Fumarate reductase is a major contributor to the generation of reactive oxygen species in the anaerobe Bacteroides fragilis

Abstract

Despite the detrimental role that endogenously generated reactive oxygen species (ROS) may play in bacteria exposed to aerobic environments, very few sources of ROS have been identified in vivo. Such studies are often precluded by the presence of efficient ROS-scavenging pathways, like those found in the aerotolerant anaerobe Bacteroides fragilis. Here we demonstrate that deletion of the genes encoding catalase (Kat), alkylhydroperoxide reductase (AhpC) and thioredoxin-dependent peroxidase (Tpx) strongly inhibits H₂O₂ detoxification in B. fragilis, thereby allowing for the quantification of ROS production. Exogenous fumarate significantly reduced H₂O₂ production in a ΔahpCΔkatΔtpx B. fragilis strain, as did deletion of fumarate reductase subunit c (frdC). Deletion of frdC also increased the aerotolerance of a strain lacking superoxide dismutase, indicating that fumarate reductase is a major contributor to ROS formation in B. fragilis exposed to oxygen.

Introduction

Life in an oxygenated world presents interesting and difficult challenges to organisms from a variety of environments. Due to its small size and unpaired outer-orbital electrons, molecular oxygen (O₂) can permeate cells and strip electrons from a wide variety of biomolecules (Fridovich, 1998, 1999; Imlay, 2003). These adventitious reactions result in the formation of partially reduced derivatives of O₂ appropriately known as reactive oxygen species (ROS). Such ROS include superoxide (${\text{O}}_{\text{2}}^{\text{–}}$), hydrogen peroxide (H₂O₂) and the hydroxyl radical, and their effects on the cell include inactivation of integral proteins and damage to DNA, eventually leading to cell death (Imlay, 2003). To survive in such a climate, an organism must attempt to limit the rate of endogenous ROS production and to detoxify ROS quickly upon formation. To this end, organisms such as Escherichia coli have evolved ROS-detoxifying enzymes such as superoxide dismutase (Sod) (McCord & Fridovich, 1969), alkylhydroperoxide reductase (AhpC) (Imlay, 2008; Parsonage et al., 2008) and catalases. E. coli can easily shift between anaerobic and aerobic environments, in part due to this robust response to oxidative stress, but many bacteria do not display such metabolic flexibility and are therefore relegated to niches containing little to no oxygen. These so called ‘anaerobes’ stop growing upon the introduction of O₂ and many die, but the nature of this sensitivity is still to be elucidated. One interesting and important model for studying the effects of oxygen on anaerobic bacteria is Bacteroides fragilis. Long regarded as a strict anaerobe, this mammalian commensal has been found to benefit from nanomolar concentrations of dissolved oxygen (≤0.05 % atmospheric), but cannot grow when O₂ levels are increased further (Baughn & Malamy, 2004). The bacterium can maintain some viability in room air (20.9 % atmospheric, 210 µM dissolved O₂ at 37 °C) even after several days of exposure, at least partially due to strong and sophisticated ROS detoxification systems. Superoxide dismutase and AhpC have been shown to protect the bacterium from O₂ exposure (Privalle & Gregory, 1979; Rocha & Smith, 1999), and KatB, the cytoplasmic catalase found in B. fragilis, has been shown to detoxify millimolar concentrations of hydrogen peroxide (Rocha et al., 1996). Additionally, B. fragilis encodes two putative rubrerythrins and several peroxidases, including a thioredoxin-dependent peroxidase (Tpx) shown to contribute to defence against organic peroxides (Herren et al., 2003). Much work has been done to characterize the components of this elaborate ROS detoxification response, but there are no studies investigating ROS production. We therefore sought to quantify endogenous ROS generation in aerated B. fragilis and to begin to identify the enzyme(s) responsible for their production. In this way, we might gain insights into the molecular mechanisms underpinning this organism’s inability to grow in room air.

Methods

Reagents.

Hydrogen peroxide, horseradish peroxidase (HRP), antibiotics and fumaric acid were purchased from Sigma, and Amplex Red (AR) was obtained from Invitrogen. Restriction enzymes were purchased from New England Biolabs.

Growth conditions.

Anaerobiosis was maintained by using a Coy anaerobic chamber containing 85 % nitrogen, 10 % hydrogen and 5 % carbon dioxide. B. fragilis was grown in brain heart infusion broth supplemented with 0.5 % yeast extract and 15 µg haematin ml⁻¹ (BHIS), or anaerobic minimal medium (AMM) containing 0.5 % glucose, as previously described (Baughn & Malamy, 2002). In some cases super anaerobic minimal medium (SAMM) plates containing 150 µg haemin ml⁻¹ were used. Because all the strains employed are thyA mutants, thymine was added to 50 µg ml⁻¹. Gentamicin (50 µg ml⁻¹), rifampicin (50 µg ml⁻¹), erythromycin (8 µg ml⁻¹), trimethoprim (80 µg ml⁻¹) and tetracycline (2.5 µg ml⁻¹) were added to plates where appropriate. E. coli was grown aerobically in L broth, and chloramphenicol (25 µg ml⁻¹) and tetracycline (10 µg ml⁻¹) were added where appropriate.

Strains and plasmids.

Strains and plasmids used in this study are listed in Table 1. E. coli strain DH5α was used for cloning, and strain HB101/pRK231 was used for mobilization of plasmids from DH5α to B. fragilis recipient strains (Godoy et al., 1993). DH5α was made competent for transformation through the use of the RbCl method previously described (Hanahan et al., 1991).

Table 1. Bacterial strains and plasmids used in this study.

Abbreviations: Cm^R, Tet^R, Tp^R, chloramphenicol, tetracycline and trimethoprim resistance, respectively. Amp^S, ampicillin sensitivity.

Strain or plasmid	Relevant characteristics	Reference or source
B. fragilis strains
ADB77	TM4000 ΔthyA TpR	Baughn & Malamy (2002)
ADB247	ADB77 ΔfrdC247	Baughn & Malamy (2003)
MBD616	TM4000thyA2Δsod	Laboratory stock
ADB247/616	ADB247Δsod	A. D. Baughn, University of Minnesota
BM28	ADB77ΔahpC	This work
BM50	BM28Δkat	This work
BM95	ADB77Δtpx	This work
BM105	BM50Δtpx	This work
BM112	BM105ΔfrdC	This work
E. coli strains
DH5α	λ Nonlysogen	Woodcock et al. (1989)
HB101	rpsL20, host strain for pRK231	Thompson & Malamy (1990)
Plasmids
pRK231	AmpS derivative of RP4, TetR Tra+	Godoy et al. (1993)
pYT102	p15A ori, CmR, RP4 oriT, B. fragilis suicide vector containing B. fragilis thyA TetR	Tang & Malamy (2000)
pADB242	pYT102 derivative, 0.35 kb BamHI–HindIII fragment replaced by 18 bp BamHI–HindIII fragment from pCR2.1-TOPO (Invitrogen) CmR	Baughn & Malamy (2003)
pYT102SD	pYT102 containing Δsod	This work
pADB247	pADB242 derivative with ΔfrdC247 allele	Baughn & Malamy (2003)
pADB261m7	pADB261 frdA : : Tn1000 (frdC+)	Baughn & Malamy (2003)
pADB242ΔkatB	pADB242 containing ΔkatB	This work
pADB242ΔahpC	pADB242 containing ΔahpC	This work
pADB242Δtpx	pADB242 containing Δtpx	This work

DNA manipulation.

Primers used in this study are listed in Table 2. Primers were designed using the B. fragilis NCTC9343 annotated sequence found on the Pedant3 webpage (http://pedant.gsf.de) and synthesized by IDT (Iowa City, IA). Genomic DNA was amplified using HotStarTaq Master Mix (Invitrogen). Plasmid and PCR product purifications were performed with QIAprep spin columns (Qiagen). Where indicated, DNA was digested with restriction enzymes purchased from New England Biolabs. Ligations were performed using T4 DNA Ligase from Invitrogen.

Table 2. Primers used in this study.

Primer	Region of homology	Sequence (5′–3′)
1843	pADB242	CCCATCGGTGATGTCGGC
61RAB	pADB242	GGCGCGCCGTAAGGAAAGTGGCTCTCAG
BAM56	ahpC1	CGATGGATCCGCAAAGGTAGGGTGAAG
BAM57	ahpC2	TGACCCATGGCCCTTTACGTCTTCGC
BAM58	ahpC3	GGCACCATGGCGAAGCAACCCTGAAAC
BAM59	ahpC4	AGGTGCGGCCGCGATGGAAGTTTCCGCAC
BAM60	ahpC5	CGACTCTTAGGTACTGG
BAM68	kat1	ATATGGATCCATCCCCCTGTGGTA
BAM69	kat2	ATATCCATGGTTAGCGCTACGCATGTT
BAM70	kat3	ATATCCATGGTTGAAGGTATCGGCTTC
BAM71	kat4	ATATGCGGCCGCCTGGGCATTTCTTTG
BAM72	kat5	CTGCACTTTACGCACTG
BAM216	tpx1	ACATGGATCCGCTTCATTAATCTGG
BAM217	tpx2	CAATCCATGGGTGGCATCGAATTTCG
BAM218	tpx3	ACATCCATGGCATGAAAGCTACCGAAG
BAM219	tpx4	CAATGCGGCCGCAAACATCGCTTTAAAG

Strain construction.

All in-frame deletions of B. fragilis genes were created using a two-step double-crossover technique (Tang et al., 1999). In order to build a deletion construct for ahpC, an N-terminal fragment was amplified using primers BAM56 and BAM57, and a C-terminal fragment was amplified using primers BAM58 and BAM59. Purified PCR products were digested with BamHI and NcoI (N-terminal) or NotI and NcoI (C-terminal), and ligated via a three-way reaction with pADB242 that had been digested with BamHI and NotI to create pADB242ΔahpC. The insert was verified via PCR with primers BAM68 and BAM71.

The suicide plasmid pADB242ΔahpC was delivered to the recipient B. fragilis strain as previously described (Thompson & Malamy, 1990). Tetracycline-resistant colonies were screened for the appropriate cointegrant event using primers 1843 and BAM59. Isolates demonstrating recombination at the ahpC locus were grown overnight in BHIS with thymine to allow for recombination events leading to resolution of the disrupting plasmid. The presence of the thyA gene on pADB242 sensitizes cointegrants to trimethoprim, while those resolvants that have excised the plasmid are trimethoprim-resistant. For this reason, the culture was plated to SAMM containing glucose, thymine and trimethoprim. Colonies arising after 3–4 days were purified and then screened for tetracycline sensitivity on the appropriate BHIS+thymine plates. Tetracycline-sensitive colonies were then used as template in a PCR with primers BAM59 and BAM60 to identify ΔahpC clones.

The deletion construct for sod was created by PCR-amplifying a region consisting of 570 bp of sod 5′ upstream sequence and 133 bp of N-terminal coding sequence to create the upstream fragment. A fragment of 150 bp of sod C-terminal coding sequence and 539 bp of downstream sequence was amplified for the downstream fragment. These pieces were ligated into pYT102 via the BamHI and HindIII cloning sites.

The deletion construct for katB was created by PCR-amplifying an N-terminal fragment with primers BAM68 and BAM69 and a C-terminal fragment with primers BAM70 and BAM71. Fragments were again digested with BamHI and NcoI (N-terminal) or NotI and NcoI (C-terminal), and ligated to the BamHI/NotI-digested pADB242 to create pADB242Δkat. The procedure described above was used to create cointegrants, which were screened by PCR with primers 61RAB and BAM72. Resolvants were screened with primers BAM68 and BAM71.

pADB242Δtrx was created by PCR-amplifying an N-terminal fragment with primers BAM216 and BAM217 and a C-terminal fragment with BAM218 and BAM219. Digestions and ligations to pADB242 were performed as above. Cointegrants were screened with primers 61RAB and BAM216 as well as 1843 and BAM219. Resolvants were screened with primers BAM216 and BAM219.

H₂O₂ detection.

The protocol for measuring H₂O₂ scavenging was adapted from the method of Seaver & Imlay (2001). Briefly, solutions of HRP and AR reagent were made in PBS (pH 7.2) (per litre: 8 g NaCl, 0.2 g KCl, 1.44 g Na₂PO₄, 0.24 g KH₂PO₄) to concentrations of 20 and 56 µg ml⁻¹, respectively. An H₂O₂ stock was made in PBS (pH 7.2) from a 30 % solution and the concentration was determined by measuring the A₂₄₀ of the solution in a spectrophotometer (ϵ = 43.6 M⁻¹ cm⁻¹). This solution was further diluted to give a 200 µM H₂O₂ stock solution. Aliquots were added to 500 µl HRP mixed with 500 µl AR in 4 ml polystyrene cuvettes and diluted to a final volume of 2 ml with PBS. These samples were briefly shaken by hand and placed in an Aminco Bowman Series 2 luminescence spectrophotometer for reading with an excitation wavelength set to 563 nm (band pass = 1) and an emission wavelength set to 587 nm (band pass = 4). Fluorescence readings were recorded and used to construct a standard curve for H₂O₂ concentration. All experiments were conducted in the presence of oxygen, except where indicated. Optical densities were measured using a Beckman DU 640 spectrophotometer and quartz cuvettes with a 1 cm path length.

H₂O₂ scavenging.

Strains were first grown anaerobically to mid-exponential phase in BHIS or AMM. Cultures were adjusted to OD₆₀₀ = 0.1 in BHIS or AMM lacking cysteine, placed in foam-stoppered flasks, and transferred to 37 °C under room air (21 % oxygen). Samples were shaken for 1 h at 250 r.p.m. then centrifuged for 10 min at 2000 g. Supernatants were removed, and the cell pellets were washed twice with an equal volume of PBS, as the growth medium was found to significantly reduce the sensitivity of the assay. Following the second wash, cell pellets were resuspended in 10 ml PBS and transferred to tubes for testing. Aliquots of H₂O₂ stock were added to the 10 ml cultures to give a starting concentration of 2.5–5 µM. Immediately after the addition of peroxide, the tubes were quickly capped and upended once to mix. A 1.1 ml volume was immediately removed to a 1.5 ml microcentrifuge tube and spun for approximately 10 s at 6000 r.p.m. in a microcentrifuge (model SD110, Clover Laboratories). A 1 ml volume of supernatant was removed to a 4 ml polystyrene cuvette (VWR), 500 µl HRP and 500 µl AR solutions were added, and the fluorescence was read. For anaerobic peroxide-scavenging experiments, cultures were transferred to conical tubes in the anaerobic chamber and sealed before removal for centrifugation. Supernatants were poured off in the anaerobic chamber, and cells were washed and ultimately resuspended in pre-reduced PBS with thymine and 0.5 % glucose. These suspensions were transferred to Hungate tubes before removal from the chamber. H₂O₂ additions and sample removal were performed by inserting a needle through the rubber septum to minimize oxygen exposure.

ROS generation.

Strains were grown for 2 days at 37 °C in 5 ml AMM. Aliquots of these dense cultures were then used to inoculate 10 ml AMM. Cultures were shaken anaerobically at 100 r.p.m. in 50 ml flasks and grown for several generations. When the OD₆₀₀ reached approximately 0.3, cultures were transferred anaerobically to conical screw-capped tubes, sealed, and centrifuged for 5 min at 2000 g. After transferring tubes to the anaerobic chamber and decanting the supernatants, cell pellets were resuspended in 10 ml AMM lacking cysteine and transferred to 125 ml foam-stoppered flasks. These flasks were shaken at 250 r.p.m. in room air at 37 °C for 1 h. Cultures were again transferred to conical screw-capped tubes and centrifuged as before. Cell pellets were washed twice with PBS and then resuspended in 20 ml PBS+0.05 % fresh glucose and thymine to OD₆₀₀ = 0.1. Flasks were shaken at 250 r.p.m. in room air at 37 °C. Samples of 1.1 ml volume were drawn at various time points as described above to determine H₂O₂ concentrations in the supernatants.

O₂ sensitivity.

Strains were grown overnight in BHIS [with or without erythromycin (±erm)] under anaerobic conditions. Aliquots of these starter cultures were used to inoculate 10 ml BHIS, and cultures were grown for several generations. When cultures reached mid-exponential phase they were transferred to 15 ml conical tubes in the anaerobic chamber. The tubes were capped, removed from the chamber, and centrifuged for 5 min at 2000 g in a tabletop centrifuge. Supernatants were poured off, and the pellets were resuspended in 10 ml oxygenated PBS (pH 7.2)+0.5 % glucose and thymine. These suspensions were transferred to 125 ml flasks stoppered with foam, and shaken at 250 r.p.m. and 37 °C under room air. Samples were taken over time and diluted in PBS. Ten microlitre volumes of diluted samples were spotted to BHIS+thymine plates (±erm) that had been placed in the anaerobic chamber overnight to reduce. After 24–48 h, c.f.u. were enumerated under a stage microscope.

Results

A B. fragilis strain missing Kat, AhpC and Tpx is severely compromised in its ability to scavenge H₂O₂

Some redox-active enzymes, including flavoproteins, can adventitiously donate electrons to O₂, thus generating superoxide or H₂O₂. The rate at which these ROS are generated depends on the accessibility of the flavin to the solvent as well as the abundance of the enzyme (Imlay, 1995; Messner & Imlay, 2002). In order to quantify the rate of this ROS formation in vivo, it is necessary to identify and inactivate the major hydrogen peroxide-scavenging enzymes. In a strain lacking such detoxifiers, H₂O₂ produced intracellularly will leak across the cell membrane and accumulate in the extracellular milieu, thus allowing quantification with AR (Seaver & Imlay, 2001). Unlike H₂O₂, superoxide cannot cross membranes at physiological pHs, but if strains encoding Sod are utilized it is converted enzymically to H₂O₂. In such strains, therefore, endogenous sources of both H₂O₂ and superoxide can be assayed using AR, thereby giving an overall picture of ROS formation.

In E. coli, deletion of ahpC and the genes encoding two catalases, katE and katG, allows for monitoring of H₂O₂ production by decreasing peroxide-scavenging rates to less than 5 % of wild-type (Seaver & Imlay, 2001). Because B. fragilis appears to encode only one catalase (katB), we hypothesized that a ΔahpCΔkatB mutant should permit quantification of endogenous ROS production. However, when cells were shaken under room air in rich medium and then assayed for H₂O₂ consumption in PBS, the ΔahpCΔkatB strain retained significant scavenging activity (Fig. 1a), potentially interfering with ROS quantification. In an attempt to identify the gene product(s) responsible for the residual scavenging activity, we deleted other putative peroxidases in the ΔahpCΔkat background. While most deletions did not appear to impair scavenging activity further, deletion of a gene encoding a putative thioredoxin-dependent peroxidase (Tpx) resulted in a strain (ΔahpCΔkatΔtpx) with approximately 20 % of the peroxide-scavenging activity of the wild-type when the cells were grown in BHIS (Fig. 1a).

Fig. 1.

H₂O₂ scavenging rates for B. fragilis cells grown in rich medium (a) and minimal medium (b). Cells were washed and resuspended in PBS to OD₆₀₀ 0.1. Aliquots of H₂O₂ were added to suspensions to start the assay and samples were taken over time. After a brief centrifugation, supernatants were assayed for H₂O₂ concentration using the AR protocol. (a) PBS (▪), ADB77 (wild-type, ▴, dashed line), BM50 (ΔahpCΔkat, ▾, dotted line), BM95 (Δtpx, •), BM105 (ΔahpCΔkatΔtpx, □). Scavenging assays were done in the presence of oxygen. (b) BM105 was assayed in the absence (solid line) and presence (dashed line) of oxygen. Shown are the mean rates of peroxide scavenging for three experiments±sem.

We also tested the H₂O₂-scavenging activity of cells grown in minimal medium (AMM), and investigated the effect of oxygen on this activity. Samples of the ΔahpCΔkatΔtpx strain were grown anaerobically, washed anaerobically, and resuspended to OD₆₀₀ 0.1 in 10 ml pre-reduced PBS in the anaerobic chamber. An aliquot of H₂O₂ was added to bring the initial concentration to ~2.5 µM, while excluding oxygen. Under such conditions, these cells were capable of scavenging H₂O₂ at a rate of ~155 nM min ⁻¹ (Fig. 1b). However, when these cells were first shaken for 1 h under room air in AMM lacking cysteine and then suspended in aerated PBS, H₂O₂-scavenging activity could no longer be detected (Fig. 1b). In fact, the H₂O₂ concentration at the start of the assay was significantly higher in these samples than in the anaerobically prepared cells, possibly due to the production of ROS during the preparation of the sample. Exposure to oxygen was clearly interfering with H₂O₂ detoxification, and we reasoned that preparation of the ΔahpCΔkatΔtpx strain under these conditions might allow us to quantify endogenously generated H₂O₂.

Fumarate reductase is a significant contributor to ROS formation when B. fragilis is aerated

When the B. fragilis ΔahpCΔkatΔtpx strain was prepared as above, adjusted to OD₆₀₀ 0.1 in 10 ml PBS containing thymine and glucose, and shaken in room air, we detected the accumulation of H₂O₂ in the buffer at a rate of 36±6 nM min⁻¹ (Fig. 2). Given such a rate, we would expect that this strain would experience significant oxidative stress in less than 30 min, as substantial damage to E. coli metabolism occurs even in the face of submicromolar H₂O₂ concentrations (Park et al., 2005).

Fig. 2.

Disrupting the fumarate reductase complex reduces the rate of ROS formation in B. fragilis. Cells grown to exponential phase were centrifuged, resuspended in AMM lacking cysteine, and shaken for 1 h at 37 °C under room air. Cells were again centrifuged, washed and resuspended in PBS+thymine to OD₆₀₀ 0.1. Freshly prepared glucose was added to 0.05 % just before the start of assay. Suspensions were shaken at 37 °C under room air throughout the course of the assay. Samples were taken over time and cells were centrifuged. Supernatants were assayed for H₂O₂ using the AR protocol. BM105 (ΔahpCΔkatΔtpx, ▪), BM112 (ΔahpCΔkatΔtpxΔfrdC, ▴, dashed line). Shown are the means of three assays±sem.

Endogenous sources of ROS have been identified in E. coli (Korshunov & Imlay, 2010), and among these, fumarate reductase (Frd) has been described as a potent generator of both superoxide and H₂O₂ due to its solvent-exposed flavin cofactor (Imlay, 1995; Korshunov & Imlay, 2006; Messner & Imlay, 2002). With this in mind, we made a deletion of the frdC gene in the B. fragilis ΔahpCΔkatΔtpx background to test whether loss of fumarate reductase activity significantly reduced the rate of ROS formation in cells challenged under room air. As illustrated in Fig. 2, we detected the accumulation of H₂O₂ at a rate of 19±1 nM min⁻¹ in strain BM112 (ΔahpCΔkatΔtpxΔfrdC), indicating that disruption of the fumarate reductase complex reduced ROS generation by approximately 47 % when B. fragilis was exposed to room air.

frdC mutants are much smaller than wild-type, and an OD₆₀₀ 0.1 in these experiments corresponded to ~4×10⁷ c.f.u. ml⁻¹ for ΔahpCΔkatΔtpx and ~2×10⁸ c.f.u. ml⁻¹ for ΔahpCΔkatΔtpxΔfrdC. To make sure that these discrepancies were not affecting our results, cells of both strains were sonicated and assayed for intracellular protein content. At this cell density, ΔahpCΔkatΔtpx and ΔahpCΔkatΔtpxΔfrdC gave mean protein contents of 4.7 and 5.4 µg protein ml⁻¹, respectively, and normalizing ROS generation rates to protein content gave similar results to those reported above. Additionally, these strains were found to scavenge peroxide at indistinguishable rates in both the presence and absence of oxygen (data not shown).

Fumarate inhibits ROS formation

If fumarate reductase is a source of superoxide or H₂O₂ when cells are aerated, then addition of fumarate should reduce ROS formation by competing with oxygen for the electrons carried on the Frd flavin group. To test this, BM105 (ΔahpCΔkatΔtpx) and BM112 (ΔahpCΔkatΔtpxΔfrdC) were prepared as above for the ROS accumulation assay. Just prior to the start of the assay, cultures were split into two flasks, and fumarate was added exogenously to one flask per strain. As shown in Fig. 3, addition of fumarate to BM105 reduced ROS production to a rate similar to that of the ΔfrdC strain, indicating that fumarate was partially inhibiting the adventitious reduction of oxygen in this strain. Fumarate also appeared to reduce ROS production in BM112 slightly. While this reduction was not statistically significant, it may indicate that fumarate is inhibiting another ROS-generating enzyme.

Fig. 3.

Exogenous fumarate inhibits ROS formation. Strains BM105 (ΔahpCΔkatΔtpx) and BM112 (ΔahpCΔkatΔtpxΔfrdC) were prepared as in Fig. 2, with cell suspensions divided into two flasks each. Just prior to sampling, 50 mM fumarate was added to one flask per strain. The AR protocol was then used to determine the H₂O₂ concentration over time. Rates are recorded as nM H₂O₂ generated min⁻¹ for cells suspended to OD₆₀₀ 0.1 in 10 ml PBS. Shown are the means of three assays±sem. A one-way analysis of variance (ANOVA) revealed significant differences between strains (P = 0.0154). Asterisks denote significant differences between means (P<0.05) in a Newman–Keuls multiple comparison post test.

Deleting frdC from a superoxide dismutase mutant partially restores aerotolerance

Given that the B. fragilis fumarate reductase plays such a major role in ROS production, we hypothesized that deleting frdC might increase aerotolerance. However, no significant differences in viability were seen when BM105 (ΔahpCΔkatΔtpx), BM112 (ΔahpCΔkatΔtpxΔfrdC), ADB247 (ΔfrdC) and ADB77 (wild-type) were exposed to room air for up to 9 h in PBS containing thymine and glucose (data not shown). This indicated that an inability to scavenge H₂O₂ did not sensitize B. fragilis to aerobic conditions during the course of these experiments. However, Fig. 4(a) shows that a Δsod mutant lost greater than three logs of viability under these conditions, suggesting that, at least initially, the major ROS generated by aerated B. fragilis is superoxide. A ΔsodΔfrdC strain (ADB247/616) showed reduced oxygen sensitivity, losing only about one log of viability during this time period, indicating that deletion of frdC restores some aerotolerance to a Δsod strain. Anaerobic controls showed that the number of c.f.u. increased slightly for the ΔsodΔfrdC strain during the course of the experiment, while the number of c.f.u. for the Δsod strain decreased slightly, resulting in an approximately threefold difference (data not shown). In separate experiments, we found that introduction of a wild-type copy of frdC in trans resensitized the ΔsodΔfrdC strain to room air (Fig. 4b), while a strain carrying an empty vector maintained significantly higher viability.

Fig. 4.

Survival of B. fragilis strains under room air. Anaerobically grown cells were suspended in PBS with 0.5 % glucose and thymine, and shaken in room air at 37 °C. Samples were taken over time, diluted, and plated to pre-reduced BHIS+thymine (a) or BHIS+thymine+erm (b). c.f.u. were enumerated after 24–48 h. (a) ADB77 (wild-type, ▾), ADB247 (ΔfrdC, ▴), MBD616 (Δsod, ▪, dotted line), ADB247/616 (ΔfrdCΔsod, •). Shown are the means of three experiments±sem. A Student’s paired t test analysing the percentages of viable c.f.u. remaining at the 6 h time point indicated a significant difference between MBD616 and ADB247/616 with P<0.5. (b) Percentage of viable c.f.u. remaining after 6 h exposure to room air for ADB247/616 (pJST61) and ADB247/616 (pADB261m7). Shown are the means of nine experiments±sem. A Student’s paired t test indicated P<0.05.

Discussion

In order to quantify ROS production rates in B. fragilis, we first constructed a strain that was incapable of scavenging any H₂O₂ formed when cells were aerated. We were surprised to uncover a potential role for Tpx in H₂O₂ scavenging, as this enzyme had previously been shown to play a role in protection against organic peroxides, but it did not appear to contribute significantly to protection against H₂O₂-mediated growth inhibition (Herren et al., 2003). Indeed, deletion of tpx alone did not affect the ability of B. fragilis to scavenge micromolar concentrations of H₂O₂ when applied exogenously. It is possible that deletion of tpx led to increased expression of Kat or AhpC, as ahpC mutants of E. coli have been shown to increase transcription of katG via activation of the transcriptional regulator OxyR (Seaver & Imlay, 2001). Alternatively, Tpx may have a lower affinity for reductant than AhpC, thus masking its activity when AhpC is present. Clearly the role of Tpx is complex and warrants further investigation.

The residual H₂O₂-scavenging activity exhibited by ΔahpCΔkatΔtpx in BHIS was lost when cells were shaken under room air in minimal medium lacking cysteine. By removing cysteine, which acts as a redox buffer in this medium, cells quickly become saturated by O₂. Presumably, O₂ is participating in reactions that consume reductant, like ROS generation or the cytochrome bd oxidase-mediated reduction of O₂ to water, and thus is competing with H₂O₂ detoxification reactions for substrate. Under such conditions, therefore, this strain cannot effectively scavenge H₂O₂, thus allowing us to dissect the pathways contributing to endogenous ROS generation in B. fragilis.

Fumarate reductase has been found to produce ROS in anaerobically grown E. coli shaken in room air (Korshunov & Imlay, 2010). We demonstrate in this work that it makes a significant contribution to ROS production in B. fragilis as well, most likely by directly reducing oxygen via its flavin cofactor. The finding that deletion of frdC restored aerotolerance to a Δsod strain suggests that in vivo under aerobic conditions Frd is generating superoxide rather than H₂O₂. Interestingly, earlier work has shown that many of the tricarboxylic acid cycle genes of B. fragilis are upregulated five- to 28-fold when the organism is aerated (Sund et al., 2008). However, transcription of frd is downregulated approximately threefold, perhaps suggesting that an adaptive response is taking place to reduce ROS generation.

The finding that fumarate slowed ROS production slightly in the ΔahpCΔkatΔtpxΔfrdC strain suggests that another ROS-generating enzyme might also utilize fumarate as its natural substrate. One such candidate would be NadB, a key component of the NAD⁺ biosynthetic pathway. This enzyme has also been shown to participate in ROS formation in E. coli (Korshunov & Imlay, 2010) and to be inhibited by fumarate. Experiments are under way to test the potential contribution of NadB to ROS production in B. fragilis.

Previous work with Bacteroides thetaiotaomicron has suggested that this organism (and, by extension, B. fragilis) might generate ROS intracellularly at a faster rate than E. coli (Pan & Imlay, 2001), potentially explaining its inability to grow aerobically. We have shown that when B. fragilis was shaken in room air, the extracellular concentration of H₂O₂ increased at a rate of 36 nM min⁻¹ for cells suspended to OD₆₀₀ 0.1. To convert this rate to reflect intracellular concentrations, we took advantage of the fact that the intracellular concentration of protein in E. coli is approximately 330 mg (ml cytoplasm)⁻¹ (J. Imlay, personal communication). Assuming a similar concentration in B. fragilis, 1 ml of cells at OD₆₀₀ 0.1 contains ~5 µg cytoplasmic protein and thus represents 15 nl of cytoplasmic volume. Converting the extracellular H₂O₂ accumulation rates thus shows intracellular rates of ROS formation to be 40 µM s⁻¹. An E. coli strain missing catalase and peroxidase activity (but encoding Sod) accumulates H₂O₂ intracellularly at a rate of 15 µM s⁻¹ (Imlay, 2008). While a direct comparison between these organisms is complicated by differences in growth media and by the fact that E. coli is actively growing under the assay conditions while B. fragilis is merely surviving, these estimates suggest that B. fragilis is producing ROS at a rate that is only ~2.5-fold faster than E. coli, a seemingly modest increase. Such an increase in ROS generation would not be without consequence, however, as even a threefold increase in the rate of superoxide production can significantly reduce the steady-state activity of crucial dehydratases in E. coli (Gort & Imlay, 1998). This suggests that the rate at which an organism generates ROS may play an important role in its ability to cope with aerobic environments.

Abbreviations:

AR

Amplex Red

HRP

horseradish peroxidase

ROS

reactive oxygen species