Quantitative analysis of mitochondrial membrane potential heterogeneity in unsynchronized and synchronized cancer cells

Abstract

Mitochondrial membrane potential (ΔΨm) is a global indicator of mitochondrial function. Previous reports on heterogeneity of ΔΨm were qualitative or semiquantitative. Here, we quantified intercellular differences in ΔΨm in unsynchronized human cancer cells, cells synchronized in G1, S, and G2, and human fibroblasts. We assessed ΔΨm using a two-pronged microscopy approach to measure relative fluorescence of tetramethylrhodamine methyl ester (TMRM) and absolute values of ΔΨm. We showed that ΔΨm is more heterogeneous in cancer cells compared to fibroblasts, and it is maintained throughout the cell cycle. The effect of chemical inhibition of the respiratory chain and ATP synthesis differed between basal, low and high ΔΨm cells. Overall, our results showed that intercellular heterogeneity of ΔΨm is mainly modulated by intramitochondrial factors, it is independent of the ΔΨm indicator and it is not correlated with intercellular heterogeneity of plasma membrane potential or the phases of the cell cycle.

1 |. INTRODUCTION

Mitochondria are essential for maintaining cellular bioenergetics and homeostasis in quiescent states and during proliferation. The constantly changing metabolic demands require fine-tuned mechanisms to control mitochondrial function in each cell. In non-proliferating cells, most of the pyruvate generated from glucose is fully oxidized by mitochondrial oxidative phosphorylation (OXPHOS), that generates more than 95% of total cellular ATP.¹ By contrast, the Warburg phenotype is characterized by enhanced aerobic glycolysis in proliferating cells with different levels of mitochondrial metabolism.^2–5 The pro-proliferative Warburg phenotype favors the formation of glycolytic intermediates not fully oxidized in mitochondria, that are used for the biosynthesis of macromolecules.⁶ The relative contribution of glycolysis and OXPHOS in cancer cells is dynamically modulated by the availability of nutrients and oxygen, proximity to neoformed and mature blood vessels, and interaction with stromal cells among other factors.⁷ Distinct metabolic and bioenergetic patterns have been found among patients with the same type of tumor, between primary and metastatic tumors, and from different regions inside the same tumor.^8,9 Heterogeneity of tumor metabolism and bioenergetics is increasingly recognized as a contributing factor to the failure of chemotherapy.¹⁰ Intercellular heterogeneity of mitochondrial membrane potential (ΔΨm), reported in several cell types,^11,12 contributes to tumor heterogeneity. Although previously recognized as a biological phenomenon, the underlying mechanisms and the adaptive relevance of ΔΨm heterogeneity in cancer cells remain undetermined.

Bioenergetic adaptations to cellular requirements in cancer cells depends on the balance between mitochondrial oxidation of respiratory substrates including pyruvate, and the rate of glycolysis. Pyruvate, fatty acyl-CoAs, glutamine, and other amino acids, that enter the mitochondrial matrix, are oxidized in the Krebs cycle, generating NADH that, in turn, fuels the electron transport chain (ETC). The flow of electrons in the ETC to the final acceptor oxygen, drives pumping of H⁺ by complexes I, III, and IV to the intermembrane space. Proton pumping results in the generation of an electrochemical gradient between the intermembrane space and the mitochondrial matrix, formed by an electrical (ΔΨm) and a chemical (ΔpH) component. The electrochemical H⁺ gradient drives the synthesis of ATP from ADP and inorganic phosphate (Pi) by the F1F0 ATP synthase (Complex V), coupling electron flow in the ETC to ATP synthesis.¹³ In addition, the potential energy stored as an electrochemical H⁺ gradient favors the transport of ions and metabolites across the inner mitochondrial membrane, as well as the import of mitochondrial resident proteins.¹⁴ The uncoupled leak of H⁺ from the intermembrane space to the matrix, is another factor influencing ΔΨm. Overall, ΔΨm, as the main component of the electrochemical H⁺ gradient, is a dynamic indicator of mitochondrial metabolism.

Mitochondrial membrane potential has been measured using different techniques, both in isolated mitochondria and in intact cells.^15–19 In live cells, most assays use lipophilic, fluorescent cations to monitor ΔΨm.¹⁹ The distribution of these hydrophobic cationic dyes between an external aqueous medium and the cytosol, and between the cytosol and the aqueous compartments of intracellular organelles, is driven by plasma membrane potential (ΔΨp) and ΔΨm, respectively. In equilibrium, these distributions are described by the Nernst equation. Accordingly, a cationic dye would accumulate ~10 -fold in the cytoplasm and ~10,000 -fold within mitochondria in a cell with ΔΨp: −60 mV and ΔΨm: −180 mV.²⁰ Fluorescence of cationic dyes is an exponential function of the differences in ΔΨp and ΔΨm when the concentration of the dye is below the limit where self-quenching occurs (non-quenching mode). Overall, mitochondrial accumulation of a potential-indicating dye, depends on the difference in potential between the mitochondrial matrix, the plasma membrane and the external medium. Furthermore, total probe accumulation and fluorescence intensity of ΔΨm-indicating probes, including tetramethylrhodamine methyl ester (TMRM), is also a function of geometry, binding, and kinetic factors related to the dye, that contribute to the practical quenching limit.²¹ In general, loading of potentiometric dyes, influenced by the magnitude of ΔΨm and ΔΨp, should be long enough to reach an equilibrium between the mitochondrial matrix, the cytosol, and the extracellular environment.^22,23 At present, most of the widely used experimental approaches assess ΔΨm changes qualitatively (mitochondria are either “polarized” or “depolarized”), or semiquantitatively by setting an arbitrary baseline value. Recently, a method has been reported to calculate absolute values of ΔΨm using time-lapse imaging of the non-quenching mode fluorescence of TMRM, and a ΔΨp indicator in intact cells that also accounts for the geometric, binding, and kinetic factors affecting TMRM fluorescence.^16,24

Here, we assessed intercellular differences in ΔΨm in unsynchronized and synchronized cancer cells, as well as in fibroblasts, using a combination of semiquantitative and absolute calibration methods. We also evaluated changes induced by pharmacological inhibition of complex III of the ETC and the ATP synthase. Our results confirmed that heterogeneity of ΔΨm was not biased by the relative contribution of ΔΨp and it was independent of the ΔΨm indicator dye used. We also demonstrated that ΔΨm remained heterogeneous in synchronized cells at G1, S, and G2. The specific inhibition of either complex III or the ATP synthase reduced intercellular ΔΨm heterogeneity, suggesting that intramitochondrial mechanisms are major contributors to ΔΨm heterogeneity in cancer cells.

2 |. MATERIALS AND METHODS

2.1 |. Materials

Antimycin A, carbonyl cyanide 3-chlorophenylhydrazone (CCCP), oligomycin, zosuquidar, tetramethylrhodamine methyl ester (TMRM), and bis (1,3-dibutylbarbituric acid) trimethine oxonol (DiBAC4(3)) were purchased from Millipore Sigma (Burlington, MA, USA). The FLIPR Membrane Potential Assay Explorer Kit (FLIPR) was from Molecular Devices (Sunnyvale, CA). Fetal bovine serum (FBS) was from Atlanta Biologicals. Penicillin, streptomycin, 100X MEM Nonessential amino acids, RPMI 1640 containing 2.05 Mm of L-glutamine, FxCycle ™ PI/RNAse Staining Solution, paraformaldehyde, and Rhodamine 123 were purchased from Thermo Fisher Scientific (Waltham, MA, USA). Eagle’s Minimum Essential Medium was purchased from the American Type Culture Collection (ATCC; Manassas, VA, USA). All other chemicals were analytical grade.

2.2 |. Cell culture

HepG2 human hepatocarcinoma cells (HB-8065), BJ1 fibroblasts, and HCC4006 human lung adenocarcinoma (CR1–2871) cells were purchased from the American Tissue Culture Collection (ATCC) (Manassas, VA). Huh7 human hepatocarcinoma cells were the generous gift from Dr Jack Wands, Brown University, Providence, RI. HepG2 cells were grown in Eagle’s minimum essential medium (EMEM); HCC4006 cells in Roswell Park Memorial Institute (RPMI) 1640; and Huh7 cells in EMEM with the addition of 1% 100X MEM Nonessential amino acids. For all cancer cell lines, media was supplemented with 10% of FBS, 100 units/mL of penicillin, and 100 μg/mL of streptomycin. Cells were maintained in 5% of CO₂/air at 37°C. BJ1 fibroblasts were grown in DMEM supplemented with 0.2% of FBS, 3% of O₂, and 10% of CO₂.

2.3 |. Assessment of relative fluorescence by confocal imaging of TMRM, Rh123, and DiBAC4(3)

Cells plated in Greiner Bio TC 4-chamber plates (Greiner-Bio-One, Monroe, NC), or 35 mm MatTek dishes (MatTek Corporation, Ashland, MA), were loaded with TMRM (200 nM) for 30 minutes in modified Hank’s balanced salt solution (HBSS) containing (in mM): NaCl 137, Na₂HPO₄ 0.35, KCl 5.4, KH₂PO₄ 1, MgSO₄ 0.81, Ca₂Cl 0.95, glucose 5.5, NaHCO₃ 25 and HEPES 20, pH 7.4, or complete growth media. After washing, subsequent incubations were performed using TMRM (50 nM) to maintain equilibrium distribution of the fluorophore as described previously.^25,26 Cells incubated in modified HBSS in a humidified 5% of CO₂/air at 37°C, were imaged using a Zeiss LSM 880 NLO inverted laser scanning confocal microscope (Thornwood, NY) with a 63X 1.4 NA plan apochromat oil immersion lens. TMRM was excited at 561 nm and emission detected with a Quasar multichannel spectral detector at 590–610 nm through a one Airy unit diameter pinhole. Alternatively, cells were loaded with Rh123 (10 nM) and zosuquidar (1 μM) for 90 minutes at 37°C in modified HBSS. Images of Rh123 were collected using a 488 nm excitation. Cells loaded with DiBAC4(3) (500 nM) for 30 minutes were imaged using a 510–540 nm band-pass filter as previously reported.²⁷ After establishing steady baseline images, CCCP (1 μM) was added to fully depolarize mitochondria while continuously recording. This enabled calibration of the measurements and confirmed that both TMRM and Rh123 were used in non-quench mode. Both TMRM and Rh123 fluorescence intensities were quantified to make relative comparisons using Photoshop CS4 software (Adobe Systems, San Jose, CA), as previously described.²⁸ A minimum of four randomly selected fields with 8–20 cells per field were imaged during the time course of at least three independent experiments. The mitochondrial network from individual cells were circled with the Photoshop lasso tool or ImageJ, to determine fluorescence intensity from each region of interest. Relative fluorescence was calculated as a percentage of the highest fluorescence value of the population. Heterogeneity was calculated by quantifying intra-experimental variances in ΔΨm, and expressed as SD of the replicates ± SE.

2.4 |. Absolute and unbiased quantification of ΔΨm in cancer cells and fibroblasts

2.4.1 |. Cancer cells

Cells were plated in glass bottom, 10-well Cellview culture slides (Greiner-Bio-One) and maintained in culture medium. Absolute millivolt potentials were calculated in single cells using time-lapse laser scanning confocal microscopy of TMRM and the ΔΨp probe FLIPR, as described with slight modifications.¹⁶ A nonfluorescent DMEM-based potentiometric imaging medium (PM) modified from the original description was prepared to assure constant TMRM and FLIPR concentrations. Briefly: 2x PM component A DMEM powder (Image Analyst Software, Novato, CA), was supplemented with NaCl, HEPES, and NaHCO₃ (in mM: 23, 20, and 1), respectively, adjusted to pH 7.4 at 37°C. Before experiments, 2xPM was further supplemented with glucose, CaCl₂, TMRM, TPB, and FLIPR (5.6 mM, 1.8 mM, 20 nM, 1 μM, and 1:200), respectively. The supplemented 2xPM was diluted 1:1 with NaCl (240 mM). The working solution (PMNa) was used to preincubate cells. After 1 hour of preincubation, half of the volume was replaced with fresh PMNa. Recordings were performed using a Zeiss LSM 880 NLO inverted laser scanning confocal microscope as described in 2.3, with wide pinhole (9.94 airy units). FLIPR and TMRM were excited simultaneously at 488 nm (2.0% laser power) and at 561 nm (0.1%), and emissions were detected at 521–558 nm and 597–668 nm, respectively. Images were analyzed using the “Mitochondrial membrane potential measurement (TMRM/FLIPR) for confocal” pipeline (Image Analyst MKII (Image Analyst Software). This pipeline, that measures mean fluorescence intensities in whole cells, was slightly modified from the generic pipeline by removal of background subtraction. Channel registration was not required for the confocal imaging setup described here. Calculation of ΔΨm depends on the number of mitochondria, ultrastructure, and probe binding. These parameters were calibrated for each cell type, determining mitochondria/cell volume fractions (7.80 ± 0.05 and 5.83 ± 0.007 for HepG2 and HCC4006 cells, respectively); and apparent activity coefficient ratio (0.18 and 0.38 for HepG2 and HCC4006 cells, respectively), as previously described.²⁴

ΔΨp was determined using the “Iterative Complete (Zero only)” approach in the “Membrane Potential Calibration Wizard,” because stepwise calibration of the ΔΨp by increasing extracellular potassium concentration as described for neurons,¹⁶ was not possible in HepG2 and HCC4006 cells. This calibration approach was used together with the “Complete” ΔΨm approach to determine both ΔΨp and ΔΨm from two-channel fluorescence time courses of mitochondrial depolarization triggered by a mitochondrial complete depolarization cocktail (MDC) containing (in μM): 1 valinomycin, 2 oligomycin, 0.1 FCCP, 2 Antimycin A, 2 Myxothiazol; followed by subsequent ΔΨp depolarization induced by a complete depolarization cocktail (CDC): 2% of paraformaldehyde and KCl (120 mM), and gramicidin (5 μg/mL). Our results for ΔΨp in human liver cancer cells were similar to low values of ΔΨp (~ − 22 mV) determined for human hepatocytes.²⁹ To improve robustness of single-cell calibration, using the above approach, we calculated the time constant of TMRM redistribution (kt) for the cell type, and then, used a uniform value to rerun the calculations using the “Complete known k” ΔΨm approach to generate single-cell potentials.

2.4.2 |. BJ1 fibroblasts

BJ1 human fibroblasts were plated in 8-well Lab-Tek chamber coverglass and cultured in DMEM supplemented with 0.2% of FBS. The PMNa composition was (in mM): NaCl 120, KCl 3.5, CaCl₂ 1.3, MgCl₂ 1, KH₂PO₄ 0.4, TES 20, NaHCO₃ 5, Na₂SO₄ 1.2, glucose 25 and glutamine 2, plus TMRM (10 nM), TPB (1 μM), and FLIPR (blue version, 1:300), at pH 7.4, 37°C. Imaging was performed on a Nikon Eclipse Ti-PFS wide-field fluorescence microscope.²⁴ Image analysis was performed by a modification of the “Mitochondrial membrane potential assay (TMRM/PMPI) with post-hoc absorbance classifier” pipeline, using automated detection of single cells based on Hoechst 33342 staining of nuclei after live-cell experiments. ΔΨp and ΔΨm calibrations were performed using “Complete with known kP (K-steps)” and “Complete” paradigms as described.²⁴ The mitochondria:cell volume fraction was 4.21 ± 0.02% (n = 4) in quiescent BJ1 fibroblasts and an apparent activity coefficient ratio of 0.41 was used for the calculations.¹⁶

The standard deviation of ΔΨm measured in a cell population reflects both intercellular heterogeneity and noise from experimental error. These two sources of heterogeneity are independent form each other. Assuming a Gaussian distribution for intercellular heterogeneity and experimental noise, the variances sum. ΔΨm calibration estimates the experimental error originating from readout noise and linear fits used for calculating calibration parameters. A corrected variance of intercellular heterogeneity was calculated as cell to cell ΔΨm variance subtracted from the predicted calibration variance. This calculation cancels instrument and calibration paradigm-specific differences between experimenters. Therefore, corrected SD (variance) values were directly comparable between fibroblasts and cancer cell lines used in this study.

2.5 |. Cell cycle synchronization

HepG2 cells (300, 000 cells/well) were plated in 6-well plates (CytoOne) for flow cytometry and, in parallel, in 4-chamber dishes (Greiner, 250, 000 cells/well) for imaging. Both sets of cells were treated equally. For synchronization in G0/G1, cells were grown in whole media for 48 hours followed by serum-free media for another 48 hours. For synchronization in S phase, cells were grown for 48 hours in growth media, before adding thymidine (2 mM) for 24 hours. After thymidine, cells were rinsed with warm PBS and exposed to 2-deoxycytidine (24 μM) for 9 hours, before addition of a second thymidine (2 mM) block in whole media for 24 hours. Cells were released from the second thymidine block by rinsing with PBS and adding fresh media for 6 hours. For synchronization in G2/M, cells were exposed to a double thymidine block followed by exposure to nocodazole (100 ng/mL) for 20 hours. After cell cycle synchronization was confirmed by flow cytometry analysis, cells plated in 4-chamber dishes were loaded with TMRM and imaged as described in 2.3.

2.6 |. Flow cytometry analysis

Analysis of cell cycle progression was determined after cells were fixed with pure ethanol at 4°C for 1 hour, washed with PBS, and incubated with FxCycle ™ PI/RNAse Staining Solution per manufacturer’s instruction for 30 minutes at room temperature. Cell size was assessed using forward and side scatter plots on the collection software BD FACSDiva. Debris and aggregates were excluded. PI uptake was analyzed by fluorescence-activated cell sorting on flow cytometry (Fortessa X-20). PI absorbance was plotted on a histogram and the percentage of cells in each phase of the cell cycle was assessed using ModFit LT 5.0.

2.7 |. Statistical analysis

Cell cycle and comparative cell line heterogeneity data were analyzed using Kruskal-Wallis test where the data were unequally varied and ranked according to mean intensity categories across multiple comparisons. Similarly, Spearman correlation-regression statistics was applied for linear regression plots because of unequal variance. Student’s t test or nonparametric equivalent was used when comparing treatments.

3 |. RESULTS

Mitochondrial membrane potential is heterogeneous in cancer cells

3.1 |. Relative quantification of TMRM fluorescence

Fluorescence live-cell imaging in environmentally controlled chambers (37°C, CO₂ 5%) is a method of choice to assess ΔΨm in intact cells under standard cell culture conditions. The widely used potential-indicator fluorophore TMRM, a cell-permeant cation that accumulates in the negatively charged mitochondrial matrix, provides a sensitive fluorescent readout of ΔΨm. Heterogeneity of ΔΨm has been previously determined by qualitative or semiquantitative methods using either confocal microscopy or sorting by flow cytometry.^30–33 Here, we quantitatively assessed ΔΨm as relative fluorescence of TMRM in three human cancer cell lines of different tissue origins (HepG2 and Huh7 human hepatocarcinoma cells; HCC4006 human lung adenocarcinoma), and BJ1 human skin fibroblasts. In all cancer cell lines, ΔΨm was heterogeneous with marked differences in the distribution of ΔΨm in individual cells (Figure 1A,B). The mean intensity of fluorescence (in arbitrary units ± SE) was different for each cell line: 25.67 ± 0.95, 29.89 ± 1.01, 36.32 ± 1.09, and 46.88 ± 0.69 for HepG2, Huh7, HCC4006, and BJ1 fibroblasts, respectively. Heterogeneity of ΔΨm was calculated by quantifying intra-experimental differences in TMRM relative fluorescence among cells. Heterogeneity of ΔΨm, calculated as mean SD of replicates ± SE, was 19.37 ± 0.95 AU, 19.30 ± 1.09 AU, and 23.37 ± 1.01 AU for HepG2, Huh7, and HCC4006, respectively. By contrast, ΔΨm heterogeneity in BJ1 cells was lower than all the cancer cell lines: 16.26 ± 0.69 AU (Figure 1A,B). We also showed that heterogeneity was not an artifact caused by the choice of focal planes since similar heterogeneity was observed in images taken 0.6 μm apart from top to bottom of each cell (Figure S1 and 3D reconstruction).

FIGURE 1.

Mitochondrial membrane potential in cancer cells and fibroblasts is heterogeneous. HepG2, Huh7, HCC4006 cancer cells, and BJ1 fibroblasts were loaded with TMRM or Rh123, as described in Material and Methods. A, Images were pseudo-colored according to the reference bar: red: maximum ΔΨm, blue: minimum ΔΨm. Red and blue arrows indicate high and low ΔΨm cells, respectively. B, The distribution of ΔΨm in each cell line is given as strip plots overlaid with box and whiskers. Each box shows the median, maximum, and minimum points at the whiskers. Values outside the whiskers are beyond 1.5 × the interquartile range. Average SDs of the data set, used as a measure of heterogeneity, are shown parallel (vertical) to the box. Each circle represents the relative TMRM fluorescence of the entire mitochondrial network of individual cells. Results represent the analysis of 1094, 524, 347, and 260 cells for HepG2, Huh7, HCC4006, and BJ1 fibroblasts, respectively. C, HepG2 cells loaded with Rh123 were pseudo-colored as described in A. D, Strip plot overlaid with box and whisker representing the distribution of Rh123 fluorescence intensity in HepG2 cells. Note the similar distribution of ΔΨm compared to TMRM. Rh123: Rhodamine 123; TMRM: tetramethylrhodamine methyl ester; A.U: arbitrary units. Data from a minimum of three independent experiments

To confirm that intercellular heterogeneity of ΔΨm was an actual biological phenomenon and not an artifact caused by the chemical behavior of TMRM, we loaded HepG2 cells with Rhodamine 123 (Rh123), another commonly used cell-permeant potentiometric dye. The relative quantification of Rh123 fluorescence intensity showed a similar intercellular distribution of ΔΨm compared to cells loaded with TMRM (mean SD ± SE: 17.75 ± 1.59 AU) (Figure 1C,D).

Our findings, in agreement with previous reports, confirmed that ΔΨm in cancer cells is different from cell to cell and independent of the potentiometric dye used.

3.2 |. Simultaneous determination of mitochondrial and plasma membrane potentials

Uptake and accumulation of the cationic probes TMRM and Rh123 is influenced by both ΔΨm and ΔΨp. Although ΔΨm is the main driver of intramitochondrial accumulation of TMRM and Rh123, ΔΨp also influences the amount of each probe accumulating in the mitochondrial matrix. To assess whether differences observed in TMRM fluorescence among individual cells were related to intercellular differences in ΔΨp, we co-loaded HepG2 cells with TMRM and the anionic probe bis (1,3-dibutylbarbituric acid) trimethine oxonol DiBAC4(3). DiBAC4(3) remains in the cytosol and does not enter mitochondria because of its negative charge. DiBAC4(3) fluorescence is inversely proportional to ΔΨp.³⁴ The mean intensities of TMRM and DiBAC4(3) fluorescence in wild-type cells co-loaded with both dyes were 43.00 ± 3.00 AU, and 51.31 ± 3.41 AU, respectively. The distribution of relative fluorescence of TMRM and DiBAC4(3) was heterogeneous for each dye but did not correlate (R² = 0.04) (Figure 2A,B).

FIGURE 2.

Heterogeneity of mitochondrial membrane potential is not driven by intercellular differences in plasma membrane potential. A, HepG2 cells were co-loaded with DiBAC4(3) and TMRM, as described in Material and Methods. B, Distribution of TMRM fluorescence and DiBAC4(3) fluorescence values showed no correlation between the heterogeneity of ΔΨm and differences in plasma membrane potential. TMRM: tetramethylrhodamine methyl ester; DiBAC4(3): bis (1,3-dibutylbarbituric acid) trimethine oxonol, A.U: arbitrary units. Data from a minimum of three independent experiments

3.3 |. Quantification of TMRM fluorescence in absolute values

To further confirm that heterogeneity of relative TMRM fluorescence was genuinely reflecting intercellular differences in ΔΨm and not differences in ΔΨp, matrix to cell volume ratios, or differential affinities for non-mitochondrial binding between cell types, we calibrated single-cell fluorescence to absolute mV. We used an internal calibration paradigm based on a mathematical modeling of cationic and anionic dyes distribution kinetics.¹⁶ We recorded fluorescence time courses (not shown) of the calibration paradigm in HepG2, HCC4006 cancer cells, or BJ1 fibroblasts co-loaded with the anionic bis-oxonol type fluorescent ΔΨp indicator (FLIPR) and TMRM (Figure 3 and not shown).

FIGURE 3.

Mitochondrial membrane potential in absolute(mV) values differ from cell to cell. Time-lapse recordings of baseline TMRM fluorescence, TMRM decay during mitochondrial depolarization, and FLIPR accumulation after complete depolarization in HepG2 and HCC4006 cells were performed as described in Materials and Methods. A, Images of both cell lines co-loaded with TMRM and FLIPR. B, Heterogeneity of ΔΨm shows intercellular heterogeneity for each cell line. C, Correlation coefficient between absolutes values of ΔΨm and ΔΨp indicates that heterogeneity of ΔΨm does not correlate with the ΔΨp in HepG2 cells. D, Mitochondrial depolarization was initiated with MDC as described in Material and Methods, followed by complete depolarization with CDC, which allowed for calibration of the system. TMRM: tetramethylrhodamine methyl ester. Data from a minimum of three independent experiments. ^#p < .05

Absolute quantification of ΔΨm in HepG2 and HCC 4006 cells showed a remarkable similar pattern of heterogeneity compared to the relative fluorescence quantification (Figure 3A,B compared to Figure 1A,B). The distribution of ΔΨm was similar in HepG2 (mean: − 131.33 ± 10.37 mV), HCC4006 (mean: − 148.79 ± 11.65 mV), and BJ1 cells (mean: −131.68 ± 2.47) (Figure 3A,B, and not shown). Heterogeneity of ΔΨm, calculated as described in 3.1.1, was: 20.73 ± 3.33 mV, 23.30 ± 4.03 mV, and 11.04 ± 1.18 mV in HepG2, HCC4006, and BJ1 fibroblasts, respectively (Figure 1). The mitochondrial complete depolarization cocktail (MDC, see Methods), depolarized mitochondria by ~90%. Subsequent addition of complete depolarization cocktail (CDC, see Methods) completely depolarized plasma and mitochondrial membranes allowing reliable calibration of subsequent parameters as described (Figure 3D and Figure S2).¹⁶ Thus, absolute single-cell calibration of ΔΨm, that is not biased by ΔΨp, cell size, background, and auto-fluorescence, showed similar intercellular heterogeneity as compared to semiquantitative relative fluorescence analysis. As it was observed using TMRM and DiBAC4(3), the absolute quantification of ΔΨm and ΔΨp using TMRM and FLIPR, did not show correlation between the extent of heterogeneity observed in ΔΨm and that contributed by ΔΨp (R² = 0.1) (Figure 3C).

Taken together, our results show that ΔΨp, although a contributing factor, is not a major driver of ΔΨm heterogeneity. We also showed that the distribution of ΔΨm was similar between relative TMRM fluorescence and absolute quantification of ΔΨm.

3.4 |. Mitochondrial membrane potential during cell cycle progression

Cancer cells maintained in growth media are unsynchronized. At any given moment, most cells are either in G1, G2, and S phases, with few cells undergoing mitosis. Accumulating evidence indicates that mitochondria change during the cell cycle, and that mitochondrial metabolism and cell cycle progression are coordinately regulated.^35,36 Therefore, we investigated if heterogeneity of ΔΨm observed in cell cultures could be caused by cell cycle progression. To determine the relationship between each phase of the cell cycle and ΔΨm, we first synchronized HepG2 cells in either G1, S, or G2 (Figure 4A). In parallel, both unsynchronized and synchronized cells were loaded with TMRM (Figure 4B). Similar to unsynchronized cells, the spread of TMRM fluorescence was broad in G1, S, and G2, with a shift in the median distribution of ΔΨm (mean intensity of fluorescence compared to unsynchronized cells in A.U.): unsynchronized: 10.96 ± 0.62; G1: 10.75 ± 3.60, S: 14.51 ± 7.91, G2: 10.44 ± 2.46) (Figure 4C). Heterogeneity of ΔΨm, expressed as mean SD ± SE, remained similar across the different phases of the cell cycle: 7.02 ± 0.19, 6.66 ± 2.46, 6.24 ± 2.93, and 6.50 ± 0.49 in unsynchronized, G1, S, and G2, respectively. We also determined the relative contribution of high (top 25 percentile of the whole population), intermediate and low (bottom 25 percentile of the whole population) ΔΨm cells to the shift in the median distribution of ΔΨm observed in the whole cell population in the different phases (not shown). Compared to unsynchronized HepG2 cells, populations arrested in G1 phase had less high ΔΨm (12.62% vs 26.17% cells), more low ΔΨm (25.74% vs 12.60%) (p < .05), and similar percentages of cells with intermediate ΔΨm. Cells arrested in S phase showed less cells with intermediate ΔΨm (48.36% vs 61.32%), more low ΔΨm cells (28.50% vs 12.66%) (p < .05), and similar percentages of cells with high ΔΨm. Cells arrested in G2 showed a slight increase in cells with low ΔΨm (17.32% vs 12.62%), a slight decrease in cells with high ΔΨm (24.74% vs 26.17%), (p > .05), and similar intermediate ΔΨm (58.00% vs 61.38%), (p > .05). Our results showed that ΔΨm remains heterogeneous in different phases of the cell cycle with a gradual increase in the proportion of cells with high ΔΨm from G1 to G2 phases.

FIGURE 4.

Mitochondrial membrane potential heterogeneity is maintained throughout the cell cycle. HepG2 cells were synchronized in G1, S, and G2 phases by flow cytometry, as described in Material and Methods. A, Flow cytometry gates for G1, S, and G2 indicating the proportion of cells in each phase. B, Images of cells loaded with TMRM were pseudo-colored according to the reference bar, as described in Figure 1.C, Distribution of TMRM fluorescence for each phase of the cell cycle with the average SD indicating heterogeneity (vertical bar next to the whiskers). Note that heterogeneity of ΔΨm was maintained throughout the cell cycle. Data from a minimum of three independent experiments. p < .05

3.5 |. Mitochondrial hyperpolarization and depolarization after inhibition of oxidative phosphorylation and ATP synthesis

To investigate if basal ΔΨm influences the response to drugs inhibiting the respiratory chain or the synthesis of ATP, we treated HepG2 and Huh7 cells with either oligomycin (OLIGO) or antimycin A. OLIGO is a specific inhibitor of the ATP synthase and antimycin A, an inhibitor of complex III of the ETC. OLIGO decreases the utilization of H⁺ for ATP synthesis increasing the electrochemical H⁺ gradient thus hyperpolarizing mitochondria. Inhibition of complex III by antimycin A, stops H⁺ pumping at this site, resulting in mitochondrial depolarization. Importantly, in the presence of antimycin A, ΔΨm is partially maintained by the ATP synthase working in reverse to hydrolyze glycolytic ATP. We assessed changes in ΔΨm, both by relative and absolute quantification of TMRM fluorescence after OLIGO or antimycin A (Figures 5 and 6). In both HepG2 and Huh7 cells, the magnitude of the increase in TMRM fluorescence in response to the hyperpolarizing effect of OLIGO (2 μM), was inversely correlated to basal values of ΔΨm (R² = 0.49 for HepG2 and 0.64 for Huh7) (Figure 5A,B). After OLIGO, relative ΔΨm increased by ~24% in HepG2 (baseline: 18.82 ± 0.15 AU, OLIGO: 24.75 ± 0.34 AU) (Figure 5C), and by ~33% in Huh7 (5.3 ± 1.48 AU, OLIGO: 7.95 ± 0.43) (not shown). Heterogeneity of relative TMRM fluorescence for individual HepG2 cells, expressed as the average SD ± SE, was 25.34 ± 1.68 AU and 16.50 ± 1.18 AU before and after OLIGO. Calculation of ΔΨm in mV also showed an increase of ~19% after OLIGO in HepG2 cells (baseline: − 131.33 ± 10.37 mV; OLIGO: − 161.60 ± 7.56 mV, p < .05). The distribution of TMRM intensities after OLIGO was shifted toward higher ΔΨm values, reducing the heterogeneity of intercellular ΔΨm (baseline: 20.73 ± 3.34 mV, after OLIGO: 16.65 ± 3.92 mV), in HepG2 cells. (Figure 5C,D).

FIGURE 5.

ATP synthase inhibition increases mitochondrial membrane potential and decreases heterogeneity. A, HepG2and Huh7 cells loaded with TMRM, as described in Material and Methods, were imaged before (baseline), and after treatment with oligomycin (OLIGO, 2 μM) for 30 minutes. B, Single-cell relative fluorescence after OLIGO was calculated as -fold increase compared to the basal value of TMRM fluorescence. C, Heterogeneity of relative TMRM fluorescence for individual HepG2 cells before and after OLIGO. D, Heterogeneity of ΔΨm in mV before and after OLIGO in HepG2 cells. The response to OLIGO was proportionally higher in low ΔΨm compared to high ΔΨm cells. Data from a minimum of three independent experiments. ^#p < .05, N.S: non-significant

FIGURE 6.

Mitochondrial depolarization induced by inhibition of the electron transport chain is more pronounced in cells with high mitochondrial membrane potential. A, HepG2 cells loaded with TMRM, as described in Material and Methods, were imaged before and after treatment with antimycin A (2 μM) for 30 minutes. B, Single-cell relative fluorescence was calculated as -fold decrease compare to basal TMRM fluorescence. C, Heterogeneity of relative TMRM fluorescence for individual cells in arbitrary units. D, Heterogeneity of ΔΨm in mV. Note that the magnitude of the response to antimycin A was slightly correlated to the basal level of ΔΨm. Data from a minimum of three independent experiments. ^#p < .05

The decrease in ΔΨm promoted by antimycin A (2 μM) in HepG2 cells was also correlated to the basal level of ΔΨm (R² = 0.14), being more pronounced in cells with high basal ΔΨm (Figure 6A,B). Determination of absolute ΔΨm confirmed the results obtained by quantification of relative TMRM fluorescence (Figure 6C). Antimycin A depolarized mitochondria by ~45% as assessed by relative fluorescence (Baseline: 27.00 ± 1.43 AU, antimycin A: 14.89 ± 2.59 AU) (Figure 6C), and by ~19% in absolute ΔΨm (Baseline: − 151.78 ± 18.27 mV, antimycin A: − 122.38 ± 14.02 mV) (Figure 6D). Antimycin A decreased heterogeneity of ΔΨm: 17.34 ± 0.64 AU vs 9.86 ± 2.51 AU, and 17.84 ± 3.92 mV vs to 13.40 ± 4.14 mV), before and after antimycin A, respectively. Addition of OLIGO before antimycin A reduced heterogeneity even more from 19.81 ± 0.58 AU at baseline to 5.3 ± 0.15 AU after antimycin A, indicative of a contribution of ATP synthase reverse activity in the maintenance of ΔΨm (Figure S3).

Our results indicated that the activity of the ETC and the ATP synthase are major contributors to the maintenance of intercellular heterogeneity of ΔΨm in cancer cells.

4 |. DISCUSSION

Intercellular and intracellular heterogeneity of ΔΨm in cancer cells under normal growth conditions and during apoptosis have been previously demonstrated using qualitative or semiquantitative methods.^12,18,37,38 Despite being a long-known phenomenon, the molecular mechanisms causing intercellular differences in ΔΨm in physiological and pathophysiological conditions are poorly understood. Rate of cell proliferation, exposure to nutrients, and metabolic reprograming, reported as influencing tumor heterogeneity, may also contribute to heterogeneity of ΔΨm in cancer cells.^7,39,40 Here, we confirmed intrinsic intercellular variations in ΔΨm, using a dual approach to quantify relative fluorescence in combination with the assessment of the absolute ΔΨm at the single-cell level. The method utilized to calculate ΔΨm in mV using fluorescence imaging allows an unbiased quantification of absolute values of ΔΨm in individual cells.^16,18,24 Until now, most studies involving ΔΨm-indicating fluorophores failed to consider important variables that impact accurate interpretation of the chemical behavior of potential-indicators dyes. Fluctuations in ΔΨp, the matrix to cell volume ratio, the dye binding affinity, and spectral shifts upon binding in the matrix/cytosol may cause differences in fluorescence independently of the intrinsic regulation of ΔΨm.^21,41

Lipophilic and cationic fluorescent probes are widely used tools for qualitative and semiquantitative determination of relative changes in ΔΨm based on their Nernstian behavior.^42,43 Here, we used both Rh123 and TMRM, because of the strong fluorescent quantum yields for these dyes, and the sensitivity to detect subtle changes in both ΔΨm and ΔΨp (Figure 1). Rh123, has been reported to inhibit the ETC, to require extended incubation times to reach equilibrium of distribution, and to be a substrate for multidrug resistant pumps (MDR).^23,31 To rule out potential artifacts associated with the use of Rh123, we used non-quenching concentrations and sufficient incubation time as confirmed by a steady baseline after 90 minutes in the presence of zosuquidar, an MDR pump inhibitor. Inhibition of MDR pumps partially reduced ΔΨm heterogeneity; indicating that differential MDR activity among these cells is a minor contributor to intercellular differences in ΔΨm. Another factor potentially influencing TMRM fluorescence is the use of the dye in quench or non-quench mode. The low ΔΨp in hepatocytes-derived cancer cell lines predictably raised the quenching limit of probes including TMRM above the often reported ~20 nM.²⁹ Here, we confirmed that we used TMRM and Rh123 in non-quench mode by adding the uncoupler CCCP, that did not cause the transient increase in whole cell fluorescence observed when TMRM is in quench mode. The use of TMRM in non-quench mode allowed quantitative and semiquantitative assessment of native ΔΨm heterogeneity among cells and linear associations during treatments, as well as limited the possibility of direct inhibition of the ETC by these dyes.

We excluded from our experimental approach the commonly used ratiometric dye 5,5,6,6-Tetrachloro-1,1,3,3,-tetraethylbenzimidazolylcarbocyanineiodide (JC-1), because of its tendency to form J-aggregates. By contrast to TMRM and Rh123, JC-1 induces an emission shift rather than a fluorescence quench above a critical threshold concentration in the matrix. When excited at 488 nm, JC-1 monomers emit green fluorescence with a maximum at 530 nm, but when the dye accumulates in mitochondria it aggregates and undergoes a spectral shift, with emission at 590 nm (red). Other factors that prevent JC-1 from being an appropriate dye for accurate quantitative assessments of ΔΨm heterogeneity include slow accumulation across the plasma membrane and high photo-sensitivity^18,41,44

In our study, the concurrent analysis of the ΔΨp indicator DiBAC4(3) and TMRM fluorescence, did not show correlation (Figure 2). One possible explanation is the limitation of the binding kinetics and affinities of the dye not linked to actual variances in ΔΨp. DiBAC4(3) binds to intracellular proteins and hydrophobic binding sites, increasing the fluorescence of the probe.⁴⁵ Furthermore, it is reasonable to expect no correlation between the two variables if differences in ΔΨp among cells are not very high compared to differences in ΔΨm, or if the binding of the dye to subcellular structures is variable, as indicated by bright subcellular aggregates (Figure 2A). Also, differences in MDR pump activity may also obscure the predicted dependence of ΔΨm on ΔΨp. This initial interpretation about the lack of correlation between ΔΨm and ΔΨp was further confirmed by using the absolute calibrated TMRM/FLIPR assay. This method cancels geometric and binding effects using an internal calibration and it also includes the P-glycoprotein MDR pump inhibitor, zosuquidar. A partial correlation between ΔΨm and ΔΨp was observed using this approach (Figure 3C). The majority of cells in a cancer cell line showed similar low ΔΨp (~20 mV) indicating that the magnitude of ΔΨm heterogeneity was not driven by intercellular differences in ΔΨp (Figure 3C).

Mitochondrial membrane potential formation depends on a complex sequence of intramitochondrial biochemical reactions, associated to the oxidation of respiratory substrates in the Krebs cycle. NADH, a major by-product of the Krebs cycle, enter the ETC as an electron donor. The flow of electrons in the ETC creates and electrochemical gradient used for ATP synthesis. Thus, ΔΨm is an indicator of overall mitochondrial function and mitochondrial metabolism, influenced among other factors by the supply of substrates and the demand for ATP. The ATP demand depends indirectly on cellular ATP consumption and the proton leak through the mitochondrial inner membrane that makes mitochondrial ATP production less efficient. In cancer cells, aerobic glycolysis, a mitochondrial-independent supplier of ATP in the Warburg phenotype, partially cope with the cellular demand for ATP, decreasing the metabolic pressure on mitochondria. Therefore, possible sources of ΔΨm heterogeneity include variable mitochondrial substrate oxidation and proton leak, relative contribution of glycolysis, and total ATP demand. An interesting observation of our study is that the magnitude of ΔΨm heterogeneity differ between cell lines raising questions about the biological meaning of the differences.¹³ Theoretically, the higher the ΔΨm, the higher the energy capacity of the inner mitochondrial membrane to generate ATP. Although previous reports showed higher ΔΨm in different cancer cells compared to nontumor counterparts, the differences in ΔΨm have been only assessed in terms of relative fluorescence and experimental bias related to ΔΨp, geometry, binding, kinetic factors may have been underestimated.^46,47

A possible role for the balance between demand and supply on the maintenance of ΔΨm can be inferred by using specific inhibitors of respiratory complexes or the ATP synthase. ATP synthase inhibition by OLIGO reduced the heterogeneity by inducing a more robust hyperpolarization in low ΔΨm compared to high ΔΨm cells. It is likely that a differential response to OLIGO between cell populations be caused by a higher ATP demand, lower glycolysis and lower mitochondrial supply of respiratory substrates in basal low ΔΨm cells. Because of the specificity of OLIGO to inhibit the ATP synthase, no modification in the proton use could cause the differential response (Figure 5). To further study possible mechanisms explaining ΔΨm heterogeneity, we inhibited the respiratory chain with antimycin A. Antimycin A depolarized ΔΨm more intensely in high ΔΨm cells and decreased overall ΔΨm heterogeneity (Figure 6). Importantly, in the presence of antimycin A, ΔΨm was maintained at a lower level by hydrolysis of glycolytic ATP. Overall, the experiments with OLIGO and antimycin A suggest that heterogeneity of ΔΨm was contributed mostly by intramitochondrial regulatory mechanisms.

The cell cycle, by imposing different metabolic demands, could also be another cause of ΔΨm heterogeneity. Since cancer cells in culture are unsynchronized, and cells during G1, S, G2, and mitosis have different metabolic needs, we hypothesized that ΔΨm heterogeneity was dependent on the phase of the cell cycle. Synchronization of cells in G1, S, and G2 allowed us to investigate ΔΨm heterogeneity in each subpopulation of cells (Figure 4). Interestingly, ΔΨm remained heterogeneous throughout the phases of the cell cycle with shifts in the ratios of high, intermediate, and low ΔΨm cells. In particular, we observed a gradual increase of ΔΨm from G1 to G2 phase. Information regarding the coordination of mitochondrial bioenergetics and cellular proliferation is still scarce. However, a recent report showing a concurrent increase of ΔΨm, mtDNA, and mitochondrial mass from early G1 to G2, partially supports our findings.⁴⁸ The current understanding of the relationship between energy supply and the cell cycle progression points to a bidirectional interplay.^35,49,50 In line with this finding, it has been shown that depolarization of mitochondria triggered a specific G1-S arrest, demonstrating that mitochondrial function is important for G1-to-S transition.⁵¹ The underlying mechanisms are unknown and certainly will drive future investigations, for example, to simultaneously assess ΔΨm and other parameters of mitochondrial function including ATP and NADH production during cell cycle progression.

A growing number of studies use ΔΨm as a sorting parameter to identify subpopulations of interest based on the metabolic activity.^52,53 Recently, shifts in ΔΨm within undifferentiated embryonic stem cell colonies have been proposed to serve as a simple visual predictor of cells undergoing the earliest stages of differentiation.⁵⁴ This suggests that ΔΨm could also be used as an indicator of stemness. Interestingly, mitochondria with high ΔΨm, seem to be the ones transmitted across generations of cells, suggesting a preferential inheritance of the most functional mitochondria.⁵⁵

In summary, formation and maintenance of ΔΨm in cancer cells depends on several factors including the ingress of respiratory substrates to mitochondria, the activity of the ETC, the demand of ATP, and the balance between mitochondria generated ATP and glycolytic ATP among others. Here, we showed that intercellular heterogeneity of ΔΨm, mainly driven by intramitochondrial factors, may eventually become a key parameter to understand differences in metabolism among cells of the same genetic background. The study of heterogeneity of ΔΨm may lead to new research avenues to better understand tumor heterogeneity as well as to the identification of pharmacological targets to modulate mitochondrial metabolism in different tumor subpopulations.

Abbreviations:

CCCP

carbonyl cyanide 3-chlorophenylhydrazone

DiBAC4(3)

bis (1,3-dibutylbarbituric acid) trimethine oxonol

ETC

electron transport chain

OLIGO

oligomycin

Rh123

Rhodamine 123

TMRM

tetramethylrhodamine methyl ester

ΔΨm

mitochondrial membrane potential

ΔΨp

plasma membrane potential