Endogenous superoxide is a key effector of the oxygen sensitivity of a model obligate anaerobe

Significance

Microbes display profound differences in their tolerance for oxygen, and this trait organizes the structure of many microbial communities. However, the molecular basis of oxygen sensitivity is not well understood. In this study we determined that Bacteroides thetaiotaomicron, an abundant member of the human intestinal flora, is incapacitated by superoxide stress when it enters a fully oxic environment. The key difference from oxygen-tolerant bacteria lies not in its defensive systems, nor in the nature of the affected enzymes, but in the rate of endogenous oxidant formation. Anaerobes thrive in oxygen-poor environments because they deploy low-potential electron-transfer pathways; these results suggest that an ancillary effect is the reactivity of these pathways with oxygen, thereby generating enough reactive oxygen species to preclude oxic growth.

Abstract

It has been unclear whether superoxide and/or hydrogen peroxide play important roles in the phenomenon of obligate anaerobiosis. This question was explored using Bacteroides thetaiotaomicron, a major fermentative bacterium in the human gastrointestinal tract. Aeration inactivated two enzyme families—[4Fe-4S] dehydratases and nonredox mononuclear iron enzymes—whose homologs, in contrast, remain active in aerobic Escherichia coli. Inactivation-rate measurements of one such enzyme, B. thetaiotaomicron fumarase, showed that it is no more intrinsically sensitive to oxidants than is an E. coli fumarase. Indeed, when the E. coli enzymes were expressed in B. thetaiotaomicron, they no longer could tolerate aeration; conversely, the B. thetaiotaomicron enzymes maintained full activity when expressed in aerobic E. coli. Thus, the aerobic inactivation of the B. thetaiotaomicron enzymes is a feature of their intracellular environment rather than of the enzymes themselves. B. thetaiotaomicron possesses superoxide dismutase and peroxidases, and it can repair damaged enzymes. However, measurements confirmed that the rate of reactive oxygen species production inside aerated B. thetaiotaomicron is far higher than in E. coli. Analysis of the damaged enzymes recovered from aerated B. thetaiotaomicron suggested that they had been inactivated by superoxide rather than by hydrogen peroxide. Accordingly, overproduction of superoxide dismutase substantially protected the enzymes from aeration. We conclude that when this anaerobe encounters oxygen, its internal superoxide levels rise high enough to inactivate key catabolic and biosynthetic enzymes. Superoxide thus comprises a major element of the oxygen sensitivity of this anaerobe. The extent to which molecular oxygen exerts additional direct effects remains to be determined.

The phenomenon of obligate anaerobiosis is the most obvious natural manifestation of oxidative stress. Many microorganisms can only grow in anoxic places. This restriction is a dominant factor in the organization of microbial ecosystems in soil and gut, where respiring organisms help to shield the majority of anaerobes from the encroachment of oxygen. In 1971, McCord et al. (1) published a survey of scavenging enzymes that implied a possible cause of obligate anaerobiosis. In contrast to oxygen-tolerant microbes, the anaerobes that they examined contained little or no superoxide dismutase (SOD) or catalase—which suggested that, upon aeration, these microbes would be poisoned by superoxide (O₂⁻) or hydrogen peroxide (H₂O₂). The table that was published has been widely circulated, and this correlation is still cited in textbooks as a likely explanation for obligate anaerobiosis.

In 1986, Carlioz and Touati (2) performed a key experimental test of the idea, by deleting the SOD genes from the facultative bacterium Escherichia coli. The resultant mutant grew at normal rates in the absence of oxygen, but upon aeration it exhibited a set of severe biosynthetic and catabolic defects. These included deficiencies in the biosynthesis of eight amino acids plus an inability to use TCA-cycle substrates as carbon sources. Analogous mutants that lacked catalase and peroxidase were generated much later, and these mutants exhibited many of the same defects (3, 4). Thus, these phenotypes confirmed the potential toxicity of reactive oxygen species (ROS), and they broadly supported the idea that anaerobes might be poisoned by endogenous oxidants.

The metabolic defects of the mutant E. coli strains were subsequently traced to damage to two types of enzymes: dehydratases that depend upon iron-sulfur clusters and nonredox enzymes that employ a single atom of ferrous iron (5–9). In both enzyme families, the metal centers are solvent exposed so that they can directly bind and activate their substrates. Superoxide and H₂O₂ are tiny molecules that cannot easily be excluded from active sites, and they have high affinity for iron. The upshot is that they directly ligand and oxidize the enzyme metal centers. The oxidized iron atoms dissociate, activity is lost, and the pathways fail.

Superoxide and H₂O₂ are continuously formed in aerobic cells because molecular oxygen adventitiously oxidizes redox enzymes (10–12). Due to its substantial titers of scavenging enzymes, WT E. coli can suppress this threat. The question remains as to whether these ROS poison obligate anaerobes. Among the bacteria whose oxygen sensitivity has received particular attention are members of the Bacteroidetes (13–18). These carbohydrate fermenters are among the dominant bacteria in the mammalian gut (19), where they grow alongside E. coli. However, in contrast to E. coli, Bacteroides species quickly stop growing upon aeration. Notably, they do so despite possessing a substantial retinue of SOD, catalase, and peroxidases (16, 20–22). Product analysis of aerated Bacteroides thetaiotaomicron showed that stoppage of growth occurs concomitant with a loss of carbohydrate catabolism (15). Two enzymes in central metabolism lose activity (Fig. 1): fumarase, a member of the iron-sulfur dehydratase family, and pyruvate:ferredoxin oxidoreductase (PFOR), a key pyruvate-dissimilating enzyme that passes low-potential electrons toward hydrogen formation and/or NAD reduction. The fumarase bottleneck is marked by a cessation of succinate production and an unusual release of lactate. When this injury was bypassed by the addition of exogenous fumarate, some succinate production was restored, but the cell instead excreted pyruvate, reflecting PFOR failure. Either block should be enough to prohibit fermentative growth.

Fig. 1.

Relevant pathways in B. thetaiotaomicron metabolism. Shown in brackets are enzymes that lose activity when cells are transferred to oxic conditions. Fumarase is critical for the redox-balancing branch of central metabolism, while PFOR initiates the energy-conserving dissimilation of pyruvate. Lower-flux pathways are represented by dashed lines. Rpe is needed for oxidative flux through the pentose-phosphate pathway; IPMI is necessary for leucine synthesis, and aconitase serves a biosynthetic role in the generation of the α-ketoglutarate family of amino acids. Not depicted: Pdf is essential for the maturation of nascent polypeptides. Acn, aconitase; Fum, fumarase; αKG, alpha-ketoglutarate; PEP, phosphoenolpyruvate; OAA, oxaloacetic acid.

In this study we tested whether O₂⁻ or H₂O₂ might be involved. Our immediate focus was drawn to fumarase, because its vulnerability to ROS is well understood (5, 23, 24). We found that aeration simultaneously inactivated other iron-sulfur dehydratases and mononuclear iron enzymes. These failures were not due to any special sensitivity of the B. thetaiotaomicron enzymes, which maintained activity when expressed in aerobic E. coli. Instead, the cellular environment of aerated B. thetaiotaomicron is much more oxidizing than that of E. coli due to a much higher rate of endogenous ROS formation. Finally, analysis indicated that O₂⁻ is the specific culprit. These results validate the original suggestion of McCord et al. (1) that O₂⁻ toxicity might underlie key aspects of obligate anaerobiosis.

Results

When B. thetaiotaomicron cultures in rich medium were aerated, growth stopped after ∼40 min (Fig. 2). The static cells remained viable; when anoxia was restored hours later, growth resumed within minutes. We previously noted that the cessation of growth was accompanied by a diminution of glucose catabolism and the parallel inactivation of fumarase and PFOR, key enzymes in central metabolism (15). Fumarase drew our attention because this enzyme belongs to the family of [4Fe-4S] dehydratases, which are vulnerable to oxygen species that can oxidize their iron-sulfur clusters (5, 25–27). Assays revealed that two other members of this enzyme family, aconitase and isopropylmalate isomerase, also progressively lost activity when B. thetaiotaomicron was aerated (Fig. 3).

Fig. 2.

Growth ceases upon aeration and resumes when oxygen is removed. Cultures growing in anoxic BHIS medium were aerated at the first arrow. At the second arrow, cells were centrifuged and then resuspended in anoxic BHIS.

Fig. 3.

Iron-sulfur dehydratases and mononuclear Fe(II) enzymes lose activity when B. thetaiotaomicron is aerated. Top three panels: Three [4Fe-4S] dehydratases. Bottom two panels: Two nonredox Fe(II) enzymes. Chloramphenicol was added to block new protein synthesis, and cells were aerated in glucose buffer. Specific protocols are described in Materials and Methods. In this and other figures, error bars represent the SEM from at least three measurements. By the final time point, P < 0.001 for inactivation of all enzymes shown.

The other family known to be vulnerable to these oxidants comprises enzymes that use solvent-exposed ferrous iron atoms to catalyze nonredox reactions (6, 7, 9). When E. coli is stripped of its scavenging enzymes, both superoxide and H₂O₂ can oxidize enzymic Fe(II) cofactors, triggering iron release, the loss of activity, and collapse of the processes to which these enzymes contribute. We examined two such enzymes in B. thetaiotaomicron: ribulose-5-phosphate 3-epimerase (Rpe) and peptide deformylase (Pdf). Both enzymes employ ferrous iron (rather than other divalent metals) in B. thetaiotaomicron (Figs. S1 and S2), and both lost activity when cells were aerated (Fig. 3).

The failure of these enzymes in aerated B. thetaiotaomicron stands in sharp contrast to aerobic E. coli, where such enzymes retain full activity in vigorously aerated cultures. We considered the possibility that oxygen-tolerant bacteria have evolved enzymes that are intrinsically less reactive with oxidants. Fumarases were purified from both B. thetaiotaomicron and E. coli, and their iron-sulfur clusters were reconstituted in anoxic buffers. In vitro, these enzymes can be inactivated by superoxide, hydrogen peroxide, and molecular oxygen itself. As shown in Fig. 4, these species inactivated the two fumarases with similar rate constants. This result indicated that aspects of the cell environment, rather than of the enzymes themselves, are determinant in whether they retain activity in oxic environments.

Fig. 4.

The B. thetaiotaomicron fumarase is no more sensitive to oxidants in vitro than is the E. coli enzyme. Inactivation rate constants were determined for the purified enzymes as described in Materials and Methods. Note the multipliers indicated above the bar graphs. No significant differences between the two enzymes were indicated for any of the three comparisons (P = 0.16 for O₂, 0.5 for H₂O₂, and 0.06 for O₂⁻, with the B. thetaiotaomicron enzyme possibly exhibiting less sensitivity). B. theta, B. thetaiotaomicron; inactiv’n, inactivation.

To test this notion more directly, the fumarase and Rpe homologs were each expressed in both bacteria. To do so, mutants were created that lacked the native enzymes. Fig. 5 A and B shows that both fumarase enzymes remained fully active in E. coli when protein synthesis was blocked and erstwhile anoxic cells were aerated. In contrast, both enzymes lost activity when the same experiment was performed using B. thetaiotaomicron. This pattern was replicated with the Rpe homologs (Fig. 5 C and D). Thus, some aspect of B. thetaiotaomicron is not conducive to the function of oxidant-sensitive enzymes.

Fig. 5.

Fumarase and Rpe enzymes from both bacteria retain activity in aerated E. coli but not in aerated B. thetaiotaomicron. (A and B) Fumarases B from E. coli and B. thetaiotaomicron were expressed in anoxic E. coli (A) or B. thetaiotaomicron (B). Chloramphenicol was added to block new protein synthesis, and activity was tracked after cells were aerated. (C and D) Analogous experiments were performed with Rpe from both sources expressed in E. coli (C) or B. thetaiotaomicron (D). Strains and protocols are detailed in Materials and Methods. B. theta, B. thetaiotaomicron. *P < 0.05.

Direct assays showed that anoxic B. thetaiotaomicron possesses 25% as much SOD activity as does E. coli (28). This difference is not a sufficient explanation for the inactivation of the enzymes, since fumarase and other iron-sulfur enzymes retain activity inside E. coli when its SOD titers are diminished to this level (29). B. thetaiotaomicron also has a catalase and three peroxidases that are devoted to scavenging H₂O₂ (22). It is difficult to quantify the internal scavenging activities of peroxidases, but B. thetaiotaomicron and E. coli are similarly effective at clearing H₂O₂ from laboratory cultures. Therefore, it is not obvious that enzymes in aerated B. thetaiotaomicron are damaged due to a particular deficiency of scavenging systems. Further, we observed that B. thetaiotaomicron is able to repair damaged enzymes. After aerated cells were returned to anoxic conditions, enzyme activities returned, even when protein synthesis was blocked (Fig. S3). In sum, although these two bacteria exhibit some qualitative and quantitative differences in cellular defenses, the differences seem unlikely to explain the disparity in enzyme fate.

When H₂O₂ is formed inside mutant strains that lack H₂O₂-scavenging enzymes, the H₂O₂ diffuses out into the medium, allowing the rate of endogenous formation to be quantified (30). Our previous work indicated that H₂O₂ is formed much more quickly in aerated B. thetaiotaomicron than in aerated E. coli: upon aeration, the OxyR H₂O₂ stress response was immediately activated only in the obligate anaerobe, and mutants lacking this response died quickly (22, 31). The H₂O₂ measurements were repeated under the condition of the present experiments. Anoxically grown cells were washed and resuspended into buffer, to avoid chemical H₂O₂ production by medium components, and then aerated. When glucose was provided as a carbon source, the rate of endogenous H₂O₂ formation was 10 times higher in B. thetaiotaomicron than in E. coli (Fig. 6A). When glucose was omitted, the rates were diminished in both strains, suggesting that catabolism is the source of ROS. Notably, the withdrawal of glucose simultaneously slowed the rates at which B. thetaiotaomicron enzymes were damaged (Fig. 6 B and C).

Fig. 6.

The rate of enzyme damage correlates with the rate of endogenous ROS formation. (A) The efflux of endogenous H₂O₂ was tracked after aeration of Hpx⁻ derivatives of B. thetaiotaomicron (ΔkatE ΔahpC, Δrbr1, Δrbr2; circles) and E. coli (ΔkatG, ΔkatE, ΔahpCF; squares). The cells were grown in BHIS (B. thetaiotaomicron) or LB (E. coli) and then washed and aerated in buffer containing chloramphenicol and either no carbon source (open symbols) or 0.2% for continued catabolism. No significant H₂O₂ accumulated in sterile buffer. (B and C) Under the conditions of A, the activities of Rpe and Pdf were monitored in B. thetaiotaomicron. Statistical analyses compare values at each time point between glucose-fed and -starved cells. *P < 0.05; **P < 0.01. B. theta, B. thetaiotaomicron; FU, fluorescence units; glc, glucose.

We do not have a good way to quantify superoxide formation inside cells. However, studies with E. coli enzymes indicate that both species evolve from the same event, the autoxidation of flavoenzymes (32). Any superoxide is subsequently converted to H₂O₂, either by SOD or by spontaneous dismutation. Therefore, it seems likely, although not definitive, that the high rate of H₂O₂ production in B. thetaiotaomicron connotes a similarly high rate of superoxide production.

The rate of endogenous H₂O₂ formation depends upon oxygen concentration, in accordance with the model that ROS are primarily formed by the adventitious oxidation of redox enzymes (31). When B. thetaiotaomicron was exposed to a range of oxygen concentrations, the amount of residual fumarase activity depended inversely upon the oxygen level (Fig. 7). This outcome is notable in that it suggests that the bacterium may retain substantial enzyme and pathway functions when oxygen is present at lower levels.

Fig. 7.

Fumarase activity in B. thetaiotaomicron depends upon the oxygen concentration. Fumarase activity was measured after 1 h of exposure to the indicated oxygen concentration. Full aeration is represented by 22% oxygen. The dotted line represents prior data (31) showing that the rate of endogenous H₂O₂ production is proportionate to oxygen concentration.

These data, plus the observation that B. thetaiotaomicron enzymes remain active in aerobic E. coli, indicate that superoxide or H₂O₂—rather than molecular oxygen—is likely to be responsible for this enzyme damage in vivo. We examined B. thetaiotaomicron mutant strains that lack either SOD or catalase/peroxidases. Fumarase lost activity more rapidly in either mutant than in WT cells, confirming the ability of these oxidants to inactivate the enzyme in vivo and showing that they are partially shielded by these scavenging enzymes in WT cells (Fig. 8). Fumarase inactivation was especially rapid in the SOD mutant. The same pattern was reproduced when Rpe activity was tracked. Strikingly, the rate of Rpe damage in SOD-deficient B. thetaiotaomicron far exceeded its rate of damage in SOD-deficient E. coli, consistent with our expectation that superoxide is formed more quickly in aerated B. thetaiotaomicron.

Fig. 8.

Scavenging enzymes slow the rate of enzyme damage in aerated B. thetaiotaomicron. (A and B) Fumarase (A) or Rpe (B) activity was monitored after the aeration of B. thetaiotaomicron strains that lacked enzymes that scavenge H₂O₂ (Hpx⁻) or superoxide (SOD⁻). B also depicts the rate of Rpe inactivation upon the aeration of E. coli SOD⁻ strains. Inactivation is much faster in the B. thetaiotaomicron strain, suggesting that superoxide levels are significantly higher. P values in A compare mutant to WT activities; P values in B compare the activities of SOD⁻ B. thetaiotaomicron with those of SOD⁻ E. coli. (C) Extracts were prepared from the B. thetaiotaomicron SOD⁻ strain at the indicated time points of aeration. Rpe activities were assayed before (gray bars) or after (black bars) treatment with Fe(II) alone or penicillamine and then Fe(II) (hatched bars). The ability to reactivate with Fe(II) alone defines the fraction of Rpe in an apoprotein form, whereas the increasing requirement for preliminary penicillamine treatment indicates that Zn(II) increasingly occupied the enzyme metal-binding site. The pattern matches that observed in superoxide-stressed E. coli (8). [As-extracted enzyme activity was not altered by either Fe(II) or penicillamine/Fe(II) treatment.] *P < 0.05; **P < 0.01; ***P < 0.001.

When superoxide oxidizes the iron cofactors of mononuclear enzymes, the dissociation of iron leaves an apoprotein whose activity can be rapidly restored by the binding of another Fe(II) atom (8). In SOD-deficient E. coli, this process of iron oxidation and rebinding can occur repeatedly, but each cycle provides some chance that a competing metal like Zn(II) will bind in place of Fe(II); this outcome diminishes activity, because in such enzymes zinc is a poor catalyst (8, 9). Interestingly, we observed the same phenomenon in SOD-deficient B. thetaiotaomicron (Fig. 8C). Activity could be restored to the lysates only by adding penicillamine, a good zinc chelator, to extract a blocking metal from the active site before the addition of Fe(II). We infer that the metal was zinc rather than manganese, because the latter metal would have furnished substantial activity (Fig. S1) and would have dissociated relatively quickly without a chelator (Materials and Methods). B. thetaiotaomicron lacks the two known bacterial manganese importers, MntH and MntABC.

The outcome is different when E. coli Rpe is oxidized by H₂O₂ (6). In the latter case, the reaction between the active-site Fe(II) and H₂O₂ generates a ferryl/hydroxyl species that has a finite chance of irreversibly damaging the polypeptide. The enzymes that are oxidized in this way cannot be reactivated. This behavior was replicated in B. thetaiotaomicron: whereas the Rpe that lost activity in SOD mutants could be fully reactivated, the enzymes that lost activity in catalase/peroxidase mutants could not be (Fig. 9). It is important then that when WT cells were aerated, the inactivated Rpe could be fully reactivated by penicillamine/Fe(II) treatment. The enzyme profile resembled Rpe that was recovered from SOD mutants that had been exposed to oxygen for 30 min. The implication is that superoxide, rather than H₂O₂, is the primary oxidant that disables these enzymes when B. thetaiotaomicron is aerated.

Fig. 9.

The reactivatibility of Rpe recovered from aerated cells suggests that superoxide rather than H₂O₂ is the damaging species. Rpe was recovered after 3 h of aeration of Hpx⁻, SOD⁻, or WT cells. Left shows that the H₂O₂-damaged enzyme can only be partially reactivated; Middle shows that superoxide-damaged enzyme can be fully reactivated after penicillamine extracts the competing metal; and Right shows that Rpe from WT cells can be fully reactivated, consistent with damage by O₂⁻ rather than H₂O₂. *P < 0.05; **P < 0.01; ***P < 0.001. ns, not significant.

To test this conclusion more directly, SOD was overproduced ∼15-fold in WT B. thetaiotaomicron. When these cells were aerated, both fumarase and Rpe were substantially protected (Fig. 10). Some damage still occurred, due either to the residual superoxide or to H₂O₂ itself, and growth did not resume (see Discussion). However, this outcome demonstrates that superoxide is the oxidant that damages these enzymes when this obligate anaerobe encounters oxygen.

Fig. 10.

Overproduction of SOD protects B. thetaiotaomicron enzymes from the aeration. (A) SOD was overexpressed 15-fold from a plasmid (psod). (B and C) Activities of fumarase and Rpe after aeration of strains containing the vector or SOD-overproducing plasmid. BHIS-grown cells were washed and then aerated in buffer containing glucose. **P < 0.01; ***P < 0.001.

Discussion

Life evolved in an anoxic world. Ferrous iron was readily available (33), and its proficiency as a surface and electron-transfer catalyst resulted in its recruitment into many enzymes. The emerging metabolic pathways became configured around the chemistry that iron can catalyze.

Two billion years later, the appearance of oxygenic photosystem II put organisms at risk, because oxygen oxidizes redox enzymes to make ROS, and the ROS can then poison the exposed iron cofactors of enzymes. It was once believed that microorganisms responded either by evolving SOD and catalase or by retreating to anoxic habitats. However, as the facility of lateral gene transfer became apparent, it seemed unlikely that bacteria would be trapped in anoxic environments simply for lack of scavenging enzymes. Indeed, further investigations revealed that even anaerobes maintain a cohort of scavenging enzymes: if not SODs and catalases, then superoxide reductases and peroxidases (34, 35). The latter enzymes were either unknown or underappreciated when McCord et al. (1) conducted their survey. It is implicit that all microbes occasionally confront enough oxygen to warrant the presence of defensive enzymes.

Why Hasn’t B. thetaiotaomicron Solved Its Oxygen Problem?

An important question has been whether the appearance of scavenging enzymes eliminated the threat of endogenous superoxide and H₂O₂, or whether these endogenous ROS are still substantial enough to exert toxic effects. The results here suggest that E. coli suffers little enzyme damage from the ROS that it generates, but B. thetaiotaomicron can still be overwhelmed. The difference lies in the amount of ROS that the two bacteria generate.

Why doesn’t B. thetaiotaomicron compensate by making higher titers of scavenging enzymes? Two ideas stand out. First, the levels of scavenging enzymes in E. coli already make these among the most abundant proteins in the cell. Even so, aerobic E. coli sits close to the verge of oxidative collapse: calculations suggest that its [4Fe-4S] dehydratases are oxidized and must be repaired every half hour or so (32). Indeed, just a fivefold increase in superoxide level is enough to reduce the steady-state activities of its [4Fe-4S] dehydratases by half (Fig. S4) (29). Since aerated B. thetaiotaomicron produces ROS at 10 times the rate of E. coli, it follows that it would require 10-fold higher titers of scavenging enzymes than E. coli to achieve the same level of enzyme stability. Such an investment of resources, including metal cofactors to activate the SOD, may be untenable.

Second, B. thetaiotaomicron features key enzymes that may be damaged by molecular oxygen per se. If oxygen can damage enzymes directly, then the value of scavenging enzymes would be limited. B. thetaiotaomicron possesses two pyruvate dissimilating enzymes: pyruvate:formate lyase (Pfl) and PFOR. Pfl is oxygen sensitive by virtue of its glycyl-radical chemistry, which is quenched by direct reaction with oxygen, a radical itself (36). PFOR may be the primary route of pyruvate consumption, and when cells are aerated its activity diminishes on a time frame similar to that of fumarase (15). The inactivation mechanism is not known, but it seems likely to involve overoxidation of its low-potential redox clusters (37). Either oxygen or ROS derived from it similarly inactivates the enzyme in vitro. Oxygen sensitivity in vitro is not enough to predict that oxygen will also poison PFOR in vivo: S-adenosylmethionine radical enzymes (38), for example, are oxygen-sensitive in vitro but function normally inside aerobic cells. However, if molecular oxygen itself does directly inactivate PFOR in B. thetaiotaomicron, then the failure of this bacterium to make higher levels of scavenging enzymes can be justified: elevated oxygen concentrations would inactivate PFOR, and thus block central metabolism, regardless of the titers of scavengers. This possibility was noted in the original report of McCord et al. (1). The mechanism is under investigation. Direct PFOR inactivation would also explain the failure of SOD overproduction to enable aerobic growth—although that outcome might also arise from the inactivation of ribonucleotide reductase. B. thetaiotaomicron relies upon an oxygen-sensitive glycyl-radical ribonucleotide reductase whose presence in B. thetaiotaomicron presumably reflects its commitment to its anoxic habitat.

What Is the Source of the Toxic ROS?

Our data show that the rapid formation of superoxide is the particular feature that condemns ROS-sensitive enzymes to inactivity when B. thetaiotaomicron is aerated. We do not know the origin of all of this ROS. Oxygen has a triplet electronic structure that constrains it to accept electrons in univalent steps (39), and so the univalent redox properties of flavoenzymes and quinones have prompted workers to focus upon respiratory chains as plausible sites of ROS production. Surprisingly, genetic studies have indicated that this chain is, at best, a minor contributor to ROS formation in aerobically grown E. coli (11). However, ROS formation in E. coli rises immediately after the aeration of anoxically grown cells, and in this specific circumstance the source was identified as the anaerobically induced respiratory enzyme fumarate reductase (12, 40). The redox flux through this flavoprotein is even greater in B. thetaiotaomicron than in E. coli, and so we conjectured that it might be the main source of ROS stress. In the related bacterium, Bacteroides fragilis, mutants that lack fumarate reductase were subsequently observed to generate lower amounts of H₂O₂ (41); interpretation was not straightforward, however, since these mutants grow much more slowly than do WT cells (42). A recent biochemical analysis showed that the B. thetaiotaomicron fumarate reductase has a distinct electronic structure that precludes reactions with oxygen, unlike that of its E. coli homolog (31). Those data did not support the idea that the respiratory chain was involved in high-rate ROS production.

Moving forward, there are some hints. The rate of H₂O₂ formation in aerated B. thetaiotaomicron is so high that it exceeds substrate fluxes through most biosynthetic pathways. We therefore suspect that the ROS may evolve from the PFOR-ferredoxin-hydrogenase/RNF (ferrodoxin:NAD oxidoreductase) route of central metabolism. After PFOR passes electrons to ferredoxin, the reduced ferredoxin delivers them either to a hydrogenase or to RNF; NADH generated by the latter then proceeds to NADH dehydrogenase of the respiratory chain (43). The redox moieties in these enzymes carry electrons at low potentials, making transfer to oxygen energetically favorable. Further, this is among the few metabolic pathways of B. thetaiotaomicron that E. coli does not share, which could explain the discrepancy in ROS rate. The gradual slowing of H₂O₂ release after aeration (Fig. 6A) plausibly reflects the progressive inactivation of PFOR. This hypothesis awaits experimental evidence.

In sum, this study shows that both classes of enzymes known to be damaged by ROS in nonscavenging mutants of E. coli are also damaged by simple aeration of WT B. thetaiotaomicron. The origin of the ROS and its involvement in the inactivation of PFOR are the next problems to solve.

Materials and Methods

Chemicals.

Brain-heart infusion (BHI) broth was purchased from Difco. Hemin hydrochloride, l-cysteine, l-cystine, fumaric acid, ferrous ammonium sulfate hexahydrate, DTT, maltose, β-lactose, 2,2′-dipyridyl, citraconate, disodium l-malic acid, dl-trisodium isocitrate, horse heart cytochrome c, xanthine, bovine xanthine oxidase, bovine liver catalase, E. coli iron-containing SOD, imidazole, antibiotics (ampicillin, erythromycin, gentamicin, and chloramphenicol), 5′-fluorodeoxyuridine, HRP, α-glycerophosphate dehydrogenase/triosephosphate isomerase from rabbit muscle, manganese(II) chloride tetrahydrate, cobalt(II) chloride hexahydrate, zinc(II) sulfate heptahydrate, ferric(III) chloride, EDTA, d-penicillamine, d-ribose 5-phosphate disodium salt, d-ribulose 5-phosphate disodium salt, d-xylulose 5-phosphate sodium salt, NADH, NAD⁺, formate dehydrogenase, formyl-Met-Ala-Ser tripeptide, thiamine pyrophosphate, isopropylthiogalactopyranoside (IPTG), 2,6-pyridinedicarboxylic acid, Tris (2-carboxyethyl)phosphine (TCEP), and 30% H₂O₂ were from Sigma. His Gravitrap was obtained from GE Healthcare. d-Glucose, Hepes, Tris⋅HCl, and MgCl₂ were purchased from Fisher. Glycylglycine was from Acros Organics. Amplex Ultrared reagent was purchased from Invitrogen through Thermo Fisher Scientific.

Cell Media and Growth.

BHI-supplemented medium (BHIS; ref. 44) contained 37 gL⁻¹ of BHI broth, 15 μM hemin chloride, 4 mM l-cysteine hydrochloride, and 22.6 mM sodium bicarbonate. Antibiotics were supplemented into the medium when required as 20 μg mL⁻¹ erythromycin, 200 μg mL⁻¹ gentamicin, 15 μg mL⁻¹ chloramphenicol, and 200 μg mL⁻¹ 5′-fluorodeoxyuridine. Luria–Bertani (LB) medium contained 10 g tryptone, 10 g NaCl, and 5 g yeast extract per liter. Defined medium was composed of minimal A salts (45) supplemented with 0.2% casein acid hydrolysate, 0.2% glucose, 5 μg/mL thiamine, 0.02% MgSO₄ heptahydrate, and 0.5 mM tryptophan. Maltose (0.5%) and 0.2% lactose were added as a carbon sources where indicated. The BHIS, LB, and defined media were adjusted to pH 7.0 before sterilization.

Anoxic growth was performed in a Coy anaerobic chamber that contained an atmosphere of nitrogen (85%), hydrogen (10%), and carbon dioxide (5%). Autoclaved media were immediately transferred to the anaerobic chamber and stored for ≥24 h to ensure the outgassing of residual oxygen. Buffers used in this study were kept in an anaerobic chamber for ≥1 wk before use.

All experiments were conducted upon exponential-phase cultures that had doubled at least four times subsequent to dilution of overnight cultures. Typically, cells were inoculated to 0.005 OD₆₀₀ into warm growth medium, and they were grown under anoxic conditions to 0.10–0.20 OD₆₀₀ before biochemical or physiological measurements were begun.

The impact of oxygen exposure and removal upon cell growth was observed by culturing WT B. thetaiotaomicron in anoxic BHIS medium to an OD₆₀₀ of 0.1. Cultures were then transferred out of the anaerobic chamber and aerated with vigorous shaking at 37 °C for ∼3 h. The cells were then centrifuged, resuspended in an equal volume of prewarmed anoxic BHIS media, and cultured once again in the anaerobic chamber at 37 °C. Growth was monitored by OD₆₀₀.

Gene Deletions.

The bacterial strains and plasmids that were used in this study are listed in Tables S1 and S2. The BT5482 Δtdk strain served as the parent strain for construction of deletion mutants using a published method (46). (The tdk gene encodes thymidine kinase, an enzyme in the salvage pathway of pyrimidine biosynthesis, and it is nonessential under the conditions of these experiments.) The B. thetaiotaomicron rpe (BT_3946) and sod (BT_0655) genes were identified by BLAST analysis using E. coli rpe and sodB sequences as queries. The consequent mutations were confirmed by genome PCR and enzyme assays. Note that under standard assay conditions B. thetaiotaomicron rpe mutants retained a few percent of the activity of WT cells; such activity has been observed in other bacteria as well (47) and is due to the promiscuity of other enzymes, which themselves lack iron and are not sensitive to oxidants. The E. coli rpe deletion mutation was obtained from the Keio collection (48) at the E. coli Genetic Stock Center. The mutation was transferred to recipient strains by P1 transduction (45), and inheritance was again verified by PCR analysis and enzyme assay. Genomic DNA was isolated from both E. coli and B. thetaiotaomicron using DNeasy blood and tissue kit (Qiagen), as instructed by the manufacturer.

Deletion of the sole fum gene (BT_2256) of B. thetaiotaomicron did not succeed until we included 20 mM fumarate in the growth media. Exogenous fumarate can be imported and used as a substrate for fumarate reductase, thereby restoring to fum mutants the redox-balancing and energetic functions of the succinate-production pathway.

Overproduction of Enzymes.

The SOD of B. thetaiotaomicron was overproduced by cloning the native gene behind its own promoter onto a multicopy plasmid. The sod (BT_0655) coding sequence, plus 520 bp immediately upstream that contains the promoter region and ribosome binding site, were amplified by PCR and inserted into pNLY-P_susA vector (22) by BamHI/SacI. The clone was screened on anaerobic BHIS plates (200 μg/mL gentamicin and 20 μg/mL chloramphenicol). Overproduction of SOD in standard BHIS medium was confirmed by assay.

To express E. coli and B. thetaiotaomicron fumarases in the B. thetaiotaomicron fum mutant strain, the coding sequences of E. coli fumB and B. thetaiotaomicron fum were amplified by PCR; the ribosome binding site (RBS) of B. thetaiotaomicron fum (21 bp) was added upstream of each gene. Primers are listed in Table S3. Each DNA fragment was inserted into the pNLY-P_susA vector by BamHI/SacI, behind the SusA promoter. Construction was confirmed by digestion. The plasmid was conjugated into the B. thetaiotaomicron fum mutant and selected in BHIS medium containing 20 mM fumarate, 200 μg/mL gentamicin, and 20 μg/mL chloramphenicol. In experiments using these strains, maltose (0.4% wt/vol) was added to BHIS medium to activate the SusA promoter.

To express fumarases in E. coli, a fumA fumB fumC mutant strain (JH400; ref. 49) lacking all three native E. coli fumarases was used as the host. Genes encoding the E. coli fumA and B. thetaiotaomicron fum were cloned behind the lac promoter of the low-copy number vector pWKS30 by HindIII/BamHI. The RBS of E. coli gapA was inserted upstream of each gene. Sequences of primers for PCR are listed in Table S3. Construction was confirmed by digestion. To enable expression, 0.2% lactose replaced glucose in the defined medium.

The E. coli rpe gene was amplified by PCR, digested with BamHI/SacI, and inserted into pNLY-P_susA plasmid. The RBS (21 bp) was included at the upstream region of the gene as pNLY-P_susA plasmid lacks it. The pNLY-P_susA-rpe plasmid was transformed into the parent B. thetaiotaomicron rpe mutant strain using a biparental mating procedure, and selected for gentamicin (200 µg/mL) and chloramphenicol (20 µg/mL) resistance. Expression was achieved by growing the recombinant plasmid containing the B. thetaiotaomicron strain in BHIS medium supplemented with 0.5% maltose in place of glucose.

The B. thetaiotaomicron rpe gene was inserted into pWKS30 (50) and transformed into an E. coli strain that lacks the rpe gene. The gene was cloned behind the lac promoter by HindIII/BamHI. The RBS of the gapA gene was again included upstream of the B. thetaiotaomicron rpe gene. The plasmid was transformed into the E. coli Δrpe strain. Expression was achieved by replacing glucose with lactose in defined minimal A medium. Inactivation of the heterologous Rpe enzymes was compared with inactivation of the native enzymes in WT strains.

Enzyme Purifications.

The E. coli and B. thetaiotaomicron fumarase coding regions were inserted into the pET16b vector (Novagen), which was transformed to BL21 (DE3). IPTG (0.5 mM) was added to 1 L log-phase culture near 0.5 OD₆₀₀ to induce the protein expression. After 3 h at 37 °C, cells were harvested by centrifugation. Fumarase was purified under aerobic conditions at 4 °C following the standard Ni-NTA resin purification protocol (Novagen). Aliquots of the purified protein were stored at −80 °C and reactivated before use.

To reactivate the purified protein, the 1-mL anaerobic reaction system contained 1–5 μM fumarase, 0.5 mM Fe(NH₄)₂(SO₄)₂, 2.5 mM DTT, 2.8 mM cysteine, and 0.1 μM purified E. coli IscS protein, in 50 mM NaPi buffer (pH 7.2) (51). The reaction was incubated at room temperature in the anaerobic chamber for ≥2 h, and then dipyridyl was added to a final concentration of 1 mM to stop the process. After 5 min, the reaction was filled to 4 mL by NaPi buffer and concentrated from 4 mL to 0.1 mL three times by 10-kDa Millipore filtering tubes. This step removed the extra Fe²⁺ and DTT. All steps were performed under anoxic conditions. Reactivated enzyme was then used without further storage.

Rpe from B. thetaiotaomicron was purified following the protocol used for its E. coli homolog (6). The rpe gene was cloned using pET16b vector, and the resulting pRpe-His₁₀ was overexpressed in the E. coli BL21(DE3) strain. The Rpe-His₁₀ protein was then purified using His Gravitrap. The His₁₀ tag was removed after purification. Purified protein was >95% pure as indicated by SDS/PAGE analysis. The enzyme was stored at 4 °C.

We found that E. coli transketolase obtained from commercial sources was contaminated with Rpe; therefore, E. coli transketolase A was purified under aerobic conditions as described in ref. 52. Cell extracts were prepared from LB culture of E. coli BL21 Δrpe strain containing the tktA overexpression plasmid pET16b-tktA. The enzyme was >95% pure as indicated by SDS/PAGE analysis. The purified enzyme was stored at −80 °C in 50 mM Tris⋅HCl, pH 8, 20% glycerol.

The E. coli enzyme IscS enables the in vitro reconstruction of [4Fe-4S] clusters on apoprotein forms of fumarase and other enzymes. IscS was purified as described (51).

Enzyme Assays.

For assays of oxidant-sensitive enzymes, cells were first centrifuged and then washed in anoxic buffers, and cell extracts were then prepared by sonication in anoxic buffers in an anaerobic chamber. For fumarase assays, washed cells were sometimes frozen and stored at −80 °C; for IMPI and aconitase assays, cells were lysed and assayed immediately. All assay reactions were assembled in the anaerobic chamber in a sealed cuvette before being moved to a laboratory spectrophotometer. Reactions were at room temperature (RT). Fumarase activity was determined in 50 mM sodium phosphate (pH 7.3) containing 50 mM l-malate; production of fumarate was monitored at 250 nm (53). Cell densities were adjusted so lysates contained ∼1 mg/mL protein. Isopropylmalate isomerase (IPMI) activity was measured by monitoring the decrease in citraconate absorbance at 235 nm in 100 mM Tris-Cl (pH 7.6) (54). The aconitase assay was performed in 100 mM Tris-Cl (pH 8.0) containing 20 mM dl-trisodium isocitrate; the formation of aconitate was measured at 240 nm (55). SOD activities were determined by the xanthine/xanthine oxidase method, under oxic conditions (56).

Rpe converts ribulose-5-phosphate to xylulose-5-phosphate, which was then detected in an enzyme-coupled assay through NADH oxidation as a decrease in A₃₄₀ (57). Rpe and Pdf were assayed immediately upon harvesting, without storage. The Rpe assay from crude extracts was performed in 50 mM glycylglycine buffer (pH 8.5) at RT. Extracts were prepared in the anaerobic chamber. Cells were washed twice with ice-cold 50 mM glycylglycine buffer containing 1 mM EDTA, pH 8.5, and resuspended in 1 mL of the same buffer without EDTA. Samples were then sonicated for 2 min (3 s on/3 s off), and lysates were cleared by centrifugation at 20,000 × g for 3 min. Assays were performed within 5 min of cell lysis to avoid metal dissociation. A typical assay (500 μL) contained 50 mM glycylglycine buffer (pH 8.5), 1 mM ribulose 5-phosphate, 1 mM ribose 5-phosphate, 1 unit of transketolase, 0.2 mM MgCl₂, 2 mM thiamine pyrophosphate, 5 mM EDTA, 1 unit of α-glycerophosphate dehydrogenase, 10 units of triosephosphate isomerase, and 0.2 mM NADH.

Pdf removes the formyl group from the model peptide formyl-Met-Ala-Ser. The released formate is then oxidized to CO₂ and H₂O by formate dehydrogenase, with reduction of NAD⁺, and NADH formation is monitored by the increase in A₃₄₀ (58). Cells were centrifuged in the anaerobic chamber at 4 °C in 50 mM Hepes buffer with 25 mM NaCl (pH 7.5) plus 1 mM EDTA, washed twice in the same buffer, and finally resuspended in 1 mL of the same buffer without EDTA. Samples were then sonicated for 2 min (3 s on/3 s off), and lysates were cleared by centrifugation at 12,000 rpm for 3 min. Assays were performed within 5 min of cell lysis to avoid metal dissociation. A typical assay (500 μL) contained 50 mM Hepes buffer with 25 mM NaCl (pH 7.5), 10 mM NAD⁺, 1 unit of formate dehydrogenase, 1 mM formyl-Met-Ala-Ser, and cell extracts, all under anoxic conditions at RT.

Transketolase was assayed by standard methods (52) in a reaction mixture containing 50 mM glycylglycine buffer, pH 8.5; 5 mM DTPA; 1 unit of α-glycerophosphate dehydrogenase; 10 units of triosephosphate isomerase; 1 mM xylulose 5-phosphate; 1 mM ribose 5-phosphate; and 0.2 mM NADH in a final reaction volume of 500 μL. Absorbance was monitored at 340 nm.

Measurement of Enzyme Activities After Aeration.

To track the effect of aeration upon native cellular fumarase, aconitase, and IPMI enzyme activities, a 150-mL culture of B. thetaiotaomicron was grown in anaerobic BHIS medium from 0.01 to 0.25 OD₆₀₀ and then centrifuged, washed once, and resuspended in 10 mL RT anoxic 50 mM NaPi buffer (pH 7.2). Bacteria were grown similarly when expressing fumarases from a plasmid, except that 100 μg/mL chloramphenicol was included. Five-milliliter aliquots of the suspended anaerobic bacteria were added to 25 mL oxic or anoxic NaPi buffer, each containing glucose and enough chloramphenicol to block protein synthesis (100 μg/mL chloramphenicol for strains without vector, 500 μg/mL chloramphenicol for strains with vector pNLY-SusA). The anoxic suspension was incubated at 37 °C in the anaerobic chamber, while the oxic suspension was transferred to an outside incubator and shaken vigorously. At time points, cultures were transferred back to chamber, and cell extracts were prepared and assayed. The aeration of cultures was conducted in buffered glucose to avoid the chemical production of H₂O₂ by components of BHIS medium, while allowing the intracellular ROS formation that occurs as a by-product of carbohydrate catabolism.

Fumarase inactivation was also examined upon aeration of an E. coli fumA fumB fumC mutant strain (JH400) that was complemented with either E. coli fumarase A or B. thetaiotaomicron fumarase expressed from plasmids. Cultures were grown in minimal A medium in which lactose replaced glucose as a carbon source, enabling expression of fum genes from the lac promoter. Ampicillin (50 μg/mL) was included to maintain the expression plasmid. As with the B. thetaiotaomicron experiments, log-phase anaerobic cultures were harvested, resuspended in aerobic buffer with lactose and chloramphenicol, and shaken under room air at 37 °C. At time points, cultures were harvested aerobically and moved back to the chamber for preparation of cells extracts and measurement of enzyme activity. A parallel anaerobic culture was harvested as an oxygen-free control.

A similar approach was taken to track fumarase activities in cells exposed to defined concentrations of oxygen. Log-phase cultures in BHIS were washed and suspended in NaPi buffer (pH 7.2, 0.2% glucose plus chloramphenicol) that was either anaerobic or that had been steadily gassed for 1 h with a selected ratio of nitrogen and air. The gassing ratio was established by mixing gas flow from nitrogen and air cylinders at a Y-intersection; the gas stream was then bubbled through a water trap (to ensure hydration) and finally through the sample tube (12). Parafilm was stretched over the tube to minimize exposure to laboratory air, which was excluded anyway because of the positive in-to-out pressure. Cultures were injected into the tube through the parafilm with a Hamilton syringe. At intervals, cultures were harvested anaerobically to prepare cell extracts and to measure enzyme activities.

To monitor the activities of Rpe and Pdf, log-phase E. coli and B. thetaiotaomicron cells were grown at least four generations to ∼0.2 OD in anoxic LB and BHIS media, respectively. Cells were then washed twice and resuspended in anoxic 50 mM glycylglycine buffer (pH = 8.5) containing 100 µg/mL chloramphenicol, either in the absence and presence of 0.2% glucose, as indicated. Cultures were either incubated in the anaerobic chamber or were transferred to an external laboratory shaker. At intervals, cultures were returned to the chamber, extracts were prepared, and enzymes were assayed. To test the ability of cells to reactivate Rpe, the air-exposed cells were returned to the anaerobic chamber, washed once with same anoxic buffer, resuspended in anaerobic BHIS buffer containing chloramphenicol (100 µg/mL), and incubated anaerobically for 1 h at 37 °C. Cell extracts were then prepared for assay.

In Vivo Repair of B. thetaiotaomicron Fumarase.

The ability of intact cells to repair oxidatively damaged fumarase was tested. WT B. thetaiotaomicron cells were aerated for 2 h in 50 mM NaPi (pH 7.2) buffer containing 0.2% glucose and 100 μg/mL chloramphenicol to inactivate fumarase. The culture was returned to the chamber. An aliquot was removed and prepared for assay. The remaining cells were centrifuged and then resuspended in anoxic 37 °C BHIS medium containing chloramphenicol. The culture was subsequently harvested and lysed, and fumarase activity was measured after 2–12 h.

Analysis of Rpe and Pdf Metallation Status.

One cannot use the metal content of purified mononuclear enzymes as an indicator of their native metallation state in vivo, because this enzyme family readily exchanges metals with the buffer during the purification process. Instead, the presence of iron in the active site can be deduced by susceptibility to H₂O₂ immediately after extract preparation (7). To test whether as-extracted Rpe and Pdf contained iron, fresh cell extracts from E. coli or B. thetaiotaomicron were treated with 100 μM H₂O₂ at 25 °C for 10 min (8). Catalase (30 U/mL) was added to stop the reaction. Rpe assays were performed before and after H₂O₂ treatment. To reactivate the enzymes in vitro, 90 μL H₂O₂-challenged cell extracts were incubated with 500 μM (final concentration) of the desired metal, and enzyme activity was remeasured. Where indicated, Pdf was preincubated for 10 min with 500 μM TCEP before the addition of the metal.

The activity and stability of Rpe metallated with Fe, Zn, and Mn were also measured. Rpe was recovered after purification in its Zn(II)-Rpe form; Zn often binds to mononuclear enzymes during purification. Zinc was removed from the purified protein by chelation using dipicolinic acid (20 mM) and EDTA (100 μM) at 25 °C for 1 h. The apoenzyme was then diluted 1,000-fold into anoxic buffer containing 100 μM of the test metal at 25 °C. Activity was then measured. To test whether differentially metallated enzyme was sensitive to H₂O₂, the enzyme was incubated with 100 μM H₂O₂ for another 5 min. The reaction was terminated by the addition of catalase (30 U/mL), and the enzyme was reassayed. In separate experiments, the spontaneous dissociation rate of the metals was determined by incubating metal-loaded enzyme in the presence of 1 mM EDTA at RT and tracking activity. Iron-, manganese-, and zinc-loaded Rpe displayed dissociation half-times of 60 min, 5 min, and >8 h, respectively, in accord with intrinsic metal softness and in close agreement with the values determined for the E. coli enzyme (6). Rpe recovered in extracts from anoxic cells lost activity in the presence of EDTA with a half-time of 60 min, matching that of iron-cofactored enzyme.

In some experiments, the metallation status of Rpe was tested after the aeration of WT, Hpx⁻, or SOD⁻ cells. Cell extracts were prepared and assayed under anoxic conditions. An aliquot of extract was then metallated with ferrous iron (1 mM) for 10 min and then reassayed; any boost in activity represents apoprotein in the original sample. Separate aliquots were tested for mismetallated Rpe; such enzyme is poorly active but can be reactivated if the inappropriate metal is extracted from the active site. To chelate metal from the active site of Rpe, cell extracts were incubated with 0.5 mM EDTA for 3 h or with 0.5 mM penicillamine for 1 h under anoxic conditions at 25 °C (8). Rpe activity was completely absent after chelation. To remetallate the enzymes, the desired metals were added to a final concentration of 1 mM, and the mixture was incubated anaerobically for 10 min at 25 °C. Activity was then remeasured. Collectively, these procedures distinguished enzyme that was correctly metallated from apoprotein, mismetallated protein, and irreversibly damaged protein.

Measurement of H₂O₂ Production by Aerated Cells.

Mutants that lack catalase and peroxidase activities are denoted as hydroperoxidase-deficient, or Hpx⁻ [katG katE ahpCF for E. coli (3) and ΔkatE ΔahpC Δrbr1 Δrbr2 for B. thetaiotaomicron (22)]. These cells cannot scavenge endogenous H₂O₂ and release it directly into the medium. Cultures were grown for more than four generations in LB (E. coli) or BHIS (B. thetaiotaomicron) to an OD₆₀₀ = 0.2. Cells were then centrifuged and washed twice with anoxic 50 mM glycylglycine buffer, pH 8.5, at RT. The cell pellets were then resuspended to OD₆₀₀ = 0.2 in anoxic 50 mM glycylglycine buffer, pH 8.5, supplemented with or without 0.2% glucose, and exposed to air with vigorous shaking at 37 °C. At 3-min intervals, aliquots were removed and centrifuged, and supernatants were stored on dry ice/ethanol until sample collection was complete. Supernatants were then thawed and immediately assayed for H₂O₂ concentration using Amplex Red/HRP (59). Standard curves were performed in the same media.

Determination of Inactivation Rate Constants.

To measure the rate constants with which H₂O₂ inactivates fumarases (5), 90 nM or 180 nM H₂O₂ was used to inactivate the purified/reconstituted enzymes. Reactions were performed with 10 nM enzyme at RT in 1 mL 50 mM NaPi in the anaerobic chamber; enzyme activities were measured at intervals. Rate constants were calculated. The reconstitution protocol had removed unincorporated iron, which was necessary to avoid chemical H₂O₂ degradation through the Fenton reaction.

For the measurement of inactivation rate constants of O₂⁻, a competition assay system was employed in vitro (27). In 1 mL 50 mM KPi, 2 nM B. thetaiotaomicron fumarase or 18 nM E. coli fumarase was exposed to O₂⁻ produced by xanthine oxidase in the presence of 0, 0.37, 1.48, or 3.7 nM iron-containing SOD (Sigma). Xanthine was 50 μM, and xanthine oxidase was 5 mU/mL. Catalase (200 U/mL) was included in the reaction mix to prevent enzyme damage by H₂O₂. The reaction components were mixed in the anaerobic chamber in a tube, which was then transferred out of the chamber and aerated by pipetting the reaction mixture down the side of the tube. At intervals, excess SOD was added to stop the reaction, and the remaining activity was determined by assay. SOD reacts with O₂⁻ at a rate of 2 × 10⁹ M⁻¹ s⁻¹. By comparing the inactivation rates of fumarase in the presence and absence of SOD, the inactivation rate constants of O₂⁻ can be calculated (27).

To measure the inactivation rate constants for molecular O₂, purified and reactivated 10–50 nM E. coli or B. thetaiotaomicron fumarases were added at RT to anoxic NaPi buffer containing 500 U/mL of both SOD and catalase. The buffer was then saturated with air (to achieve 220 μM O₂) by pipetting along the cuvette wall aerobically. The incubation extended for 20 min. Excess catalase and SOD (both 500 U/mL) were included in all assays. Samples were returned at intervals to the chamber and assayed under anoxic conditions. Activity was stable when enzymes were maintained under anoxic conditions.