Despite the considerable interest in superoxide as a potential cause of pathology, the mechanisms of its deleterious production by mitochondria remain poorly understood. Previous studies in purified mitochondria have found that the highest rates of superoxide production are observed with succinate-driven reverse-electron transfer through complex I, although the physiological importance of this pathway is disputed because it necessitates high concentrations of succinate and is thought not to occur when NAD is in the reduced state. However, very few studies have examined the rates of superoxide production with mitochondria respiring on both NADH-linked (e.g. glutamate) and complex II-linked substrates. In the present study, we find that the rates of superoxide production (measured indirectly as H2O2) with glutamate+succinate (∼1100 pmol of H2O2·min−1·mg−1) were unexpectedly much higher than with succinate (∼400 pmol of H2O2·min−1·mg−1) or glutamate (∼80 pmol of H2O2·min−1·mg−1) alone. Superoxide production with glutamate+succinate remained high even at low substrate concentrations (<1 mM), was decreased by rotenone and was completely eliminated by FCCP (carbonyl cyanide p-trifluoromethoxyphenylhydrazone), indicating that it must in large part originate from reverse-electron transfer through complex I. Similar results were obtained when glutamate was replaced with pyruvate, α-ketoglutarate or palmitoyl carnitine. In contrast, superoxide production was consistently lowered by the addition of malate (malate+succinate ∼30 pmol of H2O2·min−1·mg−1). We propose that the inhibitory action of malate on superoxide production can be explained by oxaloacetate inhibition of complex II. In summary, the present results indicate that reverse-electron transfer-mediated superoxide production can occur under physiologically realistic substrate conditions and suggest that oxaloacetate inhibition of complex II may be an adaptive mechanism to minimize this.
Mitochondria and ROS (reactive oxygen species) have emerged as central players in a wide variety of pathologies. Elevation in mitochondrial ROS production has been hypothesized to occur in Parkinson's disease and Alzheimer's disease as well as aging [1–5]. However, this belies the fact that the mechanics of mitochondrial superoxide production remain quite poorly understood.
While it is generally accepted that mitochondria are the main site of cellular ROS production, studies in isolated mitochondria have shown that the amount of H2O2 released by mitochondria (H2O2 originates from the dismutation of O2• , and is much easier to measure than O2•) under most conditions is rather modest (<0.1% electron leak). The exception to this is succinate-driven reverse-electron transfer through complex I, a condition yielding by far the highest rates of H2O2 release in isolated mitochondria [7–10]. This holds true in mitochondria isolated from diverse tissues (up to 2000 pmol of H2O2/min per mg of mitochondrial protein [7–15]). Indeed, the frequently quoted value of 1–5% of electrons passing through the chain being diverted to the formation of superoxide is only true under this condition [16,17], and the notion that mitochondria are the strongest source of cellular ROS largely derives from these measurements. In isolated mitochondria, reverse-electron transfer through complex I occurs when the ubiquinol pool is in a highly reduced state and a strong membrane potential is present, i.e. the energy of the membrane potential drives the ubiquinol (with electrons provided by succinate)-dependent reduction of NAD+ to NADH with electrons passing in the reverse direction through complex I .
However, a number of investigators have questioned whether reverse-electron transfer (and hence superoxide production emanating from it) can occur in vivo [19–21]. Data refuting this possibility have been presented by Hansford et al. . These authors showed that superoxide production decreases dramatically when the concentration of succinate is reduced below 1 mM. Further, it is thought that reverse-electron transfer-mediated superoxide production cannot occur when the final electron acceptor in this pathway, NAD+, is reduced [18,22]. Hansford et al.  reported that succinate reverse-electron transport-mediated superoxide production is dramatically decreased when the NAD-linked substrates glutamate+malate are added concomitantly with succinate . In contrast, a very recent report by Zocarato et al.  indicates that addition of glutamate/malate only modestly decreases succinate-supported superoxide production. Superoxide production is much lower when electron transport is in the forward direction, with substrates like glutamate+malate or pyruvate+malate (∼10–90 pmol of H2O2·min−1·mg−1 [8,9,12,14,15,23], and is even reported to be essentially undetectable by some studies [14,24]).
We undertook the present study (preliminary results of which were presented at the 2006 Society for Free Radical Biology and Medicine Conference ) to address experimentally the argument that reverse-electron transfer-ROS production only occurs under extreme, non-physiological substrate conditions (contrary to what Hansford et al.  reported a number of years ago, a report published while this study was in progress did indeed conclude that high rates of ROS production could occur under less extreme substrate conditions ). To this effect, we have measured the rate of superoxide production with the complex II-linked substrate succinate and NADH-linked substrates alone or in combination, over a range of substrate concentrations. We note that virtually all previous studies that have studied superoxide production in mitochondria have employed either NADH- or complex II-linked substrates alone, not in combination. In vivo, both NADH and succinate are present, with greater abundance of the former.
We observed very high rates of superoxide production when skeletal-muscle mitochondria were respiring on both succinate and the NADH-linked substrates, glutamate, pyruvate and palmitoyl carnitine. We show that these rates of superoxide production remain high even at low substrate concentrations and originate in large part from reverse-electron transfer through complex I, despite the fact that NAD was in the reduced state. Only malate was found to inhibit superoxide production with succinate, which we propose, is due to the oxaloacetate inhibition of complex II [27–29].
Taken together, our results indicate that reverse-electron transfer-mediated superoxide production can occur even when NAD is in the reduced state, and at substrate concentrations that are within the physiological range. Further, our results suggest that oxaloacetate inhibition of complex II may be a deliberate adaptation to minimize reverse-electron transfer-mediated superoxide production.
Mice were maintained in the C57 B6/J background and housed under specific pathogen-free barrier conditions. Mice were anaesthetized and killed by cervical dislocation. All procedures were approved by the IACUC (Institutional Animal Care and Use Committee) at the University of Texas Health Science Center at San Antonio. Mice used in the present study were 3–6 months of age.
All chemicals were obtained from Sigma–Aldrich (St. Louis, MO, U.S.A.) unless otherwise specified.
Isolation of skeletal-muscle mitochondria
Whole hind-limb skeletal-muscle mitochondria were purified by the method of Chappell and Perry as modified by Ernster and Nordenbrand [30,31]. Whole hind-limb skeletal muscle was excised, washed with 150 mM KCl, placed in Chappell–Perry buffer (100 mM KCl, 50 mM Tris, 1 mM EDTA, 5 mM MgCl2, 1 mM ATP, pH 7.44; composition of modified Chappell–Perry buffer: 100 mM KCl, 50 mM Tris, 0.1 mM EDTA, 5 mM MgCl2, 0.2 mM ATP, pH 7.44) with protease and minced. Homogenization was carried out by hand, in an all-glass Potter–Elvehjem-type homogenizer. The homogenate was spun for 10 min at 600 g and the supernatant was passed through cheesecloth and centrifuged at 14000 g for 10 min. The pellet was then washed once in modified Chappell–Perry buffer with 0.5% BSA and once without BSA. H2O2 release experiments were conducted immediately after mitochondrial isolation. The respiratory control ratio was ∼10 with glutamate/malate.
Preparation of respiratory substrates, inhibitors and uncouplers
Stock solutions (0.5 M) of the substrates glutamate, α-ketogluta-rate, malate and succinate were prepared in the same reaction buffer as above, their pH was adjusted to 7.44, and aliquots were stored at −80 °C. Stocks of pyruvate and oxaloacetate (0.5 M) in the same buffer were prepared fresh for each experiment, minimizing the time on ice. Palmitoyl carnitine was prepared at 20 mM stock in 10 mM Hepes (pH 3.0) and aliquots were stored at −80 °C. Respiratory inhibitors rotenone and antimycin A were dissolved in DMSO or ethanol and stored at −80 °C. Unless otherwise indicated, the working concentration for these inhibitors was 5 μM. The uncoupler FCCP (carbonyl cyanide p-trifluoromethoxyphenylhydrazone) was prepared in ethanol at 10 mM and stored at −80 °C, and the final concentration, unless stated otherwise, was 150 nM.
Measurement of mitochondrial superoxide production
Superoxide production was measured indirectly as H2O2 release from intact mitochondria. H2O2 was determined with Amplex™ Red (Molecular Probes, Eugene, OR, U.S.A.; product no. A-12212 ), as described previously [33–35]. HRP (horseradish peroxidase; 1 unit/ml) catalyses the H2O2-dependent oxidation of non-fluorescent Amplex™ Red (80 μM) to fluorescent Resorufin Red. SOD (superoxide dismutase; Sigma) was added at 30 units/ml, so as to convert all O2• into H2O2, a necessity since O2• reacts very rapidly with HRP and HRP–Compound (I) [Compound (I) is an intermediate state of peroxidases, formed by the reaction of the enzyme with H2O2]. We monitored Resorufin formation (Amplex™ Red oxidation by H2O2) at a λexcitation of 545 nm and a λemission of 590 nm using a Fluoroskan-FL Ascent Type 374 multiwell plate reader (Labsystems, Helsinki, Finland). The slope is converted into the rate of H2O2 production with a standard curve. Fluorescence remained linear with H2O2 concentration from 0 to 2 μM. The assay was performed in 96-well plates at 100 μl per well with a measuring duration of 20 ms, every 2 s for ∼10 min. We performed all assays at 37 °C, in 125 mM KCl, 10 mM Hepes, 5 mM MgCl2 and 2 mM K2HPO4 (pH 7.44), with 50–20 μg of mitochondrial protein per 100 μl of reaction buffer .
Measurement of the rates of ATP synthesis
ATP synthesis was measured using the luciferin/luciferase assay from Roche. We followed luciferin chemiluminescence using a Fluoroskan-FL Ascent type 374 multiwell plate reader (Labsystems). Mitochondria (between 10 and 5 μg) were incubated in 100 μl of 125 mM KCl, 10 mM Hepes, 5 mM MgCl2 and 2 mM K2HPO4 (pH 7.44) with substrates (glutamate, malate and succinate) and the measurement was started by the addition of luciferin/luciferase buffer containing 0.6 mM ADP. The initial slope was converted into nanomoles of ATP using the standards provided in the kit.
Measurement of the membrane potential (ΔΨ) with Safranin O
Membrane potential was monitored by fluorescence of the quench-dye Safranin O, as described by Votyakova and Reynolds , based originally on the spectroscopic method of Akerman and Wikstrom . We followed Safranin O fluorescence at a λex of 485 nm, a λem of 590 nm using a Fluroskan-FL Ascent Type 374 multiwell plate reader (Labsystems). Safranin O (5 μM) was incubated with 10–5 μg of mitochondrial protein in 100 μl of reaction buffer in 96-well black plates. We performed all assays at 37 °C, in 125 mM KCl, 10 mM Hepes, 5 mM MgCl2 and 2 mM K2HPO4 (pH 7.44).
High but variable rates of superoxide production with a combination of succinate and glutamate/malate
The vast majority of studies investigating the mechanisms of superoxide production by the electron transport chain have employed either complex I- or complex II-linked substrates, not both together. To better understand this phenomenon, we asked what would happen when electrons enter the chain through both complexes.
Results in Table 1 show the rates of superoxide production (H2O2 release) in isolated skeletal-muscle mitochondria respiring on glutamate/malate, succinate and succinate+glutamate/malate. In agreement with previous studies [7–9,14], we find that H2O2 release with glutamate+malate is low (∼30 pmol of H2O2/min per mg of mitochondrial protein) as compared with that obtained with succinate (∼400 pmol of H2O2·min−1·mg−1); we note though that the rates of H2O2 release with succinate are low as compared with the literature values (discussed below). The rate of H2O2 release with succinate+glutamate+malate was much higher than that with succinate alone, comparable with that observed in mitochondria inhibited with antimycin A (which is known to dramatically increase superoxide production). We note though that both the rates of H2O2 release with succinate alone (range: ∼30–900 pmol of H2O2·min−1·mg−1) and succinate+glutamate+ malate (range: ∼60–2600 pmol of H2O2·min−1·mg−1) were highly variable (in contrast with that observed with glutamate+ malate alone). In other words, the rates of superoxide production with succinate+glutamate+malate were sometimes as low as those observed with glutamate/malate alone, although on average, they were much higher than with succinate. We next investigated the source of this variability.
BSA stimulates superoxide production by succinate-reverse-electron transfer, but is inhibitory to superoxide production with other substrate–inhibitor combinations
In our laboratory, the rates of succinate-driven reverse-electron transfer-dependent superoxide production in skeletal-muscle mitochondria are highly variable and generally on the low end (e.g. ) of the literature values, although still higher than those with glutamate/malate. We identified the presence or absence of BSA as one source of the variability in the rates of superoxide production with succinate. In our protocol, BSA is added in the isolation procedure, but not in the reaction buffer. Some previous studies have added while others have omitted BSA in H2O2 measurement media [7–9,12,13,38]. We discovered that omitting the wash step after the addition of BSA in the isolation procedure yielded mitochondria with consistently high rates of superoxide production with succinate. We investigated the effect of BSA on superoxide production by adding different concentrations directly to the assay medium. Results in Figure 1 show that superoxide production with succinate alone is considerably increased by the addition of BSA (from ∼350 to 900 pmol of H2O2·min−1·mg−1) at low concentration (0.05%, w/v), although high concentrations were found to be inhibitory (>0.3%, w/v). Results also show that superoxide production with succinate+rotenone is essentially unchanged, while surprisingly, that with succinate+rotenone+antimycin A was dose-dependently decreased (results not shown). This was not due to BSA removing the inhibitor, because even at high concentrations of BSA (0.6%), antimycin A and rotenone still completely inhibited membrane potential formation (Safranin O) and ATP synthesis (results not shown). We performed these experiments with BSA purchased from Sigma (essentially fatty acid-free, fraction V), as well as extra-purified BSA (, a gift from Dr Paul S. Brookes (Mitochondrial Research and Innovation Group, University of Rochester Medical Center, Rochester, NY, U.S.A.), with essentially indistinguishable results (results not shown).
BSA stimulates the rate of superoxide production with succinate
Glutamate consistently stimulates while malate consistently inhibits succinate-supported superoxide production
To resolve the variability of the rates of H2O2 production with glutamate+malate+succinate, we dissected the experiment into its components and measured the rates of H2O2 release with glutamate+succinate and malate+succinate. The results in Table 2 show that H2O2 release with glutamate+succinate is consistently high (∼1000 pmol of H2O2·min−1·mg−1) and with malate+succinate is consistently low (∼20 pmol of H2O2·min−1·mg−1). This was true whether BSA was present or not, although BSA further increased the rate of H2O2 release with glutamate+succinate (from ∼1000 to ∼1300 pmol of H2O2·min−1·mg−1; Table 2). We constructed titration curves to show the effects of succinate and glutamate or malate on the rates of H2O2 release. The results in Figure 2 indicate that contrary to what is observed with succinate alone, the rates of H2O2 release with glutamate+succinate remain high even at substrate concentrations below 1 mM. In fact, greater relative stimulation of H2O2 release by glutamate was observed when succinate concentrations were low, especially in the absence of BSA (Figure 2). For example, at 1 mM succinate (in the absence of added BSA), the rates of H2O2 release were <30 pmol of H2O2·min−1·mg−1 with succinate alone, but >400 pmol of H2O2·min−1·mg−1 with the further addition of either 0.5 or 5 mM glutamate (>11-fold increase). At 10 mM succinate, adding glutamate increased H2O2 release from ∼300 to 950 pmol of H2O2·min−1·mg−1 (∼3-fold increase). The effect of the interaction of glutamate and succinate on superoxide production was ‘saturable’: at a succinate concentration of 25 mM, adding glutamate did not further increase H2O2 release in the presence of BSA; at the same time, increasing the concentration of glutamate to 25 mM did not result in any further changes in superoxide production with succinate (results not shown). Just as the stimulatory effect of glutamate was more pronounced at lower concentrations of succinate, so too was the inhibitory effect of malate. The inhibition of succinate-supported superoxide production by malate was virtually complete at concentrations of succinate below 5 mM (Figure 3); there appears to be a competitive effect between these two substrates. When succinate concentration was raised to 25 mM, the inhibitory effect of malate was alleviated (results not shown).
|Substrate(s)||BSA (%)||d[H2O2]/dt (pmol·min−1·mg−1)|
|Substrate(s)||BSA (%)||d[H2O2]/dt (pmol·min−1·mg−1)|
Superoxide production with succinate+glutamate is observed even at low substrate concentrations
Inhibition of superoxide production by malate and oxaloacetate with varying concentrations of succinate
In order to determine why glutamate was stimulatory, while malate was inhibitory, we asked how other complex I-linked substrates would affect the rate of H2O2 release with succinate. We found that pyruvate and α-ketoglutarate, as well as palmitoyl carnitine, stimulated superoxide production with succinate, to the same extent as glutamate (results not shown). We hypothesized that the inhibitory effect of malate on H2O2 release with succinate was due to the formation of oxaloacetate, a highly potent competitive inhibitor of complex II. Indeed, as results in Figure 3 show, oxaloacetate was even more effective than malate (on a concentration basis), at inhibiting H2O2 release with succinate. When employing oxaloacetate, we took precautions to minimize the time on ice, because oxaloacetate decays to pyruvate (P.S. Brookes, personal communication). We noticed that the inhibitory potency of oxaloacetate became weakened as the aliquots were kept on ice (results not shown).
Superoxide production with succinate+glutamate is dependent on membrane potential and is decreased by rotenone
To identify the site of electron leakage responsible for the high rates of superoxide production with glutamate+succinate, we asked how rotenone and FCCP, two agents that dramatically decrease reverse-electron transfer [7–9,12,13], would affect the rate of superoxide production with glutamate+succinate. Results in Figure 4 (compare white and black bars) show that just as for succinate alone, superoxide production with glutamate+succinate is completely eliminated by the addition of the uncoupler FCCP, indicating complete dependence on protonmotive force (this was the case whether BSA was present or not). Results in Figure 4 also show that superoxide production with glutamate+succinate is decreased by ∼20% by the addition of rotenone in the absence and by ∼60% decrease in the presence of BSA. The reason for this apparently stronger rotenone inhibition in the presence of BSA is that the rates of superoxide production are higher with glutamate+succinate in the presence of BSA (as compared with glutamate+succinate without BSA). The fact that superoxide production with glutamate+succinate is not further decreased can be partly explained by the fact that superoxide production in the forward-electron transfer direction is greatly stimulated by rotenone [8,9,14], a process further stimulated by imposition of a protonmotive force (as would happen when succinate is added to mitochondria respiring on glutamate but inhibited with rotenone ). Indeed, the rates of superoxide production with glutamate+succinate+rotenone are noticeably decreased by the addition of FCCP.
Superoxide production with glutamate+succinate is dependent on protonmotive force and is decreased by rotenone
In the present study, we address a contentious and unresolved question in mitochondrial ROS metabolism: can reverse-electron transfer-mediated superoxide production, the strongest source of ROS in isolated mitochondria, actually occur under substrate conditions that are physiologically plausible [8,14,40–42]? The major arguments against are that supraphysiological concentrations of succinate are required  and that it putatively cannot occur when NAD+, the terminal electron acceptor during reverse-transfer, is not available , i.e. the in vivo concentrations of succinate are low and NAD is predominantly in the reduced state).
To address these outstanding issues, we have measured the rates of reverse-electron transfer-mediated superoxide production as a function of succinate concentration and in the presence or absence of NADH-linked substrates, in isolated skeletal-muscle mitochondria. We find that high rates of superoxide production can occur at low substrate concentrations and when the NAD+ pool is reduced. While the present study was in progress, a recent article by Zoccarato et al.  in brain mitochondria reached the same conclusion. The main advance, as compared with Zoccarato et al.'s  report, is that the present study shows that certain NAD-linked substrates actually stimulate succinate-supported superoxide production, while others are inhibitory and that this can be explained by oxaloacetate inhibition of succinate dehydrogenase. Further, our study provides evidence that superoxide production with NAD-linked substrates in combination with succinate is due to reverse-electron transfer through complex I.
High rates of superoxide production with glutamate+succinate are observed even at low substrate concentration and are due to reverse-electron transfer
Succinate-supported reverse-electron transfer is, according to most studies, the strongest source of superoxide in isolated mitochondria; in our hands, it is highly variable, and on average, is in the low range of previously published estimates . During the course of the present study, we discovered that the source of this variability and lower values was due to omission of BSA in the Amplex™ Red assay medium (Figure 1). A very recent report in brain mitochondria came to the same conclusion . BSA is typically added to respiratory experiments because it increases the respiratory control ratio, which is generally felt to be due to the removal of non-esterified (‘free’) fatty acids (which can act as uncouplers [44,45]). BSA can also act as a calcium buffer . It is known that succinate-reverse-electron transfer-mediated superoxide production is extremely sensitive to even small decreases in protonmotive force, much more so than other substrate/inhibitor conditions [7–9,47,48]. Tretter et al.  concluded that BSA increases superoxide production with succinate by hyperpolarizing protonmotive force; we also find that adding BSA increases the strength of the membrane potential as measured by Safranin O with succinate, although it has no dramatic effects with other substrates (results not shown). It is debatable whether the addition of BSA is warranted: one might argue that it restores respiratory control ratios to what they would be in vivo (presumably very tightly coupled mitochondria), or that it introduces an experimental artefact. Because non-esterified fatty acids are released from mitochondria as the preparation ‘ages’ on ice , we are inclined to accept the former proposition. Because we cannot resolve this issue and to make our results easier to compare with those in the literature, we performed all subsequent experiments both in the presence and in the absence of 0.05% fatty acid-free BSA.
Even with BSA present (although more evident in its absence), we find that, in agreement with the previous study by Hansford et al. , superoxide production with succinate exhibited a strong substrate concentration dependence, becoming essentially negligible below 1 mM succinate (Figure 2). While most previous studies have used concentrations of succinate between 10 and 5 mM succinate [8,9,49], the in vivo concentration of succinate has been reported to be <0.5 mM in heart  and <0.3 mM in brain . Studies in cell culture suggest levels as low as 120 μM intracellular succinate concentration . These results would indicate that succinate concentrations are far too low under normal in vivo conditions to allow for high rates of succinate-driven reverse-electron transfer-mediated superoxide production (but see below).
Because NAD+ is the terminal electron acceptor during reverse-electron transfer from succinate , it is thought that the presence of NADH would be inhibitory towards reverse-electron transfer-mediated superoxide production. We therefore measured the rates of superoxide production with both succinate and NADH-linked substrates. Unexpectedly, we discovered that the combination of both NADH-linked substrates (glutamate/malate) and succinate resulted in higher rates of superoxide production than with succinate alone. The rates were highly variable, raging from as little as 30 to over 2000 pmol of H2O2/min per mg of mitochondrial protein (Table 1). This extreme variability could be eliminated when the glutamate+malate+succinate combination was broken down to glutamate+succinate and malate+succinate, the former combination resulting in consistently high rates of superoxide production and the latter in consistently low rates (this held true whether BSA was present or not; Table 2). We thus suggest that the apparent variability of the succinate+glutamate+malate measurement (Table 1) is the result of competing actions of glutamate (stimulatory) and malate (inhibitory).
We performed concentration dependence measures of the rate of superoxide production with glutamate+succinate and succinate+malate. While the rate of superoxide production with succinate alone decreases dramatically at lower concentrations of substrate, that with a combination of glutamate (either 5 or 0.5 mM)+succinate remains considerable even at concentrations as low as 0.25 mM succinate (Figure 2). At 0.5 mM glutamate+1 mM succinate, the rates of superoxide production were as high or higher (depending on whether BSA was present or not) as those observed with 10 mM succinate alone (compare Table 2 and Figure 2). Just as the stimulatory effect of glutamate was more evident at a low concentration of succinate, the inhibitory effect of malate on the superoxide production with succinate+malate was also more pronounced at low concentrations of succinate (Figure 3 and see the Results section).
We are aware of only two other publications comparing the rates of superoxide production with succinate versus succinate+glutamate+malate in mitochondria. In Hansford et al.'s  report, the succinate+glutamate+malate experiment was performed only once (Table 1 in ), resulting in considerably lower superoxide production than with succinate alone. In a very recent publication, Zoccarato et al.  reported that moderate concentrations (<3 mM) of glutamate+malate only partially inhibited the rate of superoxide production with succinate in brain mitochondria (in experiments performed with BSA ), thus leading them to conclude that high rates of superoxide production could indeed occur under physiologically realistic substrate conditions. This inhibition was modest unless the concentration of succinate was <1 mM, very similar to our results with malate in Figure 3, showing that malate inhibition of succinate supporter superoxide production is much more pronounced at low concentrations of succinate. Because both Hansford et al.  and Zoccarato et al.  used a combination of glutamate+malate+succinate, rather than separating the experiment into the glutamate+succinate and malate+succinate components, they were unable to tease apart the competing effects of glutamate (stimulatory) and malate (inhibitory). As far as we are aware, the present study is the first to do so. Our data could explain the contrasting results of Hansford et al.  and Zoccarato et al. . While one could invoke organ differences (rat heart compared with rat brain mitochondria), Zoccarato et al.  indicated that heart mitochondria behaved similar to brain mitochondria. Zoccarato et al.  used BSA in their assay medium, while Hansford et al.  did not; whereas Zoccarato et al.  used a maximum of 3 mM glutamate/malate, Hansford et al.  used 5 mM. Our results indicate that the inhibitory effect of malate is dose-dependent and is more evident in the absence of BSA than in its presence. Regarding the question of whether high rates of superoxide production can occur under physiologically relevant substrate conditions, we must point out that contrary to Zoccarato et al.'s  results, who showed that glutamate+malate exerted stronger inhibitory effects on superoxide production at lower (i.e. more physiological) succinate concentrations, our results demonstrate that glutamate added in combination with succinate provided stronger relative stimulation at lower succinate concentrations.
We next queried the origin of superoxide produced in the presence of both NADH-linked and complex II-linked substrates (i.e. with glutamate+succinate). In Zoccarato et al.'s  study, it was demonstrated that glutamate oxidation was still proceeding (albeit at a depressed rate) when glutamate+malate+succinate were added together, indicating that forward-electron transfer was taking place, although the authors concluded that in all likelihood, forward- and reverse-electron transfers were occurring concomitantly. Zoccarato et al.  also demonstrated that addition of nigericin (which collapses ΔpH) partially decreases H2O2 release with glutamate+malate+succinate, which is consistent with reverse-electron transfer being responsible for superoxide production (although not conclusive, because forward-electron transfer-mediated superoxide production can also be partially decreased by nigericin [7,14]). No further attempt was made to identify the site of superoxide production.
We found that as with succinate alone, glutamate+succinate supported superoxide production could be essentially eliminated by the addition of FCCP, indicating complete dependence on protonmotive force (compare glutamate+succinate in black and white bars of Figure 4). In this regard, superoxide production with glutamate+succinate behaves very similarly to succinate-mediated reverse-electron transfer ([8,9,47] and see Figure 4). Reverse-electron transfer can be inhibited to a large extent by the addition of rotenone, which prevents ubiquinol reduction of complex I [9,10]. We found that rotenone can indeed decrease the rate of superoxide production with glutamate+succinate (Figure 4). This effect was modest in the absence of BSA (∼20% decrease), but could inhibit the rate of superoxide production with glutamate+succinate by over 60% in the presence of BSA. This was due to the fact that BSA yielded higher rates of superoxide production with glutamate+succinate (although this stimulatory effect was less pronounced as compared with what we found with succinate alone, Figure 2) and led to lower rates of superoxide production with glutamate+succinate+rotenone, as compared with experiments performed in the absence of BSA. Why BSA would decrease the rate of superoxide production with glutamate+succinate+rotenone is at present unclear; we consistently observed this effect, more so with higher concentrations of BSA (results not shown). Regardless, the decrease in superoxide production with glutamate+succinate+rotenone is all the more noteworthy considering that rotenone stimulates superoxide production when electrons are added via NADH [8,9,12–15,53–55]. If one adds to that the rate of superoxide production with succinate+rotenone alone, the additive effects of the rates of superoxide production with glutamate+rotenone and rotenone+succinate can account for ∼60% of the rate of superoxide production with glutamate+succinate+rotenone (Figure 4). There appears to be a synergistic effect between superoxide production with succinate+rotenone and glutamate+rotenone (i.e. the actual measurement is higher than the sum of its individual components): this is probably the result of imposing a protonmotive force on rotenone-inhibited complex I, probably caused by the same underlying phenomenon as the observation by Lambert and Brand  that adding ATP to mitochondria treated with glutamate/malate+rotenone results in a further increase in superoxide production (by imposing a protonmotive force by hydrolysis of ATP ). Consistent with this idea, while the rate of superoxide formation with glutamate+rotenone was unaffected by the addition of FCCP, a large fraction of the rate of superoxide production with glutamate+succinate+rotenone was in fact dependent on protonmotive force (Figure 4). Based on these considerations, we argue that the 20–60% inhibition by rotenone of the rate of superoxide production with glutamate+succinate is a minimum estimate of the fraction that is due to reverse-electron transfer, and we suggest that the majority probably originates from this source.
Taken together, these results indicate that high rates of reverse-electron transfer-mediated superoxide production can occur in situations that cause even a moderate increase in succinate concentrations (even if NAD is in the reduced state). One might expect elevations in succinate concentrations during ischaemia, which could then result in high rates of reverse-electron transfer-mediated superoxide production during reperfusion. In support of this idea, administration of rotenone, which prevents reverse-electron transfer, is actually protective in mouse models of ischaemia/reperfusion injury [56,57].
The divergent effects of glutamate and malate on succinate-supported superoxide production can be explained by oxaloacetate inhibition of complex II
In addition to glutamate, other NADH-linked substrates, pyruvate, α-ketoglutarate and palmitoyl carnitine, stimulated superoxide production in combination with succinate (to a similar extent as glutamate), while citrate and isocitrate left the rate of superoxide production in combination with succinate unchanged (results not shown). Thus only malate was inhibitory towards succinate-supported superoxide production.
We propose that these contrasting effects on superoxide production can be explained by ‘relieving’ the well-known inhibition of complex II by oxaloacetate [27,28]. Succinate alone as a substrate only supports weak state 3 respiration rates  because of oxaloacetate formation, and inhibition, during continuous succinate oxidation. The addition of rotenone is necessary not only to prevent reverse-electron transfer (which would reduce P/O ratios), but also to block oxaloacetate formation: rotenone blocks oxaloacetate formation indirectly by locking the NAD pool in the reduced state, thereby inhibiting the NAD+-dependent oxidation of malate to oxaloacetate .
Why malate inhibits succinate-supported superoxide production becomes evident when considering that malate is converted into oxaloacetate by malate dehydrogenase, which accumulates unless acetyl-CoA is available. In direct support of the hypothesis that malate conversion into oxaloacetate is responsible for inhibiting succinate reverse-electron transfer superoxide production, we found that direct addition of oxaloacetate was indeed profoundly inhibitory to superoxide production with succinate (Figure 3). That oxaloacetate is a potent competitive (competitive with succinate) inhibitor of complex II/succinate dehydrogenase has been known since the 1970s, although the physiological significance of this inhibition has so far remained unclear . Based on the above results, we speculate that oxaloacetate inhibition of complex II is a deliberate mechanism evolved to minimize reverse-electron transfer-mediated superoxide production.
In this hypothesis, the stimulatory effect of pyruvate, glutamate, α-ketoglutarate and palmitoyl carnitine can be explained by the fact that, one way or another, their immediate metabolism leads to the consumption of oxaloacetate, and hence to a relief of the inhibition of succinate dehydrogenase (allowing succinate-driven reverse-electron transfer). In the case of pyruvate, oxaloacetate is consumed by citrate synthase concomitantly with acetyl-CoA generated by pyruvate dehydrogenase [44,58]. Glutamate leads to the removal of oxaloacetate because glutamate and oxaloacetate are co-substrates for transaminases yielding aspartate and α-ketoglutarate [44,58]. The α-ketoglutarate carrier exchanges malate for α-ketoglutarate , and palmitoyl carnitine generates acetyl-CoA, which is again used by citrate synthase to consume oxaloacetate [44,58]. A detailed summary of the metabolism of these classical mitochondrial substrates has been compiled by E. Gnaiger (Oroboros protocols at http://www.oroboros.at).
In summary, we have found that skeletal-muscle mitochondria respiring on both complex I- and complex II-linked substrates show high rates of reverse-electron transfer-mediated superoxide production even when substrate concentrations are low. These results counter some of the major arguments against the possibility that superoxide production by this pathway can occur in vivo. The results also suggest a physiological role for the long known phenomenon of oxaloacetate inhibition of complex II, namely the minimization of reverse-electron transfer-mediated superoxide production. The finding that reverse-electron transfer-mediated superoxide production can still occur in the presence of NAD-linked substrates has important implications for the mechanism of superoxide production by complex I as well as the mechanism of its normal catalytic turnover. The minimal interpretation of these results, we feel, is that the pathways of electron transfer in the forward and reverse direction are different and branched at one point or another.
We acknowledge the many useful discussions with Dr Paul S. Brookes and Dr Adrian Lambert (Medical Research Council Dunn Human Nutrition Unit, Cambridge, U.K.). We also thank Dr Sergey I. Dikalov (Free Radicals in Medicine Core, Division of Cardiology, Emory University School of Medicine, Atlanta, GA, U.S.A.) for directing our attention to the phenomenon of oxaloacetate inhibition. Financial support was provided by NIA (National Institute of Aging) training grant no. 5T3-AG021890-02 (F. L. M.), NIH (National Institutes of Health) grants P01AG19316 and P01AG020591 and the San Antonio Nathan Shock Aging Center (1P30-AG13319). We also thank Dr Arlan Richardson (Director, Barshop Institute for Longevity and Aging Studies) for constant encouragement and support.