The dependence of brain mitochondria reactive oxygen species production on oxygen level is linear, except when inhibited by antimycin A

Abstract

Reactive oxygen species (ROS) are byproducts of physiological mitochondrial metabolism that are involved in several cellular signaling pathways as well as tissue injury and pathophysiological processes, including brain ischemia-reperfusion injury. The mitochondrial respiratory chain is considered a major source of ROS; however, there is little agreement on how ROS release depends on oxygen concentration.

The rate of H₂O₂ release by intact brain mitochondria was measured with an Amplex UltraRed assay using a high-resolution respirometer (Oroboros) equipped with a fluorescent optical module and a system of controlled gas flow for varying the oxygen concentration. Three types of substrates were used: malate and pyruvate, succinate and glutamate, succinate alone or glycerol 3-phosphate. For the first time we determined that, with any substrate used in the absence of inhibitors, H₂O₂ release by respiring brain mitochondria is linearly dependent on the oxygen concentration. We found that the highest rate of H₂O₂ release occurs in conditions of reverse electron transfer when mitochondria oxidize succinate or glycerol 3-phosphate. H₂O₂ production by complex III is significant only in the presence of antimycin A and, in this case, the oxygen dependence manifested mixed (linear and hyperbolic) kinetics. We also demonstrated that complex II in brain mitochondria could contribute to ROS generation even in the absence of its substrate succinate when the quinone pool is reduced by glycerol 3-phosphate.

Our results underscore the critical importance of reverse electron transfer in the brain, where a significant amount of succinate can be accumulated during ischemia providing a backflow of electrons to complex I at the early stages of reperfusion. Our study also demonstrates that ROS generation in brain mitochondria is lower under hypoxic conditions than in normoxia.

Graphical Abstract

Oxidative stress plays an important role in ischemia/reperfusion injury when tissue oxygen level dramatically changes. We determined, that H₂O₂ release by respiring brain mitochondria depends linearly on oxygen concentration when substrates of complex I, II or glycerol 3-phosphate dehydrogenase are oxidized in phosphorylating (State 3) or non-phosphorylating (State 4) conditions. The highest rate of H₂O₂ release occurs at complex I in reverse electron transfer when mitochondria oxidize succinate or glycerol 3-phosphate. H₂O₂ production by complex III is significant only in the presence of antimycin A, and in this case, the oxygen dependence manifests mixed (hyperbolic+linear) kinetics. Our study demonstrates that H₂O₂ release by brain mitochondria is lower in hypoxia than in normoxia.

Introduction

The major energy-generating metabolic pathway in most mammalian cells is oxidative phosphorylation, where molecular oxygen is reduced to water by the mitochondrial respiratory chain. As a byproduct, mitochondria are also able to catalyze the one-electron reduction of oxygen to produce the superoxide radical (O₂^.-), a precursor of hydrogen peroxide (H₂O₂), which is the most common reactive oxygen species (ROS) in the cell (Loschen et al. 1974, Murphy 2009, Andreyev et al. 2015). Therefore, ROS formation is an inherent property of aerobic mitochondrial metabolism catalyzed by the respiratory chain complexes and matrix enzymes (Andreyev et al. 2005). The processes leading to the generation of ROS have gained significant attention over the last 30 years due to their involvement in apoptosis, hypoxic cellular responses, aging, and various signaling pathways (reviewed in (D’Autreaux & Toledano 2007, Schieber & Chandel 2014)). Mitochondria are commonly considered a major source of ROS, but the conditions under which mitochondrial ROS are produced is still a matter of debate (Starkov et al. 2014).

Excessive levels of ROS lead to oxidative stress, which plays a crucial role in mediating tissue damage after brain ischemia-reperfusion (Kalogeris et al. 2014, Lust et al. 2002). Brain tissue is specifically vulnerable to the deleterious effects of oxidative stress because of its low capacity for cell division and repair. Since molecular oxygen is also the substrate of ROS generation, the dependence of this reaction on oxygen concentration has been studied by several groups, but mostly in hyperoxic/hyperbaric conditions, using mitochondria from liver and heart (Boveris & Chance 1973, Oshino et al. 1975, Boveris et al. 1972, Turrens et al. 1982b, Turrens et al. 1982a). Until now, the relationship between oxygen concentration and ROS production has never been carefully analyzed in intact brain mitochondria. This knowledge is of great importance for brain ischemia/reperfusion studies where the interplay between energy failure, oxygen availability and ROS generation is critical. To our knowledge, the effect of oxygen concentration on the rate of H₂O₂ release by intact brain mitochondria has only been analyzed under limited experimental settings, and only at supraphysiological oxygen concentrations (Kudin et al. 2004).

Here, we performed a systematic analysis of H₂O₂ release from intact mouse brain mitochondria and demonstrate that, under normal respiration, the release of H₂O₂ is directly proportional to oxygen concentration regardless of the respiratory substrates used. The highest rates of H₂O₂ release were registered under conditions of reverse electron transfer (RET). Our data also demonstrate that complex III is not a major contributor to ROS generation in physiological conditions. Antimycin A significantly stimulates complex III ROS production, and its oxygen dependence manifests Michaelis-Menten-like kinetics. We also found that, even in the absence of succinate, complex II can contribute to ROS generation when the quinone pool is reduced by glycerol 3-phosphate.

Materials and Methods

Sources of chemicals, animal and, mitochondria.

The study was not pre-registered, and all experiments were performed in vitro using intact mouse brain mitochondria, no blinding was required. Most of the chemicals were purchased from Sigma, including fatty acid free bovine serum albumin (Cat# A6003) glycerol 3-phosphate (Cat# 50019), rotenone (Cat# R8875), antimycin A (Cat# A8674), myxothiazol (Cat#T5580). Pierce BCA protein assay kit (Cat# 23225), Amplex UltraRed (Cat# A36006) and horseradish peroxidase (Cat# 012001) were from Thermo Fisher Scientific. Atpenin A5 was from Cayman Chemical (Cat# 11898). All procedures were approved by the Institutional Animal Care and Use Committee of Weill Cornell Medicine and performed in accordance with the ARRIVE guidelines (Kilkenny et al. 2010). Intact mitochondria were isolated from 8–10 weeks C57Bl/6 male mice ((IMSR Cat# JAX:000664, RRID:IMSR_JAX:000664) by differential centrifugation with digitonin treatment (Rosenthal et al. 1987, Stepanova et al. 2017). The forebrain hemispheres were excised and immediately immersed into ice-cold isolation medium (225 mM mannitol, 75 mM sucrose, 20 mM HEPES-Tris, 1mM EGTA, 1mg/ml bovine serum albumin (BSA), pH 7.4). The tissue was homogenized with 40 strokes of a Dounce homogenizer with a tight pestle in 10 ml of the isolation medium, diluted twofold and centrifuged at 5,900 g for 4 min in a Beckman centrifuge at 4°C. The supernatant was supplemented by 0.02% digitonin and centrifuged again at 10,000 g for 10 min. The pellet was then gently resuspended in the isolation buffer without BSA and washed twice by centrifuging at 10,000 g for 10 min. The mitochondrial pellet was then resuspended in 0.15 ml of washing buffer and stored on ice.

Assessing respiration and H₂O₂ production in intact mitochondria.

A high-resolution respirometer (O2k, Oroboros Instruments GmbH) was equipped with home-built two-channel fluorescence optical module to monitor simultaneously oxygen concentration and H₂O₂ production as described in detail in (Stepanova et al. 2017) with a data collection interval of 0.2 s. H₂O₂ release was measured from the fluorescence of resorufin formed from Amplex UltraRed reacting with H₂O₂ in the presence of horseradish peroxidase. The fluorescence was calibrated into an H₂O₂ concentration by addition of 90nM of H₂O₂ to the chamber at the end of the study. To change oxygen levels, humidified pre-purified argon gas was equilibrated with the media of the respirometer chamber at a flow rate of 10–60ml/min. For effective gas exchange, a custom stopper for the Oroboros chamber was built that created 1 mL gas headspace above the liquid phase and contained two channels for probes. This system allows the monitoring of oxygen concentration but not mitochondrial respiration due to exchange of oxygen between the medium and the head space.

Mitochondria (0.1–0.5 mg/ml) were added to assay media containing 125 mM KCl, 14 mM NaCl, 0.2 mM EGTA, 4 mM KH₂PO₄, 2 mM MgCl₂, 20 mM HEPES-Tris, pH 7.4, supplemented with 2 mg/ml BSA, 10 μM Amplex UltraRed reagent, 4 U/ml horseradish peroxidase, and 40 U/ml Cu/Zn SOD at 37°C. No change in fluorescent signal was observed in the presence of 100 U/ml catalase. Three different substrate combinations were tested: 2 mM malate and 5 mM pyruvate, 5 mM succinate and 1 mM glutamate, and 5–40 mM glycerol-3-phosphate. The concentration of glycerol 3-phosphate was 40mM if not specified otherwise. For each substrate combination, three different experimental conditions were used: non-phosphorylating respiration (resting respiration, substrates are present, ADP is absent), state 3 phosphorylating respiration (ADP-stimulated respiration, substrates and 200 μM ADP present) and inhibited respiration (substrates and inhibitor are present, ADP is absent).

The combination of malate/pyruvate was used to generate matrix NADH for respiration by complexes I-II-IV. Here the pyruvate is oxidized by pyruvate dehydrogenase generating NADH and acetyl-CoA which further reacts with matrix oxaloacetate to shift the malate dehydrogenase reaction towards oxidation of malate and formation of NADH. Succinate/glutamate couple provide respiration via complexes II-III-IV. The succinate is the direct substrate of complex II and glutamate is used by glutamic-oxaloacetic transaminase to remove oxaloacetate, a potent inhibitor of complex II. Glycerol 3-phosphate is a substrate of mitochondrial glycerol 3-phosphate dehydrogenase (GPDH) located on the outer surface of the inner mitochondrial membrane which uses quinone as an electron acceptor providing electron transfer towards complexes III-IV (Patole et al. 1986, Kwong & Sohal 1998).

Data analysis and curve fitting.

All data calculations were performed using R (version 3.5.1) in RStudio (Version 1.1.456, Boston, MA). The rate of H₂O₂ release was calculated by differentiation of the total H₂O₂ released using the Package locfit which simultaneously performs smoothing and differentiation of the graph data by applying a local quadric fitting with a bandwidth of 10–20 data points (other parameters were kept at default). At least five separate preparations of mitochondria were used for each experimental condition. To generate the final data set of H₂O₂ rate at different oxygen concentrations, at least 5–8 traces for each experimental condition were averaged using adjusted oxygen concentrations. The averaged curves of H₂O₂ release rate versus oxygen concentration were fitted either using a linear function (Eq. 1) or a sum of linear and hyperbolic functions (Eq. 2) (Fig. 1E, F):

$$y=k\cdot x+const$$ (Eq. 1)

$$y=\frac{a\cdot x}{K_m+x}+k\cdot x+const$$ (Eq. 2)

where y is the rate of H₂O₂ release, x is the oxygen concentration, and k, a, K_m and const are fit parameters. The details of the fitting methods can be found in the Supplement section. For clarity, only values of k and K_m are shown in the figures.

Fig. 1.

Scheme of assessment of the effect of oxygen concentration on H₂O₂ release by intact mitochondria in different conditions. The reaction started with the addition of mitochondria (0.5 mg protein/ml) to the assay media containing 5 mM succinate and 1 mM glutamate (left panel) or 10 mM glycerol 3-phosphate and 1 μM antimycin A (right panel). Oxygen concentration as measured directly by the Oroboros respirometer was rapidly varied by continuously purging the headspace with argon (Ar) or air as detailed in the Materials and Methods section. (A, B) Representative registration traces of oxygen concentration (blue) and release of H₂O₂ measured with Amplex UltraRed (black). (C,D) The calculated rates of H₂O₂ release are shown as red traces. (E, F) H₂O₂ release rate vs. oxygen concentration were plotted and fitted with either a linear function (E, data from A) or a sum of both linear and hyperbolic functions (F, data from B). Corresponding equations are shown on the graphs.

All results were expressed as means ± SEM. For fit parameters, standard errors were obtained in the weighted fitting procedure. Rates of H₂O₂ release were compared using ANOVA followed by post hoc pair-wise t-tests with false discovery rate correction for multiple comparisons. Normality of data was assessed using three tests: Shapiro-Wilk, Anderson-Darling, and Kolmogorov-Smirnov. Chi-squared test for outliers was conducted. Other experimental details are described in the figure legends.

Results

Rates of H₂O₂ release by brain mitochondria at ambient oxygen concentration depend on respiratory substrates and the presence of respiratory chain inhibitors.

Our goal was to investigate the effect of physiologically relevant oxygen concentrations on ROS production in intact brain mitochondria. First, we determined the rates of H₂O₂ release from mitochondria respiring at ambient oxygen (200μM [O₂]) with different combinations of substrates. Rates of oxygen consumption and H₂O₂ release by intact brain mitochondria oxidizing malate/pyruvate, succinate/glutamate or succinate and glycerol 3-phosphate alone are summarized in Table 1. For each of three substrates combinations we measured H₂O₂ release in different respiratory states as well as in the presence of two specific respiratory chain inhibitors, rotenone and antimycin A. The highest rate of H₂O₂ release was observed in conditions of reverse electron transfer (RET) when the mitochondria have a high proton-motive force and oxidize succinate (with or without glutamate) or glycerol 3-phosphate (t-test, p<0.05, when compared to malate/pyruvate). During this process, the proton motive force drives complex I in reverse transferring electrons from the ubiquinol pool (generated from the complex II or GPDH substrates) to reduce NAD⁺. H₂O₂ release in RET was highly sensitive to the presence of ADP, which decreases the proton-motive force, or the complex I specific inhibitor rotenone, indicating that ROS release is at the level of complex I due to the RET.

Table.

Respiration and H₂O₂ emission in mouse brain mitochondria

Conditions	Substrates
	Mal/Pyr	Succ/Glu	G3P	Succ
H2O2 release, pmol/min/mg protein
Non-phosphorylating (no ADP)	143 ± 14*	1748 ± 99*	1308 ± 78*	1137 ± 86
State 3 (+0.2mM ADP)	55 ± 5*	142 ± 12	124 ± 15	134 ± 11
1μM AntA	713 ± 56	519.0 ± 32.6	1206 ± 38*	441 ± 16
1–2μM Rot	302 ± 17*	584.5 ± 8.4	505 ± 17	227 ± 12
1μM Rot + 1mM Malonate			406 ± 36
Respiration, nmol/min/mg protein
Non-phosphorylating (no ADP)	15.2 ± 1	52.0 ± 3.5	29.1 ± 1.9	48.7 ± 2.3
State 3 (+0.2mM ADP)	140.2 ± 6.7	242.4 ± 15.6	82.5 ± 5.1	186.8 ± 9.4
1μM AntA	6.2 ± 0.3	7.4 ± 0.7	3.2 ± 0.2	6.7 ± 0.5
RCR	9.5 ± 0.8	4.7 ± 0.2	2.9 ± 0.2	3.8 ± 0.3

Mal/Pyr: 2 mM malate + 5 mM pyruvate, Succ/Glu: 5 mM succinate + 1 mM glutamate, G3P: 40 mM glycerol 3- phosphate, Succ: 10 mM Succinate AntA: antimycin A; Rot: rotenone; RCR: respiratory control ratio determined by the rate of state 3 respiration divided by the rate of non-phosphorylating respiration.

^(*)– mean of the group is significantly different (p<0.05 ) from the other substrates for the same conditions.

Addition of the complex III inhibitor antimycin A to mitochondria affected the rate of H₂O₂ release differently depending on the substrate used. Three distinct outcomes where observed. When malate/pyruvate were used as substrates, the rate of H₂O₂ release increased, whereas with succinate/glutamate the rate decreased. No change was observed when glycerol 3-phosphate was used as the substrate. An exceptionally high rate of ROS production was found during oxidation of glycerol 3-phosphate in the presence of antimycin A, pointing to the existence of an additional contributor to overall ROS production, besides complex III.

Oxygen concentration affects H₂O₂ release by intact brain mitochondria.

To assess the effect of oxygen concentration on H₂O₂ release, we implemented the experimental approach detailed below. Fig. 1 shows representative traces of the fluorescence signal of Amplex UltraRed (Fig. 1A, B, black traces) when oxygen concentration in the medium of respiring mitochondria was rapidly changed (Fig. 1A, B, blue traces). The fluorescence signal, which is proportional to the total H₂O₂ released, was differentiated with respect to time to give the rate of H₂O₂ release (red traces, Fig. 1C and D). Variation of oxygen concentration (from 0 to 200 μM) strongly affected the rate of H₂O₂ release. We observed two different patterns of H₂O₂ release rates in response to changing oxygen concentration. In the absence of antimycin (Fig 1C), H₂O₂ release rate (red trace) and oxygen concentration (blue dashed line) followed the same pattern whereas, in the presence of antimycin A (Fig. 1D), the trace of H₂O₂ release rate deviated from the trace of oxygen concentration.

The recorded rate of H₂O₂ release was then plotted against oxygen concentration and fitted to the mathematical function (Fig. 1E, F). In the absence of antimycin A, we observed a linear dependence of H₂O₂ release rate on oxygen concentration (Fig. 1E), while the addition of antimycin A resulted in a non-linear curve that can be described as a sum of linear and hyperbolic functions (Fig. 1F).

H₂O₂ release by intact brain mitochondria linearly depends on oxygen concentration

We measured the oxygen dependence of H₂O₂ release by intact mitochondria oxidizing different substrates in the absence of any inhibitors of the respiratory chain. In non-phosphorylating respiring mitochondria the H₂O₂ release rate was linearly proportional to the oxygen concentration up to 200 μM for all substrates (Fig. 2A-C, open squares). The highest rate of H₂O₂ release was recorded in conditions of RET when mitochondria oxidize succinate in the presence of glutamate (Fig. 2B). Addition of ADP to respiring mitochondria stimulated oxidative phosphorylation (state 3), decreased proton-motive force (data not shown), and strongly diminished the rates of H₂O₂ release for all substrates. The residual H₂O₂ release in the presence of ADP was only 10% of the RET-supported ROS production (Fig 1B, C). The linear dependence of H₂O₂ release on oxygen concentration, however, remained (Fig. 2A-C, open triangles).

Fig. 2.

Dependence of H₂O₂ release rate on the oxygen concentration in the absence of inhibitors. H₂O₂ release during non-phosphorylating (no ADP, open squares) and phosphorylating, state 3 (0.2 mM ADP added, open triangles) respiration was measured at different [O₂] as detailed in the Materials and Methods section. Mitochondria (0.1–0.5 mg protein/ml) were added to the assay media containing the following substrates: 2 mM malate and 5 mM pyruvate (A), 5 mM succinate and 1 mM glutamate (B), and 40 mM glycerol 3-phosphate (C). At least five different isolations of mitochondria were used and 5–6 traces for each experimental condition were averaged (n=5–6). Fit parameters of Eq 1 are shown on the graphs.

Antimycin A, but not rotenone, introduces non-linearity to the oxygen dependence of the H₂O₂ release rate.

We then tested the effects of two classical inhibitors of complex I and III - rotenone and antimycin A, respectively - on H₂O₂ release rate at different oxygen concentrations. Addition of rotenone to mitochondria increased H₂O₂ release when respiring on malate and pyruvate and decreased it in conditions of RET (Table 1, Fig. 3A-C, solid triangles). In all these conditions, the linear oxygen dependences were preserved. Surprisingly, with all substrates, in the presence of antimycin A, the oxygen dependence of H₂O₂ release displayed more complex kinetics (Fig. 3A-C, solid squares). The function of oxygen dependence under these conditions is described as a sum of linear and hyperbolic terms by Eq. 2.

Fig. 3.

Dependence of H₂O₂ release rate on the oxygen concentration for mitochondria in the presence of respiratory chain inhibitors. Experimental conditions were the same as in Fig. 2. with the following substrates 2 mM malate and 5 mM pyruvate (A), 5 mM succinate and 1 mM glutamate (B) or 40 mM glycerol 3-phosphate (C). No ADP was present. The H₂O₂ release at different [O₂] in the presence of 1 μM rotenone (solid triangles) or 1 μM antimycin A (solid squares) was measured as described in the Materials and Methods. At least five different isolations of mitochondria were used and 5–9 traces for each experimental condition were averaged (n=5–9). Fit parameters of Eq 1 and Eq. 2 are shown on the graphs.

For malate/pyruvate-based respiration the quantitative analysis of the fitting curve (Fig. 3A, solid squares) shows that the addition of antimycin A results in the appearance of the hyperbolic component with an apparent K_m for oxygen of 12.7±0.4 μM. These results seem to indicate that, in the presence of antimycin A, ROS generation at complex III manifests itself with classical Michaelis-Menten kinetics for oxygen. At the same time, the linear term of the Eq. 2 quantitatively corresponded to the slope in the presence of rotenone (Fig 3A, triangles, 1.3±0.1 pmol H₂O₂/min/mg protein per μM [O₂]).

For succinate/glutamate-based respiration (Fig. 3B, solid squares), the addition of antimycin A gave rise to the hyperbolic component with an apparent K_m of 6.9±0.4 μM [O₂]. The linear term (with the slope of 1.00±0.01 pmol H₂O₂/min/mg protein per μM [O₂]) corresponded to the rate during respiration in state 3 (Fig. 2B, open triangles, 0.60±0.02 pmol H₂O₂/min/mg protein per μM [O₂]). Therefore, in the presence of antimycin A, the hyperbolic dependence originates from complex III while the linear component is provided by the activity of the respiratory chain operating in the absence of a proton motive force upstream of the block imposed by the inhibitor.

For glycerol 3-phosphate-based respiration, the addition of antimycin A also resulted in a mixed hyperbolic and linear dependence pattern (Fig. 3C, solid squares). The apparent K_m value for the first component was 5.9±0.3 μM [O₂]. The slope of the linear term was 2.2±0.1 pmol H₂O₂/min/mg protein per μM [O₂], most likely corresponding to the production of ROS by the enzymes upstream of the antimycin A block in the absence of proton-motive force.

Oxidation of glycerol 3-phosphate and H₂O₂ release.

As shown previously, mitochondrial GPDH located on the outer surface of the inner mitochondrial membrane can generate ROS at a significant rate (Patole et al. 1986, Kwong & Sohal 1998, Miwa et al. 2003, Tretter et al. 2007). Activity of GPDH is strongly dependent on the concentration of the substrate (Tretter et al. 2007, Orr et al. 2012). We also found that both oxidation of glycerol 3-phosphate (Fig. 4A) and H₂O₂ release (Fig. 4B) depended on the concentration of the substrate in the millimolar range.

Fig. 4.

Respiration and H₂O₂ release by mitochondria oxidizing glycerol 3-phosphate. (A) Effect of glycerol 3-phosphate concentration on non-phosphorylating (no ADP added, white bars) and state 3 phosphorylating respiration (0.2 mM ADP, grey bars). (B) Effect of glycerol 3-phosphate concentration on non-phosphorylating H₂O₂ release (white bars), state 3 phosphorylating H₂O₂ release (grey bars), and antimycin A-inhibited H₂O₂ release (black). (C-E) Effect of complex II specific inhibitor atpenin A5 (1.5 nM, open triangles) on the dependence of H₂O₂ release rate on oxygen concentration in the presence of 1 μM rotenone (C), 1 μM rotenone and 25 nM uncoupler SF 6847 (D) or 1 μM rotenone and 2 μM complex III inhibitor myxothiazol (E). At least five different isolations of mitochondria were used and 5–6 traces for each experimental condition were averaged (n=5–6). Fit parameters of Eq. 1 are shown on the graphs. Concentration of glycerol 3-phosphate was 40 mM.

We tested the effect of complex II inhibition on the rate of H₂O₂ release when mitochondria are oxidizing glycerol 3-phosphate. These experiments were carried out in the presence of rotenone to prevent RET-induced H₂O₂ release, and we used atpenin A5, a potent inhibitor of complex II, to block electron transfer at the ubiquinone binding site of the enzyme (Miyadera et al. 2003, Horsefield et al. 2006). In all tested conditions the rate of H₂O₂ release was decreased by the addition of atpenin (Fig 4C-E). The effect of complex II inhibition was less pronounced when the mitochondrial membrane potential was collapsed with an uncoupler (Fig. 5D) and more pronounced when quinone pool was fully reduced by inhibiting complex III with myxothizol (Fig. 4E). Complex II contribution calculated from an atpenin-sensitive fraction of total H₂O₂ release was around 21% during non-phosphorylating respiration (Fig 4C), and the addition of uncoupler decreased it to only 17% (Fig. 4D). In the presence of myxothiazol, complex II contribution was half of the total H₂O₂ release rate. Therefore, in isolated brain mitochondria, complex II can contribute to overall H₂O₂ release, even in the absence of succinate.

Discussion.

It is generally accepted that mitochondria are key organelles involved in ROS metabolism, both as producers and scavengers of ROS (Andreyev et al. 2015, Starkov et al. 2014). The relationship between ROS generation by mitochondria and oxygen concentration has been the subject of controversy in the recent literature. After the pioneering work from the Chance lab (Boveris & Chance 1973), it was accepted that hyperbaric oxygen stimulates H₂O₂ production. The H₂O₂ release rate by heart intact mitochondria was found to almost linearly correlate with oxygen concentration, between 1 and 10 atmospheric pressure of oxygen (Boveris & Chance 1973). The effect of oxygen concentration on H₂O₂ release by perfused liver was later revisited in the same laboratory (Oshino et al. 1975), and it was concluded that H₂O₂ production increases with increasing oxygen concentration, within physiological intervals of [O₂]. However, only 10–15% of H₂O₂ production in the liver can be attributed to mitochondria (Boveris et al. 1972). H₂O₂ release by intact lung mitochondria, as well as superoxide production during NADH oxidation by lung sub-mitochondrial particles in the presence of inhibitors, were also directly correlated with oxygen concentration in hyperoxic (200–800 μM [O₂]) conditions (Turrens et al. 1982b, Turrens et al. 1982a). More recently H₂O₂ production dependence on oxygen concentrations was tested using rat brain mitochondria at very high oxygen level (200–1000 μM [O₂]) (Kudin et al. 2004). The authors suggested hyperbolic concentration dependences with an apparent K_m for oxygen of approximately 1 mM for mitochondria oxidizing either malate/glutamate or succinate, while the experimental data fitted well to a linear dependence.

In this study, we showed for the first time that H₂O₂ release by intact brain mitochondria in the absence of inhibitors depends linearly on oxygen concentration in the physiological range (<200μM [O2]). This included non-phosphorylating (resting) and phosphorylating (state 3) respiration during forward electron transfer from the substrates of NAD-dependent dehydrogenases (malate and pyruvate), as well as in RET-like conditions during oxidation of succinate or glycerol 3-phosphate. We conclude that, in the absence of inhibitors, there is no specific site for oxygen binding in the reaction of mitochondrial ROS formation and this process obeys pseudo-first order kinetics and can be considered as an electron “leak” from the reduced redox center(s) of the respiratory chain. Our finding agrees with recent observations of linear dependence of superoxide generation by mammalian complex I in vitro (5–200 μM [O₂]) (Kussmaul & Hirst 2006, Vinogradov & Grivennikova 2005).

The hyperbolic dependence shown in liver mitochondria assayed with Amplex Red (Hoffman et al. 2007, Hoffman & Brookes 2009) is likely due to the different enzymatic composition of the mitochondria preparations from different tissues. As demonstrated by the work of Miwa et al. (Miwa et al. 2016), liver mitochondria possess a specific phenylmethylsulfonyl fluoride- sensitive carboxylesterase enzyme able to catalyze the deamidation of Amplex Red eventually leading to the formation of resorufin. Addition of 100 μM phenylmethylsulfonyl fluoride did not affect the H₂O₂ production rates measured in the present study, indicating that the carboxylesterase activity is not present in mouse brain mitochondria (data not shown).

As it was previously shown the highest rate of H₂O₂ release was found in conditions of RET (Treberg et al. 2011, Votyakova & Reynolds 2001, Turrens & Boveris 1980), highlighting the importance of this process in pathological conditions, such as ischemia/reperfusion. Succinate levels in the brain can be elevated up to 30 fold after periods of ischemia (Folbergrova et al. 1974, Solberg et al. 2010, Benzi et al. 1979, Sahni et al. 2017). Glycerol 3-phosphate, another substrate of RET is also increased after brain ischemia (Nguyen et al. 2007). Due to the lateral oxygenation in the penumbra area and after reperfusion, ROS production by RET can overcome the capacity of matrix ROS detoxification and significantly contribute to the oxidative stress observed in ischemia/reperfusion.

Previously published results (Hinkle et al. 1967, Turrens & Boveris 1980, Votyakova & Reynolds 2001, Treberg et al. 2011, Niatsetskaya et al. 2012b, Quinlan et al. 2013, Pryde & Hirst 2011), and results of the present study (Fig. 2), show that RET-like conditions provide the highest rate of H₂O₂ release. The sensitivity of the H₂O₂ production during oxidation of succinate or glycerol 3-phosphate to rotenone and ADP identifies complex I as a major ROS contributor in these conditions. Most likely, complex I flavin is a direct electron donor for the reduction of oxygen in superoxide-generating reaction, as suggested earlier (Turrens & Boveris 1980, Liu et al. 2002, Kudin et al. 2004, Vinogradov & Grivennikova 2005, Galkin & Brandt 2005, Kussmaul & Hirst 2006, Stepanova et al. 2017). An oxygen molecule can approach enzyme-bound flavin only via a narrow passage formed by the NADH-binding subunit (Fiedorczuk et al. 2016, Zhu et al. 2016), so the presence of the nucleotide bound to complex I restricts the access of oxygen to flavin. Kinetic measurements (Barker et al. 2007, Knuuti et al. 2013) and atomistic dynamic simulation studies (Na et al. 2018) suggest that the flavin is reoxidized and electrons are transferred downstream before NAD⁺ is released. During forward electron transfer, oxygen can approach the flavin only after NAD⁺ is released, leaving flavin in the oxidized state, so that no ROS generation can occur. However, in conditions of RET, the entire chain of iron-sulfur clusters within the enzyme is reduced from the downstream and reduction of the flavin occurs independently of the absence or presence of nucleotide in the active center. In this scenario, flavin is in the reduced form, and the NADH binding passage is accessible to oxygen, allowing it to interact with FMNH₂ to produce ROS.

The data reported here strongly reinforce this proposal and explain experimental data presented in this paper and studies by others of dual effect of rotenone-like specific inhibitors on ROS production by mitochondria oxidizing either NAD⁺-depending substrates (malate and pyruvate) or succinate (Andreyev et al. 2005, Dröse et al. 2016, Quinlan et al. 2013). In our conditions, rotenone increases H₂O₂ release with malate and pyruvate by two-fold. On the contrary, during RET rotenone decreases H₂O₂ release by around 70 and 50% with succinate or glycerol 3-phosphate, respectively. Rotenone binds close to the quinone site of complex I and interrupts electron transfer between the terminal iron-sulfur cluster and ubiquinone (Zhu et al. 2016). Therefore, when malate/pyruvate are used to generate NADH binding of rotenone increases the flavin reduction and the probability of its interaction with oxygen. In conditions of RET, rotenone prevents reduction of flavin from the downstream, leading to decreased ROS production.

Early studies have emphasized ROS production at complex III (Boveris & Chance 1973, Turrens et al. 1985). Antimycin A acts as an effective inhibitor of net electron transfer by complex III by blocking the Q_i site of the enzyme. In steady-state conditions, in the presence of antimycin A, the low-potential chain becomes reduced, eliminating heme b_L as an electron acceptor. The high-potential chain remains oxidized so that the Q_o site is controlled by an excess of quinol and oxidized Rieske protein. This favors formation of the semiquinone at the Q_o site, which can donate one electron to molecular oxygen (Crofts et al. 1999, Swierczek et al. 2010, Ksenzenko et al. 1983, Ksenzenko et al. 1984) (see also (Dröse & Brandt 2008) for a review). Without the inhibitor, the contribution of complex III to the mitochondrial ROS generation is not significant. However, the production of ROS from complex III could play a role in pathological scenarios, ranging from hereditary to acquired defects of this enzyme (Borek et al. 2015, Rana et al. 2000).

Here, we find that addition of antimycin A drastically changes the linear oxygen dependence of H₂O₂ release observed in the absence of inhibitors or in the presence of rotenone. After addition of antimycin A, the linear relationship is transformed to a more complex function which can be described as the superposition of linear and hyperbolic terms.

For the forward electron transfer (malate/pyruvate), the linear term quantitatively corresponds to the dependence in the presence of rotenone since both inhibitors block electron transfer from substrates of NAD-dependent dehydrogenases downstream. In the presence of either antimycin A or rotenone, complex I redox centers are maximally reduced, therefore antimycin A should induce “rotenone”-like H₂O₂ release in the complex I.

When antimycin A is present during succinate oxidation, the linear component is close to state 3 respiration. Based on the slope values, the linear contribution is, probably, provided by complex II operating at diminished proton-motive force (e.g., state 3 respiration). This is consistent with the fact that inhibition of complex III by antimycin A would inevitably lead to the collapse of the proton-motive force.

When mitochondria are oxidizing glycerol 3-phosphate, the linear component has an oxygen dependence slope of 2.2 pmol H₂O₂/min/mg protein per μM [O₂] (Fig. 3C). This dependence can be due to the combined activities of GPDH (Kwong & Sohal 1998, Miwa et al. 2003, Tretter et al. 2007) and complex II (Grivennikova et al. 2017, McLennan & Degli Esposti 2000, Quinlan et al. 2012, Siebels & Drose 2013).

The hyperbolic component of oxygen dependence originating in complex III in the presence of antimycin A can be characterized by an apparent K_m for oxygen in the range of 5–12 μM, depending on the substrate used. The most straightforward explanation for the hyperbolic dependence is that one-electron reduction of O₂ at the Q_o site of complex III requires a specific site for oxygen binding that can be characterized with an apparent value of K_m. However, the Q_o site is located in the hydrophobic phase, with a wide opening at the distal part allowing for interactions with the Rieske iron-sulfur protein in the water phase (Crofts et al. 1999). Since the oxygen molecule is also hydrophobic and the reaction of oxygen with semiquinone is a rapid process (Valgimigli et al. 2008), it is unlikely that the oxygen needs to bind to be reduced by the Q_o semiquinone. Furthermore, the K_m for oxygen had different values with different substrates and it is unlikely that this reflects the existence of different oxygen binding sites for different substrates. Probably, the redox poise of the quinone pool upstream complex III provided by different substrates affects the value of K_m.

The reactions rate of ROS production (ν_ROS) from a reduced center is a second-order reaction given by:

$${\mathit{\nu }}_{\text{ROS}}=\mathrm{\kappa }\times [\text{R}]\times [{\text{O}}_2]$$ (3)

where κ is the rate constant, [R] is the concentration of the reduced species donating electrons to O₂, (electron leakage) and [O₂] is the oxygen concentration. In respiring mitochondria the instantaneous flux via a redox center is much larger than the rate of electron leakage to oxygen with ROS formation. Therefore, the low rate of electron transfer from R to oxygen relative to the instantaneous flux will not affect the concentration of the reduced species and the rate of ROS production will be linear with oxygen concentration. In contrast, if the rate of generation of the reduced species is slow compared to the rate it reacts with oxygen, then there will be competition between the two processes and [R] will decrease at high oxygen concentrations leading to a hyperbolic relationship, as seen with antimycin. When complex III is inhibited with antimycin, the normal electron bifurcation cannot occur and the concentration of semiquinone at the Q_o center depends on the equilibrium between the semiquinone and quinol couple at the Q_o center and the redox potential of the cytochrome c pool. This process is not part of the normal turnover of the enzyme and so may be slow compared with the rate of electron leakage. Therefore, the ROS release from complex III in the presence of antimycin A may be explained by the effect of oxygen concentration on the semiquinone concentration at Q_o, which then participates in ROS generation. However, the exact mechanism by which antimycin induces a hyperbolic dependence of ROS production on oxygen concentration is not clear.

Finally, we found that inhibition of complex II by atpenin A5 decreases the slope of oxygen dependence in mitochondria oxidizing glycerol 3-phosphate when rotenone is present. This indicates that complex II is able to participate in ROS production in these conditions, even in the absence of succinate. The contribution of complex II H₂O₂ production is probably dependent of the oxidation state of the ubiquinone pool as the uncoupler, which will oxidize the pool due to increased rate of oxidation of ubiquinol by complex III, decreases the complex II contribution (Fig 4D) and myxothiazol, which blocks oxidation of the pool by inhibiting complex III, increases the complex II contribution (Fig 4E). Therefore, complex II can be maintained in the reduced state and contribute to ROS generation even in the absence of its substrate when quinone pool is reduced, e.g., by glycerol 3-phosphate. Our data cannot address which centers generate the ROS, but two redox centers can potentially participate: the flavin adenine dinucleotide (Orr et al. 2012, Dröse 2013) and the [3Fe-4S] iron-sulfur center (Yankovskaya et al. 2003, Grivennikova et al. 2017). Regardless of the center which generates the ROS, the contribution of complex II to ROS production is likely to be minor in physiological conditions.

In summary, our data suggest that ROS generation by brain mitochondria depends linearly on oxygen concentration obeying the law of mass action (Guldberg & Waage 1864). The highest rates of H₂O₂ release were observed in the conditions of RET. Complex III contributes to overall ROS production only in the presence of its specific inhibitor antimycin A, and H₂O₂ release from this enzyme shows hyperbolic dependence on oxygen concentration. Another important finding of our studies is that in brain mitochondria complex II can contribute to ROS generation even in the absence of succinate when the quinone pool is reduced with other substrate(s).

Although the subject of the presented study was restricted to intact brain mitochondria, there is potential clinical relevance. For example, excessive use of oxygen supplementation during resuscitation was strongly associated with greater oxidative injury and poorer neurological outcome in asphyxiated piglets and rats (Solberg et al. 2012, Koch et al. 2008). In contrast, the extent of brain injury and the acuity of cerebral oxidative damage was significantly decreased in neonatal mice reoxygenated with 18% oxygen during initial hour of reperfusion (Niatsetskaya et al. 2012a) The very distinctive linear dependence of the H₂O₂ release from intact mitochondria on the oxygen concentration in the absence of inhibitors will be very useful in critically testing any mechanisms proposed for role of mitochondria in hypoxic tissue response.

Abbreviations used:

FAD

flavin adenine dinucleotide

FMN

flavin mononucleotide

GDPH

glycerol 3-phosphate dehydrogenase

RET

reverse electron transfer

ROS

reactive oxygen species