Discrimination between two possible reaction sequences that create potential risk of generation of deleterious radicals by cytochrome bc₁

Abstract

In addition to its bioenergetic function of building up proton motive force, cytochrome bc₁ can be a source of superoxide. One-electron reduction of oxygen is believed to occur from semiquinone (SQ_o) formed at the quinone oxidation/reduction Q_o site (Q_o) as a result of single-electron oxidation of quinol by the iron–sulfur cluster (FeS) (semiforward mechanism) or single-electron reduction of quinone by heme b_L (semireverse mechanism). It is hotly debated which mechanism plays a major role in the overall production of superoxide as experimental data supporting either reaction exist. To evaluate a contribution of each of the mechanisms we first measured superoxide production under a broad range of conditions using the mutants of cytochrome bc₁ that severely impeded the oxidation of FeS by cytochrome c₁, changed density of FeS around Q_o by interfering with its movement, or combined these two effects together. We then compared the amount of generated superoxide with mathematical models describing either semiforward or semireverse mechanism framed within a scheme assuming competition between the internal reactions at Q_o and the leakage of electrons on oxygen. We found that only the model of semireverse mechanism correctly reproduced the experimentally measured decrease in ROS for the FeS motion mutants and increase in ROS for the mutants with oxidation of FeS impaired. This strongly suggests that this mechanism dominates in setting steady-state levels of SQ_o that present a risk of generation of superoxide by cytochrome bc₁. Isolation of this reaction sequence from multiplicity of possible reactions at Q_o helps to better understand conditions under which complex III might contribute to ROS generation in vivo.

1. Introduction

In biological energy conversion, cytochrome bc₁ (mitochondrial complex III) [1] plays a role of connecting the quinone redox pool with the cytochrome c redox pool. The net oxidation of hydroquinone and reduction of cytochrome c catalyzed by this enzyme according to the Q-cycle [2] through the action of two quinone oxidation/reduction sites (named the Q_o and Q_i sites) and the two chains of cofactors (the high potential c-chain and the low potential b-chain) contributes to generation of protonmotive force used for ATP synthesis. But in addition to this bioenergetic function, cytochrome bc₁ is one of the complexes of mitochondrial electron transfer chain believed to contribute to the formation of superoxide that participates in cell signaling and contributes to cellular damage in aging and disease [3–6].

The actual rate of superoxide production by cytochrome bc₁ in mitochondria is debatable. Under physiological conditions it could be low and negligible compared with the maximum rates of superoxide production from complex I, the other component of electron transfer chain [4]. However, ROS generated in hypoxia have been attributed to originate from complex III [7]. In addition, there is an indication that the rate of superoxide generation by cytochrome bc₁ may depend on the magnitude of the membrane potential, thus at high membrane potential, contribution from cytochrome bc₁ may be larger [8].

In general, conditions that severely impede the electron flow through the cofactor chains of cytochrome bc₁ induce production of superoxide. This occurs when the flow of electrons out of at least one of the chains is blocked, or severely inhibited, as for example in the presence of inhibitor antimycin which blocks the outflow of electrons from the b-chain [9,10], or in the presence of the heme c₁ knockout mutation which impedes the electron outflow from the c-chain mimicking the conditions of hypoxia [11].

The production of superoxide is generally considered to be a side effect of the action of the catalytic quinone oxidation/reduction site — the Q_o site (also referred as the Q_P site). However, the mechanism of how this occurs remains elusive. Most postulated mechanisms assume that highly unstable semiquinone formed at the site (SQ_o) is a direct source of single electrons that reduce oxygen to superoxide [10–14]. But SQ_o is very difficult to detect experimentally which so far has only been achieved under special conditions that were able to trap SQ_o at rather low occupancy levels [15,16]. This implies that SQ_o is formed, if ever, only very transiently and does not accumulate to significant levels when the electron flow through the cofactor chains proceeds unperturbedly. In addition, the multiplicity of possible reactions between the quinone/quinol and the two redox centers of the Q_o site, the FeS cluster and heme b_L, [11,14,17] makes it difficult to discern which reaction sequence/sequences might be primarily responsible for generation of superoxide.

Theory predicts that in the reversibly operating cytochrome bc₁ [18,19], SQ_o can be formed in two ways: as a part of the forward reaction toward oxidation of quinol when the oxidized FeS center withdraws one electron from quinol bound at the Q_o site or as a part of the reverse reaction toward reduction of quinone when the reduced heme b_L donates its electron to the quinone bound at that site. We name those two ways as semiforward or semireverse mechanisms, respectively.

Several studies have focused on examining the superoxide production occurring from SQ_o formed through the semiforward mechanism because of its intimate relation to the initial steps of the forward direction of the Q-cycle [14,20–24]. It was shown that the rate-limiting step for superoxide production in the antimycin-inhibited system is the same as for the Q-cycle and occurs at the level of direct one-electron oxidation of quinol to SQ_o by the FeS cluster [22,24]. However, recent studies have presented the semireverse mechanism as potent in contributing to superoxide production [11,13]. In particular, Brandt and colleagues have shown that superoxide formation is stimulated by the presence of oxidized ubiquinone, which was interpreted as an indication that the reverse reaction from reduced heme b_L to quinone and further to oxygen does occur [13]. We have provided the mechanistic description as to how this might occur in light of the multiplicity of possible reactions at the Q_o site (see below and ref. [11]).

We examined the production of superoxide in cytochrome bc₁ with mutations that severely impeded electron flow through the c-chain either by dramatically slowing the oxidation of the FeS center by cytochrome c₁ (the c₁ knockout) or by arresting the FeS head domain at the Q_o site [11], which otherwise undergoes movement between the Q_o site and cytochrome c₁ [25]. We proposed that a probability of the production of superoxide is inversely proportional to a probability of occurrence of competitive reactions engaging the cofactors of the Q_o site. In the framework of this model we described conditions that promote the production of superoxide taking into account the effects of transient changes in distances between the cofactors caused by the movement of the FeS cluster [11]. Importantly, this model was compatible with both the semiforward and semireverse mechanisms of formation of SQ_o.

Here, with the help of additional mutations and mathematical modeling, we distinguish these two mechanisms in such a way that their relative contribution to the overall production of superoxide by cytochrome bc₁ can be estimated. This allows us to describe the efficiency of single-electron reactions that under unfavorable circumstances lead to either short-circuit reactions or release of ROS.

2. Materials and methods

2.1. Preparation of cytochrome bc₁ mutants and protein samples

The single mutants + 1Ala, + 2Ala and M183K of Rhodobacter (Rba.) capsulatus cytochrome bc₁ and the genetic expression system for this bacterium were a generous gift from Prof. Fevzi Daldal, University of Pennsylvania, USA.

To obtain a double mutant M183K/+1Ala, the expression vector containing both mutations (pMTS1:M183K/+1Ala) was created in the following way: the plasmid pMTS1:M183K (containing a copy of petABC operon coding for all three subunits of cytochrome bc₁ with mutation M183K in petC) [26] was digested with XmaI–HindIII and the resulting 0.22 kbp fragment was used to replace the XmaI–HindIII portion of pMTS1 that contained mutation + 1Ala in petA [27]. pMTS1:M183K/+1Ala was conjugated into Rba. capsulatus strain MT-RBC1 devoid of petABC via triparental crosses as described previously [28]. The presence of the mutations was confirmed by sequencing of the plasmid DNA isolated from Rba. capsulatus strain.

Preparation of isolated complexes of the wild-type or mutated cytochrome bc₁ was performed as described previously [29].

2.2. Pulse EPR measurements of the FeS cluster

Pulse EPR measurements were performed according to the procedures described in ref. [30]. Prior to measurements all samples of cytochrome bc₁ were dialyzed against 50 mM bicine buffer pH 8.0, 100 mM NaCl, 1 mM EDTA, and 20% glycerol, and concentrated to approximately 200 μM of cytochrome c₁. All samples were reduced with 1 mM of sodium ascorbate and the EPR tubes containing ~ 10 μl of the sample were rapidly frozen by immersing in liquid nitrogen. Pulse EPR measurements were carried out on Bruker Elexsys-E580 spectrometer at Q-band (33.6 GHz). The spectrometer was equipped with ER5107D2/0501 Q-band resonator inserted in CF936 helium cryostat (Oxford Instruments). Electron spin echo (ESE) decay signal was recorded by measuring the amplitude of ESE excited by two-pulse sequence π/2–τ–π for τ increased by 4 ns. The phase relaxation rates were determined by fitting the stretched exponent to the ESE decay as described in detail in ref. [30]. For temperature range 12–26 K changes in phase relaxation rates of the FeS cluster were approximated by fitting the sum of third order polynomial and Lorentzian function to the data points as described in ref. [30].

2.3. Enzymatic and ROS generation assay

Steady-state enzymatic activity of cytochrome bc₁ and superoxide generation were performed as described in ref. [11]. Synthetic ubiquinol analog DBH₂ (2,3-dimetoxy-5-decyl-6-metyl-1,4-benzoquinol) and equine cytochrome c were used as substrates. The enzymatic assays were performed in 50 mM Tris buffer pH 8.0, containing 0.01% n-dodecyl-β-d-maltoside (DDM) and NaCl (100 mM or 400 mM). The final concentrations of substrates in all experiments were 20 μM. The final concentrations of cytochrome bc₁ (referred to the concentration of cytochrome c₁) were in the range 10–100 nM, depending on the general activity of the particular mutant. The reaction was monitored by recording the time trace of absorbance at 550 nm after injection of appropriate amount of cytochrome bc₁ to the reaction mixture containing the substrates. Turnover rates were calculated from the initial parts of the curves where the time dependence of cytochrome c reduction was linear.

2.4. Simulations of steady-state kinetics

Simulations were performed using the general approach and tools described in ref. [31]. Two different kinetic models were built to follow the idea of two different possible mechanisms of superoxide generation by cytochrome bc₁ as proposed in ref. [11]. The first model referred to as semiforward consisted of 11 stiff differential equations with 10 adjustable kinetic constants and the rate-limiting step in this model was the one-electron oxidation of QH₂ by the oxidized FeS center to form SQ. The second model referred to as semireverse consisted of 12 stiff equations with 11 adjustable kinetic parameters and the rate-limiting step involving a single-electron transfer from reduced heme b_L to Q at the Q_o site to produce SQ. Other assumptions of the models are described in Results. The equations were numerically integrated using Gear's algorithm. The semiforward and semireverse mechanisms were simulated separately. The simulation of the combination of both of these mechanisms would require at least 21 adjustable parameters, which would introduce too many degrees of freedom making the reliable estimation of the parameters difficult.

Initial parameters for all simulations were the same: oxidized cytochrome c 20 μM, reduced cytochrome c 0 μM, and cytochrome bc₁ 100 nM. To reduce the number of equations and free parameters the concentrations of QH₂ and Q were kept constant throughout the simulations. At the beginning of the simulation it was assumed that the FeS cluster was completely oxidized and the head domain was at Q_o site irrespectively of the mutant. The rate constants for all reaction steps were adjusted to reproduce enzymatic activity with and without antimycin in the presence and absence of added SOD.

3. Results

3.1. Distribution of positions of FeS cluster as seen in electron transfer and pulse EPR measurements

It is well established that in wild type, the movement of the FeS head domain connects the Q_o site with heme c₁ within microseconds which is fast enough not to limit the overall catalytic rate. This movement is beyond the kinetic resolution of the flash-induced measurements of electron transfer in chromatophores. The alanine insertions in the neck region linking the water-soluble FeS head domain with its membranous anchor introduce steric barrier which interfere with the normal movement to various degrees [25,27]. Fig. 1 presents schematically how those insertions influence the movement and thereby modify the distribution of positions of the FeS cluster in cytochrome bc₁.

Fig. 1.

Changes in equilibrium distribution of positions in native cytochrome bc₁ and the FeS head domain motion mutants. In the wild type (WT) the FeS head domain (FeS) moves freely between cytochrome b (light gray) and cytochrome c₁ (dark gray) subunits which leads to rather broad steady-state distribution of FeS positions [30]. Insertion of one alanine to the neck region of FeS protein (+ 1Ala) creates a barrier for diffusion and shifts the equilibrium distribution of FeS positions toward cytochrome b subunit (see Fig. 2). This barrier is drastically raised by insertion of two alanines (+ 2Ala) which almost completely arrests the FeS domain at the Q_o site. Horizontal rectangles depict the systematic increase in the density of FeS cluster at the Q_o represented in grayscale: the darker and lighter shades denote higher and lower occupancy of the FeS cluster, respectively.

When the + 1Ala insertion is present, the cluster is forced to stay significantly longer at the Q_o site (for microseconds) which is enough to effectively limit the rate of electron transfer to heme c₁ to the point that it now becomes resolvable kinetically [27]. However, even though the cytochrome c re-reduction on a millisecond timescale of flash-induced measurements is slower, it is still highly effective and complete. Consistent with those observations, the steady-state activity of isolated enzyme on a second timescale is only moderately decreased to about 80 s^− 1 from the native 100 s^− 1 range (see Fig. 3). The enzyme remains physiologically functional.

Fig. 3.

Superoxide production by different forms of isolated cytochrome bc₁. Cytochrome c reduction by wild-type cytochrome bc₁ (a), M183K single mutant (b), + 1Ala single mutant (c), and M183K/+1Ala double mutant (d), was measured in the absence (white bars) and presence (gray bars) of SOD. The measurements were performed at 100 mM and 400 mM NaCl (white and gray backgrounds, respectively) without any inhibitor (no inh), or with antimycin present (Ant). Only statistically significant amount of superoxide is given as a percentage of SOD-sensitive cytochrome c reduction¹. The activity of all these forms, as well as + 2Ala discussed in ref. [11], is abolished in the presence of stigmatellin or myxothiazol (not shown).

The + 2Ala insertion exerts much severer steric effect and arrests the FeS head domain at the Q_o site for seconds [27]. The phases of cytochrome c re-reduction in millisecond flash-induced measurements are abolished [27] and the steady-state turnover is decreased by nearly two orders of magnitude to about 2 s^− 1 [11]. This renders enzyme physiologically non-functional. Nevertheless, the residual activity of + 2Ala mutant still reflects the operation of the Q_o site as it is abolished in the presence of the Q_o site inhibitors (myxothiazol or stigmatellin) (see legend to Fig. 3). To sustain such turnover, the FeS cluster has to undergo cycles of oxidation and reduction, even on this second timescale, contrary to what was suggested in ref. [32]. Kinetically, even more severe effects can be observed by, for example, introducing thermodynamic barrier at the level of heme c₁ (in the c₁ knockout) [11]. The turnovers of less than 1 s^− 1, seen in those mutants (see also the M183K mutant described below), are also due to the operation of the Q_o site, thus again rely on oxidations and reductions of the FeS cluster. Operational Q_o site in all those cases (+ 1Ala, + 2Ala and c₁ knockouts) is a consequence of the fact that introduced mutations are remote from the catalytic site and do not alter its overall structural properties. This allows the oxidation of quinol to take place unperturbedly, as well documented in flash experiments where the native-like Q_o-site-mediated reduction of heme b_H is observed as long as substrate quinol is provided and oxidized FeS cluster is present at the site [18,33].

In a dynamic sense, it can be expected that the + 1Ala and + 2Ala mutations effectively change the density of the FeS cluster around the Q_o site and along the trajectory of its movement toward heme c₁. In other words this can be described as a change in the average equilibrium distribution of the FeS positions. Such a change can be conveniently monitored in a glass state of frozen solution by measuring the dipolar enhancement of phase relaxation of the FeS cluster with pulse EPR [30]. We have earlier observed that the relaxation of oxidized heme b_L enhances the phase relaxation of the FeS cluster in a distance-dependent manner; the closer the FeS cluster is to the Q_o site (and heme b_L), the stronger the enhancement is. This is reflected as an increase in a local maximum in temperature dependence of the phase relaxation of the FeS cluster. Using the various effects of the specific Q_o site inhibitors we were able to define the whole range of changes in the level of relaxation enhancement, correlating the changes with the shifts in the average equilibrium distribution of the FeS positions [30].

Fig. 2 compares the level of relaxation enhancement observed in the + 1Ala mutant with those reported earlier for the + 2Ala mutant and the wild type [30]. The relaxation enhancement in + 1Ala is significantly weaker than in + 2Ala, but at the same time it is much stronger than in wild type. The strongest enhancement in + 2Ala correlates with the FeS head domain occupying the Q_o site position most of the time (equilibrium distribution of positions is such that almost the entire population of FeS head domain is at the Q_o site) [30]. On the other hand the level of relaxation enhancement seen in wild type implicates existence of a broad distribution of positions between the Q_o site and heme c₁, as discussed in detail in ref. [30]. The relaxation enhancement of + 1Ala, placing itself in between those two cases, indicates that + 1Ala, in comparison to the wild type, has the density of the FeS cluster shifted toward the Q_o site. The shift is in the same direction as in + 2Ala, but not as large. It is consistent with the notion, coming from the interpretation of kinetic and biochemical data, that in + 1Ala the FeS head domain is held at the Q_o site in a similar manner as in + 2Ala, but for a significantly shorter time (milliseconds vs seconds) [27]. We note that changes in the relaxation enhancement definitely rule out the possibility that the observed kinetic impediments in + 1Ala are due to the slowing of the movement of the FeS head domain uniformly along the whole trajectory. In such a case, the average distribution of the FeS positions would not change significantly vs wild type, thus no changes in the relaxation enhancement would have been observed [30].

Fig. 2.

Comparison of the distance-dependent impact of the spin-lattice relaxation of heme b_L on electron phase relaxation of the FeS cluster. The phase relaxation rates between 12 K and 26 K were measured for ascorbate-reduced samples containing wild type (circles), + 1Ala (squares) and + 2Ala (triangles) mutants. Solid lines represent a trend that was approximated by fitting the sum of third order polynomial and Lorentzian function. The amplitude of local maximum of the curves is proportional to the density of the FeS cluster at the Q_o site [30].

Overall, the data in Fig. 2 provide direct spectroscopic demonstration that a density of the FeS cluster at the Q_o site does differ in each of the form. This density decreases in the order + 2Ala, + 1Ala, wild type.

3.2. Effect of mutations on the level of superoxide production by cytochrome bc₁

In vitro measurements with isolated cytochrome bc₁ indicate that the native form of this enzyme does not produce superoxide at detectable levels unless the electron flow through the cofactor chains is severely impeded and, consequently, the chains become saturated with electrons (i.e. cofactors remain in reduced state for a prolonged time). One classic example involves conditions where the outflow of electrons from the b-chain is blocked by antimycin at the level of the Q_i site. This traps hemes b in the reduced state. We reported earlier that isolated form of Rba. capsulatus cytochrome bc₁ under those conditions produces 17% of superoxide¹ at 100 mM NaCl or 26% at 400 mM NaCl (Fig. 3a) [11]. At the same time we noted that the generation of superoxide associated with the blockage of the Q_i site can effectively be enhanced by impeding the electron outflow from the c-chain. In the M183L mutant detectable levels of superoxide were observed even in the absence of antimycin, at higher concentrations of quinol and/or at high salt concentration (400 mM NaCl). In the presence of antimycin the levels of superoxide production were higher than the respective levels in the wild type [11].

We now extend this analysis to show that another mutant of the same type (M183K) exhibits a very similar pattern as M183L. Again, superoxide production can be observed even in the absence of antimycin at high QH₂ concentration and/or at high ionic strength: when the initial concentration of QH₂ is increased from 20 to 100 μM, superoxide is generated at 16% at 100 mM NaCl (not shown), while at 400 mM NaCl, 22.5% of superoxide is generated even with 20 μM initial QH₂ (Fig. 3b). In the presence of antimycin, the levels reach 30% at 100 mM NaCl or 35% at 400 mM NaCl (Fig. 3b) which is higher than in the wild type (Fig. 3a).

In those experiments increasing the concentration of quinol increases the extent to which the reverse Q_i to heme b_H reaction interferes with the forward reaction making the flow of electrons through the b-chain in the forward direction more difficult. On the other hand, increase in the ionic strength inhibits the interaction between cytochrome c and c₁ making the flow of electrons through the c-chain more difficult [34]. The former effect increases a probability of saturation of the b-chain with electrons that come from both the forward and reverse directions; the latter — increases a probability of saturation of the c-chain with electrons that come from the forward reaction at the Q_o site.

Clearly, testing those effects in various combinations on both forms of the c₁ knockouts (M183L and M183K) consistently indicates that the saturation of the c-chain with electrons when the b-chain is also saturated with electrons effectively increases the level of superoxide production by cytochrome bc₁.

Our earlier studies showed that + 2Ala motion knockout does not produce a statistically significant amount of superoxide under any conditions, even in the presence of antimycin. We related this with large increase in the density of the FeS cluster around the Q_o site and the resulting increase in the probability of occurrence of competitive reactions at the Q_o site that retain electrons within the protein instead of releasing them on oxygen. This was generally formulated in a model described in ref. [11]. The + 1Ala mutant, which as we show here has a density of the FeS cluster at the Q_o site decreased with respect to the + 2Ala yet significantly increased with respect to the wild type (Fig. 2), provides the very useful set of additional conditions with which the effect of change in the density of the FeS cluster on the level of superoxide production can be examined. The most straightforward prediction of our model is that the forms of cytochrome bc₁ with the + 1Ala insertion might show reduced levels of superoxide production when compared with the forms without this insertion if only the FeS domain is held at the Q_o site for sufficient time to effectively increase the contribution from competitive reactions.

Fig. 3c and d confirms this to be the case. Under the experimental conditions of this figure, the production of superoxide in + 1Ala mutant, as in the wild type, can only be observed in the presence of antimycin (Fig. 3c). However, the level of superoxide production is clearly diminished vs wild type, at both 100 mM and 400 mM NaCl (9.5% vs 17%, and 12.5% vs 26%, respectively). Likewise, an addition of + 1Ala mutation to the M183K c₁ knockout (M183K/+1Ala double mutant) generally decreases the level of superoxide production seen in the M183K mutant alone; in all cases where the production of superoxide was observed in M183K, superoxide was also produced in M183K/+1Ala albeit the overall percentage was significantly reduced (Fig. 3d). These results clearly confirm the correlation between the density of the FeS cluster at the Q_o site and the observable level of superoxide produced by cytochrome bc₁: the higher the FeS density, the lower the ROS level.

We note that the properties of + 1Ala dismiss the possibility that the suppression in ROS production is just due to the change in the steady-state quinol oxidation rate. This is because despite the substantial increase in the FeS cluster density at the Q_o site in + 1Ala, its steady-state turnover is comparable to the native system (Fig. 3), quite differently from + 2Ala, in which density of the FeS at that site is almost maximal and the enzymatic activity is much reduced.

3.3. Kinetic model of superoxide production by the Q_o site

Our kinetic model analyzed separately two sets of conditions: one encountered by the enzyme when semiquinone is formed in semiforward reaction; the other — when it is formed in semireverse reaction. The latter is shown in Fig. 4a.

Fig. 4.

Model of superoxide generation by cytochrome bc₁. a, Ubiquinol (QH₂) undergoes fast two-electron oxidation by oxidized FeS cluster and oxidized heme b_L (red arrows in the forward box). The long-lasting reduced state of heme b_L (when its reoxidation by heme b_H is impeded) allows for the single-electron reduction of quinone (Q) to semiquinone (SQ) (blue arrow in the semireverse box). This step is likely to be rate-limiting (denoted as a dotted frame of the box) and independent on the redox state and position of FeS cluster (grayed out dashed icon of FeS). Once SQ is formed (red hexagon with a dot) and the FeS head domain is at the Q_o site the reaction of SQ_o with molecular oxygen is out competed by either rapid oxidation of SQ_o by oxidized FeS cluster (blue arrow in the short circuit — SOD insensitive box) or rapid reduction of SQ_o to QH₂ by reduced FeS cluster (blue arrow in the completion of reverse reaction — SOD insensitive box). When FeS is out of the Q_o site at the time of SQ_o formation, oxygen can react with SQ_o and superoxide is formed (brown arrow in the superoxide production — SOD-sensitive box). Each of these reactions brings the enzyme to the state with oxidized heme b_L and oxidized FeS cluster and the cycle can be repeated. Color code of cofactors: red, reduced; black, oxidized. b, Quantitative analysis of the model presented in (a) requires the assumption that increasing the distance of FeS cluster from Q_o site decreases the electron transfer rates (k_ET FeS − Q_o) between the FeS cluster and Q/SQ_o/QH₂ (blue line) but increases electron exchange rates (k_ET FeS − c₁) with heme c₁ (red line). In the simulations this assumption was approximated by dividing the FeS trajectory into 3 regions: 1 — FeS exchanges electrons only with the Q_o site occupant, 2 — FeS cannot exchange electrons, and 3 — FeS exchanges electrons only with heme c₁. c, Simulated curves of steady-state reduction of cytochrome c, obtained from modeling of the semireverse mechanism as depicted in (a) and (b). In all forms, there is no SOD-sensitive cytochrome c reduction in the absence of antimycin (no Ant). In the presence of antimycin the simulated traces with (SOD) and without SOD (no SOD) diverge and percent of superoxide production can be calculated. d, Comparison of the experimentally obtained values (experiment) with those obtained from modeling semireverse or semiforward mechanisms (semireverse or semiforward, respectively). Color code of cofactors same as in (a).

We first asked a question whether the modeling can discriminate quantitatively the semiforward and semireverse mechanisms. If it cannot, we are unable to draw any conclusions about contribution of the particular mechanism to the overall production of superoxide. If, however, it does discriminate those two mechanisms and the modeling one of them follows the experimental data, then we get a strong indication as to which mechanism of superoxide production is dominant.

The model simulated steady-state kinetic curves of cytochrome c reduction by different forms of cytochrome bc₁ taking into account the following general assumptions: 1) all reactions are independent of each other; 2) the reactions in the b-chain are simplified to describe only the changes in the redox state of heme b_L; 3) the absence of antimycin is modeled by an addition of the reaction that allows rapid oxidation of heme b_L but not through the other components of the Q_o site (this reaction is omitted in the presence of antimycin); 4) changes in the density of the FeS cluster caused by the movement of the head domain are approximated to the transitions between the three states: Q_o, in-between, and c₁ states (1, 2, and 3 in Fig. 4b, respectively); 5) the transitions from each of the state (Q_o, in-between, and c₁) to its neighboring states are random (independent of each other and not influenced by the redox state of any cofactor); 6) in the Q_o state the FeS cluster can exchange electron with the occupant of the Q_o site, in the c₁ state the FeS cluster can reduce cytochrome c (because of the initial excess of substrates, oxidized cytochrome c and quinol; the electron transfer from cytochrome c back to the FeS cluster is neglected), in the in-between state the FeS cluster cannot exchange electrons with any cofactor; 7) spontaneous dismutation of superoxide is neglected; and 8) the effect of addition of superoxide dismutase in the ROS assay is modeled by eliminating the reaction of reduction of cytochrome c by superoxide anion.

Fig. 4c shows simulations obtained with the parameters found to best reproduce the whole set of experimental data. These simulations show reduction of cytochrome c by wild type and various mutants of cytochrome bc₁ when the set of conditions associated with the semireverse SQ_o formation (Fig. 4a) is considered. The model indicates no ROS generation in the “absence” of antimycin, but in its “presence” simulations clearly indicate ROS generation with the extent varying depending on which forms are analyzed. The simulated levels follow the order M183K (34%), M183K/+1Ala (21%), WT (18%), + 1Ala (9.6%), and + 2Ala (~ 1%), which is in remarkable agreement with the experimental data. More specifically, the model neatly reproduces two opposite effects: suppression of ROS production in the forms with increased density of the FeS cluster at the Q_o site (it shows almost no ROS in + 2Ala and a suppression caused by + 1Ala in both the WT and M183K backgrounds) and increase in the ROS production when the oxidation of the FeS cluster by cytochrome c₁ is suppressed (M183K).

When the set of conditions associated with the semiforward SQ_o formation was examined in the same way, the model could not reproduce the experimental data for all the mutants. In particular, as shown in Fig. 4d, the model could not produce the, revealed by experiment and obtained with the previous set of modeling, opposite effects of + 2Ala and M183K. With various parameters tested, the changes in ROS production for + 2Ala and M183K always run in the same direction (the model predicts increase in the level of ROS in both mutants).

It is thus clear that the model discriminates between semiforward and semireverse mechanisms and that the results of modeling of the semireverse mechanism are consistent with the experimental data for the wild type and all the tested mutants. This suggests that this mechanism might be dominant in the overall process of ROS generation by cytochrome bc₁.

4. Discussion

4.1. Correlation between the density of FeS at Q_o and the level of ROS

The results of this work are consistent with the notion that the presence of the FeS cluster at the Q_o site effectively diminishes probability of formation of superoxide. We show that not only the mutation (+ 2Ala) that arrests the FeS head domain for seconds at the Q_o site suppresses the ROS production to almost zero [11], but also a mutation (+ 1Ala) that prolongs the time this domain spends at the site (from native sub-milliseconds to milliseconds) [27] effectively decreases the level of ROS production when compared to the wild type (Fig. 3). We explain this correlation between the density of the FeS cluster at the Q_o site and the level of superoxide production with the model that assumes a competition between the internal reactions involving the cofactors and quinone at the Q_o site and the reaction of semiquinone with oxygen. This model was discussed in detail in ref. [11].

Inherent to this model is that it is exclusively kinetic: it assumes nothing more but changes in the reaction rates between the redox-active components of the Q_o site. Upon the FeS head domain movement, a transient increase in the distance between the FeS cluster and the remaining portion of the site slows the electron tunneling [35,36] between the FeS cluster and quinone (Q or QH or SQ) at the site. When this occurs, the probability that the FeS cluster participates in an immediate removal of SQ_o decreases, and, consequently, the probability of the reaction of SQ_o with oxygen to form superoxide increases.

The changes in electron tunneling rates constituted basis of the concept of transient “opening” of the Q_o site for the reaction with oxygen introduced originally in ref. [11]. Unlike what was suggested in ref. [24], this concept never referred to any specific gating mechanism by which enzyme would control generation of superoxide, for example by the formation of the physical gate for oxygen molecules to enter or exit the site through the FeS head domain moving out of the Q_o site. In fact, if one considers the ability of oxygen molecules to penetrate membrane and even protein matrix [37–40] it seems rather unlikely that the site can ever be completely sealed from molecular oxygen.

Our model assumes no change in the rate constant for reaction between SQ_o and oxygen. It only assumes that a contribution of other reactions at the Q_o site decreases steady-state level of SQ_o reflected as a decrease in the average level of produced ROS.

4.2. Majority of SQ_o prerequisite of ROS is formed via semireverse mechanism

Until now, the presented model has only described how the FeS cluster competes with oxygen for reaction with highly reactive SQ_o. It has not identified the reaction sequence leading to formation of SQ_o, which in the reversibly operating cytochrome bc₁ may involve either the semiforward single-electron oxidation of QH₂ by FeS or the semireverse single-electron reduction of Q by heme b_L. This important identification has proven possible when experimental data were quantitatively analyzed by the kinetic model describing either semiforward or semireverse mechanism following the logic discussed in Results.

The analysis clearly indicated that the dominant process in the superoxide production involves leaks of electrons from SQ_o formed in the semireverse reaction (Fig. 4), which is in line with recent mechanism proposed for the mitochondrial complex III [13,41]. This conclusion was drawn from the observation that only the model describing the semireverse mechanism was capable of reproducing the data for all tested conditions and mutants. Because in this mechanism the creation of SQ_o is independent on the actual position of the FeS cluster, the model predicted the progressive increase in the efficiency of competitive reactions involving the FeS cluster (decrease in the level of ROS) when its density at the Q_o site increases (+ 1Ala and + 2Ala mutants). At the same time, because only the oxidized FeS cluster can complete short-circuit reaction which adds to the measured SOD-insensitive turnover (Fig. 4a, short-circuit SOD-insensitive), the model predicted increase in the level of ROS when the oxidation of the FeS cluster is impeded (M183K).

The failure of the model describing the semiforward mechanism can be related to the fact that, unlike in the semireverse mechanism, the formation of SQ_o requires the presence of the oxidized FeS cluster at the Q_o site. Thus the rate of SQ_o generation and the rate of its elimination by the cofactors of the Q_o site (FeS and heme b_L) might be strongly correlated. As a result this model predicted similar direction of change in the level of ROS production when oxidation of the FeS cluster by cytochrome c is impeded (M183K) or the movement of FeS toward cytochrome c₁ is impeded (+ 2Ala). This, however, was not supported experimentally.

It should be noted that we do not discard the possibility that ROS production in cytochrome bc₁ is the effect of the combination of both semiforward and semireverse mechanisms. Only, as our data and modeling suggest, we maintain that the contribution of the semiforward mechanism is likely to be minor.

4.3. Concluding remarks with connotation to physiology

The models developed here are consistent with the notion that the uninhibited electron flow through the cofactors chains of cytochrome bc₁ involves the two-electron oxidation/reduction of quinol/quinone at the Q_o site which is unlikely to be a source of deleterious radicals. This is because the catalytic reactions appear to be fast enough that reaction of SQ_o with oxygen cannot effectively compete with them. However, severe inhibition of electron flow (such as that caused by antimycin or mutations shown here) that maintains long-lasting reduced states of cofactors (heme b_L and/or the FeS cluster) switches a number of additional one-electron reactions that can lead to short circuits or leaks of electrons out of the Q_o site [11,13,14,17]. Our experimental analysis of broad set of conditions combined with modeling revealed that formation of SQ_o by the reverse heme b_L to quinone electron transfer (semireverse mechanism) dominates in setting the steady-state levels of SQ_o that present a risk of generation superoxide. With constantly moving FeS head domain (kinetic studies indicate that the FeS cluster equilibrates with both of its partners on a millisecond timescale independently of its redox state and of the redox state of the partners [31]), the risk of electron escape from SQ_o to oxygen associated with an occurrence of a given heme b_L to quinone electron transfer depends on the actual position of the FeS head: it is much smaller when the FeS head domain occupies positions close/within the Q_o site and is much larger when it occupies positions remote from the site.

It follows from these scenarios, that not only states with over reduced hemes b (mimicked by antimycin) as commonly envisaged, but also states with over reduced hemes c (mimicked by the heme c₁ knockout mutants described here and in ref. [11]) present a potential risk for release of superoxide from complex III. The latter states refer not only to conditions where the oxidation of cytochrome c would be effectively inhibited at the level of cytochrome c oxidase, for example under low oxygen or inhibition by nitric oxide, but also to conditions where the pool of soluble cytochrome c is significantly reduced or damaged, for example during cytochrome c release in apoptosis ([4] and refs therein).

Obviously, all those states are interconnected with the redox state of the quinone pool and the magnitude of the membrane potential. For example, build up of QH₂ when oxidation of cytochrome c is inhibited can result in an increased pressure from the reverse reaction through the Q_i site that elevates the level of reduced hemes b. Also increase in the membrane potential may have similar effect of enforcing hemes b to remain reduced. On the other hand, as oxidized quinone is a substrate in semireverse mechanism of SQ_o formation, it becomes clear that a proportion of oxidized quinone in the membranous pool might influence the levels of ROS, even to the point that the redox state of the quinone pool could contribute significantly to the control of superoxide production by complex III and, thereby, to the control of mitochondrial redox signaling, as proposed by Brandt et al. [13,41].

It is quite possible that all those states gain physiological meaning under the resting metabolic state of mitochondria when the respiration rate slows and electrons have better chance to reside longer at individual redox-active sites. As discussed here, with the long-lasting reduced states of the FeS cluster and/or heme b_L and constantly moving FeS head domain (there is little reason to suppose that the mobility of the FeS head domain would be affected) there is a potential risk that superoxide can be generated by the Q_o site of complex III. This would be in line with a view that it is a decreased rather than an increased metabolic rate that leads to increased generation of ROS [42]. Much remains to be learned in this respect, but we believe that a dynamic description of reaction sequences at the Q_o site presented here sheds some light toward better understanding of the complex issues related with ROS generation at mitochondrial and cell levels.