|Home | About | Journals | Submit | Contact Us | Français|
The mitochondrion is a major source of reactive oxygen species (ROS). Superoxide (O2•−) is generated under specific bioenergetic conditions at several sites within the electron-transport system; most is converted to H2O2 inside and outside the mitochondrial matrix by superoxide dismutases. H2O2 is a major chemical messenger that, in low amounts and with its products, physiologically modulates cell function. The redox state and ROS scavengers largely control the emission (generation scavenging) of O2•−. Cell ischemia, hypoxia, or toxins can result in excess O2•− production when the redox state is altered and the ROS scavenger systems are overwhelmed. Too much H2O2 can combine with Fe2+ complexes to form reactive ferryl species (e.g., Fe(IV)=O•). In the presence of nitric oxide (NO•), O2•− forms the reactant peroxynitrite (ONOO−), and ONOOH-induced nitrosylation of proteins, DNA, and lipids can modify their structure and function. An initial increase in ROS can cause an even greater increase in ROS and allow excess mitochondrial Ca2+ entry, both of which are factors that induce cell apoptosis and necrosis. Approaches to reduce excess O2•− emission include selectively boosting the antioxidant capacity, uncoupling of oxidative phosphorylation to reduce generation of O2•− by inducing proton leak, and reversibly inhibiting electron transport. Mitochondrial cation channels and exchangers function to maintain matrix homeostasis and likely play a role in modulating mitochondrial function, in part by regulating O2•− generation. Cell-signaling pathways induced physiologically by ROS include effects on thiol groups and disulfide linkages to modify posttranslationally protein structure to activate/inactivate specific kinase/phosphatase pathways. Hypoxia-inducible factors that stimulate a cascade of gene transcription may be mediated physiologically by ROS. Our knowledge of the role played by ROS and their scavenging systems in modulation of cell function and cell death has grown exponentially over the past few years, but we are still limited in how to apply this knowledge to develop its full therapeutic potential. Antioxid. Redox Signal. 11, 1373–1414.
The turn of the century has seen a resurgence of interest in mitochondrial generation of reactive oxygen species (ROS), first discovered and understood during the late 1960s and early 1970s. The renewed interest derives from the newer concept that ROS are not always deleterious to cell function and from the use of more sensitive and accurate techniques and probes to understand the mechanism of ROS generation and scavenging systems. A number of commentaries and well-researched reviews have been published since 2000 on the putative mechanisms of ROS production and the complex pathophysiologic effects that ROS exert on cell function; many of these are referenced throughout this review (8, 27, 31, 48, 55, 59, 66, 72, 75, 80, 81, 113, 122, 126, 132, 133, 140, 146, 163, 179, 183, 189, 202, 209, 234, 241, 243, 244, 246, 299, 303, 307, 328). Many of these reviews address very specific aspects of mitochondrial bioenergetics or deal with nonmitochondrial ROS production or both. For this review, we hope to present a comprehensive understanding of mitochondrial ROS production sites and their regulation, ROS scavenging, and physiologic as well as pathologic effects of ROS, particularly as triggers of cell protection.
This review focuses primarily on the ROS generated by the mitochondrial respiratory complexes, the sites and conditions for ROS generation, cell regulation of ROS generation and ROS scavenging, cell-damaging effects of ROS vs. physiologic regulation of mitochondrial and cell function by ROS, and the emerging role of pharmacologic approaches to manipulate generation and scavenging of specific ROS. Other potential sources of ROS exist in mitochondria (e.g., glycerol-3-phosphate dehydrogenase, acyl-CoA dehydrogenase), but their physiological significance remains to be elucidated (8). This report concentrates on studies of mitochondria in excitable cells with high metabolic rates (i.e., mammalian myocardial and nervous system cells). Mitochondrial density is very high in muscle and nerve cells because of their higher O2 consumption rate compared with that of most other cell types. The number, size, shape, interconnections, and location of mitochondria are also different depending on cell types (91). Mitochondria also serve different functions within the same cell [e.g., in nerve cells (91) and between the interfibrillary and subsarcolemmal mitochondria in cardiac cells (213)].
This review summarizes studies on how and why ROS are generated by mitochondrial electron-transport complexes in physiologically low amounts, perhaps up to 1% or 2% of the total rate of O2 consumption (77), to modulate mitochondrial and cell function. A number of topics are addressed: How can ROS be generated when the level of the electron acceptor O2 is low? What is the source and location of “good” versus “bad” ROS? Which are the physiologically relevant ROS, and what are their targets? How are the levels of physiologic ROS modulated? How can generation or scavenging of ROS be manipulated to support cell function when the cell is stressed? Are low O2–induced cell signaling pathways initiated at the mitochondrial level, and does an increase or decrease in ROS mediate this activity?
This review does not specifically address the significant extramatrix and nonmitochondrial sources of ROS [e.g., cytochrome b5 reductase, perioxisomes, catecholamines, hydroquinones, plasma membrane oxidases such as NADPH oxidase, lipoxygenases, monoamine oxidases, xanthine/xanthine oxidase, coupled or uncoupled nitric oxide synthase, and eicosanoids pathways, among others (303)]. Most of these compounds can initially lose an electron to form the superoxide anion radical (O2•−) by autooxidation. The nonmitochondrial ROS released or formed in the cytosol are buffered generally under strong reducing conditions by intracellular thiols, particularly glutathione (GSH) and thioredoxin (TRXSH2) by the activities of their reductases (303); this topic is addressed only as applied to mitochondria-derived ROS. The free radical NO• can have profound effects on mitochondrial function (i.e., competing with O2 at mitochondrial respiratory complex IV and reacting with O2•− to form peroxynitrite). The source of the NO• may be vascular endothelium, nerve terminals, or other cytosolic sources. Controversial evidence that suggests that NO• may be generated by a mitochondrial NO synthase (23, 24, 136, 137), but this review deals primarily with respiratory complex–derived O2•− and its products.
Although much is known of the chemistry of ROS (151), much less is known about the specific molecular sites and conditions for electron leak and O2•− generation along the electron-transport pathway under pathophysiologic conditions (32, 61). Some topics addressed subsequently: How and why do singlet electrons escape during electron transfer? What are the physiologic pathways for O2•− release? How can O2•− be formed during state 3 (ADP-stimulated) respiration when the ΔΨm is less polarized because of the influx of protons? How can more O2•− be generated when O2 levels decrease? Are isolated mitochondria a proper model for assessing ROS generation? What is the relative amount of O2•− generated at sites along the respiratory complexes? Does O2•− or other ROS function as an O2 sensor to modulate enzyme activity? What distinguishes physiologic from pathologic release of O2•−? It is hoped that this review will furnish at least satisfactory answers to most of these questions.
Mitochondria are membrane-enclosed organelles (1- to 10-μm diameter) that generate most of the cell's supply of ATP; they also have a role in cell signaling and differentiation, Ca2+ buffering, apoptosis and cell death, as well as control of the cell cycle and cell growth (322). Many of these processes are triggered or mediated by Ca2+ or ROS or both. The number of mitochondria in a cell varies widely by organism and tissue type. Many cells have only a single mitochondrion, whereas others can contain several thousand mitochondria (265, 322). Mitochondrial compartments include the outer mitochondrial membrane (OMM), the inter membrane space (IMS), the inner mitochondrial membrane (IMM), and the cristae and matrix (265). The OMM has a protein-to-phospholipid ratio similar to that of the plasma membrane (~1:1 by weight) and contains large numbers of integral proteins called porins that form channels to allow molecules of more than 5,000 Daltons to diffuse freely across the OMM (265). Larger proteins also can enter the mitochondrion if a signaling sequence at their N-terminus binds to large multi-subunit proteins (translocases) that then actively transport them across the OMM. Disruption of the OMM permits proteins in the IMS to leak into the cytosol, leading to cell death (87). Because the OMM is freely permeable to small molecules, the concentrations of small molecules, such as ions and sugars, in the IMS is the same as in the cytosol (265). However, as large proteins must have a specific signaling sequence to be transported across the OMM, the protein composition of the IMS is different from the protein composition of the cytosol. One protein that is localized to the IMS in this way is cytochrome c (87).
The IMM contains proteins with several functions (225, 265): (a) redox reactions of oxidative phosphorylation, (b) ATP synthase/ATPase, (c) transport proteins that regulate metabolites across the matrix, and (d) protein-import machinery. The IMM has a 3:1 protein-to-phospholipid ratio and is rich in the phospholipid cardiolipin, which contains four fatty acids that help to make the IMM impermeable (322). Unlike the OMM, the IMM does not contain porins and is highly impermeable to all molecules, so that all ions and molecules require membrane transporters to enter or exit the matrix.
Mitochondria vary in number and location according to cell type. Liver cells have ~1,000–2,000 mitochondria per cell, making up 20% of the cell volume (2), and the IMM (+cristae) area is about 5 times greater than that of the OMM. In cardiac mitochondria, the cristae area is much larger, and mitochondrial volume can reach 30% of cell volume. Mitochondria often form a complex 3D branching network inside the cell with the cytoskeleton and tight connections to other organelles. The association with the cytoskeleton determines mitochondrial shape and function (265).
The membrane potential across the IMM is formed by the action of the respiratory enzymes and the TCA cycle (225, 322). Each pyruvate molecule produced by glycolysis is actively transported across the IMM and into the matrix, where it is oxidized and combined with coenzyme A to form CO2, acetyl-CoA, and NADH (322). The acetyl-CoA is the primary substrate to enter the TCA cycle. The enzymes of the TCA cycle are located in the mitochondrial matrix, with the exception of succinate dehydrogenase, which is bound to the IMM as part of complex II. The TCA cycle oxidizes the acetyl-CoA to CO2, and, in the process, produces three molecules of NADH and one molecule of FADH2, which are a source of electrons for transport along the respiratory complexes, and a molecule of GTP (that is readily converted to an ATP) (322). The redox energy from NADH and FADH2 is transferred to O2 in several steps via the electron transport along the respiratory complexes. The individual proteins of complexes I to V specifically interact to form defined supramolecular structures, the so-called “respiratory supercomplexes” or “respirasome” (114).
Reducing equivalents from the cytoplasm can be imported via the malate-aspartate shuttle system of antiporter proteins or fed into the electron-transport system (ETS) by using a glycerol phosphate shuttle. NADH dehydrogenase, cytochrome c reductase, and cytochrome c oxidase perform the transfer, and the incremental release of energy is used to pump protons (H+) into the IMS (322). As the proton concentration increases in the IMS, a strong electrochemical gradient is established across the IMM. The protons can return to the matrix through the ATP synthase complex, and their potential energy is used to synthesize ATP from ADP and inorganic phosphate (Pi) (322). This process, called chemiosmosis, was first described by Peter Mitchell (230), who was awarded the 1978 Nobel Prize in Chemistry for his work.
Molecular O2 (dioxygen) has two unpaired electrons with parallel spin in different antibonding orbitals. O2•− arises directly from the reduction of O2 by transfer of a lone electron to its antibonding orbital (151). The electrons are believed to escape from the mitochondrial ETS at discrete sites during duringthethe transport of electrons (Figs. 1 and and2A)2A) from NADH oxidoreductase (complex I) to cytochrome c oxidase (complex IV), the ultimate electron acceptor by which O2+4e− and 4H+ is reduced to 2H2O. Oxygen, necessary for life in aerobic organisms, is therefore simply a sink for electrons (66).
The standard reduction potentials of electron carriers in the four mitochondrial complexes span the range of redox potentials from NADH at complex I (−0.32V) to O2 at complex IV (+0.82V) (Fig. 2B). The standard reduction potential for the reduction of O2 to O2•− is about −0.16V (307), so many components of the respiratory complexes in the ETS are thermodynamically capable of irreversibly transferring an electron to O2. However, the single-electron transfers are tightly controlled and quickly coupled with the second electron transfer (107, 293, 294). The energy released as electrons are transferred along the respiratory chain is conserved into an H+ gradient directed outward through the IMM. Ejection of protons alkalinizes the matrix to create a transmembrane H+ electrochemical-potential gradient, the proton motive force (ΔμH+), comprising the membrane potential (ΔΨm, −150mV) and the smaller pH-gradient potential (ΔpHm, −30mV) (201). The energy stored in the ΔμH+ is partially dissipated as H+ reenters the matrix at complex V (ATP synthase), which rotates a protein rotor to convert ADP and Pi to ATP. This process is reversible.
The generation of ROS from mitochondria was first reported in the early 1970s (44, 218). ROS are O2 molecules alone or bound to elements C, H, or N in different states of oxidation or reduction (Table 1) (151). The primordial mitochondrial ROS is the free radical O2•− (Fig. 2B), which, as noted earlier, has an added unpaired electron. Although O2•− can reduce Fe3+ to Fe2+ and react with other ligands to form other radicals, its usual fate is rapid dismutation (Kd=1.6×109M/s) to hydrogen peroxide (H2O2) by superoxide dismutase (SOD) in the mitochondrial matrix (MnSOD) and IMS and cytosol (CuZnSOD). During the dismutation reaction, the O2•− is first protonated to the hydroperoxyl radical (HO2•) (Kd 10×107M/s), and two HO2• react (Kd 9×105M/s) to form H2O2 and H2O (151). Thus, the spontaneous dismutation rate of HO2• is ~100 times faster than that of O2•− and because it is uncharged, HO2•, unlike O2•−, can cross membranes (104). With a pKa of 4.8 for the HO2•/O2•− pair, very little O2•− exists as the protonated form in the mitochondrial matrix. It follows that dismutation of O2•− to H2O2 occurs faster when the matrix pH is lower (k=5×105M/s at pH 7.0 and 102M/s at pH 11) (151).
Not all ROS are actually free radicals (e.g., H2O2 and peroxynitrite, ONOO−) (Fig. 3). When a free radical reacts with a nonradical, the product is another free radical, whereas the interaction of two free radicals produces a nonradical (151). For example, ONOO− is a nonradical formed from the interaction of NO• and O2•−. ONOO− is stable at an alkaline pH and fairly nonreactive, but ONOO-- is readily protonated (cytosol more than matrix) at cellular pH to ONOOH (peroxynitrous acid), which is very cytotoxic, like HO•, and causes depletion of –SH groups and other antioxidants, oxidation of lipids, DNA strand breakage, as well as nitration of and deamination of DNA bases, especially guanine (151). H2O2 is not a free radical and is not so strong an oxidant as O2•−, and H2O2, like H2O, diffuses more easily through aquaporins than through lipid membranes (37, 38). H2O2 is normally maintained at low levels by catalase and glutathione peroxidase, among others, which convert H2O2 to H2O and O2.
H2O2 also can participate in the one-electron reaction with transition metal ions (e.g., Fe2+ bound to organic molecules) to generate various intermediate ferryl species (274, 275), such as Fe(IV)=O•, a powerful oxidizing agent that can lead to lipid oxidation and DNA damage, as discussed in section VI. The rate constant, k, for reactions of Fe2+ is very dependent on the ligand attached to the iron (e.g., it is 2 times greater for Fe2+-ADP than for Fe2+-ATP). Ferryl species exist only in a complex and are produced at their site-specific targets in DNA, proteins, and lipids. Other much slower reactants of H2O2 produced are Fe(III)+OH-+HO• (Fenton reaction); this reaction is 30 times slower than that of ferryl compounds with H2O2, so the hydroxyl radical (HO•) is formed in very small amounts and has no specific reactivity with biomolecules (144). ONOO− is also a stable oxidant that, in its protonated form, ONOOH, can produce oxidative changes in lipids, proteins, carbohydrates, and nucleic acids. Thus, although O2•− and NO• are the initial free radicals, H2O2 rather than O2•− is mostly regarded as the relevant biologically active metabolite.
The measurement of the various ROS is dependent on suitable techniques to assess levels of ROS. The technique should not directly produce ROS or produce artifacts. Preferably the technique should directly measure intra- versus extramatrix generation of O2•−. The most direct and selective approach to measure specific ROS, notably in the matrix, is to use electron paramagnetic resonance (EPR) (16, 129, 151). EPR is based on absorption of microwave radiation stimulated by an electromagnetic field in molecules such as free radicals and transition metal ions with unpaired electrons. Adsorption of microwave energy creates two distinct energy levels from the unpaired electrons as they transition from a lower state to a higher state. The number of unpaired electrons present in a sample is proportional to the amplitude of the ESR signal (111). Only stable ROS that accumulate to measurable levels can be identified directly at biologic temperatures, so either extreme cooling or spin traps, which involve reaction of the radical with an adduct that is stable, is used. Examples of EPR signals are shown in Fig. 4.
The majority of recent studies on mitochondrial ROS production use either the intramatrix fluorescence indicators 2'7'-dichlorofluorescein (DCF) or the fluorescent Amplex Red/horseradish peroxidase technique to assess H2O2 that is released or converted in the buffer (“cytosol”) surrounding the mitochondria or both. Figure 5 shows typical recordings of the rate of accumulation (slope) of H2O2 by the latter technique. Because of the high catalytic activities of MnSOD and CuZnSOD, most O2•− is rapidly converted to H2O2, which is readily permeable to the IMM. Therefore, the measurement of intra- versus extramatrix-generated O2•− is indirect and incomplete because of its rapid dismutation to H2O2 and the sidedness (intra- or extramatrix) of O2•− release.
In isolated tissue, other options exist to assess ROS. One is the shift in absorbance with reduction of cytochrome c by O2•−; other techniques are based on release of photons by chemiluminescence probes, such as liminol or lucigenin, on reaction with ROS (111). Dihydroethidium (DHE), also known as hydroethidium (HE), is a rather specific fluorescent marker for O2•−. O2•− nonenzymatically converts DHE to 2-hydroxyethidium (2-OH-E+) or a precursor (331), which appears to be rapidly made, is labile, and fluoresces at a slightly shorter wavelength (more red) than the heme-peroxidase oxidation product ethidium that can intercalate to bound with DNA. Thus, the 2-OH-E+ signals fluctuate with the O2•− generated in the cell, as now shown in many studies by using spectrophotometry (5, 70, 71, 181, 182, 247, 269, 271, 298, 316, 317). Figure 6 displays a typical recording of O2•- assessed by DHE fluorescence in intact hearts. Note the dynamic effects of changing temperature in the presence and absence of drugs. Caution in the use of DHE (and other probes) lies in its oxidation to ethidium by cytochrome c in the absence of ROS (33). Because the DHE product is altered by drugs that act on the mitochondrion (71), it is assumed that mitochondria are the major source of O2•− by this method in isolated tissue. A derivative of DHE is MitoSOX Red, which has a phosphonium group that selectively targets and enters mitochondria in response to the negative membrane potential; thus, imaging of the signal in cells indicates a mitochondrial source of O2•− (273).
Another method that depends on the interaction of O2•− and NO• is the formation of dityrosine from tyrosine by ONOO− (29); Fig. 7 shows an example of the dynamic changes in dityrosine formation in intact hearts during a change in temperature and drug treatment. It is assumed that the induced shift in wavelength, assessed with spectrometry, is an extracellular indicator of ONOO− (71, 247, 248). Nitration of tyrosine (no color) by ONOO− to 3-nitrotyrosine (yellow color) is another indicator of reactive nitrogen species usually identified in proteins (151).
The specific molecular sites of electron leak are not known with certainty, but a great deal of intense investigation has led to sites within complexes I and III (Figs. 1 and and2A).2A). Specific mitochondrial inhibitors (64, 197) (Table 2) are typically used to (a) force electrons to leak outside of the very tight single-electron transfer mechanism (e.g., antimycin A) leading to O2•−, or (b) to block electron leak (e.g., stigmatellin or rotenone) with glutamate as substrate. For example, the increase in ROS induced by antimycin A can be abolished by adding either stigmatellin or myxothiazol (256). In the presence of glutamate, the inhibitors rotenone, antimycin A, and stigmatellin can each increase ROS. However, in the presence of succinate as substrate, antimycin A can increase, whereas both stigmatellin and rotenone can decrease ROS (255, 256). Thus, the possible pathways have been identified primarily with substrates and inhibitors by using deductive reasoning, with the understanding that the mechanism of electron leak may be quite different in a more-relevant pathophysiologic situation.
This transmembrane complex (Fig. 1) oxidizes NADH [reduced from the transfer of electrons and H+ from tricarboxylic acid (TCA) intermediates], uses coenzyme Q10 (ubiquinone, Q) as the electron acceptor and is coupled (as are complexes III and IV) with proton pumping, thus contributing to the proton motive force, ΔμH+. The actual mechanism of proton pumping is almost completely unknown, in part to the large size of complex I and the difficulty in measuring the intermediates in the coupling reaction (201). Complex I is one of two major sites of entry for reducing equivalents; the other is complex II, succinate dehydrogenase (more precisely known as succinate ubiquinone oxidoreductase) (Figs. 1 and and2).2). Succinate (and α–glycerophosphate transferred from the cytosol to the matrix by the glycerol phosphate shuttle) alone reduces FAD to FADH2, and each molecule furnishes two electrons to the respiratory chain, as do pyruvate and TCA-cycle intermediates isocitrate, α–ketoglutarate, and malate, by reducing NAD+ to NADH.
Complex I is probably the major source of mitochondrial ROS under most physiologic conditions (335). Exogenous quinones can enhance ROS generation from isolated complex I (65). Inhibitors of complex I are useful for determining the source of ROS (96, 252). Several sites between the flavin complex and the quinone site have been proposed to generate O2•− within complex I (77, 134, 160, 199, 216, 320). One or more of the Fe-S centers is a likely source (135, 160, 251), although the ubisemiquinone (QH•) binding site (143, 160, 200), or the flavin complex (216), per se, could also be sources, as depicted in Fig. 8. It has been proposed that complex I has two redox active nucleotide-binding sites; the F site is the location for electron entry for the NADH oxidation by Q, and coupled translocation of 4 H+ for each NADH oxidized; the R site is where electrons exit during the succinate-supported NAD+ reduction, which is O2•− generating (321).
When mitochondria oxidize succinate alone (lacking other TCA substrates), the energy of the ΔμH+ is used to transfer electrons against the redox potentials of the electron carriers from reduced coenzyme Q (quinol, QH2) to NAD+, rather than forward to the final electron acceptor O2 (Fig. 9). This is called reverse electron transfer. Interestingly, a high transmatrix pH gradient appears to contribute more than the ΔΨm to O2•− generation with succinate as the substrate (Fig. 10) (201). It was suggested that the ΔpH-sensitive O2•− generation is mechanistically linked to proton pumping at complex I (201). When electrons are transferred backward from complex II (22, 158, 307), this creates the largest source of O2•− as a percentage of O2 consumption, although this does not likely occur in vivo. Reverse and forward electron flow is believed to contribute to an H2O2-production rate of ~400 and 50pmol/min/mg protein, respectively (256).
Reverse electron transfer requires a large ΔΨm or ΔpH; this occurs as electrons are passed to NAD+ until the pool is fully reduced to NADH. O2•− generation by this mechanism ceases or decreases if the mitochondria are (a) uncoupled by a proton ionophore (e.g., CCCP), (b) generating ATP during state 3 respiration, or (c) leaking protons into the matrix (matrix acidification or uncoupling proteins). Rotenone (Table 2), an irreversible inhibitor of electron transfer from the ubiquinone (Q) binding site to complex I, prevents reverse electron transfer and ceases O2•− release by this mechanism. This implies that O2•− is generated between the rotenone binding site and NAD+ in complex I. The physiological importance of succinate only–induced O2•- generation by reverse electron transfer is dubious (335). However, because succinate is a TCA-cycle intermediate, it along with NADH-linked substrates likely contributes to O2•− generation during electron transfer in vivo, particularly if electron flow is impeded. It was reported that succinate concentration–dependent H2O2 generation was only slightly reduced in the presence of NADH-linked substrates (335) (i.e., succinate-induced H2O2 production occurred under conditions of regular downward electron flow in complex I). Because NADH-linked oxidation was progressively decreased, but not abolished, by increasing succinate concentration, it was proposed that the two substrates compete for electron delivery to complex III, with succinate “pushing” electrons toward and NADH-linked substrates away from complex I (335). In this way, succinate concentration may modulate the rate of H2O2 release by controlling the QH•/Q ratio.
For forward electron transfer, it is postulated that the electrons from NADH-linked substrates (e.g., pyruvate) are moved to ubiquinone Q via the flavin complex and Fe-S centers to form QH•, and in a second step that is linked to a low ΔpH, to form QH2 (200). In the absence of rotenone, NAD+-linked substrates can also enhance O2•- generation at complex I if the ΔΨm is higher than normal (8); this condition of forward electron flow is favored by a high degree of reduction of the redox carriers proximal to the complex I proton pump. Once in a high redox state, QH• can lose its unpaired electron to O2 as all upstream redox centers are fully reduced (200). Moreover, if the ΔpH is large, QH• is longer lasting, and more O2•− may be formed. It is not known whether complex I has a Q cycle (231) like complex III (200). O2•− is believed to be released only on the matrix side of the IMM at complex I (200, 256, 295).
This complex (Fig. 11) is believed to contain a Q cycle with an inner (Qi) and outer (Qo) pool of ubiquinone (Q) facing the matrix (i) and the intermembrane space (o) (42, 43, 97–99, 211, 268, 295, 296, 307, 308). Evidence for the Q cycle (231) arose when it was reported that addition of O2 to anaerobic mitochondria caused a transient reduction of cytochrome b rather than an expected oxidation; this suggested that at least two different sites for electron transfer existed (43). In the subsequent model, QH• is believed to form at both the Qi and Qo sites. First, ubiquinone (Q) is fully reduced to QH2 in the inner side of the IMM and migrates to the outer side, releasing 2 H+ and transferring one electron to cytochrome c1 (Rieske Fe-S protein) to form the first QH• and Q (the electron moves on to cytochrome c and cytochrome c oxidase (complex IV). The second electron reduces cytochrome b so that electrons are moved from Q at the Qo site to the Qi site, and Q is reduced to QH2, completing the cycle. Cytochrome c and cytochrome c oxidase accept only single electrons in sequence. Thus, the complete reduction of Q at the Qi site requires that the Qo site must oxidize two QH2 molecules in two successive turnovers. This powers the complex III proton pumps, but the second bifurcation reaction cannot occur unless the first does (138). Rather than transfer an electron to cytochrome c1, the prolonged lifetime of QH• is believed then to allow it to undergo autooxidation by releasing a singlet electron to be attacked by O2, forming O2•−.
Myxothiazol is a complex III inhibitor (296) that binds at the Qo site of the Q pool to block electron transfer from QH2 at site Qo to Fe-S clusters and to cytochrome b2 (Table 2, Fig. 11). Stigmatellin blocks transfer of the first electron to the Fe-S center (center P, QH2 oxidation site Qo) (96). Antimycin A binds to the Qi site to block electron transfer of the second electron to the Qi site. Thus, antimycin A stimulates electron leak by inhibiting QH2 formation (center N, Q reduction site) so that Q• accumulates at the Qo site, whereas myxothiazol prevents formation of Q• at the Qo site (Fig. 11). The finding that antimycin A did not affect H2O2 generation when added after myxothiazol indicated that myxothiazol stimulates H2O2 generation at a site in the Qo center proximal to the site that is inhibited by antimycin A (296). Myxothiazol and stigmatellin inhibit the ROS-inducing effect of antimycin A, as noted earlier. The QH• radical is fleeting and highly reactive and is not readily detected (320).
These studies suggested that O2•− generated at complex III would be released into the IMS rather than into the matrix space. Evidence for this was found in mitoplasts devoid of the OMM (152). This would further suggest that CuZnSOD is responsible for converting O2•− to H2O2 in the cytosolic space and that MnSOD would not have much importance for protecting the matrix from ROS damage. However, the Qi site can also be a site for ROS generation, particularly with limited electron transfer into the Qo center of complex III (263). Moreover, the MnSOD (matrix SOD) gene knockout is lethal (204, 214), whereas the CuZnSOD (extramatrix SOD) gene knockout is not, although the life span is shortened (118), and oxidative stress is elevated by twofold to threefold (235). This indicates the importance of high MnSOD activity to convert O2•− to H2O2, which unlike O2•−, can easily exit the matrix. The relative amounts and conditions for ROS generated by complex III on either side of the IMM that play a physiologic role remains to be resolved. This information will be important because the balance between antioxidant capacities and O2•− generation within and outside the matrix that contribute to overall ROS emission is not clear.
Most studies indicate that ROS are generated, especially at complex I, during state 4, when the redox state and ΔΨm are high (Fig. 12), but not during state 3 (ADP phosphorylation). In the presence of succinate plus rotenone, O2•− can be generated by forward electron transfer via complex III, although at much lower rates than for reverse electron transfer with succinate alone, as noted earlier. This too requires a high redox state. However, it is possible to observe enhanced ROS at complex III with pyruvate/malate or succinate/rotenone during state 3 conditions (276) (high respiration, ATP synthesis, slightly decreased redox state, and ΔΨm) when O2 availability is high (Fig. 13) or protein concentration is low (276), or when complex I is blocked with rotenone (199) (Fig. 14) and antimycin A is given to block the Qi site (308). It is not known how substantial ROS are formed pathophysiologically (i.e., hypoxia) if state 3 conditions are maintained, but small amounts of ROS clearly can initiate protective pathways, as discussed in sections VII and VIII.
Tissue damage during hypoxia and reperfusion after ischemia has long been known to be associated with increased levels of various ROS (6, 15, 19, 123, 170, 290, 309, 318). Reperfusion after ischemia increases markers of ROS. Administration of scavengers of O2•− and H2O2, but not of O2•− alone, reduces the extent of injury (70). Isolated hearts exhibit a hypothermia-dependent linear increase in both O2•− (Figs. 6 and and15)15) and ONOO− (Fig. 7) (71). Although nonmitochondrial sources of ROS exist during ischemia [e.g., NAD(P)H oxidase (141, 188, 328) and xanthine oxidase (25) in vasculature], the majority of ROS are likely derived from the mitochondrial respiratory complexes, as shown by the damage to respiratory complexes (212, 213, 245) and the effect of mitochondrial inhibitors (5, 28, 71, 81–86). The increase in H2O2 generation in mitochondria isolated after cardiac ischemia/reperfusion injury was enhanced by either rotenone or antimycin A (84); this suggested that both complexes I [Q and/or N2 (final) Fe-S center] and III were damaged and capable of producing O2•− after ischemia/reperfusion injury. The source of O2•− generation during cell hypoxia leans toward complex III, in part because a decrease in oxidant stress during hypoxia in cytochrome c–null cells mimics the actions of myxothiazol and stigmatellin (145, 146), which in effect prevent electron transfer to cytochrome c (Fig. 11). Loss of cytochrome c locks the Rieske Fe-S complex and cytochrome c1 in a reduced state and prevents oxidation of QH2 by the Fe-S complex to QH• (145, 146).
In addition to induction of ROS by hypoxia, ischemia, hypothermia, and mitochondrial toxins, ROS per se may lead to even greater ROS generation in a self-amplifying manner (46, 336). Photoexcitation of individual cardiac myocyte mitochondria caused an initial slow increase in ROS that culminated in a large burst of ROS (Fig. 16) that accompanied an abrupt loss of ΔΨm and opening of the mitochondrial permeability transition (MPT) pore (336). This phenomenon, called ROS-induced ROS release (336), may be associated with Ca2+ overload and may play a role in initiating apoptosis, but whether normal stimuli initiate it is not known. It seems unlikely that Ca2+ overload can itself induce ROS generation because it dissipates ΔΨm and reduces the redox state; but it may hinder the ROS scavenger system so that more ROS are liberated (8). During early ischemia in isolated hearts, ROS increase along with mitochondrial [Ca2+] and the redox state (NADH), as shown in Fig. 17 (5, 70, 298), and treatment with exogenous intra- and extramatrix ROS scavengers reduces ROS and mitochondrial [Ca2+] and better maintains the redox state (NADH) on reperfusion after ischemia, as exemplified in Fig. 18 (70).
Another example is ROS generated by cytosolic NAD(P)H oxidase in a loop with ROS generation by mitochondria and then again by NAD(P)H oxidase (112). In this recent study, it was found that angiotensin II led to mitochondrial dysfunction by activating vascular NAD(P)H oxidase via a protein kinase C (PKC)-dependent pathway. The O2•− produced by NAD(P)H oxidase appeared to increase mitochondrial O2•− generation as well as to decrease NO• bioavailability. Moreover, the H2O2 generated by mitochondria was proposed to cause a feed-forward activation of more O2•− by further activating NAD(P)H oxidase (112).
Because ischemia and hypoxia are marked in the cell not only by ROS release but also by Ca2+ loading, it is not clear whether one leads to the other or whether they are completely independent events (11). Again, it seems unlikely that increased Ca2+ can induce O2•− generation because it can also induce MPT pore opening to abolish ΔΨm (59, 297). Ca2+-transport systems are sensitive to redox conditions, so damage to Ca2+ import and export systems by ROS is more likely to lead to Ca2+ overloading (72). Moreover, ROS-triggered MPT pore opening is potentiated by Ca2+ overload (59). Nevertheless, it has been postulated that Ca2+ can induce ROS physiologically (59), a topic discussed in more detail in section VIII.
An effect of cytochrome c release during MPT pore opening and rupture of the OMM is the activation of caspase proteases. The collapse of ΔΨm does not occur before these caspases are activated and released (267), and cytochrome c may function to shunt any free electrons released outside the matrix to complex IV, thus initially reducing O2•− generation (8). Activated caspase-3 can disrupt electron transfer at complexes I and III (but not at complex IV) to induce O2•− generation as a feed-forward pathway, leading to the collapse of ΔΨm and initiation of apoptosis (267). This is discussed in more detail in section VII.
During the last decade, it has become increasingly clear that ROS are produced not only during reperfusion and reoxygenation after ischemia or hypoxia, but also during ischemia (Figs. 17–19) (5, 28, 70, 181–183, 247, 269, 270, 298, 317) and hypoxia (316). Figures 17 through 19 show examples of ROS fluctuations during ischemia and reperfusion during conditions of ischemia preconditioning, ROS scavenging, and hypothermia in isolated hearts. This seemingly paradoxic situation of ROS emission during ischemia (145, 146) can be explained because cells never truly become anoxic, and so O2 remains available to form O2•− radicals (78). The rate of O2•− generation is chemically proportional by mass action to the O2 concentration times the rate of electron leak (310), but in isolated liver mitochondria, an [O2] >50μM does not alone result in greater O2•− generation, as shown in Fig. 20 (161). Mitochondria can respire normally at a very low po2; only when po2 decreases to <5–7 torr does respiration begin to be limited by the O2 supply (80). The Km for O2 of cytochrome c oxidase is <1μM, and mitochondrial function is independent of a po2 down to <2 torr. If O2 levels are forced to decrease toward zero, electron transfer (respiration) through the respiratory complexes becomes markedly slowed, and the electron carriers operate at a more reduced state. How this occurs is not known, but it has been proposed (145, 146) that if the gene for cytochrome c expression were deleted, this would prevent cytochrome c1 from giving up its electron at that site, and in turn, at the Reiske Fe-S protein site and in QH2 at the Qo site, so that the QH• radical, and thus O2•− would not be generated, and these sites would remain locked in a reduced state.
During conditions of blocked electron transfer via the respiratory complexes and a highly reduced redox state (5, 70, 298), the QH• radical may exist too long or be incapable of full reduction to QH2 because of changes in protein conformation (61, 146, 221), so that electrons leak to combine with O2 in a thermodynamically favorable reaction. Other possibilities are that a low [O2] at complex IV leads to cytochrome c reduction and so limits its capability to scavenge O2•− (63, 292), and that access of QH• to O2 is improved at low O2 levels (146).
Despite the accumulating evidence that ROS are indeed formed during ischemia and hypoxia, and mostly from the mitochondrial respiratory complexes, a recent direct study refutes that hypoxia enhances ROS. In recent, detailed experiments on this apparent paradox (161), it was demonstrated in liver mitochondria that between 20 and 200μM O2 concentration, respiration, H2O2 generation, and cytochromes a/a3 redox state were unchanged; the H2O2 production rate was lower during state 3 than during state 4. Below 20μM, [O2] respiration and H2O2 generation (Fig. 20), both decreased toward zero, and the cytochromes a/a3 redox state became more reduced. It was of interest, however, that H2O2 generation as a percentage of the respiratory rate increased fourfold (state 4) and sixfold (state 3) from 20μM to near zero [O2]. It was concluded (161) that (a) hypoxia in intact cells may elicit ROS from nonmitochondrial sources, (b) probes used are not specific for ROS, (c) cytosolic pathways are required for low O2 sensing and ROS generation, (d) inhibition of complex IV by NO• is required, (e) a hypoxia-induced decrease in tonal release of ROS downregulates signaling pathways involved with ROS, and (f) the fractional increase in ROS generation as a percentage of respiration acts as a hypoxia signal. It is, moreover, possible that hypoxia and ischemia produce derangements in mitochondrial electron transfer because of changes in substrate utilization (e.g., succinate) and redox state that lead to generation of O2•− independent of the effect of a reduced [O2] to accept electrons at complex IV.
Endogenous NO• modulates respiration by its effect to compete reversibly for the O2 binding site on cytochrome c oxidase (complex IV) (45, 58, 119). In endothelial cells, blocking the effect of NO• to compete for O2 at complex IV with an NOS inhibitor enhanced the rate of O2 consumption (respiration) at very low O2 concentrations (89). This indicated that NO• inhibition of O2 binding might be responsible for the inability of mitochondria to consume O2 readily at low O2 concentrations. Endothelial cells are a natural source of NO•, and NO• may be generated in mitochondria (137, 139), and with O2•−, to produce ONOO− (and ONOOH), which can modulate cell function, as addressed in section III. However, whether the enzyme NO• synthase actually resides in the IMM is very controversial (54, 302).
Obviously, much work remains to de done to understand why, how, and from where ROS are generated in isolated cell or organ systems during hypoxia or ischemia but apparently not in isolated hypoxic mitochondria (161) without the use of electron-transport inhibitors. It is not known whether the ratio of O2•− generated from complex I and from the Qi versus Qo sites changes during hypoxia and reoxygenation or during ischemia and reperfusion, and how the activities of MnSOD versus CuZnSOD and other cellular antioxidant systems modulate the emission (generated minus scavenged) of distinct ROS. The specific sites, sidedness, relative amounts of O2•− generation, and its products in these complexes remain to be elucidated. It is possible that the source of O2•− generation and the particular kinds and amounts of ROS change during the course of ischemia and reperfusion, as suggested by the observed time-dependent phasic changes in NADH and FAD redox state, mitochondrial [Ca2+], and O2•− levels (DHE, ETH) during ischemia (examples shown in Figs. 17 through 19) in isolated heart studies (5, 7, 70, 181, 183, 269, 298). Note, for example, in Fig. 19, that MnTBAP, given during brief ischemic pulses before the longer ischemic period, reversed the marked decrease in ROS during late ischemia and reperfusion induced by ischemia preconditioning (IPC), suggesting that O2•− is required to initiate IPC.
Cellular antioxidant defenses depend on the reduction potential of the electron carriers and the reducing capacity of linked redox couples in the matrix (NADH/NAD+ and FADH2/FAD) and cytoplasm (10). Severe cell stress can induce ROS formation that exceeds the capacity of antioxidant enzymes, so that the net emission of ROS is increased. Up to a limit, mitochondrial and cell antioxidant systems are capable of neutralizing excess ROS. These include the intra- and extramatrix SODs (171, 263), and glutathione (GSH) (253, 279) and thioredoxin (TRXSH2) (171) systems (Fig. 21), catalase in the cytosol, and cytochrome c in the intermembrane space (discussed in section VII). GSH, a tripeptide with the thiol (-SH) group of cysteine as the active site, is an abundant source of reducing equivalents; it reduces phospholipid hydroperoxides (PHPs) and H2O2, among other peroxides, via PHP glutathione peroxidase, an enzyme essential for life (8). GSH peroxidase, TRXSH2, and TRXSH2 reductase gene knockouts are also embryonically lethal (234). GSH and TRXSH2 are maintained in a highly reduced state by their reductases; this allows them to reduce effectively H2O2 and lipid peroxides to H2O. The resulting oxidized forms, GSSG (glutathione disulfide) and TRXSS (thioredoxin disulfide), rely on the NAD(P)H/NAD(P)+ redox state to again become reduced. Thus, efficient mitochondrial bioenergetic function is required for the antioxidant activity of these systems.
The activities of antioxidant redox pairs do not appear to be constant. For example, a transcriptional coactivator, (PPAR)-γ coactivator 1-α (PGC-1α), a major regulator of oxidative metabolism and mitochondrial biogenesis, is believed to regulate the mitochondrial antioxidant defense system (153, 312). Endothelial cells that overexpressed PGC-1α upregulated the activity of oxidative stress–protective genes (mRNA for MnSOD, Prx3, Prx, TRXSH2, TRXSH2 reductase, UCP-2, and catalase) and resulted in reduced accumulation of ROS, increased ΔΨm, and reduced apoptotic cell death. This work (312) indicates that increased mitochondrial demand for energy production is met by activation of PGC-1α, which also enhances the mitochondrial antioxidant defenses.
Normally, mitochondria likely serve as a net sink rather than a net source of ROS because of these very efficient scavenger systems (8), but ROS release can become excessive. During ischemia, O2•− and H2O2 levels increase as antioxidant defenses are overwhelmed and complex oxygen intermediates are formed (77, 240). One effect of an excessive initial release of ROS could be more ROS formation by ROS-induced ROS release by a mechanism that may involve MPT pore opening (336) or by caspase-induced modification of respiratory complexes during initiation of apoptosis (267). A high cytosolic GSH/GSSG ratio indicates a large reducing capacity to detoxify ROS and prevent activation of inner-membrane anion channels (IMACs), which precedes MPT pore opening (10). The overabundance of ROS during Ca2+ overload with a collapsed ΔΨm could reflect more the inability of the GSH, Pr-SSG, and TRXSH2 systems to regenerate the reduced state (thus less H2O2 removal) than of more ROS to be generated (8).
Superoxide, as a charged species, is relatively impermeable to membranes but can pass through anion (e.g., IMAC) channels (11, 152). NO• is not ionized and so is much more membrane permeable and can compete for O2•− with SOD to produce the nonradical ONOO− (29), which protonates at a low pH to ONOOH, which is highly reactive to tissues. NO• can also react with O2 to form nitrogen dioxide, NO2• (brown gas pollutant). The fate of O2•− (depending on NO• availability) is typically dismutation to H2O2, which can then be converted to H2O by catalase or to site-directed metal radicals [e.g., (Fe=O)2+] in the presence of transition metals bound within numerous organic molecules. It is the stable and membrane-permeable H2O2 that is the most abundant reactant that, in excess, likely leads to damage to cell structure and function via these ferryl reactants.
Because “good” ROS are regulators or modulators of normal cell function, it is difficult to draw the line on which ROS effects or amounts are beneficial and which are deleterious. The so-called diverse, non–receptor mediated, “bad” effects of ROS are mentioned here, but arbitrarily. It must be emphasized that the net effect of ROS is dependent not only on how much was produced, but also more important, on how much was not inactivated by antioxidant defenses. Thus, the gene-regulated GSH/GSSG, TRXSH2/TRXSS, and Pr-SSG/PrSH redox system, as well as other scavengers that play important roles in protection against excess ROS-induced injury, are the same factors that probably modulate physiologic ROS signaling.
Inhibition of MPT pore opening with cyclosporin A could prevent loss of the carrier cytochrome c so that an e− is normally transferred to complex IV from cytochrome c1, and QH• is oxidized to Q rather than O2 being reduced to O2•−. As discussed also in section VII, MPT pore opening with release of oxidized [Fe(III)] cytochrome c, an O2•− scavenger, may reduce ROS emission in the IMS by oxidizing O2•−. Inhibiting MPT pore opening may also prevent the loss of enzyme scavengers of ROS within the matrix that normally neutralizes ROS in the matrix (220). The independence or interrelation of matrix and extramatrix ROS release and scavenging and Ca2+ loading with MPT pore opening remains unclear.
A sensitive target of excess ROS is mitochondrial DNA (Fig. 22), which is subject to an oxidation rate 10- to 20-fold higher than that of nuclear DNA (140, 142, 262). This is due in part to the lack of histone protection and the proximity to the ETS. For example, ROS intermediates can react with desoxyguanosine to form 8-hydroxydesoxyguanosine (288), which over time interferes with DNA duplication and RNA replication (125, 140). Mitochondrial proteins are also subject to attack by ROS (88, 121, 157, 257, 318). ROS can damage TCA-cycle enzymes, especially aconitase and α-ketoglutarate dehydrogenase (62, 168, 249, 250, 277). Other damaging effects of ROS are on the respiratory complexes, most at complex I and least at complex IV (66). Amino acids are also subject to oxidizing attack by ROS, and the GSH, TRXSH2 and other redox systems are important not only for neutralizing ROS but also for repairing damage due to oxidation of proteins by virtue of their thiol groups (189). Protein oxidation can lead to unfolding and result in loss of catalytic function and degradation (289) [e.g., ROS can damage the Na/K ATPase complex (287)].
Phospholipids are major targets of ROS (48); Transition metal radicals like (Fe=O)2+ can initiate lipid peroxidation cascades in membranes to generate a complex mixture of short-chain aldehydes (Fig. 23), many of which are believed to be toxic. Accumulated ROS damage with aging may damage the lipid composition required for complex I activity. It was found (319) that liver mitochondria of old versus young rats showed a decrease in respiratory rate and reduced activity of complex I but not complex III. It was postulated that an increase in somatic mtDNA mutations would affect the hydrophobic subunits of complex I that are essential for CoQ binding and energy conservation, or that age-related defects of complex I, such as direct alterations of the protein or lipid environment, particularly cardiolipin, which is required for complex I activity, may play a role.
Cardiolipin is a phospholipid found only in the IMM that anchors the mobile electron carrier cytochrome c to the IMM and optimizes the activity of electron-transport complexes, especially complex IV. Cardiolipin can be especially damaged by ROS because of these important roles (213, 258, 260, 261). Submitochondrial particles that exhibit loss of cardiolipin and cytochrome c during ischemia exhibit enhanced H2O2 release (83), which suggests their importance in protection against ROS. Excess ROS are primarily accountable for initiating MPT pore opening with the subsequent swelling, OMM rupture, and release of cytochrome c (187) and feedback generation of ROS (336). Accumulation of ROS over time is believed to be wholly or partially responsible for aging (210), but recently this has been questioned for vertebrates (234).
From the foregoing material, it is evident that, to decrease ROS-mediated cell damage, MPT pore opening (35, 109, 110), and apoptosis due to hypoxia, ischemia, or toxins, one could attempt either to reduce the ROS generation or to enhance ROS scavenging so that overall ROS emission is diminished. Maintaining a large pool of reductants (like the GSH system) requires bioenergetically stable mitochondria to regenerate the reduced state after detoxifying the ROS. Supplying exogenous GSH may be protective but only if sufficient NAD(P)H is also available. Other cytosolic and matrix antioxidant systems exist, as described in section V. The endogenous mechanism of PGC-1α–mediated activation of redox cycling systems also was discussed in section V.
Chance et al. (77) proposed antioxidant capability must be found near sites of ROS production. An antioxidant mechanism, located at the site of electron transfer to complex IV and generation of O2•− in complex III, is carried out by cytochrome c. Cytochrome c can accept or donate an electron, depending on the valence of its heme (Fe) state. It was shown that adding exogenous cytochrome c to cytochrome c–depleted mitochondria reduced O2•− levels by sevenfold to eightfold (325, 332). In its reduced form (Fe2+), cytochrome c normally generates a proton motive force (ΔμH+) during its oxidation by cytochrome c oxidase (complex IV), as one O2 molecule is fully reduced to two molecules of H2O by addition of 4e− (8, 332). In its oxidized form [Fe3+, or Fe(III)], cytochrome c forms O2 and Fe2+ by virtue of the reducing ability (e− donation) of O2•− with Fe(III), thus neutralizing superoxide. Interestingly, Fe2+ cytochrome c, but not Fe(III) cytochrome c, can suppress H2O2 levels by giving an e− to H2O2 to form H2O (332, 333). This is called the “alternative electron-leak pathway,” (i.e., reduction of preexisting H2O2 to H2O by reduced cytochrome c). This mechanism can balance e− leak to O2 (forming O2•−) with an e− leak to H2O2 (forming H2O) and so function to protect against too much ROS formation.
Because Fe(III) cytochrome c is a very efficient scavenger of O2•− within the matrix, when Fe(III) cytochrome c is released into the IMS during MPT pore opening and initiation of apoptotic signaling, it could act as an ideal extramatrix antioxidant (292). However, the loss of cytochrome c to the IMS may allow matrix O2•− to increase because of an absence of cytochrome c (220).
Treatment with superoxide dismutases and catalase is another approach. However, effective delivery of these enzymes into the cytosol and matrix as a therapy is quite problematic. Overexpression of these enzyme systems could be successful in enhancing antioxidant defenses, but because ROS also play a significant physiologic role (described later), the effects may be deleterious, as indicated by a six- to 10-fold overexpression of MnSOD, which caused reduced fertility and abnormal development in mice (264), and the lack of any benefit to prolonging life span in mice overexpressing CuZnSOD (166). However, overexpression of catalase localized to mitochondria increased median life span in mice by 5 months (280). We reported that administration of the chemical SOD mimetic MnTBAP alone actually worsened cardiac function after ischemia in isolated hearts, whereas addition of glutathione and catalase with MnTBAP elicited the best protection (70); Fig. 18 shows the changes in ROS and mitochondrial Ca2+ during and after ischemia with these treatments. Our study suggested that enhanced generation of matrix O2•− during ischemia must be coupled with its dismutation to H2O2 and its oxidation to H2O to offer the best protection.
Another perhaps more practical approach is to modify mitochondrial bioenergetics in a way that leads to reduced capability for producing O2•−. As discussed earlier, it may be possible to generate O2•− during state 3 or when ΔΨm is less polarized (291, 276, 324), but most experimental studies show that even a slight decrease in ΔΨm (191, 201, 324) or transmatrix ΔpH (201) results in marked reduction or cessation of ROS production. Therefore, temporary partial uncoupling of respiration from phosphorylation by inducing an extrinsic proton “leak” may be therapeutic. Essentially, this process short-circuits the gated passage of H+ through complex V to make ATP, so that respiration increases at the expense of no gain in phosphorylation (mild uncoupling) (47, 49, 60). It takes a greater electron-transfer rate to maintain the ΔΨm when an H+ leak is present. Oxidative phosphorylation is not wholly efficient, as up to 25% of basal H+ flux into the matrix could be outside of complex V, as suggested by the heat produced as a fraction of the standard metabolic rate (Fig. 24) (55). The net effect of H+ leak is to stimulate respiration (electron transfer) and produce heat at the expense of ADP phosphorylation.
H+ leak is maximal when ΔΨm is highly polarized, and little or no leak occurs when ATP is being generated. In vitro, state 4 respiration occurs only when no substrate ADP exists, ATP is not being consumed, or complex V is blocked (oligomycin). Thus, H+ leak may be an intrinsic mechanism to attenuate electron leak and O2•− generation when the ΔΨm is large. At a high ΔΨm with a greater probability to generate O2•−, the maximal capability to have a proton leak would tend to decrease ΔΨm and thus O2•−. The mechanism of basal H+ leak is unknown but is thought to be protonophoric (i.e., H+ is carried across the IMM) (55). H+ leak is assessed indirectly by any increase in respiratory rate or decrease in ΔΨm or both during inhibition of complex V (Fig. 25).
A small class of “uncoupling” proteins, called UCPs 1–4, are believed to induce an inward proton “leak” in charged mitochondria (49, 55, 176, 300). AMP may also act allosterically on adenine nucleotide translocase (aka adenine nucleotide transporter, ANT; or ADP, ATP carrier, AAC) to induce an H+ leak (67). Proton leak could also be elicited by cycling of protonated/unprotonated nonesterified fatty acids (Fig. 24) (55, 132, 175), and by repetitive gating of the MPT pore by protons (53, 169). It is feasible that UCPs and AMP could play a role in protecting against ischemia and reperfusion injury, particularly if they could be activated or stimulated before the insult to reduce ΔΨm and before the redox state increases because of inhibited electron transfer. Mitochondria isolated from perfused rat hearts subject to ischemic preconditioning had a greater H+ leak than did ischemia controls (Fig. 25) (238); this H+ leak was completely abolished by the UCP inhibitor GDP, or by the ANT inhibitor carboxyatractyloside (CAT) (238). ATP also inhibits UCPs (174). However, ANT, rather than UCP2, may be most responsible for the antioxidant mechanism in heart muscle mitochondria (238). It was suggested (238) that the smaller H+ leak induced by ischemic preconditioning is mediated by UCP, because the H+ leak in ischemia controls was blocked only weakly by GDP, but strongly by CAT, whereas the larger leak in ischemia/reperfusion alone is mediated by ANT.
Some lipid peroxidation products, such as 4-hydroxy-trans-2-noneal (HNE) may induce partial uncoupling of mitochondria through UCPs and are thought to initiate protective mechanisms (41, 116). HNE can also induce uncoupling of oxidative phosphorylation by enhancing H+ leak through other membrane proteins such as ANT if ΔΨm is high (14). This H+ leak, although preventable by CAT, did not interfere with ANT inhibition (14), so it was suggested that HNE causes a conformation change in ANT. Although ROS can activate UCPs (117) and ROS and HNE can directly cause a small proton leak (55), whether UCPs are activated by either endogenous ROS or HNE has been rigorously questioned (75). Because ubiquinone (Q) is not required for activation of UCPs (174), it is unlikely that ROS from complexes I or III exert a significant role. Together these studies indicate that mild uncoupling by H+ leak reduces ROS formation in state 4 and that UCPs can induce a small H+ leak, but evidence that mitochondrial ROS activate UCPs and the UCPs mediate protection against ROS-induced cell damage is not yet convincing.
It remains unclear how free fatty acids and alkenals like HNE activate proteins such as UCPs and ANT (14). One possibility is that a protonated fatty acid may cross the IMM into the matrix and dissociate, while the fatty acid anion is translocated back out of the matrix by the protein. Flip-flopping of the protonated and unprotonated forms would cause a net flux of H+ into the matrix; other mechanisms are proposed (14). For alkenals (nonfatty acids), it was proposed recently that H+ conductance via ANT occurs by formation of covalent adducts only at a high ΔΨm, thereby exposing sulfhydryl groups and lysine residues to attack by alkenals and causing a conformation change in ANT. This was based on the finding that CAT blocked H+ leak via ANT if given before but not after HNE, whereas blockade of the ANT transport mechanism remained; this that implied HNE caused a permanent conformational change in ANT distinct from its translocation role (14).
Brookes (55) proposed that ROS and H+ leak comprise a loop not requiring UCPs to operate, but rather that the leak is dependent on the ΔΨm alone. A high ΔΨm would generate ROS (191), and the ROS would in turn induce an H+ leak to reduce the ΔΨm in a feedback manner. The generated O2•− could induce an H+ leak indirectly through a lipid oxidation product or possibly via protonation of O2•− in the acidic intermembrane space to HO2• (215), which is membrane permeable and is deprotonated in the alkaline matrix (104). A decrease in the phosphorylation to respiration (P/O) ratio (i.e., uncoupling), may come at the expense of higher ΔΨm and ROS. The higher, but less efficient, metabolic rate associated with H+ leak may be linked to the aging process (55). ROS could cause an H+ “slip” rather than an H+ leak. H+ slip is described as direct reduction of cytochrome c by the O2•− generated into the intermembrane space, thereby bypassing the Qi site (55), as discussed earlier in this section. Instead of 8H+ pumped per 4e− at complex III, only 4H+ would be pumped per 4e−; this effectively represents a 50% reduction in ATP synthesis capacity at a given level of respiration.
Despite rather efficient electron transfer along the respiratory system (i.e., little electron leak), one might ponder why the mitochondrion and its surrounding environment are incapable of removing all the ROS that are produced despite the well-developed and necessary scavenging systems. Or one could argue that natural selection allowed a system that uses ROS for cell signaling and even protection against damage. Although the mitochondrion is largely a sink for O2•− and H2O2 due to its ROS scavenger systems, it likely releases small amounts of H2O2 as a natural chemical messenger in the modulation of mitochondrial and cellular function (27, 113, 126, 183, 203, 206). It was noted that the MnSOD knockout is lethal (204, 214), that MnSOD overexpression is associated with developmental abnormalities (264), and that ROS are required to trigger the apoptotic mechanism (68, 172, 207, 323), which is useful to actively eliminate poorly functioning cells and is required during embryonic development. It is now well known that ROS trigger or mediate ischemic and pharmacologic preconditioning of hearts (15, 27, 102, 181, 183, 248, 254, 303). However, it is difficult to determine how much, from where, and what kind of ROS is beneficial versus detrimental. Although a therapeutic goal may be to reduce ROS emission, particularly during oxidative stress, too much scavenging or the wrong kind of scavenging may eradicate protective cell-signaling mechanisms, as we have shown (70) and several reviews (118, 163, 203, 235, 303) have suggested.
The complex and extensive scavenging systems, the ROS/H+ leak-feedback relation, and the possible role of UCPs in H+ leak are several selected areas of research in which attempts have been made to understand the pathophysiologic regulation of ROS emission. It is evident that the bioenergetic state of mitochondria is important, not only for generation of ROS but also for ROS scavenging, and therefore the regulation of ROS emission. For example, free fatty acids function as natural mild uncouplers by preventing the transmembrane electrochemical H+ potential difference (ΔμH+) from being above a threshold critical for ROS formation by complexes I, III, or both (190).
NO•, a well-known modulator of cell function, is not reviewed in detail here because its generation within mitochondria is debatable, as mentioned in section IV. However, the interaction of NO• and O2•− to form ONOO− and ONOOH as cell-toxic NO• derivatives is well known; but low concentrations of ONOO− (0.1mM) were found to protect neurons against NO•-mediated apoptosis by activating the phosphoinositide-3-kinase (PI3K)/Akt antiapoptotic signaling pathway (106). The activation was accompanied by an increase in oxidized phosphoinositide phosphatase (PTEN), indicating that activation of PI3K/AKt inhibited PTEN and the NO•-mediated apoptotic pathway. ONOO− also was shown to stimulate pentose phosphate pathway (PPP) activity and the accumulation of NADPH, an essential cofactor for glutathione regeneration, and to activate glucose-6-phosphate dehydrogenase (G6PD), an enzyme that catalyzes the first rate-limiting step in the PPP (128). These and other studies [reviewed in (39)] indicate that low levels of ONOO−, like low levels of H2O2, have a potential cytoprotective effect that could be explored therapeutically.
An additional area of focus recently is the role of cation exchangers and channels in modulating ROS and cell signaling. In developing the chemiosmotic theory of energy coupling, Mitchell (229) recognized that mechanisms must exist to exchange anions for OH− and cations for H+ in the IMM; otherwise, in creating the ΔΨm, the mitochondrion would swell and lyse, as only cations could leak in and anions leak out. Thus, mitochondrial cation antiporters (exchangers) were found to be necessary to regulate an osmotic differential across the IMM that would result from the high H+ electrochemical gradient (34, 131). The chemiosmotic hypothesis requires several electroneutral cation antiporters for H+ (NHE, Na+/H+ exchanger; and KHE, K+/H+ exchanger) and a low permeability to K+ and Na+ (34, 131, 239). In this way, countercations to H+ would enter and leave the matrix along their electrochemical gradients, thus preventing osmotic swelling. Modulation of extramatrix and matrix pH affects the protonation of O2•− and ONOO−, as discussed in section III.
The K+ cycle (Fig. 26) is believed to be an important element in modulating mitochondrial function (34, 130, 131, 133) and may modulate ROS release (9, 159). The importance of regulated pathways for both K+ uptake and K+ efflux may be that this flux regulates a very fine tuning of mitochondrial volume that affects the rate of respiration (149, 150). K+ may be taken up via one or more putative mitochondrial K+ channels on a stimulus of low ATP levels or increased Ca2+ (KATP, KCa). In isolated cardiomyocytes, it was reported that increases in ROS and NO• are downstream effects of application of a drug thought to open KATP channels (205), but we have found it difficult to observe significant increases in ROS with pharmacologic preconditioning agents, including K+ channel openers and anesthetics in isolated hearts (182, 183, 298). Figure 27 (from ref. 182) shows a small increase in ROS (ETH) in isolated hearts exposed to sevoflurane that was blocked by MnTBAP but not by the putative KATP antagonist 5-HD.
We have postulated that mitochondrial KCa channels can modulate ROS production up or down by altering ΔpH independent of ΔΨm (158, 159). We proposed that opening KCa channels with NS1619 produces a small H+ leak (Fig. 28) (159). Others have proposed that K+ cycling can account for no more than a small H+ leak (49, 50). The increase in K+ influx may be rapidly matched by K+ efflux and H+ influx via KHE driven by the pH gradient. As K+ enters the matrix electrophoretically via K+ channels (e.g., during ischemia due to Ca2+ loading and low ATP during ischemia) K+ may be exchanged for the energetically more favorable entry of H+ (leak) because of the large ΔpH. If the inward H+ leak is small, a small increase in proton pumping by complexes I, III, and IV enhances the respiratory rate without decreasing ΔΨm, while promoting ROS release (Fig. 29). K+ flux could be a mechanism by which increased matrix Ca2+ modulates respiration. When the H+ leak is large, ΔΨm (and O2•− generation) would decrease, as discussed with other mechanisms of the H+ leak (section VI).
If K+ flux mediates an H+ leak through KHE, then blocking KHE (quinine) should increase matrix K+ but not H+ levels. In preliminary experiments in isolated cardiac mitochondria, we observed that valinomycin, the putative KCa channel opener NS1619, and buffer Ca2+ increased matrix K+ only in the presence of KHE blockade and that matrix pH decreased only in the absence of KHE (3, 4). A caveat to our work is that NS1619 has been reported (74) also to act as a protonophore, in which case, a small direct H+ leak, rather than K+ flux and KHE, might lead to the increase in H2O2 if ΔΨm remains unchanged. Thus, although a large proton leak would be expected to act as a brake on further ROS generation, a small proton leak that stimulates respiration without reducing ΔΨm may actually slightly increase ROS to modulate cell-signaling pathways. Moreover, in a recent study in which K+ flux across the matrix was assessed indirectly by swelling and respiratory changes induced by valinomycin (K+ ionophore) and nigericin (KHE activator), the putative K+ channel openers NS1619 and diazoxide could not be shown to enhance K+ flux (30).
A putative mitochondrial KATP channel opener was also reported to increase ROS production (Fig. 30), but by an alternative mechanism (9). In the presence of ATP, diazoxide increased ROS in isolated mitochondria, and the effect was inhibited by 5-hydroxydecanoate (5-HD); it was hypothesized that matrix K+ influx causes matrix alkalinization, which retards electron transfer at complex I to cause electron leak and ROS formation (9).
The mitochondrial effects of drugs can be different, depending on the substrates used, their concentrations, and the bioenergetic state. For example, in isolated cardiac mitochondria, we found that two putative KATP channel openers (diazoxide and pinacidil) differentially attenuated mitochondrial respiration and that KATP channel antagonists (5-HD and glibenclamide) had no effect on this (270). But when ATP synthase was inhibited by oligomycin, both KATP channel openers accelerated respiration, which was abolished by the KATP channel inhibitors. Many drugs also have biphasic effects, largely because they have concentration-dependent effects on mitochondrial energetics. For example, the various putative types of mitochondrial K+ channel openers at low concentrations accelerate respiration and ROS generation without altering ΔΨm or redox state (9, 159), whereas larger concentrations accelerate respiration even more, depolarize ΔΨm, and decrease the redox state and ROS generation (105, 130, 159). Moreover, these openers may have primary (74, 162) or secondary (159) effects to induce a proton leak into the matrix and may not actually open mitochondrial K+ channels.
Thus, drugs that act on mitochondria can have biphasic as well as nonspecific effects, which complicate our understanding of mitochondrial function. As noted elsewhere, all these findings in isolated mitochondria are highly condition dependent and may not represent what occurs in intact tissue preparations. Further studies are necessary to determine the mechanism and extent to which specific mitochondrial K+ channels do or do not modulate ROS at physiologic levels. That these channels are sensitive to either low ATP levels or high Ca2+ levels suggests that they could have important roles in mitochondrial regulation. Although many of the drugs used to assess ion channels in cell membranes may have specificity for a particular channel, the drugs used to assess mitochondrial channel function (e.g., diazoxide or NS1619) do not appear to be specific for these channels but nevertheless exert protective function by an uncoupling mechanism and or by stimulation of ROS-dependent pathways.
Does Ca2+ directly modulate ROS? Ischemia and reperfusion result in increases in both Ca2+ and ROS, but it is unlikely that Ca2+ itself modulates ROS physiologically for the following reasons: (a) in pathologic situations, excess ROS results in Ca2+ loading, but not the other way around (72); (b) excess Ca2+ loading will itself cause MPT pore opening and dissipation of the ΔΨm (72), but ROS-induced MPT pore opening does not require Ca2+ loading (186); (c) a Ca2+-induced increase in oxidative phosphorylation (and decreased ΔΨm) tends to decrease ROS, not to increase ROS (59, 72, 243, 296); (d) excess Ca2+ increases ROS only when mitochondrial inhibitors are used; otherwise, ROS decreases (59, 65); (e) blocking mitochondrial Ca2+ flux does not interfere with bursts of ROS produced by mitochondria during a local laser flash (11); and (f) O2•−-induced mitochondrial swelling is Ca2+ independent, but Ca2+-induced swelling is ROS dependent (127, 194). However, the possibility that Ca2+-induced MPT pore opening, leading to loss of cytochrome c, differentially alters the matrix and IMS levels of ROS is discussed in section V.
As discussed in section III, it is well known from O2-consumption experiments in mitochondria that the state 3 to state 4 transition is usually accompanied by enhanced ROS production. Uncoupling (increased respiration without increased phosphorylation) typically leads to reduced ROS production, so it is possible that a change in the rate of oxidative phosphorylation modulates the rate of O2•− generation. Moderate increases in m[Ca2+] induced by cell-receptor stimulation have long been thought to stimulate TCA-cycle substrate dehydrogenases and ATP synthase (108, 227, 228). However, it appears that this Ca2+-induced stimulation is a slow process, because the dynamic, rapid stimulation of oxidative phosphorylation is not inhibited by partially blocking mitochondrial Ca2+ entry (148). Oxidative phosphorylation is activated with a time constant of seconds for the creatinine kinase system with an increase in ATP hydrolysis (313) that is too fast for Ca2+ to enter the matrix (52, 232). The prevailing view has been that oxidative phosphorylation is not governed by feedback control (17, 18, 51, 52, 180, 226). Computer simulations of increased cardiac work demands indicate that neither direct activation of ATP use alone nor a direct activation of both ATP use and substrate dehydrogenation, including Ca2+-dependent TCA-cycle dehydrogenases, can account for the constancy of [ATP], [PCr] [Pi], and [NADH] during an increase in O2 consumption in hearts in vivo (192, 193). Rather it was proposed that a so-called “each-step-activation” (or parallel-activation) mechanism is the explanation, in which all oxidative phosphorylation complexes are directly activated by some cytosolic factor related to muscle contraction in parallel with activation of ATP use and Ca2+-dependent TCA-cycle dehydrogenases. This is a broad concept difficult to disprove. Others have proposed that oxidative phosphorylation in the heart is not likely regulated by quickly diffusing ADP and Pi in a simple feedback-control system (314). But another computer simulation (26), which incorporated each respiratory complex, substrate transporters including ANT, and cation fluxes, obtained reasonable fits to published data if Pi-dependent activation of dehydrogenase activity and the electron-transport system (especially complex III) were incorporated. Further development of this model (329) predicted that NAD is a more important regulator than ADP of the TCA-cycle dehydrogenases, and that a decrease in cytosolic pH decreases ΔΨm and the ability to synthesize ATP.
Phosphorylation of complex IV at high matrix ATP/ADP ratios is known to decrease the H+/e− stoichiometry (H+“slip”) of the complex inducing an intrinsic form of uncoupling in which the efficiency of proton pumping is reduced (see also section VII). Respiratory control is generally defined as the increase in respiration in the presence of ADP and its decline once ADP is phosphorylated. A possible mechanism for control of ROS release by oxidative phosphorylation may arise from a “second mechanism of respiratory control,” in which a high matrix ATP/ADP ratio is thought to allosterically inhibit complex IV (219). This inhibition of cytochrome c oxidase is switched on by cAMP-dependent phosphorylation and switched off by Ca2+-activated dephosphorylation (179). cAMP-dependent phosphorylation of cytochrome c oxidase may optimize the efficiency of oxidative phosphorylation by maintaining a low ΔΨm via the second mechanism of respiratory control. The effect of NO• to compete with O2 for binding at complex IV is discussed in section IV. Extrinsic uncoupling (H+ “leak”) of oxidative phosphorylation (e.g., UCPs, HNE) is described in section VII.
Preconditioning is defined by the removal of the stimulus (brief ischemia or drug) some time before the onset of ischemia. Thus, the stimulus does not directly induce the cardioprotection but rather some downstream factors do. Much circumstantial evidence suggests that mitochondria-derived ROS play an important role to initiate ischemic preconditioning (IPC) (Fig. 19) and pharmacologic preconditioning (PPC) (90, 100, 101, 217, 271, 306), which is effected by protein kinase pathways (223, 224, 247). Earlier studies examined the role of KATP channel openers and inhibitors to elicit and to block PPC. A common denominator of these studies was the finding that PPC, induced by the putative mitochondrial KATP channel opener diazoxide, could be blocked by ROS scavengers such as N-acetylcysteine (124) or N-mercapto-propionyl-glycine (254). A mild ROS stress, per se, triggered cardioprotection by activating protein kinases, but ROS also were reported to activate sarcolemmal KATP channels by modulating ATP binding at this channel, as this effect was blocked either by glibenclamide or by ROS scavengers (304, 305). Conversely, IPC requires ROS independent of KATP channel activity (120). Both KATP channel opening and H2O2 inhibit MPT pore opening (94). These and other studies suggested that a feedback loop may exist between KATP channels and ROS to produce cardioprotection (205, 237, 272, 298, 299), but much of this remains unresolved, in part because many of the drugs used as probes have other effects on cell or mitochondrial function (103, 154–156, 259). Kukreja (198) addressed the controversy in this area.
Mitochondrial Ca2+-dependent K+ channel (KCa) as well as KATP channel openers may play major roles not only in modulating mitochondrial bioenergetics but also in cardioprotection against ischemia/reperfusion injury. Most is reported about the putative KATP channel (1, 130, 131, 178, 195, 236). Mitochondrial KCa openers, like KATP channel openers, appear to mediate their effects through ROS-dependent mechanisms, as ROS scavengers, such as MnTBAP, block the protection (Figs. 18 and and19)19) (181, 183, 298). Figure 31 shows that paxilline (PX, an inhibitor of KCa channels) and MnTBAP (TB, an O2•− dismutator) blocked the ROS (DHE) and redox (NADH) reducing effect of NS119 (putative KCa channel opener) during ischemia and reperfusion of isolated hearts. Diverse cardiac-preconditioning drugs given before ischemia and reperfusion in isolated hearts result in improved function along with improved tissue redox state (NADH and FAD), reduced cytosolic and mitochondrial Ca2+ loading, and reduced production of O2•− and ONOO− during ischemia and reperfusion (76, 181, 248, 298). If the observed changes in isolated mitochondria by mild K+ flux that promotes the increase in O2•− (observed via dismutation to H2O2) also occur in vivo, then modulation of mitochondrial bioenergetics by K+ flux might ultimately be a key initiator of mitochondrial and cell protection.
Another approach to protect cells undergoing ischemia is to inhibit complex I during ischemia (81). Rotenone or amobarbital (a reversible inhibitor of complex I) given just before the onset of cardiac ischemia resulted in improved oxidative phosphorylation and retention of cytochrome c in mitochondria isolated during ischemia or on reperfusion (82, 85). We found subsequently that amobarbital, per se, when given for 1min before ischemia, arrested hearts, increased O2•− emission (and NADH, with no change FAD), and reduced mitochondrial [Ca2+] (Fig. 32) (5); this led to reduced O2•− generation and mitochondrial [Ca2+] during ischemia and reperfusion. These ex vivo results suggested that amobarbital blocks electron transfer upstream of the rotenone binding site but distal to the NAD+ binding site. During reperfusion after ischemia, we observed a more reduced NADH redox state, decreased O2•− generation, and reduced mitochondrial [Ca2+] in this intact heart model (5). These mitochondria-protective effects were accompanied by improved cardiac function and smaller infarct size.
Intracellular signaling pathways (e.g., Ca2+, cAMP, protein phosphorylation–dephosphorylation cascades) are essential for modulating cell metabolism and function. Many of these pathways are linked to extracellular receptor ligands at the plasma membrane, but many are derived from within the cell (e.g., the feedback link between the cell's metabolic rate and mitochondrial respiratory activity). Accumulating evidence indicates that ROS are not always deleterious but are essential participants in cell signaling (113, 126, 163, 303). Whether ROS are “bad” or “good” could simply be a matter of quantity or rate of production, or it may also depend on the specific species, the site of generation, and the site of ROS release. The important and well-examined role of non–mitochondria-derived ROS (O2•− and NO•) in vascular sensing and control (328) is not addressed here. The “pathologic” roles of ROS and Ca2+ loading during cellular stress [e.g., ischemia, hypoxia that lead to induction of cell necrosis and apoptosis (68, 207, 323, 330)] were addressed in section III.
ROS-induced cell signaling involves two general mechanisms: alterations in the intracellular redox state (e.g., GSH, TRXSH2, and other redox systems) (Fig. 21) and oxidative modification of proteins (92, 303). ROS appear to regulate a large number of signaling pathways that modulate cell function, but the actual molecules targeted by the ROS are unclear in most cases. Evidence for a physiological role of ROS emission are based in part on the effects of antioxidants and inhibitors of ROS generation to prevent growth factor and cytokine-activated signals, or some other physiological effect, and the effects of exogenously applied oxidants to activate these responses.
Synchronized oscillations in ΔΨm, the redox states of NADH, GSH/GSSG and other redox pairs, and ROS can occur across isolated cardiomyocytes (11) and were postulated to result from the balance between O2•− efflux through IMM anion channels (IMAC) and the intracellular ROS scavenging capacity. This was based on a computational model (93) of mitochondrial energetics and Ca2+ handling that reproduced the observed oscillations, which could be modulated by ROS scavengers or by the rate of oxidative phosphorylation. Oxidation of thiol groups appears to govern the sequential opening of IMAC (before MPT pore opening occurs), based on the GSH/GSSG redox status (10). Such oscillations may play a role in physiologic timekeeping or redox signaling across the cell or both.
The role of ROS in initiating preconditioning protection, particularly well known in the heart, is a vivid example of how ROS modulate cell signaling by mitochondrially generated ROS. To be a modulator of physiological function, the oxidizing effect of ROS should be transient and reversible, generally by other reducing agents. Broad examples of ROS (typically H2O2)-induced signaling (266) are inhibition of tyrosine phosphatase (278, 315) leading to cell proliferation; translocation and activation of serine/threonine kinases, such as protein kinase C (247, 301) and tyrosine hydroxylase mRNA, the rate-limiting enzyme in catecholamine biosynthesis (196); activation of the MAPK family of protein kinases (including MAPKs ERK, JNK, p38) that mediate mitogen and stress-activated signals; NF-κB, a transcription factor that regulates gene expression involving immune and inflammation responses; and AP-1, a transcription complex (303).
One example of a regulatory feedback loop between ROS signaling and enzyme modification is the following: pyruvate was shown to activate JNK1 indirectly via pyruvate-induced mitochondrial H2O2 release (242); in turn, JNK1 inhibited the activity of the metabolic enzymes glycogen synthase kinase 3β, allowing increased activity of glycogen synthase so that glucose was stored as glycogen rather than undergoing glycolysis. Thus substrate metabolism can be tied directly to mitochondrial ROS. Ion-channel function is also subject to modification by thiol reducing and oxidizing agents (73, 163, 164).
How ROS modify the structure and function of these and other signaling proteins is not well understood, but several mechanisms are described (163, 303), particularly those involving the key thiol groups in regulatory proteins (20). One is the reversible oxidation of the sulfhydryl group (−SH) in the cysteine residue to form SOH, SO2H, or SO3H derivatives that alter the activity of an enzyme if the cysteine is located in a catalytic domain or DNA-binding site. For example, H2O2-induced oxidation of a cysteine residue located in its catalytic site precisely and rapidly inhibits tyrosine phosphatase (278, 315). The modification of –SH to –SO2H prevents the enzyme from further oxidation and promotes reversibility to its active form (except for SO3H) by thiols (278, 315).
Another mechanism for posttranslational protein modification by ROS (Fig. 7) is a conformational change in the structure of protein kinases by formation of intramolecular disulfide bridges in cysteine linkages. Another is protein dimerization by intermolecular disulfide linkages of monomers to form an active protein, or dissociation of an inactive protein complex linked by disulfide linkages into an active protein. Several proteins can also become cross-linked because of active dityrosine formation from two tyrosine molecules by H2O2 peroxidase–dependent reactions. Mixed-function oxidases can facilitate metal-catalyzed oxidation by O2•− of proteins with 4Fe-4S tetranuclear cores to mark them for ubiquitination and so alter their stability (303). To qualify as cell-signaling pathways, these oxidation reactions should be reversible by using the cell's redox systems (e.g., GSH and TRXSH2), as discussed previously.
Precursor ROS do not always oxidize proteins directly but rather through an oxidized and reactive phospholipid intermediary (Fig. 23). Oxidized lipids can activate cell-signaling pathways by non-covalent bonding to a receptor, by covalent binding with direct modification of the protein, and by activating pathways that induce ROS formation from other sources (334). The electrophilic lipid peroxidation product 4-HNE (section VI) causes ROS emission that activates MAPK pathways (311), and 4-HNE selectively inactivates thiol containing proteins, such as α-ketoglutarate dehydrogenase and pyruvate dehydrogenase, and thus inhibits NADH-dependent respiration (complex I) (168).
A physiologic response to low O2 requires an O2 sensor coupled to a signal-transduction system (80). For example, ATP synthesis is tightly proportional to O2 consumption in the presence of adequate O2, so the ability of the cell to sense a critically low level of O2 would appear to be fundamental to restoring O2 levels for ATP synthesis (45). But the molecular identity or mechanism of an O2 sensor has not been forthcoming. Several O2-sensor mechanisms have been proposed and reviewed (80): O2-sensitive heme proteins; O2-sensitive ion channels; O2-sensitive NADPH oxidase; O2-sensitive complex IV; and O2-sensitive ROS generation. Many of the proposed O2 sensors do not appear to have the sensitivity to act directly as O2 sensors, but they could all be linked to hypoxia-induced changes in redox state or ROS generation. For example, the capacity of complex IV to produce H2O from O2, electrons, and protons is not impaired at even very low po2 values, but enzyme activity (Vmax) is reduced so that reduced electron transfer and a more reduced redox state ensue (80).
A principal regulator of the response to hypoxia is the hypoxia-inducible factor (HIF) family of transcription factors (31, 61, 115, 165, 177, 185, 221, 282–286, 316, 327). Although constitutively expressed, the HIF-1α subunit has a very short half-life in normoxic conditions because of its rapid hydroxylation by prolyl-4 hydroxylases (PH1–3) (12, 36). During hypoxia, the HIF-1α subunit becomes uninhibited by PHs, so that HIF-1α accumulates and transfers to the nucleus to initiate HIF-1α-mediated transcription of more than 70 genes involved in protein stability, such as heat-shock proteins (HSPs) (40). Therefore, low O2 would appear to control activation of HIF-1α by inhibiting PH activity. The mitochondrion would seem to be a natural sensor for O2 because complex IV is where O2 binding and O2 consumption occur. However, blocking complex III with antimycin A and cyanide to block complex IV did not inhibit the HIF-1α in isolated cardiomyocytes in response to hypoxia (79). Moreover, inhibition of PH did not occur until O2 levels were below 5%, with maximal activation at 0.5% O2 (177). Rather than O2 being the actual sensor, it was proposed that the actual sensor is the redox changes upstream from complex IV (146).
The signaling consequences of hypoxia are probably modified by NO•. Nitric oxide, and other inhibitors of mitochondrial respiration, can prevent stabilization of HIF-1α during hypoxia as a result of an increase in PH-dependent degradation of HIF-1α (147). During inhibition of mitochondrial respiration by hypoxia, O2 was found to redistribute toward nonrespiratory O2-dependent targets such as PHs, so that they no longer registered hypoxia (147). It was concluded that NO• acts as an endogenous regulator of intracellular O2 availability; inhibition of mitochondrial O2 consumption may create a paradox of increased O2 availability for prolyl hydroxylation of HIF-1α, so that the cell may fail to register hypoxia. In a related study (222), lower concentrations of NO• (<400nM) caused a rapid decrease in HIF-1α stabilized by exposing cells to 3% O2. This effect of NO• was dependent on inhibition of mitochondrial respiration, because the NO• effect was mimicked by other inhibitors of mitochondrial respiration, including those not acting at complex IV. It was suggested that although stabilization of HIF-1α by high level of NO• may have pathologic consequence, the inhibitory effect of low levels of NO• may be physiologically relevant.
It has been suggested that O2 sensing by the cell is actually carried out by mitochondrial ROS (31, 61, 80, 146, 221), including NO• (57). One proposal is that the increase in mitochondrial redox state secondary to the reduced Vmax of complex IV during hypoxia creates a more favorable condition for O2•− generation and downstream ROS production (Fig. 33) (80). Thus, the common denominator for O2 sensing by ROS would be the heightened redox state and ΔΨm or both. However, the mechanism for the reduction in Vmax by low O2 levels is not known.
Support for a role of ROS in HIF-1α stabilization derives from studies showing that the absence of cytochrome c or blocked Rieske Fe-S proteins under hypoxic conditions blocks HIF-1α stabilization (61, 221). How H2O2 inhibits PH activity is not yet known, but likely involves one of the ROS- induced protein modifications discussed earlier. When iron chelators were given to block Fe2+ from reacting on PH, or HIF-1α hydroxylation was prevented by 1-dimethyloxallyl glycine (DMOG) to compete at PH, HIF-1α accumulated and induced gene expression (173). These mechanisms depend on the observed increase in mitochondrial ROS in cells when O2 levels decrease during hypoxia and ischemia. However, this must be reconciled by the finding in isolated mitochondria that hypoxia does not itself increase ROS (Fig. 20) (161). Thus, it has been proposed that the hypoxic signal is a decrease in ROS, not an increase (13, 184, 233, 326). Studies in intact cells, tissue and isolated hearts indicate that ROS emission increases during hypoxia and ischemia, as discussed earlier (section III), so the controversy on the actual hypoxia-sensing mechanism continues.
Peroxides produced by mitochondria can create nonenzymatic shunts in the TCA cycle (56, 121). Succinate dehydrogenase (complex II) is very resistant to oxidation by H2O2 or tert-butyl hydroperoxide (tert-BuOOH) (250), but α-ketoglutarate dehydrogenase activity is markedly decreased (168, 250) because of its susceptibility for modification by peroxides of its thiol groups (167, 249). Although this decreases the enzymatic conversion of α-ketoglutarate to succinyl Co-A and succinate by decarboxylation and blocks production of reducing equivalents (NAD+ to NADH) at this step, these oxidants nonenzymatically (chemically) decarboxylate α-ketoglutarate to succinate and thus reduce the level of peroxides (Fig. 34) (121). This bypass depends on a supply of α-ketoglutarate by transamination because its precursor, aconitate, is also inactivated by peroxides (62, 249). This TCA bypass increases the contribution of succinate to the total energy supply (FADH2 dependent) while decreasing the NADH-dependent energy supply (168). Because succinate can contribute to reverse electron transfer and generation at complex I under certain conditions, control of succinate levels during acute cell stress may be beneficial. Another peroxide bypass, the electron-leak pathway mediated by cytochrome c (332, 333), was discussed in section VII.
Hypoxia, increased succinate, and decreased α-ketoglutarate can each lead to inhibition of PH (208, 281), which permits translocation of HIF-α to the nucleus to support HIF-α–dependent transcription of the multitude of genes responsible for O2 transport, vascularization, and anaerobic energy production (208). Addition of succinate improved oxidative phosphorylation after ischemia/reperfusion injury (69), so the role played by succinate is not well understood. These studies again point to H2O2 (and other peroxides), rather than O2•− per se, as the transduction factor in physiologic regulation by ROS (61). Modulation of a key TCA intermediate, in this case succinate, by ROS is a prime example for demonstrating how mitochondria and cells respond to oxidative stress by altering gene expression. If ischemia or hypoxia induces reverse electron transfer by a TCA shunt to the FAD-linked substrate succinate because of inhibition of the NADH-linked substrates, the resulting increase in O2•− generation at complex I may signal measures to alter mitochondrial function. It was noted earlier that inhibiting complex I with amobarbital improved cardiac function after ischemia and reperfusion and improved mitochondrial function (5, 85). Thus, it is possible that reverse electron transfer may be, at least in part, responsible for some of the increase in ROS during ischemia. As discussed in section III, NADH-linked substrates do not appreciably block succinate-induced H2O2 emission via reverse electron transfer (335).
Although H2O2, and in some situations, O2•- (21), are found to trigger a multitude of signaling pathways (most unknown) via their reactants such as peroxides, the precise mechanisms by which mitochondrial ROS initiate the signal-transduction pathways are mostly unknown, despite the TCA succinate bypass example given earlier. One difficulty in a better understanding of ROS modulation of cell signaling is the short half-life and reactive state of many ROS. The source or compartmentalization of the ROS is difficult to discern in most situations. The original oxidation product may itself be oxidized to a product that is the active signaling molecule. A surge in ROS or changes in redox capacity may have local or broad effects on many regulatory proteins simultaneously, so determining specificity is difficult. The ROS can originate from extracellular, cytosolic, or a nonmitochondrial organelle, or from a combination of extra- and intramitochondrial compartments. Moreover, redox regulation of ROS occurs at multiple levels. All of these complicating factors make it quite difficult to sort out the exact signaling mechanisms of ROS. It would appear that Nature has developed a highly sophisticated system to regulate specifically the generation and scavenging of ROS and to modulate the downstream effects of physiologically induced ROS emission on mitochondrial and cell activity.
From this review, it is clear that much has been learned about mitochondrial sources of ROS, the mechanisms of ROS generation and scavenging, and the pathologic effects of “bad” ROS on cell processes. However, knowledge of the source, amount, generation site, and specific “good” reactive species that are involved in cellular and organelle protection and physiologic modulation of cellular activity, even in one cell type (e.g., the cardiac myocyte), remains in its infancy. We hope that future studies will refine and detail our specific understanding of cell-signaling pathways mediated by mitochondrial ROS by using more physiologic approaches in normal cell settings. From these studies, it is hoped that novel therapeutic targets and drugs can be discovered by which to treat mitochondrial and cellular stresses, or at least to assist the organism's intrinsic protective mechanisms.
Reviewing Editors: Hossein Adrehali, Juan Bolaños, Enrique Cadenas, Sergey Dikalov and Rodrigue Rossignol
We thank Mohammed Aldakkak, M.D., and Johan Haumann, M.D., for their careful review and critique of this manuscript. This work was supported in part from the National Institutes of Health (HL089514 to D.F.S. and HL073246 to A.K.S.C.), VA Research Merit Award 8204-05P (to D.F.S.), and a Grant-in Aid from the American Heart Association (0855940G to D.F.S.)
2-OH-E+, 2-hydroxyethidium; 5-HD, 5-hydroxydecanoic acid; ΔμH+, proton motive force, transmembrane electrochemical H+ potential difference; ΔΨm, mitochondrial transmembrane potential; ΔpHm, mitochondrial pH gradient potential; ANT, adenine nucleotide translocase; CAT, carboxyatractyloside; CCCP, carbonyl-cyanide-m-chlorophenylhydrazenone; DCF, 2'7'-dichlorofluorescein; DHE, dihydroethidium; DMOG, 1-dimethyloxallyl glycine; ETS, electron-transport system; FADH2, flavin adenine dinucleotide (reduced); GSH, glutathione (reduced); H2O2, hydrogen peroxide; HIF, hypoxia inducible factor; HNE, 4-hydroxy-trans-2-noneal; IMAC, inner membrane anion channel; IMM, inner mitochondrial membrane; IMS, intermembrane space; IPC, ischemic preconditioning; KATP, ATP-sensitive K+ channel; KCa, Ca2+-sensitive K+ channel; KHE, K+/H+ exchange; L-NAME, NG-nitro-l-arginine methyl ester; MAPK, ERK, JNK, p38, family of protein kinases; MnTBAP, Mn(II)tetrakis(4-benzoate) porphyrin chlorine; MPG, N-(2-mercaptopropyonyl)glycine; MPT, mitochondrial permeability transition; NADH, nicotinamide adenine dinucleotide (reduced); NHE, Na+/H+ exchange; NO•, nitric oxide radical; O2•−, superoxide anion radical; OMM, outer mitochondrial membrane; ONOO–, peroxynitrite; PH, prolyl-4-hydroxylase; PHP, phospholipid hydroperoxide; PPAR-γ, peroxisome proliferator activated receptor gamma; PPC, pharmacologic preconditioning; Prx3, Prx5, peroxirredoxins 3 and 5; Pr-SSG, glutathione-protein mixed disulfides (oxidized); Q, coenzyme Q10, ubiquinone, quinine; ROS, reactive oxygen species; SOD, superoxide dismutase; TCA, tricarboxylic acid; TRXSH2, thioredoxin (reduced); UCP, uncoupling proteins.
No competing financial interests exist.