If you don't remember your password, you can reset it by entering your email address and clicking the Reset Password button. You will then receive an email that contains a secure link for resetting your password
If the address matches a valid account an email will be sent to __email__ with instructions for resetting your password
East Carolina Diabetes and Obesity Institute, East Carolina University, Greenville, North Carolina, USADepartments of Physiology, Brody School of Medicine, East Carolina University, Greenville, North Carolina, USA
East Carolina Diabetes and Obesity Institute, East Carolina University, Greenville, North Carolina, USADepartments of Physiology, Brody School of Medicine, East Carolina University, Greenville, North Carolina, USA
East Carolina Diabetes and Obesity Institute, East Carolina University, Greenville, North Carolina, USADepartments of Physiology, Brody School of Medicine, East Carolina University, Greenville, North Carolina, USADepartments of Biochemistry and Molecular Biology, Brody School of Medicine, East Carolina University, Greenville, North Carolina, USA
Biological energy transduction underlies all physiological phenomena in cells. The metabolic systems that support energy transduction have been of great interest due to their association with numerous pathologies including diabetes, cancer, rare genetic diseases, and aberrant cell death. Commercially available bioenergetics technologies (e.g., extracellular flux analysis, high-resolution respirometry, fluorescent dye kits, etc.) have made practical assessment of metabolic parameters widely accessible. This has facilitated an explosion in the number of studies exploring, in particular, the biological implications of oxygen consumption rate (OCR) and substrate level phosphorylation via glycolysis (i.e., via extracellular acidification rate (ECAR)). Though these technologies have demonstrated substantial utility and broad applicability to cell biology research, they are also susceptible to historical assumptions, experimental limitations, and other caveats that have led to premature and/or erroneous interpretations. This review enumerates various important considerations for designing and interpreting cellular and mitochondrial bioenergetics experiments, some common challenges and pitfalls in data interpretation, and some potential “next steps” to be taken that can address these highlighted challenges.
). Recent technological advances and widespread commercial availability of instrumentation and reagents have facilitated an explosion of investigations into the consequences of gene variance in metabolic systems, with emphasis placed on processes related to cellular energy transduction through mitochondrial oxidative phosphorylation (OxPhos) and substrate-level phosphorylation in glycolysis (
Though there has been great utility in the widespread application of bioenergetics assays to biomedical science, some assumptions regarding cellular metabolic organization persist that may ultimately impede scientific progress. These assumptions can be nuanced and likely persist because they can be made to be observable by experimental conditions. As examples, the “Warburg effect” in cancerous and noncancerous cells and mitochondrial “dysfunction” in metabolic diseases (e.g., type II diabetes mellitus) can both be detected depending on conditions, but both have physiological implications that remain disputed (
). Additionally, common bioenergetics measurements have significant caveats that can dramatically impact the quality of their interpretations. For example, the ECAR to OCR ratios that are often used to “diagnose” the Warburg effect are subject to several modifiers that can impact their proportionality to the metabolic processes they are meant to represent (
The purpose of this article is to provide an overview of cellular bioenergetics assay design, emphasizing key conceptual and practical considerations, as well as perspectives on some common pitfalls in data interpretation. Some key considerations in flexible bioenergetics assay design are covered in the first section, followed by additional discussion of three topic areas of general importance: 1) cellular energy and ATP, 2) glycolysis, ECAR, and energy partitioning, and 3) mitochondrial dysfunction.
Designing bioenergetics experiments
Nearly all bioenergetics studies take advantage of the principle of steady-state mass balance to determine metabolic fluxes. In this scheme, external and/or internal fluxes are measured, then the net flux is fit to a model of the metabolic system(s) under study, and the unknown fluxes are then estimated. The main discriminating factor among different approaches lies in the degree of molecular “resolution,” or the ability to distinguish accurately between pathways or individual reactions. The current experimental modality with the highest molecular resolution is metabolic flux analysis (MFA) (
). Two alternative methods include respirometry and extracellular flux analysis, which involve time-resolved measurement of external fluxes only, paired with metabolic substrate and inhibitor protocols to reductively interpolate fluxes through specific reactions/pathways. These methods are low molecular resolution compared with MFA but are capable of measurements over short timescales, which is a limitation of MFA studies due to the typically long incubation times required to achieve an isotopic steady state (
). Because of technical challenges, MFA studies also tend to be performed by researchers with extensive experience in biochemistry/metabolism. Respirometry and extracellular flux analysis will be the continued focus of this review because of their broad use by researchers that do not necessarily specialize in bioenergetics research.
Measuring mitochondrial metabolic fluxes via oxygen consumption rate (i.e., respirometry) has been used for decades and has been essential in the development of the modern understanding of mitochondrial physiology (
). In these experiments, cells/tissues/organelles in an enclosed chamber are exposed to defined conditions and oxygen tension in the media is measured over time. The structure and coupling stoichiometries of the electron transfer system (ETS) and energetically coupled transport processes are well characterized in a variety of cell types (
). This allows the interpolation of metabolic flux through mitochondrial metabolic pathways (e.g., OxPhos) from OCR data. There are two common methods currently employed to measure OCR: 1) a potentiometric method that uses a platinum hydrogen (Clark) electrode, and 2) quenching of the fluorescence lifetime of metal-porphyrin complexes (MPCs) by the presence of elemental oxygen (
), although that information has not been made explicitly available to the scientific community. MPCs can also be prepared/purchased in soluble form and their molecular design and function have been described (
). ECAR is another external parameter that can be measured and interpolated to metabolic flux. Like the methods employed in respirometry, this can be measured using fluorescence lifetime quenching of lanthanide-based small-molecule probes (
Several additional measurements are commonly used in conjunction with measured external fluxes. For example, measurement of the electrochemical potential across the inner mitochondrial membrane (IMM) is proportional to the energetic state of the ETS and can be measured potentiometrically or with fluorescent organic cations (e.g., tetramethyl rhodamine methyl ester) (
). The redox potential of the reduced pyridine nucleotide pool (NAD(P)H) within specific subcellular compartments can be proportional to the energetic state of specific pathways and can be monitored via autofluorescence (ideally fluorescence lifetime) (
). Additionally, enzyme activity assays are cost-effective and relatively simple to perform and can be used in conjunction with other assays to provide information regarding specific reactions and/or to reveal allosteric interactions (
). There has also been significant advancement in design of genetically encoded probes that can report additional changes in internal conditions, for example, intracellular pH, GSH/GSSG ratio, NAD(P)H redox state, etc. (
A major consideration in the design of bioenergetics assays is the buffer (or media) chemistry. The media/buffer composition for intact cell versus permeabilized cells/isolated mitochondria is essentially opposite in ionic composition (i.e., intact high sodium and calcium/low potassium; permeabilized-high potassium/low sodium and calcium). Intact cells placed in buffers designed for permeabilized cells/isolated mitochondria will likely be lysed due to the hypotonic nature of the buffer (and vice versa). Both ionic strength and osmolarity should be considered to avoid significant effects on the kinetics and energetics of metabolic pathways (
). Additionally, certain ions may modulate metabolic fluxes on their own, which could necessitate the use of salts with alternative (inert) ions. For example, sodium alters isolated mitochondrial respiration rates, which can be circumvented by using tris-based salts (
To further emphasize the impact of assay media conditions on tonicity and osmolarity, consider that many metabolic intermediates are multivalent ions, which contribute substantially to both parameters. For example, a 5 mmol/l addition of sodium citrate, imparts 20 mOsmol/l to the osmolarity and 30 mmol/l to the ionic strength of the buffer. Additionally, sodium citrate is a salt of a weak acid that may raise the pH and buffering capacity of the media, particularly in low or unbuffered media that is often used for ECAR measurements. To address this type of issue, reported buffer/assay media descriptions should include information regarding changes in these parameters over the course of the assay. Additionally, controls should be included to account for the relative contributions of these changes, for example: 1) metabolically inert sugar alcohols (e.g., mannitol) to account for osmotic effects, 2) salts with alternative counterions in accordance with the type of preparation, or 3) inert weak acids with a pKa similar to the metabolite under investigation to account for pH effects.
Because respirometry and EFA are limited by low molecular resolution, experiments are typically designed to compensate by reductively separating out components of individual reactions or pathways and by measuring multiple related parameters to reduce ambiguity (
). The most common general strategy is known as substrate-inhibitor titration (SUIT). In this strategy, external flux is measured while “input” metabolite (substrate) for a given pathway is provided in increasing concentration until the saturating substrate concentration is reached. To obtain additional resolution, specific inhibitors of steps within the pathway are titrated to determine the degree of control exerted by those steps under that specific set of experimental conditions (
Importantly, there is no single way to approach SUIT design, and flexible protocol design can be generalized to a few key steps: 1) define the pathway (metabolic network model), 2) identify experimentally measurable fluxes, 3) identify candidate substrate/inhibitor combinations to isolate flux through specific pathway components, and 4) determine the most appropriate statistical design for the specific comparisons of interest. Notably, each of these steps must be very carefully considered to avoid ambiguity in the resulting interpretations. Several articles and books have been published that serve as excellent starting points for SUIT design in isolated mitochondria, permeabilized cells/tissue, and intact cells (
Choice of substrates (i.e., metabolic intermediates that support catabolic pathway flux) is significantly influenced by the model of the metabolic network under study. In permeabilized cells/isolated mitochondrial preparations, the metabolic systems are separated from their normal physiological context. This necessitates careful consideration of substrate choices because multiple metabolites may need to be present to fully support the metabolic network or prevent feedback inhibition. As an example, fatty acid oxidation (FAO) in isolated skeletal muscle mitochondria cannot proceed in the absence of low (∼0.1 mM) concentrations of malate, but can in liver mitochondrial preparations due to a different mechanism of relieving feedback inhibition by acetyl-CoA (
). Additionally, substrate transport may be influenced by several other factors including energy dependency of the transport process, substrate ionization state, substrate concentration, or cofactors required for symport/antiport mechanisms (
). For intact cells, transport of metabolites across the plasma membrane also adds an additional layer of complexity and some metabolites may require specific cofactors or growth factor stimulation for uptake (
). Ultimately the network model chosen by the experimenter must account for these factors, which will manifest differently depending on preparation type, as well as subculture conditions or cell type/tissue of origin.
There are a large number of possible inhibitors that can be used to target specific reactions within metabolic pathways and reports/reviews that may aid in identifying appropriate compounds are readily available in the literature (
). There are a few considerations to keep in mind when choosing inhibitors for bioenergetics experiments: 1) not all inhibitors are permeable to all preparation types (e.g., intact cells versus permeabilized cells/isolated mitochondria), 2) some inhibitors may be metabolized or transported out of cells, which can result in unstable kinetic effects, and 3) minimal effective inhibitor concentrations should be determined using titrations and curve fitting to determine the appropriate experimental concentration from the dose–response kinetics. Importantly, the details involved in choosing the inhibitor concentration should be reported when possible.
Another critical factor in the design of bioenergetics experiments is the choice of normalization factor(s). Metabolic flux through a given pathway will depend on the amount/activity of the enzymes and metabolic intermediates present in the sample, neither of which necessarily scales with other variables (e.g., cell volume). Cell counts are sometimes used for normalization in intact cells. However, some experimental designs may require comparisons among cells that are physically different sizes and would thus have a different total cell mass for the same number of cells. A popular way to circumvent this challenge is to normalize flux data to protein concentration, which can be measured using a bicinchoninic acid assay (
). This strategy may present challenges in situations where enzymes participating in the metabolic pathway under study are outnumbered by other soluble proteins, which act as background noise in the assay. The same considerations apply to isolated mitochondria studies, with the additional challenge that cellular proteins may contaminate the mitochondrial preparation, which may skew normalization when comparing different tissue/cell types or conditions (
). Ultimately, there is no single optimal normalizing factor. Instead, normalization must be carefully considered early in the experimental design and several options should be tested to determine the best approach.
Standardization and quality control
All bioenergetics experiments are subject to variability due to the mode of preparation. For intact cells, this may include growth conditions of the cells (e.g., media formulation, incubation conditions), cell viability (especially in drug studies), and cell type/tissue of origin. For permeabilized cells and tissues, significant variability can be introduced by the permeabilization agent and procedure during tissue isolation/separation. Additionally, isolated mitochondria preparations are very sensitive to the isolation conditions, and obtaining undamaged mitochondria is not trivial. Finally, variation may be introduced due to drift in the instrumentation being used. To account for potential sources of variation, standardization practices should be incorporated into all bioenergetics designs (and reported when possible). The simplest way to standardize the effects of preparation on bioenergetics experiments is to develop abbreviated assay protocols that are not anticipated to vary so long as underlying conditions remain the same. Several previously described assay designs can be adapted for use as standard assays, including the popular “stress tests” often implemented in the xF analyzer platforms or the respiratory control ratios described for isolated mitochondria studies (
). These can be performed repeatedly throughout the course of experiments to detect drift in preparation or instrumentation. To standardize drift in instrumentation, positive and negative control conditions can be created that test the dynamic ranges of the measurement. For example, glucose/glucose oxidase can be used to induce a high noncellular rate of oxygen consumption and cyanide can be used to completely inhibit most biological respiratory processes (
Additionally, due to the variability that preparation quality can introduce, it is important to include quality control assessments in assay design whenever possible. For example, damaged mitochondria may still respire, but at very low coupling efficiency with OxPhos, which can be assessed by using respiratory control ratios that determine the degree of kinetic control of OxPhos over total respiration (
). To account for mitochondrial damage, a cytochrome c titration is often performed to test the integrity of the outer mitochondrial membrane. Cytochrome c may also be limiting under certain assay conditions or physiological circumstances, necessitating its addition in the assay buffer (
). For intact cells, several cost-effective cell viability assay strategies are available and should be used in conjunction with qualitative morphological analysis to ensure that cells are healthy and viable. Note that some viability assays such as luciferase-based ATP assays or diaphorase activity-dependent assays (e.g., resazurin or tetrazolium dyes) are sensitive to changes in cellular metabolism and should not be used as indicators of viability. Necrosis (permeability)-based assays such as ethidium homodimer-2 or propidium iodide are recommended.
Interpreting bioenergetics experiments
The preceding section highlighted key considerations regarding general design of bioenergetics studies. The following section addresses a few issues that commonly impact the interpretation of cellular bioenergetics studies. Experimental interpretations are not just sensitive to the caveats intrinsic to the measurements being made, but also to some of the assumptions that underlie the questions being asked. The following section will focus on four areas in which both issues have and continue to influence data interpretation: 1) cellular energy and the meaning of ATP measurements, 2) energy partitioning between glycolysis and OxPhos, 3) challenges underlying interpretation of ECAR measurements, and 3) whether differences in mitochondrial functional parameters are indicative of pathological “dysfunction,” as well as strategies to discern when (or if) that is the case.
Cellular energy and ATP measurements
Though there are many energy transducing reactions within cells, adenosine triphosphate (ATP) is generally considered to be the premier energy intermediate (
). A very common strategy for approximating cellular energetic state is to associate a change in metabolic flux (e.g., OCR and/or ECAR) with a reduction in total ATP concentration (often qualitatively denoted ATP “levels”), which is commonly measured using luciferase-based enzyme coupled reaction schemes. This association is often interpreted as a failure of metabolic flux to support a given physiological ATP concentration (
). However, there are several problems with this interpretation. First, ATP levels reflect the steady-state balance between the rates of ATP synthesis and ATP hydrolysis, and thus both must be considered. Second, there is a common misconception regarding how ATP acts as an energy intermediate in cells; specifically, that the energy “stored” by ATP is held within the gamma phosphate bond. The gamma phosphate bond is sometimes referred to as “high-energy,” which is physically correct but colloquially misleading. The term “high-energy” refers to both charge repulsion and the fact that inorganic phosphate is resonance stabilized in aqueous solution (
). However, it is not the case that the bond itself “contains” energy, and it is certainly not the case that energy is released by breaking of the bond (a process that requires energy).
From a thermodynamic perspective, energy is transduced in proportion to the direction and extent of displacement of the ATP hydrolysis reaction rather than through the ATP molecule itself. Quantifying this relationship requires measuring concentrations of both the product (ATP) and reactants (ADP+Pi). The free energy of ATP hydrolysis can be expressed as a phosphate group transfer potential and quantitated by the Gibbs free energy of the hydrolysis reaction (ΔGATP, kJ/mol) (
ΔGATP is empirically derived from the equilibrium (Keq) and observed (Γ) mass action ratios of the reaction. is the Gibbs energy at physiological pH and temperature and R and T are the gas constant and absolute temperature, respectively. This equation demonstrates that the energy available from ATP hydrolysis is quantitated from its relative displacement from thermodynamic equilibrium, which in biological systems is held at a steady-state value dependent on the energetic and kinetic states of both catabolic (supply) and anabolic (demand) reactions (
In extreme situations, total ATP may be an indicator of cell health. In fact, the most popular luciferase-based assay, the cell-titer glo assay (Promega), is billed by the manufacturer as a viability assay, not as an energetic assay. A reduction in cellular ATP may indicate a reduction in the total adenylate pool due to purine degradation during energetic crisis or other physiologically relevant scenarios (
). However, if the energy status of the cell is what is being investigated, then a more comprehensive approach is required. The theoretical ATP hydrolysis mass action ratio is challenging to measure due to the many ionic forms of reactants and products involved in the gross reaction (
). This (rudimentary) approach offers an improvement over measuring ATP levels alone, as it leverages the same concept as commercial luciferase kits. It also accounts for the ATP hydrolysis mass action ratio and is not significantly more difficult to perform. Alternatively, the ATP hydrolysis mass action ratio can be determined using high (or ultra) performance liquid chromatography (
). Note that in this case, free ADP is difficult to measure because it is present in very low concentrations, and extraction methods often capture ADP that was not actually free within the cell, but rather, bound to cellular proteins. Approaches that leverage the known equilibrium constant of the creatine kinase reaction, along with HPLC measurement of other phosphagens, may be used to calculate free ADP concentrations in some tissues (
The rate of ATP turnover (JATP) is another representation of the cell’s energetic phenotype that compliments static measurements of the ATP mass action ratio. It is directly proportional to the power transferred through the phosphate potential (Box 1), which if measured over the duration of a cellular process is proportional to the total work required by the process itself (
). Importantly, there is a major caveat to this approach: i.e., the quality of the estimation is only as good as the assumed reaction coupling stoichiometries for a given metabolic pathway, and the model of the pathway is ultimately up to the experimenter. It can be very difficult to determine the coupling stoichiometries explicitly in intact cells because it is not always possible to isolate specific substrate oxidation patterns. For example, a sample of intact cells may oxidize stored carbohydrates and fatty acids simultaneously, which are predicted to have different ATP/O stoichiometries in OxPhos (
). In this case, SUITs could be designed to isolate oxidation patterns of specific substrates with known stoichiometries in mitochondria isolated from the intact cells, and the ATP/O stoichiometry could be measured directly under the same substrate conditions (
). Parallel experiments could then be performed to measure oxidation patterns in the intact cells using nutrient-restricted media or inhibitors designed to reduce ambiguity in the substrates being oxidized (
): 1) organic molecules (i.e., fuels) transfer electrons through a series of coupled redox reactions that terminate on the reduction of O2 to H2O; 2) this series of redox reactions is coupled with proton translocation to the intermembrane space, resulting in a rapid equilibration of other permeant ions and counter ions that constitutes an electrochemical potential; 3) the electrochemical potential (or proton motive force, pmf) is dissipated through membrane-bound enzyme complexes (e.g., ATP synthase) in order to perform useful work. For additional visualization, this concept has also been described through several didactic analog models (
The potential energy of each step is constrained by the maximum transfer potential of the previous step, which depends on both the magnitude of the “driving” force as well as the stoichiometry of the process (
This equation demonstrates the maximal work transfer from the combined redox reactions of the ETS to the proton motive force (pmf) for a two-electron transfer from NADH to oxygen. O indicates mole O2, F is Faraday’s constant (converts from Volts to J/mol). The stoichiometric coefficient 10 comes from assuming a proton pumping stoichiometry of 4, 4, and 2 for respiratory complexes I, III, and IV, respectively (
). ADP phosphorylation by the mammalian FoF1 ATP synthase enzyme complex is thought to phosphorylate ∼3 ADP from ∼10 protons transferred from the intermembrane space (IMS) toward the matrix, resulting in a final stoichiometry of ∼3.33 mol H+∙mol−1 ATP (
The Gibbs energy differential from pmf to ADP phosphorylation can be determined by:
Thus, the estimated maximum work that the full redox span from NADH/NAD+ to O2/H2O can transfer to the ATP phosphate group transfer potential (for this specific value of ΔGredox) has a magnitude of 3.33∙17.8 kJ mol−1 = 59.2 kJ mol−1. If the magnitude of driven force drops due to net ATP hydrolysis (demand), the driving force will tend toward this value, so long as the ETS redox potential (supply) remains unchanged.
Thermodynamics describe the direction and extent of biochemical reactions, but these expressions do not contain any kinetic information. Steady-state kinetics of mitochondrial energy transduction are similarly constrained by flux through preceding and succeeding steps as well as the coupling stoichiometries of the individual reactions that comprise the system (
). Work (or energy) transferred per unit time is denoted as power and has SI units of Watts (J∙s−1). Power is an extremely useful parameter because it encompasses both kinetic and energetic information and is capable of supplying information regarding the actual quantities of work required to perform cellular processes (i.e., by integrating the power over the time course of the process) (
). If a sample of a known quantity of isolated mitochondria or permeabilized cells respires on pyruvate/malate, then electron transfer to the CoQ/CoQH2 redox couple can be attributed to only NADH linked redox transfer through respiratory complex I (
), then an equivalent +59.2 kJ mol−1 work transfer is required to maintain it. If mitochondria respiring against a ΔGATP = −59.2 kJ∙mol−1 have a measured steady-state respiration rate of JO2 = 4.5 nmol O2∙s−1 mg−1 mitochondrial protein (
). This results in only 6 H+/O for the entire redox span, and the ATP/O2 ratio changes to:
To maintain the same rate of ATP phosphorylation (i.e., the same power output), the estimated respiration rate would need to be approximately 7.5 nmol O2∙s−1 mg−1. This is congruent with experimentally determined values using isolated mitochondria from mouse hearts (Fig. 7) (
). Together these examples illustrate a critical bioenergetic concept: two different catabolic substrates exhibit different rates of respiration but support the same power output. Even though it may be tempting to assign “dysfunction” to any condition that results in slower respiration rates, these observations may simply represent a common metabolic strategy that ensures consistent power output (
To the best of our knowledge, no similar approach is available for isolating substrate level phosphorylation stoichiometries in glycolysis, which would likely necessitate the use of isotopomer-based metabolic flux analysis (
). For these reasons, ECAR measurements must be very carefully interpreted, which is discussed in greater detail in the next section.
Cellular energy partitioning between glycolysis and OxPhos
Despite the importance of the ATP hydrolysis reaction in the function of many cells, the number of biochemical pathways that result in net phosphorylation of ADP are surprisingly sparse. The two most well-understood energy transducing reaction schemes are OxPhos and substrate level phosphorylation in glycolysis. There is a well-known difference in molar efficiency between OxPhos and substrate-level phosphorylation (i.e., ∼38 mol of ATP/mol glucose via OxPhos, 2 mol ATP/mol glucose in substrate level phosphorylation in glycolysis), which is often used as a basis for classifying the energetic economy of cells. EFA technologies have facilitated an explosion in studies that compare the relative contributions of these two pathways to cellular energetic states. The comparisons made often include multiparametric “stress” tests that use SUITs to assess several flux parameters (
). However, the interpretation of some of these parameters, particularly OCR/ECAR, can be quite problematic. The problems arise from two important issues: 1) OCR/ECAR is not quantitatively related to any metabolic fluxes, and 2) the physiological meaning of a shift in flux between these two pathways is complicated and can be artifactually created by the assay conditions.
Challenges underlying interpretation of ECAR measurements
Assumptions about the meaning of OCR/ECAR stem from the so-called “Warburg Effect” but may not be meaningful
In the 1920s, Otto Warburg famously observed that Ehrlich ascites tumors produced venous blood lactate under normoxic conditions (
). He later hypothesized that this observation could be caused by intrinsic damage to cellular respiration, citing the observation that other common causes of respiratory damage result in a similar metabolic phenotype (i.e., ischemia/reperfusion and exposure to respiratory poisons) (
). At the time, it was a very reasonable hypothesis, because fermentation was being heavily studied in yeast and skeletal muscle, neither of which performs extensive fermentation under aerobic conditions. Warburg’s hypothesis has persisted (inconclusively) for nearly 70 years and has continued to be broadly associated with cancer despite the lack of consensus model, mechanism, or consistent evidence of therapeutic potential (
). The acceptance of the Warburg hypothesis as an “effect,” seems to have also inadvertently led to persistent generalizations regarding the relationship between central carbon metabolism and cellular energy partitioning even in noncancerous cells (
). More specifically, it has reinforced the notion that some cells may “prefer” either aerobic or anaerobic glycolysis in support of their energy metabolism, and that this “preference” indicates metabolic inefficiency if the “anaerobic” pathway prevails under normoxic conditions (
). In summary, a hexose (typically glucose) is phosphorylated, isomerized, and enzymatically cleaved, ultimately resulting in 2 M equivalents of triose phosphate. In “lower” glycolysis, these triose phosphate equivalents undergo two major electron rearrangements, which provide the net driving force for substrate-level phosphorylation of ADP. Mass action ratios of glycolytic intermediates and calculations of pathway energetics have been assessed under resting conditions in erythrocytes (which notably lack mitochondria) and other subcultured cell lines (
). In those studies, substrate-level phosphorylation supported a steady-state ΔGATP between approximately −50 kJ∙mol−1 and −59 kJ mol−1, which are similar (albeit lower on average) compared with ΔGATP supported by OxPhos (
). This supports the notion that, under certain conditions (e.g., hypoxia or ETS inhibition), homolactic fermentation may support the energy metabolism of some cell types. For example, this phenomenon has been well described in ρ° cells that have been depleted of mitochondria via chronic inhibition of mtDNA replication (
Sole reliance on fermentative energy metabolism would certainly impose significant kinetic constraints on the required flux through carbohydrate metabolism, but such an effect may only be meaningful when high-power output is required over sustained periods of time. Indeed, this likely explains why genetic diseases that affect mitochondria disproportionately impact tissue types that undergo sustained increases in power output as part of their physiological function (e.g., striated muscle and neurons) (
). Thus, the limitations on cellular function imposed by energy homeostasis depend in large part on the physiological activities of the cell type, as well as the specific metabolic constraints determined by the cell’s microenvironment (e.g., nutrient availability and/or regulation of nutrient uptake) (
). The extent to which this limitation applies to cell/tissue types that do not depend on high-power output is less understood. Unfortunately, it is not uncommon for assumptions about the energetic requirements of cellular processes to be implied with very little (if any) supporting evidence. The energetic requirement of cell proliferation in cancer is one prominent example and is often assumed to be highly dependent on ATP demand (
). However, findings from several recent investigations indicate that cancer energetics are much more nuanced, and that other energy intermediates such as reduced pyridine nucleotides may be more limiting than ATP in essential macromolecular synthesis reactions that limit proliferation rate (
A more modern view of the balance in energy partitioning between OxPhos and substrate-level phosphorylation hinges on several important observations. First, both OxPhos and glycolysis transduce chemical energy while simultaneously providing molecular “backbones” and reducing equivalents for anabolic reactions (termed amphibolism) (
). This is important because either of these requirements may constrain flux depending on the context/conditions. For example, the “warburg effect” in cancer has been proposed to be a by-product of maintaining high flux through the pentose phosphate and 1-carbon metabolism pathways, which may be more important for mitosis than supporting a high rate of ATP turnover (
). Second, cells likely balance the energetic and kinetic trade-offs between the two pathways to ensure consistent power output and may use subcellular localization of mitochondria or glycolytic “metabolons” to support specific processes such as maintenance of transmembrane ion gradients (
). Finally, flux through glycolysis and OxPhos are integrated through electron “shuttling” pathways such as the malate-aspartate shuttle (MAS). Though the MAS is often depicted as a way to deliver “excess” electron equivalents to the ETS, this process is actually energy-dependent due to electrogenic transport of neutral glutamic acid for anionic aspartate (
). Manipulating MAS has been shown to induce complex effects in both OxPhos and glycolysis in several cell types, further highlighting that electron shuttling reaction schemes serve as points of integration between the two pathways (
Challenges of interpreting ECAR and ECAR/OCR ratio data and some suggested strategies for overcoming these limitations
Homolactic fermentation produces two net lactate anion equivalents from a hexose equivalent. Monocarboxylate transporters on the plasma membrane may then transport lactate out of the cell via an electroneutral lactate/H+ symport mechanism, which results in net acidification of the external microenvironment (
). Thus, in minimally buffered media, ECAR is proportional to the rate of lactate efflux resulting from homolactic fermentation (Fig. 1A). This measurement is typically included in EFA studies, but is very often misinterpreted as being a direct measure of “glycolytic flux” (
). Importantly, the proportionality of ECAR with glycolytic flux can be confounded by other sources of extracellular acidification. Additionally, typical ECAR units (mpH/min) are not stoichiometrically related to any metabolic fluxes (i.e., how many ATP/mpH/min are expected to result from glycolysis?) (
It is generally assumed that all lactate produced in glycolysis will be effluxed (Fig. 1B). However, a recent radioisotope labeling study in mice determined that circulating lactate turnover was the highest of all blood metabolites, resulting in labeling of a variety of TCA cycle intermediates in all tissues tested (
). Lactate dehydrogenase reactions tend to favor lactate formation from pyruvate at equilibrium under physiological conditions (ΔG°’LDH = −23.4 kJ mol−1) and LDH isoforms generally exhibit high activity (
). At relevant concentrations of reaction intermediates in vivo, the LDH reaction operates close to thermodynamic equilibrium, implying that the direction and extent of the reaction are sensitive to modulation by changing lactate, pyruvate, or NADH/NAD+ concentrations. This is particularly important for studies involving subcultured cells because it suggests that substrate conditions will likely affect the oxidation rate of lactate and thus skew the interpretation of EFA studies.
ECAR units can be made to be more stoichiometrically useful by conversion to a proton efflux rate (JH+; a.k.a proton production rate- PPR) by empirically determining the buffer capacity of the media via titration of known amounts of strong acid (
). In other words, Q represents the percentage of JH+ that is linked to substrate level phosphorylation. This can be difficult to determine because it can be influenced by several factors (discussed above). However, a good place to start would be to begin the assay in the absence of exogenous lactate and measure the effluxed lactate at the end of the assay using an enzyme coupled reaction (e.g., the hydrazine sink method) (
). This does, however, leave two remaining sources of error. First, in some cell types (e.g., mammalian spermatozoa), the LDH reaction produces lactate from exogenous pyruvate independent of flux through lower glycolysis (Fig. 1C) (
). The extent to which this occurs in other cell types is not known. Thus, control experiments may need to be performed in the presence of an inhibitor that targets reactions of glycolysis (e.g., α-chlorohydrin or 3-bromopyruvic acid; 2-deoxyglucose is not recommended due to confounding osmotic stress) or in the absence of exogenous pyruvate. Second, as mentioned above, lactate produced from homolactic fermentation may be oxidized in mitochondria, which would have the effect of “uncoupling” JH+ from glycolytic flux. Whether or not a given cell type retains the ability to oxidize lactate in support of cellular respiration is testable using isolated mitochondria or permeabilized cell preparations (
). Unfortunately, controlling for the magnitude of this effect in the intact cell to correct the ECAR measurement does not appear to be possible without altering the integration of OxPhos and glycolysis, because inhibiting lactate oxidation would likely alter glycolytic flux. Thus, further study would be helpful in this area.
Another caveat to ECAR measurements is that Lactate/H+ symport may not be the only source of H+ exchange (Fig. 1D). In a closed system near physiological pH, 1 mol of glucose will produce 6 mol of CO2 if completely oxidized. Each mole of CO2 generated will produce a molar equivalent of H+, therefore the CO2 produced by oxidation of substrates in the TCA cycle can contribute substantially to ECAR through reaction equilibrium with bicarbonate and carbonic acid (Fig. 1E) (
). The CO2 contribution to ECAR is far from trivial and reported increases in ECAR are very often accompanied by increases in OCR that may confound the ECAR interpretation. This effect may be accounted for by leveraging a mathematical correction factor, described by Mookerjee et al. (
). In a final word of caution, some cell types (e.g., adipocytes, polarized-epithelial cells, osteoclasts, etc) acidify their environments as part of their normal physiological function, and many cells exchange free acid for cations (e.g., Na+/H+ antiport), which may necessitate the use of additional controls (
Nonphysiological assay conditions may create artifactual glycolytic flux states
A major challenge in bioenergetics studies is that assay conditions themselves often create artifactual conditions that demonstrate questionable physiological relevance. Glycolysis has been extensively studied in a variety of cells and tissues, and nearly every enzyme in the pathway has been identified in one study or another as either energetically or kinetically limiting to overall pathway flux (
). Importantly, under normal physiological conditions, glycolysis does not appear to be regulated by cellular energetic demand. A recent study by Tanner et al. systematically overexpressed each catalyzing step in the glycolytic pathway and combined metabolic flux analysis with metabolic control analysis in two mouse cell lines (
). It was determined that glycolytic flux control under basal conditions was exerted at only four reactions: glucose transport, the hexokinase reaction, the phosphofructokinase reaction, and lactate transport (i.e., via the monocarboxylate transporter). Interestingly the ADP phosphorylating reactions of “lower” glycolysis did not exert control under resting cellular conditions. This is similar to another metabolic control study performed in primary rat hepatocytes that determined the majority of glycolytic flux control was exerted by glycolytic enzymes (i.e., ∼66% by enzymes, ∼47% by ATP hydrolysis rate) (
However, control of flux in energy transducing systems is highly sensitive to conditions, and inhibition of a step within a given pathway can result in redistribution of control to the inhibited step (
). Under extreme conditions, such as exposure to the F-Type ATP synthase inhibitor Oligomycin A, glycolytic flux exhibits an increase in net driving force as well as flux through the pathway, indicating sensitivity to increased energetic demand similar to kinetic regulation by adenylates observed in the OxPhos system (
). Increased ECAR is commonly observed response to OxPhos inhibitors (e.g., oligomycin, antimycin A, Rotenone, etc.), which is interpreted to represent compensatory flux through aerobic fermentation in support of the phosphate potential. However, this condition does not strictly represent any physiologically meaningful steady state and instead creates artifactual changes in flux that do not necessarily provide information about physiological energy partitioning. Additionally, in cells with active electron-shuttling systems (e.g., malate aspartate shuttle), inhibiting the ETS may affect glycolytic flux through the cytosolic NAD+/NADH redox potential, which may change glycolytic flux kinetics independent of energetic demand (
With these considerations in mind, how can assumptions about energy partitioning among OxPhos and glycolytic substrate-level phosphorylation be addressed moving forward? Abandoning the simplified view that cellular energy transducing pathways are independent is a recommended start. Instead, research should focus on metabolic system integration and adaptation under experimental conditions conducted in the most physiologically relevant micronutrient environment that can be reasonably achieved (
). However, buzzwords such as “max glycolytic capacity” and OCR/ECAR index simply are not meaningful and should not be used to compare experimental treatments or group effects unless this practice can be clearly justified under the circumstances.
Mitochondrial function and “dysfunction”
The details regarding the measurement of specific parameters related to mitochondrial function and OxPhos are quite numerous/complex and have been the subject of several books and reviews (
). Standardized respiratory states are commonly achieved via saturating substrate conditions in the absence of ADP (State 4; Low respiration rate- JO2), in the presence of ADP (State 3; High JO2), or via addition of ATP synthase inhibitors and protonophore uncouplers such as Oligomycin and Carbonyl cyanide-p-trifluoromethoxyphenylhydrazone (FCCP; state u; High JO2). Additional resolution is often achieved by using combinations of substrates and inhibitors to isolate activities of specific ETS complexes (
Intact cell measurements became fairly standardized with the advent of EFA and typically involve basal respiration as well as “max” or “uncoupled” respiration induced by a cell permeable protonophore (e.g., FCCP), which are often reported with calculated “ATP-linked” respiration (i.e., basal – oligomycin) and “respiratory reserve capacity” (i.e., “max” – basal) (
). As was the case for OCR/ECAR, interpretation of mitochondrial functional parameters hinges on several important assumptions about the organization of the system as well experimental caveats that should be carefully considered. This is perhaps most relevant when attributing physiological effects to mitochondrial pathology or the more nebulous (and arguably less useful) term “dysfunction.”
Assumptions about mitochondrial dysfunction stem from observed changes/differences in functional parameters that may not be meaningful
Bioenergetics studies often identify functional changes in mitochondrial metabolism (e.g., reduced respiration, membrane potential, etc) and then interpret these changes through a lens of mitochondrial pathology, leading to the conclusion that many diseases are mediated by some degree of mitochondrial dysfunction (
). However, there are a few problems with this practice. First, mitochondrial dysfunction is not really a testable condition because it lacks a consensus definition. Second, and perhaps more importantly, changes in mitochondrial functional parameters often simply reflect the underlying energetic and kinetic flexibility of the system rather than a bona fide mitochondrial pathology (
A good way to illustrate this point is through brief discussion of the debate regarding the role of mitochondrial dysfunction in the etiology of skeletal muscle insulin resistance (IR) and glucose intolerance that arise as a consequence of obesity and type II diabetes mellitus (
). This hypothesis stipulated that ectopic accumulation of proinflammatory lipids (e.g., acyl-CoAs, diacylglycerol, and ceramides) occurs because of dysfunctional mitochondrial β-oxidation, which ultimately results in stress-dependent serine kinase activity that antagonizes insulin signaling (
). The proposed mitochondrial dysfunction was suggested to be the result of age- or environment-related decline based on observations of reduced rates of TCA cycle flux and ADP phosphorylation at rest in elderly IR patients versus young weight-matched controls (
). However, from simple bioenergetic principles, a decrease in mitochondrial function at rest should have minimal impact on the ability to oxidize fatty acids because the capacity of mitochondria to oxidize substrates is far in excess of what is needed at rest. Experimentally, several transgenic models in rodents have recapitulated loss-of-function in muscle mitochondrial β-oxidation, which results in ectopic lipid accumulation but not IR (
). Additionally, reductions in state 3/state u respiration were not detected at the onset of IR in humans, and observations in rodent models have demonstrated that in vitro reductions in respiratory kinetics and changes in mitochondrial ultrastructure/content follow (rather than precede) development of IR, indicating that that these changes are simply adaptive in nature (
The connection between overnutrition and the development of IR is more likely mediated by substrate competition among fatty acid and glucose oxidation pathways, which is a manifestation of straightforward metabolic flux balance and allosteric feedback regulation of reaction kinetics by metabolic intermediates (
). In other words, these effects can be reasonably explained by the core principles of bioenergetics, rather than the presence of underlying mitochondrial pathology. More specifically, overnutrition results in an “oversupply” of catabolites, and mitochondrial energy transduction systems are energetically and kinetically limited by demand when supply is not limiting (
). Reduced rates of respiration can be normal (or even beneficial), rather than pathological, but to appreciate why requires an understanding of a few fundamental bioenergetics concepts (Box 1).
Overnutrition stimulus has been shown to be followed by adaptive reorganization of mitochondrial enzyme activity and/or content that manifests as a reduction in intrinsic mitochondrial functional parameters (i.e., in isolated mitochondria that are removed from nutritional conditions of the cell) (
). But why would an in vivo shift in substrate oxidation patterns cause this effect? Though the ATP/O ratios for carbohydrate oxidation and β-oxidation are similar (∼2.78 versus ∼2.45 respectively), the molar ATP equivalents produced from complete β-oxidation are nearly fourfold higher (
). This also means that the number of electrons donated to the ETS per mole of substrate is also about fourfold higher. Additionally, β-oxidation donates ∼2.4 times more electron equivalents to the CoQ/CoQH2 redox couple, which is not directly linked to the thermodynamic “backpressure” of proton pumping through complex I. These consequences of metabolic organization result in the production of superoxide because the demand remains unchanged, but the input “pressure” from supply is increased (
). Reactive oxygen species (ROS) have been demonstrated to induce rapid changes in various ETS enzyme activities through posttranslation modifications, and Guaras et al. showed that complex I is degraded under conditions that result in a hyperreduced CoQ/CoQH2 redox couple (
). Both effects would be expected to manifest in conditional reductions in respiratory kinetics in isolated mitochondria or intact cell preparations, but these changes should be considered homeostatic rather than pathological.
There is no doubt that mitochondria play a critical role in the physiology of many cell types. Mitochondria-derived pathologies have been experimentally induced by genetic disruption of mitochondrial ultrastructure (
). Though similar associations have been inserted into the etiology of several diseases (e.g., cardiovascular disease, neurodegeneration, and cancer), the interpretation of those data generally suffer from the same ill-defined causality as that promoted for the association of mitochondrial dysfunction with insulin resistance (
). To confidently conclude pathology, experiments must be designed to test the hypothesis that changes in mitochondrial functional parameters or cellular energy status precede the development of the disease phenotype.
Challenges involved with interpreting mitochondrial functional parameters and some suggested strategies for overcoming these limitations
Different types of preparations (i.e., isolated mitochondria/permeabilized cells (IM/PCs) versus intact cells) have major differences in both experimental design and interpretation. IM/PCs offer more control over the specific biochemical conditions that regulate mitochondrial metabolism, but lack physiological context (
), which can be useful for determining rate limiting steps when combined with inhibitor titrations but may also be susceptible to misinterpretations because the conditions are not reasonably recapitulated in live cells. Intact cell conditions are more physiologically relevant but limit unambiguous interpretation due to the complex and integrated nature of whole cell metabolic systems. Thus, interpreting data from either preparation type alone involves trade-offs between physiological relevance and experimental tractability (Fig. 2).
Nonphysiological metabolic states and artifacts
As mentioned earlier (Box 1), mitochondrial energy transduction can be modeled as an energetic and kinetic balance among four network control nodes (or modules) (
). These include: 1) substrate oxidation by matrix dehydrogenases (energy supply), 2) coupling of the ETS redox reactions ETS with proton “pumping,” 3) the electrochemical potential (ΔΨm; energy intermediate), and 4) ΔΨm dissipating processes (energy demand). Importantly, these nodes are kinetically and energetically linked, and a change in a parameter that reports on only one node cannot be interpreted unambiguously without having additional information about the other nodes. Thus, the ideal approach to defining mitochondrial function under a given set of conditions requires a multidimensional analysis that can account for the net behavior of the entire system. In fact, this concept is not new. In their widely read article on assessing mitochondrial function in cells, Nicholls and Brand highlighted the use of the respiratory control ratio (RCR) as a practical single parameter indicator of mitochondrial function, but also cautioned that a more detailed modular analysis should be performed to avoid ambiguity in the interpretation of RCRs (
One of the most prevalent limitations in the interpretation of bioenergetics studies is that mitochondrial functional parameters are measured under extreme assay conditions that (by design) elicit large effects. For example, maximal state 3 or state u respiration is often used in conjunction with SUIT strategies to identify rate or energetically limiting steps in OxPhos. These measurements are then interpreted as a diagnosis of respiratory complex deficiency or more broad mitochondrial limitation and/or dysfunction. However, this logic is flawed both conceptually and experimentally. Conceptually, a limitation measured at capacity is only relevant if the physiological system reaches that capacity, which is rarely (if ever) demonstrated. Additionally, measurements made under extreme assay conditions do not inform the potential limitations of a system functioning below maximal capacity, which more closely represents the physiological state of cells. From an experimental perspective, the conditions in which typical SUIT strategies are performed include: 1) saturated substrate transport (potentially superseding carrier mediated transport), 2) collapsed membrane potential induced by saturating ADP or uncoupling agents, and 3) adenylate mass action ratios that are opposite to those that occur in live cells, which likely understates the importance of thermodynamic and allosteric control mechanisms (
). These SUIT strategies have been time tested and are useful but were explicitly developed to interrogate the mechanisms that underlie OxPhos, not to investigate the complex roles that OxPhos and other energy-linked processes play in cellular physiology. To accomplish the latter, new iterative testable models derived from multiplexed measurements in both intact cells and IM/PCs must be developed.
The most common EFA assay applied to intact cells is sometimes referred to as a “mito stress test” and is comprised of oxygen consumption rate measurement combined with a brief SUIT protocol from which several parameters are often calculated including: basal, ATP linked, non-ATP linked (leak), uncoupled (reserve), and nonmitochondrial respiration (
). There are, however, several important considerations that can influence the interpretation of this assay and its individual parameters (Fig. 3). First, different net substrate oxidation patterns can exhibit different rates of respiration in support of the same power output (Box 1) and are difficult to control in intact cells. Second, ATP-linked respiration is determined from additions of Oligomycin, but there are several complex caveats that impact the quality of this assumption (
). Though this may seem a reasonable interpretation, empirical correlation between uncoupled respiration rates and intact tissue respiratory reserve has not been demonstrated outside of very specific circumstances (e.g., in resting versus exercising skeletal muscle) (
). The concept likely holds in skeletal muscle because myofibers undergo large changes in energy demand over short timescales during exercise. However, the extent to which this concept applies to other cell types or cells growing in subculture is not known. Finally, chemical uncoupling agents disrupt membrane transport processes that depend on voltage potential, which may result in a biased interpretation of substrate utilization due to confounding effects of these compounds alone (
). This assay design is indeed a good “starting point” and does provide some information regarding respiratory control by OxPhos. However, some of the measured and subsequently calculated parameters are limited by ambiguities or lack a clear physiological meaning (Fig. 3).
Many of the limitations discussed above can be explored further using IM/PC preparations. For example: 1) Kinetic parameters involving specific substrate oxidation pathways can be more precisely assessed by isolating specific ETS-linked reactions (
), 2) OxPhos coupling can be targeted with other inhibitors such as carboxyatractyloside or bongkrekic acid (ATP/ADP carrier inhibitors) to provide additional resolution on the degree of coupling with OxPhos (
), and 3) removing the plasma membrane barrier relieves some of the transport related ambiguity that occurs in intact cells. Despite these considerations, the lack of physiologically relevant adenylate ratios presents a major limitation to interpretation of these measurements due to the absence of normal energetic and allosteric regulatory mechanisms exerted by physiological ΔGATP. Additionally, ΔΨm measurements can only be made under extreme states involving depolarization or hyperpolarization (i.e., state 4 or state 3/u respectively), where this measurement has limited physiological relevance.
Fortunately, a collection of pioneering studies over the last few decades has led to the development of an elegant in vitro IM/PC-based system that utilizes purified creatine kinase enzyme to “clamp” the free ATP/ADP concentration ratio (and thus, ΔGATP) at fixed values by taking advantage of the known equilibrium constant of the reaction (K’CK) (Fig. 4A) (
This can be used to assess respiratory kinetics over a physiologically relevant range of phosphate potential values and allows unprecedented examination of the energetic–kinetic relationships that support OxPhos (
). Another advantage of this system is that the phosphate potential can be increased (i.e., more negative ΔGATP; analogous to reduced cellular energetic demand) through the creatine kinase reaction equilibrium by titration of known amounts of phosphocreatine (Fig. 4, B and C) (
). Conductance is the inverse of resistance, and as such, an observed change/difference in conductance among experimental groups implies that one or more source(s) of resistance in the mitochondrial energy transduction system (i.e., the four control nodes) are responsible (Fig. 4D). It is important to note that the empirical relationship becomes nonlinear at extreme values of ΔGATP and the interpretation of this assay becomes complex. Thus, the conductance is best interpreted in the “linear” range of the curve and must be optimized for different tissue/cell types to ensure that this is the case. Additionally, the free energy values are calculated from the assay conditions, which is not trivial. To aid researchers in designing and interpreting experiments, an open-source online calculator is available (Bioenergetic Calculators (dmpio.github.io)) (
The goal of force-flow analysis is to simultaneously account for the driving and driven energetic forces that influence the kinetics of mitochondrial respiration under conditions that more closely recapitulate the living cell. A change/difference in conductance implies alteration in the apparent coupling efficiency of OxPhos, which is also related to the ATP/O stoichiometry. The stoichiometries of proton translocation by the ETS and ATP synthase are generally considered to be fixed (though may vary by taxonomy) (
). Thus, the two most likely sources of variation in conductance in IM/PC preparations are derived from substrate oxidation stoichiometries (Box 1) or non-OxPhos coupled respiration. Though a change/difference in conductance does not indicate the specific source of altered “efficiency,” this can be further investigated by including additional measurements (