Sites of Superoxide and Hydrogen Peroxide Production by Muscle Mitochondria Assessed ex Vivo under Conditions Mimicking Rest and Exercise*

Background: Ten mitochondrial sites of superoxide/H2O2 generation are known, but their contributions in vivo are undefined. Results: We assessed their rates ex vivo in conditions mimicking rest and exercise. Conclusion: Sites IQ and IIF generated half the signal at rest. During exercise, rates were lower, and site IF dominated. Significance: Contributing sites ex vivo probably reflect those in vivo. The sites and rates of mitochondrial production of superoxide and H2O2 in vivo are not yet defined. At least 10 different mitochondrial sites can generate these species. Each site has a different maximum capacity (e.g. the outer quinol site in complex III (site IIIQo) has a very high capacity in rat skeletal muscle mitochondria, whereas the flavin site in complex I (site IF) has a very low capacity). The maximum capacities can greatly exceed the actual rates observed in the absence of electron transport chain inhibitors, so maximum capacities are a poor guide to actual rates. Here, we use new approaches to measure the rates at which different mitochondrial sites produce superoxide/H2O2 using isolated muscle mitochondria incubated in media mimicking the cytoplasmic substrate and effector mix of skeletal muscle during rest and exercise. We find that four or five sites dominate during rest in this ex vivo system. Remarkably, the quinol site in complex I (site IQ) and the flavin site in complex II (site IIF) each account for about a quarter of the total measured rate of H2O2 production. Site IF, site IIIQo, and perhaps site EF in the β-oxidation pathway account for most of the remainder. Under conditions mimicking mild and intense aerobic exercise, total production is much less, and the low capacity site IF dominates. These results give novel insights into which mitochondrial sites may produce superoxide/H2O2 in vivo.

is the sum of rates from different sites that prematurely reduce oxygen to superoxide or H 2 O 2 . Each site has its own unique properties. The sites can be distinguished and studied in situ by providing electrons from appropriate substrates and using specific electron transport inhibitors to isolate them pharmacologically. In this way, at least 10 distinct sites have been characterized in rat skeletal muscle mitochondria (16). These sites are represented as red circles in Fig. 1. In order of their maximum capacities, they are as follows: III Qo 4 in complex III (17); I Q (18,19) and II F (20) in complexes I and II; O F , P F , and B F in the 2-oxoglutarate, pyruvate, and branched-chain 2-oxoacid dehydrogenase complexes (16); G Q in mitochondrial glycerol phosphate dehydrogenase (21); I F in complex I (16); E F in ETF/ETF:Q oxidoreductase (22); and D Q in dihydroorotate dehydrogenase (23).
The rates of superoxide/H 2 O 2 production under native conditions (i.e. in the absence of added inhibitors) are much less than these maximum capacities (24). Therefore, the maximum capacities of the sites are not necessarily related to the actual engagement and rate of each site under native conditions. To establish the native rate from each site, we devised novel methods based on measurements of the redox states of endogenous reporters in isolated mitochondria oxidizing conventional substrates (24,25). These studies led to two striking conclusions. First, the overall rates of H 2 O 2 production differed 5-fold between different conventional substrates; they were much higher with succinate than with glutamate plus malate as sub-strate. Second, the relative contribution of each site was completely dependent on the substrate being oxidized. With succinate as substrate, site I Q was dominant; with glutamate plus malate, sites I F , III Qo , and O F all contributed; with palmitoylcarnitine, site II F was also important; and with glycerol 3-phosphate, five sites contributed significantly, including G Q (25). Thus, the relative and absolute contribution of each specific site to the production of superoxide/H 2 O 2 in isolated mitochondria depends very strongly on which conventional substrate is being oxidized.
In complex systems, such as intact cells, different substrates are metabolized simultaneously. The main groups of substrates oxidized by mitochondria are represented in Fig. 1 and Table 1 by colored boxes. In skeletal muscle, oxidation of ketone bodies, amino acids, tricarboxylic acid cycle intermediates, glycerol 3-phosphate, and fatty acids feeds electrons into multiple sites in the matrix dehydrogenases and electron transport chain. Therefore, in vivo, it is likely that several sites produce superoxide/H 2 O 2 simultaneously at different rates.
The conditions experienced by muscle mitochondria within cells differ substantially between rest and exercise. In particular, ADP supply, free Ca 2ϩ , cytosolic pH, and the availability of different substrates are very different, and this will have profound effects on the mitochondrial production of superoxide and H 2 O 2 . During exercise, the levels of both reactive oxygen species and reactive nitrogen species are increased (26). The production of such species is crucial for the beneficial effects of exercise (27)(28)(29) and for force development (30). However, excess production is detrimental for skeletal muscle performance (30,31). Mitochondria are leading candidates for the increased production of these reactive species during exercise (30,32), although others argue for non-mitochondrial sources (26,32).
Several different probes can be used in intact cells to report changes in reactive oxygen species (33,34), but they cannot reliably distinguish the mitochondrial sites active under particular physiological or pathological conditions. This is because pharmacological or genetic manipulation of particular mitochondrial sites invariably alters the redox states of other sites, changing their contributions to overall production of superoxide/H 2 O 2 and making interpretation unreliable. Also, these probes invariably compete with the endogenous antioxidant defenses and measure steady state levels rather than rates of radical production. On the other hand, the rate from each site can now be quantified using isolated mitochondria (24,25). However, given the strong substrate dependence of their superoxide/H 2 O 2 production, isolated mitochondria oxidizing conventional substrates are not sufficiently physiologically relevant.
To improve the physiological relevance of the more amenable mitochondrial system, in the present work, we develop an ex vivo system in which isolated muscle mitochondria are incubated acutely in novel complex media. These media contain the measured physiological concentrations of all metabolites and effectors assumed to be relevant in skeletal muscle cytosol at rest and during mild and intense aerobic exercise. Using this system, we quantify the contribution of each mitochondrial site to total H 2 O 2 production to gain novel insights into the rates and sites of mitochondrial superoxide/H 2 O 2 production in skeletal muscle during rest and exercise.

EXPERIMENTAL PROCEDURES
Animals, Mitochondria, and Reagents-Female Wistar rats were from Charles River Laboratories, age 5-10 weeks, and fed chow ad libitum with free access to water. Mitochondria were isolated from hind limb skeletal muscle at 4°C in Chappell-Perry buffer (CP1; 100 mM KCl, 50 mM Tris, 2 mM EGTA, pH 7.4, at 4°C) by standard procedures (35) and kept on ice during the assays (Ͻ5 h). Protein was measured by the biuret method. The animal protocol was approved by the Buck Institute Animal Care and Use Committee in accordance with IACUC standards. Reagents were from Sigma except for Amplex UltraRed, from Invitrogen.
Oxygen Consumption and Superoxide/H 2 O 2 Production-Skeletal muscle mitochondria (0.3 mg of protein⅐ml Ϫ1 ) were incubated at 37°C for 4 -5 min in the appropriate "basic medium" (Table 1) mimicking the cytosol of skeletal muscle during rest (basic "rest" medium plus oligomycin), mild aerobic exercise (basic "mild aerobic exercise" medium with no further additions), or intense aerobic exercise (basic "intense aerobic exercise" medium plus glucose and sufficient hexokinase after titration (about 0.08 units⅐ml Ϫ1 ) to give 90% of the maximum state 3 oxygen consumption rate). ATP (6 mM) was injected into the chamber, and then after 1 min, the appropriate "complex substrate mix" was added (Table 1). Oxygen consumption rates were measured using a Clark-type electrode fitted in a waterjacketed chamber (Rank Brothers, Bottisham, UK). Rates of superoxide/H 2 O 2 production were measured collectively as rates of H 2 O 2 production as two superoxide molecules are dismutated by endogenous or exogenous superoxide dismutase to yield one H 2 O 2 . H 2 O 2 was detected using 5 units⅐ml Ϫ1 horseradish peroxidase and 50 M Amplex UltraRed in the presence of 25 units⅐ml Ϫ1 superoxide dismutase (36) in a Varian Cary Eclipse spectrofluorometer ( excitation ϭ 560 nm, emission ϭ 590 nm) with constant stirring and calibrated with known amounts of H 2 O 2 in the presence of all relevant additions because some of them quenched fluorescence (36). NAD(P)H and Cytochrome b 566 Redox State-The reduction state of endogenous NAD(P)H was determined by autofluorescence in mitochondria incubated as described for H 2 O 2 production (most of the signal is from bound NADH in the matrix, and NADPH hardly changes in the present experiments, but for transparency we call it "NAD(P)H") using a Shimadzu RF5301-PC dual wavelength spectrophotometer at excitation ϭ 365 nm, emission ϭ 450 nm (18,24). NAD(P)H was assumed to be 0% reduced after 5 min without added substrate. 100% reduction was established internally by adding saturating conventional substrate (e.g. 5 mM malate plus 5 mM glutamate) and 4 M rotenone at the end of each run. Intermediate values of NAD(P)H reduction were measured at 3-4 min after the addition of the appropriate mix of substrates and were determined as percentage of reduced NAD(P)H relative to the 0 and 100% values. The reduction state of endogenous cytochrome b 566 was measured using 1.5 mg of mitochondrial protein⅐ml Ϫ1 with constant stirring at 37°C in an Olis DW 2 dual wavelength spectrophotometer at A 566 nm -A 575 nm (17). This signal reports ϳ75% cytochrome b 566 and ϳ25% cytochrome b 562 (17,37). Cytochrome b 566 was assumed to be 0% reduced after 5 min without added substrate. 100% reduction was established in separate cuvettes with 5 mM succinate and 2 M antimycin A. Intermediate values of cytochrome b 566 reduction were measured over 45 s, ϳ20 s after the addition of the appropriate complex substrate mix, and were determined as percentage of reduced cytochrome b 566 relative to the 0 and 100% values.
Correction for Matrix Peroxidase Activity-H 2 O 2 production rates in Figs. 3 and 12 (but not in other figures) were corrected for losses of H 2 O 2 caused by peroxidase activity in the matrix to give a better estimate of actual superoxide/H 2 O 2 production rates (38). Rates were mathematically corrected to those that would have been observed after pretreatment with 1-chloro-2,4-dinitrobenzene (CDNB) to deplete glutathione and decrease glutathione peroxidase and peroxiredoxin activity, using an empirical equation, where is the rate of H 2 O 2 production in pmol of Due to the non-linearity of the curve, the correction is different at different overall rates. Therefore, the total H 2 O 2 production rates were first corrected using Equation 1. The corrected rate from each site (and its S.E.) was then back-calculated based on its relative contribution in a given condition (Table 3). All sites produce superoxide/H 2 O 2 exclusively into the mitochondrial matrix, except for III Qo and G Q , which generate ϳ50% of the superoxide/H 2 O 2 to the cytosol (24). For these sites, only 50% of the rate was corrected.
Curve Fitting-Data were fit by exponential functions in Figs. 5C, 6C, and 10 (C and F) to yield the parameter values in Equations 2-5, respectively, where H 2 O 2 is the rate of H 2 O 2 production.
Statistics-The calibration curves in Figs. 5C, 6C, and 10 (C and F) were used to calculate the rates of superoxide/H 2 O 2 production at sites I F and III Qo , both directly when assessing the rates from these sites and indirectly when correcting for the effects on these sites of inhibition at other sites ( Table 3). The errors in the calibration curves were taken into account when calculating the S.E. values of the assessed rates. This error was calculated by error propagation as described previously (24).
Because error propagation was used to include the uncertainty from the calibration curve in the assessed rates, the mean Ϯ S.E. values plotted do not represent individual values. Therefore, the statistics were calculated using the averaged values Ϯ S.E. and the number of observations. The significances of differences between the total measured rates and the assessed rates were calculated using Welch's t test. For the regular multiple comparison tests in Figs. 2 and 4, a one-way ANOVA was used, followed by Tukey's post hoc test. When comparing the values with omission of single substrates with the total values, Student's t test was used. p Ͻ 0.05 was considered significant.

Media Mimicking Skeletal Muscle Cytosol during Rest, Mild
Aerobic Exercise, and Intense Aerobic Exercise ex Vivo-In vivo, the substrates for mitochondrial metabolism come primarily from the catabolism of sugars, proteins, and fats. In the cytosol, they are available to the mitochondria as metabolites that can be categorized into five groups according to their origins: ketone bodies, amino acids, tricarboxylic acid cycle intermediates, glycerol 3-phosphate, and acylcarnitines ( Table 1). The major metabolic fates of these substrates are indicated by the colored boxes leading to the mitochondrial dehydrogenases in Fig. 1. These metabolites connect to the 10 sites in skeletal muscle mitochondria known to have significant capacity to produce superoxide or H 2 O 2 . In Fig. 1, these sites are grouped in planes reflecting their operating redox potentials. The top plane represents the NADH/NAD ϩ isopotential group, containing the dehydrogenases that reduce NAD ϩ and the sites in complex I that oxidize NADH. From complex I, the electrons drop down in energy to a more positive redox potential in the ubiquinone pool. The bottom plane represents the QH 2 /Q isopotential group, containing the ubiquinone oxidoreductases that reduce ubiquinone and the sites in complex III that oxidize ubiquinol. The electrons are then transferred to cytochrome c and on to the final acceptor, O 2 , to generate H 2 O (not shown). Table 1 lists the consensus concentrations in rat skeletal muscle cytosol of all metabolites and effectors thought to be potentially relevant to mitochondrial electron transport and production of superoxide/H 2 O 2 . The values were taken from the extensive literature on rat skeletal muscle obtained mostly by freeze-clamp followed by enzymatic or chromatographic quantitation, both in vivo and in isolated muscle preparations. Where appropriate, they were corrected using consensus values for extracellular contamination and for the estimated compartmentation between the mitochondrial matrix and the cytosol. Values are listed for resting muscle and for two exercise conditions: submaximal stimulation, representing mild aerobic exercise, and extensive stimulation, representing intense aerobic exercise. We assumed that the exact K ϩ and Cl Ϫ concentrations were unimportant and therefore used KCl to bring the osmolarity to the physiological value of 290 mosM.
These literature values enabled us to prepare three different media for mitochondrial incubations ex vivo, containing the relevant metabolites and effectors at the concentrations that would be encountered by mitochondria in vivo during rest and mild and intense aerobic exercise. ATP turnover and hence steady-state ADP concentrations were set based on respiration rates as described below. These media are the first that we know of to be carefully designed to mimic the substrate and effector mix in the cytosol of skeletal muscle cells during "rest," "mild aerobic exercise," and "intense aerobic exercise" (Table 1). They enabled us to assess the production of H 2 O 2 from isolated mitochondria in an ex vivo system that closely mimicked the relevant aspects of the cytosol of rat muscle cells in vivo.
Setting ATP Demand during "Rest", "Mild Aerobic Exercise," and "Intense Aerobic Exercise"-To deplete endogenous substrates, rat skeletal muscle mitochondria were incubated for 4 -5 min in the appropriate basic medium lacking all respiratory substrates (Table 1). ATP was added 1 min before respiration was initiated, to minimize the time available for ATP hydrolysis. To initiate respiration, the appropriate complex mix of substrates mimicking rest or mild or intense aerobic exercise (Table 1) was then added ( Fig. 2A).
"Rest"-In resting muscle in vivo, the rates of respiration and ATP synthesis are low (39). To mimic rest ex vivo, a low rate of ATP synthesis by mitochondria incubated in the "rest" medium ( Table 1) was achieved by including oligomycin in the medium to fully inhibit the mitochondrial ATP synthase. This was not ideal, because it implies no ATP turnover at rest, but it was necessary because the relatively high contaminating ATPase activity present in the mitochondrial preparation caused an intermediate respiration rate. The rate of oxygen consumption at "rest" was less than 10% of the maximum rate (Fig. 2, A and B, red).
"Mild Aerobic Exercise"-During exercise in vivo, respiratory rates increase due to increased ATP demand and altered substrate supply and concentrations of effectors (pH, Ca 2ϩ ). Mild exercise induces Յ45% of whole body maximal O 2 consumption rate, VO 2max (39) . To mimic mild aerobic exercise ex vivo, mitochondria were incubated in the basic "mild aerobic exercise" medium in the absence of oligomycin, followed by ATP and then the appropriate complex mix of relevant substrates ( Table 1). The initial fast rate of respiration required to rephosphorylate ADP formed from the added ATP ( Fig. 2A) was ignored. After this fast phase was complete and ATP/ADP settled to a steady state value, ATP demand and respiration rate ran at an intermediate rate limited by the supply of ADP from contaminating extramitochondrial ATPases. In this phase, the respiratory rate was 22% of the maximum rate (Fig. 2, A and B, green).
"Intense Aerobic Exercise"-Metabolic demand increases with exercise intensity. During intense exercise, respiration is Ն65% of VO 2max (39). ADP level increases significantly but is still an order of magnitude lower than the ATP level (40). Cytosolic pH drops to 6.7 due to increased CO 2 and lactate production. To mimic intense aerobic exercise ex vivo, mitochondria were incubated in the basic "intense aerobic exercise" medium, followed by ATP and then the appropriate complex mix of relevant substrates. To achieve high respiratory rates but keep ADP levels relatively low, in separate experiments, the O 2 consumption rate was titrated to 90% of the maximum rate using hexokinase in the presence of glucose to set the rate of extramitochondrial ATP hydrolysis high but still submaximal in each mitochondrial preparation, and this amount of hexokinase  Table 1: ketone bodies, amino acids, tricarboxylic acid (TCA) cycle, glycerol 3-phosphate (GP) shuttle, and ␤-oxidation. Unfilled boxes represent intermediate metabolites not added in the media. Electrons from the oxidation of these reduced substrates enter the mitochondrial electron transport chain through different isopotential groups of redox centers, denoted by the two planes, each operating at about the same redox potential (E h ): NADH/NAD ϩ at E h ϳϪ280 mV and QH 2 /Q at E h ϳϩ20 mV (59). The flow of electrons through NADH to the Q pool at complex I and from the Q pool to cytochrome c (cyt c) at complex III is indicated by the large green arrows dropping down through the isopotential planes. Enzymes that feed electrons into each isopotential group are represented as ovals, electron transport inhibitors are drawn with red blunted arrows, and CN-POBS, a suppressor of electron leak at site I Q that does not inhibit electron transport, is drawn with a green blunted arrow. Electrons from NAD-linked substrates enter the NADH/NAD ϩ pool through appropriate NAD-linked dehydrogenases (DH), including those for branched-chain 2-oxoacids (BCOADH), pyruvate (PDH), 2-oxoglutarate (OGDH), and others that are grouped together because there is no evidence that they produce superoxide/H 2 O 2 . Electrons from NADH flow into complex I (site I F ) and then drop down via site I Q to QH 2 /Q in the next isopotential pool, providing the energy to generate protonmotive force (pmf). Q oxidoreductases, including complex II, mitochondrial glycerol 3-phosphate dehydrogenase (mGPDH), ETF:QOR, and dihydroorotate dehydrogenase (DHODH), can also pass electrons into the Q pool. Electrons flow from QH 2 through complex III to cytochrome c and finally to oxygen (not shown), again pumping protons and generating pmf. The redox state of NADH (outlined in blue) reports the redox state of the first isopotential group. The redox state of cytochrome b 566 (outlined in blue) reports the redox state of the second isopotential group (24). (about 0.08 units⅐ml Ϫ1 ) was included in the "intense aerobic exercise" medium ( Fig. 2, A and B, blue). Maximum phosphorylating respiration was set by adding excess hexokinase (Fig. 2B, white).
Mitochondrial H 2 O 2 Production during "Rest," "Mild Aerobic Exercise," and "Intense Aerobic Exercise"-The rate of generation of superoxide and H 2 O 2 was measured under the three conditions described above as the rate of extramitochondrial H 2 O 2 production in the presence of exogenous superoxide dismutase to convert any superoxide to H 2 O 2 for assay. Fig. 2C shows that rates were linear for 2-3 min after the addition of the complex substrate mixes and then decreased. Some metabolites were present at very low concentrations (Table 1) and were likely to be consumed quickly, so ideally we would measure initial rates, but because the system was not at steady-state over the first minute in the "exercise" medium ( Fig. 2A), we calculated the rates as the pseudolinear rate between 1 and 3 min (dotted boxes in Fig. 2, A and C). Fig. 2D shows that the highest rate of H 2 O 2 production was observed when mitochondria were incubated in the medium mimicking the cytosol of skeletal muscle at rest (336 pmol of H 2 O 2 ⅐min Ϫ1 ⅐mg of protein Ϫ1 ). In the media mimicking mild aerobic exercise and intense aerobic exercise, the rates were significantly lower.
Unphysiological concentrations of conventional substrates and the presence of appropriate electron transport inhibitors favor high rates of mitochondrial superoxide/H 2 O 2 produc-tion. These rates can be up to 2% of the total respiration rate of uninhibited resting mitochondria (2,41). Under more realistic non-inhibited (native) conditions with conventional substrates, the percentage of electron leak to H 2 O 2 is only ϳ0.15% (42,43). Table 2 shows the % electron leak when mitochondria were incubated under conditions mimicking rest, mild aerobic exercise, and intense aerobic exercise. At "rest," the electron leak was 0.35%. During "exercise," it was 10 -35-fold lower: only 0.03% for "mild aerobic exercise" and 0.01% for "intense aerobic exercise." The ex vivo data in Fig. 2D and Table 2 suggest that skeletal muscle mitochondria in vivo are unlikely to contribute to the overall increase in reactive oxygen species production observed during exercise, supporting the conclusions of others (26,32). Instead, the mitochondrial production rate may decrease during exercise, because the redox centers that donate electrons to O 2 become more oxidized during exercise (see below).
Sites of Superoxide/H 2 O 2 Production-Ten mitochondrial sites associated with the tricarboxylic acid cycle and electron transport chain are known to produce superoxide or H 2 O 2 (Fig.  1). Their maximum capacities are shown in Fig. 3, on the same scale as the ex vivo rates from Fig. 2D after all of the values have been corrected for consumption of intramitochondrial H 2 O 2 by matrix peroxidases. Clearly, the maximum capacities of several sites are much higher than the ex vivo rates, making it impossible to predict a priori which sites contribute most to the total signal ex vivo or in vivo. The rates of H 2 O 2 production measured under the different conditions in Fig. 2D are the sums of the rates from different sites. Our goal here was to determine the contribution of each site during "rest," "mild aerobic exercise," and "intense aerobic exercise" in the tractable ex vivo system because no adequate methods exist to address this question in vivo or in intact cells. We assume that the sites contributing ex vivo in media containing the physiological cytosolic concentrations of all substrates and effectors thought to be relevant will approximate the sites contributing in vivo.
Conventionally, complexes I and III are thought to be the dominant sources of mitochondrial superoxide/H 2 O 2 in isolated mitochondria and cells (10,13). Complex I produces superoxide at two distinct sites: site I Q , a high capacity site, most obviously active during reverse transfer of electrons from ubiquinol to NADH (18,19,44,45); and site I F (18), now known to have much lower capacity (16). Complex III can produce superoxide at high rates from site III Qo (2,17). The rate of superoxide production from site I F depends on the redox state of the complex I flavin, which in turn has a unique relationship to the redox state of mitochondrial NAD(P)H. Similarly, the rate of superoxide production from site III Qo is related to the redox state of cytochrome b 566 (17). The redox states of NAD(P)H and cytochrome b 566 were therefore used as endogenous reporters of the rates of superoxide production at sites I F and III Qo , respectively (24,25). Fig. 4 and Table 3 show the reduction levels of NAD(P)H and cytochrome b 566 when mitochondria were incubated in the three media. During "rest," NAD(P)H was nearly 100% reduced, and cytochrome b 566 was 30% reduced. During "exercise," both were significantly more oxidized. To assess the rate of superoxide production by sites I F and III Qo , calibration curves were built to establish the relationships between superoxide production from these sites and the reduction levels of the endogenous reporters.
Flavin of Complex I, Site I F , during "Rest"-NAD(P)H redox state was calibrated as a reporter of the rate of superoxide production at site I F , as described previously (24). Fig. 5A shows the observed rate of H 2 O 2 production from site I F as a function of the concentration of malate in the presence of rotenone (with

TABLE 2
Electron leak during "rest," "mild aerobic exercise," and "intense aerobic exercise" Rates were calculated per nmol e Ϫ by multiplying the rates of H 2 O 2 production in Fig. 2D in pmol of H 2 O 2 ⅐min Ϫ1 ⅐mg of protein Ϫ1 by 0.002 and rates of oxygen consumption in Fig. 2B (16)). Fig. 5B shows the NAD(P)H reduction state under the same conditions. Fig. 5C replots the data from Fig. 5, A and B, to show the dependence of the measured rate of H 2 O 2 production arising from superoxide production at site I F on the NAD(P)H reduction state. As shown in Fig. 4, NAD(P)H during "rest" was almost 100% reduced. From Fig. 5C, the contribution of site I F during "rest" was assessed to be 65 Ϯ 12 pmol of H 2 O 2 ⅐min Ϫ1 ⅐mg of protein Ϫ1 (mean Ϯ S.E.; n ϭ 8). The contri-butions of site I F and other relevant values developed below are summarized in Table 3.
Site I F accounted for 20 Ϯ 3% of the total rate of H 2 O 2 production measured during "rest" (Fig. 5D). Therefore, other sites also produced superoxide/H 2 O 2 at rest.
Outer Ubiquinone Binding Site of Complex III, Site III Qo , during "Rest"-Cytochrome b 566 redox state was calibrated as a reporter of the rate of superoxide production at site III Qo as described previously (24). Fig. 6A shows the observed myx-  (Table 1). A, rate of superoxide production from site I F , defined by the presence of 4 M rotenone, 1.5 mM aspartate, and 2.5 mM ATP, at different malate concentrations. B, dependence of NAD(P)H reduction level on malate concentration under the same conditions (100% reduction was established by adding 5 mM malate plus 5 mM glutamate). C, calibration curve obtained by combining the y axes from A and B, showing the dependence of the rate of superoxide production from site I F on the NAD(P)H reduction level. D, total rate of H 2 O 2 production at "rest" (from Fig. 2D) and the contribution of site I F assessed from the reduction state of NAD(P)H in Fig. 4   Corrections applied for changes in redox state of NAD(P)H and cytochrome b 566 , and rate of superoxide/H 2 O 2 production from each site during "rest," "mild aerobic exercise," and "intense aerobic exercise" ⌬ control values in columns B and C were used to correct the observed changes in H 2 O 2 production rate in column A for changes in the redox states of NAD(P)H and cytochrome b 566 , as shown in the last column. a 2.5 M CN-POBS decreased I Q superoxide production by only 65%; therefore, values were scaled to account for CN-POBS potency. b 10 M CN-POBS reduced I Q superoxide production by only 75%; therefore, values were corrected for CN-POBS potency. c The relative contribution was calculated based on the CDNB corrected rates for each site. The S.E. was scaled according to the internal error for each site. DHAP, dihydroxyacetone phosphate; nd, not determined.
othiazol-sensitive rate of H 2 O 2 production from site III Qo as a function of the ratio of added succinate and malonate in the presence of rotenone. At each point, the rate of H 2 O 2 production from site I F was calculated from the observed NAD(P)H reduction level using Fig. 5C and subtracted from the total observed rate to give the rate specifically from site III Qo (Table  3). Fig. 6B shows the cytochrome b 566 redox state under the same conditions. Fig. 6C replots the data from Fig. 6, A and B, to show the dependence of the rate of H 2 O 2 production arising from superoxide production at site III Qo on the cytochrome b 566 reduction state. Fig. 4 and Table 3 show that cytochrome b 566 was 30% reduced during "rest." Using the calibration curve in Fig. 6C, the contribution of site III Qo was 55 Ϯ 7 pmol of H 2 O 2 ⅐min Ϫ1 ⅐mg of protein Ϫ1 (mean Ϯ S.E.; n ϭ 12) (Fig. 6D). Site III Qo accounted for 15 Ϯ 2% of the total rate of H 2 O 2 production measured during "rest" (Fig. 6D and Table 3). After accounting for the contributions of sites I F and III Qo to the total rate of H 2 O 2 production, there was still a significant difference between the total rates measured and assessed ( Fig. 6D). Therefore, other sites also produced superoxide/ H 2 O 2 at rest.
Ubiquinone Binding Site of Complex I, Site I Q , during "Rest"-Site I Q has a high capacity for superoxide production during reverse electron transport when succinate is oxidized (19, 43) (Fig. 3) and during forward electron flow from NAD-linked substrates under particular conditions (44). However, the importance of this site in vivo and under semiphysiological conditions (25,46,47) is unknown. The contribution of site I Q during "rest" ex vivo was assessed by adding rotenone to inhibit electron transport at this site, followed by correction for consequent changes in the rate of superoxide production from sites I F and III Qo . Rotenone was added 1 min prior to the substrate mix and after correction significantly decreased the rate of superoxide/H 2 O 2 production at "rest" (Fig. 7A). The rate attributed to site I Q was 79 Ϯ 23 pmol of H 2 O 2 ⅐min Ϫ1 ⅐mg of protein Ϫ1 (mean Ϯ S.E.; n ϭ 6).
CN-POBS at 2.5 M selectively blocks 65% of superoxide production from site I Q with no effects on oxidative phosphor-  (Table 1). A, rate of superoxide production from site III Qo , defined as the rate in the presence of 4 M rotenone sensitive to 4 M myxothiazol, at different ratios of succinate/malonate (total concentration 5 mM). Data were corrected for changes in the contribution of site I F using changes in the NAD(P)H redox state and the calibration curve in Fig. 5C (Table 3) (24). B, dependence of cytochrome b 566 reduction state on the succinate/malonate ratio under the same conditions (100% reduction was established by adding 5 mM succinate plus 2 M antimycin A). C, calibration curve obtained by combining the y axes from A and B showing the dependence of the rate of superoxide produced from site III Qo on the cytochrome b 566 reduction level. D, total rate of H 2 O 2 production at "rest" (from Fig. 2D) and the contributions of sites I F (from Fig. 5D) and III Qo (assessed from the reduction state of cytochrome b 566 in Fig. 4 (Table 1). A, total rate of H 2 O 2 production and the rates in the presence of a I Q electron transport inhibitor (4 M rotenone) or a suppressor of I Q superoxide production (2.5 M CN-POBS) after correction for changes in the rates of sites I F and III Qo assessed by changes in NAD(P)H and cytochrome b 566 redox state ( Table 3). The yellow hatched area represents the contribution of site I Q . B, total rate of H 2 O 2 production at "rest" (from Fig. 2D) and the contributions of sites I F (from Fig. 5D), III Qo (from Fig. 6D), and I Q (from A). Inverted error bars indicate the propagated errors for each site, and the conventional error bar indicates the propagated sum of these errors. C, total rate of H 2 O 2 production and the rate in the presence of 4 M malonate to inhibit the flavin site of complex II, after correction for changes in the rates of sites I F , I Q , and III Qo ( Table 3). The red hatched area represents the contribution of site II F . D, total rate of H 2 O 2 production at "rest" (from Fig. 2D) and the contributions of sites I F (from Fig. 5D), III Qo (from Fig. 6D), I Q (from A), and II F (from C). Inverted error bars indicate the propagated errors for each site, and the conventional error bar indicates the propagated sum of these errors. Values are means Ϯ S.E. (error bars) (n ϭ 3-20). **, p Ͻ 0.005; ***, p Ͻ 0.0001 by one-way ANOVA followed by Tukey's post hoc test. *, p Ͻ 0.05 by Student's t test. Welch's t test was used to determine p ϭ 0.003 and p ϭ 0. 39. ylation, although at higher concentrations, it is less selective (48). CN-POBS was added 1 min prior to the substrate mix and significantly decreased the rate of superoxide/H 2 O 2 production at "rest." The rate attributed to site I Q after correction for the 65% effect of CN-POBS was 66 Ϯ 29 pmol H 2 O 2 ⅐min Ϫ1 ⅐mg of protein Ϫ1 (mean Ϯ S.E.; n ϭ 6) (Fig. 7A), indistinguishable from the estimate made using rotenone.
From the data obtained with CN-POBS and corroborated using rotenone, site I Q accounted for 23 Ϯ 11% of the total rate of H 2 O 2 production measured during "rest" (Fig. 7B and Table  3). This is the first evidence in any system that site I Q may be an important contributor to mitochondrial superoxide/H 2 O 2 production ex vivo and in vivo. The sum of the rates of superoxide/ H 2 O 2 production from sites I F , III Qo , and I Q was significantly different from the total rate of H 2 O 2 production measured during "rest" (Fig. 7B), suggesting that yet other sites also produced superoxide/H 2 O 2 at rest.
Flavin of Complex II, Site II F , during "Rest"-The flavin site of complex II, site II F , has a high capacity for superoxide/H 2 O 2 production (20) (Fig. 3). When downstream electron transport is prevented, there is a peak of superoxide/H 2 O 2 production from this site at succinate concentrations close to the physiological range (20), indicating that site II F might also contribute to H 2 O 2 generation at rest. The contribution of site II F during "rest" ex vivo was assessed by adding malonate to inhibit electron transport at this site, followed by correction for consequent changes in the rate of superoxide production from sites I F , III Qo , and I Q . The addition of malonate significantly decreased the rates of H 2 O 2 production at "rest" ( Table 3). The change in rate attributed to site II F after correction was 80 Ϯ 35 pmol H 2 O 2 ⅐min Ϫ1 ⅐mg of protein Ϫ1 (mean Ϯ S.E.; n ϭ 7) (Fig.  7C and Table 3). Site II F accounted for 24 Ϯ 10% of the total rate of H 2 O 2 production measured during "rest" (Fig. 7D and Table  3). This is the first evidence in any system that site II F may be an important contributor to mitochondrial superoxide/H 2 O 2 production ex vivo and in vivo.
The sum of the rates of superoxide/H 2 O 2 production from sites I F , III Qo , I Q , and II F was not significantly different from the total rate of H 2 O 2 production measured during "rest" (Fig. 7D). However, we still assessed the potential contributions of the remaining sites.
Electron-transferring Flavoprotein (ETF) and ETF:Q Oxidoreductase (ETF:QOR), Site E F , during "Rest"-␤-Oxidation of fatty acids, primarily palmitoylcarnitine, is an important source of ATP for resting skeletal muscle. During oxidation of palmitoylcarnitine, electrons are transferred partly to NAD ϩ and partly to ETF and then through ETF:QOR to the Q-pool. Superoxide/H 2 O 2 production in the ETF/ETF:QOR system (site E F ) probably occurs at the flavin site of ETF (22). Site E F produces superoxide/H 2 O 2 at a maximum rate of 210 pmol of H 2 O 2 ⅐min Ϫ1 ⅐mg of protein Ϫ1 (22). However, during oxidation of palmitoylcarnitine plus carnitine as the sole added substrate under native conditions without inhibitors, the observed H 2 O 2 production was mainly from sites I F , III Qo , and II F , leaving a small (but not statistically significant) shortfall that might have been from other sites, such as ETF (22). The contribution of site E F during "rest" ex vivo was assessed by omitting palmitoylcarnitine from the substrate mix, followed by correction for consequent changes in the rate of superoxide production from sites I F and III Qo . Palmitoylcarnitine omission significantly decreased the rate of H 2 O 2 production (Fig. 8A). There was no consequent change in the rates from sites I F and III Qo ( Table 3). The possibility that palmitoylcarnitine withdrawal could alter the rate of superoxide/H 2 O 2 production from site II F was tested by adding malonate in a mix without palmitoylcarnitine, which decreased H 2 O 2 production to the same extent as in the presence of palmitoylcarnitine, showing that palmitoylcarnitine omission did not affect the rate from site II F (data not shown). The rate attributed to site E F was 46 Ϯ 21 pmol of H 2 O 2 ⅐min Ϫ1 ⅐mg of protein Ϫ1 (mean Ϯ S.E.; n ϭ 8) (Fig. 8A and Table 3). Site E F accounted for 13 Ϯ 6% of the total rate of H 2 O 2 production measured during "rest" (Fig. 8B and Table 3).
Other Sites, P F , O F , G Q , B F , and D Q , during "Rest"-The pyruvate dehydrogenase complex, site P F , and the 2-oxoglutarate dehydrogenase complex, site O F , have the greatest capacities for superoxide/H 2 O 2 production in the NADH/NAD ϩ isopotential group (16) (Figs. 1 and 3). Omission of pyruvate from the substrate mixture slightly slowed the rate of H 2 O 2 production during "rest," although this effect was not statistically significant (p ϭ 0.5). Fig. 8C shows that after correcting for consequent changes at sites I F , III Qo , and I Q (Table 3), site P F was not significantly active at rest. Arithmetically, omission of pyruvate gave small negative rates for site P F after correction ( Fig. 8D and Table 3), presumably reflecting noise in the assays and small errors in the assumptions.
Similarly, omission of 2-oxoglutarate or the glycerol 3-phosphate/dihydroxyacetone phosphate couple from the substrate mixture resulted in non-significant small increases in the total rates of H 2 O 2 production after correction for consequent changes in other sites ( Fig. 8C and Table 3). Therefore, these sites were also unlikely to contribute to the measured rates of H 2 O 2 production. Again, these assessments gave small negative rates for sites O F and G Q after correction( Fig. 8D and Table 3).
Site D Q in dihydroorotate dehydrogenase (23) was considered to make zero contribution to total H 2 O 2 production in our analysis because this site has low maximum capacity, the enzyme has low activity in skeletal muscle (49), and its substrate, dihydroorotate, is present at very low concentrations in rat skeletal muscle and was therefore not included in the substrate mix (Table 1). Similarly, site B F of the branched-chain 2-oxoacid dehydrogenase (16) was considered to make zero contribution because (i) branched-chain 2-oxoacids are present at very low concentrations in rat skeletal muscle mitochondria at rest and were therefore not included in the substrate mix (Table 1) and (ii) the addition of the transaminase inhibitor aminooxyacetate at 1 mM did not alter the measured rate of H 2 O 2 production (data not shown), indicating that transamination of the added branched-chain amino acids to generate branched-chain 2-oxoacids and produce superoxide/H 2 O 2 at site B F was not a major pathway. Fig. 8D shows the final sum of the assessed rates of superoxide/H 2 O 2 production from all of the sites, plotted both as a stack of positive and notionally negative contributions and as the sum of these values. This final value was indistinguishable from the total measured rate of H 2 O 2 production (Fig. 8D and Table 3), so the observed rate of H 2 O 2 production was entirely accounted for, within experimental error, by the sum of the assessed rates from each site.
In summary, when rat skeletal muscle mitochondria were incubated in a complex medium mimicking the cytosol of rat skeletal muscle at rest, containing physiological concentrations of all substrates and effectors thought to be relevant, the total rate of H 2 O 2 production was 337 Ϯ 21 pmol of H 2 O 2 ⅐min Ϫ1 ⅐mg of protein Ϫ1 (mean Ϯ S.E.; n ϭ 11). Ranked by magnitude, the sites that contributed to this H 2 O 2 production were sites II F (24 Ϯ 10%), I Q (23 Ϯ 11%), I F (20 Ϯ 3%), III Qo (15 Ϯ 2%), and E F (13 Ϯ 6%).
"Mild Aerobic Exercise"-During mild aerobic exercise, substrate concentrations are different, and the free Ca 2ϩ concentration is 20-fold higher compared with rest (Table 1). When mitochondria were incubated in the medium mimicking mild aerobic exercise, ATP turnover was driven by extramitochondrial ATPases, respiration ran at 22% of the maximum rate (Fig.  2, A and B), and NAD(P)H and cytochrome b 566 reduction lev-els were significantly lower than at "rest" (Fig. 4 and Table 3). Fig. 2, C and D, shows that the measured rate of H 2 O 2 production during "mild aerobic exercise" was only a quarter of the rate at "rest." The contribution of each site to superoxide/H 2 O 2 production during "mild aerobic exercise" was assessed exactly as described above for "rest" ( Table 3). The contributions of sites I F and III Qo were assessed using the calibration curves in Figs. 5C and 6C. Fig. 9 shows the final sum of the assessed rates of superoxide/H 2 O 2 production from all sites, plotted both as a stack of positive and notionally negative contributions and as the sum of these values. This final value was indistinguishable from the total measured rate of H 2 O 2 production ( Fig. 9 and Table 3), so the observed rate of H 2 O 2 production was entirely accounted for, within experimental error, by the sum of the assessed rates from each site. Compared with "rest," the absolute contribution of each site decreased. The decrease was most marked for the sites in or connected to the QH 2 /Q isopotential  (Table 1). A, total rate of H 2 O 2 production and the rate in the absence of palmitoylcarnitine after correction for changes in the rates of sites I F and III Qo ( Table 3). The olive hatched area represents the contribution of site E F . B, total rate of H 2 O 2 production at "rest" (from Fig. 2D) and the contributions of sites I F (from Fig. 5D), III Qo (from Fig. 6D), I Q (from Fig. 7A), II F (from Fig. 7C), and E F (from A). Inverted error bars indicate the propagated errors for each site, and the conventional error bar indicates the propagated sum of these errors. C, total rate of H 2 O 2 production and the rates in the absence of pyruvate, 2-oxoglutarate, or glycerol 3-phosphate plus dihydroxyacetone phosphate (DHAP) to assess the contributions of sites P F , O F , and G Q . D, total rate of H 2 O 2 production at "rest" (from Fig. 2D) and the contributions of sites I F (from Fig. 5D); III Qo (from Fig. 6D); I Q (from Fig. 7A); II F (from Fig. 7C); E F (from A); and G Q , P F , and O F (from C). Inverted error bars indicate the propagated errors for each site. Because some of the assessed contributions were negative, the sum of the contributions by each site was also plotted (error bar indicates the propagated sum of the errors for each assessed site) and used for the statistical test. Values are means Ϯ S.E. (error bars) (n ϭ 3-20). *, p Ͻ 0.05 by Student's t test. NS, not significantly different using one-way ANOVA followed by Tukey's post hoc test. Welch's t test was used to determine p ϭ 0.42 and p ϭ 0.15. group (Fig. 1), reflecting their sensitivity to the reduction level of the Q pool, which dropped by 50% (Fig. 4). Site I Q is also very sensitive to the protonmotive force (19), which will have decreased when ATP turnover increased. The absolute contribution of the only site in the NADH/NAD ϩ isopotential group that made a large contribution to H 2 O 2 production at rest, site I F , decreased much less, because the reduction level of NAD(P)H dropped by only 20% (Fig. 4). As a result, the relative contribution of site I F to the rate of superoxide/H 2 O 2 production increased significantly from 20% during "rest" to 44% during "mild aerobic exercise" (Table 3).
In summary, when rat skeletal muscle mitochondria were incubated in a complex medium mimicking the cytosol of rat skeletal muscle during mild aerobic exercise, containing physiological concentrations of all substrates and effectors thought to be relevant, the total rate of H 2 O 2 production decreased to about 25% of the rate in the medium mimicking rest. The sites that contributed to this H 2 O 2 production were, ranked by magnitude, sites I F (44 Ϯ 4%), I Q (18 Ϯ 15%), III Qo (15 Ϯ 4%), and maybe E F (12 Ϯ 15%) and G Q (5 Ϯ 7%).
"Intense Aerobic Exercise"-During intense aerobic exercise, free Ca 2ϩ concentration is high; ATP concentration is maintained almost constant (ϳ6 mM); ADP concentration, which is more than 1 order of magnitude lower than ATP concentration, can increase 5-fold; and creatine phosphate hydrolysis generates an increased concentration of P i (40). CO 2 and lactate production increase, lowering intracellular pH (40,50,51), and substrate and effector concentrations also change (Table 1). When mitochondria were incubated in the medium mimicking intense aerobic exercise, ATP turnover was driven by sufficient extramitochondrial hexokinase to cause respiration to run at 90% of the maximum rate (Fig. 2, A and B), and NAD(P)H and cytochrome b 566 reduction levels were significantly lower than at "rest" (Fig. 4 and Table 3). NAD(P)H reduction level was also significantly lower than during "mild aerobic exercise" (Fig. 4 and Table 3). Fig. 2, C and D, shows that the measured rate of H 2 O 2 production during "intense aerobic exercise" was only one-sixth of the rate at "rest," but there was no significant difference in rates between "mild aerobic exercise" and "intense aerobic exercise." The contribution of each site to superoxide/H 2 O 2 production during "intense aerobic exercise" was assessed as described above for "rest" and "mild aerobic exercise" ( Table 3). The contributions of sites I F and III Qo were assessed using the new calibration curves in Fig. 10, C and F, because we found that the calibrations were different in this medium compared with the other two media (Figs. 5C and 6C), probably because of the lower pH. Fig. 10, A and B, show the dependences of the observed rate of H 2 O 2 production from site I F and the NAD(P)H redox state on the concentration of malate in the medium mimicking intense aerobic exercise. Fig. 10C replots the data from Fig. 10, A and B, to show the dependence of the measured rate of H 2 O 2 production arising from superoxide production at site I F on NAD(P)H reduction state in this medium. Similarly, Fig. 10D shows the observed rate of H 2 O 2 production from site III Qo as a function of the ratio of added succinate and malonate in the presence of rotenone and myxothiazol. At each point, the rate of H 2 O 2 production from site I F was calculated from the observed NAD(P)H reduction level using Fig. 10C and subtracted from the total observed rate to give the rate specifically from site III Qo . Fig. 10E shows the cytochrome b 566 redox state under the same conditions. Fig. 10F replots the data from Fig.  10, D and E, to show the dependence of the rate of H 2 O 2 production arising from superoxide production at site III Qo on the cytochrome b 566 reduction state in this medium. Fig. 11 shows the final sum of the assessed rates of superoxide/H 2 O 2 production from all sites, plotted both as a stack of positive and notionally negative contributions and as the sum of these values. This final value was indistinguishable from the total measured rate of H 2 O 2 production (p ϭ 0.46) (Fig. 11 and Table 3), so the observed rate of H 2 O 2 production was entirely accounted for, within experimental error, by the sum of the assessed rates from each site. However, because of the low rate of H 2 O 2 production during "intense aerobic exercise," the experimental errors were relatively high, particularly for the sites making smaller contributions (Table 3). Under this condition, site I F was the major contributor (99 Ϯ 26%) and possibly the only one (Fig. 11). However, we could not discard the possibility that other sites were also active. Note that the total measured rates of H 2 O 2 production were indistinguishable between "mild aerobic exercise" and "intense aerobic exercise" (Fig. 2D), but the sites contributing under each condition appeared to differ (Table 3).
In summary, when rat skeletal muscle mitochondria were incubated in a complex medium mimicking the cytosol of rat skeletal muscle during intense aerobic exercise, containing physiological concentrations of all substrates and effectors thought to be relevant, the total rate of H 2 O 2 production decreased to about 15% of the rate in the medium mimicking rest. The main site that contributed to this H 2 O 2 production was site I F (99 Ϯ 26% of the total measured rate and 42 Ϯ 10% of the sum of assessed rates).  Table 1). The total rate of H 2 O 2 production (from Fig. 2D) and the contributions of sites I F , III Qo , I Q , E F , G Q , P F , II F , and O F were assessed in the same way as for "rest" in

Rates of H 2 O 2 Production after Correction for Losses of H 2 O 2 in the Mitochondrial
Matrix-Peroxidases in the mitochondrial matrix degrade some H 2 O 2 before it can escape and be registered by the extramitochondrial horseradish peroxidase/ Amplex UltraRed assay. The losses can be greatly decreased by depleting mitochondrial glutathione. Previous work has established the relationship between observed rates of H 2 O 2 production and total rates after correcting for H 2 O 2 losses (24,38), so we corrected the data presented above using Equation 1 to give a more realistic estimate of the true total rates and the rates from each site. Most of the sites produce superoxide/H 2 O 2 exclusively to the matrix, so their relative contributions will not change substantially after correction. Importantly, however, ϳ50% of the superoxide/H 2 O 2 from sites III Qo and G Q is produced toward the cytosol and avoids matrix peroxidase scavenging (17,21). Therefore, the contribution of these sites becomes relatively smaller after this correction. Fig. 12 shows the corrected data and represents our best estimate of the total rate of mitochondrial superoxide/H 2 O 2 production ex vivo and the contribution of each site during rest and mild and intense aerobic exercise. At "rest," more than 50% of the total rate of FIGURE 10. Calibration of the reporters of sites I F and III Qo during "intense aerobic exercise." Mitochondria were incubated in the basic medium mimicking intense aerobic exercise (Table 1). A-C, construction of the calibration curve showing the dependence of the rate of superoxide production from site I F on NAD(P)H reduction level exactly as in Fig. 5, A-C, but using the basic medium mimicking intense aerobic exercise. D-F, construction of the calibration curve showing the dependence of the rate of superoxide production from site III Qo on the cytochrome b 566 reduction level exactly as in Fig. 6, A-C, but using the basic medium mimicking intense aerobic exercise. Values are means Ϯ S.E. (error bars) (n ϭ 3). FIGURE 11. Contributions of different sites during "intense aerobic exercise." Mitochondria were incubated in medium mimicking intense aerobic exercise ( Table 1). The total rate of H 2 O 2 production (from Fig. 2D) and the contributions of sites I F , III Qo , II F , E F , O F , G Q , I Q , and P F were assessed in the same way as for "rest" in Fig. 8 using the calibration curves in Fig. 10, C and F. Inverted error bars indicate the propagated errors for each site. Values are means Ϯ S.E. (error bars) (n ϭ 3). Because some of the assessed contributions were negative, the sum of the contributions by each site was also plotted (error bar indicates the propagated sum of the errors for each assessed site) and used for the statistical test. Welch's t test was used to determine p ϭ 0.46.
superoxide/H 2 O 2 production was shared between site I Q and site II F , previously not considered to be physiologically relevant. During "exercise," superoxide/H 2 O 2 production decreased substantially, and the low capacity site I F accounted for half or more of the superoxide produced.

DISCUSSION
Mitochondria can produce superoxide or H 2 O 2 from at least 10 different sites (16,23). Which sites are the major producers in vivo is unknown. Currently, there are no methods available that can be used in intact cells or tissues or in vivo to identify unambiguously which sites are active or to measure the rates of production from each site. However, we recently developed methods to identify and quantify the rates from these sites in isolated mitochondria (24,25). From these studies, we concluded that the overall rates and relative contributions of each site are strongly dependent on the substrate being oxidized. As a consequence, it is unreasonable to extrapolate from isolated mitochondria to tissues in vivo when the results in vitro are obtained using single conventional substrates at unphysiological concentrations.
The work described here represents a first step toward characterizing the mitochondrial sites of superoxide and H 2 O 2 production in skeletal muscle in vivo. We analyzed the sites ex vivo in media designed to mimic the in vivo concentrations of all substrates and effectors we thought likely to be important in determining mitochondrial superoxide and H 2 O 2 production in muscle at rest and during exercise. With this approach, extrapolations from ex vivo to in vivo are much more reasonable. Of course, the approach can still be criticized for assuming that other effects, such as fragmentation of mitochondria during isolation, do not substantially affect mitochondrial function and that other effectors that are not included in the media, such as components of unknown signaling pathways from the cytosol that do not work by altering the concentrations of the explicit metabolites, can be ignored. We used the approach specifically for rat skeletal muscle mitochondria, but it could be used with mitochondria from any cell type or any tissue from control or disease models as long as all metabolites and effectors are present at their physiological or pathological concentrations.
In the current paper, we identified and quantified for the first time all mitochondrial sites that are likely to contribute to mitochondrial superoxide and H 2 O 2 production in skeletal muscle at rest and under mild aerobic and intense aerobic exercise. We found that the maximal capacities of the sites did not correlate with their native rates of superoxide and H 2 O 2 production ex vivo. At "rest," half of the total rate of H 2 O 2 production was from sites I Q and II F , two sites whose relevance in vivo was not previously appreciated. Also, the low capacity site I F produced H 2 O 2 as fast as the highest capacity site, III Qo .
In isolated mitochondria, site I Q can produce superoxide/ H 2 O 2 at high rates during both the forward and reverse reactions (43,44). The contribution of site I Q in cells and in vivo is unknown. In some cells, the addition of rotenone decreases cellular production of reactive oxygen species (29,(52)(53)(54), which is consistent with a substantial contribution of site I Q . However, if the cells were oxidizing predominantly NADlinked substrates, the addition of rotenone would block reduction of ubiquinone and decrease superoxide/H 2 O 2 production from sites in the QH 2 /Q isopotential group, particularly sites III Qo and II F . If the cells were running reverse electron transport (29), the addition of rotenone would block reduction of NAD ϩ and decrease superoxide/H 2 O 2 production from sites in the NADH/NAD ϩ isopotential group, particularly sites I F , P F , and O F . These alternative explanations greatly weaken any conclusion from rotenone inhibition experiments that site I Q is active in cells. Conversely, in many other cells, rotenone increases cellular production of reactive oxygen species (55)(56)(57), which appears to be inconsistent with a substantial role for site I Q . However, if the cells were oxidizing primarily NAD-linked substrates, the addition of rotenone would cause reduction of the NADH/NAD ϩ isopotential group and increase superoxide/ H 2 O 2 production from sites I F , P F , O F , etc., masking any decrease in the rate from site I Q . In general, whether rotenone addition causes an increase or decrease in observed production of reactive oxygen species may depend upon the balance of rates and capacities of sites in the NADH/NAD ϩ and QH 2 /Q isopotential groups and not only on the activity of site I Q itself. These considerations highlight the problems of valid interpretation when using electron transport chain inhibitors or genetic manipulations to define mitochondrial sites of superoxide/ H 2 O 2 production in cells if changes in the redox states of all other relevant sites are not properly accounted for as in Table 3.
Here, we used two different approaches to assess the contribution of site I Q . First, we used rotenone to inhibit the Q-site of complex I and corrected the observed overall changes in H 2 O 2 production for the changes caused by alterations in the redox states of the NADH/NAD ϩ isopotential pool (site I F ) and QH 2 /Q isopotential pool (site III Qo ), exposing the contribution of site I Q . Second, we used CN-POBS, which specifically suppresses superoxide production from site I Q without inhibiting electron transport (48) or altering the redox state of the NADH/ NAD ϩ and QH 2 /Q isopotential pools (Table 3). Both approaches gave the same result (Fig. 7A). This is the first evidence that site I Q is active under a semiphysiological condition and therefore may also be significantly active in vivo.
The physiological contribution of superoxide/H 2 O 2 production from site II F has also been unappreciated, mostly because this site was not recognized as an important potential source until recently. However, the capacity of site II F to produce superoxide/H 2 O 2 in muscle mitochondria is great and is second only to site III Qo (Fig. 3) (20). When succinate is used as a single substrate by isolated skeletal muscle mitochondria, the rate of production of superoxide/H 2 O 2 by site II F is highly dependent on the concentration of succinate, rising to a peak at about 400 M (20), a concentration that approximates the physiological cytosolic level in skeletal muscle (Table 1), and then falling away at higher concentrations as succinate becomes inhibitory for this reaction. Our data indicate that at "rest," ϳ25% of the total H 2 O 2 produced by muscle mitochondria originates from site II F , indicating that this site is an important source of superoxide/H 2 O 2 in vivo. Although the raw data are sparse, the concentration of succinate in skeletal muscle may rise from about 200 M at rest to about 300 M during exercise (Table 1). However, this increase ex vivo was not associated with a corresponding increase in the contribution of site II F in "aerobic exercise" (Fig. 12), presumably because the activatory effect of higher succinate concentration was more than compensated for by the inhibitory effect of oxidation of the QH 2 /Q pool during "exercise" (Fig. 4). Nonetheless, under conditions in which succinate concentration rises, such as hypoxia-reperfusion (58), but the QH 2 /Q pool may not become oxidized, the contribution of site II F in cardiac muscle or brain may be even greater.
The mechanisms underlying the decrease in total mitochondrial superoxide/H 2 O 2 production during "aerobic exercise" are clear from our results. Despite increases in the concentrations of several substrates and effectors in the media mimicking exercise, including citrate, malate, pyruvate, succinate, glycerol 3-phosphate, acetylcarnitine, and free Ca 2ϩ (Table 1), the redox centers in both the NADH/NAD ϩ and QH 2 /Q isopotential groups became more oxidized (Fig. 4 and Table 3). This shows that the dominant effect of the media mimicking exercise was increased supply of ADP caused by turnover of ATP, leading to an increased respiration rate and a more oxidized electron transport chain. Because of the steep dependence of superoxide/H 2 O 2 production on the redox state of electron transport chain components (Figs. 5, 6, and 10), this led to the steep decline in total mitochondrial superoxide/H 2 O 2 production seen in "exercise" in Fig. 2.
Similarly, the mechanisms underlying the decreased relative contributions of some sites are also clear. Because the oxidation of the NADH/NAD ϩ pool was less severe than the oxidation of the QH 2 /Q pool ( Fig. 4 and Table 3), the decrease in the rate of superoxide production by site I F was less than the decrease at the other major sites, which are all linked to the QH 2 /Q pool, and the relative contribution of site I F increased. Although complex I has a high capacity for superoxide production at site I Q , the flavin site (site I F ) has one of the lowest capacities in skeletal muscle mitochondria (Fig. 3). During "aerobic exercise" when the protonmotive force decreases and the NADH and QH 2 pools become more oxidized, other sites contribute much less to overall H 2 O 2 production, allowing the low capacity site I F to dominate (Fig. 12).

CONCLUSIONS
There is a lack of information about the specific mitochondrial sites of superoxide and hydrogen peroxide production that are physiologically or pathologically active in vivo or in intact cells. Our data provide the first realistic estimate of the sites active in skeletal muscle in vivo under three physiological conditions, rest, mild aerobic exercise, and intense aerobic exercise, and provide the first evidence that in addition to the sites currently recognized (sites III Qo and I F ), sites I Q and II F may also be very important. These results illuminate the specific sites that may need to be normalized to prevent excessive mitochondrial superoxide and H 2 O 2 production both physiologically and in many disease states.