Forces, fluxes, and fuels: tracking mitochondrial metabolism by integrating measurements of membrane potential, respiration, and metabolites

Abstract

Assessing mitochondrial function in cell-based systems is a central component of metabolism research. However, the selection of an initial measurement technique may be complicated given the range of parameters that can be studied and the need to define the mitochondrial (dys)function of interest. This methods-focused review compares and contrasts the use of mitochondrial membrane potential measurements, plate-based respirometry, and metabolomics and stable isotope tracing. We demonstrate how measurements of 1) cellular substrate preference, 2) respiratory chain activity, 3) cell activation, and 4) mitochondrial biogenesis are enriched by integrating information from multiple methods. This manuscript is meant to serve as a perspective to help choose which technique might be an appropriate initial method to answer a given question, as well as provide a broad “roadmap” for designing follow-up assays to enrich datasets or resolve ambiguous results.

INTRODUCTION

Our understanding of mitochondrial function has largely reflected the methods used for its evaluation. Historical methods relied on biochemical techniques such as enzymology, respirometry, and absorption or fluorescence spectroscopy (1–4). Consequently, our perspective of mitochondrial function as centered mainly on electron transfer reactions and energy transduction has remained constant for decades. Increasingly, however, metabolomics and stable isotope tracing are emerging as frontline techniques (5–8), further broadening our understanding of mitochondrial function beyond oxidative phosphorylation. It is now clear that altering flux through specific metabolic pathways, without changing bioenergetic rates per se, can have substantial impacts on physiological and disease processes as varied as oncogenesis (9–13), heart failure (14, 15), immune cell activation (16–20), and neuronal excitability (21, 22). Moreover, our understanding of the TCA cycle has evolved beyond electron transfer and biosynthesis, and now includes an emerging role of TCA cycle intermediates as signaling metabolites with far-reaching effects on cell function and fate (23). As such, merging bioenergetic readouts with mass spectrometry-based approaches is a powerful, comprehensive approach for analyzing mitochondrial function in cell-based systems.

Several manuscripts and protocols are available describing the utility of respirometry (24–28), mass spectrometry (5, 6, 8, 29), and membrane potential assays (30–32) to study mitochondrial function in cells. As an addition to this work, this manuscript aims to provide a perspective for how to integrate these measurements for a thorough assessment of mitochondrial metabolism. Additionally, we hope to provide a framework for 1) deciding which technique would be an appropriate initial method for studying mitochondrial function in a given system, as well as 2) designing follow-up experiments to build on results previously obtained from other assays.

Table 1 lists some relative strengths and weaknesses for each method, and they are briefly discussed here. Respiration is a powerful technique to measure mitochondrial function because oxygen consumption is coupled to ATP synthesis through the mitochondrial membrane potential (3, 33). The potential energy harvested from nutrient oxidation and electron transport chain activity is used to drive the ATP synthase (34). Interventions that reduce oxygen consumption slow ATP synthesis and vice versa. The experimental implications of this are powerful—the oxygen consumption rate can therefore be used to detect a change in any process that either generates ATP (i.e., the oxidation of sugars, amino acids, and fatty acids) or consumes ATP (e.g., biosynthesis, ion homeostasis, proliferation, motility, etc.) (24). Practically, respirometry is a sensitive, quantitative measurement encompassing dozens of oxidative pathways into a single readout, and real-time analysis on live cells provides kinetic responses to acute compound addition. Follow-up measurements using pathway-specific inhibitors (35) and/or the use of reductionist systems such as permeabilized cells (36) or isolated mitochondria (37) allow evaluation of maximal capacities for specific metabolic pathways. Techniques include multi-well, plate-based approaches with oxygen-sensitive fluorescent dyes (32, 38, 39) as well as more traditional, Clark-type platinum-based electrode setups (25).

Table 1.

Comparison of methods to measure mitochondrial function

	Oxygen Consumption (Respirometry)	Metabolomics and Stable Isotope Tracing	Fluorescent Membrane Potential Indicators
What it measures	Oxygen consumption is the last step in the electron transport chain and an integrated measurement of oxidative mitochondrial metabolism. Since respiratory chain activity is coupled to ATP synthesis, oxygen consumption is a direct measurement of any process that generates mitochondrial ATP as well as an indirect measurement of processes that consume ATP. Platforms include platinum-based electrodes and multi-well, fluorescence-based measurements.	Abundances of intracellular and extracellular metabolic intermediates. Upfront separation methods, such as gas or liquid chromatography, will often determine the breadth of metabolites that can be detected. When conducted with stable isotope tracers (e.g., 13C6-glucose or 13C16-palmitate), the enrichment of these substrates into metabolites provides information about relative metabolic rates. Sophisticated mathematical modeling (“metabolic flux analysis”) can be used to calculate absolute fluxes.	The relative mitochondrial membrane potential across the inner membrane, which is the potential energy that is used to drive ATP synthesis and other processes. Dyes such as TMRE are lipophilic cations (membrane permeable and positively charged) and will therefore accumulate exponentially in the mitochondrial matrix in a manner dependent on the membrane potential.
Advantages	Allows real-time kinetic responses to exogenously added effectors; 24- or 96-well plate formats allow several groups per experiment; most treatments changing the rate of cellular ATP production or ATP utilization will manifest in an altered oxygen consumption rate; analysis is conducted on live cells; reductionist systems such as permeabilized cells or isolated mitochondria can be used to quickly identify precise sites of dysfunction (see Fig. 2 and 3).	Analysis provides depth far beyond other methods presented here, even for the most basic metabolomic analysis; can measure activity of specific reactions that often do not manifest in altered respiration rates or membrane potential; measurements are typically conducted under native, steady-state conditions; use of isotopically labeled substrates gives information about pathway-specific fluxes without requiring reductionist systems or inhibitors/effectors (see Fig. 1); can be extended to in vivo analysis.	Live cell analysis that provides an integrated readout of mitochondrial activity; amenable to any fluorometric live cell readout such as microscopy, flow cytometry, or a multi-well plate reader; is suitable for relatively small quantities of cells; does not require specialized equipment; potential-sensitive indicators detected by mass spectrometry or PET enable in vivo measurements.
Limitations	Pathway of interest must be rate controlling for oxygen consumption to measure differences (see Fig. 1); substrate-specific analysis often requires use of inhibitors or nonphysiological conditions that may not accurately reflect native biology; requires a relatively large number of cells, particularly when using platinum-based electrodes; Agilent Seahorse XF analyzers can be cost prohibitive for many individual laboratories.	Difficult to measure fast, real-time metabolic changes such as those occurring during cellular activation; inability to quickly iterate and relative lack of throughput compared to 96-well-based approaches that are experimentally and analytically less intensive; destructive, end point measurement rather than live cell analysis; common applications require a large number of cells; cost-prohibitive equipment and lack of specialized training in instrument operation and data analysis can be high barriers of entry.	Meaningful changes in cellular and mitochondrial metabolism can often occur without detectable changes in the mitochondrial membrane potential; there can be low sensitivity in bulk fluorescent measurements; similar to respiration, substrate-specific analysis requires reductionist or nonphysiological conditions; quantitative measurements are exceedingly difficult for the nonspecialist and rely on several assumptions.
Interpretation pitfalls	Because so many processes can alter oxygen consumption rates, pinpointing the precise metabolic change often requires reductionist systems where physiological regulation can be lost; an unaltered oxygen consumption rate does not necessarily mean there is no metabolic change—pronounced changes in substrate utilization can occur while global rates of energy metabolism remain broadly constant (Fig. 1).	Changes in metabolite abundance do not necessarily provide information about metabolic rates or directionality: steady-state metabolite abundance is dependent on both its production (“flux in”) as well as its consumption (“flux out”); inhibition of distinct electron transport chain components or enzymes involved in ATP synthesis can present similarly (Fig. 2); it can be difficult to separate compartment-specific changes between cytoplasmic and mitochondrial metabolites; care must be taken to ensure the experimental system is at metabolic and isotopic steady state.	Healthy and harmful changes in mitochondrial function can present similarly (see Figs. 3A and 4); bulk membrane potential readings can be confounded by changes in mitochondrial content (Fig. 5), or by changes in the plasma membrane potential, which controls the concentration of dye available for mitochondrial uptake; high concentrations of TMRE are self-quenching: fluorescence increases directly with membrane potential at low concentrations, but at high concentrations—where the dye self-quenches in the matrix—there is an inverse relationship between fluorescence and membrane potential; ATP hydrolysis by the ATP synthase can maintain membrane potential in response to respiratory chain dysfunction.
Complementary assays	Metabolomics and stable isotope tracing provide a breadth and depth of metabolic analysis that cannot be matched by respiration rates, which reflect the coordinated activity from dozens of metabolic enzymes into a single readout; while respirometry requires nonphysiological conditions to measure pathway-specific fluxes, this is possible under basal conditions with stable isotope tracing.	Conducting respirometry in conjunction with metabolomics and stable isotope tracing confers several advantages. These include experiments to supplement existing findings (real-time responses to perturbations and measurement of maximal pathway activities) or inform future experiments (leveraging 96-well formats for range-finding experiments or surveying metabolic pathways for targeted, follow-up mass spectrometry analysis).	Almost all interventions that change the membrane potential will also result in an altered respirometry profile, which is often a more user-friendly technique. Membrane potential assays are therefore often supportive or secondary if other assays are available. However, they are important front-line assays when sample size is prohibitively small for other approaches or when microscopy/flow cytometry are preferred.

TMRE, tetramethylrhodamine ethyl ester.

Similarly, the membrane potential also provides an integrated readout of mitochondrial function. The steady-state mitochondrial membrane potential (ΔΨ_m) is a balance between processes that generate it–mainly the electron transport chain–and processes that consume it, largely the ATP synthase but also proton leak pathways, calcium cycling, metabolite uptake, and enzymes such as the NAD(P)H transhydrogenase (40). All measurements exploit the fact that lipophilic cation probes will accumulate exponentially in the alkaline mitochondrial matrix in a manner dependent on ΔΨ_m (1) and are amenable to a range of detection systems including fluorescence spectroscopy (41), mass spectrometry (42), and radioactivity (43). As with oxygen consumption, measurements of the membrane potential integrate dozens of cellular processes into a single readout, are readily adapted to multi-well plate readers, and can be used with reductionist systems. However, changes that affect ΔΨ_m will invariably change the respiratory rate as well, and oxygen consumption is very often a more convenient and sensitive readout. As such, the use of membrane potential as a first-line measurement to study mitochondrial function is often driven by practical decisions, including limitations in sample size that preclude other approaches or preferences for microscopy or flow cytometry.

The strengths of these two techniques as integrative measurements, however, also pose a fundamental drawback: the inability to study specific metabolic pathways in-depth or under basal, unperturbed conditions. Reductionist approaches such as permeabilized cells or isolated mitochondria measure maximal capacities of specific pathways, and the use of pathway-specific inhibitors can measure the dependency of energetics on a particular substrate. However, these approaches cannot assess cellular nutrient preference under native conditions. Moreover, reductionist systems strip away cytoplasmic effects, lack the physiological relevance of studying an intact cell, and only test the maximal capacities of pathways using a group of predetermined substrates.

Almost all of these shortcomings, however, can be addressed by measuring intracellular metabolic intermediates with mass spectrometry. Even the most basic metabolomic analysis provides analytical depth unmatched by other bioenergetic techniques, particularly when combined with the use of stable isotope tracers such as ¹³C₆-glucose, ¹³C₅-glutamine, or ¹³C₁₆-palmitate. The measurements are typically conducted under basal, physiologically relevant conditions and can detect perturbations in metabolic pathways whose activity often is not reflected in integrative measurements, such as branched-chain amino acid oxidation (15, 44). Moreover, stable isotope tracing measures relative fluxes for individual metabolic reactions that are often critical to cell physiology, such as pyruvate carboxylase activity in hepatocyte mitochondria (45) or the reductive carboxylation of α-ketoglutarate (α-kg) that supports lipogenesis during hypoxia (46).

Of course, mass spectrometry-based metabolomics to study mitochondrial function is not without its drawbacks. Unlike live cell analysis using respiration or membrane potential as a readout, mass spectrometry is a destructive, “snapshot-in-time” end point measurement. It is therefore difficult to measure real-time responses to metabolic or physiologically relevant effectors, particularly given the need to conduct measurements under metabolic and isotopic steady-state conditions (5). The relative lack of technical and analytical throughput can prohibit rapid experimental iteration, and calculating absolute fluxes requires sophisticated mathematical modeling with a metabolomics approach (47). Moreover, as most approaches measure metabolite abundances and average labeling patterns across all cellular compartments, determining specific changes localized to mitochondria can be complicated (5, 48). Similar to oxygen consumption and membrane potential measurements, recent approaches have applied stable isotope tracing techniques to permeabilized cells (49) and isolated mitochondria (50), though the limitations regarding limited physiological relevance and measurement of only maximal capacities apply here as well.

Here, we compare and contrast the use of respirometry, mass spectrometry, and mitochondrial membrane potential measurements by highlighting their relative strengths and weaknesses in the context of four phenotypes: 1) altered cellular substrate preference, 2) respiratory chain inhibition, 3) acute cell activation, and 4) changes in mitochondrial content.

ALTERED CELLULAR SUBSTRATE HANDLING

Almost all healthy cells and tissues readily switch between different oxidizable substrates in response to environmental cues such as nutrient availability or hormonal stimulation (51–54). As such, cells can often change the balance between glucose, amino acid, and fatty acid oxidation without appreciably changing bioenergetic rates or even maximal capacities. Here, we provide two examples of how metabolomics and stable isotope tracing can easily identify changes in substrate handling that are difficult to reveal with oxygen consumption.

To model genetic interventions or disease states that reprogram cellular fuel preference without altering bioenergetic rates, we examined A549 lung adenocarcinoma cells treated for 24 h with either UK5099 or aminooxyacetate (AOA). UK5099 is a covalent inhibitor of the mitochondrial pyruvate carrier (MPC) (55), the protein complex that catalyzes the transport of cytoplasmic pyruvate into the mitochondrial matrix (56, 57). Although UK5099 blocks the mitochondrial uptake and subsequent oxidation of glucose-derived pyruvate, the respiratory profile is unchanged in A549 cells treated with the compound for 24 h (Fig. 1A). The same phenomenon is apparent in cells treated with AOA, a competitive inhibitor of transaminases (58). Despite AOA blocking the interconversion of amino acids and α-keto acids (e.g., conversion of pyruvate and glutamate to alanine and α-kg), 24-h treatment in A549 cells does not alter oxygen consumption rates (Fig. 1A). Maximal respiratory rates in response to FCCP are also unaffected by compound addition (Fig. 1, A and B), suggesting a high degree of metabolic plasticity sufficient to sustain maximal energy demands despite severe reductions in mitochondrial pyruvate uptake or cellular transaminase activity. An experimental implication of this metabolic flexibility is that integrative methods such as cellular respiration often do not identify metabolic changes, which remodel flux through specific pathways but leave global bioenergetic rates unchanged.

Figure 1.

Stable isotope tracing and metabolomics can easily identify changes in cellular substrate handling that oxygen consumption rates cannot. A: kinetic trace of A549 cells offered glucose, glutamine, and pyruvate in the experimental medium and treated for 24 h with either 5 µM UK5099 or 500 µM aminooxyacetate (AOA); (n = 5 technical replicates). B: respiratory parameters calculated from the trace in A. C: A549 cells treated as in A were permeabilized and offered either pyruvate/malate (Pyr/Mal), glutamate/malate (Glu/Mal), or succinate/rotenone (Succ/Rot); (n ≥ 4 technical replicates). D: schematic showing incorporation of uniformly labeled ¹³C₆-glucose into the TCA cycle and its inhibition by UK5099. E: metabolite abundances of citrate, pyruvate, and aspartate upon 24-h treatment with 5 µM UK5099 in A549 cells. F: enrichment from glucose-derived carbon into the TCA cycle intermediates citrate, α-ketoglutarate (α-kg), succinate (succ.), fumarate (fum.), and malate in A549 cells with treatment as in E. G: abundance of leucine and α-ketoisocaproate (α-KIC) after 24-h treatment with 500 µM aminooxyacetate in A549 cells. H: aspartate abundance following treatment as in G. I: enrichment of glucose-derived carbon into alanine upon treatment as in G. E–I: are all n = 3 technical replicates. All data are presented as means ± SD. Schemes of aminotransferases are presented with the directionality observed in the experiment, though each of these enzymes are bidirectional. ALT, alanine aminotransferase; AST, aspartate aminotransferase; BCAT, branched-chain aminotransferase; MPC, mitochondrial pyruvate carrier.

The use of permeabilized cells can help reveal changes in specific oxidative pathways, but this technique also has its limitations. In response to UK5099, for example, pyruvate-driven respiration is reduced in permeabilized A549 cells whereas no change is observed in respiration driven by glutamate or succinate (Fig. 1C). Although this analysis detects a defect in either mitochondrial pyruvate uptake or oxidation, it is dependent on the fact that UK5099 makes a covalent adduct on the MPC and is retained during the permeabilization protocol (36). On the other hand, there is no difference in respiration in permeabilized A549 cells offered glutamate/malate after 24-h AOA treatment, despite the reliance of this pathway on the aspartate aminotransferase (AST) to generate α-kg to fuel the TCA cycle (Fig. 1C and Supplemental Fig. S1). The result is likely attributable to i) the dilution of AOA to the experimental medium during the permeabilization protocol, ii) α-kg formed via glutamate dehydrogenase (GDH) to sustain maximal TCA cycle flux, or iii) oxidation of malate itself supported by high rates of malic enzyme activity. More broadly, though, the result demonstrates that respirometry analysis is dependent on the user selecting the appropriate experimental conditions. If a defect occurs in a pathway that is generally not studied in standard analyses [e.g., branched-chain amino acid (BCAA) oxidation] or a pathway that is not rate controlling for the overall oxygen consumption rate, then respiration may provide an inconclusive result.

Changes in cellular substrate preference, however, are readily detected with metabolomics and stable isotope tracing. In the example of UK5099, changes in both metabolite abundance and enrichment with uniformly labeled ¹³C₆-glucose (Fig. 1D, expanded scheme in Supplemental Fig. S1) can readily identify profound alterations in intracellular metabolism despite unchanged rates of oxygen consumption. For example, MPC inhibition decreases the abundance of citrate, which indicates reduced mitochondrial pyruvate oxidation (Fig. 1E) (21, 59). Accumulation in the steady-state levels of pyruvate (due to reduced pyruvate oxidation) and aspartate (due to increased glutamine oxidation and aspartate aminotransferase activity) are also readily detected (Fig. 1E and Supplemental Fig. S1). In addition to metabolite levels, enrichment from isotopically labeled substrates can also readily reveal a defect in pyruvate oxidation from UK5099. Indeed, incorporation of carbon from ¹³C₆-glucose into TCA cycle intermediates was clearly reduced upon drug treatment (Fig. 1, D and F).

Similarly, metabolomics and stable isotope tracing can easily identify a block in transaminase activity from AOA in a way that respiration rates in intact and permeabilized cells cannot. Abundances of different reactants or products of transaminase pairs are altered with AOA treatment. For example, inhibition of the branched-chain aminotransferase (BCAT) causes the accumulation of the BCAA leucine and a reduction in its corresponding α-keto metabolite α-ketoisocaproate (α-KIC), demonstrating that AOA blocks BCAA breakdown (Fig. 1G). Similarly, a drop in aspartate levels suggests reduced AST activity (Fig. 1H). Finally, stable isotope tracing can also clearly reveal changes in transaminase activity that are undetectable by respirometry. As an example, tracing with ¹³C₆-glucose shows that roughly half of the alanine pool in A549 cells consists of glucose-derived carbon from alanine aminotransferase (ALT) activity. AOA treatment entirely blocks ALT activity and the transamination of (glucose-derived) pyruvate into alanine, as shown by the wholesale disappearance of alanine labeled with three heavy carbons (“m + 3”) (Fig. 1I). Altogether, the results highlight how metabolomics and stable isotope tracing can reveal alterations in metabolic pathways difficult or impossible to localize by even sophisticated oxygen consumption measurements.

RESPIRATORY CHAIN INHIBITION

In the same way that metabolomics and stable isotope tracing have distinct advantages over respirometry (and membrane potential measurements) to pinpoint changes in cellular substrate preference, the converse is true for interventions that affect the mitochondrial respiratory chain or ATP utilization. The mitochondrial membrane potential links respiration to ATP synthesis (Fig. 2A). This coupling means that anything that blocks the production or consumption of the membrane potential, directly or indirectly, will slow rates of oxidative metabolism. As a simple model for how reductions in both substrate oxidation (ΔΨ_m production) as well as ATP synthesis (ΔΨ_m consumption) can present similarly in mass spectrometry-based approaches, we treated A549 cells with either the complex I inhibitor rotenone or the complex V (ATP synthase) inhibitor oligomycin. Both compounds sharply reduce basal rates of oxygen consumption (Fig. 2B) in a manner indistinguishable without follow-up analysis. The same is largely true for metabolomics and stable isotope tracing approaches. Inhibition of complex I and complex V both reduce steady-state levels of the TCA cycle intermediates citrate, α-kg, and succinate (Fig. 2C). Similarly, enrichment into each of these metabolites from uniformly labeled ¹³C₆-glucose is decreased as well (Fig. 2D). Given the qualitative similarity of the responses to rotenone and oligomycin, in the absence of other information, it is difficult to determine whether the reduced rates of oxidative metabolism arise from impaired substrate oxidation or a reduction in ATP synthesis/demand. Moreover, the interpretation can be further complicated by compartment-specific changes to cytoplasmic and mitochondrial metabolites that are merged in a single readout (48, 60).

Figure 2.

Inhibition of different sites of oxidative phosphorylation cannot be easily distinguished with metabolomics and stable isotope tracing. A: schematic depicting the mitochondrial respiratory chain. Respiratory complexes I–IV are depicted in blue and generate the mitochondrial membrane potential. The ATP synthase (complex V) is depicted in red and consumes the membrane potential. B: basal mitochondrial respiration in A549 cells in response to inhibition of complex I with rotenone (blue) or complex V with oligomycin (red); (n = 5 technical replicates). C: abundances of the TCA cycle intermediates citrate, α-ketoglutarate (α-KG), and succinate upon complex I or complex V inhibition in A549 cells; (n ≥ 3 technical replicates). D: incorporation of glucose-derived carbon into TCA cycle intermediates in response to inhibition of either complex V or complex I in A549 cells; (n ≥ 3 technical replicates). All treatments are for 24 h; [oligomycin] = 30 nM; [rotenone] = 300 nM. All data are presented as means ± SD. IMS, intermembrane space; Δψ_m, mitochondrial membrane potential.

However, distinguishing between these differences is readily apparent with measurements of membrane potential or oxygen consumption measurements that extend beyond basal rates. For example, plate-based fluorescent measurements with tetramethylrhodamine ethyl ester (TMRE) clearly show that rotenone and oligomycin have opposing effects on the steady-state membrane potential: complex I inhibition reduces it—due to decreased ΔΨ_m production—whereas complex V inhibition increases it because of decreased ΔΨ_m consumption (Fig. 3A). A substantial membrane potential is still present upon complex I inhibition and is mostly attributable to “reverse” hydrolysis by the ATP synthase. When the membrane potential drops below a thermodynamic threshold, the ATP synthase will hydrolyze ATP and pump protons into the intermembrane space to maintain it (3). Standard oxygen consumption measurements can also identify whether lowered oxygen consumption rates (as in Fig. 2B) are due to reductions in substrate oxidation or ATP synthesis. Specifically, these two can be easily distinguished using measurements in response to the uncoupler FCCP, which separates (or “uncouples”) respiratory chain activity from ATP synthesis. Respiration that drives ATP synthesis is entirely absent in response to both rotenone and oligomycin (Fig. 3B). However, when respiratory chain activity is disengaged from the ATP synthase and FCCP is added to consume the membrane potential (see Fig. 1A for an example), complex I inhibition continues to block respiration, whereas complex V inhibition is largely indistinguishable from controls (Fig. 3B). It is therefore clear that complex I inhibition blocks substrate oxidation, whereas complex V inhibition affects ATP synthesis or utilization.

Figure 3.

Mitochondrial membrane potential measurements and respirometry profiles can discriminate between different sites of respiratory chain inhibition. A: relative mitochondrial membrane potential measurements in response to acute (30 min) inhibition of respiratory complex I or V in A549 cells measured by plate-based TMRE fluorescence; (n = 6 technical replicates). B: ATP-linked and uncoupler-stimulated rates of respiration in response to 24-h inhibition of complex I or V. C and D: kinetic trace (C) and respiratory parameters (D) showing the effect of complex I or V inhibition on respiration driven by glutamate oxidation. E and F: kinetic trace (E) and respiratory parameters (F) showing the effect of complex I or V inhibition on respiration driven by succinate oxidation. All data for B–F are n = 5 technical replicates with 24-h drug treatment in A549 cells (concentrations as in Fig. 1). All data are presented as means ± SD. TMRE, tetramethylrhodamine ethyl ester.

Although the intact cell measurements in Fig. 3, A and B, can segment inhibitors based on whether they block production or consumption of the membrane potential, they cannot identify precise sites of dysfunction. For example, inhibition of respiratory complex I or III would provide largely similar results in the intact cell assays conducted thus far (61). Use of permeabilized cells, however, allows direct provision of pathway-specific substrates to in situ mitochondria for precisely this analysis (62, 63). For example, inhibition of complex I or V exhibited starkly different respiratory profiles in permeabilized A549s offered glutamate with malate, a substrate pair that generates NADH and therefore requires complex I for oxidation. Although glutamate-driven respiration is blocked upon complex I inhibition for all respiratory states, complex V inhibition only affects ADP-linked respiration and not FCCP-linked respiration (Fig. 3, C and D). Measuring succinate-driven respiration, a pathway where FADH₂ transfers electrons to the quinone pool and therefore does not require complex I activity, can pinpoint a defect in complex I activity. Although oligomycin again blocks ADP-linked respiration, rotenone has no effect on succinate oxidation, which requires activity of complexes II–IV and therefore localizes the effect to complex I (Fig. 3, E and F).

CELL ACTIVATION

Cellular activation is almost always associated with coordinated changes in metabolism to support this enhanced or altered function. Examples range from metabolic alterations that fuel biosynthesis in activated T cells (64), restore ion gradients in excited neurons (65), drive heat production in adrenergically stimulated brown adipocytes (66), or generate signaling metabolites in LPS-activated macrophages (67, 68). In many cases, the increased ATP requirements of cell activation are associated with enhanced oxidative phosphorylation, resulting in both increased production and consumption of the membrane potential. In Fig. 3A, we show how changes in the mitochondrial membrane potential do not always correlate with the bioenergetic status of cells, as inhibition of respiratory chain complexes I and V have opposing effects on the membrane potential, despite both sharply reducing oxygen consumption rates (Fig. 2B). Here, we build on that principle, showing that models of cell activation that increase oxygen consumption rates can also have varying effects on the membrane potential depending on cell physiology.

Figure 4A shows that both pancreatic β-cells and brown adipocytes increase oxygen consumption rates in response to activation. In the β-cell, unique metabolic features allow a structure whereby blood glucose levels are inextricably linked to insulin secretion (69). Most cell types adjust glucose uptake and oxidation to meet the energy needs of the cell. β-cells, however, “sense” glucose and adjust their oxidative metabolism in response to the external glucose concentration to facilitate glucose-stimulated insulin secretion (GSIS) (70). In the canonical model, increased glucose oxidation in response to increased supply results in a high ATP/ADP ratio, which closes plasma membrane K⁺-ATP channels. The depolarization of the plasma membrane triggers voltage-dependent calcium influx and insulin exocytosis (71, 72). In the brown adipocyte, stimulation of β₃-adrenergic receptors with norepinephrine (NE) during the physiological cold response triggers an increase in both glucose and fatty acid oxidation (66). The release of free fatty acids from NE-induced lipolysis activates uncoupling protein 1 (UCP1), which dissipates the mitochondrial membrane potential to generate heat at the expense of ATP production (73, 74). This “short-circuiting” of mitochondrial energy metabolism is a major mechanism by which brown adipose tissue drives nonshivering thermogenesis in newborn and hibernating mammals.

Figure 4.

Changes in mitochondrial membrane potential do not always track with oxygen consumption during cellular activation. A, left: basal oxygen consumption in INS-1 832/13 cells in response to low (2.8 mM) and high (17.5 mM) glucose. Right: basal oxygen consumption in primary brown adipocytes in response to acute treatment with 1 µM norepinephrine (NE); (n ≥ 4 technical replicates). B, left: schematic showing glucose sensing by pancreatic β-cells increases production of the membrane potential at a level greater than its consumption. Right: schematic showing that NE-stimulated brown adipocytes increase consumption of the membrane potential, largely by uncoupling protein 1 (UCP1), at a rate greater than its production. C, left: relative mitochondrial membrane potential in INS-1 cells in response to low and high glucose as in A, measured by plate-based TMRE fluorescence per cell; (n = 6 technical replicates), Right: relative mitochondrial membrane potential of primary brown adipocytes in response to either norepinephrine or FCCP (1 µM each). Each data point represents TMRE fluorescence of an individual cell from a single preparation of adipocytes as measured by confocal microscopy. D: representative image of a brown adipocyte before and after treatment with NE. TMRE, tetramethylrhodamine ethyl ester. [TMRE] = 15 nM for all experiments. All data are presented as means ± SD.

Although activation of both pancreatic β-cells and brown adipocytes substantially increases oxidative mitochondrial metabolism (Fig. 4A), this has very different effects on the mitochondrial membrane potential based on the relative balance of production versus consumption (Fig. 4B). In the β-cell, substrate oxidation and membrane potential generation exceed the corresponding change in consumption, resulting in an increase in the steady-state membrane potential [Fig. 4C (left)]. These increases in ΔΨ_m correlate well with increased insulin secretion and suggest that there are mitochondrial signals in addition to ATP production that act as triggers for GSIS (72, 75). In the brown adipocyte, however, the membrane potential is almost completely depolarized despite a fivefold increase in the respiratory rate. Membrane potential consumption by UCP1 far outstrips the increase in substrate oxidation upon brown adipocyte activation, phenocopying the chemical uncoupler FCCP [Fig. 4, C (right) and D] and physiologically contributing to thermoregulation upon cold exposure (76). From a practical standpoint, these results in concert with Fig. 3A show changes in membrane potential alone say little about cell physiology; increases or decreases can stem from both inhibition of oxidative metabolism or healthy physiological alterations.

MITOCHONDRIAL BIOGENESIS

Finally, we conclude with a brief example about how interventions that change mitochondrial content and biogenesis affect measurements of respiration and mitochondrial membrane potential. Of course, it goes without saying that none of the three techniques covered in this manuscript are a genuine measure of mitochondrial mass, and thus are no substitution for western analysis (77), mitochondrial:nuclear DNA ratios (78), proteomics (79), or quantitative microscopy (80). However, changes in mitochondrial content can be readily inferred from readouts of oxygen consumption and membrane potential. Additionally, when measuring membrane potential with bulk fluorescence per cell, it is essential to discriminate between whether the primary driver of a phenotype is an increase in the potential itself or rather increased mitochondrial mass.

As an example, Fig. 5 measures the bioenergetic changes elicited from treating primary bone marrow-derived macrophages (BMDMs) with the anti-inflammatory cytokine interleukin-4 (IL-4). Treatment with IL-4 elicits a well-established metabolic phenotype marked by a substantial increase in mitochondrial biogenesis and a heightened capacity for oxidative mitochondrial metabolism (81, 82). Plate-based measurements show a clear increase in bulk TMRE fluorescence per cell from IL-4 treatment (Fig. 5A). However, this analysis alone is insufficient to distinguish between i) similar numbers of mitochondria that increase their respective potentials, or ii) an increase in mitochondrial biogenesis with individual mitochondria having roughly similar potentials. Quantitative, high-resolution microscopy illustrates this point. IL-4 clearly increases the bulk TMRE fluorescence per cell [Fig. 5, B (left) C (top)], though a similar fold change is observed in mitochondrial content using MitoTracker Green [Fig. 5, B ( middle) and C (bottom)]. As such, the increased TMRE signal suggesting enhanced membrane potential is entirely due to increased mitochondrial mass, and the membrane potential per mitochondrion is unchanged [Fig. 5B (right)].

Figure 5.

Changes in mitochondrial content can present as changes in mitochondrial membrane potential. A: plate-based measurements of relative TMRE fluorescence per cell for bone marrow-derived macrophages (BMDMs) treated for 48 h with 20 ng/mL interleukin-4 (IL-4); (n = 6 technical replicates). B: all data were collected using confocal microscopy, with each point representing an individual cell from a single preparation of murine BMDMs. Left: total TMRE fluorescence per cell. Mitochondrial area per cell (middle), defined as area positive for MitoTracker Green staining. Right: quotient of the two gives the mitochondrial membrane potential per unit area. C: representative images of primary BMDMs treated with either vehicle or IL-4 for 48 h. Top: mitochondria are stained with TMRE (red) and nuclei with Hoescht 33342 (blue). Bottom: a mask using the area positive for MitoTracker Green staining shows the mitochondrial area for the 48-h vehicle and IL-4 treatments. D: standard respirometry profile of murine BMDMs treated with either vehicle or 20 ng/mL IL-4 for 48 h. Left: representative kinetic trace; (n = 5 technical replicates). Right: calculations of ATP-linked and maximal respiration (n = 2 biological replicates; data presented as means ± spread). [TMRE] = 15 nM for all experiments. [MitoTracker Green] = 200 nM. All data are presented as means ± SD unless otherwise noted. TMRE, tetramethylrhodamine ethyl ester.

In the absence of microscopy analysis that can require specialized instrumentation or training, respirometry analysis can highlight probable changes in mitochondrial biogenesis. The IL-4 response in BMDMs is a hallmark profile of mitochondrial biogenesis [Fig. 5D (left)], with a marginal change in basal respiration but a far more pronounced change in the maximal respiratory rate. The profile reflects the foundational principle that the basal and ATP-linked respiratory rates are largely limited by the energy demands of most cell types (24, 83). Put another way, increasing mitochondrial content or providing additional substrates does not markedly change initial rates of respiration because cells will not make more ATP than what is necessary to match the energetic requirements (e.g., proliferation, macromolecule biosynthesis, ion homeostasis, motility, etc.). In fact, the slight increase in ATP-linked respiration observed upon IL-4 treatment [Fig. 5D (right)] is likely attributable to an increase in ATP-consuming processes such as de novo lipid synthesis and proliferation (82, 84, 85). However, after addition of oligomycin and FCCP, the respiratory chain is pushed to its maximal rate and can operate independently of any restraints imposed by the ATP demand. Under these conditions, the increased oxidative capacity afforded by mitochondrial biogenesis results in a substantial increase in the FCCP-stimulated rate, as is observed in BMDMs with IL-4 treatment [Fig. 5D (right)]. The result does not explicitly identify mitochondrial biogenesis as the cause of the altered respiratory profile, but it provides support for more direct, follow-up measurements of mitochondrial content.

CONCLUSIONS AND FUTURE PROSPECTS

Given the remarkable advancements in our understanding of mitochondrial metabolism in cell physiology over the past 10 years, it is exciting to imagine what the future decade may bring. Undoubtedly this will include enhancements to make measurements under more physiologically relevant conditions that resemble the in vivo microenvironment (86, 87). Methodologically, as tempting as it may be to dream of technology that can encompass metabolomics as well as rate-based bioenergetics, there remains an enormous amount of progress that can be made in back-end analysis. As an example, uniform protocols, experiments, and outputs have helped standardize oxygen consumption assays, broadening their use to nonspecialists and fostering discoveries that may not have materialized without this ease of use. It is therefore tempting to consider the possibility of substantially lowering the barrier of entry for measuring mitochondrial metabolism with mass spectrometry with similar standardization, and recent software-based approaches toward this end are exciting developments (88, 89). Of course, there will always be a struggle between the rapidity and ease-of-use of “turnkey” methods and their potential for misuse without in-depth, individualized analysis. Nonetheless, it is clear that merging bioenergetic techniques with pathway-specific mass spectrometry approaches is a powerful combination that will continue to enrich our understanding of how mitochondrial metabolism can control cell physiology.

SUPPLEMENTAL DATA

Supplemental Fig. S1 and detailed methods for all studies: https://doi.org/10.6084/m9.figshare.13078025.v1.

GRANTS

This work was generously supported by National Institutes of Health Grants R35GM138003 (to A.S.D.), P30DK063491 (to A.S.D.),R01DK099618 (to O.S.S.), and T32CA009056 (UCLA Tumor Cell Biology Training Grant award to A.E.J.).

DISCLOSURES

A. S. Divakaruni has previously served as a paid consultant for Agilent Technologies. None of the other authors has any conflicts of interest, financial or otherwise, to disclose.

AUTHOR CONTRIBUTIONS

M.V. and A.S.D. prepared figures; A.E.J., A.A., and A.S.D. drafted manuscript; A.E.J., L. Sheng, A.A., M.V., O.S.S., L. Stiles, and A.S.D. edited and revised manuscript; A.E.J., L. Sheng, A.A., M.V., O.S.S., L. Stiles, and A.S.D. approved final version of manuscript.