Oxygen Sensitivity of Mitochondrial Reactive Oxygen Species Generation Depends on Metabolic Conditions

The mitochondrial generation of reactive oxygen species (ROS) plays a central role in many cell signaling pathways, but debate still surrounds its regulation by factors, such as substrate availability, [O2] and metabolic state. Previously, we showed that in isolated mitochondria respiring on succinate, ROS generation was a hyperbolic function of [O2]. In the current study, we used a wide variety of substrates and inhibitors to probe the O2 sensitivity of mitochondrial ROS generation under different metabolic conditions. From such data, the apparent Km for O2 of putative ROS-generating sites within mitochondria was estimated as follows: 0.2, 0.9, 2.0, and 5.0 μm O2 for the complex I flavin site, complex I electron backflow, complex III QO site, and electron transfer flavoprotein quinone oxidoreductase of β-oxidation, respectively. Differential effects of respiratory inhibitors on ROS generation were also observed at varying [O2]. Based on these data, we hypothesize that at physiological [O2], complex I is a significant source of ROS, whereas the electron transfer flavoprotein quinone oxidoreductase may only contribute to ROS generation at very high [O2]. Furthermore, we suggest that previous discrepancies in the assignment of effects of inhibitors on ROS may be due to differences in experimental [O2]. Finally, the data set (see supplemental material) may be useful in the mathematical modeling of mitochondrial metabolism.

The production of reactive oxygen species (ROS) by mitochondria has been implicated in numerous disease states including but not limited to sepsis, solid state tumor survival and diabetes (1). In addition, mitochondrial ROS (mtROS) play key roles in cell signaling (reviewed in (2;3)). There exist within mitochondria several sites for the generation of ROS, with the most widely studied being complexes I and III of the electron transport chain (ETC). However, there is currently some debate regarding the relative contribution of these complexes to overall ROS production (4)(5)(6)(7)(8)(9), and the factors which may alter this distribution. One such factor considered herein, is [O 2 ]. Estimates of physiological [O 2 ] within tissues (i.e. interstitial [O 2 ]) range from 37 µM down to 6 µM at 5-40 µm away from a blood vessel (10). More recently, EPR oximetry has estimated tissue [O 2 ] to be in the 12-60 µM range (11). In addition, elegant studies with hepatocytes have shown that O 2 gradients exist within cells, such that an extracellular [O 2 ] of 6-10 µM yields an [O 2 ] of ~5µM close to the plasma membrane, dropping to 1-2 µM close to mitochondria deep within the cell (12). In cardiomyocytes, at an extracellular [O 2 ] of 29 µM, intracellular [O 2 ] varied in the range 6-25 µM (13). Clearly, different tissues consume O 2 at different rates, so these gradients can vary considerably between tissue and cell types.
By definition, the generation of reactive oxygen species by any mechanism, is an O 2dependent process. However, measurements in intact cells have indicated that mtROS generation increases at lower O 2 levels (1-5 % O 2 ) (14). Proponents of an increase in mtROS in response to hypoxia suggest that under such conditions reduction of the ETC results in increased leakage of electrons to O 2 at the Q O site of complex III (14). Such a model posits that increased hypoxic ROS is a mitochondria-autonomous signaling mechanism, i.e. it is an inherent property of the mitochondrial ETC. Therefore mtROS generation should increase in hypoxia regardless of the experimental system being studied, including isolated mitochondria. In contrast to this hypothesis, we and others have demonstrated that ROS generation by mitochondria is a positive function of [O 2 ] across a wide range of values (0.1-1000 µM O 2 ) (15)(16)(17)(18), suggesting that signaling mechanisms external to mitochondria may be required to facilitate the increased hypoxic mtROS production observed in cells.
One limitation of our previous work (15) was that only a single respiratory condition was studied, namely succinate as respiratory substrate (feeding electrons into complex II) plus rotenone to inhibit backflow of electrons through complex I (5;7). The possibility exists that under different metabolic conditions, which may lead to differential redox states between the cytochromes in the ETC (19;20) (21)). Fig. 1 shows a schematic of the mitochondrial ETC, highlighting sites of electron entry resulting from various substrates, binding sites of inhibitors, and major sites of ROS generation. Fig. 2 shows the specific details of each experimental condition, indicating the predicted sites of ROS generation resulting from the use of each substrate/inhibitor combination. The legend to Fig. 2 (23). Authentic H 2 O 2 was added at the end of each experimental run to internally calibrate the fluorescent signal. Such a method ensures that the obtained signal truly reflects the net H 2 O 2 production and is not affected by scavenging due to enzymes such as catalase.
Incubations were carried out in mitochondrial respiration buffer (15), with oligomycin (1 µg/ml) present to enforce state 4 respiration. Where indicated, mitochondrial substrates and inhibitors were used at the following concentrations: glutamate (10 mM), malate (5 mM), succinate (10 mM), palmitoylcarnitine (1 µM), rotenone (1 µM), antimycin A (10 µM), malonate (2 mM), and were present from the beginning of incubations before mitochondrial addition. SOD (80U/ml) was present in all incubations to ensure rapid dismutation of O 2 •to H 2 O 2 , and to avoid scavenging of the former by reaction with nitric oxide (NO • ). This was a precaution, despite our previous observations that additional SOD was not necessary in this regard, and that NO • scavenging of O 2 •-(which would lead to peroxynitrite mediated tyrosine nitration) was not occurring in hypoxia (15). The latter is also unlikely because the K M for O 2 of all NOS isoforms is very high (6-24 µM) ( Table S1. VO 2 varied considerably between metabolic substrates. For example, a higher VO 2 was observed with complex II substrates (condition E) than with complex I substrates (condition A). The consensus view is that because fewer H + are pumped across the inner membrane when electrons enter at complex II, the ETC has to work faster in condition E (and thus consume more O 2 ) to maintain the same H + gradient as in condition A.
The  Furthermore, it is insufficient to merely subtract ROS as calculated above, since that flux was calculated from mitochondria respiring on succinate, and the flux through the respiratory chain (i.e., VO 2 ) was lower with glutamate plus malate. This lower electron flux has two opposing effects on ROS generation by complex III Q O : First, less electrons reach complex III, as shown by the VO 2 in condition A ( Fig. 2A) which was 39.3% of that in condition E (Fig. 2E). Second, this slower electron flux through complex III results in an increased dwell-time for the ubisemiquinone radical at the Q O site, which enhances ROS generation (32). To correct for this second effect, it is necessary to determine the relationship between the percentage of electrons diverted to ROS and the total electron flux (VO 2 ). Fig. 5  Notably, Fig. 5, which shows that the percentage of electron flux diverted to ROS increases as respiration slows down, might be misconstrued as demonstrating that mitochondrial ROS generation increases at low respiration rates (such as those caused by low [O 2 ]). However, as we previously discussed (15), even though a greater percentage of electrons may be diverted to ROS, the absolute number of electrons flowing through the respiratory chain and thus available for diversion, decreases by a far greater magnitude, such that the absolute number of ROS generated is lower. From both ROS detection and cell signaling perspectives, the parameter which matters is not the percentage of electrons diverted to ROS, but the absolute amount of ROS, which always decreases at low [O 2 ].
Complex I Backflow: The backflow of electrons through complex I causes ROS production at its downstream ubiquinone binding site (5;9), which can be calculated in two ways: First, ROS generation in the presence of succinate alone (condition D) includes ROS from both forwards and backwards electron flow. The contribution of backflow can be quantified by subtracting data obtained under the forwards flow only condition (i.e., succinate plus rotenone, condition E).
Secondly, palmitoyl-carnitine feedings electrons into the Q-pool can serve as an alternative source of electrons for complex I backflow. Similarly to the succinate data (condition D vs. E), the presence or absence of rotenone to prevent backflow can also be applied to palmitoyl-carnitine linked ROS generation (i.e., condition K vs. L), to infer the rate of ROS from backflow. Averaged data from these two calculations are shown in Fig. 4C, indicating a V MAX of 135, and an apparent K M of 0.9 µM O 2 . The occurrence and physiological importance of ROS from complex I backflow remains unclear (5;7;9), and the data in Fig. 4C indicate the importance of [O 2 ] in regulating this phenomenon. In addition, the ratio of complex I vs. complex II substrates is expected to play an important regulatory role in vivo, since forwards electron flow through complex I will effectively prohibit backflow (21;33;34).

ETFQOR:
The ETFQOR of β-oxidation is known to generate ROS (5). In the presence of palmitoyl-carnitine (condition K) ROS generation is from 3 sites: the ETFQOR, complex III Q O site, and complex I backflow. To account for backflow, it is necessary to consider the rate of palmitoylcarnitine linked ROS generation in the presence of rotenone (condition L). To account for ROS from the complex III Q O site under these conditions, a similar calculation is performed as above for the complex I FMN site, i.e. scaling of the complex III Q O site data using a correction factor that considers the effects of respiration rate on ROS generation at this site: Electron flux through complex III in the presence of palmitoyl-carnitine plus rotenone (VO 2 2.8, Fig. 2L) is 14.3% of the flux with succinate plus rotenone (VO 2 19.5, Fig.  2E). Furthermore, by reference to Fig. 5, this lower electron flux results in a 2.37 fold increase in the percentage of electrons donated to ROS (vs. that observed at maximal flux). Combining these correction factors, it is necessary to subtract 34% of the ROS generation from the complex III Q O site (condition E), to reveal the residual ROS from the ETFQOR. Data from this calculation are shown in Fig. 4D,  Notably, it has been observed that in skeletal muscle mitochondria, ROS generation under dual electron entry far exceeds the sum of rates with either complex I or complex II substrates alone (7). This may be due to different substrate preferences between muscle vs. liver mitochondria, which may in turn be related to differential expression levels of the various proteins of the oxidative phosphorylation machinery (35). Together these results highlight that patterns of ROS generation vary greatly between different tissues, and that tissue oxygenation is another factor which may influence specific pathways of mitochondrial ROS generation.

Effects of inhibitors:
The effect of mitochondrial inhibitors on ROS generation has been widely studied, but the influence of O 2 on these effects has not. In the current investigation two inhibitors, rotenone and antimycin A, were examined: Rotenone binds at the downstream Q site within complex I, increasing ROS generation at the upstream FMN site. However, another effect of rotenone is to block electrons from exiting complex I and proceeding via the Q pool to complex III (27). Thus rotenone decreases ROS generation from the complex III Q O site. The effects of rotenone on overall ROS generation will manifest as a balance of these two effects, and herein we estimated that ROS generation at these two sites is differentially O 2 sensitive (Figs. 4A vs. 4B). Comparing the rates of ROS generation in mitochondria respiring on complex I linked substrates in the presence (condition B) and absence (condition A) of rotenone (i.e. subtracting A from B) reveals an interesting pattern for the effect of rotenone on ROS generation as a function of [O 2 ] (Fig. 4F) Fig. 4G, indicating that the foldincrease in ROS resulting from antimycin A addition is greater at high O 2 levels (3-fold at 20 µM O 2 vs. 2-fold at 2 µM O 2 ). This is in agreement with a previous observation that the effect of antimycin A on mtROS was far greater under hyperoxic conditions (16 In addition to hyperoxia, the current data may have implications for the role of mitochondria in cell signaling during hypoxia. Specifically, mitochondria within cells have been proposed to increase their ROS generation during hypoxia, leading to the stabilization of hypoxia inducible factor 1α (HIF-1α) and the downstream expression of hypoxia-sensitive genes (47)(48)(49)(50)(51). The current data indicate that, under all of the metabolic conditions examined, mtROS decreased at low Interestingly, the site of ROS generation which has received most attention within the context of hypoxic signaling is complex III (14). It is possible that the position of complex III in the respiratory chain adjacent to the terminal cytochrome c oxidase, may render it more sensitive to redox events at the terminal oxidase (e.g. ischemia/hypoxia), compared to upstream complexes which are relatively more "cushioned" from such events (52). However, our results (Fig.  4) suggest that complex III is one of the least likely sites for ROS generation under hypoxic conditions. Thus we hypothesize that in hypoxia, ROS may originate from an as-yet unidentified mitochondrial source. Alternatively, since the current experiments highlight that the property of increased ROS in response to low O 2 is not autonomous to the mitochondrial respiratory chain, an additional signal external to the mitochondrion may be required to facilitate a hypoxic increase in mtROS within cells. Such a signal is likely absent from our isolated mitochondrial incubations. A third possibility is that differences in methodology (e.g. choice of fluorescent ROS probes) and definitions of "hypoxia" may account for the varied reports of hypoxic ROS generation in the literature (see (15) for discussion).
In  Fig. 1. Mitochondrial pathways of electron flow resulting from the substrates and inhibitors used in this study. Substrates used were: glutamate/malate -generates NADH via the TCA cycle, feeding into complex I; succinate -feeds electrons directly into complex II; palmitoyl-carnitine -feeds electrons into the ETC via acyl-CoA dehydrogenase as well as through the β-oxidation pathway. (For a more thorough explanation, refer to (21). Inhibitors used were: rotenone -inhibits at the downstream Q binding site of complex I (9); malonate -a competitive inhibitor of complex II (25;26); antimycin A -a complex III inhibitor which prevents electron flow to the Q I site of complex III, thus stabilizing QH • at the Q O site (6;28).    •are required to make one H 2 O 2 , so data on the y-axis were calculated by dividing 2 x ROS generation rate by respiration rate. Data are from mitochondria respiring on succinate plus rotenone (condition E), and are taken from the tables in the supplemental data. Error bars are eliminated for clarity. The right-most point is state 4 respiration, with the remaining points on the curve originating from changes in VO 2 due to titration of O 2 levels. Arrows highlight specific values of VO 2 referred to in the text, and the extrapolated values of % electrons diverted to ROS.