Riding the tiger – physiological and pathological effects of superoxide and hydrogen peroxide generated in the mitochondrial matrix

Abstract

Elevated mitochondrial matrix superoxide and/or hydrogen peroxide concentrations drive a wide range of physiological responses and pathologies. Concentrations of superoxide and hydrogen peroxide in the mitochondrial matrix are set mainly by rates of production, the activities of superoxide dismutase-2 (SOD2) and peroxiredoxin-3 (PRDX3), and by diffusion of hydrogen peroxide to the cytosol. These considerations can be used to generate criteria for assessing whether changes in matrix superoxide or hydrogen peroxide are both necessary and sufficient to drive redox signaling and pathology: is a phenotype affected by suppressing superoxide and hydrogen peroxide production; by manipulating the levels of SOD2, PRDX3 or mitochondria-targeted catalase; and by adding mitochondria-targeted SOD/catalase mimetics or mitochondria-targeted antioxidants? Is the pathology associated with variants in SOD2 and PRDX3 genes? Filtering the large literature on mitochondrial redox signaling using these criteria highlights considerable evidence that mitochondrial superoxide and hydrogen peroxide drive physiological responses involved in cellular stress management, including apoptosis, autophagy, propagation of endoplasmic reticulum stress, cellular senescence, HIF1α signaling, and immune responses. They also affect cell proliferation, migration, differentiation, and the cell cycle. Filtering the huge literature on pathologies highlights strong experimental evidence that 30-40 pathologies may be driven by mitochondrial matrix superoxide or hydrogen peroxide. These can be grouped into overlapping and interacting categories: metabolic, cardiovascular, inflammatory, and neurological diseases; cancer; ischemia/reperfusion injury; aging and its diseases; external insults, and genetic diseases. Understanding the involvement of mitochondrial matrix superoxide and hydrogen peroxide concentrations in these diseases can facilitate the rational development of appropriate therapies.

Keywords:

• Mitochondria
• matrix
• ROS
• disease
• redox signaling
• antioxidants
• superoxide
• hydrogen peroxide
• mCAT
• mitoTEMPO
• mitoTEMPOL
• peroxiredoxin
• thioredoxin
• S1QEL
• S3QEL
• SOD2

Introduction

Aerobic life must deal with a dilemma. The most abundant and accessible source of energy on Earth is sunlight, and that energy is stored by plants primarily as reduced carbon (–CH₂–, –CHOH–, etc.). The best way to release the energy is to pass electrons from the reduced carbon substrates to the most oxidizing acceptor that is widely available, oxygen. However, electrons are very reactive and hard to control, so riding the tiger of abundant energy release carries the risk of premature electron escape to oxygen to form reactive oxygen species (ROS) that can drive damage and disease – the tiger may bite you.

Mitochondria take electrons from reduced carbon substrates such as fatty acids, amino acids and tricarboxylic acid cycle intermediates using appropriate substrate dehydrogenases. They pass these electrons down an electron transport chain, and use them two pairs at a time to reduce molecular oxygen (O₂) to water (2H₂O). The redox energy that is released by substrate oxidation is coupled to proton pumping from the mitochondrial matrix to the intermembrane space, and the resulting protonmotive force across the mitochondrial inner membrane, consisting of a membrane potential and a pH gradient, is used to drive ATP synthesis during oxidative phosphorylation. Occasionally, however, electrons leak prematurely from the respiratory chain to reduce O₂ singly to form the superoxide radical anion (O₂^•−), or in pairs to form hydrogen peroxide (H₂O₂) (Boveris and Chance 1973; Chance et al. 1979). Superoxide and hydrogen peroxide are the two primary ROS from which others, such as hydroxyl radical and peroxynitrite, are formed.

There is a large literature on the roles of mitochondrially-generated superoxide and hydrogen peroxide in cellular signaling (Cadenas 2004; Starkov 2008; Hamanaka and Chandel 2010; Ray et al. 2012; Sena and Chandel 2012; Guo et al. 2013; Dai et al. 2014; Holmstrom and Finkel 2014; Shadel and Horvath 2015; Brand 2016; Diebold and Chandel 2016; Siauciunaite et al. 2019; Watson et al. 2019; Sies and Jones 2020). An even larger literature invokes mitochondrial ROS production as a cause of oxidative damage and pathology (Cadenas and Davies 2000; Raha and Robinson 2000; Lenaz 2001; Turrens 2003; Brand et al. 2004; Brookes et al. 2004; Cadenas 2004; Adam-Vizi 2005; Andreyev et al. 2005; Balaban et al. 2005; Brookes 2005; Jezek and Hlavata 2005; Zorov et al. 2006; Starkov 2008; Kowaltowski et al. 2009; Lambert and Brand 2009; Murphy 2009; Brand 2010; Jastroch et al. 2010; Rigoulet et al. 2011; Drose and Brandt 2012; Sena and Chandel 2012; Chen and Zweier 2014; Holmstrom and Finkel 2014; Sies 2014; Andreyev et al. 2015; Orr et al. 2015; Brand 2016; Brand et al. 2016; van ‘t Erve et al. 2017; Crooks et al. 2018; Murphy and Hartley 2018; Watson et al. 2019; Sies and Jones 2020). The present review outlines the conditions that favor superoxide and hydrogen peroxide production by electron leak from different sites in the mitochondrial electron transport chain, and then discusses the evidence supporting roles for superoxide and hydrogen peroxide in subsequent biological effects: physiological signaling and pathological damage.

Sites of mitochondrial production of superoxide and hydrogen peroxide and their relative importance

At least eleven distinct sites in the mitochondrial electron transport chain and associated substrate dehydrogenases have been characterized and shown to generate superoxide or hydrogen peroxide at measurable rates using isolated mitochondria. The characteristics of these sites are reviewed elsewhere (Brand et al. 2004; Brand 2010; Quinlan et al. 2013; Brand 2016; Wong et al. 2017, 2019). Each site generates superoxide on the matrix side of the mitochondrial inner membrane, or within the matrix itself, and some of them may also generate hydrogen peroxide as the primary product. Superoxide dismutase-2 (SOD2; MnSOD) in the matrix converts superoxide to hydrogen peroxide, so all sites can contribute to superoxide and hydrogen peroxide levels in the mitochondrial matrix. Hydrogen peroxide (but not superoxide) can diffuse out of the mitochondria to the cytosol, so in principle all sites can also contribute to cytosolic hydrogen peroxide levels. Two sites, one in complex III and one in mitochondrial glycerol 3-phosphate dehydrogenase, produce about half of their superoxide directly into the intermembrane space (St-Pierre et al. 2002; Muller et al. 2004; Brand 2016) and can therefore contribute to superoxide levels in the cytosol as well as in the mitochondrial matrix. Each of these eleven sites is a potential source of superoxide or hydrogen peroxide in cells and in vivo, and might contribute to redox signaling and overt pathology. At one time it was reasonable to question whether mitochondria make any significant contribution to cellular hydrogen peroxide levels (Brown and Borutaite 2012), but there is now good evidence that they do: in resting cultured C2C12 myoblasts and myocytes, mitochondria are the largest single contributor to the pool of hydrogen peroxide in the cytosol (Wong et al. 2018; Goncalves et al. 2020), and in a range of cell types they contribute about 30% of the total (Fang et al. 2020).

Measurements of the rates and relative importance of different mitochondrial sites of superoxide and hydrogen peroxide production in cells are sparse, but four sites, sites I_Q and I_F in complex I, II_F in complex II, and III_Qo in complex III (where subscripts ‘Q’ and ‘F’ denote the quinone and flavin sites, respectively), dominate in mitochondria isolated from rat skeletal muscle and incubated in a medium mimicking the cytosol in muscle at rest (Goncalves et al. 2015, 2020). S1QELs (Brand et al. 2016) and S3QELs (Orr et al. 2015) are small molecules that suppress electron leak from the respiratory chain to oxygen to form superoxide and hydrogen peroxide at sites I_Q and III_Qo, respectively. Importantly, they do so without inhibiting forward or reverse electron transport in the respiratory chain (Wong et al. 2019) and without inhibiting oxidative phosphorylation. Measurement of the acute effects of S1QELs and S3QELs on total cellular hydrogen peroxide production has enabled extension of these ex vivo observations (Goncalves et al. 2015, 2020) to intact cells and organisms, and has shown that sites I_Q and III_Qo dominate the mitochondrial contribution to cytosolic hydrogen peroxide levels in cultured C2C12 myoblasts and myocytes (Wong et al. 2018; Goncalves et al. 2020). Just one site, site I_Q, is responsible for at least half of the superoxide and hydrogen peroxide generated in the mitochondrial matrix in a range of cell types cells, and most of the remainder originates from site III_Qo (Wong et al. 2018; Fang et al. 2020). Site I_Q is also a significant contributor to hepatic and cardiac oxidative burden in vivo (Wong et al. 2020). By extrapolation, the physiological and perhaps the pathological effects caused by matrix production of superoxide or hydrogen peroxide that are discussed below may be driven largely by superoxide or hydrogen peroxide derived from sites I_Q and III_Qo.

Mitochondrial levels of superoxide and hydrogen peroxide

What reactions lead to elevated physiological or pathological levels of superoxide or hydrogen peroxide derived from a particular mitochondrial site? The concentration of superoxide in the mitochondrial matrix will be set by the rate of its production (reaction 3 in Figure 1) and the rate of its removal (primarily reaction 4 in Figure 1). When superoxide is produced faster than it is removed, the concentration of superoxide will rise progressively and reaction 4 will be driven progressively faster until removal occurs at the same rate as production and a steady state is achieved. In the same way, the concentration of hydrogen peroxide in the mitochondrial matrix at steady state will be set by the rate of its production (reaction 4 in Figure 1, plus any direct production from mitochondrial sites) and the rate of its removal (primarily reactions 5 and 6 in Figure 1). Other factors that are commonly discussed, such as tissue specificity, protein expression level, electron transport rate, rate of oxygen consumption, mitochondrial dysfunction, respiratory chain inhibition, substrate preference, substrate concentration, calcium uptake, mitochondrial membrane potential, uncoupling, mitochondrial permeability transition pore, NAD redox state, ubiquinone redox state, allosteric modification of proteins, post-translational modification of proteins by phosphorylation or acetylation, antioxidant status, and so on, must always operate solely through changes in these proximal rates. Elevated steady-state levels of matrix superoxide or hydrogen peroxide are caused by increased production rates or decreased activity of removal pathways, or both.

Figure 1. Pathways of mitochondrial superoxide and hydrogen peroxide production and removal. X*⁻ is the reduced form of one of eleven redox centers (X) in the mitochondrial electron transport chain and associated substrate dehydrogenases; ‘*^-’ denotes the species that reacts with O₂, usually its single-electron reduced form, such as a semiquinone in complex I or complex III. Reactions 1-6 are discussed in the text. SOD2: superoxide dismutase-2; PRDX3: peroxiredoxin-3; GPX1: glutathione peroxidase-1; mCAT: catalase expressed transgenically in the mitochondrial matrix; *OH: hydroxyl radical (see color version of this figure at www.tandfonline.com/ibmg).

(i) Rates of production and removal of mitochondrial superoxide

What determines the underlying rates of production and removal? The reduction of O₂ to superoxide (1-electron reduction) or hydrogen peroxide (2-electron reduction) by a relevant electron donor (X*⁻) normally follows simple second-order chemical kinetics. Findings of little dependence on oxygen level at physiological oxygen tension (Hoffman et al. 2007) have been superceded (Grivennikova et al. 2018; Stepanova et al. 2019). Therefore, the rate of production of superoxide (or of hydrogen peroxide, if it is a direct product of reaction 3 in Figure 1) depends only on the concentration of X*⁻, on the local concentration of O₂, and on the rate constant k for their reaction.

The concentration of X*⁻ is determined by the size of the pool of X (for example, by the amount of the relevant respiratory complex), by the rate of X*⁻ production by upstream reductants (or nominally downstream reductants driven back up to X by reverse electron transport) (reaction 1 in Figure 1) and by the kinetics of X*⁻ consumption by downstream oxidants (reaction 2 in Figure 1). The rate of reaction 1 can be altered by the supply of electrons (which is determined, for example, by succinate or pyruvate concentration, or by matrix NADH/NAD⁺ ratio) and by the activities of the enzymes or electron transport chain components that pass electrons from such substrates to X. These enzymes can be regulated (for example, by the level of free Ca²⁺ in the mitochondrial matrix, which affects the activity of pyruvate, 2-oxoglutarate and isocitrate dehydrogenases), or their activities can be decreased by appropriate enzyme or electron transport chain inhibition. The rate of reaction 2 can be altered by changes in the downstream electron transport chain activity (for example, by the addition of respiratory poisons, by changed energy demand [which alters the redox state of downstream chain components], or by altered rate constants for electron transport). The local concentration of O₂ can be altered by hypoxia or hyperoxia at the site (caused, for example, by altered oxygen supply or delivery to the tissue, or by altered respiration rate, which changes the local oxygen tension and intracellular oxygen gradients). The rate constant, k, for the reaction of X*⁻ with O₂ can be altered in many ways, for example by mutations, conformational changes, post-translational modifications, or small molecules such as S1QELs and S3QELs.

Based on these principles, at any given O₂ tension, the rate of superoxide and hydrogen peroxide production by site I_F is determined by the amount of active complex I, the size of the matrix NADH plus NAD⁺ pool, and the matrix NADH/NAD⁺ ratio (Kussmaul and Hirst 2006; Treberg et al. 2011; Quinlan, Treberg, et al. 2012; Quinlan et al. 2014). The rate of superoxide and hydrogen peroxide production by site I_Q is determined by the amount of active complex I, the matrix NADH/NAD⁺ ratio, the redox state of the ubiquinone pool (QH₂/Q ratio), the mitochondrial membrane potential, and the mitochondrial pH gradient (Lambert and Brand 2004; Lambert et al. 2010; Treberg et al. 2011; Robb et al. 2018). The rate of superoxide and hydrogen peroxide production by site II_F is determined by the amount of active complex II, the QH₂/Q ratio, and the succinate and other dicarboxylate concentrations (Quinlan, Orr, et al. 2012; Quinlan, Treberg, et al. 2012). The rate of superoxide production by site III_Qo is determined by the amount of active complex III, the QH₂/Q ratio, and the mitochondrial membrane potential (Quinlan et al. 2011; Quinlan, Treberg, et al. 2012).

Very few of the electrons flowing down the respiratory chain in cells or in vivo leak prematurely to O₂ to form superoxide or hydrogen peroxide. We estimate that only 0.35% do so in skeletal muscle at rest, dropping to 0.01% during exercise (Goncalves RL et al. 2015). These numbers give a sense of scale to the biological fluxes of electrons, but they are not useful measures of the rate of production of superoxide or hydrogen peroxide. The rate of electron transport through the electron transport chain (Figure 1) affects superoxide and hydrogen peroxide production only through its effects on the concentrations of various X*⁻ species (and through any effects on intracellular oxygen gradients). The level of a particular X*⁻ can be increased either by increasing reaction 1 and raising respiration rate, or by decreasing reaction 2 and lowering respiration rate. For this reason, the rate of respiration does not predict the rate of superoxide or hydrogen peroxide production. The proportion of electrons flowing down the respiratory chain that end up forming superoxide depends completely on the manipulations made during the measurements, and changes in this proportion are uninformative about the intensity of production of superoxide or hydrogen peroxide in physiology or pathology; the absolute rates or levels are much more informative.

The rate of removal of matrix superoxide (reaction 4 in Figure 1) is determined primarily by the rate of dismutation catalyzed by SOD2, the only superoxide dismutase normally present in the matrix (Weisiger and Fridovich 1973). Spontaneous dismutation can also be significant (and is the dominant pathway when matrix superoxide levels rise in SOD2 null cells). Rates of reaction with other species are normally quantitatively minor (although potentially of great biological significance), except for the reaction with nitric oxide (NO) to form peroxynitrite, which can be a significant flux when NO production rate is high.

(ii) Rates of production and removal of mitochondrial hydrogen peroxide

Production of hydrogen peroxide in the matrix is normally catalyzed mostly by SOD2, with some contribution from spontaneous dismutation (together comprising reaction 4 in Figure 1) plus any direct production from mitochondrial sites. Any diffusion into the mitochondrial matrix from the rest of the cell will also contribute; when this happens extensively mitochondria can be considered to be sinks of hydrogen peroxide, potentially acting as a buffer of cytosolic hydrogen peroxide levels (Mailloux 2018). There are several mechanisms for removal of hydrogen peroxide from the matrix. The main pathway is thought to be catalyzed by peroxiredoxin-3 (PRDX3), which is targeted to the mitochondrial matrix (Chae et al. 1999) and has been calculated to clear 90% of matrix hydrogen peroxide (Cox et al. 2009), using reducing equivalents from thioredoxin-2 generated by thioredoxin reductase-2 using matrix NADPH. Under conditions of stress, peroxiredoxin 6 may also be trafficked to the mitochondrial matrix (Eismann et al. 2009). Glutathione peroxidases, primarily GPX1, also play a role, using reducing equivalents from matrix glutathione generated by glutathione reductase using matrix NADPH. Together these make up reaction 5 in Figure 1. Diffusion of matrix hydrogen peroxide to the cytosol (reaction 6 in Figure 1) is also quantitatively and biologically significant. Reactions with other species are normally quantitatively minor (although potentially of great biological significance). Since superoxide in the matrix is converted almost quantitatively into hydrogen peroxide by SOD2 or spontaneous dismutation, the addition of pure SOD mimetics will not alter the rate of hydrogen peroxide production or the matrix concentration of hydrogen peroxide (unless NO is very high, allowing a competition for superoxide to ensue) but will lower the steady-state concentration of superoxide (see below). The kinetics of hydrogen peroxide removal from the matrix can be altered in several ways, for example by addition of catalase mimetics, by changed glutathione levels in the matrix (Mari et al. 2009), or by mutations or inhibitors that affect other matrix antioxidant systems. Some of the best evidence for a significant role for mitochondrial hydrogen peroxide in driving different phenotypes comes from experiments using genetically engineered expression of catalase in the mitochondrial matrix (mCAT), where it is not normally found (Dai et al. 2014, 2017).

Mechanisms driving biological effects of mitochondrial superoxide or hydrogen peroxide

The mechanism by which a physiological change or a pathological insult increases the mitochondrial concentrations of superoxide and hydrogen peroxide derived from a specific site or sites and leads to biological effects will be by altering one or more of the variables discussed above. In ischemia-reperfusion injury the mechanism is thought to be altered substrate supply, as succinate built up during ischemia is used on reperfusion to drive reverse electron transport to reduce sites I_Q and I_F, leading to increased production of superoxide and hydrogen peroxide, increased levels of these species, and reperfusion injury (Chouchani, Pell, et al. 2016). In various mitochondrial diseases it may be simply inhibition of the reoxidation of X*⁻ caused by a loss-of-function mutation in downstream electron transport complexes or the ATP synthase, or it may be a gain-of-function change in the rate constant for superoxide production, as in certain succinate dehydrogenase mutations in Caenorhabditis elegans (Adachi et al. 1998). In acetaminophen toxicity it is largely the decrease in removal pathways initiated by depletion of glutathione by direct adduct formation, and inhibition of mitochondrial glutathione peroxidases, both caused by reactive metabolites of acetaminophen (Du et al. 2016).

Altering the kinetics of superoxide and hydrogen peroxide removal using transgenes and antioxidants

A widespread problem in the literature is the lack of precision in setting up hypotheses, in particular, the discussion of various reactive oxygen species as if they were all equivalent and interchangeable – they are not (Murphy et al. 2011; Sies and Jones 2020). When the nature and reactions of a particular reactive oxygen species are considered, it becomes obvious that different manipulations, such as SOD2 overexpression, mCAT expression, addition of the targeted lipid peroxidation chain-breaker mitoQ, or addition of the glutathione precursor N-acetylcysteine are not equivalent and cannot meaningfully be lumped together as “anti-ROS” or “antioxidant” treatments (Davis et al. 2001; Murphy et al. 2011; Lark et al. 2015; Sies and Jones 2020). It is important to emphasize that decreases (including complete knockout), or increases (even to saturating levels) in the activities of the enzymes that remove superoxide and hydrogen peroxide do not usually have much effect on the rates of production of superoxide or hydrogen peroxide by mitochondria. Instead, such changes alter the steady-state concentrations of these species in the mitochondrial matrix.

Consider Figure 1 again. The superoxide concentration in the mitochondrial matrix is set by a steady state in which its rates of production and consumption are equal. Following some arbitrary imposed increase in mitochondrial superoxide production (from complex I or other sites facing the matrix; reaction 3), the matrix superoxide concentration will rise until, in a new steady-state, it drives a new rate of consumption that is equal to the new rate of production. Because they operate very far from equilibrium, the reactions that generate superoxide are essentially irreversible and insensitive to superoxide concentration, so there will be no direct secondary or compensatory change in their rates in response to the rising concentration of their product, superoxide. Some superoxide will react with other molecules (proteins, lipids, NO, mtDNA etc.; the minor fluxes) but, overwhelmingly, the major consumers are the conversion of superoxide to hydrogen peroxide by SOD2-catalyzed and spontaneous dismutation (reaction 4). Under cellular conditions these consumer reactions are again essentially irreversible and independent of the concentration of their product, hydrogen peroxide.

These properties have two important consequences. First, because the dismutation rate (reaction 4) is so much higher than the rates of the other reactions (minor fluxes), the superoxide that is generated is converted almost quantitatively to hydrogen peroxide, at a rate determined by the superoxide generation rate, and (because of spontaneous dismutation) independent of the activity of SOD2. Second, the steady-state concentration of superoxide (and hence the rates of the minor side-reactions) is very sensitive to the activity of SOD2. If SOD2 is activated (or pure SOD mimetics are added), matrix superoxide concentration will drop until the steady state is reestablished at almost the same consumption rate and hydrogen peroxide production rate as before the change, but at a lower level of superoxide and hence lower rates of the quantitatively minor side-reactions. If SOD2 activity is decreased, such as in a SOD2 heterozygous animal, matrix superoxide concentration will be higher but hydrogen peroxide production will hardly be affected (it will slow very slightly as the side-reactions from superoxide increase and carry a little more of the total flux). In the limit, if SOD2 is removed completely, as in a SOD2 knockout, matrix superoxide will rise until it reaches a high level sufficient to drive all the flux to hydrogen peroxide by spontaneous dismutation (and the relatively minor rates of the side-reactions with other molecules will be maximal). Thus, modification or mimicry of SOD2 activity will change matrix superoxide concentration, with potentially significant effects of changes in rate of the superoxide side-reactions, but will have almost no effect on matrix or cellular hydrogen peroxide concentration. See (Gardner et al. 2002) for a simple quantitative model of how altered SOD2 activity can increase, not affect, or decrease hydrogen peroxide production; only in the few exceptional cases in which endogenous pathways that consume matrix superoxide without forming hydrogen peroxide run at a significant rate will increased SOD2 activity cause increased hydrogen peroxide production (by successfully competing for the pool of superoxide). Thus, the major fluxes from superoxide to hydrogen peroxide will be essentially independent of manipulations in the kinetics of reaction 4. Instead, the steady state level of superoxide and therefore the minor fluxes of its other reactions will be very sensitive to manipulations in the kinetic properties of reaction 4, for example by altered expression of SOD2 or the addition of SOD mimetics.

Similar considerations apply to the steady state that determines the matrix hydrogen peroxide concentration. The arbitrary imposed increase in mitochondrial superoxide production will drive a matching rate of hydrogen peroxide production, and the matrix hydrogen peroxide concentration will rise until it drives a new rate of consumption, mostly by peroxiredoxin-3 as discussed above, that is equal to the new rate of production. Other than diffusion to and from the cytosol, the producers and consumers of hydrogen peroxide (SOD2 and spontaneous dismutation; peroxiredoxin-3, mCAT, glutathione peroxidases, catalase mimetics) run far from equilibrium and are therefore essentially irreversible and product-independent, so the same rules apply: modulations of the consumers (reaction 5) will affect the hydrogen peroxide concentration in the matrix and the rate of escape of hydrogen peroxide from the mitochondria (reaction 6), but will have virtually no effect on hydrogen peroxide production rate (reaction 4) or on matrix superoxide levels. Thus, modification of peroxiredoxin-3 activity or expression of mCAT will change matrix (and therefore cytosolic) hydrogen peroxide concentration, with potentially significant effects of changes in rate of the hydrogen peroxide side-reactions, but will have no effect on matrix superoxide production rate or concentration.

Different antioxidants that act in the mitochondria will have different effects depending on their properties. Those that react directly with superoxide will lower superoxide level by adding a new superoxide consumer to the steady state. This will decrease the rates of the superoxide side-reactions, and lower hydrogen peroxide level by diverting superoxide flux. Those that react directly only with hydrogen peroxide will not affect superoxide concentration, but will decrease hydrogen peroxide concentration (by adding a new hydrogen peroxide consumer to the steady state) and decrease the rates of the hydrogen peroxide side-reactions. Antioxidants that affect downstream reactions, such as those that act as chain-breakers in lipid peroxidation reactions, will affect lipid peroxidation and downstream reactions, but will have no direct effect on superoxide and hydrogen peroxide steady state levels or the biological effects they drive. S1QELs and S3QELs will decrease premature electron flow to O₂ and lower the levels of both superoxide and hydrogen peroxide, as will oxidation of respiratory chain components. If a particular phenotype is caused by matrix superoxide and not by matrix hydrogen peroxide, etc. (see Figure 1), then its responses to different “antioxidants” may be very different, sowing confusion unless the nature of “ROS” is more thoughtfully specified.

It follows from these considerations that different manipulations of “ROS” levels can be much more diagnostic of which species is causal than is commonly assumed. If a phenotype is prevented by SOD2 overexpression and S1QEL or S3QEL addition but not by peroxiredoxin-3 overexpression or mCAT expression, then it is caused by matrix superoxide and not by matrix hydrogen peroxide. Similar arguments apply to other combinations of responses to genetic and pharmacological interventions. The possible responses of a particular phenotype to relevant experimental manipulations and the interpretations are shown in Figure 2. Thus, different “antioxidants” can have profoundly different effects, and general conclusions that take no account of this property can be very misleading, and miss important implications that could be extracted from a particular dataset.

Figure 2. Identification of mitochondrial sources of superoxide and hydrogen peroxide that cause a given phenotype using different manipulations that affect that phenotype.

Proximal and more distal targets of matrix superoxide and hydrogen peroxide

What are the immediate targets of superoxide and hydrogen peroxide in the mitochondrial matrix? The chemical reactivities of superoxide and hydrogen peroxide are very different, so they have different molecular targets (Collin 2019; Sies and Jones 2020). The negatively charged superoxide radical anion, O₂^•−, is reactive and a good oxidant of positively charged Fe-S centers in proteins. The classic example is the [4Fe-4S]²⁺ center of the tricarboxylic acid cycle enzyme aconitase in the mitochondrial matrix, which is very sensitive to inactivation by superoxide (and 10,000 times less sensitive to hydrogen peroxide) (Gardner 2002; Imlay 2006; Castro et al. 2019). Its inactivation may lead directly to inappropriate lipogenesis (Crooks et al. 2018). Succinate dehydrogenase (complex II) in the matrix is also extremely sensitive to superoxide (Melov et al. 1999; Powell and Jackson 2003; Lustgarten et al. 2011; Brand et al. 2016), probably by inactivation of one of its Fe-S centers. The protonated form of superoxide, perhydroxyl radical (HOO^•), is uncharged and lipid-soluble. It reacts with polyunsaturated fatty acyl side-chains of mitochondrial phospholipids, particularly cardiolipin (because it tethers four fatty acyl side-chains in close proximity), to initiate cascades of lipid hydroperoxide formation that alter the functions of the mitochondrial inner membrane, and also react with mitochondrial DNA and membrane proteins.

In contrast to superoxide, hydrogen peroxide is relatively stable, but can rapidly oxidize specific reactive protein thiolates (which are more prominent in the mitochondrial matrix because of its alkaline pH), to the sulfenic and sulfinic acids, and lead to the formation of disulfides. An important thiolate target of hydrogen peroxide in the mitochondrial matrix may be cysteine residues on pyruvate dehydrogenase kinase 2 (PDHK2). Their oxidation by hydrogen peroxide inhibits PDHK2 activity, resulting in activation of the pyruvate dehydrogenase complex and increased oxidation of pyruvate (Hurd et al. 2012). Hydrogen peroxide is membrane-permeant, so it will diffuse to the cytosol where, mostly through modification of protein thiols, it can orchestrate changes in protein activity through activation of MAP protein kinases (Siauciunaite et al. 2019) and the HIF1α and other pathways (Klimova and Chandel 2008; Diebold and Chandel 2016), and changes in gene transcription through redox-regulated transcription factors such as c-Jun, ATF2, ATF4, NFkB, Nrf2 and p53 (Siauciunaite et al. 2019).

These reactions of superoxide and hydrogen peroxide generate secondary species that are even more reactive than the parent species. Oxidation of Fe-S centers by superoxide releases Fe²⁺ into the matrix, which in turn catalyzes Haber-Weiss chemistry to generate ^•OH radical from superoxide and hydrogen peroxide, and Fenton chemistry to generate ^•OH radical from hydrogen peroxide. ^•OH radicals are highly reactive but have only weak specificity, and react at diffusion-limited rates with proteins to inactivate them, mtDNA bases to cause mutation, and polyunsaturated fatty acyl side-chains of mitochondrial phospholipids. Matrix superoxide can also activate mitochondrial uncoupling proteins (Echtay, Murphy, et al. 2002; Echtay, Roussel, et al. 2002), perhaps through lipid peroxidation reactions (Brand et al. 2004). HOO^• reacts rapidly with nitric oxide to limit its concentration and signaling potential, forming peroxynitrite (ONOO^-), a powerful tyrosine-nitrating and methionine-oxidizing agent. Hydrogen peroxide reacts with bicarbonate to form peroxymonocarbonate (HCO₄⁻), which reacts with thiols much more readily than the parent hydrogen peroxide (Collin 2019; Sies and Jones 2020).

Biological effects

As laid out in the Introduction, there are large and unwieldy bodies of literature on the roles of mitochondria-derived superoxide and hydrogen peroxide in cellular signaling and pathology. Table 1 lists phenotypes that are probably driven by such signaling pathways. Pathologies that may be driven by mitochondrial production of superoxide or hydrogen peroxide are listed in Table 2.

Table 1. Published evidence that particular physiological processes are driven or modified by mitochondrial production of superoxide and/or hydrogen peroxide, according to the numbered criteria in Figure 3.

	1. S1QELs and S3QELs	2. SOD2	3. PRDX3 and mCAT	4. Mitochondria-targeted antioxidants – mitoTEMPO	5. Mitochondria-targeted chain-breakers – mitoQ, SkQ, SS-31
Apoptosis	(Orr et al. 2015) (Brand et al. 2016) (Mezera et al. 2017)	(Wong et al. 1989) (Li and Oberley 1997) (Kokoszka et al. 2001) (Van Remmen et al. 2001) (Epperly et al. 2003) (Dolgachev et al. 2005) (Kasahara et al. 2005) (Strassburger et al. 2005) (Raimundo et al. 2012)	(Wonsey et al. 2002) (Chang et al. 2004) (Arita et al. 2006) (De Simoni et al. 2008) (Dai et al. 2010) (Baek et al. 2017)	(Liang et al. 2010) (Wang et al. 2013) (Lin et al. 2019)	(Kelso et al. 2001) (Zhao et al. 2005) (Pletjushkina et al. 2006) (Thomas et al. 2007) (Mizuguchi et al. 2008) (Ojano-Dirain and Antonelli 2012) (Rehman et al. 2016) (Hao et al. 2018) (Zhou et al. 2018)
Autophagy		(Zhang et al. 2010) (Case et al. 2011) (Li et al. 2013)	(Dai and Rabinovitch 2011) (Jiang et al. 2017)	(Jong et al. 2015) (Ye et al. 2017) (Park and Chung 2019)	(Xiao et al. 2017) (Kim, Choi, et al. 2018) (Chen et al. 2019)
Cell Cycle, Proliferation, Growth	(Pohl et al. 2018) (Amoedo et al. 2020)	(Liu et al. 1997) (Zhong et al. 1997) (Li et al. 1998a) (Li et al. 1998b) (Rodriguez et al. 2000) (Weydert et al. 2003) (Kim et al. 2004) (Ough et al. 2004) (Ozeki et al. 2004) (Zhong et al. 2004) (Hodge, Peng, et al. 2005) (Hodge, Xiao, et al. 2005) (Venkataraman et al. 2005) (Kalen et al. 2006) (Weydert et al. 2006) (Menon et al. 2007) (Lee et al. 2009) (Kim et al. 2010) (Sarsour et al. 2012) (Wang et al. 2012)	(Rodriguez et al. 2000) (Wonsey et al. 2002) (Yang et al. 2007) (Chua et al. 2010) (Goh et al. 2011) (Nonn et al. 2003) (Chua et al. 2010) (D’Souza et al. 2013) (Liu, Hu, et al. 2016)	(Shimura et al. 2016) (Lopez-Acosta et al. 2018) (Lian et al. 2019) (Yang et al. 2020)	(Schauen et al. 2006) (Velando et al. 2019) (Yang et al. 2020)
Cell Migration	(Amoedo et al. 2020)	(Connor et al. 2007) (Liu et al. 2015)	(Liu, Hu, et al. 2016)	(Porporato et al. 2014) (Lian et al. 2019) (Yang et al. 2020)	(Porporato et al. 2014) (Yang et al. 2020)
Cellular Senescence		(Velarde et al. 2012)	(Dai et al. 2010) (Wu et al. 2016)	(Mistry et al. 2013) (Konig et al. 2017) (Chen et al. 2018)	(Victorelli et al. 2019) (Vizioli et al. 2020)
Differentiation		(Gao J et al. 2018)	(Nemoto et al. 1990) (Lee, Tak, et al. 2011)	(Lee, Tak, et al. 2011)	(Lee, Tak, et al. 2011) (Kim, Choi, et al. 2018)
Endoplasmic Reticulum Stress Propagation	(Orr et al. 2015) (Brand et al. 2016) (Fang D et al. 2017) (Mezera et al. 2017)	(Zhong et al. 2016)	(Bronner et al. 2015)	(Liu, Wang, et al. 2018)	(Escribano-Lopez et al. 2019)
Gut Microbiome			(Yardeni et al. 2019)
Hormetic Antioxidant Defense		(Van Raamsdonk and Hekimi 2009) (Schaar et al. 2015) (Cox et al. 2018)
Hypoxia – HIF1α Pathway	(Orr et al. 2015)	(Wang et al. 2005) (Bullock et al. 2009) But (Brunelle et al. 2005)	(Wang et al. 2005) (Guo et al. 2012)	(Nazarewicz, Dikalova, Bikineyeva, Ivanov, et al. 2013) (Chowdhury et al. 2018)	(Sanjuan-Pla et al. 2005) (Guzy et al. 2008)
Hypoxic Pulmonary Vasoconstriction	(Sommer et al. 2017)			(Sommer et al. 2017)	(Dallas et al. 2009)
Immune Response (Adaptive)	(Oberkampf et al. 2018)	(Case et al. 2011)	(Oberkampf et al. 2018)		(Case et al. 2011) (Sena et al. 2013)
Immune Response (Innate) see also Table 2: Inflammation	(Ahmed et al. 2019)	(Wang, Jin, et al. 2017)	(West et al. 2011) (Bronner et al. 2015) (Lee et al. 2016)	(Ding et al. 2016) (Ding et al. 2017) (Ahmed et al. 2019)	(Bulua et al. 2011) (Formentini et al. 2017) (Han, Xu, et al. 2018) (Wang, Sun, et al. 2018)
Sexual Signaling					(Cantarero and Alonso-Alvarez 2017)
Sleep		(Hill et al. 2018)
Synaptic Pruning see also Table 2: Angelman Syndrome					(Calkins et al. 2011) (Sidlauskaite et al. 2018)
Thermogenesis		But (Silva et al. 2005)			(Chouchani, Kazak, et al. 2016)
Wound Healing		(Luo et al. 2004) (Marrotte et al. 2010) (Muliyil and Narasimha 2014)		(Kim and Park 2019)	(Demianenko et al. 2010) (Demyanenko et al. 2015) (Demyanenko et al. 2017) (Tamer et al. 2018)

Table 2. Published evidence that particular pathologies are driven or modified by mitochondrial production of superoxide and/or hydrogen peroxide, according to the numbered criteria in Figure 3.

	1. S1QELs and S3QELs	2. SOD2	3. PRDX3 and mCAT	4 Mitochondria-targeted antioxidants – mitoTEMPO	5. Mitochondria-targeted chain-breakers – mitoQ, SkQ, SS-31	6. Gene Associations
2.1 Aging and aging-associated diseases
Aging, Lifespan	(Watson et al. 2020)	(Kokoszka et al. 2001) (Sun et al. 2002) (Harris et al. 2003) (Curtis et al. 2007) (Hu et al. 2007) (Paul et al. 2007) (Cabreiro et al. 2011) (Treiber et al. 2011) But (Van Remmen et al. 2003) (Jang et al. 2009) (Perez et al. 2009) (Cabreiro et al. 2011)	(Schriner et al. 2005) but (Treuting et al. 2008) (Dai et al. 2017)		(Anisimov et al. 2008) (Skulachev et al. 2009) (Anisimov et al. 2011) (Siegel et al. 2013) (Braakhuis et al. 2018)	(Stessman et al. 2005) (Soerensen et al. 2009) But (De Benedictis et al. 1998) (Taufer et al. 2005) (Khabour et al. 2009) (Gentschew et al. 2013)
Aging – Cardiovascular Disease (see also Atherosclerosis, Cardiomyopathy)		(Strassburger et al. 2005) (Kwak et al. 2015) (Vendrov et al. 2017)	(Schriner et al. 2005) (Dai et al. 2009) (Dai et al. 2010) (Canugovi et al. 2019)	(Vendrov et al. 2015) (Owada et al. 2017) (Vendrov et al. 2017) (Olgar et al. 2018) (Olgar et al. 2020)	(Gioscia-Ryan et al. 2014) (Manskikh et al. 2015) (Gioscia-Ryan et al. 2018)
Aging – Cognitive Decline (see also Alzheimer’s Disease)		(Paul et al. 2007)	(Chen et al. 2014) (Csiszar et al. 2019)	(Tarantini et al. 2019)	(McManus et al. 2011) (Stelmashook et al. 2015) (Tarantini et al. 2018) (Olesen et al. 2020)	(Salminen et al. 2017)
Aging – Gastrointestinal System (see also Cancer - Colon; Inflammatory Bowel Disease)				(Koç et al. 2019)
Aging – Kidney Dysfunction (see also Diabetic Nephropathy; Kidney Disease)					(Sweetwyne et al. 2017)
Aging – Skin (see also Radiation Injury)		(Treiber et al. 2011) (Velarde et al. 2012) (Scheurmann et al. 2014)	(Baek et al. 2017)	(Park et al. 2018)	(Victorelli et al. 2019)	But (Gao W et al. 2017)
Aging – Thymus Involution					(Obukhova et al. 2009) (Skulachev et al. 2009)
Muscle Atrophy – Age, Disuse Denervation		(Lee et al. 2009) (Treiber et al. 2011) but (Kuwahara et al. 2010) (Lustgarten et al. 2011)		(Zhou et al. 2019)	(Min et al. 2011) (Powers et al. 2011) (Talbert et al. 2013) (Szeto and Birk 2014)
Ocular Diseases – Cataract		(Reddy et al. 2004) But (Van Remmen et al. 2003) (Wolf et al. 2005)	(Schriner et al. 2005) (Wolf et al. 2005)		(King et al. 2004) (Neroev et al. 2008) (Skulachev et al. 2009) (Skulachev 2013) (Cai et al. 2015) (Babizhayev 2016)	Not (Celojevic et al. 2013)
Olfaction (Age-related Loss)		(Paul et al. 2007) (Hussain et al. 2018)
Osteoporosis		(Treiber et al. 2011) (Kim et al. 2017)	(Bartell et al. 2014) (Almeida et al. 2017) (Ucer et al. 2017)	(Domazetovic et al. 2017) (Song et al. 2018)	(Skulachev et al. 2009) (Song et al. 2018) (Kang et al. 2020)	(Deng et al. 2011) (Botre et al. 2015) But (Mlakar et al. 2012)
Reproductive Aging – Female (see also Placenta; Pre-eclampsia)		But (Raineri et al. 2001)		(Yang et al. 2018) (Yang et al. 2019)	(Tsybul’ko et al. 2010) (Ding et al. 2019) (Marei et al. 2019)	(Ruiz-Sanz et al. 2011) (Pournourali et al. 2016) (Moshtaghi et al. 2017) (Liu, Liu, et al. 2019)
Reproductive Aging – Male		But (Raineri et al. 2001)	(Selvaratnam and Robaire 2016)	(Lu et al. 2018) (Zhang, Lu, et al. 2019)	(Tsybul’ko et al. 2010) (Fang L et al. 2014) (Liu, Wang, et al. 2016) (Jiang et al. 2018) (Ibrahim et al. 2019) (Zhang, Bao, et al. 2019)	(Ji et al. 2012) (Faure et al. 2014) (Yan L et al. 2014) But (Zhu et al. 2017)
Sarcopenia: see Muscle Atrophy
Spinal deformities – Scoliosis, Kyphosis		(Treiber et al. 2011)			(Shabalina et al. 2017)
2.2 Cancer (incidence, growth, metastasis)
Cancer Incidence, Risk	(Reddy et al. 1993)	(Van Remmen et al. 2003)	(Wonsey et al. 2002) (Treuting et al. 2008) (Woo et al. 2012)	(Shetty et al. 2019)	(Agapova et al. 2008) (Skulachev et al. 2009)	(Ambrosone et al. 1999) (Ratnasinghe et al. 2000) (Mitrunen et al. 2001) (Wang et al. 2001) (Stoehlmacher et al. 2002) (Woodson et al. 2003) (Cai et al. 2004) (Hung et al. 2004) (Liu, Zhou, Park, et al. 2004) (Liu, Zhou, Wang, et al. 2004) (Millikan et al. 2004) (Tamimi et al. 2004) (Wang et al. 2004) (Bergman et al. 2005) (Taufer et al. 2005) (Cox et al. 2006) (Slanger et al. 2006) (Ergen et al. 2007) (Kang et al. 2007) (Landi et al. 2007) (Choi et al. 2008) (Ezzikouri et al. 2008) (Mikhak et al. 2008) (Wheatley-Price et al. 2008) (Bica et al. 2009) (Chan et al. 2009) (Cheng et al. 2009) (Gallagher et al. 2009) (Hernandez-Saavedra and McCord 2009) (Wang, Wang, et al. 2009) (Zejnilovic et al. 2009) (Mao, Qiu, et al. 2010) (Kucukgergin, Sanli, Amasyali, et al. 2012) (Kucukgergin, Sanli, Tefik, et al. 2012) (Aynali et al. 2013) (Sun et al. 2013) (Atilgan et al. 2014) (Parlaktas et al. 2015) (Zhang, Wang, et al. 2017) (Wang, Zhu, et al. 2018) But (Bag and Bag 2008) (Wang, Wang, et al. 2009)
Cancer Growth	(Amoedo et al. 2020)	(Church et al. 1993) (Li et al. 1995) (Yan et al. 1996) (Zhong et al. 1996) (Kiningham and Clair 1997) (Liu et al. 1997) (Stclair et al. 1997) (Zhong et al. 1997) (Li et al. 1998b) (Zhang et al. 1999) (Weydert et al. 2003) (Ough et al. 2004) (Ridnour et al. 2004) (Venkataraman et al. 2005) (Weydert et al. 2006)	(Wonsey et al. 2002) (Goh et al. 2011) (Woo et al. 2012)	(Weinberg et al. 2010) (Nazarewicz, Dikalova, Bikineyeva, Ivanov, et al. 2013)	(Agapova et al. 2008) (Diebold and Chandel 2016)	(Martin et al. 2006) (Bewick et al. 2008)
Cancer Metastasis		(Safford et al. 1994) (Yan et al. 1996) (Weydert et al. 2003) (Ough et al. 2004)	(Yan et al. 1996) (Goh et al. 2011) (Fatemie et al. 2012)	(Weinberg et al. 2010) (Porporato et al. 2014) (Diebold and Chandel 2016)	(Porporato et al. 2014)	(Martin et al. 2006) (Bewick et al. 2008)
Cancer – Bladder						(Hung et al. 2004) But (Terry et al. 2005) (Kucukgergin, Sanli, Amasyali, et al. 2012)
Cancer – Breast	(Lubet et al. 1997)	(Li et al. 1995) (Zhang et al. 1999) (Weydert et al. 2006)	(Goh et al. 2011) (Fatemie et al. 2012)	(Luo, Shang, et al. 2018)		(Ambrosone et al. 1999) (Zhang et al. 1999) (Mitrunen et al. 2001) (Cai et al. 2004) (Millikan et al. 2004) (Bergman et al. 2005) (Taufer et al. 2005) (Martin et al. 2006) (Slanger et al. 2006) (Bewick et al. 2008) (Bica et al. 2009) (Wang, Wang, et al. 2009) But (Egan et al. 2003) (Tamimi et al. 2004) (Gaudet et al. 2005) (Cox et al. 2006) (Bag and Bag 2008) (Sun et al. 2009)
Cancer – Colon			(Woo et al. 2012)			(Stoehlmacher et al. 2002)
Cancer – Esophagus						(Cheng et al. 2009) (Sun et al. 2013) But (Sun et al. 2009)
Cancer – Kidney						(Atilgan et al. 2014)
Cancer – Larynx						(Aynali et al. 2013) But (Coskun et al. 2016)
Cancer – Leukemia					(Agapova et al. 2008) (Skulachev et al. 2009)	But (Banescu et al. 2014)
Cancer – Liver				(Shetty et al. 2019)		(Ezzikouri et al. 2008)
Cancer – Lung	(Lam et al. 2002) (Amoedo et al. 2020)	(Yan et al. 1996)				(Ratnasinghe et al. 2000) (Wang et al. 2001) (Liu, Zhou, Park, et al. 2004) (Liu, Zhou, Wang, et al. 2004) (Wang et al. 2004) (Gallagher et al. 2009) (Hernandez-Saavedra and McCord 2009) (Zejnilovic et al. 2009) (Sun et al. 2013) But (Lin et al. 2003)
Cancer – Melanoma		(Church et al. 1993)		(Nazarewicz, Dikalova, Bikineyeva, Ivanov, et al. 2013)
Cancer – Mesothelioma						(Landi et al. 2007)
Cancer – Pancreas		(Weydert et al. 2003) (Ough et al. 2004)				(Wheatley-Price et al. 2008)
Cancer – Prostate		(Li et al. 1998b)				(Woodson et al. 2003) (Taufer et al. 2005) (Ergen et al. 2007) (Kang et al. 2007) (Choi et al. 2008) (Mikhak et al. 2008) (Bica et al. 2009) (Chan et al. 2009) (Wang, Wang, et al. 2009) (Mao, Qiu, et al. 2010) (Sun et al. 2013) (Parlaktas et al. 2015) (Zhang, Wang, et al. 2017) (Wang, Zhu, et al. 2018) But (Tefik et al. 2013)
2.3 Cardiovascular diseases (hypertension, atherosclerosis, cardiomyopathy, heart failure)
Atherosclerosis, Coronary Artery Disease (see also Cardiomyopathy)		(Du et al. 2001) (Ballinger et al. 2002) (Ohashi et al. 2006) (Brown et al. 2007) (Wang, Shi, et al. 2012) (Zhou et al. 2012) (Vendrov et al. 2017)	(Wang, Wang, et al. 2017)	(Vendrov et al. 2015) (Vendrov et al. 2017)	(Han et al. 2009) (Hulsmans et al. 2012) (Eirin et al. 2014) (Zhang, Zhao, et al. 2017)	(Kakko et al. 2003) (Gottlieb et al. 2005) (Tian et al. 2012) (Souiden et al. 2016) (Yeh et al. 2018)
Cardiomyopathy – Fibrillation			(Xie et al. 2015)	(Sovari 2011)
Cardiomyopathy – Heart Failure with preserved Ejection Fraction HFpEF (Diastolic)			(Dai et al. 2010) (Sverdlov et al. 2016)	(Owada et al. 2017)	(Wang, Sun, et al. 2018)
Cardiomyopathy – Heart Failure with reduced LV Ejection Fraction HFrEF (Dilated CM) (Systolic) (see also Diabetic Cardiomyopathy)		(Li et al. 1995) (Lebovitz et al. 1996) (Van Remmen et al. 2001) (Strassburger et al. 2005) (Nojiri et al. 2006) (Richters et al. 2011) (Kang et al. 2015) (Almomani et al. 2020)	(Matsushima et al. 2006) (Treuting et al. 2008) (Dai et al. 2010) (Dai, Johnson, et al. 2011) (Dai and Rabinovitch 2011) (Dai et al. 2012) (Hoshino et al. 2014)	(Hoshino et al. 2014) (Ku et al. 2015) (Ku and Park 2017) (Owada et al. 2017) (Olgar et al. 2018) (Olgar et al. 2020)	(Dai, Chen, et al. 2011) (Dai et al. 2013) (Manskikh et al. 2015) (Yancey et al. 2015) (Sabbah et al. 2016) (Daubert et al. 2017) (Rivera-Barahona et al. 2017) (Ribeiro Junior et al. 2018) (Goh et al. 2019) (Kim et al. 2020) But (Yancey et al. 2015)	(Hiroi et al. 1999)
Hypertension – Cardiac		(Rodriguez-Iturbe et al. 2007) (Dikalova et al. 2010) (Dikalov and Nazarewicz 2013) (Kroller-Schon et al. 2014) (Itani et al. 2016) (Dikalova et al. 2017) But (Fennell et al. 2002)	(Dai, Chen, et al. 2011) (Dai, Johnson, et al. 2011) (Itani et al. 2016) (Dikalova et al. 2017)	(Dikalova et al. 2010) (Nazarewicz, Dikalova, Bikineyeva, Dikalov 2013) (Dikalov et al. 2014) (Dikalova et al. 2015) (Park et al. 2016) (Dikalova et al. 2017) But (Nautiyal et al. 2019)	(Graham et al. 2009) (Dai, Chen, et al. 2011) (Nazarewicz, Dikalova, Bikineyeva, Dikalov 2013) (McLachlan et al. 2014) (Vilaseca et al. 2017) (Rossman et al. 2018)	(Nakagawa et al. 2013)
Hypertension – Pulmonary Arterial Hypertension (see also Pulmonary Fibrosis; Chronic Obstructive Pulmonary Disease)	(Diolez et al. 2019)	(Rodriguez-Iturbe et al. 2007) (Archer et al. 2010) (Fu et al. 2020) (Song et al. 2020) But (Adesina et al. 2015)	(Adesina et al. 2015)	(Zhang, Ma, et al. 2016) (Song et al. 2020)	(Pak et al. 2018) (Suresh et al. 2018) (Suresh et al. 2019)	(Xu et al. 2017)
Pre-eclampsia			(Li et al. 2010)	(McCarthy and Kenny 2016) (Vaka et al. 2018)	(Vaka et al. 2018) (Yang et al. 2020)	(Luo, Julien, et al. 2018) (Teimoori et al. 2019)
2.4 External insults (drugs, radiation, trauma, infection)
Chagas’ Disease		(Dhiman et al. 2013) (Wen and Garg 2018)				(Lassen et al. 2018)
Drug-induced Liver Injury			(Bae et al. 2012)			(Lucena et al. 2010)
Drug-induced Toxicity – Acetaminophen	(Ansher et al. 1983) (Warnet et al. 1989b) (Warnet et al. 1989a)	(Fujimoto et al. 2009) (Yoshikawa et al. 2009) (Ramachandran et al. 2011)		(Du et al. 2017) (Du et al. 2019) (Kim, Lee, et al. 2019)
Drug-induced Toxicity – Alcohol. Liver Cirrhosis		(Wheeler, Nakagami, et al. 2001) (Mansouri et al. 2010)	(Bai and Cederbaum 2001) (Mari et al. 2002)	(Alfonso-Loeches et al. 2014) (Mohan et al. 2015) (Kumar et al. 2019)	(Chacko et al. 2011) (Hoyt et al. 2017) (Hao et al. 2018) (Turkseven et al. 2020)	(Nahon et al. 2005) (Sun et al. 2009) (Srivastava et al. 2010) (Nakagawa et al. 2013) (Gitik et al. 2016) (Roy et al. 2016) But (Stewart et al. 2002)
Drug-induced Toxicity – Aminoglycosides (see also Hearing Loss)				(Esterberg et al. 2016) (Alassaf et al. 2019)	(Plotnikov et al. 2011) (Jankauskas et al. 2012) (Ojano-Dirain and Antonelli 2012) (Ojano-Dirain et al. 2014) (Dirain et al. 2018; Fujimoto and Yamasoba 2019)
Drug-induced Toxicity – Anesthesia-induced Cognitive Impairment					(Wu et al. 2017) (Zhao et al. 2019)
Drug-induced Toxicity – Anthracyclines		(Yen et al. 1996) (Chaiswing et al. 2005) (Cho et al. 2008)	(Gilliam et al. 2016)	(Rocha et al. 2016) (Fang X et al. 2019)	(Chandran et al. 2009) (Smith and Murphy 2010) (Gilliam et al. 2012) (Bayeva et al. 2013) (Szeto and Birk 2014) (Min et al. 2015) (Guigni et al. 2018) (Montalvo et al. 2020)	(Visscher et al. 2015) But (Rajic et al. 2009)
Drug-induced Toxicity – Antipsychotics, Tardive Dyskinesia (see also Schizophrenia)					(Phillips et al. 2017)	(Hori et al. 2000) (Galecki et al. 2006) (Hitzeroth et al. 2007) (Bakker et al. 2008) (Al Hadithy et al. 2010) But (Hori et al. 2000) (Zhang et al. 2002) (Pae et al. 2007) (Zai et al. 2010) (Wang et al. 2015)
Drug-induced Toxicity – AZT and other NRTIs		(Jiang et al. 2007) (Kohler et al. 2009) (Glover et al. 2014)	(Jiang et al. 2007) (Kohler et al. 2009) (Nacarelli et al. 2016)	(Liu et al. 2014)	(Nacarelli et al. 2016)
Drug-induced Toxicity – CCl4 (see also Nonalcoholic Steatohepatitis)	(Ansher et al. 1983) (Ukai, Shiraishi, et al. 1988) (Yang et al. 2017)			(Jang et al. 2015)	(Rehman et al. 2016) (Vilaseca et al. 2017)
Drug-induced Toxicity – Diclofenac and other NSAIDs				(Mazumder et al. 2016)
Drug-induced Toxicity – Fluoroquinolone Antibiotics	(Pouzaud et al. 2004)				(Lowes et al. 2009)
Drug-induced Toxicity – Herbicide Toxicity					(Reddy et al. 2011) (Szeto and Birk 2014)	(Fong et al. 2007)
Drug-induced Toxicity – Iodinated Contrast Agents				(Lin et al. 2019)	(Duan et al. 2013) (Szeto and Birk 2014)
Drug-induced Toxicity – Lead					(Maiti et al. 2017)
Drug-induced Toxicity – Methamphetamine		(Maragos et al. 2000)		(Li et al. 2012) (Xu et al. 2018)		(Nakamura et al. 2006) (Lai et al. 2019)
Drug-induced Toxicity – Multiple Chemical Sensitivity						(Cui et al. 2013)
Drug-induced toxicity – Organophosphates					(Wani et al. 2011)
Drug-induced Toxicity – Platinum Anti-cancer Drugs (see also Hearing Loss; Radiation Injury)		(Davis et al. 2001) (Chen, Cheng, et al. 2015)	(Marullo et al. 2013) But (Davis et al. 2001)	(Marullo et al. 2013) (Kong et al. 2018) (Kong et al. 2019) (Mapuskar et al. 2019)	(Mukhopadhyay, Horvath, Zsengeller, Zielonka, et al. 2012) (Toyama et al. 2014) (Tate et al. 2017) (Toyama et al. 2018; Fujimoto and Yamasoba 2019)	(Brown et al. 2015) But (Lui et al. 2018)
Drug-induced toxicity – Smoking					(Sukjamnong et al. 2018) (Chen et al. 2019) (Li, Chan, et al. 2019)
Drug-induced toxicity – Statins				(Gouni-Berthold et al. 2013) (Bouitbir et al. 2016) (Lotteau et al. 2019)
Hemorrhagic Shock				(Fink et al. 2007) (Macias et al. 2007)	(Powell et al. 2015)
Hepatitis c Infection (see also Nonalcoholic Steatohepatitis)					(Gane et al. 2010)	(Houldsworth et al. 2014)
Hyperthermia		(Li and Oberley 1997) (Venkataraman et al. 2004)		(Wang et al. 2013) (Wang, Shen, et al. 2019)
Leprosy Susceptibility						(Ramos et al. 2016)
Lung – Ventilation-induced injury (see also Acute Respiratory Distress Syndrome)				(Song et al. 2016)	(Hudson et al. 2015) (Smuder et al. 2018)
Ocular Diseases – Conjunctivitis					(Wei et al. 2019)
Radiation Injury (see also Cancer; Drug-induced Toxicity – Platinum Anti-cancer Drugs)		(Wong 1995) (Epperly et al. 1999) (Epperly et al. 2000) (Epperly et al. 2002) (Kanai et al. 2002) (Takada et al. 2002) (Epperly et al. 2003) (Guo et al. 2003) (Dolgachev et al. 2005) (Kalen et al. 2006) (Niu et al. 2008) (Yan S et al. 2008) (Veldwijk et al. 2009) (Niu et al. 2010) (Hosoki et al. 2012) (Khan et al. 2018)	(Epperly et al. 2009) (Liao et al. 2013) (Parihar et al. 2015) (Chmielewski et al. 2016) (Baek et al. 2017)	(Jiang et al. 2009) (Shimura et al. 2016) (Ku et al. 2018) (Park et al. 2018)	(Cline et al. 2018) (Gan et al. 2018) (Ibrahim et al. 2019)	(Ambrosone et al. 2005) (Martin et al. 2006) (Burri et al. 2008) (Ho-Pun-Cheung et al. 2011) But (Green et al. 2002) (Ahn et al. 2006)
Subarachnoid Hemorrhage					(Zhang, Xu, et al. 2019)
Trauma – Severe Burn Trauma					(Lee, Kaneki, et al. 2011) (Righi et al. 2013) (Wu et al. 2020)
Trauma – Traumatic Brain Injury. Mechanical Trauma		(Sullivan et al. 1999) (Lewen et al. 2001) (Xiong et al. 2005) (Shi et al. 2018)	(Hu, Dang, et al. 2018)	(Szarka et al. 2018)	(Isaev et al. 2012) (Constantinou et al. 2016) (Isaev et al. 2016) (Zhou et al. 2018) (Genrikhs et al. 2019) (Czigler et al. 2019) (Chen et al. 2020)
2.5 Genetic diseases
Angelman Syndrome (see also Table 1 – Synaptic Pruning)					(Santini et al. 2015)
Ataxia-Telangiectasia			(D’Souza et al. 2013)		(Mercer et al. 2012)
Barth Syndrome			But (Johnson et al. 2018)	(He et al. 2014) (Wang, McCain, et al. 2014) (Chowdhury et al. 2018)
Friedrich’s Ataxia		(Melov et al. 1999)			(Jauslin et al. 2003) (Zhao et al. 2017)
Hereditary Hemochromatosis						(Valenti et al. 2004)
Huntington’s Disease		(Andreassen et al. 2001) But (Underwood et al. 2010)		(Xun et al. 2012)	(Yin et al. 2016) (Pinho et al. 2020)
Leber’s Hereditary Optic Neuropathy (LHON)		(Qi et al. 2003) (Qi et al. 2004) (Qi et al. 2007)				But (Ishikawa et al. 2005)
Leigh’s Syndrome		(Qi et al. 2004) But (Brunelle et al. 2005)
Mitochondrial Disease		(Melov et al. 1999)	(Dai et al. 2010)		(Shabalina et al. 2017) (Karaa et al. 2018) But (Stealth Biotherapeutics 2019)
Propionic Acidemia					(Rivera-Barahona et al. 2017)
Spinocerebellar Ataxia Type 1					(Stucki et al. 2016)
Wilson Disease						(Gromadzka et al. 2015)
X-linked Adrenoleukodystrophy (XALD) and Adrenomyeloneuropathy (X-AMN)						(Brose et al. 2012)
2.6 Inflammatory and autoimmune diseases
Acute Respiratory Distress Syndrome (see also Lung – Ventilation-Induced Injury)			(Li et al. 2007) (Li, Kaifu, et al. 2009)
Asthma (see also Inflammation)				(Jaffer et al. 2015)		(Despotovic et al. 2015) But (Gurel et al. 2004) (Kinnula et al. 2004) (Holla et al. 2006) (Yang et al. 2011)
Chronic Obstructive Pulmonary Disease (see also Hypertension; Pulmonary Fibrosis; Inflammation)				(Li, Xu, et al. 2018)	(Wiegman et al. 2015) (Chen et al. 2019)	(Pietras, Szemraj, et al. 2010)
Emphysema: see Chronic Obstructive Pulmonary Disease; Pulmonary Fibrosis
Inflammation, Macrophage Activation, Alveolar Epithelial Cell Death, Emphysema (see also Table 1 – Immune Response)	(Soberanes et al. 2019)	(Li, Wang, et al. 2016) (Song et al. 2020)	(Li et al. 2007) (Treuting et al. 2008) (Li, Kaifu, et al. 2009)	(Heid et al. 2013) (Varga et al. 2013) (Alfonso-Loeches et al. 2014) (Park et al. 2015) (Powell et al. 2015) (Shahzad et al. 2015) (Traba et al. 2015) (Ding et al. 2016) (Li, Wang, et al. 2016) (Nadatani et al. 2016) (Wen et al. 2016) (Ding et al. 2017) (Lawana et al. 2017) (Li, Xu, et al. 2018) (Piippo et al. 2018) (Xu et al. 2018) (Kurundkar et al. 2019) (Liao et al. 2019) (Lin et al. 2019) (Soberanes et al. 2019) (To et al. 2019) (Wang, Shen, et al. 2019) (Xu et al. 2019) (Song et al. 2020) but (Bordt and Polster 2014)	(Lowes et al. 2008) (Bulua et al. 2011) (Yang et al. 2012) (Dashdorj et al. 2013) (Kelly et al. 2015) (Chelombitko et al. 2017) (Hoyt et al. 2017) (Fock et al. 2018) (Chen et al. 2019) (Wei et al. 2019) (Wu et al. 2020) But (Formentini et al. 2017)	(Montano et al. 2012)
Inflammatory Bowel Disease	(Watson et al. 2020)			(Wang, Keita, et al. 2014)	(Dashdorj et al. 2013) (Chu et al. 2017)	(Costa Pereira et al. 2017) But (Li, Song, et al. 2016)
Kidney Disease – Acute Kidney Injury (AKI), Sepsis-induced AKI (SAKI), End-stage Kidney Disease (ESRD), Chronic Renal Insufficiency (CRI) (see also Drug-induced Toxicity – Platinum-containing Anti-cancer Drugs; Diabetic Nephropathy; Sepsis)			(Hwang et al. 2019)	(Liang et al. 2010) (Patil et al. 2014) (Sims et al. 2014) (Ding et al. 2016) (Ding et al. 2017) (Zhang et al. 2017) (Liu, Wang, et al. 2018)	(Bakeeva et al. 2008) (Mizuguchi et al. 2008) (Skulachev et al. 2009) (Plotnikov et al. 2010) (Mitchell T et al. 2011) (Plotnikov et al. 2011) (Szeto et al. 2011) (Eirin et al. 2012) (Parajuli et al. 2012) (Zhao et al. 2013) (Sims et al. 2014) (Tabara et al. 2014) (Szeto and Birk 2014) (Dare, Bolton, et al. 2015) (Gu et al. 2015) (Szeto et al. 2015) (Szeto et al. 2016) (Hamed 2017) (Saad et al. 2017) (Szeto et al. 2017) (Liu, Murphy, et al. 2018)	(Crawford et al. 2011) (Chao et al. 2016) (Jerotic et al. 2019) But (McKnight et al. 2012) (Kidir et al. 2016)
Multiple Sclerosis					(Mao et al. 2013)
Ocular diseases – Uveitis						(Lan et al. 2009)
Osteoarthritis		(Koike et al. 2015)	(Collins et al. 2016)	(Farnaghi et al. 2017)	(Delco et al. 2018) (Bartell et al. 2020)	(Rego-Perez et al. 2008) (Rego et al. 2010)
Periodontitis		(Li, Wang, et al. 2016)		(Li, Wang, et al. 2016)		(Kazemi et al. 2015) (Lee et al. 2019)
Pulmonary Fibrosis (see also Chronic Obstructive Pulmonary Disease)			(Kim, Cheresh, et al. 2016)
Rheumatoid Arthritis				(Valcarcel-Ares et al. 2014)	(Andreev-Andrievskiy et al. 2016)	(Mattey et al. 2000)
Salivation (Hyposalivation, Dry Mouth, Xerostomia, Sjögren syndrome)	(Epstein et al. 1983) (Ukai et al. 1984) (Ukai, Taniguchi, et al. 1988) (Ukai et al. 1989) (Glenert 1992) (Hamada et al. 1999) (Nagano and Takeyama 2001)
Sepsis (see also Kidney (SAKI))				But (Rademann et al. 2017)	(Lowes et al. 2008) (Supinski et al. 2009) (Zang et al. 2012) (Lowes et al. 2013) (Wu et al. 2015) (Li, Wu, et al. 2016) (Prauchner 2017) (Fock et al. 2018) (Li, Ren, et al. 2019) But (Rademann et al. 2017)	(Elsakka et al. 2007) (Paludo et al. 2013) (Majolo et al. 2015)
Systemic Lupus Erythematosus					(Buskiewicz et al. 2016) (Fortner et al. 2020)
Vitiligo						(Laddha et al. 2013) (Tuna et al. 2017)
2.7 Ischemia/reperfusion injuries
Ischemia/Reperfusion Injury – Heart and artery	(Brand 2016) (Brand et al. 2016) (Boucard et al. 2019) (Detaille et al. 2019) (Diolez et al. 2019) (Piao et al. 2020)	(Chen et al. 1998) (Asimakis et al. 2002) (Loor et al. 2011)	(Zaha et al. 2016) But (Loor et al. 2011)	(Pung et al. 2012) (Escobales et al. 2014) (Jang et al. 2017)	(Zhao et al. 2004) (Adlam et al. 2005) (Cho J et al. 2007) (Neuzil et al. 2007) (Bakeeva et al. 2008) (Skulachev et al. 2009) (Smith and Murphy 2010) (Kloner et al. 2012) (Pung et al. 2012) (Birk et al. 2013) (Brown et al. 2014) (Dai et al. 2014) (Szeto and Birk 2014) (Zhang et al. 2018) But (Gibson et al. 2016)
Ischemia/Reperfusion Injury – Intestine					(Hu, Ren, et al. 2018)
Ischemia/Reperfusion Injury - Kidney: see Kidney Disease
Ischemia/Reperfusion Injury – Limb				(Miura et al. 2017)	(Cai et al. 2018)
Ischemia/Reperfusion Injury – Liver		(Wheeler, Katuna, et al. 2001)		(Han, Choi, et al. 2018)	(Mukhopadhyay, Horvath, Zsengeller, Batkai, et al. 2012) (van Golen et al. 2019)
Ischemia/Reperfusion Injury – Tissue and Cell Transplantation	(Orr et al. 2015)	(Lehmann et al. 2003) (Baldari et al. 2016)		(Lian et al. 2019)	(Thomas et al. 2007) (Mitchell T et al. 2011) (Parajuli et al. 2012) (Zhao et al. 2013) (Szeto and Birk 2014) (Dare, Logan, et al. 2015) (Szeto et al. 2015) (Hamed 2017) (Sui et al. 2018)
Ocular Diseases – Ischemia-Reperfusion		(Liu et al. 2012)			(Wei et al. 2019)
Placenta, Birth Complications – Gestational Hypoxia			(Li et al. 2008) (Wu et al. 2016)		(Phillips et al. 2017) (Aljunaidy et al. 2018) (Nuzzo et al. 2018) (Ganguly et al. 2019)	(Giusti et al. 2012) (Poggi et al. 2012)
Stroke (and Excitotoxicity)	(Huang S et al. 2020)	(Keller et al. 1998) (Murakami et al. 1998) (Fujimura et al. 1999) (Kim et al. 2002) (Maier et al. 2006) (Bullock et al. 2009) (Fukui and Zhu 2010) (Xu L et al. 2010) (Mehta et al. 2011)		(Li, Sun, et al. 2018)	(Cho S et al. 2007) (Bakeeva et al. 2008) (Skulachev et al. 2009) (Plotnikov et al. 2010) (Szeto and Birk 2014) (Kim et al. 2015) (Silachev et al. 2015) (Isaev et al. 2016) (Kim, Stepanova, et al. 2018) (Silachev et al. 2018)	(Pascotini et al. 2018)
2.8 Metabolic diseases (diabetes, obesity and their complications)
Diabetes – Pancreatic β-cell Dysfunction (Type 1 Diabetes, T1D)	(Fouqueray et al. 2011) (Pacini et al. 2015) (Perry et al. 2016) (Diolez et al. 2019)	(Kang et al. 2014)	(Gurgul et al. 2004) (Wolf et al. 2010) (Lortz et al. 2013)	(Lim et al. 2011)	(Lim et al. 2011) (Escribano-Lopez et al. 2019)	But (Pociot et al. 1993)
Diabetes – Hyperglycemia and Insulin Resistance (Type 2 Diabetes, T2D) (see also Metabolic Syndrome, Obesity)	(Fouqueray et al. 2011) (Gero et al. 2013) (Vial et al. 2015) (Detaille et al. 2016) (Fontaine et al. 2016) (Gero et al. 2016) (Diolez et al. 2019) But (Perry et al. 2016)	(Du XL et al. 2000) (Nishikawa et al. 2000) (Du XL et al. 2001) (Hoehn et al. 2009) (Boden et al. 2012) (Kang et al. 2012) (Liu et al. 2013) But (Kang et al. 2014) (Lark et al. 2015) (Han et al. 2016)	(Houstis et al. 2006) (Chen et al. 2008) (Anderson et al. 2009) (Lee et al. 2010) (Huh et al. 2012) (Kang et al. 2012) (Lark et al. 2015) (Paglialunga et al. 2015) (Lee et al. 2017) (Olson et al. 2017)	(Kizhakekuttu et al. 2012) (Ni et al. 2016)	(Anderson et al. 2009) (Feillet-Coudray et al. 2014) (Fink BD et al. 2014) (Fouret et al. 2015) (Coudray et al. 2016) (Escribano-Lopez et al. 2018) (Li, Chan, et al. 2019) (Marin-Royo et al. 2019) But (Paglialunga et al. 2012) (Szeto et al. 2016) (Fink BD et al. 2017) (Ding et al. 2019) (Bond et al. 2019) (Reynolds et al. 2019)	(Nomiyama et al. 2003) (el-Masry et al. 2005) (Houstis et al. 2006) (Chen, Yu, et al. 2012) (Ramprasath et al. 2012) (Vats et al. 2015) (Zahreddine et al. 2016) (Yahya et al. 2019)
Diabetic Cardiomyopathy (see also Cardiomyopathy)		(Shen et al. 2006)	(Arkat et al. 2016) (Luptak et al. 2019)	(Luo et al. 2013) (Ni et al. 2016)	(Escribano-Lopez et al. 2016) (Escribano-Lopez et al. 2018)	(Jones et al. 2010) (Ramprasath et al. 2012)
Diabetic Embryopathy		(Zhong et al. 2016)
Diabetic Nephropathy (see also Kidney Disease)		(Craven et al. 2001)	(Hwang et al. 2019)	(Shahzad et al. 2015) (Chen et al. 2018)	(Chacko et al. 2010) (Hou et al. 2016) (Ward et al. 2017) (Xiao et al. 2017) (Han, Xu, et al. 2018) (Ducasa et al. 2019) (Wang, Tang, et al. 2019)	(Chistyakov et al. 2001) (Nomiyama et al. 2003) (Mollsten et al. 2007) (Mollsten et al. 2009) (Tian et al. 2011) (McKnight et al. 2012) (Ascencio-Montiel Ide et al. 2013) (Mohammedi et al. 2014) (Houldsworth et al. 2015) But (el-Masry et al. 2005) (Hovnik et al. 2009) (Abbasi et al. 2018) (Yahya et al. 2019)
Diabetic Neuropathy: see Neuropathy
Diabetic Peripheral Arterial Disease (PAD)			(Ryan et al. 2016)			(Strauss et al. 2018)
Diabetic Retinopathy: see Ocular Disease - Retina
Metabolic Syndrome (see also Diabetes; Obesity)					(Mao, Kraus, et al. 2010) (Mercer et al. 2012) (Li, Chan, et al. 2019) (Marin-Royo et al. 2019)
Muscle Weakness(/Strength), Exercise Intolerance		(Kinugawa et al. 2005) (Bruton et al. 2008) (Lustgarten et al. 2009) (Kuwahara et al. 2010) But (McClung et al. 2010)	(Li, Lai, et al. 2009) (Lee et al. 2014) (Umanskaya et al. 2014) (Zhang, Rao, et al. 2016)	(Elnakish et al. 2015) (Laitano et al. 2016)	(Siegel et al. 2013) (Campbell et al. 2019)	(Ben-Zaken et al. 2013) (Ahmetov et al. 2014) (Jowko, Gromisz, et al. 2017) (Jowko, Gierczuk, et al. 2017)
Nonalcoholic Fatty Liver Disease (NAFLD) and Steatohepatitis (NASH) (see also Drug-induced Toxicity – CCl4; Hepatitis c Infection)	(Wong et al. 2020)	(Li, Huang, et al. 1995) (Wong et al. 2020) But (von Montfort et al. 2012)			(Mao, Kraus, et al. 2010) (Rehman et al. 2016) (Fock et al. 2018)	(Namikawa et al. 2004) (Al-Serri et al. 2012)
Obesity (see also Diabetes; Metabolic Syndrome)	(Diolez et al. 2019)		(Huh et al. 2012) (Townsend et al. 2020)		(Mao, Kraus, et al. 2010) (Marin-Royo et al. 2019)	(Montano et al. 2009) (Hiroi et al. 2011) (Becer and Cirakoglu 2015)
Ocular Diseases – Diabetic Retinopathy (see also Ocular Diseases – Retina)		(Kowluru et al. 2006) (Kanwar et al. 2007) (Goto et al. 2008)			(Huang et al. 2013) (Alam et al. 2015)	(Kimura et al. 2000) (Petrovic et al. 2008) (Hovnik et al. 2009) (Kangas-Kontio et al. 2009) (Tian et al. 2011) (Haghighi et al. 2014) (Haghighi et al. 2015) (Huang et al. 2017) (Naser et al. 2019) But (Vanita 2014)
2.9 Neurological diseases (cognitive, motor, degenerative, vision, hearing, neuropathies)
Alzheimer’s Disease		(Li et al. 2004) (Esposito et al. 2006) (Dumont et al. 2009) (Massaad et al. 2009) (Massaad et al. 2010) (Ma et al. 2011)	(Chen, Yoo, et al. 2012) (Mao et al. 2012) (Chen et al. 2014)	(Carvalho et al. 2014) (Hu and Li 2016)	(Manczak et al. 2010) (Calkins et al. 2011) (Ma et al. 2011) (McManus et al. 2011) (Szeto and Birk 2014) (Isaev et al. 2016) (Jia et al. 2016) (Stefanova et al. 2016) (Reddy et al. 2017) (Reddy et al. 2018) (Young and Franklin 2019)	(Wiener et al. 2007) (Gamarra et al. 2015) (de Mendonca et al. 2016) But (Ventriglia et al. 2005) (Spisak et al. 2014)
Amyotrophic Lateral Sclerosis		(Andreassen et al. 2000)	(Muller et al. 2008) (Pehar et al. 2014)		(Petri et al. 2006) (Miquel et al. 2014) (Szeto and Birk 2014)	But (Tomkins et al. 2001)
Anxiety and Depression		But (Dumont et al. 2009)	(Olsen et al. 2013) (Vicente-Gutierrez et al. 2019)	(Wu et al. 2019)	(Nussbaumer et al. 2016)	(Galecki et al. 2010) (Pietras, Witusik, et al. 2010) (Wigner et al. 2018) (da Cruz Jung et al. 2020) But (Pae et al. 2006) (Cumurcu et al. 2013)
Dementia: see Aging – Cognitive Decline; Alzheimer’s Disease
Epilepsy		(Liang et al. 2000) (Liang and Patel 2004) (Liang et al. 2012)			But (Goodman 2016)	(Kegler et al. 2019) But (Esih et al. 2017)
Hearing Loss – AHL and NIHL (see also Drug-induced Toxicity – Aminoglycosides; Drug-induced Toxicity – Platinum Anti-cancer Drugs)		(Tuerdi et al. 2017) But not AHL (Kinoshita et al. 2013)	(Someya et al. 2009) (Chen et al. 2013)	(Esterberg et al. 2016) (Alassaf et al. 2019)	(Fujimoto and Yamasoba 2019) (Kim, Baek, et al. 2019)	(Fortunato et al. 2004) (Li, Li, et al. 2009) (Liu et al. 2010) (Iqbal et al. 2013) (Nolan et al. 2013) (Brown et al. 2015) (Wang, Li, et al. 2017)
Migraine						(Palmirotta et al. 2015) (Saygi et al. 2015)
Neuropathy		(Vincent et al. 2007)	(Hattori et al. 2003)	(Zhan et al. 2018)	(Toyama et al. 2014) (Escribano-Lopez et al. 2016) (Toyama et al. 2018)	(Zotova et al. 2003) (el-Masry et al. 2005) (Isikli et al. 2018)
Obsessive-compulsive Disorder						(Orhan et al. 2012)
Ocular Diseases – General	(Diolez et al. 2019)	(Sandbach et al. 2001) (Reddy et al. 2004) (Justilien et al. 2007) (Kasahara et al. 2005) (Kanwar et al. 2007) (Goto et al. 2008) (Usui et al. 2009) (Liu et al. 2012) (Brown et al. 2017) But (Van Remmen et al. 2003) (Wolf N et al. 2005)	(Wolf N et al. 2005) (Usui et al. 2009)		(King et al. 2004) (Neroev et al. 2008) (Skulachev et al. 2009) (Huang et al. 2013) (Skulachev 2013) (Szeto and Birk 2014) (Alam et al. 2015) (Brzheskiy et al. 2015) (Cai et al. 2015) (Iomdina et al. 2015) (Babizhayev 2016) (Petrov et al. 2016) (Wei et al. 2019) But (Vlachantoni et al. 2011)	(Kimura et al. 2000) (Hovnik et al. 2009) (Kangas-Kontio et al. 2009) (Lan et al. 2009) (Tian et al. 2011) (Haghighi et al. 2014) (Haghighi et al. 2015) (Zhou Y et al. 2015) (Huang L et al. 2017) (Zhang, Zhang, et al. 2017) (Naser et al. 2019) But (Celojevic et al. 2013) (Celojevic et al. 2014) (Vanita 2014)
Ocular Diseases – Glaucoma		(Jiang et al. 2016)			(Neroev et al. 2008) (Skulachev et al. 2009) (Skulachev 2013) (Iomdina et al. 2015) (Babizhayev 2016) (Rao et al. 2019)	(Zhou Y et al. 2015) (Zhang, Zhang, et al. 2017) But (Celojevic et al. 2014)
Ocular Diseases – Retina – Macular Degeneration, Retinal Degeneration, Retinitis Pigmentosa (see also Ocular Diseases – Diabetic Retinopathy)	(Wang, Liu, et al. 2009) (Diolez et al. 2019)	(Sandbach et al. 2001) (Kasahara et al. 2005) (Justilien et al. 2007) (Usui et al. 2009) (Lee, Usui, et al. 2011) (Brown et al. 2017)	(Usui et al. 2009) (Lee, Usui, et al. 2011)	(Liao et al. 2019)	(King et al. 2004) (Neroev et al. 2008) (Skulachev et al. 2009) (Skulachev 2013) (Brzheskiy et al. 2015) (Petrov et al. 2016) But (Vlachantoni et al. 2011)	(Kimura et al. 2000) (Kowalski et al. 2010) But (Kondo et al. 2009)
Opioid Addiction						(Boroumand et al. 2018)
Parkinson’s Disease	(Diolez et al. 2019)	(Klivenyi et al. 1998) (Andreassen et al. 2001) (Callio et al. 2005) (Filograna et al. 2016) (Biosa et al. 2018)	(De Simoni et al. 2008) (Perier et al. 2010)	(Kim, Kim, et al. 2016) (Burbulla et al. 2017) (Lawana et al. 2017)	(Yang et al. 2009) (Ghosh et al. 2010) (Solesio et al. 2013) (Jin et al. 2014) (Xi et al. 2018) But (Snow et al. 2010)	(Shimoda-Matsubayashi et al. 1996) (Fong et al. 2007) (Wang et al. 2010) (Liu, Fang, et al. 2019) But (Farin et al. 2001) (Yu et al. 2014)
Schizophrenia (see also Drug-induced Toxicity – Antipsychotics; Placenta)					(Phillips et al. 2017)	(Akyol et al. 2005) (Galecki et al. 2006) (Zhang et al. 2014) (Wang et al. 2015) (Zhang, Rao, et al. 2016) But (Hori et al. 2000) (Zhang et al. 2002) (Ventriglia et al. 2006) (Hitzeroth et al. 2007) (Pae et al. 2007) (Boskovic et al. 2013) (Wang et al. 2015)

Much of the evidence for involvement of mitochondrial superoxide and/or hydrogen peroxide in a particular signaling or pathological phenotype is circumstantial, making the conclusions tentative. First, most of the studies cited in the reviews listed above are associative rather than mechanistic; they show that some signal attributed to mitochondrial ROS (such as the expression of an antioxidant protein, or the response of a fluorescent probe) is increased when the physiological signaling pathway or pathology is activated, or decreased when it is inactive. Such associations are suggestive, but they do not distinguish between causal and bystander or downstream roles of mitochondrial ROS. Second, when these studies probe mechanism by manipulating mitochondrial ROS production they often use mitochondrial poisons or genetic manipulations that alter not only mitochondrial superoxide and hydrogen peroxide production but also cellular energetics and metabolite concentrations. This makes such studies theoretically unreliable, because it is hard, if not impossible, to unscramble effects of mitochondrial superoxide and hydrogen peroxide from the simultaneous effects of altered energetics and metabolism caused by the manipulation. Third, many of the cited studies use cellular probes for reactive oxygen species that suffer from many serious drawbacks, such as lack of specificity for particular ROS (or even for ROS over other redox-active reactants), interference with the redox status of the target cell (for example by acting as pro-oxidants), and inadequate controls for effects of changes in plasma membrane potential and mitochondrial membrane potential on the accumulation of the mitochondria-targeted probes, etc. (Wardman 2007; Murphy et al. 2011; Kalyanaraman et al. 2012, 2017; Xiao and Meierhofer 2019), making them experimentally unreliable.

Strength of different types of evidence that mitochondria-derived superoxide or hydrogen peroxide drives biological effects

Given the difficulties with association studies, what types of evidence can give robust support to the hypothesis that mitochondrial production of superoxide or hydrogen peroxide is an important driver of a particular physiological signaling pathway or a particular pathology, or, conversely, can provide strong evidence against that hypothesis? Ideally, supportive evidence should show that production is both necessary (i.e. preventing a rise in superoxide or hydrogen peroxide concentration in the mitochondrial matrix prevents the engagement of the signaling pathway or the induction of the pathology in response to its normal initiators) and sufficient (i.e. artificially triggering a rise in the level of matrix superoxide or hydrogen peroxide triggers the signal generation or pathology). In general, pharmacological or genetic interventions that decrease or increase concentrations of superoxide or hydrogen peroxide specifically in the mitochondrial matrix without affecting other reactions such as oxidative phosphorylation are required. Each of these interventions has advantages and drawbacks.

The advantage of pharmacological interventions is that they can be applied to essentially any experimental system without the need for complex genetic preparation, and they work acutely (within seconds to hours), without necessarily giving the system time to make compensatory changes in gene expression. The disadvantage is that they require a well-characterized pharmacologically-active molecule with specificity for the target and with suitable properties such as solubility, chemical and biological stability, and membrane permeability. The chief potential drawback of pharmaceutical approaches is always non-specificity, in this case the possibilities of residual inhibition of the electron transport chain or completely off-target effects at high concentrations always have to be borne in mind. Although short-term responses to well-characterized pharmaceuticals can be interpreted with little ambiguity, long-term treatment with a pharmacological reagent allows time for compensatory signaling and gene expression changes (such as down-regulation of endogenous antioxidant pathways) that can complicate interpretations.

The advantage of genetic manipulations is specificity; with appropriate design, a particular protein can be overexpressed or knocked down in a specified tissue and subcellular compartment, at a time of the experimenter’s choosing. The chief drawback of genetic approaches is always the poor time resolution. Relatively long times (several hours, days or weeks) are needed to allow overexpression, or turnover of existing protein in knock-down experiments, which allows a myriad of compensatory changes in the expression of other genes, or changes in development, long-term damage etc. These consequent changes always generate ambiguity in the interpretation of phenotypes caused by making changes in the expression of a gene of interest. In addition, overexpressed proteins may aggregate, or bind and alter partner molecules, causing unexpected secondary effects, and the target protein might have other quite separate unforeseen functions, complicating the analyses.

Figure 3 summarizes the different manipulations that can provide robust support, ranked according to how proximal they are to the primary variables of interest: matrix superoxide and hydrogen peroxide concentrations. These manipulations are used below as criteria to judge the strength of published evidence in supporting the hypothesis that mitochondrial production of superoxide and/or hydrogen peroxide drives physiological signaling pathways (Table 1) or pathologies (Table 2).

Figure 3. Criteria for assessing whether mitochondrial production of superoxide and/or hydrogen peroxide drives a biological effect, ranked by reliability. For details and justifications see the text.

• 1. The inhibition of a phenotype by addition of well-characterized S1QELs or S3QELs is the best evidence that the phenotype is caused by mitochondrially-generated superoxide or hydrogen peroxide. The general problem with directly manipulating the production of superoxide (or hydrogen peroxide) (reaction 3 in Figure 1) is that such manipulation using conventional reagents modifies the flow of electrons in the respiratory chain, causing changes in energetics that can dominate the observed effects. For this reason effects of mitochondrial respiratory poisons such as rotenone, myxothiazol, or antimycin, effects of uncouplers of oxidative phosphorylation, or effects of genetic modulations to the activity of respiratory complexes, such as mutations in complex I or the Rieske FeS center of complex III, are extremely unreliable guides to the biological roles of mitochondrial superoxide and hydrogen peroxide. This problem has been overcome by the discovery, using chemical screens, of small molecules that suppress electron leak to oxygen without inhibiting electron transport in the respiratory chain. S1QELs (Brand et al. 2016) suppress site I_Q electron leak, and S3QELs (Orr et al. 2015) suppress site III_Qo electron leak (Orr et al. 2013, 2015; Brand 2016; Brand et al. 2016; Wong et al. 2017). Specific suppressors of the other sites of superoxide and hydrogen peroxide production have not yet been published. By directly decreasing formation of superoxide (and perhaps hydrogen peroxide), S1QELs and S3QELs decrease matrix levels of both superoxide and its product, hydrogen peroxide. Unlike conventional antioxidants they have the advantages of site-selectivity and of potentially fully-preventing superoxide production from a particular site rather than mopping up after superoxide or hydrogen peroxide have had effects (Watson et al. 2019). The best of the published S1QELs and S3QELs have high potencies (low IC₅₀s) in mitochondrial assays: S1QEL1.1, 70 nM (Brand et al. 2016); S3QEL3, 350 nM (Orr et al. 2015), and cause no inhibition of cell growth or the electron transport chain at 20 times higher concentrations. Other S1QELs with lower potencies have also been discovered: anethole trithione (5-[p-methoxyphenyl]-1,2-dithiole-3-thione, Sulfarlem^®, OP2113), IC₅₀ = 10–26 µM (Boucard et al. 2019; Detaille et al. 2019), and Imeglimin, IC₅₀ = ∼100 µM (Detaille et al. 2016). However, it is unclear whether the S1QEL activities of anethole trithione (a known slow-release H₂S donor (Wallace et al. 2018)) or Imeglimin are the only cause of their biological effects.

• 2. Excellent evidence for involvement of mitochondrially-generated superoxide in a phenotype comes from manipulating its removal (reaction 4 in Figure 1). As outlined above, SOD2 is the only superoxide dismutase normally present in the matrix (Weisiger and Fridovich 1973) and it provides the only significant catalyzed pathway for removal of matrix superoxide; the other pathway that can reach high rates is spontaneous dismutation. Overexpression of SOD2 will lower the steady-state superoxide concentration in the matrix, usually without affecting matrix hydrogen peroxide production or concentration, and will decrease effects caused by matrix superoxide. One caveat to SOD2 overexpression studies is that there needs to be an adequate supply of Mn and lack of excessive Fe, otherwise the overexpressed apoenzyme may incorporate Fe instead of Mn, making it a peroxidase with damaging pro-oxidant effects (Ganini et al. 2018). Conversely, partial or complete SOD2 knockout will raise matrix superoxide concentration, amplifying its effects. Excellent evidence for a significant physiological or pathological role of matrix superoxide comes from prevention of a phenotype by SOD2 overexpression, or exacerbation by SOD2 knockdown. Constitutive knockout of SOD2 is neonatally lethal in mice (Li, Huang, et al. 1995), illustrating that excessively raised matrix superoxide concentrations can exert dramatic pathological effects.

• 3. Excellent evidence for involvement of matrix-derived hydrogen peroxide in a phenotype comes from manipulating its removal (reaction 5 in Figure 1). As outlined above, PRDX3 may normally remove 90% of matrix hydrogen peroxide (Cox et al. 2009). Overexpression of PRDX3 will lower the steady-state hydrogen peroxide concentration in the matrix, without affecting matrix superoxide production or concentration, and will decrease effects caused by matrix hydrogen peroxide. Conversely, partial or complete PRDX3 knockout will raise matrix hydrogen peroxide concentration, amplifying its effects. Other systems also remove matrix hydrogen peroxide, in particular diffusion to the cytosol and to other organelles (reaction 6 in Figure 1), where it can be degraded by catalases and peroxidases. GPX1 is also active in the matrix, but the majority (about 90%) is located in the cytosol and it also catalyzes the reduction of various other hydroperoxides (Asayama et al. 1994; Esworthy et al. 1997; Ho et al. 1997), making it a poor candidate for driving specific alterations in matrix hydrogen peroxide concentration. Strong evidence for a significant physiological or pathological role of matrix hydrogen peroxide comes from prevention of a phenotype by PRDX3 overexpression, or exacerbation by PRDX3 knockdown. However, PRDX3 is not fully specific for hydrogen peroxide, but also removes other hydroperoxides (Peskin et al. 2010), making interpretations less secure than those for superoxide based on manipulations of SOD2. Some of the best evidence for a significant role for mitochondrial hydrogen peroxide in driving different phenotypes comes from experiments using genetically engineered expression of catalase in the mitochondrial matrix (mCAT) (Dai et al. 2014, 2017). Expression of mCAT will decrease hydrogen peroxide but not superoxide concentrations in the matrix, and has been used extensively to provide high-quality evidence for involvement of matrix-derived hydrogen peroxide in numerous pathologies (see Table 2).

• 4. Good evidence for the involvement of mitochondrially-generated superoxide in a phenotype comes from protective effects of mitochondria-targeted SOD mimetics (and exacerbation by mitochondria-targeted superoxide-generators). mitoTEMPO and mitoTEMPOL are piperidine nitroxides attached to a lipophilic cation, enabling them to pass through lipid bilayers and accumulate several hundred-fold in mitochondria. These and related or equivalent molecules are mitochondria-targeted SOD mimetics. However they have sub-optimal chemical specificity because they are also peroxidases and lipid peroxidation chain terminators (Soule et al. 2007) and they interact with the electron transport chain at ubiquinone (Trnka et al. 2008), making them less definitive than we would wish. Nonetheless, experiments with mitoTEMPO provide some of the most extensive pharmacological evidence in the literature for the roles of matrix superoxide in physiology and pathology (see Tables 1 and 2). Pro-oxidants such as mitochondria-targeted paraquat, mitoPQ (Robb et al. 2015) should also be useful, but their specificity is unclear and they have not yet been used extensively.

• 5. Mitochondria-targeted antioxidants that work more distally from superoxide and hydrogen peroxide production can also provide useful evidence for the involvement of generic matrix oxidants in different phenotypes. MitoQ and SkQ and their derivatives are quinones attached to a lipophilic cation, enabling them to pass through lipid bilayers and accumulate several hundred-fold in mitochondria (Skulachev 2013; Murphy 2016). They probably act as primarily as lipid peroxidation chain terminators (Asin-Cayuela et al. 2004; Murphy 2016), preventing many of the downstream effects of matrix superoxide and hydrogen peroxide. Because they are redox cyclers, there is always the possibility of pro-oxidant effects, and at higher concentrations the triphenylphosphonium targeting group can inhibit electron transport. SS-31 (elamipretide, Bendavia) is a tetrapeptide that appears to act as an targeted antioxidant through its interaction with cardiolipin, which causes SS-31 to accumulate in mitochondrial membranes (Szeto 2006, 2014; Szeto and Birk 2014). Its mode of action is unclear, but it may modulate mitochondrial membrane surface charge (Mitchell et al. 2020) and decrease cytochrome c-catalyzed cardiolipin peroxidation (Birk et al. 2013). Such mitochondria-targeted antioxidants have been used extensively to show protective effects in many pathologies (Smith and Murphy 2010; Dai et al. 2014; Zielonka et al. 2017; Szeto and Liu 2018), providing a large body of relevant evidence (see Table 2).

• 6. In principle, gene-association studies are a powerful way to identify pathologies that are modified by genetic variants in relevant proteins, specifically SOD2 and PRDX3 (and to a much lesser extent because of its dual matrix and cytosolic location and lower quantitative contribution to hydrogen peroxide removal from the matrix, GPX1). Gene-association studies have the great advantage that, unlike the cell and animal studies that underlie the criteria above, they can directly link mitochondrial matrix superoxide and hydrogen peroxide levels to specific human diseases. Many such gene-association studies are cited in Table 2. However, it is essential that such studies are adequately powered to identify small effects in high background noise. This is seldom the case, and many of the meta-analyses fail to confirm the initial reports, or show only small effects. For this reason, gene-association studies are not highly rated as a criterion in Figure 3.

There are also weaker suggestive criteria that have not been used in Tables 1 and 2:

• 7. Inhibition of a phenotype by knock-in or activation by knock-down of other mitochondrial antioxidant proteins can suggest that the phenotype is driven by ROS (from unspecified sites). The NADP-specific isocitrate dehydrogenase (IDH2) is not essential for citric acid cycle flux (Williamson and Cooper 1980), so unlike other citric acid cycle enzymes it can be knocked down or out to affect matrix NADPH and glutathione redox state (Kim, Baek, et al. 2019) without necessarily causing confounding changes in citric acid cycle flux and bioenergetics. Lowering the level of reduced glutathione in the mitochondrial matrix will raise hydrogen peroxide levels and is associated with numerous pathologies (Mari et al. 2009). Similarly, alterations in the thioredoxin system by manipulating the matrix enzymes thioredoxin-2 (TRX2) (Spyrou et al. 1997) or thioredoxin reductase-2 (TRXR2) (Lee et al. 1999) will have effects. However, the effects of these manipulations on superoxide and hydrogen peroxide levels in the matrix are rather indirect. GPX1, PRDX5 and GPX4 are present in several compartments, making manipulations of their activities less definitive in the present context. In addition, PRDX5 detoxifies mostly alkyl hydroperoxides and GPX4 detoxifies mostly membrane hydroperoxides (Collin 2019; Sies and Jones 2020).

• 8. General antioxidants provide weaker circumstantial evidence. Inhibition of a phenotype by untargeted antioxidants such as N-acetylcysteine (NAC), and SOD/catalase mimetics such as the salen Mn complexes (EUKs) (Doctrow et al. 2012; Batinic-Haberle et al. 2018; Bonetta 2018), suggests that the phenotype may be driven by cellular ROS, although the exact targets and mechanisms are very unclear.

• 9. The levels of the mRNA or the enzymic activities of various matrix antioxidant enzymes, such as SOD2, PRDX3, or GPXs, are often reported, as are the intensities of various markers of ROS levels, such as fluorescent probes, protein oxidation, or engagement of downstream ROS-driven signaling pathways. Association of these signals with a phenotype is helpful in drawing attention to possibilities or in strengthening the circumstantial case for ROS involvement, but by themselves they do not distinguish causes from effects, so are excluded from Tables 1 and 2. Effects of pharmacological inhibition or genetic manipulation of respiratory chain proteins, or addition of uncouplers, are deeply confounded by potential effects on energetics and cannot be relied upon, so are also excluded.

Redox signaling

There is widespread agreement that cells use mitochondrial ROS as a redox signal in many different physiological contexts (see the Introduction). Using the criteria set out in Figure 3 as a guide, what is the evidence that superoxide and/or hydrogen peroxide generated in the mitochondrial matrix acts as a redox signal to drive different physiological responses? The relevant literature is collated in Table 1 and ranked and made easier to overview in Figure 4. Note that the division between physiological responses (Table 1) and pathologies (Table 2) is somewhat arbitrary. For example, cell proliferation and cell migration are required for normal cell growth and development, but are also features of pathologies such as cancers. Inflammation is required for responses to infection, but chronic inflammation plays a part in many pathologies.

Figure 4. Physiological processes in which mitochondrial superoxide and/or hydrogen peroxide are implicated. Ranked representation of Table 1. Entries are ranked in order of the criteria shown in Figure 3 (and then alphabetically for entries with the same score). Green cells indicate a preponderance of papers reporting an effect, orange cells (marked “.....”) indicate a preponderance of papers reporting no effect, and white cells indicate that no relevant papers were identified (see color version of this figure at www.tandfonline.com/ibmg).

Table 1 and Figure 4 show that there is considerable evidence that mitochondrial generation of superoxide and hydrogen peroxide has an important role in a range of physiological responses of cells. Many of these responses are related to cellular stress management, and overlap with each other. They include propagation of ER stress, apoptosis, autophagy, cellular senescence and wound healing, HIF1α signaling, and the adaptive and innate immune responses. A second group of physiological responses is related to the cell cycle, cell proliferation and growth, cell migration, and differentiation. There is sporadic evidence for several other responses, some of which may be related to the first two groups (hormetic defense is related to stress responses, synaptic pruning is related to apoptosis, and effects on the gut microbiome may be related to immunity).

Pathology

There is widespread agreement that mitochondria are able to generate damaging amounts of ROS in many different pathological contexts (see the Introduction). What is the strength and weight of the evidence that superoxide and/or hydrogen peroxide generated in the mitochondrial matrix drives a particular pathological response? The relevant literature (following the criteria set out in Figure 3) is diverse and comes partly from cell models, partly from animal models, and partly from clinical studies. It is collated in Table 2 and ranked and made easier to overview in Figure 5.

Figure 5. Pathologies in which mitochondrial superoxide and/or hydrogen peroxide are implicated. Ranked representation of Table 2. Entries are ranked in order of the criteria shown in Figure 3 (and then alphabetically for entries with the same score). Green cells indicate a preponderance of papers reporting an effect, pale green cells indicate mixed evidence, orange cells (marked “….”) indicate a preponderance of papers reporting no effect, and white cells indicate that no relevant papers were identified (see color version of this figure at www.tandfonline.com/ibmg).

Table 2 and Figure 5 list more than 100 pathologies that may be driven by mitochondrial superoxide or hydrogen peroxide according to the literature organized using the criteria in Figure 3. Some of the pathologies overlap and the divisions are somewhat arbitrary, but for 30-40 of these pathologies there are multiple lines of strong experimental evidence for a role of mitochondrial superoxide or hydrogen peroxide in cell and animal models, and sometimes in clinical studies. The top half of the entries in Figure 5 can be crudely grouped into a smaller number of general categories. Most of the entries in the bottom half of Figure 5 also fit into these general categories. The categories are:

• Metabolic diseases (diabetes, obesity and their complications)

• Cardiovascular diseases (hypertension, atherosclerosis, cardiomyopathy, heart failure)

• Inflammatory diseases

• Cancer (incidence, growth, metastasis)

• Neurological diseases (cognitive, motor, degenerative, vision, hearing, neuropathies)

• Ischemia/reperfusion injuries

• Aging and aging-associated diseases

• External insults (drugs, radiation, trauma, infection)

• Genetic diseases

As with signaling responses, many of these pathologies overlap and the divisions between them are sometimes arbitrary. For example, inflammation is an underlying theme in most of the other categories such as metabolic diseases, neurological diseases, and aging. The greatest risk factor for most of the categories is aging, underscoring the connections between them. In addition, the presence of one disease often precipitates or exacerbates another, e.g. diabetic cardiomyopathy. External insults often cause damage that triggers other pathologiess (radiation and drugs can cause aging and inflammation, heart disease, cognitive decline, etc.). Nonetheless, it is clear that mitochondrial superoxide and hydrogen peroxide have a strong causal connection to a wide range of interconnected pathologies.

There are pathologies for which there is literature showing effects of manipulation of SOD2, but little or no literature showing effects of PRDX3 or expression of mCAT (stroke, ischemia/reperfusion injury, some cancers, acetaminophen and methamphetamine toxicity, some eye diseases, periodontitis, etc.). This asymmetry suggests the possibility that these conditions are driven by matrix superoxide but not by matrix hydrogen peroxide (see above, Figure 2), although it may simply reflect a lack of studies reporting effects of manipulations that alter matrix hydrogen peroxide concentration. Conversely, there are pathologies for which there is literature showing effects of manipulation of PRDX3 or expression of mCAT but no literature showing effects of manipulation of SOD2 (pre-eclampsia, heart failure with preserved ejection fraction, ataxia-telangiectasia, arterial and lung diseases, reproductive aging, etc.) suggesting the possibility that these conditions are driven by matrix hydrogen peroxide but not by matrix superoxide (with the same caveat about absent literature).

Conclusions

Within the huge literature on ROS, signaling, and disease, much of the evidence is pertinent and strong, showing clearly that elevated mitochondrial matrix superoxide and/or hydrogen peroxide concentrations are both necessary and sufficient to drive a wide range of physiological responses and pathological outcomes. Necessary, at least partially, because lowering matrix superoxide or hydrogen peroxide concentration by suppressing their formation or enhancing their disposal (or their downstream effects) is protective. Sufficient, at least partially, because raising matrix superoxide or hydrogen peroxide concentrations by compromising their disposal can recapitulate the phenotype. This conclusion reflects the opening paragraph of the present review: riding the tiger of abundant energy release carries the risk of premature electron escape to oxygen to form ROS that can drive damage and disease – the tiger may bite you. Increased understanding of the roles of matrix superoxide and hydrogen peroxide in these pathologies can direct and illuminate the development of therapeutic approaches to treat them while avoiding untoward interference with the beneficial signaling pathways. There is a great opportunity for improved S1QELs, S3QELs and suppressors of electron leak to oxygen at other sites, improved mitochondria-targeted SOD and catalase mimetics, and improved modifiers of downstream effects such as lipid peroxidation and FeS and thiol modification to be developed to treat the pathologies listed in Table 2. Perhaps we can eventually cage our tiger.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Additional information

Funding

Work in my laboratory is supported by Calico Life Sciences LLC (South San Francisco, CA) and the Buck Institute for Research on Aging (Novato, CA).