of maximal expression, the E. coli fumarate reductase can produce as much as 20 nmol superoxide per min per 1012 cells. This mechanism, however, differs from that described for E. faecalis, in which superoxide is generated extracellularly, mutation of the fumarate reductase complex causes no decrease in superoxide production and the quantity of superoxide formed by live E. faecalis exceeds that for E. coli by approximately 1000-fold (17.2 nmol extracellular superoxide per min per 109 CFU for E. faecalis compared with 20 nmol intracellular superoxide per min per 1012 cells for E. coli).

A wide variety of carbohydrates are fermented by enterococci with 13 different sugars metabolized by all species (d-glucose, d-fructose, d-mannose, ribose, galactose, N-acetyl-glucosamine, amygdalin, arbutin, salicin, cellobiose, maltose, lactose and gentiobiose) (Devriese et al., 1993; Farrow et al., 1983). Sugar fermentation yields ATP and pyruvate through intermediary carbohydrate metabolism, with pyruvate ultimately catabolized by enterococci to CO2, acetate, ethanol, formate, acetoin, and/or lactate depending on growth conditions (Snoep et al., 1991). We found that extracellular superoxide production by live E. faecalis required a fermentable sugar (glucose, gentiobiose, cellobiose or d-mannose). A non-fermenting sugar (xylose) did not support superoxide production. These results suggest that the reducing equivalents for superoxide originate from sugar oxidation. Consistent with this hypothesis are observations that superoxide is produced by purified E. faecalis membranes, but only upon addition of NADH or l-lactate. Reducing equivalents in the form of NADH and l-lactate are formed through pyruvate metabolism by pyruvate dehydrogenase and l-lactate dehydrogenase respectively. Presumably, E. faecalis reoxidizes NADH and lactate through demethylmenaquinone via membrane-associated NADH or l-lactate oxidoreductases. ESR data obtained using purified E. faecalis membranes suggest the presence of these enzymes, but neither has yet to be characterized for this microorganism.

To withstand the self-imposed oxidative stress of robust superoxide production, E. faecalis necessarily expresses multiple antioxidant defences. For example, freely diffusable hydrogen peroxide that arises from the dismutation of superoxide can either be eliminated by a flavin-dependent, OxyR-regulated NADH peroxidase that utilizes a cysteine-sulphenic acid redox centre (Claiborne et al., 1992; Ross and Claiborne, 1997) or catalase that, like cytochrome bd, is dependent on exogenous haematin for activity (Pugh and Knowles, 1983). Dismutation of any intracellular superoxide is catalysed by an inducible manganese superoxide dismutase (Britton et al., 1978). In addition to these reactive oxygen species scavengers, E. faecalis is unusual as a Gram-positive bacterium in the ability to synthesize glutathione and express glutathione reductase (Patel et al., 1998; Sherrill and Fahey, 1998). Glutathione is a key reductant for several different classes of enzymes that can reduce hydrogen peroxide or hydroperoxides to form water and corresponding alcohols. Finally, superoxide-sensitive intracellular targets like [4Fe-4S]-containing aconitase (Fridovich, 1997; Gort and Imlay, 1998) are lacking because E. faecalis does not express the tricarboxylic acid cycle.

Nutritional conditions in the mammalian intestinal tract appear to favour a mode of respiration by E. faecalis that results in superoxide formation and, as shown by ESR, other reactive oxygen species such as hydroxyl radical. These findings suggest a novel microbiological source of oxidative stress for the intestinal epithelium. The implications are potentially profound for intestinal diseases such as colon cancer that may arise through mechanisms involving reactive oxygen species (Babbs, 1990). Studies are underway to determine the impact of free radical production by E. faecalis on colonic epithelial cells.

Experimental procedures

Bacterial strains and growth conditions

The plasmid-free E. faecalis strain OG1RF was previously shown to produce extracellular superoxide (Huycke et al., 1996). Enterococcus faecalis strains were cultured at 37°C in BHI (Difco) in closed, rotating tubes overnight or on BHI agar. E. coli strains were cultured in Luria–Bertani medium or Terrific Broth with aeration at 37°C (Sambrook et al., 1989). For plasmid and transposon selection, chloramphenicol and erythromycin were added to media at 50 µg ml−1. Chemically defined medium at pH 7.4 was filter-sterilized and consisted of (per litre): K2HPO4, 500 mg; KH2PO4, 120 mg; NaCl, 2 g; (NH4)2SO4, 1 g; N-[2-hydroxyethyl]piperazine-N′-[2-ethanesulfonic acid], 6 g; glucose, 10 g; thiamine, 10 mg; d-biotin, 10 mg; pantothenic acid, 10 nicotinic acid, 1 mg; riboflavin, 1 mg; folic acid, 0.1 mg; pyridoxine hydrochloride, 10 mg; uracil, 20 mg, thymine, 20 mg; adenine, 20 mg; xanthine, 20 mg; inosine, 1 mg; pyridoxal hydrochloride 10 mg; orotic acid, 1 mg; vitamin B12, 0.01 mg; paraaminobenzoic acid, 10 mg; lipoic acid, 1 mg; MgSO4, 250 mg; CaCl22H2O, 45 mg; ZnCl2, 4 mg; CuSO4, 10 mg; MnCl24H2O, 80 mg; FeCl36H2O, 1.3 mg; and amino acids at 20 mg l−1 (Sigma). Phenylalanine, tryptophan, tyrosine, haematin, NADH, l-lactate, zinc protoporphyrin IX, protoporphyrin IX and manganese superoxide dismutase (Sigma) were included in specific experiments as indicated.

Extracellular superoxide analysis

E. faecalis strains were grown in BHI to mid-logarithmic phase, washed, and resuspended in phosphate-buffered saline (PBS) (pH 7.4). The rates of extracellular superoxide production for whole bacteria were assayed at 37°C by measuring ferricytochrome c reduction as described previously (Huycke et al., 1996) after addition of glucose to a final concentration of 5 mM.

Transposon mutagenesis

E. faecalis was transformed with the transposon delivery vector pTV1(Ts) by electroporation (Weaver and Clewell, 1987). Growth at 42°C was used to limit plasmid replication, and erythromycin included at 50 ng ml−1 to induce Tn917 transposition. Erythromycin-resistant chloramphenicol-sensitive colonies were tested for decreased rate of ferricytochrome c reduction.

Electron spin resonance (ESR)-spin trapping

Superoxide and hydroxyl radical were measured by ESR by resuspending cells to 109 CFU ml−1 in 25 mM Tris (pH 7.4) with 7 mM glucose. 5-(diethoxyphosphoryl)-5-methyl-1-pyrroline-N-oxide (DEPMPO) at 20 mM (Frejaville et al., 1995), or 5,5-dimethyl-1-pyrroline N-oxide (DMPO) at 40 mM (Buettner and Mason, 1990), was added and spectra immediately recorded. Scans were performed at ambient temperature on a Bruker ER300E ESR spectrometer using the following parameters: 100 kHz field modulation, 20 mW microwave power, modulation width 1.0 Gauss, field sweep 100 Gauss per 84 s, and time constant 164 msec. Spectra were simulated using O'Zone Software epr simulation Program (Version 3.0). Enterococcus faecalis membranes were prepared by cell disruption and collection as described previously (Winstedt et al., 2000). Protein concentrations were determined using a Bio-Rad protein assay kit.

Recombinant DNA and homologous recombinational mutagenesis

DNA ligations, restriction endonuclease digestions, agarose gel electrophoresis and E. coli transformations were performed using standard techniques (Sambrook et al., 1989). Chromosomal and plasmid DNA were purified using commercially available reagents (Qiagen) according to the manufacturer's specifications. Electroporation of E. faecalis with purified plasmid DNA was performed as described previously (Shepard and Gilmore, 1995). Oligonucleotides were synthesized and polymerase chain reaction (PCR) amplification performed as reported previously (Winters et al., 1998). Chromosomal DNA flanking Tn917 sequences was amplified by inverse PCR (Triglia et al., 1988). Southern blot analyses were performed using Schleicher and Schuell nylon transfer membranes, according to the manufacturer's instructions. Internal fragments of targeted genes approximately 500–900 bp in length were generated by PCR (Table 1) and cloned into the shuttle vector p3erm that cannot replicate in E. faecalis (Callegan et al., 1999). Recombinant plasmids were amplified in E. coli XL2-Blue, electroporated into OG1RF, and transformants arising from a single cross-over event were selected using erythromycin. Typically, 5–15 colonies were obtained per electroporation, indicating a recombination frequency of approximately 10−8. Single cross-over recombinations were confirmed by Southern blotting.

Table 1. Primers for amplifying internal gene fragments in E. faecalis strain OG1RF by PCR.
GeneForward and reverse primers

DNA sequencing

Cloned PCR fragments in pBluescript (Stratagene) were purified from E. coli DH5α transformants, sequenced using the SequiTherm Longread Cycling Sequencing Kit (Epicentre) and analysed using a Model 4200 Longread Infrared Sequencer (LI-COR). Nucleotide sequences were edited and assembled using the Genetics Computer Group's sequence software package, version 9.0 (Devereux et al., 1984). Nucleotide and deduced amino acid sequences were compared with those in databases using blast search protocols (Altschul et al., 1997).

Chromatography and spectroscopy

Thin-layer chromatography of E. faecalis lipid extracts was performed as described previously, with modifications (Collins et al., 1980). First, 20 ml of Folch reagent was added to 200 mg of lyophilized bacteria, and the mixture stirred overnight. After removing cellular debris, 0.2 volume of water was added to the supernatant. The lipid phase was dried under nitrogen, resuspended in chloroform and resolved on silica gel plates (Sigma) using hexane:ether:acetic acid (80:20:1 [v:v:v]) as the mobile phase. Spots between 0.6 and 0.8 RF were located by iodine staining, scraped from undeveloped plates and re-extracted. Lipids were dissolved in CH2Cl2 and identified by fast atom bombardment mass spectroscopy (VG Model ZAB) using 3-nitrobenzyl alcohol as the matrix. Difference spectra for cell membranes (sodium dithionite reduced minus air oxidized) were recorded at room temperature using a Johnson Research Foundation DBS-3 scanning dual wavelength spectrophotometer, as previously described (Winstedt et al., 2000). Succinate was quantified enzymatically using a Biopharm kit.

Intestinal colonization model

E. faecalis strains OG1RF-SS and PW18-SS were spontaneously derived streptomycin- and spectinomycin-resistant mutants of OG1RF and the p3erm allelic replacement menB knockout respectively. The intestinal tracts of male weanling Wistar rats were colonized with these strains as described previously using streptomycin- and spectinomycin-containing water (each at 500 mg l−1) to facilitate colonization (Huycke et al., 1992). Enterococci in stool were enumerated using bile-esculin-azide agar plates (Difco).


We thank L. Henson, H. J. Harmon and M. Chilakapati for technical assistance. Sequencing of E. faecalis was accomplished with support from the National Institutes of Health. This work was supported in part by a Merit Review Program from the Department of Veterans Affairs (to M.M.H.) and the Frances Duffy Endowment.