Single-cell time-lapse imaging of intracellular O2 in response to metabolic inhibition and mitochondrial cytochrome-c release.

The detection of intracellular molecular oxygen (O2) levels is important for understanding cell physiology, cell death, and drug effects, and has recently been improved with the development of oxygen-sensitive probes that are compatible with live cell time-lapse microscopy. We here provide a protocol for the use of the nanoparticle probe MitoImage-MM2 to monitor intracellular oxygen levels by confocal microscopy under baseline conditions, in response to mitochondrial toxins, and following mitochondrial cytochrome-c release. We demonstrate that the MitoImage-MM2 probe, which embeds Pt(II)-5,10,15,20-tetrakis-(2,3,4,5,6-pentafluorophenyl)-porphyrin as oxygen sensor and poly(9,9-dioctylfluorene) as an O2-independent component, enables quantitative, ratiometric time-lapse imaging of intracellular O2. Multiplexing with tetra-methyl-rhodamine-methyl ester in HeLa cervical cancer cells showed significant increases in intracellular O2 accompanied by strong mitochondrial depolarization when respiratory chain complexes III or IV were inhibited by Antimycin A or sodium azide, respectively, and when cells were maintained at 'physiological' tissue O2 levels (5% O2). Multiplexing also allowed us to monitor intracellular O2 during the apoptotic signaling process of mitochondrial outer membrane permeabilization in HeLa expressing cytochrome-c-eGFP, and demonstrated that mitochondria post cytochrome-c release are able to retain their capacity to respire at physiological O2 despite a decrease in mitochondrial membrane potential.

Aerobic organisms require a constant supply of molecular oxygen (O 2 ) to produce ATP through oxidative phosphorylation by mitochondria, a process that also leads to the formation of reactive oxygen species (ROS). 1 The response to O 2 levels in mammalian tissues is tightly regulated by specific genes and signaling pathways in order to maintain cell bioenergetics and survival. 2 Severe fluctuations in O 2 levels may lead to anoxia (no oxygen), hypoxia (decreased availability of O 2 ) or hyperoxia (increased O 2 levels), each condition capable of inducing cell and tissue damage. Because of uncontrolled cell proliferation, cancer cells are often exposed to tissue hypoxia. Many cancer cells are therefore specifically equipped to adapt and survive hypoxic periods. 3,4 Similar to hypoxic conditions, mitochondrial cytochrome-c (cyt-c) release during apoptosis also induces a bioenergetics crisis, as cyt-c shuttles electrons between complexes III and IV. [5][6][7] Many cancer cells are resistant to caspase activation, 8 and when caspase activation is compromised, cancer cells may survive the bioenergetics crisis induced by cyt-c release, as the fraction of cyt-c remaining in the intermembrane space after equilibration with the cytosolic compartment may still be able to contribute to respiratory chain activity. [9][10][11] This enables mitochondria to sustain intracellular ATP in the absence of further mitochondrial degradation. This process is facilitated through enhanced extracellular glucose uptake, another key bioenergetics alteration of cancer cells. 9 Because of the key role played by the mitochondrial respiratory chain in the control of cell survival during apoptosis, O 2 sensing represents an important method for the study of cancer energy metabolism and bioenergetics responses to metabolic inhibitors or mitochondrial cyt-c release. 12 Therefore, the development of new O 2 sensing and imaging protocols that enable measurements of oxygen levels in single living cells and during asynchronous, apoptotic cell death relative to other physiological parameters is of great interest to the cell death and bioenergetics community.
Significant progress has been made in the field of molecular O 2 detection by optical sensing. 13 The advantages of this technique are its sensitivity, accuracy and non-invasive nature. 14 Quenching of phosphorescence has become an important method for measuring O 2 by optical sensing. 15 Phosphorescence quenching relies on the ability of O 2 to quench the emission of excited triplet state molecules. In biological systems, phosphorescence quenching is highly specific to O 2 , since oxygen is the only small molecule dynamic quencher present in sufficiently high concentrations. 16 Advantages of phosphorescent probes include high specificity, fast response, high sensitivity, stable calibration and various readout parameters such as intensity and lifetime. However, most of the probes developed still could not satisfy all the requirements for O 2 measurement in high-resolution imaging modalities in long-term experiments because of lack of compatibility with other probes, requirement of special imaging hardware, limited uptake into cells, or significant phototoxicity. 17 Detection techniques such as the Whalenstyle platinum electrode 18 allow for the measurement of O 2 consumption at the single cell level, 19 but only deliver data for one cell at a time. Other optical intracellular oxygen sensing probes and techniques including Clark-type oxygen chips 20,21 often require highly specialized equipment such as fluorescence life time microscopy technologies. 21,22 As many laboratories routinely use confocal or epifluorescence timelapse imaging, there is a significant need for the development of probes for these applications. 23,24 In this study, we evaluated the utility of a nanoparticle-based phosphorescent probe, MitoImage-MM2, consisting of the O 2 -sensitive phosphorescent reporter dye (PtTFPP) and the O 2 -insensitive component (PFO) embedded in a cationic polymer, 25 for confocal time lapse imaging. We demonstrate that MM2 senses changes in oxygen concentration at single cell level in response to hypoxia, metabolic inhibitors and mitochondrial cytochrome-c release, and demonstrate its suitability for multiplexing with additional single cell fluorescent probes. Importantly, we also demonstrate that cell culture environment needs to be adjusted to more physiological conditions, 26 to more reliably detect alterations in cellular O 2 levels and mitochondrial bioenergetics.

Results
MM2 accurately measures O 2 concentration. The MM2 probe has been previously shown to be sensitive to O 2 and compatible with specialized detection platforms such as FLIM systems and time-resolved fluorescence (TR-F) readers. 25,27 Here we evaluated the capability of the ratiometric MM2 probe to monitor changes in O 2 concentration in HeLa cells loaded with MM2 nanoparticles, using a confocal live cell imaging setup. First, we explored the response of the probe alone under different O 2 concentrations. To this end we measured the MM2 ratio signal R = (F PFO /F PtTFPP ) using 10 μg/ml of the probe in glass bottom cell culture dishes at varying O 2 concentrations on the live cell imaging setup, using the stage incubator described in Materials and Methods. In order to account for the variability of R (t = 0 min) between experiments, we normalized R to its start value R 0 = R(20% O 2 ). The on-stage calibration indicated a strong, linear correlation between the normalized MM2 intensity ratio R/R 0 and the oxygen concentration equilibrated to 20, 10, 5 and 2% O 2 (Figures 1a and b). The dependence of the PtTFPP intensity on the O 2 concentration is clearly visible in Figure 1a, while the PFO intensity remained stable during the calibration with 2, 5, 10 and 20% oxygen in the stage incubator atmosphere. Fitting to an exponential function provided the calibration equation O 2 [μM] = 1/0.0045 exp (R/R 0 /38.01), with an r 2 = 0.9997, representing an optimal fit. To confirm that the un-quenching of PtTFPP and therefore the drop in R/R 0 was reversible, O 2 in the buffer was again equilibrated to ambient O 2 at the end of each calibration experiment (Figures 1a and b).
MM2 is effectively loaded into HeLa cells. We next conducted a series of experiments to establish the best uptake conditions in HeLa cells. We incubated HeLa cells with 10 μg/ml MM2 probe in medium containing 10, 5, 2 and 1% of fetal bovine serum (FBS) for 16 h. This was required as the concentration of heat inactivated serum in the culture medium has been shown to affect the intracellular concentration of many probes and nanoparticles. 28 We identified 1%  Effect of mitochondrial chain inhibitors on intracellular O 2 levels of HeLa cells maintained at 'supra-physiological' (20%) or 'physiological' (5%) ambient O 2 . We next explored the effect of mitochondrial electron transport chain (ETC) inhibitors on intracellular oxygen levels and ΔΨ M as detected with TMRM. It has been shown that HeLa cells in a glucose-deprived medium shift their energy metabolism predominantly towards oxidative phosphorylation. 29 All experiments were therefore carried out with HeLa cells in glucose-free medium supplemented with 2 mM sodium pyruvate. HeLa cells loaded with MM2 probe and incubated with 30 nM TMRM were treated with ETC inhibitors, Antimycin A (AA; 10 μM), or NaN 3 (0.2 mg/ml). AA and NaN 3 inhibit complexes III and IV, respectively. As expected, inhibition of complex IV (Figures 2a and b MitoImage-MM2 probe reveals respiration in single HeLa cells which underwent MOMP. We next explored whether multiplexing with the MM2 probe would also allow for simultaneous time lapse imaging of both cyt-c-eGFP redistribution and ΔΨ M using TMRM during apoptosis. Previous studies have suggested that oxidative phosphorylation activity and mitochondrial ATP production is still abundant after outer mitochondrial membrane permeabilization (MOMP) when caspase activation is compromised and glucose is available. 9,11,30 HeLa cells over expressing cyt-c-eGFP were pre-treated with 100 μM zVAD-fmk and then equilibrated at 2% O 2 for 60 min to mimic borderline normoxic/hypoxic conditions in tumors. Cells were then treated with the apoptosis-inducing broad-spectrum kinase inhibitor STS to induce the release of cyt-c-GFP. We hypothesized that a compromised ETC turnover after cyt-c release would cause an increase in intracellular O 2 .    Figure 4bii). A quantification of these three types of responses is provided in Figure 4c. In order to investigate whether this was also determining the outcome of respiration changes after glucose addition, we separately analyzed the three response groups (Figure 4d). Interestingly, the majority of cells with a single-cell history of increased respiration after MOMP demonstrated decreased respiration after glucose addition. In contrast, cells which showed no response or a decrease in respiration after MOMP were less likely to decrease respiration after glucose addition. Furthermore, 149 of the 228 cells studied showed a recovery of ΔΨ M (as indicated by an increase of TMRM average intensity) after the addition of 25 mM glucose.

Discussion
In this report, we evaluated the potential of MitoImage-MM2 to probe alterations in intracellular O 2 levels and exploited the multiplexing capability of MM2. We found that exposure of HeLa cells loaded with MM2 to different oxygen levels accurately detected intracellular O 2 concentrations similar to ex vivo experiments. To explore whether MM2 reported alterations in O 2 levels as a consequence of changing mitochondrial respiration, we exposed HeLa cells to Antimycin A and NaN 3 and detected alterations in O 2 levels in particular when cells were maintained at more physiological tissue O 2 levels (2-5%). The complete O 2 consumption inhibition with NaN 3 enabled us to quantify the cellular O 2 concentration in cells kept in standard culture conditions. This then allowed us to calibrate single cell O 2 kinetics, and to simultaneously measure ΔΨ M kinetics.
MM2 is a second-generation probe for O 2 imaging, characterized by its PFO and PtTFPP, suitable for confocal and two-photon imaging. 25 In contrast to the loading of cells with nanoparticles incorporating PtTFPP alone, 27 MM2 can be used for ratiometric measurements. The dye combination used in MM2 also circumvents detector sensitivity issues with an emission wavelength above maximum at 650 nm, 32 making it suitable for confocal time lapse imaging. 33 MM2 was photostable with little bleaching and could be calibrated to estimate actual oxygen levels within cells. MM2 is compatible with fluorescent proteins, FITC-and rhodamine-based optical probes. Applications of second-generation phosphorescent probes for O 2 sensing adds speed, 19,34 and probes can be used for plate reader-based assays in parallel. 35 Interestingly, a cell-impermeable analog of MM2, PtP-C343, modified with polyethylene glycol residues, has recently also been shown to measure O 2 in brain microvessels when administered intravenously. 22 Our confocal microscopy experiments indicated that MM2 was taken up and internalized in HeLa cells. While we have not yet defined the intracellular structures that incorporate MM2, a previous study using similar nanoparticles has identified uptake into endosomes in proximity to mitochondria. 27,36 Using the settings described in our study, we did not detect toxicity of the probe for up to 24 h after a 16-h incubation period, or phototoxic effects triggered by the image acquisition settings in control time lapse experiments. The detection principle used (quenching of the phosphorescence signal by O 2 ) will potentially generate ROS. To avoid this, acquisition settings have to be tested and adjusted to the highest sensitivity and lowest excitation light intensity and exposure time. This comes at the cost of increased noise, limiting the detection of subtle changes of intracellular O 2 . However as we demonstrate in this report, MM2 is a highly interesting probe for detection of the kinetics of intracellular O 2 concentration by light microscopy, and enables the detection and classification of response types of intracellular O 2 changes even at low    Figure 4.
Previous findings have shown that atmospheric O 2 , cell density, respiration rate and its dynamics are the major factors influencing the oxygen-sensitive signaling pathways. 37 We show that alterations in oxygen levels in transformed HeLa cancer cells could be adequately detected at an oxygen level of 53 and even 215 μM (5% O 2 or 20% in the atmosphere with 5% CO 2 ). This represents O 2 levels higher than found in the human body, which ranges between 9 and 21 μM in healthy   38 but still represents an extracellular O 2 level much lower than atmospheric oxygen (215 μM). At atmospheric O 2 levels, mitochondrial respiration inhibition with sodium azide did not induce significant alterations in intracellular oxygen levels. This may be due to the fact that dissolved O 2 at a high concentration (20%) can diffuse and equilibrate rapidly into the cells, rendering it difficult to detect changes in ETC O 2 consumption. Importantly, as oxygen consumption is central to ATP production via the respiratory chain, this implies that any analysis of cellular bioenergetics, such as mitochondrial membrane potential, ATP levels, ROS production, or NADH/NADPH levels should be performed under controlled oxygen levels.
Because mitochondria are the primary consumer of molecular O 2 and also control adaptive responses to hypoxia, 39-41 cellular O 2 levels can be an indicator of O 2dependent metabolic activities, such as aerobic respiration or oxygen-dependent synthesis and degradation of cellular components, 33,42 and acts as a potential site for O 2 sensing in the cell. 43,44 Indeed, we found that simultaneously detected TMRM uptake to be different at 20 and 2% O 2 when cells fully rely on ETC to maintain their energy metabolism and fully depend on mitochondrial ATP. However, findings from isolated mitochondria suggest that mitochondrial O 2 availability does not become critically low and ATP production through ETC remains stable until the oxygen concentration falls near anoxic conditions (o0.3%). 45,46 Tumor cells harbor the 'Warburg effect' which describes a switch from mitochondrial respiration to glycolysis in the presence of oxygen. However, recent studies have highlighted that this 'switch' is not complete and tumor cells are well capable of using mitochondria for aerobic ATP production. 47,48 Indeed, we show that inhibition of respiratory chain activity with Antimycin A or NaN 3 significantly increased intracellular O 2 in cells kept under physiological O 2 concentration.
Cytochrome-c is the main enzyme involved in transporting electrons and binds to oxygen in the ETC. 49 This, in turn, polarizes mitochondrial membrane potential, which drives proton motive force for ATP synthase to produce ATP. 50 In our experiments, cyt-c release during apoptosis depolarized ΔΨ M but had little effect on intracellular O 2 levels in the majority of cells, confirming previous reports that released cyt-c is still accessible for mitochondrial respiration. 11 Our study also demonstrates that, when glucose became available in cells which underwent MOMP, most cells responded with an increase in intracellular O 2 , indicating that cancer cells are able to flexibly adapt to extracellular glucose increases post-MOMP with an increased glycolytic activity. Of note, our data also show a significant heterogeneity of bioenergetics responses post cyt-c release and glucose addition. This heterogeneity was not unexpected, as previous single cell imaging and mathematical modelling studies from our laboratory indicated a strong heterogeneity in mitochondrial respiration post MOMP, with cells operating different modes of complex V/ATP synthase activity in a given population of cancer cells, depending on the amount of respirationaccessible cyt-c and the degree of glycolytic ATP production. 9 This may also explain our findings that cells with increased respiration after MOMP (excessive respiration accessible cyt-c) were more likely to respond to an increase in extracellular glucose with a decrease in respiration, while cells with limited respiration (and limited accessible cyt-c) showed no decrease in respiration as a response to glucose addition.
The direct link between increased glycolytic activity and oxygen consumption in cancer cells that underwent MOMP may also be important in the context of ROS production and HIF-dependent and -independent hypoxia signaling, as both are strongly influenced by intracellular O 2 consumption and may control demise or recovery of cells post-MOMP. 51,52 The findings reported here may therefore also have important implications for tumor cell survival and resistance to therapy.   Monitoring changes in O 2 concentration using the ratiome-tricMM2 probe. HeLa cells were loaded with 10 μg/ml of MM2 (or co-loaded with TMRM for 30 min in the dark) in RPMI medium supplemented with 1% FBS medium for 16 h at 37°C. MM2 probe intensity ratio was recorded at different oxygen concentrations (20, 10, 5 and 2% O 2 ) in a Willco dish mounted on LSM 5 live Duoscan confocal microscope (Carl Zeiss, Jena, Germany) equipped with a × 40, 1.3 NA Plan-Neofluar oil-immersion objective and a thermostatically regulated chamber (Pecon, Erbach, Germany) at 37°C in a humidified atmosphere of 5% CO 2 /95% air. CO 2 and O 2 levels (%) in the stage incubator were regulated using the CTI controller 3700 Digital in combination with an O 2 controller (Pecon, Erbach, Germany). This unit requires N 2 to displace O 2 from the incubation atmosphere. The MM2 probe was excited using 2% of the 30 mW 405 nm DPSS Laser, and the emission was collected through a 415-480 nm band pass and a 570 nm long pass filter using a 565 nm secondary dichroic to split the emission between the two detectors of the LSM 5 live. TMRM was excited at 561 nm, and the emission was collected through a 570-640 nm band pass filter. In HeLa cells expressing cyt-c-GFP, GFP was excited with a 489 nm DPSS laser and the emission was detected using a 495-555 nm band pass filter. All images were processed using ImageJ (version 1.45, Wayne Rasband, NIH, Bethesda, MD, USA) and MetaMorph Software version 7.5 (Molecular Devices, Wokingham, Berkshire, UK). The intensity ratio images between the PFO and the PtTFPP (F PFO /F PtTFPP ) were calculated for all areas of the image with PFO and PtTFPP fluorescence above background noise and after background subtraction. Mitochondrial membrane potential (ΔΨ m ) changes were measured using the average pixel intensity of TMRM per cell.
Sodium azide (NaN 3 ) and AntimycinA (AA) treatment at physiological (5%) and hyperoxic (20%) O 2 . HeLa cells cultured on Willco dishes and incubated with the MM2 probe (10 μg/ml) for 16 h were co-loaded with TMRM (30 nM) at 37°C in the dark for 30 min. The Willco dishes with cells were mounted on the stage of an LSM 5 live confocal microscope (Zeiss, Jena, Germany) equipped with a × 40, 1.3 NA Plan-Neofluar oil-immersion objective and a thermostatically regulated chamber (Pecon, Erbach, Germany) at 37°C in a humidified atmosphere of 5% CO 2 /95% air. Cells were maintained at 20% O 2 for 10 min to allow for equilibration and then reduced to 5% O 2 using the equipment described above. After 60 min of equilibration the cells were exposed to NaN 3 (0.2 mg/ml) or AA (10 μM). TMRM and MM2 probe fluorescence intensities were recorded. Experiments were terminated by addition of D-(+)-Glucose (100 mM) and glucose oxidase (100 μg/ml) to completely deplete the available oxygen. In another set of experiments, Hela cells were maintained at 20% O 2 . After a 10-min equilibration period, cells were exposed to sodium azide (NaN 3 , 0.2 mg/ml) or AA (10 μM) and TMRM and MM2 probe intensity ratio was recorded as described above.
Induction of apoptosis and simultaneous measurement of oxygen consumption and GFP redistribution. HeLa cyt-c-GFP cells were cultured in glass bottom dishes and incubated with 10 μg/ml MM2 probe for 16 h followed by TMRM (30 nM) and zVAD-fmk (100 μM) for 30 min to inhibit caspase activation. After treatment with STS, caspase activation would otherwise trigger mitochondrial demise following MOMP due to cleavage of complex I. 9,11,30 Cells were mounted on stage as described above and maintained in 20% O 2 for the required 10 min of equilibration. Then O 2 was reduced to 2% using the equipment described above. After 60 min cells were treated with 3 μM STS. During the subsequent time lapse imaging experiments, glucose was added to the medium as indicated. All images were processed using ImageJ (1.45) and MetaMorph Software version 7.5 as described above. The distribution of cyt-c-GFP was calculated using the S.D. of the average pixel intensity as described previously. 31  Statistics. Data are given as means ± S.E.M. For statistical comparison, one-way analysis of variance, paired samples t-test, Wilcoxon test and Friedman's test were carried out using SPSS software (SPSS version 23, IBM). Where the p-value waso0.05, groups were considered to be significantly different. All experiments were carried out in triplicate unless indicated otherwise.
Conflict of Interest D.P. is a stakeholder of Luxcel Biosciences, a company which produces the oxygen sensor used in this study. The other authors have no conflict of interest.