ROS Defense Systems and Terminal Oxidases in Bacteria

Reactive oxygen species (ROS) comprise the superoxide anion (O2•−), hydrogen peroxide (H2O2), hydroxyl radical (•OH), and singlet oxygen (1O2). ROS can damage a variety of macromolecules, including DNA, RNA, proteins, and lipids, and compromise cell viability. To prevent or reduce ROS-induced oxidative stress, bacteria utilize different ROS defense mechanisms, of which ROS scavenging enzymes, such as superoxide dismutases, catalases, and peroxidases, are the best characterized. Recently, evidence has been accumulating that some of the terminal oxidases in bacterial respiratory chains may also play a protective role against ROS. The present review covers this role of terminal oxidases in light of recent findings.


Introduction
Reactive oxygen species (ROS) are partially reduced oxygen derivatives. They include the superoxide anion (O 2 •− ), hydrogen peroxide (H 2 O 2 ), hydroxyl radical ( • OH), and singlet oxygen ( 1 O 2 ). ROS can be produced within the cell as an unavoidable consequence of bacterial metabolism or derived from the environment. ROS are generated by the host innate immune system in response to bacterial colonization. Invading pathogens are recognized by pattern recognition receptors located on the surface of a phagocyte. As a result, in the course of phagocytosis, the production of ROS and reactive nitrogen species (RNS) is triggered to generate bactericidal oxidative stress [1]. O 2 •− is generated by the phagocyte NADPH oxidase. O 2 •− can then undergo dismutation to form H 2 O 2 spontaneously or enzymatically by superoxide dismutase. H 2 O 2 is also generated by many microorganisms at concentrations sufficient to kill their nearby competitors. For instance, arginine-replete Streptococcus gordonii monocultures can maintain H 2 O 2 concentrations within 20-30 µM throughout exponential growth [2]. In exponentially growing Escherichia coli (E. coli) cells, H 2 O 2 production was estimated to occur at rates of 9-22 µM/s using strains lacking intracellular scavenging enzymes and grown on a variety of growth substrates [3]. H 2 O 2 permeates freely across bacterial membranes and can react with Fe 2+ , producing a very powerful oxidant through this Fenton reaction, • OH. One more extremely dangerous ROS, 1 O 2 , can be generated by endogenous photosensitizers, such as flavins, quinones, porphyrins, and rhodopsins [4]. All these ROS, particularly • OH and 1 O 2 , can damage bacterial DNA, RNA, proteins, and lipids. To protect themselves against ROS-induced oxidative stress, bacteria utilize different ROS defense mechanisms, of which the enzymatic ROS scavengers, such as superoxide dismutases, catalases, and peroxidases are the best characterized [1,5]. Superoxide dismutases catalyze the dismutation of 2O 2 •− into H 2 O 2 and O 2 with the participation of 2H + as co-substrate. The decomposition of H 2 O 2 is usually conducted by catalases or peroxidases. Catalases catalyze the disproportionation of 2H 2 O 2 into 2H 2 O and O 2 . Peroxidases catalyze the reduction of H 2 O 2 (and/or organic hydroperoxides) by a wide variety of organic and inorganic substrates that serve as electron donor.
In the case of E. coli, the most studied species of bacteria, the following enzymes are used to degrade H 2 O 2 in vivo: the KatG and KatE catalases [6], the NADH peroxidase AhpCF [7], and the periplasmic cytochrome c peroxidase YhjA (also denoted as Ccp) that possesses quinol peroxidase activity [8,9].
Recently, evidence has been accumulated indicating that some of the enzymatic complexes of the terminal segment of the O 2 -dependent respiratory chains, terminal oxidases, may also contribute to ROS defense mechanisms in bacteria. These enzymes catalyze the four-electron reduction of O 2 to 2H 2 O using quinol or cytochrome c as the electron donor [10][11][12][13][14][15]. The membrane-embedded terminal oxidases include the superfamily of heme-copper oxidases [13,14,[16][17][18][19][20][21][22][23][24] and the family of copper-lacking bd-type oxidases (cytochrome bd) [11,[25][26][27][28][29]. All these oxidases couple the catalytic redox reaction to the generation of a proton motive force [30][31][32]. Unlike cytochrome bd [33][34][35], the hemecopper oxidases create the proton motive force not only due to the transfer of protons and electrons to the catalytic site from different sides of the membrane but also due to a unique mechanism of the proton pumping [36,37]. This is a likely reason why the proton to electron stoichiometry (characteristic of the bioenergetic efficiency) of the heme-copper oxidases is 1.5-2 times higher than that of cytochrome bd [30,38]. Heme-copper oxidases are divided into families A, B and C based on the constituents of their proton channels [39][40][41]. Cytochrome bd, in turn, can be classified into two subfamilies, S and L, based on the size of a hydrophilic region between transmembrane helices 6 and 7 of subunit I, denoted as the Q-loop [42,43]. A heme-copper oxidase usually carries three or four redox centers depending on whether it is a quinol oxidase or cytochrome c oxidase (COX). In addition to the electron entry subunit that carries a binuclear Cu A center, some COXs (caa 3 , cbb 3 ) have an additional domain, the substrate cytochrome c [44][45][46]. A distinctive feature of the heme-copper oxidase superfamily is an active site, called the binuclear center (BNC), which consists of a high-spin heme (a 3 , b 3 , or o 3 ) and a copper ion (Cu B ) close to the heme-iron. In the binuclear center, O 2 is reduced to two molecules of H 2 O. All cytochrome bds known to date are quinol (ubiquinol or menaquinol) oxidases. A typical cytochrome bd has three redox centers, hemes b 558 , b 595 , and d but no copper. The high-spin heme d is the site in which the oxygen chemistry takes place. Sometimes heme d is replaced by heme b [47]. Cytochrome bd usually reveals a much higher affinity for O 2 than heme-copper oxidases [48][49][50][51].
In this review, we discuss the contribution of the bd-type oxidases and other terminal oxidases to oxidative stress defense mechanisms in bacteria in light of recent findings.

The bd-Type Oxidases by Fast O 2 Scavenging Protect O 2 -Labile Enzymes from Oxidative Inactivation and Reduce Intracellular ROS Levels
Possibly due to the lack of proton-pumping machinery, cytochrome bd generally consumes O 2 much more rapidly than heme-copper oxidases. In E. coli and Azotobacter vinelandii, the bimolecular rate constant for O 2 reaction with the bd enzyme approaches diffusion control [82]. This trait allows the bd oxidase to play a crucial role in "respiratory protection" of nitrogenase, the O 2 -labile N 2 -fixing enzyme complex, even under aerobic conditions [83] (Figure 1). The prevention of O 2 inhibition of nitrogenase activity by cytochrome bd was shown in Azorhizobium caulinodans [84], A. vinelandii [83], Klebsiella pneumoniae [85]. This is in agreement with the fact that mutant strains lacking cytochrome bd are not able to fix nitrogen in the air [86]. Due to the presence of the bd enzyme, some bacteria classified as strict anaerobes, e.g., Bacteroides fragilis [87] and Desulfovibrio gigas [88,89], can survive at low O 2 concentrations. In this case, apart from protection against the deleterious effects of O 2 , cytochrome bd provides the bacteria with the proton motive force to drive ATP synthesis and dissipates excess reducing equivalents via the O 2 -dependent respiratory chain. Consistently, in the anoxygenic phototroph Rubrivivax gelatinosus, the bd oxidase is used to reduce the environmental O 2 pressure [90]. This expands the physiological range of ambient O 2 tensions for this bacterium under which photosynthesis can be initiated. In E. coli, a facultative anaerobic bacterium, cytochrome bd inhibits the production of intracellular H 2 O 2 by reduced fumarate reductase. This is observed when anaerobic cultures of an E. coli strain devoid of canonical H 2 O 2 -scavenging enzymes KatG, KatE, and AhpCF are abruptly aerated [91]. An underlying mechanism for this phenomenon upon aeration is likely the action of cytochrome bd as an electron sink. The bd enzyme pulls electrons away from fumarate reductase via the quinone pool. As a consequence, the rate at which fumarate reductase generates H 2 O 2 decreases [91].

Bacterial Mutants Devoid of Cytochrome bd Show Higher Sensitivity to H 2 O 2 . Cytochrome bd Expression Increases in the Presence of H 2 O 2
Cytochrome bd plays a role in protecting bacterial cells against oxidative stress caused by H 2 O 2 . E. coli mutant cells devoid of cytochrome bd-I (encoded by the cydAB operon) are extremely sensitive to H 2 O 2 exposure [92][93][94]. Consistently, expression of cytochrome bd-I in E. coli K-12 increases in the presence of external H 2 O 2 [94]. In uropathogenic E. coli, the doubling time of strains lacking either cytochrome bd-I or cytochrome bd-II (encoded by the cyxAB operon) increases considerably following treatment with 1 mM H 2 O 2 [66]. Such a protective function of the bd enzyme is not limited to E. coli strains. In the case of A. vinelandii cells, 0.15 mM H 2 O 2 appeared to be more toxic to the mutant strain devoid of the bd oxidase than to the wild-type strain [95]. The mutant strain of the sulfate-reducing bacterium Alishewanella sp. WH16-1, deficient in cytochrome bd, is also more sensitive to H 2 O 2 than the wild type and complemented strain [96]. Similarly, Brucella abortus mutants lacking the bd oxidase activity show higher sensitivity to added H 2 O 2 [97]. This sensitivity is reversed after the introduction of a plasmid (pSEK102) that contains a copy of the cydAB operon. Overexpression of superoxide dismutase and catalase can also alleviate the loss of cytochrome bd [97], emphasizing that the antioxidant properties of these enzymes are of similar importance. In Porphyromonas gingivalis involved in the pathogenesis of periodontitis, the absence of the bd oxidase leads to an increase in the susceptibility of exponentially growing bacteria to 0.5 mM H 2 O 2 [98]. The complementation of the P. gingivalis mutant with the native cydAB genes partially restores the resistance of the cells to H 2 O 2 treatment. Small et al. [99] reported the catalase-independent hyper-resistance to H 2 O 2 in Mycobacterium tuberculosis cells overexpressing the bd enzyme. The hypersensitivity of the cydAB mutants to exogenous H 2 O 2 was also documented for Mycobacterium smegmatis [100]. Consistently, in Staphylococcus aureus, the cydAB genes are strongly (by 8-9-fold) induced upon 20 min of exposure to H 2 O 2 [101]. Altogether, these data suggest that at least in a few bacteria, including pathogenic strains, cytochrome bd contributes to mechanisms that provide bacterial defense against H 2 O 2 -induced oxidative damage.

Catalase-Like Activity of Cytochrome bd
Apart from the above-described ways by which cytochrome bd can decrease intracellular ROS levels indirectly, the enzyme was reported to be able to metabolize H 2 O 2 directly. Borisov et al. [102] reported that the addition of Н 2 О 2 to the isolated as-prepared cytochrome bd-I from E. coli results in the О 2 evolution in a sealed respirometry chamber ( Figure 2, main panel). The observed rate of О 2 evolution is proportional to the enzyme concentration. The reaction rate also increases linearly with the Н 2 О 2 concentration, up to 0.2-0.5 mM of the reactant. At higher [Н 2 О 2 ], however, the dependence exhibits somewhat saturation behavior ( Figure 2, inset), which may be due to partial inactivation of cytochrome bd-I by ROS. In this reaction, there is the evolution of approximately one О 2 molecule per every two Н 2 О 2 molecules decomposed, implying the catalase-like reaction mechanism. A series of experiments show that the reaction is indeed associated with the bd-I enzyme [102]. After the thermal inactivation of cytochrome bd-I, the О 2 evolution is no longer detected. Hence, the possible presence of trace amounts of adventitious transition metals cannot be the reason for the observed О 2 evolution. The addition of NO, even at a concentration of 20 µM, does not affect the rate of O 2 formation. At the same time, NO was reported to inhibit bona fide catalase with K i of~0.18 µM [103]. Furthermore, if the bd-I enzyme is reduced completely with dithiothreitol (DTT) and 2,3-dimethoxy-5-methyl-6-(3methyl-2-butenyl)-1,4-benzoquinone (Q 1 ), the catalase-like activity is lacking. However, if bona fide catalase is then added to the chamber, the О 2 evolution resumes. It is hard to imagine that a contaminant catalase, if present, would be redox (DTT/Q 1 )-sensitive, especially as many catalases are not reducible with as strong a reducing agent as dithionite [104], even in the presence of a mediator [105]. Thus, the latter two findings suggest that the isolated untagged cytochrome bd-I, rather than a potential presence of a native catalase as a contaminant, is responsible for the observed activity. It should be noted that this conclusion is not consistent with the data of Al-Attar et al. [106]. They reported that the isolated His 6 -tagged cytochrome bd-I of E. coli does not perform a catalase-like activity as the addition of 1 mM Н 2 О 2 to the enzyme does not lead to О 2 generation [106]. Al-Attar et al. proposed that the catalase-like activity of cytochrome bd-I shown by Borisov et al. [102] might be due to impurities that include an unknown membraneassociated catalase. However, such an activity is also detected in vivo [102]. Substantial rates of О 2 production are observed if H 2 O 2 is added to respiring E. coli UM2 cells devoid of KatE and KatG but overexpressing the bd-I enzyme ( Figure 3, red line). If cytochrome bd-I is not overexpressed, the reaction is not seen (Figure 3, blue line). This can only happen if "an unknown membrane-associated catalase" in the cells is cytochrome bd-I. This discrepancy may be attributed to the differences between the protein forms (untagged vs. hexahistidine-tagged) or other experimental conditions used for protein expression and purification that Al-Attar et al. also do not exclude. Additional work is needed to resolve the controversy.  The molecular mechanism underlying the catalase-like activity of cytochrome bd-I remains unclear. To try to identify the enzyme site responsible for the observed reaction, a few compounds targeting different sites were tested [102]. Antimycin A (167 µM), which inhibits the bd-I oxidase via interaction with the quinol binding site [107], does not affect the О 2 evolution. Consistently, 250 µM oxidized Q 1 also does not inhibit the reaction. Hence, the quinol binding site does not participate in the activity. Similarly, the rate of О 2 formation is not affected by 20 µM N-ethylmaleimide, a small organic electrophile that blocks cysteine thiols through covalent modification [108]. This suggests that the enzyme thiol groups are also not involved in the reaction. Neither 20 µM NO nor 2 µM CO inhibits the О 2 evolution. The canonical О 2 reductase activity of cytochrome bd-I was reported to be blocked by NO and carbon monoxide (CO) with K i of 100 [57] and 40 nM [109], respectively. Since both NO and CO do this through binding to heme d, the participation of this heme in the catalase-like activity is not very likely. This conclusion is also supported by the fact that the catalase-like and the heme d-based О 2 reductase activities do not seem to compete with each other. The reaction is also insensitive to its product, О 2 , as the rates of О 2 evolution at 3 and 255 µM О 2 are virtually identical. Notwithstanding this, two small molecules were found to effectively inhibit the catalase-like activity, cyanide and azide.
These ligands are known to block heme-containing enzymes by targeting ferric heme-iron. The О 2 evolution is inhibited by cyanide with a K i of 2.5 µM. Consistently, 100 µM azide inhibits the activity almost completely-by 98%. The catalase-like activity appeared to be approximately three orders of magnitude more sensitive to these ligands than the heme d-based О 2 reductase one. This indicates that a heme, but not heme d, is involved in the reaction. The site at which the catalase-like chemistry occurs could be heme b 595 (Figure 4). It is pentacoordinate high-spin and therefore can potentially bind an external ligand, such as H 2 O 2 [110]. It also cannot be ruled out that this catalytic role is played by heme b 558 . Although this is a hexacoordinate low-spin heme, the bond between its sixth axial ligand Met393 and the iron ion is weak and can be replaced with a stronger external ligand [111]. Surprisingly, the addition of cyanide to the as-prepared cytochrome bd-I at a concentration (50 µM) that fully inhibits the catalase-like activity induces small absorption changes as if the ligand reacts with only some small population of heme b. If this is the case, only a fraction of the enzyme (2-4%) is involved in the reaction but with an apparent turnover number greater than 3000 s −1 [102]. The catalase-like activity of cytochrome bd-I could be induced in vivo in response to the oxidative stress by post-translational protein modification, proteolysis, protein truncation in the translation process, or interaction of the enzyme with other cellular components. . Proposed catalase-like activity of cytochrome bd-I and cytochrome bd-II from E. coli. Shown is the scheme for bd-type enzyme arrangement in the E. coli membrane bilayer based on the solved bd-I structure [28,29]. The oxidase consists of four different subunits, CydA, CydB, CydX, and CydH. CydA carries three hemes, b 558 , b 595 , and d.
Preparations of untagged cytochrome bd-II isolated from E. coli also show high catalaselike activity. Similar to cytochrome bd-I, NO at a high concentration (20 µM) does not affect the activity ( Figure 5, top panel). The observed О 2 evolution is also susceptible to the bd-II enzyme redox-state. When cytochrome bd-II is converted into the fully reduced state following the consumption of all O 2 in turnover with excess DTT and Q 1 , the H 2 O 2induced catalase-like activity is no longer observed. However, if a bona fide catalase is subsequently added, the reaction proceeds ( Figure 5, bottom panel). Further studies will show how this discovered activity of cytochrome bd-II (Figure 4) contributes to the bacterial defense mechanisms against oxidative stress in vivo. In this regard, a very recent report by Chanin et al. [112] on the role of cytochrome bd-II-mediated aerobic respiration of E. coli during intestinal inflammation deserves attention. In the course of the inflammatory process, the host produces antimicrobial products including O 2 •− to impede bacterial growth. The O 2 •− molecules generated by the Nox1 NADPH oxidase undergo rapid dismutation to H 2 O 2 and O 2 by superoxide dismutase. Using chemical and genetic murine models of noninfectious colitis, Chanin et al. showed that cytochrome bd-II provides a fitness advantage for E. coli during anaerobic growth in the presence of H 2 O 2 in the inflamed murine intestine. In the absence of Nox1, this fitness advantage is ablated. To do this, the bd-II enzyme may use H 2 O 2 or its breakdown product O 2 generated by the catalases KatE and KatG, as the substrate. It turned out that in the absence of KatE and KatG, at 5 µM H 2 O 2 , the wild-type strain outcompetes the mutant strain devoid of cytochrome bd-II. For this reason, Chanin et al. concluded that O 2 produced by catalase-mediated degradation of H 2 O 2 serves as the terminal electron acceptor for the bd-II oxidase [112]. However, given the observed catalase-like activity of cytochrome bd-II ( Figure 5, top panel), the possibility that at higher H 2 O 2 concentrations, cytochrome bd-II could also metabolize H 2 O 2 in vivo, contributing to the O 2 pool formation in the inflamed gut, cannot be excluded. Whatever the exact mechanism is, detoxification of the host-derived ROS through cytochrome bd-II allows E. coli to respire in an otherwise anaerobic environment, promoting bacterial outgrowth [112]. all O 2 is consumed and cytochrome bd-II (12.8 nM) is converted into the fully reduced state by 10 mM DTT and 250 µM Q 1 . Subsequent addition of bona fide bovine catalase (2 µg/ml) restores the reaction. Cytochrome bd-II was isolated from E. coli strain MB37 as described [38]. Changes in O 2 concentration were recorded using a high-resolution respirometer (Oxygraph-2k, Oroboros Instruments). Assays were performed at 25 • C in 50 mM Na/phosphate buffer (pH 7.0) containing 0.1 mM ethylenediaminetetraacetate (EDTA), supplemented with 0.02% dodecyl-β-D-maltoside.
Reduced catalase-like activity was determined in cell-free extracts of A. vinelandii when comparing the mutant strain MK5 devoid of the bd oxidase and the wild-type strain UW136 [95]. In Alishewanella sp. WH16-1, cytochrome bd is also suggested to catalyze the decomposition of H 2 O 2 via the catalase-like reaction (see Figure 7 in [96]). A dramatic increase in resistance of M. tuberculosis to Н 2 О 2 upon the overexpression of cytochrome bd reported by Small et al. [99] could be explained, at least in part, by the ability of the bd oxidase to perform the catalase-like reaction [113].

Peroxidase-Like Activity of Cytochrome bd
Borisov et al. reported [114] that the isolated untagged cytochrome bd-I from E. coli displays a peroxidase-like activity. Under aerobic conditions, the enzyme can catalyze the oxidation of guaiacol (o-methoxyphenol), benzohydroquinone, ferrocene, and ferrocyanide upon the addition of H 2 O 2 . Using guaiacol as the electron donor, the effect of a few inhibitors of the O 2 reductase activity of cytochrome bd-I on the peroxidase-like activity was studied. It turned out that 2-n-heptyl 4-hydroxyquinoline-N-oxide (HQNO), pentachlorophenol, and cyanide inhibit both activities at similar concentrations [114]. Based on the inhibitory analysis, it was concluded that guaiacol binds and donates electrons to the quinol binding site of cytochrome bd-I. The electrons are then transferred to the heme d site at which H 2 O 2 is bound and reduced to 2H 2 O. Although an apparent turnover number for the guaiacol peroxidation reaction is as low as about 4 s −1 , it was suggested [53] that this value could be much higher in vivo where the natural quinol is used as the electron donor.
Consistent with this, Al-Attar et al. later reported [106] that, under anaerobic conditions, the isolated His 6 -tagged cytochrome bd-I of E. coli shows significant peroxidase-like activity. As the electron donor, decyl-ubiquinol (dQH 2 ) was used and the oxidation of dQH 2 by H 2 O 2 was measured spectrophotometrically by monitoring the absorption change at 260 nm. The average dQH 2 /H 2 O 2 ratio appeared to be 1.05 ± 0.19, which is consistent with the peroxidase reaction mechanism. The k cat and K M values were reported to be 101 ± 10 s −1 and 6.6 ± 1.1 mM H 2 O 2, respectively. This gives a specificity constant k cat /K M of 1.5 × 10 4 M −1 s −1 [106]. In contrast to the catalase-like activity, the dQH 2 peroxidase reaction is promptly, but reversibly, inhibited by NO ( Figure 6). This suggests that the heme d site is directly involved in the binding and reduction of H 2 O 2 (Figure 7). The reaction is also inhibited by HQNO (50% inhibition is measured at about 10-15 µM HQNO), emphasizing that dQH 2 injects electrons directly into the quinol binding site of cytochrome bd-I. The observed high rates of the reaction indicate that it may have physiological significance in E. coli. Figure 6. Inhibition of decyl-ubiquinol (dQH 2 ) peroxidase activity of the isolated cytochrome bd-I from E. coli by NO. The reaction is monitored spectrophotometrically under anaerobic conditions. The addition of 6 µM NO promptly inhibits the enzymatic oxidation of 0.2 mM dQH 2 by 10 mM H 2 O 2 . The inhibition is reversible as the activity gradually resumes due to the disappearance of NO. The latter is probably due to the reaction between NO and dQH 2 . Reprinted from [106].

Figure 7.
Proposed peroxidase-like activity of cytochrome bd-I from E. coli. Shown is the scheme for the enzyme arrangement in the E. coli membrane bilayer based on the solved bd-I structure [28,29]. The oxidase consists of four different subunits, CydA, CydB, CydX, and CydH. CydA carries three hemes (b 558 , b 595 , d) and the quinol binding site at which the electron donor dQH 2 ) is likely bound.

ROS and Heme-Copper Oxidases
The main function of heme-copper respiratory oxidases in mitochondria and most bacteria is highly efficient energy conversion and generation of the membrane potential (the proton motive force) due to the redox energy of O 2 reduction to water [115,116]. The unique ability of heme-copper oxidases to pump protons through the membrane determines their distinctive features: the presence of a special device for a redox-coupled proton pump and intra-protein proton-conducting pathways arranged in a special way [37]. Each of the single-electron steps in the catalytic cycle of COX during the O 2 reduction in the BNC (heme a 3 /Cu B ) is associated with the transfer of~1 pumped proton through the membrane. The catalytic cycle of heme-copper oxidases is a highly coordinated system of individual electrogenic stages of electron transfer from cytochrome c on the P-side of the membrane and substrate protons on the N-side through the protein matrix to the BNC, as well as the transfer of pumped protons from the N-side of the membrane through temporary loading proton sites to the external water phase [32].
The BNC of COX is designed by nature to avoid, during the reduction of O 2 , producing of free forms of ROS, which would be released to the bulk phase. After binding of the oxygen molecule to heme a 3 in the reduced BNC, the O-O bond is broken and four electrons are transferred to O 2 in virtually one step. The heme a 3 iron gives up two electrons and is oxidized to an oxidation state of +4, while Cu B and the redox-active tyrosine residue give the other two electrons for complete reduction of the oxygen atoms to produce two molecules of water. The resulting P M catalytic intermediate is homologous to compound I of peroxidases. The P 0 compound corresponding to compound 0 in horseradish peroxidase with the bound primary H 2 O 2 adduct of the heme moiety was not time-resolved in the case of COX of mitochondria and other heme-copper oxidases of the A family. P M has an oxoferryl state of heme a 3 with the oxidized tyrosine residue (the radical form) whose reduction by an electron from cytochrome c (the third electron in the COX catalytic cycle) and protonation of the hydroxyl bound to Cu B lead to the F state. The F state is homologous to compound II of peroxidases. In the heme-copper oxidases of the B family, only the intermediate state P was kinetically resolved [117]. The intermediate state F was observed only in stationary measurements during prolonged incubation with excess H 2 O 2 (for details, see [118]). For the heme-copper oxidases of the C family, only computer calculations were reported. According to these calculations, the P M state is not energetically favorable and is not formed [119].
In addition to the main reaction, for COX from mitochondria, peroxidase-like and catalase-like activities were demonstrated. It was found that COX can catalyze the reduction of H 2 O 2 in the presence of cytochrome c, i.e., cytochrome c peroxidase-like reaction [120].
The catalase-like activity (dismutation of H 2 O 2 ) was observed initially by monitoring spectrophotometrically how the mitochondrial COX reduces the concentration of added H 2 O 2 in the absence of an external electron donor [121]. This catalase-like activity of COX was described as dismutation of H 2 O 2 with a turnover number of about 100 min −1 .
Recently, a second-order rate constant of 60-200 M −1 ·s −1 for the catalase-like activity of the bovine COX was obtained in more accurate measurements using an H 2 O 2 -sensitive electrode [122,123].
In the course of the reaction with H 2 O 2 , the BNC of COX goes through the same intermediates (P M and F), which are resolved during the O 2 reduction. The P M and F intermediates of the mitochondrial COX (with different relative ratios) can be obtained in a steady state in the presence of H 2 O 2 . The pre-steady state measurements showed that the interaction of the BNC with two H 2 O 2 molecules leads to the sequential formation of P M and the reduction of P M to F by the second H 2 O 2 molecule with the production of O 2 •− [124]. During the reaction of the mitochondrial COX with H 2 O 2 at a high concentration, two molecules of H 2 O 2 reduce the P M state formed upon the binding of the first H 2 O 2 to heme a 3 . Two molecules of O 2 •− are formed in the BNC and undergo dismutation into the new H 2 O 2 molecule [124]. At submillimolar concentrations of H 2 O 2 , its decomposition occurs at least at two sites: (i) the catalytic heme a 3 −Cu B center where H 2 O 2 is reduced to water via the P M and F states, and (ii) the surface-exposed lipid-based radicals generated due to the migration of radicals formed initially in the catalytic heme a 3 −Cu B center [125].
The mitochondrial COX can oxidize various aromatic compounds including some pharmacologically and physiologically active substances via the peroxidase mechanism [122]. Noticeably, the rates of both catalase-like and peroxidase-like activities of the mitochondrial COX are several orders of magnitude less than those for the true catalases and specific peroxidases (10 7 M −1 ·s −1 ). Hence, against the background of the specialized enzymes designed to scavenge ROS, the "parasitic" reactions (peroxidase-like and catalase-like activities) of the mitochondrial COX can be characterized as side reactions. For this reason, they are unlikely to be of physiological significance in the ROS detoxification in mitochondria. However, COX is present at a high concentration in all tissues in the body, and often there are tissues, such as the myocardium, in which there is no peroxidase at all against the background of large numbers of mitochondria. Additionally, specific localization of the enzyme in the mitochondrial membrane promotes the accumulation of hydrophobic aromatic substances. Thus, the nonspecific peroxidation catalyzed by COX via the peroxidase mechanism should be taken into account in some cases (e.g., metabolism of hydrophobic medicinal or cardiotoxic compounds) [122]. It should be noted that cytochrome c, which possesses peroxidase-like activity, could protect against ROS production in mitochondria [126].
Even though in mitochondria the function of direct ROS detoxification, a kind of "manual" work, is performed very effectively by specialized enzymes (peroxidases, catalases, superoxide dismutase, and glutathione reductases), COX nevertheless participates in the control of ROS but at a higher level of organization, through an indirect mechanism of ROS regulation in which COX performs signaling, rather than a catalytic function. The mechanism of reversible "allosteric ATP-inhibition" of dimeric COX keeps the ROS production and heat generation low in mitochondria by maintaining low values for the mitochondrial inner membrane potential [127]. This ability of COX to prevent oxygen radical formation and cellular damage is canceled by increased intracellular calcium, as a consequence of stress, which dephosphorylates and monomerizes COX.
The decomposition of H 2 O 2 by the prokaryotic aa 3 -type cytochrome c oxidases from Rhodobacter sphaeroides and Paracoccus denitrificans (homologous to the mitochondrial COX) occurs at a rate of ten or more times faster as compared to the enzyme from mitochondria (up to 2800 and 3300 M −1 ·s −1 , respectively) [122,128]. In contrast to the bovine enzyme, the observed rate of H 2 O 2 decomposition by the bacterial COXs is too high to be explained by the catalytic cleavage of H 2 O 2 in the oxygen reducing center, since the rate of H 2 O 2 binding to the BNC is significantly smaller (500-800 M −1 ·s −1 ) than the catalase-like activity. This may indicate the protective significance of these "parasitic" reactions in bacteria. There is reason to believe that the Mg ion located in the A family COXs near the proposed protonreleasing pathways (for references, see [129]) can be replaced by the Mn ion, depending on the environment in which the bacteria exist, and this ion can perform a catalytic function [122]. Meanwhile, the Mn ions are known to be part of the catalytic center of peroxidases and very good catalysts for the peroxidase reaction.
It is known that inhibition of mitochondrial respiration by NO (targeting COX) and its derivatives stimulates ROS and RNS production by mitochondria, which have signaling roles in the heart but may also contribute to cell death [130]. In contrast to the A family mitochondrial COX, which is inhibited by NO, the NO reductase activity is observed for the B family heme-copper oxidases, e.g., the ba 3 oxidase from Thermus thermophilus [131]. It is suggested that this activity may be related to the higher Cu B affinity of these enzymes for gaseous ligands. It is known that the activity of NO reductase, an enzyme related to heme-copper oxidases, provides resistance of some bacteria to the immune response of macrophages [132]. The presence of the NO reductase activity in prokaryotic heme-copper oxidases may provide pathogenic bacteria with the antioxidant capacity to protect against ROS and RNS in the course of an immune response and develop resistance against these harmful species.
Finally, in bacteria, the heme-copper oxidases of the C family (cbb 3 -type enzymes), which are expressed in low-oxygen environments, can also perform a protective function against ROS, and are in some cases very effective. The high O 2 affinity cytochrome cbb 3 , along with the bd oxidase, plays an important role in the protection of O 2 -sensitive nitrogenase in A. caulinodans by quickly consuming O 2 . The A. caulinodans mutant strain devoid of both terminal oxidases is no longer capable of fixing N 2 [84]. Akin to cytochrome bd, the cbb 3 -type oxidase is necessary to reduce the environmental O 2 pressure before anaerobic photosynthesis. Accordingly, in contrast to the wild-type R. gelatinosus strain, the double mutant lacking both cbb 3 and bd oxidases can initiate photosynthesis only in a deoxygenated medium [90]. The C family heme-copper oxidases have been much less studied than the oxidases of the other families. For the oxidases of the B and C families, variability in the stoichiometry of proton pumping was reported. How this could be related/correlated to their activity to be expressed under low O 2 conditions, as well as to the ability to suppress ROS, remains to be elucidated.

Concluding Remarks
Bacteria have evolved elaborate strategies to defend themselves from ROS and minimize oxidative damage. Many specialized detoxifying enzymes, such as superoxide dismutases, catalases, and peroxidases, have been extensively characterized. In this review, according to recent data, we report that terminal oxidases in bacterial respiratory chains may also play a protective role against ROS (Figure 8). Being efficient O 2 scavengers, both copper-lacking cytochrome bd and the heme-copper oxidase cbb 3 protect nitrogenase, the O 2 -labile enzyme complex responsible for catalyzing N 2 fixation, from inactivation by O 2 , as documented in A. caulinodans, A. vinelandii, and K. pneumoniae. The bd and cbb 3 oxidases also reduce the environmental O 2 pressure, thereby expanding the physiological range of O 2 tensions for the anoxygenic phototroph R. gelatinosus, which allows photosynthesis to start. The bd-type enzyme gives B. fragilis and D. gigas, classified as strict anaerobes, the ability to survive in low-oxygen environments. Furthermore, the E. coli cytochrome bd-I pulls electrons away from ROS-producing fumarate reductase, which leads to a reduced amount of ROS. Finally, cytochrome bd-I and cytochrome bd-II from E. coli may directly metabolize H 2 O 2 through the catalase mechanism. The former cytochrome can apparently catalyze ROS removal through another mechanism as well, acting as a quinol peroxidase.

Conflicts of Interest:
The authors declare no conflict of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, or in the decision to publish the results.