Synchronized whole-cell oscillations in mitochondrial metabolism triggered by a local release of reactive oxygen species in cardiac myocytes

Reactive oxygen species (ROS) and/or Ca2+ overload can trigger depolarization of mitochondrial inner membrane potential (DeltaPsim) and cell injury. Little is known about how loss of DeltaPsim in a small number of mitochondria might influence the overall function of the cell. Here we employ the narrow focal excitation volume of the two-photon microscope to examine the effect of local mitochondrial depolarization in guinea pig ventricular myocytes. Remarkably, a single local laser flash triggered synchronized and self-sustained oscillations in DeltaPsim, NADH, and ROS after a delay of approximately 40s, in more than 70% of the mitochondrial population. Oscillations were initiated only after a specific threshold level of mitochondrially produced ROS was exceeded, and did not involve the classical permeability transition pore or intracellular Ca2+ overload. The synchronized transitions were abolished by several respiratory inhibitors or a superoxide dismutase mimetic. Anion channel inhibitors potentiated matrix ROS accumulation in the flashed region, but blocked propagation to the rest of the myocyte, suggesting that an inner membrane, superoxide-permeable, anion channel opens in response to free radicals. The transitions in mitochondrial energetics were tightly coupled to activation of sarcolemmal KATP currents, causing oscillations in action potential duration, and thus might contribute to catastrophic arrhythmias during ischemia-reperfusion injury.

mitochondrial permeability transition, but on the other acting as second messengers that protect cells against injury (4)(5)(6). Mitochondria are a major site of physiological ROS production in the cardiomyocyte, with approximately 1 to 5% of the electrons flowing through the electron transport chain leaking into the production of ROS (7,8). The negative effects of ROS on metabolism are evident in several studies showing rapid and spatiotemporally heterogeneous discharge of ∆Ψ m in response to oxidative stress (1,2) and/or Ca 2+ overload (9), including protocols employing laserinduced photo-oxidation (10,11) and mitochondrial ROS-induced ROS release (12). In light of our previous work showing that substrate-deprivation can initiate synchronized oscillations of mitochondrial redox and membrane potential (13), and that a diffusible cytoplasmic messenger may be involved (14), the present study tests whether similar global self-organizing behavior can be triggered by a highly localized perturbation of a few mitochondria among the thousands packed within the cardiac myocyte. We demonstrate that ROS release and mitochondrial depolarization in less than 1% of the volume of the cell can trigger spatiotemporally synchronized oscillations in ∆Ψ m , ROS production, and mitochondrial redox potential throughout the entire volume of the cell. Close coupling of the metabolic responses to cardiac electrical excitability highlights the importance of this form of intraorganellar communication in determining whole-cell function.

Experimental Procedures
Cardiomyocyte isolation. All experiments were carried out at 37°C on freshly isolated adult Retention of calcein in the mitochondrial matrix was tested by loading myocytes for 20 min with 2 µM calcein-AM at room temperature. After dye loading the cells were resuspended in the experimental solution used for imaging. Intracellular esterase action then resulted in loading of both the cytoplasmic and mitochondrial compartments of the cell. Calcein-AM desterification was allowed to proceed for at least 1 h before imaging. To reduce the contribution of the cytoplasmic component to the fluorescence images, myocytes were patch-clamped with fluorophore-free pipet solution, which permitted diffusion of the cytoplasmic calcein into the large pipet volume.
Image acquisition and analysis. Images were recorded using a two-photon laser scanning microscope (Bio-Rad MRC-1024MP) with excitation at 740nm (Tsunami Ti:Sa laser, Spectra-Physics). Owing to the overlap in the cross sections for two-photon excitation of the three fluorophores of interest (20) (NADH, CM-DCF, TMRE), this wavelength permitted recording of redox potential, ROS production and ∆Ψ m simultaneously. The red emission of TMRE was collected at 605±25nm and the green emission of CM-DCF was recorded at 525±25nm. NADH emission was collected as the total fluorescence <490nm. At 3.5s intervals, 512 x 512 pixel 8-bit grayscale images of the three emission channels were collected simultaneously and stored. The total illumination time was 3.0-s per image, unless otherwise specified.
Average power from the Ti:Sa laser was 1000mW and the pulse bandwidth was ~12 nm, corresponding to <60 fs pulse duration at 80MHz repetition rate. This excitation was attenuated by 6 state excitation of local fluorophores (21,22). Local photon-induced ROS elaboration, reinforced by the restricted irreversible depolarization of ∆Ψ m , allowed us to perturb a small region of the myocyte to look for propagating effects. In the absence of the local perturbation, with few exceptions (see results), myocyte behavior was stable over the duration of the experiments.
For visualization of the spatio-temporal responses of TMRE and CM-DCF presented in figures 1,2,4,6,7 a 2 to 3 pixel wide line was drawn along the length of the myocyte (as shown in To quantitatively determine whether mitochondrial depolarizations were spatially synchronized in the presence of glucose or inhibitors (Fig. 1), we applied grid analysis to the 2D images. A binary mask of the cell TMRE fluorescence was made and the cell area, excluding nuclei, was divided into small squares approximately the size of individual mitochondria (~2µm x 2µm).
The average fluorescence within each grid object was measured and histograms were made of the distribution of fluorescence in polarized and depolarized mitochondria, giving two major peaks of fluorescence intensity. A cutoff value halfway between these peaks was then used to determine if a given grid object was classified as "polarized" and the fraction of polarized mitochondria with respect to the total number of objects at time zero was calculated for the image series. The initial value of ~80% in most experiments reflects an underestimation of the total number of polarized mitochondria due to overlap between the histogram distributions, causing some objects in the polarized population to fall below the cutoff.

Global transitions of mitochondrial energetics triggered by local perturbation of mitochondrial function
Under otherwise normal physiological conditions (normoxic cells in the presence of external substrates), we tested whether a highly localized metabolic perturbation could have widespread effects on the mitochondrial network of the myocyte. The thin optical section excited by the twophoton laser (23) was employed to perturb a small fraction (<0.3% of the total cell volume) of the mitochondrial population while monitoring the behavior of the remainder of the cell. After collecting 10-20 control images, an 8.7 x 8.7 µm region of the cell was excited in a single flash ( Fig. 1A). This resulted in a rapid, but not instantaneous (exponential time constant 18.2s) (see Fig.   2C), depolarization of ∆Ψ m in the flashed region (see white square in Fig. 1A). Thereafter, ∆Ψ m remained depolarized in the flashed area throughout the experiment (Fig. 1A). Unflashed control cells in the same field remained polarized throughout the experiment (Fig. 1C, D).
In contradistinction to the response in the flashed region, mitochondria throughout the rest of the cell were initially unaffected, but began to oscillate in a synchronized manner after a substantial delay. The spatiotemporal pattern of mitochondrial depolarization and repolarization can be readily appreciated from the time-line images created from the entire image sequence ( Fig. 1B; the time-line image is not a "line-scan" image, but is a 2D representation of the full 3D time series of the experiment, as described in methods) and by inspection of the image montage for a single oscillation ( Fig. 2A; from the oscillation marked with an asterisk in Fig. 1B).
Plots of whole-cell fluorescence illustrate that synchronized and periodic transitions in ∆Ψ m and the NADH redox pool (Fig. 2D) were associated with bursts of mitochondrial ROS production Close inspection of individual cycles of ∆Ψ m depolarization-repolarization revealed that the rapid depolarization phase (4.2±0.4 s; n=15; 10 experiments) was followed by an exponential repolarization with fast (14±1.34 s; n=14) and slow (111±4.9 s; n=6) components, until another sudden depolarization initiated the next cycle (Fig. 2D). The average period of oscillation was determined to be 104±5 s (n=66 periods analyzed in 15 cells; 10 experiments). By analyzing myocytes with fully polarized mitochondria before the flash, and almost fully depolarized during a transition, we calculated that the relative decrease in ∆Ψ m during depolarization was ~60%.

Role of mitochondrial ROS in the initiation and propagation of cell-wide mitochondrial oscillations
The correlation between ROS and mitochondrial instability was supported by the observation that a distinct threshold of ROS accumulation was necessary to observe global oscillations in ∆Ψ m . Figure 3 shows the normalized ROS production as a function of time for representative cells. In the majority of myocytes exposed to a flash (Fig. 3A), total cellular CM-DCF fluorescence soon rose to a level more than ~20% above baseline, and oscillations in ∆Ψ m occurred contemporaneously. In the few cells that were flashed, but showed no oscillations ( Consistent with the results obtained with rotenone, and in accord with previous studies (18,25,26), inhibition of the oxidation of ubiquinol to ubisemiquinone and electron transfer to the FeSIII center of complex III with myxothiazol (see Q cycle of Fig. 4A) also resulted in suppression of triggered ∆Ψ m oscillations (Fig. 4D), and significantly decreased of ROS production (Fig. 4E), either added acutely or after preincubation.
Antimycin A, which is known to block normal electron flow through complex III, but enhance O 2 .production as a result of accumulation of ubisemiquinone(7,18) (Fig. 4A), markedly increased ROS accumulation (by ~2-fold) in response to a flash (Fig. 4E). In the presence of antimycin, the flash induced a rapid release of TMRE from the mitochondrial matrix into the cytoplasm ( Fig. 4C; evident as an increase in the diffuse distribution of the dye) and, within minutes, caused rigor contracture, indicating a drop in cellular ATP to µM levels (27). This response supports the abovementioned evidence that ROS generated at complex III can mediate mitochondrial uncoupling, although in this case, the ROS concentrations were overwhelming and the depolarization irreversible.
Through a mechanism similar to that of myxothiazol, inhibition of O 2 .generated by the Qcycle can also be achieved by reduction of the FeSIII electron acceptor of complex III, which can occur upon downstream inhibition of the electron transport chain (18,24,26). This was tested by blocking cytochrome oxidase (complex IV) with cyanide ( Fig. 4A). As predicted, acute addition of cyanide abruptly stopped flash-triggered oscillations, allowed ∆Ψ m to be restored (Fig. 4D), and significantly decreased ROS production ( Fig. 4E), again supporting the conclusion that complex III was the major site of ROS production during metabolic oscillation.
The results obtained with cyanide indicated that flux through the electron transport chain, and a ready supply of downstream electron acceptors were necessary for the mitochondrial ROS bursts occurring during the oscillations. Hypothetically, other inhibitors that decrease electron transport and push respiration toward state 4 should also decrease ROS production in this context.
This held true when respiration was slowed by inhibiting the F 1 F 0 ATPase with oligomycin ( Fig.   4A,E) or when the adenine nucleotide translocator (ANT) was inhibited with bongkrekic acid (BKA; Taken together, these findings suggest that the self-sustaining bursts of ROS production triggered by a prior local oxidant stress require O 2 .production at complex III, and that mitochondrial ROS are intimately involved in the oscillatory and propagation mechanisms. Consistently, BKA blocked the cell-wide mitochondrial oscillations by keeping ROS production below threshold levels. By inhibiting ANT, which may be part of a permeability transition pore (PTP) complex, BKA has been used previously as a PTP blocker. Therefore, we further investigated whether the PTP contributed to the observed mitochondrial ∆Ψ m depolarizations.

Absence of PTP opening and mitochondrial Ca 2+ overload in mechanism of cell-wide mitochondrial oscillation
The results described above indicate that beyond a threshold level, mitochondrially-produced The oxidation of CM-DCF by ROS should ostensibly cause an irreversible increase in fluorescence and record the history of ROS production in the cell. However, in some cases, we observed a decrease in CM-DCF fluorescence after each burst of ROS production (e.g., Fig. 2E).
This anomalous behavior of the ROS signal was absent, and CM-DCF fluorescence increased in a stepwise manner, when TMRE was not loaded during the experiment, indicating that it was due to fluorescence resonance energy transfer (FRET) (19) between CM-DCF and TMRE as donor and acceptor fluorophores, respectively (see online supplemental material). This result also showed that the CM-DCF is retained inside the mitochondrial matrix during ∆Ψ m oscillations, while the TMRE moves in and out.
Ca 2+ overload has been reported to be a cofactor in the opening of the PTP (see (1) for a review), and it also may be a messenger that can propagate mitochondrial transitions (9). Although no sarcomere shortening was evident in the present experiments, and the myocytes were studied under quiescent, minimally Ca 2+ -loaded conditions, we tested whether suppression of sarcoplasmic reticulum (SR) or mitochondrial Ca 2+ handling influenced flash-induced mitochondrial oscillations.
Ruthenium red, an inhibitor of the Ca 2+ -uniporter in mitochondria (31) and of the SR Ca 2+ -release channel (32), did not prevent the flash-induced or spontaneous synchronized transitions in mitochondrial activity, either acutely, or when the flashes were given after 30 min of preincubation with the inhibitor (see supplemental material).
Thapsigargin, an inhibitor of the SR Ca 2+ ATPase which effectively depletes intracellular Ca 2+ stores in myocytes, also did not stop mitochondrial oscillations (see supplemental material).
Finally, extensive buffering of intracellular Ca 2+ with 1 mM EGTA did not affect flash-induced oscillations (see Fig. 7), confirming that Ca 2+ was not a key factor in triggering the mitochondrial transitions in our experiments.

Anion channel inhibitors block initiation of cell-wide mitochondrial oscillations, but enhance local matrix ROS accumulation in the flashed area
To investigate whether a channel distinct from the classical PTP mediated mitochondrial depolarization and, perhaps, O 2 .efflux from the matrix, we tested the effects of reported inhibitors of inner membrane anion channels (IMAC) on the response to local oxidant stress.
The isoquinoline carboxamide PK11195 specifically binds to the peripheral benzodiazepine receptor that localizes to mitochondria, and is one among several classes of amphipathic inhibitors of IMAC (33). DIDS, which also inhibits IMAC (34), is a stilbene-2,2'-disulfonate most frequently used as an inhibitor of Clchannels and/or anion transporters. Either of these inhibitors prevented the cell-wide synchronized ∆Ψ m depolarizations and ROS bursts after a flash ( Fig. 6A and 6C), despite the fact that, within the flashed zone, mitochondria depolarized irreversibly and ROS production was accentuated ( Fig. 6B and 6D). In fact, the normalized CM-DCF signal in the flashed region was

Oscillations in mitochondrial energetics drive cyclical changes in the cardiac action potential
The oscillatory uncoupling of mitochondria depletes cellular ATP levels and drives the activation of ATP-sensitive K + (K ATP ) channels in the sarcolemma. This will, in turn, produce cyclical changes in the action potential of the cardiomyocyte (13). We directly demonstrated this effect on cellular electrical excitability by recording action potentials or K + currents using whole-cell patch clamp while simultaneously imaging myocytes during flash-triggered oscillations in ∆Ψ m (Fig. 7). Action potentials (recorded in current clamp mode) shortened dramatically in coincidence with cell-wide depolarization of ∆Ψ m (Fig. 7A) and relengthened during the recovery phase of the cycle (Fig. 7B). Under voltage clamp conditions, the current-voltage relation of the oscillatory sarcolemmal current was consistent with the activation of K ATP current (i.e., weakly inwardly rectifying and having a reversal potential near the equilibrium potential for K + ) and was synchronized with the mitochondrial depolarizations ( Fig. 7C and 7D). Localized transitions in ∆Ψ m (e.g., limited to the flashed area) did not induce the drastic changes in action potentials (not shown).

Discussion
The main contributions of the present work are that: does not immediately spread to neighboring mitochondria, but pushes the system towards instability; eventually evolving into a synchronized oscillatory response.

Events occuring during a metabolic oscillation
The cyclical nature of the triggered mitochondrial response allowed us to examine the dynamic relationship between ∆Ψ m , NADH, and ROS production to gain insight into the underlying mechanism. Close inspection of the signals revealed that ∆Ψ m depolarization and a rapid increase in the rate of mitochondrial ROS production occurred in concert with oxidation of the redox pool ( Fig.   2D and 2E). The peak rate of ROS production occurred when mitochondria were rapidly uncoupling

Sites of mitochondrial ROS production and release
Both the rate of ROS production and the synchronized cell-wide transitions in ∆Ψ m were suppressed by interventions that would be expected to decrease O 2 .generation from the Q-cycle at complex III of the respiratory chain. It is interesting to note that ROS generation from complex I appeared to play little role in the setting of oxidative stress in the intact normoxic cells, based on the observation that rotenone or reduction of the electron transport chain by downstream inhibition (myxothiazol, cyanide, oligomycin, or ANT inhibition) suppressed ROS production ( Fig. 4A and 4E). These findings are consistent with earlier studies demonstrating that reduction of the redox chain has opposing effects on ROS production at complex I versus complex III (18,24); i.e., reduction of complex I enhances ROS production from a highly electronegative component of this complex (38)  radical is involved in the synchronization mechanism (Fig. 4F). While O 2 .appears to be strongly implicated in the oscillatory mechanism, many other molecules are likely to be changing simultaneously (e.g., P i , pH, ATP/ADP ratio, etc.), and may also be implicated in the response. This possibility, and the role of other cytoplasmic factors including substrate selectivity and ROS scavenging will require further investigation.
Since is trapped in the matrix when anionic pathways are blocked ( Fig. 6B and 6D), a possibility that has been suggested previously (6). An alternative explanation for suppression of propagation of the response outside of the flashed region could be that the anion transport inhibitors blocked the target of ROS-induced depolarization (which we hypothesize is the IMAC) in the rest of the cell.

Inner membrane anion channels and ∆Ψ m depolarization.
Our findings indicate that neither Ca 2+ nor the classical CsA-sensitive, large conductance  Fig. 4E).
We also reported a lack of effect of CsA on spontaneous oscillations in ∆Ψ m in substratedeprived myocytes (45), which were also shown to be reversibly suppressed by the benzodiazepine receptor ligand PK11195 (46), suggesting a mechanistic link with the present results. PK11195, along with many other amphipathic compounds (e.g., amiodarone, amitriptyline, dihydropyridines) and anion transport inhibitors, inhibit IMAC in isolated mitochondria (33,34). Patch-clamp studies of isolated mitoplasts demonstrated that an outwardly rectifying current was the predominant component of background conductance (47), and numerous single channel studies have provided evidence that anion channels are present on the inner membrane, the most common being the 108 pS (or "centum-picosiemen") anion channel (48). Inhibitors of IMAC in isolated mitochondria (33,34) also block inner membrane anion channel activity in single channel recordings(48). In particular, Kinnally et al, have shown that both the 108 pS anion channel and the larger "multiconductance" channel (MCC), argued to be the single channel equivalent of the PTP, are blocked by mitochondrial benzodiazepine receptor (mBzR) antagonists(48). Discrimination between these channel types is complicated by the similar actions of the mBzR compounds; however, the MCC is exquisitely sensitive to cyclosporin A (49), which had no effect in our experiments. The mBzR has been reported to be an 18-kDa protein that is tightly associated with the outer membrane voltage-

Significance of synchronized oscillations in mitochondrial energetics for cardiac physiology.
Previous evidence indicated the existence of oscillations in K ATP membrane currents and in the redox state in metabolically-compromised cardiac cells (13,14). Clearly, cell-wide transitions in mitochondrial energetics drive the cyclic activation of ATP-sensitive potassium currents, producing dramatic cyclic changes in the duration of the cardiac action potential during flash-induced oscillations in cardiomyocytes ( Fig. 7A and 7C).
This effect will potentially introduce both temporal and spatial electrical dispersion in individual cells or regions of the myocardium, greatly increasing the susceptibility to fatal ventricular arrythmias (13,46). Therefore, these results are likely to be relevant to the behavior of heart cells during ischemia/reperfusion, where ROS production plays a key role in cell injury.
Conditions would particularly favor mitochondrial oscillation during reperfusion, when substrate is restored and a burst of ROS production occurs, and, in fact, oscillations in whole heart NADH have been observed upon reoxygenation (52). The results also point out that the opening of the IMAC channel under conditions of oxidative stress could possibly be a precursor to the opening of the PTP.
If this early step can be interrupted or reversed, it could allow cells to avoid taking the irreversible path leading to necrotic or apoptotic cell death following ischemia and reperfusion.
In a broader context, our studies contribute to understanding how oxidative phosphorylation is coordinated among mitochondria during oxidative stress in spatially-structured metabolic networks (53,54). Since the mechanisms described should be active in all cell types containing mitochondria, the implications for normal as well as for pathophysiological cell function are universal (55-59).         ROS production eventually slows in the depolarized state as the source of electrons (NADH) is depleted (Fig. S1B: arrow 3). During repolarization, the process then reverses as the TMRE reenters the matrix, and the CM-DCF signal declines since its fluorescence energy is transferred to the acceptor (Fig. S1B: arrow 4).
During large ∆Ψ m depolarization, the exit of TMRE from mitochondria into the cytoplasm leads to quenching of the dye by an incompletely understood mechanism. The rapid redistribution of TMRE into the cytoplasm after depolarization of ∆Ψ m can be transiently detected in the nucleus (used as a convenient location to measure the nonmitochondrial intracellular component) (Fig. S2).
In a control experiment, we triggered whole cell mitochondrial oscillations with a laser flash in isolated cardiomyocytes in the absence of TMRE and CM-DCF. The rationale of this control was to demonstrate that the observed oscillations are due to intrinsic metabolic mechanisms and rule out the putative role of artifacts due to the photosensitizing effects of the fluorophores. Figure S3 depicts the whole cell oscillations in NADH after a laser flash. Figure S4 illustrates the absence of an effect of Ruthenium Red or Thapsigargin on the flash-triggered whole cell mitochondrial ∆Ψ m oscillations.
The link activated by Figure S5 shows a movie of a typical experiment of a flashinduced whole cell mitochondrial oscillations. Figure S1. Fluorescence resonance energy transfer between TMRE and CM-DCF, and the relationship between mitochondrial ROS production and ∆Ψ m depolarization.