Photo-Induction and Automated Quantification of Reversible Mitochondrial Permeability Transition Pore Opening in Primary Mouse Myotubes

Opening of the mitochondrial permeability transition pore (mPTP) is involved in various cellular processes including apoptosis induction. Two distinct states of mPTP opening have been identified allowing the transfer of molecules with a molecular weight <1500 Da or <300 Da. The latter state is considered to be reversible and suggested to play a role in normal cell physiology. Here we present a strategy combining live-cell imaging and computer-assisted image processing allowing spatial visualization and quantitative analysis of reversible mPTP openings (“ΔΨ flickering”) in primary mouse myotubes. The latter were stained with the photosensitive cation TMRM, which partitions between the cytosol and mitochondrial matrix as a function of mitochondrial membrane potential (ΔΨ). Controlled illumination of TMRM-stained primary mouse myotubes induced ΔΨ flickering in particular parts of the cell (“flickering domains”). A novel quantitative automated analysis was developed and validated to detect and quantify the frequency, size, and location of individual ΔΨ flickering events in myotubes.


Introduction
The mitochondrial permeability transition pore (mPTP) is a non-selective channel located in the mitochondrial inner membrane (MIM) and its opening (''permeability transition'') was first described 40 years ago [1][2][3][4]. In the open state, the mPTP allows ions and solutes up to a size of 1500 Da to passively diffuse over the MIM, leading to a rapid collapse of the highly inside-negative electrical potential (DY) across this membrane. The probability of mPTP opening is increased by elevated calcium concentrations in the mitochondrial matrix ([Ca 2+ ] m ), but other factors such as reactive oxygen species (ROS), pH, and DY also regulate this [5]. Permeability transition is involved in apoptotic and necrotic cell death, for instance during ischemia-reperfusion and muscular dystrophies due to collagen VI or laminin-2 deficiencies [6][7][8].
Currently, the molecular identity of the mPTP remains obscure. Although the adenine nucleotide translocase (ANT) and voltage-dependent anion channel (VDAC) were suggested as potential mPTP structural proteins, genetic studies disproved such a function for VDAC and revealed that ANT acts as a regulator of permeability transition [9][10][11]. Similarly, the mitochondrial phosphate carrier (PiC) was proposed as a mPTP structural protein although decreasing PiC expression up to 80% by RNA interference strategies did not affect mPTP opening [12]. Recent studies suggest that the mitochondrial F o F 1 -ATP synthase constitutes a structural component of the mPTP [13,14]. A well-characterized mPTP modulator is the mitochondrial matrix protein Cyclophilin D (CypD), which increases the probability of Ca 2+ -dependent mPTP opening. The immunosuppressant cyclosporin A (CsA) is known to inhibit mPTP opening via CypD and thereby desensitizing the mPTP to Ca 2+ -stimulated opening [15]. This property makes CsA a widely used experimental tool to demonstrate involvement of mPTP opening in mitochondria-associated cellular phenomena.
Interestingly, during opening the mPTP can assume either a low-or a highconductance state. In the low-conductance state the mPTP has a MW cut-off below 300 Da and thus only allows passage of small ions including H + and Ca 2+ . Additionally, when in low-conductance mode, the mPTP opens transiently (''flickering'') and mitochondrial swelling is absent [16,17]. In the highconductance state the mPTP displays a much higher cut-off (below 1500 Da) and opening is permanent, resulting in sustained DY depolarization, mitochondrial swelling/rupture and cell death [16,17].
Various methods have been described to study mPTP opening. For instance, permeability transition can be monitored in isolated mitochondria by quantifying the extent of mitochondrial swelling (measuring absorbance), mitochondrial Ca 2+ retention capacity (CRC) or DY depolarization [15,18,19]. Although mitochondria are highly accessible by the above strategies, a major limitation of these techniques is the lack of a cellular context. This means that cytosolic factors that potentially regulate mPTP opening are absent. Moreover, isolation of mitochondria from tissues and cells significantly alters their structure, electrical connectivity and function [20]. In intact cells, permeability transition can be monitored using cationic fluorescent probes such as methyl (TMRM) or ethyl (TMRE) esters of tetramethylrhodamine, which accumulate in the mitochondrial matrix in a DY-dependent manner [21][22][23][24].
Although mPTP-dependent flickering of DY has been observed in studies of isolated mitochondria and intact cells [25][26][27][28], to the best of our knowledge no automated quantitative method for their combined spatial and temporal analysis exists. This precludes unbiased statistical comparison of reversible mPTP opening with respect to their spatiotemporal properties under varying experimental conditions. Here we present an integrated experimental and computational approach for automatic detection, spatial visualization and frequency analysis of DY depolarization and repolarization in primary mouse myotubes. Both the number of DY depolarizations/repolarizations were dramatically reduced by the mPTP inhibitor CsA, strongly suggesting that they reflect reversible mPTP opening. In contrast, inhibition of the F o F 1 -ATP synthase by oligomycin A increased the number of DY depolarizations/repolarizations. Both inhibitors reduced the spatial dimensions of the DY depolarization/repolarization region, suggesting that they affect the electrical connectivity of (the) mitochondrial (sub)network(s).

Animals and housing conditions
Mice were bred with a mixed 129/sv x C57BL/6 background. All animals received a standard rodent diet ad libitum and were maintained at 21.0˚C and 60% humidity, and with a light/dark (12 h/12 h) cycle. All breeding and experiments were approved by the Animal Experimentation Committee at the Radboud University Medical Center, in accordance with Dutch laws and regulations regarding animal experimentation.

Primary myoblast cell culture
Coverslips (24 mm) were first coated with 20 mg/ml fibronectin (Roche Diagnostics GmbH, Mannheim, Germany) for one hour and subsequently with 10 ml Matrigel to create a ''Matrigel-spot'' for the cells. To induce fusion, 5,000 primary mouse myoblasts were seeded onto the Matrigel-spot in 10 ml DMEM-HG culture medium and after attachment (10 min) cultured with 3 ml DMEM-HG differentiation medium supplemented with 2% HS and 1% p/s (both PAA Laboratories) for 3 days at 37˚C (95% air, 5% CO 2 ).

Microscopy imaging of TMRM-stained mouse myotubes
To visualize mitochondria and detect DY changes, myotubes were loaded with 30 nM tetramethylrhodamine methyl ester (TMRM, Invitrogen) in DMEM-HG differentiation medium for exactly 25 minutes. Under these conditions TMRM operated in non-quenching mode. Then cells were washed and transferred to a HEPES-Tris (HT) medium (132 mM NaCl, 4.2 mM KCl, 1 mM CaCl 2 , 1 mM MgCl 2 , 5.5 mM D-glucose and 10 mM HEPES, pH 7.4). The coverslips were mounted in an incubation chamber and placed on the stage of an inverted microscope (Axiovert 200 M, Carl Zeiss, Germany). For noise analysis, empty coverslips were mounted on the microscope and HT-buffer containing different concentrations of TMRM (0-80 nM) was used. TMRM was excited at 540 nm using a monochromator (Polychrome IV, TILL Photonics, Gräfelfing, Germany) and a Zeiss 406/1.3 NA Plan NeoFluar objective. Fluorescence light was directed using a 560DRLP dichroic mirror (Omega Optical Inc, Brattleboro, VT, USA) and a 565ALP emission filter (Omega Optical Inc.) onto a CoolSNAP HQ monochrome CCD-camera (Roper Scientific, Vianen, The Netherlands). The microscopy hardware was controlled using Metafluor 6.0 software (Universal Imaging Corporation, Downingtown, PA, USA) and images were recorded every 2 seconds for a period of 20 minutes using a 300 ms image acquisition time.
Electron microscopy 5,000 mouse primary myoblasts were seeded onto the Matrigel-spot in 10 ml DMEM-HG culture medium and after attachment (10 min) cultured with 3 ml DMEM-HG differentiation medium for 3 days at 37˚C (95% air, 5% CO 2 ). Then myotubes were washed with PBS and fixed in 2% glutaraldehyde buffered with 0.1 M sodium cacodylate pH 7.4. After 1 hour of fixation the myotubes were postfixed in 1% osmium tetroxide in Palade's buffer pH 7.4 with 1% potassium hexacyanoferrate (III)-trihydrate. After dehydration in ethanol and propylene oxide, tubes were embedded in Epon. Semi-thin, 1 mm thick transverse sections were stained with 1% toluidine blue. Ultrathin sections were stained with uranyl acetate and lead citrate and examined in a JEOL TEM1010 transmission electron microscope.

Data analysis
Automated image processing and analysis were carried out using Image Pro Plus 6.3 software (Media Cybernetics, Silver Spring, MD, USA). Additional data processing, visualization and statistical analysis were performed using custom scripts written in Matlab 6.1 (Mathworks, Natick, Massachusetts, U.S.A.). Unless stated otherwise, data is presented as mean ¡ SD (standard deviation). The pvalues reported in the results section correspond either to a two-way ANOVA (with two factors: manual vs. automated analysis and CsA treated vs. untreated) or one way ANOVA (1 factor with three levels: untreated, CsA or OLI treated cells). In the latter case, the ANOVA was followed by a Tukey post-hoc procedure.

Photo-induced DY flickering in primary mouse myotubes
Current evidence supports a mechanism in which TMRM illumination stimulates the cyclic photo-generation of mitochondrial singlet oxygen molecules ( 1 O 2 ), which increase the probability of mPTP opening [28]. As a consequence, a cycle of DY depolarization and repolarization is induced leading to cyclic mitochondrial TMRM efflux and re-uptake (Fig. 1A). To monitor this cycle, primary myotubes were stained with the fluorescent cation TMRM and visualized using an epifluorescence video microscope (Fig. 1B). TMRM partitioning between the cytosol and mitochondrial matrix primarily depends on the magnitude of the electrical membrane potential (DY) across the mitochondrial inner membrane (MIM). Fields of views (FOVs) were chosen in such a way that they mainly contained myotubes to prevent interference from myoblasts. Myotubes are easily identified because they contain multiple nuclei ( Fig. 1B; indicated by ''N''), whereas myoblasts do not. For each time lapse recording (''image stack'') 600 images were routinely acquired using an interval of 2 s and an illumination time of 300 ms. The recorded images had a bit-depth of 12 and were converted to TIFF format for subsequent visualization and processing. Visual inspection highlighted the presence of localized TMRM intensity fluctuations (''flickering''; Movie S1), which typically were initiated within 20 seconds after the start of recording (i.e. after 10 images were acquired). The total integrated cellular and nuclear TMRM signal both slowly decreased (Fig. 1C) to 75-80% with respect to their initial TMRM signal. This suggests that part of the TMRM intensity is lost due to photo bleaching. The fact that both the mitochondrial and nuclear TMRM intensity decreased at a similar rate indicates that the TMRM ''flickering'' is fully reversible and not associated with permanent TMRM repartitioning (i.e. sustained DY depolarization).

Visualization of DY flickering and its association with mPTP opening
A typical depolarization (blue) and repolarization event (red) for a specific region of interest (ROI) are shown in figure 1D. To better visualize the TMRM intensity changes and minimize interference by TMRM photo bleaching, difference images were calculated between each image (N) and its preceding image (N-1) in the image stack. This yields a new image stack (''DIF'') in which DY depolarizations and repolarizations appear as regions of dark and light pixels, respectively (Fig. 1E). This allows graphical visualization of the average signal changes within a ROI as a function of time (e.g. Fig. 1F). To demonstrate that these depolarization and repolarization events arise from mPTP opening, a manual analysis was  [28]. Following TMRM staining (#1) and its accumulation in the mitochondrial matrix (#2) cells are illuminated leading to photo-induced generation of singlet oxygen ( 1 O 2 ) within the mitochondrial matrix (#4). As a consequence opening of the mPTP is stimulated (#5) leading to DY depolarization (#6) and TMRM efflux from the mitochondrial matrix (#7). The latter prevents photo-induced 1  performed on DIF stacks recorded from control myotubes and myotubes treated with the mPTP-inhibitor Cyclosporin A (CsA; 5 mM, 2 h) (Fig. 2). To allow faithful event detection, an ROI of identical size was used and placed along the whole myotube using the raw frames (see Fig. 2A left panel for a part of a typical myotube). ROI size was empirically chosen in such a way that it was large enough to minimize the number of ROIs required, but small enough to still pick up low magnitude intensity changes. Then these ROIs were used in the DIF stack ( Fig. 2A; right panel) for temporal intensity analysis of the complete myotube. In figure 2B, traces of ,110 ROIs of one control myotube are shown. Clearly, many sharp downward peaks (corresponding to DY depolarization) were observed in control myotubes. Additionally, lower amplitude upward peaks (corresponding to DY repolarization within the same ROI) were also observed. Visual inspection of the DIF stack confirmed the ROI-dependent presence of DY depolarization (dark pixels) and DY repolarization (light pixels). The same strategy with equal ROIs was used on myotubes pre-treated with CsA. CsA pre-treatment greatly reduced the occurrence of depolarization and repolarization events ( Fig. 2C; ,100 ROIs of two myotubes). This strongly suggests that the observed DY flickering arises from reversible mPTP openings. Although quantitative, a major drawback of manual ROI analysis is its time-consuming nature. Moreover, due to the size of the ROIs and their (sometimes overlapping) positioning to cover the entire cell, some events can be simultaneously detected within multiple ROIs (indicated by the black arrows in Fig. 2A right panel), thereby over-estimating their number. To address these issues we developed an automated analysis strategy.

Evaluation of the shot noise influence
To demonstrate that the changes in TMRM fluorescence intensity is a result of a biological process (i.e. DY flickering) and not due to instrumental background noise, the influence of shot noise was estimated using a protocol similar proposed by Blinova et al. [30]. The fluorescence signal of TMRM solutions at different concentrations (0 to 80 nM) was recorded using identical microscopic settings as for the cells. The theoretical Poisson distribution and experimental distribution were compared, as shown in Figure 3. At high TMRM concentration, the observed noise deviated from the theoretical distribution, likely due to uneven illumination across the FOV and temperatures effects. However, when using the difference signal as calculated in our protocol for DY analysis (see above) the noise distribution was normally distributed and displayed a limited spread across the range of TMRM concentrations used. This indicates that the changes in TMRM fluorescence intensity are a consequence of a biological process in the myotubes and not a consequence of instrumental background noise.
An automated protocol for quantification of DY flickering frequency and domain size A flow scheme of the automated strategy is presented in Figure 4. The experimental part (yellow) and image analysis part (blue) including the creation of the DIF stack were already described above. For automated analysis we manually defined a ROI that encompassed a single myotube. We aimed to design an algorithm compatible with the spatial resolution (i.e. considering the smallest observable mitochondrial size) and quality (i.e. signal-to-noise ratio) of the microscopy images. Figure 5A depicts a typical electron microscopy (EM) picture of a mitochondrion in a mouse myotube. Quantification of mitochondrial dimensions in multiple EM images yielded average values for mitochondrial area (0.460 mm 2 ), length (1.6 mm) and diameter (0.333 mm) (Fig. 5B, N531). A reticular was used to calibrate the microscopy images and showed that these dimensions are within the same range as the lateral (XY) dimensions (i.e. 0.3 mm). This means that, in theory, a single pixel could correspond to a small mitochondrial object. However, a single pixel intensity change could also be reflecting a noise-related artifact. Considering mitochondrial length we assumed that mitochondrial objects are longer than 3 pixels (i.e. 0.9 mm). It was further assumed that individual mitochondria can be randomly oriented so a 363 square region with an area of 0.81 mm 2 was used. Although this area exceeds that of an average mitochondrion (Fig. 5B), average mitochondrial length is larger than 3 pixels. Therefore, we defined a mitochondrial object as a region with a size larger than 363 pixels displaying TMRM flickering.
The next challenge is to carry out a threshold operation on the DIF stack that faithfully separates ''real'' signal changes due to DY flickering and signal changes due to random noise. To automatically determine the optimal threshold value, a decision algorithm was applied based on the assumption that DY flickering is absent during the first 10 frames of the stack. This assumption was proven valid by visual inspection of the acquired DIF stacks. Next, the fluctuations in the DIF stack signal during the first 10 frames were analyzed to calculate the standard deviation (SD) value (s), which was used to define various threshold values. For example, a value of 1.64 times s corresponded to 90% of the area of a Gaussian distribution (Fig. 5C), meaning that signals in the DIF stack exceeding this value can be considered to represent DY flickering events with 90% certainty. Using this threshold, three depolarization events and one repolarization event were highlighted in the partial recording depicted in Figure 5C. Obviously, the number of detected DY flickering events decreases with increasing threshold values. To find the optimal threshold value we determined the number of DY depolarization peaks as a function of s in untreated (Fig. 5D) and CsA-treated myotubes (Fig. 5E). Comparing this number with that obtained by manual analysis (dotted horizontal lines), the value 1.64 (90% confidence) and not 1.5 (87% confidence) times s was chosen as the most optimal threshold, since the former value gives a higher confidence. By applying this optimal threshold value, two new image stacks were generated from each DIF stack (Fig. 4). One with below-threshold signal changes (BIN2; reflecting DY depolarizations) and another with above-threshold signal changes (BIN+; reflecting DY repolarizations).
In addition to the temporal properties of the DY flickering events, we also wanted to quantify their spatial properties in the BIN2 and BIN+ images. Our results (e.g. Fig. 1 and 2) demonstrate that DY depolarization is relatively fast and associated with simultaneous intensity changes in regions with a size of multiple pixels (''flickering domains''). As explained above we first discarded flickering domains that were smaller than 9 contiguous pixels (forming a 363 square region) from our analysis. Moreover, when closely juxtaposed ROIs displayed synchronized DY flickering events we assumed that they constituted part of the same flickering domain. The latter was included in our algorithm in that a 15 pixels wide radius (0.45 mm), around each flickering domain was scanned for the presence of another domain with synchronous flickering. If so, this domain was considered part of the same flickering domain. Application of the threshold and size criteria on the BIN2 and BIN+ stacks yields two new binary stacks (Fig. 4) displaying the position of DY depolarizations (DEP) and repolarizations (REP). The result of this automated strategy was validated by manual analysis. For this purpose multiple DIF stacks of control myotubes (N58 in 3 experimental replicates) and CsA-treated (N57 in 3 experimental replicates) myotubes were visually inspected frame-by-frame; each DY flickering and its position was manually scored. The localization (data not shown) and number (Fig. 5F) of DY depolarizations obtained by manual and automated analysis was identical and the latter was also significantly reduced by CsA treatment (p#0.001) (Fig. 5F). Finally, we evaluated the noise-sensitivity of our algorithm by artificially adding different amounts of (white) Gaussian noise to each image in a typical stack. Subsequent application of the analysis algorithm (Fig. 4) revealed that noise addition increased the percentage of missed peaks (Fig. 5G). This result demonstrates that our strategy requires image stacks with a sufficiently high signal-to-noise ratio.

Spatiotemporal properties of DY flickering in control myotubes
Applying our analysis protocol (Movie S2) revealed that DY depolarizations occur in specific regions ( Fig. 6A; blue). Depolarizations were typically paralleled   stack for a given myotube, two images can be calculated to visualize the total number of depolarizations ( Fig. 6B; calculated from DEP stack) and repolarizations ( Fig. 6C; calculated from REP stack). These images also reveal the position and domain size of the flickering domains. Multiple events were observed within the same domains, suggesting that mitochondria within these domains are electrically coupled (Fig. 6B; red). Certain regions within the myotube displayed more flickering events than other, suggesting the existence of DY flickering 'hotspots'. In contrast, DY flickering was not detected in other regions of the myotubes during the time course of the experiment (Fig. 6; noncolorized pixels). Comparing the data obtained for DY depolarizations (Fig. 6B) and repolarizations (Fig. 6C) revealed highly similar patterns. Furthermore, we estimated the co-localization of both types of events as the ratio between the number DY depolarizations and repolarizations per pixel, as shown in Fig. 6D. The average co-localization, as reflected by the Manders coefficient [31], equaled 0.93. This indicates that the large majority of the observed events reflect reversible DY flickering that represents mPTP opening.

Spatiotemporal properties of DY flickering in myotubes treated with CsA and oligomycin A
Recent evidence suggests that the mitochondrial F o F 1 -ATP synthase (complex V or CV) constitutes (part of) the mitochondrial permeability transition pore [13,14]. It was further observed that CypD, which mediates CsA-induced mPTP inhibition, interacts with the OSCP (oligomycin-sensitivity conferring protein) subunit of CV. To demonstrate whether inhibition of CV activity affects photoinduced DY flickering we analyzed mouse myotubes in the absence and presence of the specific CV-inhibitor oligomycin A (OLI; 10 mM). Interestingly, relative to untreated myotubes, CsA treatment greatly diminished DY flickering frequency as expected. However, acute OLI treatment 3-fold increased DY flickering frequency within 15 min (Fig. 7A-C), suggesting that short-term CV inhibition acutely stimulates DY flickering. The increase of the DY flickering upon OLI treatment could not be reduced by CsA. During OLI treatment, the increase in flickering frequency was paralleled by complete disappearance of the 'hotspots' observed in untreated cells (Fig. 7D) and depolarizations occurred throughout the myotube (Fig. 7F). The fact that the median area of the detected flickering domains was ,70 pixels, i.e. 6.3 mm 2 in untreated myotubes (Fig. 7G), suggests that such a domain consists of multiple mitochondria displaying synchronized DY flickering or an electrically connected mitochondrial (sub)network. Inter-mitochondrial electrical coupling appears to be lost in OLI-treated cells. The size of the DY flickering domains was 2-fold reduced in OLI-treated cells (Fig. 7G). CsA-induced inhibition of DY flickering was also associated with a 2-fold reduction in size of the DY flickering domains.

Discussion
This study presents an integrated experimental and computational approach allowing detection and spatiotemporal analysis of light-induced DY flickering events in TMRM-stained primary mouse myotubes. After validation, this strategy was applied to demonstrate that DY flickering is associated with mPTP opening.

Photo-induction of DY flickering in TMRM-stained primary mouse myotubes
We previously discussed that TMRM is well suited for DY quantification given its low toxicity, fast equilibration kinetics across the MIM and low a-specific binding to mitochondrial membranes [32]. However, it is important to realize that mitochondrial accumulation of this dye above a certain concentration leads to autoquenching; making DY to appear depolarized [33]. However, acute application of the mitochondrial uncoupler FCCP (p-trifluoromethoxy carbonyl cyanide phenyl hydrazine) induced an immediate decrease in mitochondrial TMRM fluorescence (data not shown). Hence TMRM autoquenching is absent under our experimental conditions. Evidence was provided that illumination of TMRM and subsequent 1 O 2 formation drive mPTP opening [28,34]. Compatible with this 1 O 2 formation we observe TMRM photobleaching in our experiments. However this photobleaching is not associated with changes in TMRM mitochondrial/nuclear distribution ratio, irreversible DY depolarization or aberrations in cell shape. This finding agrees with previous results demonstrating that TMRM illumination does not release toxic amounts of 1 O 2 within the mitochondrial matrix space [35].
In principle, a sudden decrease in TMRM intensity can also result from mitochondria moving out of focus instead of DY depolarization. To address this potential issue a DY-insensitive mitochondrial co-stain could be applied to normalize the number and size of DY depolarizations/repolarizations. In this study we refrained from using a second mitochondrial stain for several reasons: (i) it might affect mitochondrial physiology, possibly altering DY and intramitochondrial TMRM concentration, and (ii) it might act as a photosensitizer thereby likely altering experimental settings (e.g. illumination time or TMRM loading concentration). However, in our experiments the number of DY depolarizations and repolarizations are equal (Manders overlap coefficient of 0.93) making it unlikely that the observed DY flickering results from mitochondrial movement.
DY flickering activity is virtually abolished by CsA, supporting the conclusion that this flickering represents reversible mPTP opening events. With respect to mitochondrial CRC, a previous study suggests that this parameter increases with TMRM irradiation times up to 30 s (suggesting reduced mPTP opening) and subsequently decreases (suggesting increased mPTP opening) for a TMRM irradiation time .30s [34]. These results were obtained using isolated mouse liver mitochondria with a fully inhibited electron transport chain (ETC), which prevents DY repolarization. However, we also observed a slight increase in mPTP opening frequency at longer illumination times (e.g. Fig. 7A).

A method for quantitative analysis of reversible DY flickering
Mitochondrial DY flickering was previously demonstrated in isolated mitochondria and intact cells [25][26][27][28]. However, quantitative analysis relied on manual analysis by placing small ROIs over (parts of) single mitochondria in different cell-types, which is time-consuming and might lead to biased quantitative analysis. More importantly, this method is susceptible to mitochondrial movement and focus drift. The novel method presented here addresses these points in myotubes and can also be more easily implemented in high-content (HCS) and/or high-throughput (HTS) microscopy screening strategies. In essence, our algorithm consists of calculating a stack of difference images (DIF) from an image stack (RAW) obtained by fluorescence microscopy. This approach has already been used before to visualize DY depolarization in complete astrocytes [26]. However, our method is able to automatically detect DY depolarizations and repolarizations in regions of a myotube and provide information about their spatial localization, area, and frequency. This requires a proper mitochondriaspecific fluorescence signal to allow faithful discrimination between this organelle and the remainder of the cell. From the DIF stack two new image stacks (DEP and REP) are calculated by applying an automatically determined threshold operation. In this case, the optimum is to use 1.64 s (equaling 90% of the area of a normal distribution) in the first 20 seconds of the experiments. Other options seem reasonable based on the results provided in figures 5D and 5E. In this sense, a value of 1.5 s (reflecting 87% of the area of a normal distribution) yielded similar results. However, here we chose to use 1.64 s since this value best matched the manual analysis while providing the highest (90%) confidence. The DEP and REP stacks are then used for numerical analysis of DY depolarization frequency and domain size (DEP stack) and DY repolarization frequency and domain size (REP stack). It was previously discussed that mPTP opening can induce full DY depolarization within 5 ms [28]. This time period is much smaller than our image illumination time (300 ms) and acquisition interval (2 s). However, following mPTP closure DY is restored by ETC action, as reflected by a slowly increasing mitochondrial TMRM fluorescence (data not shown) [28]. In this way, our algorithm is still able to reliably detect DY depolarizations. Quantitative comparison indicates that the numbers of detected DY depolarizations and DY repolarizations are identical, meaning that both types of events are reliably detected. In theory, the fact that our algorithm reported no DY flickering events in subcellular regions of control myotubes could be due to: (i) that these events are truly absent, (ii) that they originate from mitochondrial objects smaller than 363 pixels that are excluded from the analysis, or (iii) given the axial dimensions of the myotubes, mitochondrial objects might be out of focus. When treated with OLI, the algorithm reported an increased number of DY flickering events that originated from DY flickering domains with a substantially reduced size. This observation argues against mitochondria in control cells being so small that they evade detection. OLI-treated cells also displayed DY flickering domains throughout the complete myotube. This result argues against mitochondrial objects being out of focus. In the light of the above, we conclude that our algorithm reliably detects DY flickering events and that the latter are restricted to subcellular DY flickering domains in control myotubes.
Cyclosporin A and oligomycin A affect the spatiotemporal properties of reversible mPTP openings associated with DY flickering Our results demonstrate that DY flickering frequency is reduced by CsA, but increased by OLI. In contrast, both inhibitors reduced the size of DY flickering domains suggesting that they both reduce the electrical connectivity of (the) mitochondrial (sub)network(s). This means that a reduced mitochondrial electrical connectivity can be associated with either a decreased or increased mPTP opening frequency. Interestingly, a previous study demonstrated that mitochondrial matrix volume and CRC were reduced when isolated mitochondria were placed in a medium with low osmolarity (250 mOsM) whereas mitochondrial volume and CRC were increased in a medium with high osmolarity (400 mOsM) [36]. This suggests that a reduction in mitochondrial matrix volume inhibits Ca 2+ -induced mPTP opening. Therefore, the CsA-induced inhibition of mPTP opening observed in this study might be (partially) mediated by a reduction in mitochondrial size, compatible with smaller DY flickering domains.
Recent studies support the idea for a structural role of F O F 1 -ATP synthase in mPTP opening and in particular the c-subunit [13,14,37]. Oligomycin A, a potent inhibitor of the F O F 1 -ATP synthase, was shown to inhibit mPTP opening and TNF-a/ementine-induced cell death in HeLa cells [38]. Surprisingly, acute OLI treatment dramatically increased DY flickering frequency, reduced the size of DY flickering domains and triggered DY flickering events throughout the entire myotube. Although our aim and method is not to induce terminal depolarization and apoptosis, the observed phenomenon might lead eventually to apoptosis. However, OLI is able to increase the rate of apoptosis in eosinophils, suggesting that the effect of OLI could depend on the cell type used [39].
In conclusion, a novel quantitative automated analysis was developed to detect and quantify the frequency, size, and location of individual DY flickering events in myotubes as a consequence of mPTP opening.

Supporting Information
Movie S1. Typical movie of TMRM-stained myotubes. In the myotubes (similar to figure 1B) several localized disappearances (DY depolarizations) and reappearances (DY repolarizations) of TMRM fluorescence intensity can be observed. doi:10.1371/journal.pone.0114090.s001 (WMV) Movie S2. Application of the analysis protocol on a TMRM-stained myotube. The automated analysis protocol detects many DY depolarizations (pixels colored blue) and repolarizations (pixels colored green) in the myotube. The automated analysis protocol also calculates for each frame the number (# flashes) and size of the DY depolarizations (green line) and repolarizations (blue line), which are shown in two separate graphs. doi:10.1371/journal.pone.0114090.s002 (WMV)