Oxygen Toxicity: A Radical Explanation

The atmosphere of planet earth was anaerobic until the advent of water-splitting, O₂-evolving photosynthesis. The accumulation of O₂ changed the environment for, and therefore changed the selection pressures on, all living organisms. It also increased the mutation rate and therefore hastened subsequent evolution. Advantages could be gained by using the O₂ to increase the useful energy derivable from foodstuffs, to carry out novel metabolic transformations, to solubilize and detoxify numerous compounds and even to generate heat and light.

But there was a price to pay for these benefits and that was to provide defenses against the considerable toxicity of this paramagnetic gas. Those organisms that succeeded in developing the requisite defenses could reap the benefits, and they gave rise to the enormous variety of aerobic life forms that are now so evident on earth. Those that could not accommodate to the challenge of O₂ toxicity evolved into the sensitive microscopic anaerobes now restricted to those anaerobic niches that remain even on a thoroughly aerobic planet. So, why is O₂ toxic and what sorts of defenses have been evolved to blunt that toxicity?

Rotating electrical charges generate magnetic fields. This applies to electrical current in a coil of wire or to a single spinning electron. The pairing of electrons with opposite spin states neutralizes this effect. Most substances are not influenced by imposed magnetic fields because the electrons they contain are all in spin-opposed pairs. Such substances are diamagnetic. O₂ is unusual in being paramagnetic, and that implies unpaired electronic spins. Indeed, O₂ contains two unpaired electrons and they have the same spin state.

The reduction of O₂ to 2H₂O requires four electrons. Hence, intermediates will be encountered on this univalent pathway and these are superoxide (O₂⁻ ), hydrogen peroxide (H₂O₂) and hydroxyl radical (HO^●). It is these intermediates that are responsible for the toxicity of O₂, and defenses against that toxicity must include minimizing their production to the maximum extent possible and eliminating those whose production cannot be avoided.

Most of the O₂ consumed by respiring cells is reduced by cytochrome c oxidase which, by virtue of two ferrihemes and two Cu(II) prosthetic groups, manages the four-electron reduction of O₂ to 2H₂O without releasing intermediates. But there are enzymes that reduce O₂ to H₂O₂, and there are both enzymic and spontaneous processes within cells that produce O₂⁻. It has been estimated that only approximately 0.1% of the O₂ reduced by Escherichia coli is reduced to O₂⁻ (Imlay and Fridovich, 1991). Nevertheless, so great is the rate of O₂ utilization by these cells that, were the O₂⁻ stable, this would correspond to the production of approximately 5 μmol l⁻¹ intracellular O₂⁻ per second. Similarly, in mitochondria, a small fraction of total O₂ reduction gives rise to O₂⁻ (Gardner and Boveris, 1990).

Cells are rich in reductants, such as thiols and enediols, and these are able to reduce Fe(III) to Fe(II), thus obviating the need for reaction 2. None the less, O₂⁻ does collaborate with H₂O₂ in producing HO^●within cells and it does so by oxidizing the [4Fe–4S] clusters of dehydratases, such as aconitase, causing the release of Fe(II). In this way, O₂⁻ increases the availability of iron for reactions 3a–3d. This mechanism had been proposed (Liochev and Fridovich, 1994) and was subsequently experimentally verified (Keyer et al. 1995; Keyer and Imlay, 1996).

A common selection pressure applied to a varied biota is apt to call forth multiple adaptations. In this light, it is not surprising that we find multiple defenses against O₂⁻. These are the superoxide dismutases (SODs) which catalyze reaction 6 at diffusion-limited rates. These supremely efficient catalysts are abundant in aerobic cells and they keep the steady-state level of O₂⁻ in the 10⁻¹⁰ mol l⁻¹ range (Imlay and Fridovich, 1991). It is useful to point out that with [SOD] at approximately 10⁻⁵ mol l⁻¹ and [O₂⁻ ] at 10⁻¹⁰ mol l⁻¹, an O₂⁻ is 10⁵ times more likely to encounter a molecule of SOD than it is to encounter another O₂⁻. Add to this the fact that the rate constant for the reaction of O₂⁻ with SOD is k₂≈2×10⁹ l mol⁻¹ s⁻¹, while that of its uncatalyzed reaction with another O₂⁻ at neutral pH is k≈2×10⁵ l mol⁻¹ s⁻¹. Hence, the lifetime of an O₂⁻ would be shortened by a factor of 10⁹ by the SOD. This would be the case if the spontaneous and the SOD-catalyzed dismutations were the only fates open to O₂⁻. However, there are targets that would be attacked by O₂⁻, were it not removed by SOD. It has recently been estimated that the SODs in E. coli provide approximately 95% protection for all targets susceptible to O₂⁻ attack in that cell (Liochev and Fridovich, 1997). Returning to the multiplicity of SODs, we note that there are SODs that depend for their activity on active sites containing Cu and Zn, Mn, Fe and even Ni. There are SODs that are cytosolic, localized to specific subcellular organelles and also secreted from the cell. We will consider these in turn.

How is it possible for an enzyme to react with its substrate at a rate of approximately 2×10⁹ l mol ⁻¹ s⁻¹, which is close to the diffusion limit? This question arises because the Cu site, at which reactions with O₂⁻ occur, represents less than 1% of the surface of the enzyme. On the basis of random collisions, one would expect more than 99% of the collisions between O₂⁻ and the SOD to be fruitless. Koppenol (1982) was the first to consider this problem, and he suggested electrostatic guidance as a possible answer. Subsequently, the superoxide dismutases were seen to utilize electrostatic facilitation on the basis of ionic strength effects (Benovic et al. 1983; Cudd and Fridovich, 1982). Site-specific modifications have been used to enhance electrostatic facilitation and thus to make a more active SOD (Getzoff et al. 1992).

An extracellular SOD (ECSOD), which is a homotetrameric glycoprotein with a high affinity for heparin sulfate, has been described (Tibell et al. 1993). It presumably functions to scavenge the O₂⁻ that is released from the surfaces of cells. The discovery of the signaling roles of NO, and the extremely rapid reaction of NO with O₂⁻, increases our appreciation of the importance of extracellular SOD. Knockout mice, which lack ECSOD, appear superficially normal under ordinary conditions, but they exhibit enhanced sensitivity towards hyperoxia (Carlsson et al. 1995). Parasitic nematodes have been found to make extracellular SOD (Tang et al. 1994; James et al. 1994). In the case of such parasites, the ECSOD may provide a defense against O₂⁻ produced by host leukocytes. Microorganisms have also been seen to produce extracellular SODs, and these may similarly act as pathogenicity factors. Thus, Mycobacterium avium make an extracellular MnSOD (Escuyer et al. 1996), as does Nocardia asteroides (Alcendor et al. 1995).

Mutations in the cytosolic CuZnSOD have been associated with the familial form of amyotrophic lateral sclerosis (FALS). To date, approximately 50 different amino acid replacements have been seen in the CuZnSOD of FALS patients, and the activities of these mutant enzymes have ranged from 0.1 to 100% of normal. That range of activities, as well as the genetic dominance of FALS, suggested that a toxic gain of function, rather than a loss of SOD activity, was the problem. This was established to be the case when transgenic mice expressing FALS-associated mutant CuZnSODs were seen to develop paralysis (Dal Canto and Gurney, 1994). Transgenic mice expressing normal human CuZnSOD did not develop symptoms of paralysis.

These enzymes, which are as active as the CuZnSODs, but are unrelated as judged by sequence, may be dimeric or tetrameric. They contain one Mn(III) per subunit and their structures have been determined by X-ray crystallography (Wagner et al. 1993). The E. coli MnSOD is dimeric and it is not ordinarily produced when the cells are growing anaerobically. However, it is induced under aeration and is further induced by compounds that can increase intracellular O₂⁻ production. Compounds such as viologens, quinones, pyocyanine and a host of synthetic dyes are in this category (Hassan and Fridovich, 1979). This induction of the E. coli

MnSOD is controlled by the soxRS regulon, which will be further discussed below.

There is a homotetrameric MnSOD in the matrix of mitochondria, which is closely related, in terms of sequence, to the prokaryotic enzyme. It is fascinating that this eukaryotic organelle should contain a SOD very similar to the prokaryotic enzyme and totally unrelated to the SOD in the surrounding cytosol. This is certainly supportive of the endosymbiotic origin of these organelles. Knockout mice, which are unable to make this mitochondrial MnSOD, are severely affected and live only a few days after birth (Li et al. 1995). The deleterious consequences of knocking out the cytosolic CuZnSOD are notable but much less dramatic (Reaume et al. 1996; Kondo et al. 1997).

It behoves a facultative enteric organism such as E. coli to contain a constitutive SOD, in addition to the inducible MnSOD, so that it will not be devoid of defense against O₂⁻ when faced with the abrupt transition from anaerobic to aerobic conditions. In E. coli, this standby defense is provided by an FeSOD which, although specific for the metal it contains, is related by sequence to the MnSOD. There are some SODs, such as those found in Bacteroides fragilis (Gregory and Dapper, 1983) and Propionibacterium shermanii (Meier et al. 1982), which can be active with either Fe or Mn at the active site. These are called cambialistic SODs. The organisms that contain these cambialistic SODs insert Fe under anaerobiosis and Mn under aerobiosis. This is a sensible adaptation since environmental Fe(II) would autoxidize to the much less soluble Fe(III) and thus be less available in the presence of oxygen. Mn, in contrast, remains Mn(II), and soluble, under both conditions. This, of course, leaves open the question of why these organisms insert Fe when anaerobic.

Lactobacillus plantarum, which ordinarily lives in the Mn-rich environment provided by fermenting plants, shows a fascinating adaptation. It accumulates large amounts of Mn(II), approximately 25 mmol l⁻¹, which serves as a functional replacement for SOD. Inorganic Mn(II) is much less efficient than SOD in eliminating O₂⁻, but at 25 mmol l⁻¹ its activity is sufficient to meet the defensive needs of this aerotolerant organism. When grown in medium relatively poor in Mn(II), L. plantarum became intolerant of O₂ (Archibald and Fridovich, 1981).

Some comment about SOD activity in anaerobes is needed. Why should an anaerobe such as Bacteroides fragilis need a SOD? O₂⁻ will not be encountered under anaerobiosis. However, it is virtually impossible always to avoid some exposure to O₂. Even methanogens have been shown to contain SOD, which is probably needed to survive transient exposure to O₂ while en route from one anaerobic niche to another. The consequences of lacking SOD have been clarified with E. coli. Thus, mutants unable to make either the FeSOD or the MnSOD exhibited oxygen-dependent nutritional auxotrophies, an increased mutation rate, hypersensitivity to paraquat and slow growth (Carlioz and Touati, 1986). Yet these sodA sodB mutants were not really devoid of SOD since they retained the periplasmic CuZnSOD. A sodA sodB sodC mutant of E. coli has not yet been reported. It should be noted that the periplasmic CuZnSOD of E. coli is present in small amounts relative to the MnSOD and FeSOD (Benov and Fridovich, 1994) and it is unusual in being monomeric (Battistoni et al. 1996). There is enough sequence homology between prokaryotic and eukaryotic CuZnSODs to indicate a distant common evolutionary origin (Imlay and Imlay, 1996).

A homotetrameric nickel-containing SOD has recently been described from Streptomyces griseus, which also contains a homotetrameric FeSOD. The NiSOD had a subunit mass of 13 kDa, while that of the FeSOD was 22 kDa, and these two enzymes showed no serological cross reactivity. Our appreciation of the diversity of the SOD family thus continues to grow.

The catalases, which dismute 2H₂O₂ into O₂ + 2H₂O, and the peroxidases, which use diverse reductants to reduce H₂O₂ to 2H₂O, are the enzymes that deal with H₂O₂. Some of the peroxidases can also reduce alkyl hydroperoxides to the corresponding alcohols. As was the case with the SODs, the catalases and peroxidases constitute a diverse family of enzymes. Most of these are ferriheme enzymes, and their action involves the divalent oxidation of the heme to an Fe(IV) π cation radical by H₂O₂, followed by divalent reduction by H₂O₂, in the case of catalase, and by two successive univalent reductions by the organic substrate, in the case of the peroxidases (Dolphin et al. 1971).

Mammalian catalases are homotetrameric ferriheme-containing enzymes whose subunit mass is approximately 60 kDa. These enzymes are most efficient when dealing with relatively high concentrations of H₂O₂ because their K_m for H₂O₂ lies in the millimolar range. Hence the packaging of catalase into peroxisomes, along with many H₂O₂-producing enzymes. Mammalian catalase can also act as a peroxidase towards a few small molecules such as methanol, ethanol, nitrite and formate. Thus, it can use H₂O₂ to oxidize these substrates, which are small enough to gain access to the heme iron. The structure, as determined by X-ray crystallography, indicates that the heme lies deeply buried in the protein and is thus accessible only to small substrates (Reid et al. 1981). Catalase contains tightly bound NADPH, which may function to prevent the accumulation of an inactive Fe(IV) form of the enzyme (Kirkman et al. 1987). Yet the structure shows no interaction between the NADPH and the heme (Fita and Rossman, 1985).

E. coli makes two catalases, which have been named hydroperoxidases (HP) I and II (Claiborne and Fridovich, 1979; Claiborne et al. 1979). HP-I is unusual in that it is active as a catalase and as a peroxidase, which can utilize large reducing substrates such as dianisidine. HP-II, in contrast, is a catalase without this peroxidase activity. The significance of the peroxidase activity of HP-I is not known since it had no peroxidase activity towards any dialyzable component of E. coli extract. HP-I is induced as a member of the oxy R regulon in response to H₂O₂, while HP-II is induced as cells enter stationary phase by the sigma factor rpos (Visick and Clarke, 1997). The structures of the genes coding for HP-I and HP-II of E. coli have been elucidated (Loewen and Stauffer, 1990; von Ossowski et al. 1991). HP-I and HP-II are differently localized in E. coli, with HP-I appearing in the periplasm and associated with the cytoplasmic membrane, while HP-II is cytosolic (Heimberger and Eisenstark, 1988).

The importance of catalase is nicely illustrated by the behavior of L. plantarum, which cannot synthesize heme. When heme is available in the medium, this organism makes a heme-containing catalase. When heme is not available, it makes a different catalase which contains manganese (Kono and Fridovich, 1983). This manganese catalase was called pseudocatalase because, unlike the more familiar heme-containing catalases, it was not inhibited by CN⁻ or N₃⁻. Such is the need for catalase that L. plantarum contrives to supply itself with this activity whether or not heme is available.

Enzymes that use a variety of electron donors to reduce H₂O₂ to 2H₂O are widespread. Thus, yeast contains a cytochrome c peroxidase, plants contain ascorbate peroxidase as well as peroxidases acting on a variety of phenols and amines, and E. coli makes an NADPH peroxidase which is called alkyl hydroperoxide reductase. The principal peroxidase in mammals is glutathione peroxidase (GSH Px). Of course, GSH is oxidized to the corresponding disulfide during its action, but glutathione reductase converts that back to GSH, using NADPH as the reductant. GSH Px is important not only for the elimination of H₂O₂. Its specificity encompasses alkyl hydroperoxides as well, and it reduces them to the corresponding alcohols. As was the case with the SODs and the catalase, there are several GSH Pxs, one of which is secreted and found in the extracellular space (Perry et al. 1992), and another acts specifically on phospholipid hydroperoxides (Ursini et al. 1985). All have proved to be seleno-enzymes. GSH is present and abundant in most cells but is not found in trypanosomes which contain, in its place, diglutathionyl spermidine. This compound has been named trypanothione, and the organisms that contain it also produce a trypanothione peroxidase and a trypanothione reductase. The latter enzyme seems to be a pathogenicity factor in Leishmania (Dumas et al. 1997).

Phagocytic leukocytes, when stimulated, increase their rate of O₂ consumption 15-to 20-fold in what has been called the respiratory burst. This is due to the activation of a membrane-bound NADPH oxidase which reduces O₂ to O₂⁻ (Curnutte et al. 1974). The respiratory burst is an important component of the armamentarium used by these leukocytes to kill invading microorganisms, and its genetic lack results in a hypersusceptibility to infection termed chronic granulomatous disease (Jendrossek et al. 1997). The O₂⁻ produced during the respiratory burst is converted to H₂O₂ by the dismutation reaction, and the H₂O₂ is used to oxidize Cl⁻ to hypochlorite under the catalytic influence of myeloperoxidase (Klebanoff, 1996). Hypochlorite is bacteriocidal and is, moreover, the precursor of N-chlorotaurine, which is also an effective antibacterial compound.

Removal of O₂⁻, by the superoxide dismutases, and of H₂O₂, by the catalases and peroxidases, is important; but a well-rounded defense requires more. This is beautifully illustrated by the response of E. coli to conditions that increase intracellular O₂⁻ production. Under this stress, E. coli activates the soxRS regulon, which is a family of approximately 12 coordinately regulated genes whose products provide the needed defenses. The SoxR protein is the sensor. It is a [2F–2S] protein that can exist in a reduced and an oxidized state. It is the oxidized state that transcriptionally activates the soxS gene, and the SoxS protein in turn then activates all the other genes in the regulon (Gaudu and Weiss, 1996). What are the members of this regulon?

MnSOD is one and of course its role is to remove O₂⁻. Glucose-6-phosphate dehydrogenase is another and it serves to supply the NADPH needed by glutathione reductase and alkyl hydroperoxide reductase, among others. Fumarase C is part of this regulon and it serves as an O₂⁻-stable replacement for fumarases A and B, which are rapidly inactivated by O₂⁻. Aconitase A is controlled by SoxRS because the aconitases too are rapidly inactivated by O₂⁻ (Gruer and Guest, 1994), and more synthesis is needed to replace the activity lost to O₂⁻. Ferredoxin-flavodoxin reductase is a member of this regulon probably because reduced ferredoxin/flavodoxin function in the reductive reactivation of fumarases A and B, aconitases A and B and the other [4Fe–4S]-containing dehydratases that are inactivated by O₂⁻. Endonuclease IV is a member because it serves in the repair of oxidatively damaged DNA. This incomplete listing of the members of the SoxRS regulon gives some appreciation of what it takes to blunt the negative impacts of O₂⁻. A recent review provides more detail (Nunoshiba, 1996).

The stress imposed upon E. coli by H₂O₂ turns on a distinct group of genes under the control of oxyR, which is transcriptionally active only in its oxidized state. The proteins whose production is positively regulated by OxyR include hydroperoxidase I, alkyl hydroperoxide reductase and glutathione reductase. There are at least half a dozen more that have yet to be identified. The OxyR regulon has been reviewed (Iuchi and Weiner, 1996). Although the proteins whose production is known to be regulated by OxyR make perfect sense in terms of defense against H₂O₂, it is surprising that the soxRS regulon is activated by O₂⁻ and the oxyR by H₂O₂. This is because one would think that a source of O₂⁻ would necessarily also be a source of H₂O₂ because of the dismutation reaction. One might therefore expect that a source of O₂⁻ should activate both soxRS and oxyR.

The multiplicity of the defenses that have arisen to deal with oxidative stress serves as an index of the gravity of this stress. Yet no defense can be perfect. This is specially the case in living cells, which must balance finite resources to meet diverse needs. In fact, we have just enough defense to allow reproductive success, which is what evolution requires. We are thus subject to low-level, but chronic, oxidative damage. It is therefore not surprising that oxidative damage contributes to ‘spontaneous’ mutation, to senescence and to numerous pathologies. We may anticipate great strides in moderating these types of damage, or at least in understanding their bases, now that the role of oxidative stress is coming to light.

This work was supported by grants from the Council for Tobacco Research – USA, the National Institutes of Health and the United States Army Medical Research Command.