Address correspondence and reprint requests to Ian J. Reynolds, Department of Pharmacology, University of Pittsburgh, W1351 Biomedical Science Tower, Pittsburgh PA 15261, USA. E-mail: email@example.com
Mitochondria are widely believed to be the source of reactive oxygen species (ROS) in a number of neurodegenerative disease states. However, conditions associated with neuronal injury are accompanied by other alterations in mitochondrial physiology, including profound changes in the mitochondrial membrane potential ΔΨm. In this study we have investigated the effects of ΔΨm on ROS production by rat brain mitochondria using the fluorescent peroxidase substrates scopoletin and Amplex red. The highest rates of mitochondrial ROS generation were observed while mitochondria were respiring on the complex II substrate succinate. Under this condition, the majority of the ROS signal was derived from reverse electron transport to complex I, because it was inhibited by rotenone. This mode of ROS generation is very sensitive to depolarization of ΔΨm, and even the depolarization associated with ATP generation was sufficient to inhibit ROS production. Mitochondria respiring on the complex I substrates, glutamate and malate, produce very little ROS until complex I is inhibited with rotenone, which is also consistent with complex I being the major site of ROS generation. This mode of oxidant production is insensitive to changes in ΔΨm. With both substrates, ubiquinone-derived ROS can be detected, but they represent a more minor component of the overall oxidant signal. These studies demonstrate that rat brain mitochondria can be effective producers of ROS. However, the optimal conditions for ROS generation require either a hyperpolarized membrane potential or a substantial level of complex I inhibition.
In addition to their critical role in ATP synthesis, mitochondria are also the major source of reactive oxygen species (ROS) in most cell types. Generated by the incomplete reduction of molecular oxygen during the process of oxidative phosphorylation, superoxide is the main form of ROS produced by mitochondria (Sorgato et al. 1974; Boveris and Cadenas 1975). It has been suggested that 2% of the oxygen consumed by mitochondria is converted to superoxide (Boveris and Chance 1973). In turn, superoxide is converted by manganese superoxide dismutase to H2O2, which is more stable and more lipid soluble, and thus can be more readily released by mitochondria. Although it is possible that mitochondrially derived ROS serve a signaling function in cells, it is more widely believed that ROS are harmful as the result of the oxidative modification of proteins, nucleic acids and lipid membranes.
The mechanisms responsible for the generation of ROS by the electron transport chain have been extensively investigated, primarily in mitochondria derived from heart muscle (Loschen et al. 1971; Boveris and Chance 1973; Cadenas and Boveris 1980; Turrens et al. 1985; Korshunov et al. 1997). The principal source appears to be the redox cycling ubiquinone in complex III (Boveris et al. 1976; Cadenas et al. 1977). An additional source of superoxide is complex I, which is also endowed with a number of redox centers (Cadenas et al. 1977; Takeshige and Minakami 1979; Turrens and Boveris 1980). In isolated mitochondrial preparations it is possible to demonstrate ROS generation from these two major sites when mitochondria respire on substrates that drive either complex I (glutamate and malate) or complex II (succinate). It is interesting to note that the relative magnitude of the ROS signal deriving from either substrate varies quite considerably between tissues. For example, in heart mitochondria complex I substrates generate the largest ROS signal (in the presence of appropriate inhibitors) (Boveris and Chance 1973), while in brain mitochondria complex II substrates appear to be quantitatively more important (Cino and Del Maestro 1989). However, the specific, endogenous mechanisms responsible for altering ROS production by mitochondria are poorly understood.
It is widely believed that ROS contribute to the pathogenesis of a number of neurodegenerative diseases (Halliwell 1992; Beal et al. 1997), and given the high rate of oxygen consumption by the brain, it may be reasonable to assume that mitochondria are responsible for the majority of the ROS burden under both normal and pathophysiological situations. However, the mechanisms by which brain mitochondria produce ROS have been investigated less than in other tissues. Indeed, the earliest investigation of this topic concluded that brain mitochondria do not produce ROS (Sorgato et al. 1974). More recent studies have conclude that peroxide production can be detected from brain mitochondria from several species, and that the ROS derive from both complex I and ubiquinone (Patole et al. 1986; Zoccarato et al. 1988; Cino and Del Maestro 1989). Several studies in intact neurons have used oxidation-sensitive fluorescent dyes to detect ROS generation following glutamate receptor activation, and have concluded that mitochondria are the likely source of the ROS signal (Reynolds and Hastings 1995; Dugan et al. 1995; Bindokas et al. 1996). This suggests that glutamate- and calcium-mediated alterations in mitochondrial function might contribute to the acute injury of neurons, as well as to chronic neurodegenerative states. However, there is a major gap in the understanding of the mechanisms that link glutamate receptor activation to the alteration of mitochondrial ROS production.
We undertook the present study to characterize the mechanisms responsible for ROS production by brain mitochondria using the peroxidase substrates scopoletin and Amplex red to detect peroxide production by mitochondria. In particular, we have investigated the influence of the mitochondrial membrane potential on ROS production because this has been reported to modify certain forms of ROS production by mitochondria in some tissues (Loschen et al. 1971; Korshunov et al. 1997), and may also be a key variable in the injury of neurons by glutamate (Nieminen et al. 1996; Schinder et al. 1996; White and Reynolds 1996; Vergun et al. 1999). We report here that there are mitochondrial membrane potential dependent and independent mechanisms for the generation of ROS by isolated rat brain mitochondria.
Materials and methods
Isolation of rat brain mitochondria
All procedures using rats were approved by the Institutional Animal Care and Use Committee of the University of Pittsburgh, and are consistent with guidelines provided by the National Institutes of Health. Rat brain mitochondria were isolated from the cortex of adult Sprague–Dawley strain rats by conventional differential centrifugation as described by Rosenthal et al. (1987) with minor modifications. After removal, brains were placed in isolation media which contained: 225 mm mannitol, 75 mm sucrose, 5 mm HEPES buffer (pH adjusted to 7.3 with KOH), 1 mg/mL BSA and 0.5 mm tetrapotassium EDTA. After centrifugation at 10 500 g, the first mitochondrial pellet was treated with digitonin (40 : l of 10% w/v digitonin solution per brain) with a final concentration of 0.013%. Mitochondria were then resuspended and centrifuged in isolation medium without digitonin. All the isolation procedures were carried out at 0 ± 2°C. Prior to experimentation, mitochondria were stored on ice at final concentration of 20–30 mg protein/mL in isolation medium. The protein concentration was determined by the Bradford method (Bradford 1976). Mitochondria prepared in this way were active for 5–6 h, as determined by their ability to maintain a transmembrane potential in the presence of oxidizable substrates.
Fluorescence measurements of H2O2 production and mitochondrial transmembrane potential ΔΨm
Fluorescence measurements were performed in a Shimadzu RF5301 spectrofluorimeter in a stirred cuvette maintained at 37°C. Mitochondria were added to a standard incubation buffer that contained: 125 mm KCl; 2 mm K2HPO4; 5 mm MgCl2; 10 mm HEPES (pH adjusted to 7.0 with KOH); 10:M EGTA; and 0.2 mg/mL mitochondrial protein. Substrates (5 mm of succinate or 5 mm of glutamate plus 5 mm malate) and adenine nucleotides were added separately as indicated on figures. Hydrogen peroxide was measured using either scopoletin or Amplex red in the presence of 1 U/mL of horseradish peroxidase (HRP). The concentration of both dyes was 2 µm. Measurements were carried out at excitation/emission wavelengths of 365 nm (slit 3 nm)/460 nm (slit 5 nm) for scopoletin and 560 (slit 1.5 nm)/590 (slit 3 nm) for Amplex red, respectively. Since hydrogen peroxide was measured by a decrease of fluorescence in the case of scopoletin, this method is less sensitive to low concentrations of H2O2 and may underestimate slow rates of release by mitochondria. On the other hand, the advantage of this method is the opportunity to follow long-lasting reactions by addition of a second aliquot of dye upon the complete oxidation of the initial dye addition. In the case of Amplex red, fluorescence is increased with the generation of H2O2. This method is more sensitive to low concentrations of H2O2 (Mohanty et al. 1997; Zhou and Panchuk-Voloshina 1997). We tested both Amplex red and its oxidation product, resorufin, on mitochondrial function using polarography and found that neither agent at 4 µm had any effect on the respiratory control ratio with either glutamate and malate or succinate as substrate. Note also that the addition of superoxide dismutase to the assay medium did not further increase the signal with either scopoletin or Amplex red (data not shown) suggesting that the endogenous dismutase capacity was sufficient to metabolize all of the superoxide produced in this system. The signals detected by both scopoletin and Amplex red were decreased by 85% by the inclusion of 800 U/mL catalase in the reaction mixture. ΔΨm was estimated using fluorescence quenching of the cationic dye safranine O which is accumulated and quenched inside energized mitochondria (Akerman and Wikstrom 1976). The excitation wavelength was 495 nm (slit 3 nm) and emission 586 nm (slit 5 nm), and the dye concentration used was 2.5 µm. We attempted to calibrate the safranine O signal using media with different KCl concentrations and valinomycin. Using rat liver mitochondria we obtained a good linear correlation between values calculated from the Nernst equation and measured changes in dye fluorescence within the range of 48–150 mV (data not shown). However, using similar conditions with rat brain mitochondria we were not able to record significant changes in safranine O fluorescence in the calibration buffer because the mitochondria appeared to spontaneously depolarize. This could be due to depletion of potassium from the matrix during the tissue isolation process (in potassium free buffer), or it is also possible that brain mitochondria could be more sensitive to valinomycin-induced swelling. However, this limitation prevented precise calibration of the safranine O signal, and instead the results for this dye are reported in fluorescence units or as a percent of maximal values.
H2O2 generation was calibrated by constructing standard curves using known H2O2 concentrations in the presence of the standard incubation buffer, the appropriate dye and horseradish peroxidase but without mitochondria. In this study we show either traces reflecting the fluorescence values measured during the experiment or rates determined from the slope of the fluorescence change and converted to H2O2/min/mg protein. Statistical analysis was performed using Prism (GraphPad Software, San Diego, CA, USA). Typically data were subjected to an anova followed by an appropriate post hoc test. Differences were considered significant when p < 0.05.
Amplex red was obtained from Molecular Probes (Eugene, OR, USA). Scopoletin, safranine O, and HRP were obtained from Sigma (St Louis, MO, USA). Myxothiazol was purchased from Fluka (Buchs, Switzerland). All other reagents and inhibitors were purchased from Sigma.
ROS production by mitochondria metabolizing succinate
We first investigated the characteristics of mitochondrial ROS production in rat brain mitochondria incubated with the complex II substrate succinate (Fig. 1). In the absence of succinate the rate of oxidation of scopoletin was very low. However, mitochondria aerobically incubated in media with succinate begin to release hydrogen peroxide after a short lag period of some 30 s. This lag period coincides with the time necessary for mitochondria to gain their maximal value of ΔΨm(Fig. 1a). Interestingly, the ROS generation supported by succinate is highly sensitive to small changes in ΔΨm(Fig. 1b). To determine the threshold of ΔΨm loss necessary to block ROS production, mitochondria were titrated with uncoupler in small increments. Figure 1(b) shows that 15 nm of carbonyl cyanide p-(trifluoromethoxy) phenyl hydrazone (FCCP) caused 50% inhibition of ROS production while producing very small changes in ΔΨm. The minimum concentration of uncoupler that fully blocks the ROS generation was found to be 40–50 nm, which caused decreased the safranine O signal by only 2–3%, while a complete depolarization was observed at 80–100 nm depending on mitochondrial preparation (Fig. 1c). A similar result was obtained with the uncoupler 2,4-dinitrophenol (2,4-DNP) which is known to have a different uncoupling mechanism (Skulachev 1998). The ROS generation was inhibited by 95% at 4 µm of 2,4-DNP while the corresponding decrease of ΔΨm was as little as 5% (data not shown). It is important to note that the measurements of membrane potential with safranine O may underestimate hyperpolarized values of ΔΨm which, in turn, might underestimate the magnitude of the depolarization in the presence of low uncoupler concentrations (Akerman and Wikstrom 1976). The concentration of uncoupler that abruptly and completely depolarizes mitochondria varies for different mitochondria preparations, and this is reflected at greater standard error at the end of the graph for membrane potential (Fig. 1c). Nevertheless, these data clearly indicate that succinate driven ROS production is very sensitive to changes in ΔΨm.
The physiological relevance of small depolarizations of ΔΨm are illustrated in Fig. 2. Providing ADP together with succinate results in phosphorylating, or state 3 respiration. The synthesis of ATP depletes ΔΨm, and this depletion is sufficient to decrease ΔΨm to the point that ROS generation is fully inhibited. The subsequent addition of oligomycin to prevent oxidative phosphorylation results in hyperpolarization of ΔΨm and the resumption of ROS production. This principle can also be illustrated by the addition of a limited concentration of ADP (40 µm). This completely stopped ROS generation concurrently with a depolarization of mitochondria by some 30–40% of the safranine O signal (Fig. 2b). After the ADP was fully consumed and ΔΨm recovered to a maximum value, the ROS production resumes. When ADP was present in excess and mitochondria are permanently in state 3, no ROS production was observed up to 30 min of incubation (Fig. 2a and data not shown). A low concentration of oligomycin (100 nm) inhibited the ATPase and restored ΔΨm and then re-initiated ROS generation when the membrane potential reached its maximal value. These data show the reversibility of the process of free radical generation supported by succinate. It is worth noting that the changes of ΔΨm at the transition from state 4 to state 3 (ATP synthesis switched off and switched on, respectively) measured by safranine O are greater than the depolarization necessary for complete inhibition of ROS production by an uncoupler. This suggests that brain mitochondria with an uninhibited respiratory chain engaged in ATP synthesis are protected from ROS generation mediated by complex II substrates.
Mechanism of ROS generation mediated by succinate
Previous studies have shown that in mitochondria derived mainly from tissues other than brain, ROS originate in two sites of the respiratory chain. Superoxide can be generated from complex I as a result of reverse electron transfer at high membrane potential values (Hinkle et al. 1967) (Croteau et al. 1997) (Korshunov et al. 1998), and also from the Q-cycle of complex III (see Turrens 1997). We used a series of selective inhibitors to evaluate the site(s) of the respiratory chain of rat brain mitochondria responsible for succinate supported ROS production.
Addition of the complex I inhibitor rotenone to mitochondria oxidizing succinate decreased ROS production (Fig. 3, Table 1). ΔΨm remains high in this case, because proton pumping can still occur at complexes III and IV (data not shown). This suggests that the principle effect of rotenone is to block reverse electron transfer from complex II to complex I, and that this pathway is the major route of succinate-driven ROS generation. The complex III inhibitor antimycin also substantially inhibited ROS generation (Fig. 3), and also rapidly decreased ΔΨm. However, it is evident from Fig. 3 that there is a smaller and delayed oxidation of scopoletin that occurs following the addition of rotenone or antimycin. This remaining ROS production is likely to be due to free radical formation in the Q-cycle. Myxothiazol was reported to inhibit ROS generation in Q-cycle of bovine heart submitochondrial particles (Turrens et al. 1985). In our experiments, the addition of myxothiazol alone inhibited ROS generation by 97–98%. Application of myxothiazol after rotenone or antimycin results in the same rate of ROS generation, supporting, supporting the suggestion of Q-cycle as a source of the residual ROS signal when reverse electron transport to complex I is blocked.
Table 1. Effect of electron transport inhibitors and uncoupler on ROS generation by isolated mitochondria
Rate of peroxide production (pmol/min/mg protein)
Glutamate + malate
Values are the mean ± SEM (N in parentheses) of determinations from different mitochondrial preparations. Peroxide production was determined using the scopoletin/horseradish peroxidase technique. Concentrations of drugs: antimycin 1 µm; rotenone 1 µm; myxothiazol 1 µm; FCCP 150 nm. Drugs were added together when indicated by +, and sequentially when indicated by ‘then’. –, condition not tested.
1388 ± 59 (25)
275 ± 18 (11)
228 ± 6 (20)
46 ± 2 (3)
59 ± 7 (4)
1266 ± 98 (3)
174 ± 18 (8)
434 ± 9 (19)
56 ± 18 (3)
40 ± 2 (6)
82 ± 11 (4)
450 ± 20 (5)
67 ± 7 (3)
FCCP then antimycin
1105 ± 130 (3)
254 ± 39 (3)
53 ± 12 (3)
85 ± 9 (3)
Antimycin then FCCP
463 ± 29 (6)
283 ± 8 (6)
79 ± 11 (3)
FCCP then rotenone
148 ± 7 (3)
Rotenone then FCCP
155 ± 15 (8)
Interestingly, when mitochondria are uncoupled with FCCP, antimycin stimulates Q-cycle ROS generation to an even greater extent. The rate of H2O2 generation is dependent on the ratio of [fumarate]/[succinate] in the mitochondrial matrix with maximum at 10 : 1 (Ksenzenko et al. 1984). Thus, building up fumarate as a result of succinate oxidation favors H2O2 production. For this reason, both the order and timing of additions of inhibitor and uncoupler is important. When inhibitor comes first the amount of fumarate accumulated is less compared to the case when mitochondria were first uncoupled. In our conditions, when succinate is present in excess, the later antimycin was added the higher rate of H2O2 was observed (not shown). This process was also substantially inhibited by myxothiazol (Fig. 3, Table 1).
ROS production supported by NADH-linked substrates
We next examined the characteristics of ROS generation by mitochondria utilizing glutamate and malate to drive NADH-linked respiration at complex I. The rate of endogenous hydrogen peroxide release was much lower when rat brain mitochondria oxidize NADH-linked substrates glutamate and malate and was undetectable using the scopoletin method (Fig. 4a, Table 1). However, the addition of rotenone, antimycin and even myxothiazol increased the rate of scopoletin oxidation (Fig. 4a, Fig. 5, Table 1).
The mechanism(s) of NADH-supported ROS generation
We further investigated the properties of ROS production triggered by the addition of inhibitors to mitochondria utilizing glutamate and malate. However, the short lag period observed between the addition of the inhibitor and the initiation of scopoletin oxidation added an ambiguity to these experiments, because this might reflect either a delay in ROS production or else a threshold below which scopoletin was unable to detect a ROS signal. To avoid this problem we detected ROS with the Amplex red method instead. We first investigated the concentration dependence of ROS generation triggered by rotenone. The aim of these experiments was to find out minimal degree of complex I inhibition which causes ROS generation in the respiratory chain and whether any correlation between ROS generation and membrane potential value existed (Fig. 6). Treatment of mitochondria with increasing rotenone concentrations resulted in a gradual increase in the rate of ROS generation (Fig. 6a). The minimal rotenone concentration causing noticeable hydrogen peroxide release was 20 nm (equal to 100 pmol rotenone per mg mitochondrial protein). ROS generation by rotenone-inhibited mitochondria was not linear and typically increased during the experiment. The initial rate of ROS production (Vi) gradually rose with increasing inhibitor concentration and showed saturation at 500 nm, while the maximal rate (Vm) reached a plateau at a lower concentration (200 nm) (Figs 6a and c). Complex I contains several types of redox centers including eight or nine Fe–S clusters, flavin mononucleotide (FMN) and the tightly bound ubiquinone pool (Onishi 1998). The accelerated rate of peroxide production may be due to the progressive reduction of the upstream redox groups upon complete block of electron flow at rotenone binding site. Alternatively, increase in rate may reflect the point at which the endogenous antioxidant systems become depleted.
The kinetics of depolarization upon administration of different rotenone concentrations (Fig. 6b) are biphasic. These two components are likely to be attributable to a combination of the removal of the proton-motive force upon inhibition by rotenone, and the non-ohmic inherent proton leak of the inner membrane because the remaining respiratory chain activity is no longer able to compensate this passive leak. These experiments do not allow us to determine which component is attributable to which phase of the response. The apparent similarity between the kinetics of ROS production and loss of membrane potential can be explained by the fact that both processes are dependent on the degree of inhibition at complex I. However, in this case, membrane potential does not control the process of free radical production (see below).
To determine the extent of respiratory chain inhibition necessary to start free radical generation we performed polarographic measurements of oxygen consumption (Fig. 6c). There is an interesting discrepancy between the concentration dependence of the inhibition of respiration compared to ROS generation. Twenty nanomolar rotenone inhibits respiration by some 50%, while minimally altering H2O2 release. In the presence of 50 nm of rotenone 90% of respiration is inhibited, whereas Vm of ROS generation is only half maximal. Likewise, rotenone at 200 nm inhibited respiration by 98%, and Vm also reached a maximum, although the initial rate of ROS generation, Vi, is still submaximal at this concentration.
The addition of rotenone to mitochondria clearly decreases ΔΨm in addition to increasing ROS generation. To determine whether ROS production was driven by the loss of ΔΨm we performed experiments with FCCP. Depolarization of mitochondria itself with FCCP does not cause ROS production (Fig. 4a, Table 1). Indeed, uncoupler FCCP partially inhibited ROS generation induced by rotenone. Similarly, addition of rotenone to mitochondria previously exposed to FCCP results in ROS generation at a rate comparable to the experiment with the reversed order of drug addition. These data show that dissipation of ΔΨm alone is insufficient to trigger ROS production. Additionally, in rotenone-induced ROS production with glutamate and malate as substrate there is an uncoupler sensitive component. The latter is consistent with previous reports using mitochondria derived from tissues other than brain (Vinogradov et al. 1995).
To further probe the source of ROS in rotenone-inhibited mitochondria we tested the effects of antimycin and myxothiazol. Addition of antimycin to rotenone-poisoned mitochondria (Fig. 4b) accelerates H2O2 release. However, this enhancement is quite transient, because within 1–2 min the rate returns to the same level as observed with rotenone alone. In addition, myxothiazol has minimal effects on ROS generation in rotenone-inhibited mitochondria (Fig. 4a). The transient effect of antimycin may be due to small amounts of succinate present in the system from the transamination of glutamate and oxaloacetate, giving α-ketoglutarate. As this source of succinate donating electrons to Q-cycle is limited (analogous to the succinate experiments shown in Fig. 3) the addition of antimycin causes only transient acceleration of ROS production before the succinate is consumed. The limited effects of myxothiazol in this case indicates that upon glutamate-malate oxidation ROS generation does not involve the ubiquinone pool from complex III.
We also examined the effects of antimycin alone on ROS generation supported by glutamate and malate. For these studies we used Amplex red to ensure that we detected the full range of the ROS signal. Where rotenone provided a clear concentration dependence for ROS generation and mitochondrial depolarization, antimycin essentially produced an all-or-none effect with the critical concentration being between 40 and 50 nm(Fig. 7). Like rotenone, the effects of antimycin on both ROS production and ΔΨm were only apparent when respiration had been almost completely inhibited (Fig. 7c). Increasing the antimycin concentration from 50 nm up to 1 µm did not further increases in the rate of ROS generation, though mitochondria lost membrane potential faster (Figs 7a and b). Similar data were obtained with scopoletin (not shown). The antimycin-induced ROS generation by mitochondria was almost completely inhibited by myxothiazol, and resumed after the subsequent addition of rotenone (Fig. 5, Table 1). The fact that myxothiazol suppressed the ROS production supported by glutamate and malate substrates in the presence of antimycin suggests that the hydrogen peroxide derives from Q-cycle under these conditions. However, in the presence of rotenone ROS production is likely to originate from complex I. Consistent with the involvement of the Q-cycle in antimycin-induced ROS production is the slight stimulatory effect of FCCP (Table 1).
Administration of myxothiazol to mitochondria oxidizing glutamate and malate caused a slow ROS production (Table 1). This is an interesting observation because myxothiazol is known to inhibit free radical production in Q-cycle (Turrens et al. 1985) and can be used to identify the Q-cycle as the source of ROS. A recent study also reported a slow ROS production in the presence of myxothiazol (Starkov and Fiskum 2001). Using another inhibitor of complex III, stigmatellin, the authors concluded that myxothiazol induces ROS in complex III, but at a site different from that of antimycin A action. In our experiments addition of rotenone to myxothiazol-poisoned mitochondria increased ROS production, though antimycin had no effect in this condition. The rates of ROS generation induced by rotenone alone and rotenone added to myxothiazol-poisoned mitochondria were the same indicating that myxothiazol block does not affect the ability of complex I to generate radicals.
In this study we have investigated the regulation of mitochondrial ROS generation by ΔΨm in a brain mitochondria preparation. There is clear evidence for ROS production originating from at least two different sites in the electron transport chain. In addition, mitochondria clearly produce ROS using both ΔΨm-dependent and -independent mechanisms.
Detailed information on the mechanism of ROS generation by respiratory chain in mitochondria and submitochondrial particles has been accumulated for tissues other than brain. In the majority of studies investigating intact mitochondria, it has been shown that oxidation of succinate gives the most effective production of ROS (Loschen et al. 1971; Boveris et al. 1972; Boveris and Chance 1973; Croteau et al. 1997; Korshunov et al. 1998; Kwong and Sohal 1998). Data on brain mitochondria are rather less, and have been summarized in Table 2. Our data along with data of Cino and Del Maestro (1989) and Kwong and Sohal (1998) clearly indicate succinate as the most effective ROS generating substrate for intact mitochondria, while others (Herero and Barja 1997; Barja and Herero 1998) failed to measure significant ROS production by rat mitochondria in the presence of this substrate. Arnaiz (Arnaiz et al. 1999) found that for mice brain mitochondria succinate was almost as effective as NADH-linked substrates. There are clearly several critical variables in these studies; the choice of incubation media is important, as well as the method of ROS detection. However, perhaps the most important factor is the ability of the mitochondria to maintain a sufficiently hyperpolarized membrane potential to support reversed electron transport. Some of the studies summarized in Table 2 used uncouplers or antimycin in their experiments, which would result in different results than those reported here.
Table 2. Prior studies on ROS generation by brain mitochondria
The substantial succinate driven ROS production observed in previous studies and in our experiments is likely to be the result of reversed electron transport from complex II to complex I. Reversed electron transport has been shown to be very sensitive to ΔΨm (Korshunov et al. 1998), and the present studies show that a very small degree of depolarization are sufficient to completely inhibit ROS generation by this pathway. Thus, low concentrations of the uncouplers FCCP and dinitrophenol, active oxidative phosphorylation, and the complex III inhibitors are each sufficient to depolarize ΔΨm to the extent that ROS production is greatly reduced in succinate-oxidizing mitochondria. The finding that rotenone blocks this ROS signal, even though it does not depolarize ΔΨm is also consistent with this conclusion, because it prevents electrons from reaching the redox centers associated with complex I responsible for superoxide generation. This means that the redox center(s) responsible for ROS production during reverse electron transport are located upstream of rotenone block. According to Hansford and colleagues (Croteau et al. 1997) all free radicals generated upon succinate oxidation originate in complex I.
Our data, along with experiments on rat heart mitochondria (Korshunov et al. 1998), clearly indicate that there is a secondary source of ROS in mitochondria respiring on succinate. This signal is likely to be derived from the ubiquinone cycle, because it is enhanced by antimycin and blocked by myxothiazol, and in the presence of antimycin is stimulated by the addition of uncouplers. This is consistent with the model of superoxide generation proposed in previous studies (Boveris et al. 1976; Cadenas et al. 1977; Cadenas and Boveris 1980; Turrens and Boveris 1980; Turrens et al. 1985). Notably, this pathway appears to be insensitive to changes in ΔΨm. Reverse electron transfer is a function that requires tightly coupled mitochondrial membranes capable of maintaining hyperpolarized potentials.
The contribution of complex III in overall ROS signal may depend on the concentration of CoQ in mitochondrial membrane, which is type and tissue specific. As Turrens has shown (Turrens et al. 1982), there is strong correlation between rates of H2O2 release and ubiquinone content in mitochondria of different sources. In addition, the impact of complex III and complex I in the overall ROS signal associated with succinate respiration depends on other factors, such as functional integrity of mitochondrial membrane and the potential, the redox state of the electron transport complexes under the appropriate experimental conditions, relative activity of the complexes themselves, and also the activity of ROS scavenging systems.
Rat brain mitochondria respiring on NADH-linked substrates produce a very small ROS signal in the absence of electron transport chain inhibitors. This is distinct from mitochondria derived from other species and tissues [mouse heart, kidney, brain (Kwong and Sohal 1998); pigeon heart (Boveris and Chance 1973); rat liver (Boveris et al. 1972); porcine lung (Turrens et al. 1982); and mouse brain (Arnaiz et al. 1999)], where the uninhibited, complex I-derived signal is some 50–70% of that obtained with succinate. The basis for these intertissue differences remains unclear.
Significant ROS generation is triggered by the addition of rotenone which promotes superoxide generation by the redox center(s) of complex I. Our data are in agreement with those of Takeshige and Minakami (1979) who concluded that rotenone induces ROS only from complex I in NADH oxidizing bovine heart submitochondrial particles. It has subsequently been demonstrated that there are redox centers in complex I that are capable of generating superoxide though the precise location of the site(s) has not been identified (Kang et al. 1983; Krishnamoorthy and Hinkle 1988; Onishi 1998; Vinogradov 1998). These centers are apparently upstream of the site of rotenone blockade, and ROS production at this site(s) would account for the lack of sensitivity of this signal to myxothiazol or antimycin beyond that which can be accounted for by succinate generation (Fig. 3).
When either antimycin or rotenone were present with complex I substrates, the ROS production appears to be largely independent of ΔΨm. Although the addition of these inhibitors substantially alters ΔΨm, the changes in membrane potential occur independently of ROS generation, as indicated by the failure of FCCP to trigger a ROS signal in mitochondria respiring on glutamate and malate (Fig. 6). Interestingly, in the presence of rotenone and FCCP, antimycin does have a more substantial accelerating effect on ROS generation (Fig. 6a) than the transient effect of antimycin in the presence of rotenone alone (Fig. 6b), although the mechanism underlying this effect is not clear. Free radical production supported by succinate in the presence of antimycin block has previously been shown to be dependent on the redox potential of the fumarate/succinate pair (Loschen et al. 1973; Ksenzenko et al. 1984). The dependence had a bell shape with maximum at about 40 mV which is close to Em of ubiquinones. These experiments were carried out on submitochondrial particles since fumarate poorly penetrates into mitochondria. Similar results were obtained for brain mitochondria, where dependence of ROS production in the presence of antimycin on the concentration of succinate had the same bell-shaped form (Patole et al. 1986; Zoccarato et al. 1988). Another confirmation of this notion was recently reported, where the problem of poor delivery of fumarate was overcome by using alamethicin-permeabilized mitochondria (Starkov and Fiskum 2001).
It is important to notice that both types of substrates we used in this work, succinate and glutamate with malate, cannot be considered ‘pure’ substrates of a particular complex. The oxidation of succinate in the Krebs cycle results in the production of fumarate and then malate, which can fuel complex I. In turn, some quantity of succinate is built up in the process of malate oxidation as well as from the transamination of glutamate. The transient effect of antimycin on rotenone treated mitochondria may be due to the to oxidation of succinate formed in this way.
It is widely believed that oxidative stress contributes to the demise of neurons in a number of neurodegenerative diseases, and it is likely that mitochondria are quantitatively the most important source of ROS in the brain. There have been numerous reports of impairment of the function of electron transport complexes in association with disorders like Parkinson's disease and Alzheimer's disease, conditions that are also associated with increases in markers of oxidative stress (Halliwell 1992; Beal et al. 1997). It is tempting to speculate that the partial inhibition of the electron transport chain reported in these diseases results in a feed forward effect, such that mitochondrial ROS generation is enhanced by the mechanisms evident when mitochondria respire on glutamate and malate. There are several potential limitations to this speculation, including the finding that very substantial inhibition of respiration is required prior to observing increased ROS generation, in other words the mitochondrial machinery may have a broad margin of safety. There is little evidence that the inhibition of respiration reaches this extent in any chronic neurological disorder. Perhaps a more reasonable speculation is that there may be conditions whereby mitochondria are excessively hyperpolarized under certain conditions, so that succinate-driven ROS generation occurs. This would not require extensive impairment of electron transport, but would imply that endogenous mechanisms that limit membrane potential [such as uncoupling proteins or endogeneous fatty acids, for example, Korshunov et al. (1998); Ricquier and Bouillaud (2000)] are dysfunctional. It is also possible that inhibition of ATP synthesis could generate a sufficiently hyperpolarized state to support succinate-driven ROS generation in intact tissues. Given the magnitude of the ROS signal from this mechanism and the liability this oxidant burden would impose on neurons, these are clearly issues which require further investigation.
One circumstance of mitochondrial involvement in neuronal injury that has received particular attention is excitotoxicity, where neurons die as the result of excessive activation of NMDA receptors and substantial cellular calcium accumulation. This process is associated with mitochondrial depolarization (Nieminen et al. 1996; Schinder et al. 1996; White and Reynolds 1996; Vergun et al. 1999), mitochondrial calcium accumulation (Kiedrowski and Costa 1995; White and Reynolds 1995; White and Reynolds 1997) and ROS generation (Lafon-Cazal et al. 1993; Dugan et al. 1995; Reynolds and Hastings 1995; Bindokas et al. 1996). Moreover, preventing mitochondrial calcium accumulation protects neurons from injury (Budd and Nicholls 1996; Stout et al. 1998). The mechanisms linking glutamate-induced mitochondria calcium accumulation to neuronal death remain unclear. One candidate mechanism is the generation of ROS. Previously, it has been suggested that mitochondrial calcium and sodium accumulation could promote ROS generation by isolated mitochondria (Dykens 1994). However, this paper reported a lack of ROS production in mitochondria respiring on succinate, and an increase in ROS generation under conditions where mitochondria would be depolarized following the addition of calcium. These claims are in direct contrast to the present findings, as well as those of Cino and Del Maestro (1989). The basis for this substantial discrepancy remains unclear. Indeed, it is not easy to account for the effects of glutamate or calcium on the basis of the results presented here. The depolarization of ΔΨm that is associated with mitochondria calcium cycling (Nicholls and Akerman 1982) is sufficient to completely inhibit succinate-driven ROS generation. In addition, the experiments reporting glutamate-induced ROS generation in intact neurons is obviously performed without the addition of complex I or complex III inhibitors. In fact, Dugan et al. (1995) reported that glutamate-induced oxidation of dihydrorhodamine was prevented by rotenone. There are rather few reports of mitochondrial ROS generation in intact neurons that can be compared to the present study, although Budd et al. (1997) did find that antimycin increased the oxidation of dihydroethidium in cerebellar granule cells. Perhaps the most reasonable way to account for the consequences of NMDA receptor activation in intact neurons is to propose a calcium-mediated inhibition of complex I or complex III. This might be accomplished by the production of nitric oxide, which is an effective inhibitor of complex III (Brown 1999; Stewart et al. 2000), although NOS inhibitors did not prevent the oxidation of either dichlorofluorescin or dihydroethidium (Reynolds and Hastings 1995; Bindokas et al. 1996). The actual mechanism of calcium-mediated mitochondrial ROS generation in neurons awaits further investigation.
This work was supported by USAMRMC Neurotoxin Initiative (grant number DAMD17–98–1-8627). We would like to thank Lauren Richards for help in some experiments, Dr Teresa Hastings for access to an oxygen electrode, Kirk Dineley for reading the manuscript and Dr Alexei Permin for helpful discussion of these findings.