Energy metabolism design of the striated muscle cell

Abstract

The design of the energy metabolism system in striated muscle remains a major area of investigation. Here, we review our current understanding and emerging hypotheses regarding the metabolic support of muscle contraction. Maintenance of ATP free energy, so called energy homeostasis, via mitochondrial oxidative phosphorylation is critical to sustained contractile activity, and this major design criterion is the focus of this review. Cell volume invested in mitochondria reduces the space available for generating contractile force, and this spatial balance between mitochondria acontractile elements to meet the varying sustained power demands across muscle types is another important design criterion. This is accomplished with remarkably similar mass-specific mitochondrial protein composition across muscle types, implying that it is the organization of mitochondria within the muscle cell that is critical to supporting sustained muscle function. Beyond the production of ATP, ubiquitous distribution of ATPases throughout the muscle requires rapid distribution of potential energy across these large cells. Distribution of potential energy has long been thought to occur primarily through facilitated metabolite diffusion, but recent analysis has questioned the importance of this process under normal physiological conditions. Recent structural and functional studies have supported the hypothesis that the mitochondrial reticulum provides a rapid energy distribution system via the conduction of the mitochondrial membrane potential to maintain metabolic homeostasis during contractile activity. We extensively review this aspect of the energy metabolism design contrasting it with metabolite diffusion models and how mitochondrial structure can play a role in the delivery of energy in the striated muscle.

CLINICAL HIGHLIGHTS

The design of the energy metabolism system has major implications for striated muscle function in health and disease. The ability of the muscle to maintain its available energy during alterations in workload is critical for normal function.

The inability to maintain adequate energy supply for muscle contraction has been implicated in several pathologies including heart failure, age-related dysfunction, arrythmias, muscle atrophy, ragged fiber disease, metabolic syndrome, and a host of other chronic and systemic diseases.

Beyond the content and composition of mitochondria in tissue, the structure of mitochondria into interconnected reticulum networks is proposed to be critical in the distribution of energy across the muscle cell. The role of mitochondrial structure in the metabolic design of muscle is now recognized as a critical element in the balance of energy with work in this tissue. However, the role of mitochondrial structure in different disease states has not been extensively evaluated and could be a contributing factor to many energy-related chronic and acute clinical conditions.

1. INTRODUCTION

The purpose of this review is to update the information and hypotheses regarding the mechanisms of distributing energy converted by intermediary metabolism across the mammalian cardiac and skeletal muscle cell to support muscle contraction. To understand the design of energy conversion and distribution in the muscle cell, the entire process needs to be evaluated as it pertains to its physiological function. This includes the vascular supply of basic substrates, carbon sources and oxygen, the cellular energy conversion process, primarily glycolysis and oxidative phosphorylation (OxPhos), and finally the structural and biochemical methods to control and distribute the potential energy across the cell generated by the energy conversion processes. The energetics and physiological regulation of the different elements of muscle energy conversion have been a topic of many prior articles in Physiological Reviews illustrating the continuing interest in this basic physiological process. Some titles include the following: Ventricular Energetics by Suga (1), Energetics of Muscular Contraction by Mommaerts (2), Regulation of Increased Blood Flow (Hyperemia) to Muscles During Exercise: A Hierarchy of Competing Physiological Needs by Joyner and Casey (3), Myocardial Fatty Acid Metabolism in Health and Disease by Lopaschuk et al. (4), Muscle K_ATP Channels: Recent Insights to Energy Sensing and Myoprotection by Flagg et al. (5), Myocardial Substrate Metabolism in the Normal and Failing Heart by Stanley and Lopaschuk (6), Coronary Physiology by Feigl (7), and Myoglobin-Facilitated Oxygen Diffusion: Role of Myoglobin in Oxygen Entry into Muscle by Wittenberg (8).

With such a broad scope, we have attempted to focus our attention on the performance of the muscle approaching maximum workloads under normal physiological conditions. Maximum work load is a major design criterion of the muscle in that it must be able to support maximum workload output in terms of blood flow, intermediary metabolism, and mechanics. Weibel and coworkers (9–11) have also put forward the concept of symmorphosis where all the elements associated with a given physiological performance are tuned to just meet the maximum work output requirements of the tissue with little or no excess. Simply put, symmorphosis proposes that biology optimizes its use of space and protein to accomplish the task at hand without a large, potentially wasteful, excess protein for “reserve” or as part of the regulatory system. This design concept has been called “enough but not too much” by Diamond (12), who also pointed out the importance of optimizing space in biological systems. Although details of this hypothesis have been challenged (14, 15), this concept provides a theoretical framework to focus on the structural and biochemical design of heart and skeletal muscle function approaching maximum workload conditions.

What are the maximum workload conditions for heart and skeletal muscle? For the heart, the question is rather straightforward since the heart works primarily in a sustainable manner providing constant workload over a long period of time supported, in the steady state, by coronary blood flow and OxPhos. Thus we consider the maximum work of the heart to be the maximum sustainable cardiac output under physiological conditions.

The skeletal muscle is much more complex as it has at least two types of maximum power requirements as well as many different muscle types. The two conditions we will be considering are peak nonsustainable power, where metabolic recovery capacity may be more important than the sustainable rate of energy conversion, while the other is maximal sustained power, commonly called critical power (16–18), where the energy conversion of intermediary metabolism can support muscle contractile power for a sustained period of time. Of these two measures, critical power is similar to the maximum sustainable cardiac output for the heart. With the remarkable variety of skeletal muscle fiber types, we will be seeking common elements consistent across most muscle fiber types.

The energy conversion design of the heart and skeletal muscle will be discussed separately. Due to the very different design constraints on skeletal muscle and heart, we will focus any comparisons between heart and skeletal muscle to the maximum sustained power conditions. Skeletal muscle will be evaluated simply as fast-twitch, glycolytic and slow twitch, oxidative fiber types, with specific qualifications so as to not overly simplify the true spectrum of skeletal muscle fiber types where contractile and metabolic phenotypes may diverge. Additionally, for specific elements in the energy distribution system such as myoglobin, creatine kinase, and the mitochondrial reticulum, the historical and theoretical understanding of these systems will be covered when they are introduced. A more detailed summary of the overall muscle designs will be provided in sect. 4 of the review. Finally, while it is well appreciated that deficits in energy conversion capacity within the striated muscle cell have been implicated in many pathologies, the focus of this review will be on the energy metabolism design of the healthy striated muscle cell. Reviews on the role of muscle energy metabolism in heart failure (19–21), metabolic syndrome (22–24), aging (25, 26), arrhythmias (27, 28), and atrophy (29, 30) can be found elsewhere.

2. METABOLIC HOMEOSTASIS IN STRIATED MUSCLE

The muscle is a remarkable machine, converting chemical potential energy into contractile work. It does this in a rather efficient manner in human skeletal muscle running at ∼20% total efficiency (31); for review see Ref. 32. The muscle performs work primarily using the energy conversion processes of cytosolic glycolysis and mitochondrial OxPhos, which are precisely controlled to match the energetic needs of the muscle. It is now well established that the primary energy intermediate in this process is ATP, as initially taught to us by Lohmann [see historical review published just before his death in 1978 (33)], supporting both muscle contraction as well as ion transport. Early on in working out the energy sources for muscle contraction, it was appreciated that the concentration of ATP, though clearly the potential energy intermediate between carbon substrates and work, is remarkably constant within the cell despite changes in ATP hydrolysis. Hill in his “Challenge to Biochemists” in 1950 (34) was puzzled with the role of ATP from data in intact muscle: “It may very well be the case, and none will be happier than I to be quit of revolutions, that the breakdown of ATP really is responsible for contraction or relaxation: but in fact there is no direct evidence that it is. Indeed, no change in the ATP has ever been found in living muscle except in extreme exhaustion, verging on rigor. This is explained by supposing that as soon as ATP is broken down into ADP and phosphate it is promptly restored”…….“If this happens after each stimulus, then the smallness of the changes involved and their quickness make it extremely difficult to gain any direct evidence on the subject.” At the time, Hill was focused on the Lohmann reaction (creatine kinase), discussed below. However, it took the inhibition of both creatine kinase and metabolism to demonstrate a change in ATP with muscular work from the Davies laboratory a decade later (35, 36) resolving this challenge. Thus the homeostasis of ATP can be maintained by the creatine kinase reaction or metabolism, which was subsequently demonstrated in the creatine kinase knockout mouse models as discussed further below. This phenomenon of a near constant ATP in skeletal muscle with physiological workloads was also observed in the heart for decades using conventional freeze clamp extraction assays (37–39). However, it was unclear what was happening to the phosphorylation potential since both the ADP and phosphate (Pi) were difficult to assay using these extraction techniques (40–42). The advent of ³¹P-NMR to noninvasively measure intact tissue phosphate metabolites (43) together with the use of enzyme equilibrium measures to estimate the chemical activity of ADP (44), allowed the extensive study of the effect of workload on phosphate metabolites using noninvasive physical methods. Approaching maximum respiratory rates, the entire pool lifetime of ADP is only 40–170 ms, while ATP is 2–10 s and mitochondrial NADH is 70–250 ms, making the freeze-clamp approach challenging in active tissues (45). In both heart and skeletal muscle, the original observation that the concentration of ATP was extremely well maintained has been well replicated in the literature. In most skeletal muscle types, a modest decrease in phosphocreatine (PCr) and increase in Pi is accompanied with an increase in work resulting in a concomitant increase in the calculated ADP and decrease in free energy for ATP hydrolysis while the ATP concentration ([ATP]) remains near constant below exhaustion levels of contraction (31, 46, 47). Although the creatine kinase reaction can transiently maintain ATP using PCr for a few seconds (48), the support of muscle [ATP] for sustained striated muscle work must come from metabolically generated ATP synthesis. Thus the regeneration of ATP via intermediary metabolism, as originally shown by Davies laboratory (35, 36), is critical for maintaining ATP constant within the cell. It is important to point out that with exhaustive exercise, decreases in [ATP] have been reported in some human subjects (32–34) associated with decreases in muscle oxygen content (49) consistent with compromised metabolism due to hypoxia during sustained intense exercise.

In the heart, metabolic homeostasis occurs over a wide range of physiological workloads. Focusing on in vivo cardiac studies, with physiological increases in workload, not only is ATP held constant but PCr and Pi are also unchanged, at the sensitivity of ³¹P-NMR, until near maximum experimental workloads are achieved across mammalian species (50–55). These studies have also been reproduced in healthy human volunteers (56, 57).

The ability of the heart, and to a lesser extent the skeletal muscle, to maintain metabolic homeostasis during increases in work is now a well-established phenomenon (58–60). The small alterations observed in even highly aerobic fibers when compared with the heart might be related to the large dynamic range of skeletal muscle energy demands compared with heart. However, the mechanism by which metabolic homeostasis is maintained in all muscle cells is still unknown. The classical model of a simple feedback of ADP and Pi to drive oxidative respiration and glycolysis (61–63) does not seem to be adequate to drive these processes without severe alterations in the kinetics of ATP and Pi utilization (for examples, see Refs. 64, 65) or compartmentation of substrates (see example Ref. 66). Other proposals suggest that there are parallel activation schemes where muscle contractile activity and metabolism are activated in parallel potentially via intracellular calcium (67–71). However, the dominance of calcium as a parallel regulator of OxPhos has been challenged by the survival and modest effect of removing the putative mitochondrial calcium transporter (MCU) in heart and skeletal muscle in the mouse (72–74), although MCU knockout mice are not born at the expected Mendelian ratios and some have indeed reported significant metabolic effects in the MCU knockout mouse (75, 76). However, the fact that a mouse whose heart operates near its maximum capacity at rest survives the removal of MCU, as well as creatine kinase and myoglobin discussed below, with modest changes in metabolic regulation suggests that the signaling network that supports muscle metabolic homeostasis is complex and likely does not depend on only one or two factors but an ensemble of signaling processes to maintain this critical process (77–79). It is clear that to accomplish this task a coordinated modulation of blood flow, substrate entry, substrate oxidation, and the entire OxPhos cascade needs to be activated to accomplish this rather remarkable bioengineering task. Fell and Thomas (77) and Korzeniewski (80–82) have presented theoretical models supporting this notion that the modulation of ATP production needs to occur over most of the steps in the reaction network. No single signaling mechanism like Ca²⁺ (69), AMPK (83), or even local metabolite (Pi, ADP, etc.) feedback (84) can likely coordinate these processes alone. Likely an integrated signaling network is responsible that we are just beginning to understand and will not be the focus of this review. Intrinsic to this process is the distribution of potential energy through the cell, which will be discussed further below.

The observed metabolic homeostasis first discussed by Hill (34) also has consequences with regard to the distribution of free energy in the muscle cell. That is, the kinetics of metabolically converted free energy distribution needs to match, or exceed, the rate of muscle cell ATP hydrolysis/synthesis and be done in such a way that the concentration of ATP does not change significantly. The complication here is that changes in metabolite diffusion rates require proportional changes in concentration gradients, although small diffusion distances, facilitated diffusion, and specialized gradient barriers could explain some of this phenomenon. However, the fact that large changes in ATP, ADP, and Pi flux occur in striated muscle without significant alterations in the concentration of these metabolites brings into question the role of metabolite diffusion in the regulation of overall energy conversion rates. Much of this review will focus on hypotheses involving a mitochondrial reticulum to distribute free energy in the form of the mitochondrial membrane potential that rapidly conducts through the muscle cell via a regulated mitochondrial reticulum thereby avoiding significant metabolite diffusion requirements.

3. METABOLIC SUPPORT DESIGN OF THE MUSCLE CELL

The heart needs to function, in the steady state, over a wide range of workloads and associated metabolic rates. This steady-state work is almost exclusively supported by the energy conversion occurring in OxPhos as illustrated by the small energy conversion capacity of glycolysis in the heart (85, 86). With this limitation, the maximum rate of ATP production by OxPhos should be matched to the maximum need for energy conversion in support of sustained myocardial contraction (i.e., myofibril activity) and control (i.e., ion transport) as well as basal maintenance of the myocyte [up to 30% of resting heart metabolism (87)]. Supporting this notion, the heart maximum rate of mitochondrial OxPhos, estimated from cytochrome oxidase (COX) content, essentially matches the maximum rate of respiration determined in the exercising animal (88, 89). These observations suggests that the design of the heart permits the delivery of substrates, oxygen and carbon metabolites, to not be rate limiting for OxPhos at maximum rates with little reserve capacity of OxPhos. These observations are consistent with the concept of symmorphosis (9–11) discussed above.

In contrast to cardiac muscle, which is designed to maintain energy homeostasis across nearly the entire range of contractile demands in the constantly beating heart, the design of the intermittently contracting mammalian skeletal muscle places greater emphasis on maximal contractile power at the expense of reducing oxidative energy metabolism. Consistently, whereas oxygen delivery is largely nonlimiting under physiological conditions in the heart due to high capillarization, relatively smaller cell sizes, and stable po₂ gradients as discussed above, lower capillary densities and larger cell sizes in skeletal muscle fibers contribute to larger and more dynamic oxygen gradients from the capillary to the mitochondrion (90–95). As a result, the likelihood for oxygen delivery limitations is much greater in skeletal muscle cells and even more so in glycolytic compared with oxidative muscle fibers. Indeed, endurance exercise training in both rodents and humans has been reported to result in a twofold increase in skeletal muscle mitochondrial content, yet only ∼15% change in maximal oxygen uptake (96, 97). These data suggest that while mitochondrial content correlates well with performance on submaximal endurance tests (96, 98), mitochondrial content does not appear to be a major limitation to maximal oxygen uptake by the skeletal muscle cell (99, 100). The exact contributions of specific limiting factors to maximal oxygen uptake remain hotly debated, however (101–106). It is important to note that critical power, or the maximal sustainable skeletal muscle work rate mentioned above, generally occurs at ∼70–80% of the rate of maximal oxygen uptake in humans and thus may not face the same oxygen supply limitations. The primary components of the oxygen delivery system interacting directly with the muscle fiber, capillaries, and myoglobin will be discussed below. Recent in-depth reviews of oxygen delivery pathways upstream from the capillaries can be found elsewhere (3, 90, 107–109).

3.1. Major Muscle Cell Energy Conversion Processes

To understand the distribution of potential energy across the muscle, the sources of energy conversion to ATP need to be evaluated along with the topology of these sites in the cell. The primary sources of ATP during muscle contraction are creatine kinase, glycolysis, and OxPhos (FIGURE 1). Glycolysis and OxPhos provide the large majority of ATP during contractions lasting beyond a few seconds and will be the major focus of this section. However, discussion of creatine kinase will be included below as part of the facilitated diffusion section.

FIGURE 1.

Major adenosine triphosphate (ATP) production and utilization processes in striated muscle. Muscle contraction and maintenance of ion gradients across membranes are the major ATP utilizing work processes in striated muscle cells. Metabolic homeostasis is maintained by ATP production through the creatine kinase, glycolysis, and oxidative phosphorylation energy conversion processes. PCr, phosphocreatine; Cr, creatine; ADP, adenosine diphosphate; HK, hexokinase; PGI, phosphoglucose isomerase; PFK, phosphofructokinase; ALDO, aldolase; GAPDH, glyceraldehyde 3-phosphate dehydrogenase; TPI, triose phosphate isomerase; PGK, phosphoglycerate kinase; PGM, phosphoglycerate mutase; ENO, enolase; PK, pyruvate kinase; LDH, lactate dehydrogenase; G6P, glucose 6-phosphate; F6P, fructose 6-phosphate; F1,6BP, fructose 1,6-bisphosphate; GA3P, glyceraldehyde 3-phosphate; DHAP, dihdroxyacetone phosphate; 1,3BPG, 1,3-bisphosphoglycerate; 3PG, 3-phosphoglycerate; 2PG, 2-phosphoglycerate; PEP, phosphoenolpyruvate; β-Ox, fatty acid β-oxidation; TCA, tricarboxylic; CI, complex I (NADH dehydrogenase); Q, ubiqinone; CIII, complex III (ubiquinol-cytochrome c reductase); c, cytochrome c; CIV, complex IV (cytochrome oxidase); CV, complex V (ATP synthase); ANT, adenine nucleotide translocase.

With regard to the enzymes involved in the major energy conversion processes, it is reasonable to assume that protein content of an enzyme is a reflection of its relative maximum velocity based on the simplified Michaelis-Menten kinetics, where V_max is assumed to be proportional to the concentration of the enzyme as V_max∼k_cat[E], where k_cat is the enzyme rate constant that can depend on many parameters including pH, ions, allosteric interactions, and posttranslational modifications. However, with all of these modulators being equal, the enzyme concentration can be used to estimate the “potential” V_max of the system (110–112). This is important for this review to understand the relative capacity of the different muscles to produce ATP together with the localization within the cell. The maximum velocity of enzymes, especially putative rate-limiting enzymes, has been used in the past to estimate the maximum ATP production of glycolysis [glyceraldehyde-3-phosphate dehydrogenase (113, 114), phosphofructokinase (115), and enolase (112)]. Some models have suggested other glycolytic enzymes could be rate limiting (114); however, at this stage, the essentially irreversible phosphofructokinase remains the primary candidate as a rate limiting step (116–118). While estimates of OxPhos have been made from enzymes of the citric acid cycle [oxoglutarate dehydrogenase (113)] or the electron transport chain [cytochrome oxidase (89, 119)], there does not appear to be a single rate limiting enzyme for OxPhos. Instead, control of flux through the OxPhos process is shared among nearly all of the enzymes (68, 120–122). It is important to note that the use of enzyme concentrations to estimate V_max assumes that the k_cat is the same for different tissues at maximumly stimulated workloads, but this is often complicated by different tissue isoforms with different kinetic properties as well as the artificial environment an enzyme is exposed to for in vitro studies (123). Even so, the value of this type of data has been well appreciated in the literature as numerous studies on the protein composition or enzyme activities of muscle tissues have been created (111, 124–135).

A challenge in the field has been to rely on the more difficult determination of absolute protein content rather than the relatively easy enzymatic activity of tissue extracts. Given the dependence of enzyme activity on the assay conditions and purification conditions (136), we have selected to use protein content measured with quantitative antibody or, preferably, physical measures such as mass spectrometry or optical spectroscopy to characterize the flux capacity of different muscle systems. Despite the extensive work on the protein content of muscle, a relational quantitative study of protein content in the three muscle types we are considering in this review, heart, slow-twitch, and fast-twitch muscles from a single species, is not in the literature. To help our analysis to assimilate data from many different sources, we created a relational database of the proteins of the rabbit heart, soleus (98% slow-twitch), and gracilis [99% fast-twitch (137)] muscles using standard contemporary mass spectrometry labeling techniques. While we only focus on the creatine kinase, glycolytic, and OxPhos enzymes here, although the other detected proteins will be provided in Supplemental Table S1 (see https://doi.org/10.6084/m9.figshare.14109524.v1). We selected rabbit since it provided adequate sample for analysis, is one of the most extensively characterized muscle systems in the literature, and contains highly homogenous fast- and slow-twitch muscles (gracilis and soleus, respectively). In this comparison, we found over 3,000 proteins across the different tissue types using mass spectrometry (data presented in Supplemental Table S1). Again, we focus on the protein programming of these muscle cells for the metabolic energy conversion systems including putative facilitated diffusion systems such as creatine kinase. This analysis will use the extensive prior literature data together with the relational database in the rabbit to illustrate the protein programming associated with the overall energy conversion in striated muscle cells.

3.1.1. Glycolysis.

Glycolysis is an important energy source both in the form of ATP as well as for providing pyruvate for OxPhos. The relative database created for this review can be converted into a quantitative analysis of glycolysis using the baseline concentrations of glycolytic enzymes provided by Maughan et al. (125) of the free cytosolic proteins in the fast-twitch psoas muscle of the rabbit. This analysis is presented in TABLE 1. It has been noted by Maughan et al. (125), and others (111), that their values are considerably higher than prior work using primarily enzyme activities in extracts (138). We have selected to use Maughan et al. (125) for these analysis as they are the only quantitative physical measure of the glycolytic enzymes of the rabbit currently available.

Table 1.

Concentrations of glycolytic enzymes in rabbit muscles

Glycolytic Enzyme	Soleus, µmol/L cell volume	Gracilis,* µmol/L cell volume	Heart, µmol/L cell volume
Glycogen Phosphorylase	8	43	5.6
Phosphoglucose isomerase	5	29	9
Phosphofructose kinase	5	35	8
Aldolase	8	54	6
Triose-phosphate isomerase	20	82	13
Glyceraldehyde-3-phosphate dehydrogenase	16	74	10
Phosphoglycerate kinase	30	131	20
Phosphoglycerate mutase	16	69	12
Beta enolase	15	108	12
Pyruvate kinase	11	78	7
Lactate dehydrogenase†	25	76	40

^*Gracilis data equated to rabbit psoas data from Ref. 125. †Lactate dehydrogenase concentrations are based on the sum of both the A and B isoforms.

Summing the mass of these proteins by grams of protein per liter cell, we find that the gracilis muscle has over 60 g of protein per liter cell devoted to glycolysis while the soleus and heart muscle have only 9 and 10 g/l, respectively. Using a protein density of 1.37 g/ml (139), this corresponds to ∼9% of the gracilis muscle cell volume is devoted to glycolytic enzymes while only 1–2% of the cell volume of the soleus and heart. Note that the remaining space is primarily consumed by myofilaments and mitochondria as discussed below, as well as sarcoplasmic reticulum (SR). These rather large differences reflect the known heavy dependence of the gracilis, a glycolytic fast twitch muscle, on the production of ATP from glycolysis when compared with soleus or heart muscle.

What is the maximum ATP conversion rates of glycolysis in these muscles? It is tempting to take the concentration of phosphofructokinase and its specific activity to arrive at a maximum velocity for glycolysis. However, this approach in the past has grossly overestimated the physiological fluxes observed. For example, Newshome and Crabtree (113) estimated from phosphofructokinase activity the maximum ATP production rate in the rat heart at 60 µmol ATP/min/g (corrected for temperature to 37°C). However, Kobayashi and Neely (85) found only ∼5 µmol ATP/min/g wt in rat heart while in the rabbit heart maximum glycolytic ATP production rate is ∼6 µmol ATP/min/g wt (140) [both calculations assume a wet/dry weight of 5.9 (141)] Similarly, the maximum glycolytic ATP production rates estimated from PFK activity in male and female human tibialis anterior muscle [∼180 and 140 µmol ATP/min/g wt, respectively (142)] exceed the measured maximal glycolytic rates in male and female human tibialis anterior muscles [∼60 and 30 µmol ATP/min/g wt, respectively (143)] by severalfold. Thus there is either significant reserve capacity in these enzymes under some conditions (115) or the extract assays do not reflect the true activities of these cytosolic enzymes (113, 115). This is likely even more important for many of the OxPhos enzymes that reside in a membrane with a very complex reaction environment including a membrane potential etc. As a result, we will consider the proteomics data to provide relative V_max capacities, but they cannot, at this time, be used to estimate absolute fluxes.

3.1.2. Oxidative phosphorylation.

The mitochondrial protein content of the heart is clearly much higher than the other muscle tissues. Summing all of the mitochondrial proteins, we find that for the 304 highest abundance mitochondrial proteins, the ratio of mitochondrial proteins compared with heart was 36 ± 2 and 4 ± 0.1 for heart/gracilis and heart/soleus, respectfully. Due to the remarkable consistency in the ratios across tissues for these mitochondrial enzymes, the protein content for OxPhos has been simplified down to the levels of COX, the only irreversible step in the entire cascade from NADH to the terminal reduction of oxygen to water. Thus the content of COX serves as a good estimate of the OxPhos capacity of a system. We used the quantitative data on COX content in the rabbit heart from Phillips et al. (89) at 35 µmol/L cell volume and from Glancy and Balaban (144) at 12.2 and 1.9 µmol/L cell volume for the rabbit soleus and gracilis, respectively (TABLE 2). For reference, the relative COX abundance in each muscle type can also be calculated using our relational proteome database (Supplemental Table S1). Assessing the 14 COX subunits detected in each tissue yields a 21.7 ± 0.1- and 4.4 ± 0.0-fold greater COX concentration in the heart compared with the gracilis and soleus, respectively.

Table 2.

Concentrations of cytochrome c oxidase in striated rabbit muscle

Protein	Soleus,† µmol/L cell volume	Gracilis,† µmol/L cell volume	Heart*, µmol/L cell volume
COX	12.2	1.9	35

COX, cytochrome c oxidase. *Heart value from Phillips et al. (89). †Soleus and gracilis values from Glancy and Balaban (144).

These data support the notion that the heart has the largest capacity for OxPhos when compared with the other striated muscle systems. Note that most of the enzymes associated with OxPhos are upregulated in the heart and soleus suggesting that the entire network cascade is required to provide more ATP from OxPhos. Moreover, it appears that all of the enzymes within mitochondria seem to be in proportion across the different muscle types as illustrated in FIGURE 2, A–C, for the top 152 abundant mitochondrial proteins, normalized to total tissue protein. The ratio of mitochondrial protein content is remarkably consistent with very few exceptions. The average ratio of mitochondrial protein expression was 0.3 + 0.01 for soleus/heart, 0.1 + 0.02 for gracilis/heart, and 4.8 + 0.2 for soleus/gracilis. The gracilis/heart data are plotted as heart/gracilis to condense the ratio axis. The tabulated data for these plots are presented in Supplemental Table S2 (see https://doi.org/10.6084/m9.figshare.14109521.v1). The similarities in mitochondrial protein composition among striated muscles as shown here for the rabbit has also been observed in the pig with nearly identical protein expression profiles as assessed by two-dimensional (2-D) differential in gel electrophoresis (89, 144). Moreover, while endurance exercise training results in an increase in skeletal muscle mitochondrial content, the OxPhos protein content per mitochondrion remains relatively unchanged in both rats and humans (96, 97). In sum, these data are consistent with the notion that, with the possible exception of fuel preference (137, 144), the protein programming per mitochondrial volume is nearly constant across mammalian striated muscle, implying that content and structure are the primary methods of adapting to different maximum OxPhos velocity requirements in different muscle types. Possibly, the protein composition for ATP production is maximized per mitochondrion, and all that is left is modulation of content and location to regulate cellular ATP production capacity.

FIGURE 2.

Mitochondrial protein composition is remarkably similar across striated muscle types. Ratio of mitochondrial protein contents normalized to total protein content between tissues from whole muscle mass spectroscopy data. One-hundred fifty-two of the highest abundant proteins are plotted with labels applied to the proteins with large increases beyond the mean. All proteins are identified along with tabulated data in Supplemental Table S1. A: soleus muscle vs. heart. B: soleus vs. gracilis. C: heart vs. gracilis.

3.1.3. Specific power of energy conversion processes in muscle.

As muscle is a moving organic system, the specific power, or weight efficiency, of the energy conversion process used is important as moving the “weight” of the energy conversion system becomes one of the workloads of the muscle. Thus the higher the specific power of the energy conversion system, the more efficient the muscle can be in performing work. For example, the high specific power of the internal combustion engine, in comparison with the steam engine, made airplane flight possible (145). Using the data in the rabbit heart, we can total the glycolysis enzymes for ATP production to be ∼10 mg/gr wet weight (extrapolated from Ref. 125) and our relational proteomic data (Supplemental Table S1) that can support a realized rate of 5 µmol ATP/min/g heart resulting in 0.5 mmol ATP/min/g glycolytic enzymes. How does this compare with the weight/volume efficiency of OxPhos? The maximum rate of ATP production by OxPhos per gram of mitochondria (including associated water) can be estimated from the cytochrome oxidase (COX) content and maximum activity in intact heart mitochondria of ∼700 nmol O₂/nmol COX/min (88, 146) or assuming 5.6 ATP/O₂ then ∼4 µmol ATP/nmol Cox/min, which correlates with maximum in vivo rates of canines (88). In heart mitochondria, there are ∼1 nmol COX/mg protein (89) resulting in 4 µmol ATP/min/mg mitochondria protein. The mitochondria require an intact compartmentalized structure to function, in contrast to glycolysis; thus, to get the wet weight of mitochondria, the mass of mitochondrial lipid, water, and solutes must be accounted for. Mitochondrial protein makes up 25% of the total mass of mitochondria (147), so 1 mg protein is equivalent to 4 mg of mitochondrial wet weight. This results in 1 µmol ATP/min/mg wet weight or 1 mmol ATP/min/g compared with 0.5 mmol ATP/min/g in glycolysis calculated above. Assuming a free energy of hydrolysis (ΔG_ATP) of –64 kJ/mol of ATP, these values correspond to a specific power (kW/kg) of 1.1 and 0.5 for aerobic respiration and glycolysis in the heart, respectfully. The specific powers of skeletal muscle glycolysis (3.8 kW/kg) and aerobic respiration (1.7 kW/kg) are of similar magnitude (skeletal muscle calculations),¹ although glycolysis appears to have greater specific power than OxPhos in the skeletal muscle. For reference the specific power of a V8 internal combustion engine is ∼0.65 kW/kg or close to the specific power of glycolysis in the heart. The weight efficiency of OxPhos and glycolysis are clearly very close within a factor of 3 taking into account the potential errors in this type of calculation including ignoring the water space associated with glycolysis.

Therefore, why pick localized mitochondrial OxPhos over the more distributed glycolysis or vice versa? The selection of energy source is much more complex than the simple specific power of the enzyme system. The complete oxidation of glucose to CO₂ makes more efficient use of the entry of glucose and does not result in the generation of a high potential metabolic end product, lactate, that needs to be regenerated into glucose in the liver and kidney at a deficit to the net energy balance of the animal. That is, if the muscles only used glycolysis and lactate was converted back to glucose in liver and kidney for subsequent reuse, the body actually losses 4 ATP for each glucose oxidized (148) which is not sustainable. Thus lactate must be excreted, which is a huge loss of potential energy, or used as reducing equivalents in other regions of the body for OxPhos. This later constraint is likely a reason that the heart prefers plasma lactate for oxidation over almost all other substrates when present in high concentration in the blood (149, 150). Another reason may well reside in the structural requirements of glycolysis versus OxPhos. Glycolysis is believed to be partially compartmentalized but is clearly in the freely soluble cytosolic fraction permitting tight packing within the myofilaments. Also, being free in the cytosol reduces the impact of internal shear forces of the muscle that the mitochondria could be susceptible to as complex organelles. As the muscle reduces longitudinal length and thickens upon contraction, shear on the highly compliant mitochondria is potentially generated in both the short and long axis. One compensatory mechanism is to place mitochondria where internal shear is minimized such as the I-band, Z-disk, or subsarcolemmal regions of the cell. Another is to have the mitochondria move with the myofilaments in both dimensions by being directly attached to the myofilaments, thus shortening and thickening with each contraction. Evidence is available that cardiac mitochondria do remain attached to the sarcomere, shortening and thickening with each contraction and retaining their relationship with the sarcomere (151). The connection to the sarcomere may be through the intermediate fibers of desmin together with plectin (152). The impact of this reshaping of the mitochondria on each contraction has not been extensively explored. Overall, the mitochondria require dedicated space that is difficult to locate within the muscle without the sacrificing cross-sectional area of the contractile apparatus needed to maximize power, and specific placement of mitochondria may result in exposure to shear forces as mentioned above. Finally, OxPhos resulting from fatty acid fuel sources, as is most common in the heart (153, 154), also requires displacement of myofibrils or other cellular structures to make room for lipid droplets, which are often tightly coupled to mitochondria. This aspect of the topology of the energy conversion systems will be discussed further in section 3.2.3.

3.2. Energy Distribution Processes of the Muscle Cell

In addition to the content and composition of the energy conversion processes as discussed above, the energetic design of a cell must also take into account the distribution of the inputs, intermediates, and outputs involved in the energy conversion processes (155). Thus, in the relatively large and densely packed muscle cell, the distances between the participating structures as well as the processes regulating the movement of metabolites and ions between them play a particularly important role in determining the energetic support capacity of the cell. To limit the scope of this section of the review, we will focus on the delivery of oxygen from the capillary to mitochondria, the distribution of electrical potential energy within mitochondria, and the delivery of ATP from mitochondria to cellular ATPases. Thorough discussion of substrate metabolism in muscle cells can be found elsewhere (4, 6, 156–158).

3.2.1. Capillary structure and blood flow.

One of the primary requirements to support OxPhos is to provide oxygen. To support the supply of oxygen, the heart is highly vascular with a remarkable density of capillaries between 2,500 and 4,500 capillaries/mm² (159–164), which are nearly evenly spaced every 12 to 22 μm (162, 164–166) consistent with the average diameter of a heart cell (163). These densities suggest that each heart cell is in contact with one or more supporting capillaries. This is consistent with the three-dimensional electron microscopy studies (167, 168) revealing a capillary associated region within most cardiac myocytes. Adjacent to these capillaries is a slightly higher density of mitochondria (159) within the paravascular pool (167), although not as well developed as in the skeletal muscle (169). Thus the capillary network is well designed to deliver blood and associated oxygen and substrates to essentially every cell in the heart wall.

To support the high level of OxPhos, the coronary blood flow is relatively high and linearly balanced with the workload (170) maintaining a low venous po₂ (∼18 Torr) over much of the workload range (171). The mechanism of how coronary flow is precisely matched to oxygen demand remains a major question in cardiac physiology and likely relies on multiple and redundant mechanisms across varying spatial scales for this absolutely critical function in the heart and the body (7, 171–173). The net result is a highly dynamic system providing an appropriate amount of oxygen, maintaining a near constant venous po₂, and meeting the oxygen delivery needs of the heart through a dense feeding capillary network.

The capacity of capillaries to deliver oxygen to skeletal muscles has been of great interest (174–178) since August Krogh’s Nobel Prize winning work (179) on the regulation of oxygen diffusion in muscles roughly a century ago. Capillary density can vary greatly in skeletal muscles from ∼250 to 1,500/mm² of muscle volume with oxidative muscles containing up to twice the number of capillaries per unit volume and ∼33% higher capillary to muscle fiber ratio (∼2.0 vs. 1.5) (175, 177, 178, 180, 181). While capillary-to-fiber ratio and capillary density have been commonly measured on 2-D images for decades and often, but not always, correlate well with muscle oxidative capacity (174–178, 180, 182–184), how the capillary interacts with the muscle fiber also plays a major role in the diffusion capacity between these structures. Indeed, measures that incorporate the geometry of capillary-fiber interactions, either in 2-D or 3-D, have become more prevalent and better correlate with mitochondrial content across both oxidative and glycolytic muscle types across a variety of conditions (175, 176, 185–192). Thus the overall structural capacity for oxygen delivery from capillaries to within the skeletal muscle fiber is dependent on 1) the number of capillaries in proximity to the cell, 2) the direct surface area contact between capillaries and the muscle fiber, and 3) the cross-sectional area of the cell. As mentioned above, more oxidative muscles are smaller, have greater capillary densities, and have greater capillary surface area contact than more glycolytic muscles. However, most muscles, particularly in humans, are comprised of heterogeneous groups of both oxidative and glycolytic muscle fibers with the relative proportion of each determining the overall metabolic phenotype of the whole muscle (182, 193). This heterogeneity results in adjacent oxidative and glycolytic muscle fibers each with different demands for oxygen delivery along their shared borders, and often, oxidative fibers can be surrounded by several glycolytic fibers. The incongruous demand for oxygen to oxidative fibers adjacent to glycolytic fibers may be met by placing capillaries in grooves in the sarcolemma of the more oxidative cell (194). Embedding capillaries atop only one of the two adjacent fibers allows the capillary surface area contact to be increased to one but not the other fiber (FIGURE 3, A–D) and thus offers another level of control when designing the structural capacity for oxygen delivery to the muscle cell within heterogeneous tissues. Much remains to be discovered regarding these capillary grooves such as how and when they form during muscle development as well as more precise information on their contribution to oxygen delivery capacity to muscle cells. It is clear, however, that these grooves in the sarcolemma are true physical structures and not just the result of capillary compression of the cell membrane, as sarcolemmal grooves remain visible in isolated live muscle fibers even after removal of the adjacent capillaries (FIGURE 3C). How or whether the composition of the sarcolemma is altered in this region has yet to be determined, although we speculate that receptors or transporters, which interact with substrates carried in the blood may be upregulated in sarcolemmal grooves relative to the rest of the membrane to increase the efficiency of intercellular signaling and/or transport processes.

FIGURE 3.

Capillary embedding within sarcolemmal grooves provides additional control over oxygen delivery capacity to specific muscle cells. A: schematic figure of a capillary located between a glycolytic (top) and oxidative (bottom) muscle fiber. Mitochondria and nuclei locate laterally to the capillary sitting in a sarcolemmal groove and increasing surface area contact with the oxidative fiber. Adapted from Ref. 194. B: 3-dimensional (3-D) rendering of mitochondrial structure in an intact Tibialis anterior muscle cell showing voids where capillary grooves (C) and nuclei (N) are located. Adapted from Ref. 169. C: 3-D rendering of mitochondrial structure in an isolated, live soleus muscle fiber. Grooves in the sarcolemma remain even after removal of capillaries through collagenase digestion. Adapted from Ref. 169. D: transmission electron microscopy image of a capillary (white arrow; IMF, intermyofibrillar; PV, paravascular) embedded in a sarcolemmal groove in the upper but not lower muscle fiber. RBC, red blood cells. Adapted from Ref. 194.

Similar to the dynamic nature of skeletal muscle contraction, the rate of blood flow can increase more than an order of magnitude from rest to exercise (195) and occurs fast enough that oxygen delivery to the muscle does not appear limiting at the onset of exercise in healthy populations (195–198). Despite the large, rapid increase in oxygen delivery at the onset of exercise, the increased contractile rate of the muscle cells during even moderate exercise results in a fall in the mean po₂ in both the capillaries (∼35 to 24 Torr) and interstitial space between capillaries and muscle fibers (∼16 to 7 Torr) (90, 92). Together with the similar magnitude fall reported for intramuscular po₂ from ∼10–15 Torr at rest to 2–5 Torr at exercise intensities ranging from moderate to severe (91, 199–201), it has been suggested that exercise does not alter the oxygen pressure gradient from the capillary to mitochondria as previously thought (90, 202). Instead, the diffusion constant for oxygen (do₂), analogous to its conductance, appears to increase severalfold to account for the increased oxygen flux (90, 202), although the precise mechanisms behind this change remain to be resolved. What is clear is that blood flow to the skeletal muscle is tightly regulated through several related and likely overlapping mechanisms, which provide precise control over oxygen delivery across the wide range of skeletal muscle contractile demands.

3.2.2. Cytosolic facilitated diffusion processes.

Due to the metabolic rates and distinct compartmentation of the ATP-utilizing and -producing machinery in muscle, as well as the observed metabolic homeostasis limiting large metabolite gradients, it has often been suggested that simple metabolite diffusion is inadequate to support this highly active structure (203). The two major cytosolic facilitated diffusion systems that have been the focus of extraordinary effort are creatine kinase for ATP and ADP (204–208) and myoglobin for oxygen (8, 209–212). These two mechanisms will be discussed below as they represent the classical view of energy distribution in the muscle cell.

3.2.2.1 myoglobin.

In addition to the dense capillary network with only 20 or so micrometers separating each capillary run, the heart also contains myoglobin that can serve as a parallel facilitated diffusion pathway for oxygen (8). Although myoglobin in diving animals is believed to provide an oxygen reserve for diving (213), the low concentration in land animals would only support OxPhos for seconds during a hypoxic event (214, 215). Thus a role as an oxygen reserve in most nondiving mammals is believed to be very small. However, the relatively high concentration of myoglobin relative to oxygen suggested it might play a role in facilitated diffusion. The demonstration of facilitated oxygen diffusion by an oxygen binding protein was first described for hemoglobin and/or myoglobin by several investigators early in the 1960s (209–212) with an early suggestive abstract by Wittenberg (209). This was then developed into a detailed model of myoglobin facilitated diffusion within muscle in a classic paper in Physiological Reviews (8). The concept of myoglobin facilitated diffusion is basically that myoglobin can provide a parallel diffusion pathway for oxygen through the gradients in oxy- and deoxymyoglobin. Myoglobin facilitated diffusion depends on several factors: first, myoglobin affinity for oxygen is intermediate between the relatively low nonlinear affinity of hemoglobin and high-affinity cytochrome oxidase, thus facilitating the transfer from low- to high-affinity sites reflected in hemoglobin/myoglobin and myoglobin/cytochrome oxidase transfers of oxygen. Second, myoglobin is relatively small and is nearly freely diffusible in the cytosol (216, 217). Third, depending on the tissue (218), myoglobin is in sufficient concentration, significantly higher than molecular oxygen, to provide a viable diffusion driven oxygen transfer mechanism. However, based on the concentration and diffusion coefficient measured by two independent physical measures, fluorescence (216, 219) or NMR (220), the contribution of myoglobin diffusion to oxygen transport has been suggested to be <10% even under extreme conditions (203, 219, 220). One of the key aspects of the facilitated diffusion mechanism with myoglobin is that gradients in oxy-and deoxymyoglobin, as well as oxygen, need to exist for this passive facilitated pathway to contribute to oxygen distribution in the muscle cell. We will address this key measure, in vivo, below.

The normal venous po₂ of the heart, depending on the impact of arterial-venous shunting (221), can represent an upper limit for the end capillary po₂, which would put the oxygen tension well above the consensus affinity for cytochrome oxidase under most conditions (7). Indeed, taking the structure of the capillary system and assuming very simple Krogh type diffusion models (222, 223), it has been predicted that within the heart cell, the oxygen tension should greatly exceed the COX P₅₀ for oxygen with a capillary to muscle cell gradient of only 1 or 2 Torr. However, several lines of evidence suggest that the tissue po₂ may be much lower than the capillary geometry, venous po₂, and models would imply. The initial evidence was with small oxygen-sensitive electrodes plunged into the tissue and recording the distribution of oxygen tensions over a large number of insertions (224). These results provided evidence that a significant fraction of the tissue was well below the po₂ of the venous blood. The interpretation of these results ranged from selective barriers to oxygen at the capillaries (222, 225), capillary perfusion heterogeneity (226), and various measures in isolated perfused hearts with questionable extrapolation to in vivo conditions (227). Numerous studies have shown that the saline-perfused mammalian heart with this low-oxygen content but high flow rate, is hypoxic under most experimental conditions using a variety of techniques (228, 229). Since this review is focused on the in vivo condition, the saline perfused mammalian heart will not be considered further here.

Two independent studies in 1999 using noninvasive NMR (230) or optical spectroscopy (223) measures in the heart of the average myoglobin oxygenation, in vivo, detected little or no deoxymyoglobin under normal physiological conditions even approaching maximum workloads. With ischemia or low flow conditions, deoxymyoglobin was easily detected (223, 230). These independent studies supported two important conclusions: 1) that the average oxygen tension in the heart in vivo exceeded the oxygen P₅₀ of myoglobin, consistent with earlier model predictions; and 2) that myoglobin facilitated diffusion of oxygen does not participate in oxygen distribution in heart muscle under normal conditions since no deoxymyoglobin was detected for a driving a diffusion gradient for oxy-myoglobin. These data suggest that the earlier invasive microelectrode studies were not reflecting the true average tissue oxygen tension while the saline perfused hearts were intrinsically oxygen deprived (224, 228, 229). However, under conditions where the capillary po₂ drops, deoxymyoglobin is formed consistent with a potential role for myoglobin facilitated diffusion under these conditions. These observations are consistent with several models that suggested myoglobin facilitated diffusion would only occur under severe anoxic conditions (216, 231). These data also suggest that the heart was designed to supply all the oxygen the mitochondria need to support near maximal work without being oxygen limited.

More recently, genetic knockout experiments have demonstrated that the presence of myoglobin is not necessary for near normal function in the mouse (232, 233). This is particularly surprising in the mouse that operates within a factor of two of its maximum cardiac output while at rest suggesting it is operating near its maximum demand for oxygen nearly all the time (234). The degree of molecular adaptation to myoglobin knockout seems to be controversial with modifications ranging from changes in capillary density (233) and nitric oxide handling (235) to no change in capillary structure and protein expression in heart and skeletal muscle (194). Thus, if any modifications occurred in the myoglobin knockout mouse, they are likely small.

Based on these different lines of evidence, the role of myoglobin in the normal distribution of oxygen in the heart is believed to be minimal. Again, even at maximum workloads, there is little evidence that myoglobin plays a critical role in the delivery of oxygen across the heart (223, 230) even under hypertrophic heart conditions (236). These results coupled with new experiments on other roles of myoglobin have led to other hypotheses for the large investment in myoglobin protein in the mammalian heart (237) and oxidative muscle with the most prevalent being the metabolism of free radicals or nitric oxide (237–239). These new roles for the high concentration of myoglobin in the heart and skeletal muscle are currently an active area of research.

The fact that deoxymyoglobin has not been detected in the working heart in vivo suggests that the vascular-cellular geometry, together with the highly regulated coronary flow results in an oxygen tension in the cell that far exceeds the consensus affinity for oxygen for OxPhos. Thus it is our conclusion that oxygen is not rate limiting for energy conversion under most physiological conditions in the heart. This suggests that the extensive capillary network and regulation of blood flow is capable of keeping the cellular oxygen content well above the P₅₀ of myoglobin and cytochrome oxidase under most conditions using pure oxygen diffusion mechanisms. A specialized system to support molecular oxygen delivery beyond the extensive capillary system does not seem to be necessary in the heart.

The putative oxygen facilitated diffusion system through myoglobin discussed above for the heart is also present to some degree in skeletal muscle fibers (240). Oxidative fibers contain large concentrations of myoglobin, albeit less than in the heart, capable of maintaining a viable diffusion driven oxygen transfer mechanism whereas glycolytic fibers are generally considered to have insufficient myoglobin for facilitated diffusion to contribute significantly (240). Unlike the cardiac tissue, deoxygenated myoglobin can begin to be detected in humans by NMR during exercise at ∼50% of maximal oxygen uptake (moderate exercise) (91, 201) implying oxy/deoxymyoglobin gradients may be present within the skeletal muscle cell. However, increasing the intensity of exercise in these NMR studies all the way up to 100% of maximal oxygen uptake either under normoxia or hypoxia did not lead to further deoxygenation of myoglobin. Although a similar NMR study did find progressive deoxygenation of myoglobin with exercise intensity (200), both groups reported ∼50% myoglobin deoxygenation during exercise at maximal oxygen uptake resulting in a calculated intracellular po₂ of ∼3 Torr similar to the values obtained using biochemical methods (94). Even with the lower intracellular po₂ compared with cardiac cells, the po₂ values in skeletal muscle cells appear to remain above the threshold, which would begin to limit oxygen consumption by mitochondria (P₅₀ <1 Torr) (93, 241–243). Indeed, in vivo spectroscopic measurements resolving myoglobin and hemoglobin individually in mouse skeletal muscle revealed that oxygen consumption was not limited by oxygen availability across the physiological range of intracellular po₂ values (93) supporting the conclusion that oxygen does not play a regulatory role in skeletal muscle respiration under physiological conditions. Together with the lack of effect of myoglobin knockout on exercise capacity in mice (232), these data suggest that myoglobin does not play a significant role in oxygen distribution in skeletal muscle cells under normal conditions. However, further investigations into the myoglobin knockout mice have revealed an important role of myoglobin in nitric oxide signaling (244, 245), which supports the hypothesis that myoglobin may act as a protective shield for mitochondrial respiration by scavenging nitric oxide, a known inhibitor of cytochrome oxidase (246, 247).

In summary, current evidence suggests that the oxygen concentration gradients driving oxygen delivery through the striated muscle cell do not result in oxygen concentrations that become rate limiting for OxPhos under submaximal conditions in the heart or skeletal muscle. It should be stressed that a near maximum workloads, difficult to replicate in the laboratory (54), myoglobin-facilitated diffusion may contribute to oxygen delivery as direct measures of myoglobin oxygenation under conditions of maximum exercise-induced cardiac output, in vivo, have not been attained in contrast to contracting skeletal muscle (91, 200, 201).

3.2.2.2. creatine kinase.

Creatine phosphate (PCr) was one of the first “high energy” phosphate metabolites found in muscle (248, 249), and its breakdown with muscle contraction (248) and ability to support muscle action in the absence of glycolysis (250) supported the early notion that PCr could actually be the high energy molecule supporting muscle contraction. However, it quickly became apparent that PCr had these functional properties due its rapid exchange of high energy phosphate bonds with ADP creating the primary source of muscle contraction, ATP, through the reversible Lohman (251) reaction catalyzed by creatine kinase:

MgADP + PCr + H⁺ MgATP + Cr (Keq = 100: 1 mM Mg2, pH 7.0) (252)

This reaction is believed to be operating near equilibrium in heart and skeletal muscle based primarily on two independent measures: 1) The concentrations of reactants and cofactors (i.e., PCr, ATP, H⁺, Cr, and Mg²⁺) appear to be poised near the creatine kinase equilibrium (44, 253, 254); and 2) the unidirectional flux through this “dead end” (i.e., Cr and PCr are not rapidly used for any other reaction) measured using ³¹P-magnetization transfer methods, far exceeds the calculated ATP turnover rate (255–258). Thus ATP, ADP, Cr, and PCr are believed to be at near equilibrium concentrations in mammalian muscles under most experimental conditions. Additionally, creatine kinase is one of the most abundant proteins in the cytosol, although its concentration varies according to muscle fiber type (TABLE 3). With the high concentrations of creatine kinase, Cr, and PCr coupled with this rapid exchange process, it quickly became clear that the PCr pool could serve as a short term spatial and temporal “buffer” for ATP when ATP production could not keep up with ATP utilization (260). With [PCr] only exceeding [ATP] by twofold in heart (261) and fourfold in muscle (262), this buffer capacity is quite small considering that the turnover of ATP at high workload is only a few seconds in heart (263). Thus, for short temporal mis-matches in ATP production and utilization, the rapid creatine kinase reaction could maintain ATP (264) even with the net production of Pi reducing the free energy of ATP hydrolysis (265, 266).

Table 3.

Concentrations of creatine kinase in striated rabbit muscle. Values are µmol/L cell volume

Protein	Soleus,* µmol/L cell volume	Gracilis,† µmol/L cell volume	Heart,† µmol/L cell volume
Creatine kinase	10.4	20.6	7.0

*Rabbit soleus value from Illg and Pette (259). †Rabbit heart and gracilis calculated from relational proteome database using the abundance ratios of the sum of cytosolic creatine kinase isoforms.

Once the buffering power of PCr was realized to be minimal, another more sustained impact of creatine kinase on maintaining the spatial distribution of free energy in ATP was proposed by Bessman and coworkers in a series of papers (204–208) describing the creatine kinase shuttle where the topology of the creatine kinase enzyme at the mitochondria and the myofibrils facilitated the conversion of ATP + Cr to PCr + ADP at the mitochondria and PCr + ADP to ATP + Cr at the myofibrils. Due to the aforementioned higher concentrations of PCr when compared with ATP, and especially Cr when compared with ADP, this resulted in a potential for facilitated diffusion through the cytosol (267). That is, PCr facilitated the diffusion of the high energy phosphate bond from ATP created at the mitochondria to the myofibrils while the diffusion of the low energy substrate ADP, in micromolar concentrations, was greatly enhanced by the mM concentration of Cr. This proposal suggested that the creatine kinase reaction is not just a short-term buffer for high-energy metabolites but also plays a role in the distribution of the potential energy and reaction intermediates required for energy conversion in the muscle cell. The creatine kinase shuttle has been a major focus in muscle energetics since the Bessman proposal and has been the topic of numerous reviews (204, 263, 267–273), although many have questioned the extent of cytosolic energy distribution dependent on this shuttle, especially in the heart, for various quantitative and experimental considerations (203, 267, 271, 274–277).

Numerous investigators have repeated the classical work of Jacobus and Lehninger (278) demonstrating that isolated muscle mitochondria will preferentially generate PCr when given ADP, Pi, and creatine at reasonable physiological concentrations (279). This is a fundamental element of the creatine kinase shuttle that ATP is converted to PCr at the mitochondria and both the ATP and PCr are available for diffusion through the cytosol in a parallel path. There is a muscle creatine kinase isoform in the mitochondrial intermembrane space to perform this task (280, 281). At the myofibrils, the creatine kinase converts the ADP generated by the contractile and ion transport ATPase to Cr. Due to the equilibrium of the creatine kinase reaction, most of the ADP will be rapidly transferred into Cr providing a potentially large gradient in Cr from the ATPase sites to the mitochondria to be used in the mitochondrial creatine kinase reaction. Although this is an attractive theory, and simply based on facilitated diffusion arguments (267) should supplement the distribution of energy across the muscle cell, several observations bring the net importance of this process into question. In the heart, over large variations in workload in vivo, there were little or no changes in PCr concentration ([PCr]) or by mass balance arguments Cr concentration ([Cr]), until near maximum cardiac workloads were attained (50, 282–288). No changes were also observed transiently with work jumps (282). Diffusion, or a metabolite shuttle, relies on changes in the concentration gradients of the metabolites to change the rate of flux. Since [Cr] and [PCr] do not change significantly over physiological workload in heart, in vivo, this suggests that alterations in these facilitated diffusion pathways are not playing a dynamic role in matching the rate of ATP production with ATP utilization as first predicted by Hill (34). A more likely hypothesis is that the energy distribution system of the cell greatly exceeds the energy requirements throughout the cell below V_max without requiring significant changes in steady state concentrations of any metabolite. However, as maximum workloads are approached, numerous systems become limiting from blood flow to energy conversion (9–11).

Another line of in vivo evidence against a baseline role for the creatine kinase shuttle mechanism of energy distribution has been the evaluation of different types of creatine kinase (289–295) or creatine (274) knockouts in mice. The mouse provides an excellent model for evaluating energy distribution in the heart due to its remarkably high baseline metabolic rate (89, 234) setting energy demands near maximum even at rest. Creatine kinase knockout studies have revealed very modest changes in the performance or protein programming of these animals suggesting a potentially secondary role of creatine kinase shuttle in supporting energy distribution in the mammalian striated muscle system. This conclusion is strongly supported by the recent review from the Dr. Weiss laboratory that concluded: “…….³¹P MRS studies in CK-deficient hearts suggest that CK does not play an obligate intracellular transport role in the normal mouse heart and that its presence is not critical for baseline contractile function in normal hearts.” (275). We agree with this conclusion based on theoretical concerns and the experimental results including biophysical measures and genetic knockout experiments.

3.2.3. An alternative to metabolite facilitated diffusion: the muscle mitochondrial reticulum.

As discussed above, the concept that facilitated diffusion of metabolites or oxygen supporting normal muscle function has been challenged over the last several decades based on theoretical ground, genetic modifications and the observation of a metabolic homeostasis of the metabolites involved. Another possible mechanism to rapidly distribute energy across the cell from OxPhos is the mitochondrial reticulum with its high conductance only requiring small gradients in potential energy throughout the cell (169, 296). Thus the actual structure of the mitochondrial network may provide an additional mechanism for intracellular distribution of potential energy within the muscle cell. The following sections will address our current understanding of muscle mitochondrial network structure and several remaining questions surrounding how the mitochondrial reticulum may contribute to and regulate cellular energy distribution in striated muscle.

3.2.3.1. initial observations of muscle mitochondrial structures.

Mitochondria in striated muscle were first visualized roughly 180 years ago by Schwann (297) and Henle (298) who each described them as “granules” located between the myofibrils as well as near the nuclei (FIGURE 4A) of skeletal muscle cells. The description of these interstitial granules was extended by Kolliker (305, 306) who found that they varied across different species and muscle types in addition to demonstrating their capacity for swelling under different osmotic conditions and speculating a relationship with metabolism. Knoll (307) provided perhaps the most detailed descriptions of interstitial granules within cardiac muscle cells around this time, demonstrating the alignment of many granules between the myofibrils in a variety of species. In 1890, the terms “bioblast” and “sarcosome” were coined by Altmann (308) and Retzius (309), respectively, to describe what are now known as mitochondria in the muscle, with Altmann considering the bioblast to have metabolic and genetic autonomy within the cell. The term mitochondria was coined by Benda in 1898 (310) to describe both the thread-like and granular nature he observed for these organelles, and soon after mitochondria, bioblasts, and sarcosomes were all confirmed to be the same as the granules initially described by Schwann (311, 312). However, the term sarcosome persisted into the 1960s (313–315) and was sometimes described as a separate structure from mitochondria (316). It is also interesting to note that many investigators in this era believed that mitochondria could transform directly into myofibrils, although this was largely attributable to nonspecificity of the staining techniques used at the time (317). Overall, the early studies up through 1920 found that mitochondria in striated muscles were generally located around the nuclei, between the myofibrils along the anisotropic (A)-bands, and/or at the end of each sarcomere in the isotropic (I)-band region of the muscle fiber with A-band mitochondria being the predominant type in oxidative and cardiac muscle and I-band mitochondria more abundant in glycolytic muscle types.

FIGURE 4.

A visual history of muscle mitochondrial reticulum structure. A: interstitial granules within a muscle cell imaged through a light microscope in 1839 (297). B: cross-sectional 2-dimensional image of the mitochondrial reticulum in rat diaphragm oxidative muscle imaged by transmission electron microscopy (TEM) in 1966 (322). C: 3-dimensional (3-D) rendering of the rat diaphragm oxidative muscle mitochondrial reticulum imaged by serial section TEM in 1978 (299). D: scanning electron microscopy (SEM) image of the rat hindlimb oxidative muscle mitochondrial reticulum from 1985 (300). CN, thin mitochondrial column; IM, I-band limited mitochondria; Z, Z-disk. E: high-voltage TEM projection image of rat hindlimb muscle mitochondrial reticulum in 1986 (301). F: 3-D rendering of horse muscle mitochondrial reticulum imaged by serial section TEM in 1988 (302). G: 3-D rendering of the oxidative mouse muscle mitochondrial reticulum imaged by focused ion beam (FIB)-SEM in 2015 (169). H: 3-D renderings of the mitochondrial reticulum in glycolytic (left), oxidative (middle), and cardiac (right) mouse muscles imaged by FIB-SEM in 2018 (303). IFM, intrafibillar mitochondria; PVM, paravascular mitrochondria. I: 3-D rendering of the human oxidative muscle mitochondrial reticulum imaged by FIB-SEM in 2019 (304).

3.2.3.2. origins of the muscle mitochondrial reticulum.

Though little attention was paid to muscle mitochondrial structure in the 1920s, 30s, and 40s, the advent of histochemical staining (318, 319) and electron microscopy (320) together with new methods to isolate mitochondria for functional testing (315, 321) brought muscle mitochondria back into focus in the 1950s. Andersson-Cedergren (322) provided the first look at muscle mitochondrial structure in 3-D in 1959 by using serial section electron microscopy. However, she focused primarily on the motor end plate region of a mouse intercostal skeletal muscle where mitochondria appeared as individual, globular shapes. The initial notion of a mitochondrial reticulum, or network, in skeletal muscle was reported by Bubenzer in 1966 (323) where he showed using 2-D transmission electron microscopy (TEM) that the thin (likely oxidative) fibers of the rat diaphragm contained mitochondrial grids around the sarcomeric I-bands (FIGURE 4B), which were connected through longitudinal mitochondrial tubules along the A-band to form what he proposed was a 3-D mitochondrial framework. Furthermore, Bubenzer showed a fiber type dependence of the mitochondrial network structure as the thicker (likely more glycolytic) fibers of the diaphragm contained smaller grids of mitochondria around the I-bands with fewer connecting mitochondrial tubules along the A-bands. It was around this time that Skulachev first proposed the hypothesis that energy could be transported throughout the cell along mitochondrial membranes in the form of the mitochondrial membrane potential (324). The idea of mitochondria as electric power transmitting cables was suggested to be of particular importance in large cells with “giant” mitochondria, which could cover distances on the scale of cell sizes. In mammalian systems, the muscle fiber was considered likely to contain the proposed electrically coupled mitochondrial energy distribution system due to the large diameter of muscle fibers, the very high-energy demands to be met, and the presence of what appeared in 2-D images to be highly connected networks linking the round mitochondria near the sarcolemma to the more grid-like mitochondrial profiles in the intrafibrillar space (325).

To confirm that mitochondria formed three dimensional networks in rat diaphragm muscle as suggested by the 2-D images of Bubenzer, Skulachev’s group (299) stitched serial section electron microscopy images together in 1978 to provide the first 3-D look at intrafibrillar muscle mitochondrial structure. Indeed, in this seminal work, they confirmed Bubenzer’s suggestions of highly connected mitochondrial networks in the rat thin diaphragm fibers and also showed that these networks could span across the entire cross-sectional diameter of the cell. Moreover, they found that the mitochondrial reticulum of the rat diaphragm contained many intermitochondrial junctions between the inner and outer mitochondrial membranes of adjacent mitochondrial profiles and suggested that electrical conductivity may propagate across these structures analogous to gap junctions. It was further suggested that these junctions may act as regulatory sites of conductivity within the network and may prevent spreading of de-energization throughout the network in response to localized damage. However, the tools to test these many of the newly generated hypotheses were lacking at the time in addition to scarce technology available to visualize the 3-D mitochondrial structures. In fact, all mitochondrial structures were transposed on to dental wax models and photographed to make the 3-D figures for that paper (FIGURE 4C) which, in several cases, made it difficult to fully appreciate the connectivity and complex shapes within the mitochondrial reticulum. Soon thereafter, Skulachev’s group (326) used a similar serial section approach to report that all of the mitochondria within cardiomyocytes were connected through intermitochondrial junctions into a singular mitochondrial reticulum again suggesting that this would allow for cell-wide transmission of electrical potential energy. Furthermore, they demonstrated that mitochondria often appeared physically coupled across the cell membranes at gap junction sites between cardiomyocytes further speculating that the putative intracellular power transmission system could extend across cells through these regions. Many of Skulachev’s original hypotheses regarding the distribution of energy through mitochondrial networks would eventually be tested further once the appropriate technology was developed and will be discussed further below.

Several years later, Ogata and Yamasaki (300) took a different approach to visualize muscle mitochondria in 3-D by using scanning electron microscopy (SEM) on samples from rat hindlimb red, white, and intermediate muscle fibers where the contractile apparatus had been chemically dissolved away after fixation (FIGURE 4D). While this SEM approach did not allow for quantification of 3-D mitochondrial structures, the striking images clearly showed the fiber type dependency of mitochondrial network connectivity in mammalian hindlimb muscles. Mitochondrial tubules in the intrafibrillar regions were connected across both the parallel and perpendicular axes in the red (oxidative) muscle and were limited to primarily the perpendicular axis in the white (glycolytic) muscle whereas in the intermediate muscle characteristics fell in between as the name would imply. The specific distribution and content of mitochondria in the striated muscle cell was also recognized as a potential source of different muscle disorders, specifically in ragged red fiber disease (327). The fiber type dependency of mitochondrial reticulum structure was additionally supported using high-voltage electron microscopy to create projection images of thick (∼0.5 µm) muscle sections (328, 329) allowing for quantitative analyses of surface area and volume in these pseudo-3-D images (301, 330) (FIGURE 4E). Using this technique, the Brooks laboratory (301, 330) also demonstrated that rat hindlimb muscle mitochondrial networks were responsive to environmental conditions as elevated metabolic activity through endurance exercise training resulted in a large increase in the size of the mitochondrial reticulum. Conversely, Kayar et al. (302) in the Weibel laboratory used serial section electron microscopy in 1988 (FIGURE 4F) to show that while the subsarcolemmal and intrafibrillar mitochondria were directly connected in horse hindlimb muscle, they did not find evidence of a highly connected mitochondrial reticulum throughout the entire cell. These authors pointed out that caution should be taken in modeling oxygen utilization given this structure and “……we need more information, particularly including a quantitative estimate of mitochondrial connectivity, before it is possible to construct a comprehensive model for cellular level aerobic respiration.” We completely agree with these concerns raised in 1988. This latter study, as well as the dearth of available 3-D visualization tools and the prevalence of 2-D images of circular mitochondrial cross sections all likely contributed to the lack of wide acceptance of a highly physically connected muscle mitochondrial reticulum around this time.

3.2.3.3. recent investigations into the muscle mitochondrial reticulum.

There were relatively few studies examining muscle mitochondrial network structure during the 1990s and the early 2000s (331). However, the emergence of mitochondria as sometimes dynamic organelles (332, 333), the adaptation of fluorescent proteins to tag and track specific cellular molecules (334, 335), and the commercial availability of confocal and multiphoton light microscopes all converged to bring a new wave of interest into mitochondrial structure in all cell types, including muscle (336–339). In 2010, both mitochondrial fission (division) (340) and mitochondrial fusion (341) were shown to regulate individual mitochondrial morphology as well contribute to maintaining normal skeletal muscle contractile function using transgenic mouse approaches. Subsequently, fission and fusion proteins were also found to be essential for maintaining mitochondrial morphology, respiratory capacity, and contractile function in the heart (342, 343). Yet, it still remained unclear exactly how dynamic mitochondria actually were in striated muscles due to the physical constraints set in place by the surrounding contractile apparatus. Application of transgenic photoactivatable fluorophores to mouse muscle mitochondria by several groups soon followed (344–349) and allowed for demonstration of matrix content exchange by dynamic mitochondria in live cardiac and skeletal muscle fibers. However, the rates of mitochondrial fission and fusion reported in striated muscle in vivo were orders of magnitude slower than previously shown in other cell types in culture often requiring up to 30 min to detect any dynamic changes in mitochondrial morphology (345, 346). For example, the rate of mitochondrial fusion in adult rat muscle fibers was shown to be more than 10-fold slower than in rat myotubes while the rate of mitochondrial fission in the adult fibers was an additional 3-fold slower than the rate of fusion (344). Indeed, the expression levels of mitochondrial dynamics proteins involved in fission, fusion, and motility are all downregulated in mature muscles relative to cultured myotubes as well as neonatal muscle in the mouse (133, 135). Overall, the studies of mitochondrial dynamics in healthy, mature striated muscle to date reveal a structurally stable network with the impact of mitochondrial fission and fusion on muscle function likely due to the importance of mitochondrial dynamics as a chronic or developmental process, rather than acute, adaptation toward maintaining mitochondrial quality control in muscle (350–355). However, dysfunctional mitochondrial dynamics processes may play a critical role in pathological states such as with mtDNA deletions, mitochondrial myopathies, and aging among others (346, 356–358).

The morphological differences between mitochondria interspersed between the myofibrils and mitochondria in the subsarcolemmal regions near capillaries and nuclei have been discussed since the earliest studies on muscle mitochondrial “granules” (297, 298). However, there has recently been renewed focus on the connectivity of mitochondria between these two regions of a muscle fiber. Picard and colleagues (359) used SEM to show qualitative differences between subsarcolemmal and intermyofibrillar mitochondrial structures as well as performed a rigorous quantitative analysis of 2-D TEM images reiterating the physical connectivity between the two mitochondrial populations as described by earlier studies (301, 323). This group also reiterated the need to separately analyze the vastly different mitochondrial structural appearances between longitudinal and transversely cut 2-D sections and showed that interactions between mitochondria increase after a 3-h exercise bout without a change in expression of fusion proteins (360). The following year, the Prats group (361) used light microscopy image deconvolution to demonstrate the fiber type dependency of the 3-D human muscle mitochondrial network in agreement with a previous SEM study (331). Furthermore, they used focused ion beam-scanning electron microscopy (FIB-SEM) to visualize the direct continuity of individual mitochondria from the subsarcolemmal to the intermyofibrillar space. Thus human muscle mitochondria (361) had similar morphological characteristics and connectivity to previously published rodent muscle data using scanning block face scanning electron microscopy, although the mitochondrial data was not discussed in the earlier study which focused on the neuromuscular junction (see Video S1 from Ref. 362).

Around this same time, we found that the large pools of many mitochondria located near the cell periphery were always associated with capillaries that embedded into persistent grooves (FIGURE 3) in the sarcolemma of the mouse muscle fiber by using 3-D multiphoton microscopy on live mice (194). Accordingly, we chose to describe these large pools as paravascular mitochondria rather than subsarcolemmal mitochondria due to the more descriptive nature of the term as well as the fact that there are many mitochondria that are located just under the sarcolemma but not within a large pool of mitochondria and instead are in direct contact with the myofibrils appearing to be structurally the same as intermyofibrillar mitochondria. The large paravascular pools of mitochondria appeared to taper and get smaller in diameter when moving away from the capillaries and often stretched as far as 50 µm from the nearest vessel (194). Thus there was a significant fraction of mitochondria located near the cell periphery that were as far away from a capillary as the mitochondria in the middle of the interior of the muscle fiber arguing against the notion that mitochondria are placed near the sarcolemma primarily for oxygen diffusion reasons. As a result, we hypothesized that mitochondria were functionally coupled throughout these large pools with the energetic advantage of being near the capillary shared across the entire pool through conduction of the mitochondrial membrane potential similar to Skulachev’s proposal nearly 45 yr earlier (324). Indeed, we followed a similar path to Skulachev by next performing 3-D electron microscopy studies to assess muscle mitochondrial connectivity but with the benefit of several decades of technological advances. Structurally, we also found a highly connected mitochondrial reticulum comprised of many mitochondria aligned both parallel to the muscle fiber between the myofibrils and perpendicular to the muscle fiber surrounding the I-bands (169) (FIGURE 4G). Moreover, roughly one-fifth of the paravascular mitochondria were directly connected to the interior of the cell through projections into the intermyofibrillar space, and nearly all mitochondria in both regions of the cell were connected to adjacent mitochondria through intermitochondrial junctions. While this work confirmed the structural connectivity of the muscle mitochondrial reticulum, it remained unknown whether physically connected skeletal muscle mitochondria were indeed functionally connected. Skulachev’s group (363) had reported in 1988 that damaging a single mitochondrion in cultured neonatal cardiomyocytes could lead to depolarization of the membrane potential of adjacent mitochondria within a physically connected cluster; however, it was unclear how these in vitro results would translate to mature muscle fibers within an animal.

To test the functional connectivity of the mitochondrial reticulum in adult muscle fibers, we utilized a photoactivatable, mitochondrial specific depolarizing agent, MitoPhotoDNP, which had been developed a few years earlier (364). MitoPhotoDNP is named after its three constitutive parts: 1) a lipophilic cation, triphenylphosphonium, which ensures the probe concentrates in the cell according to voltage and, thus, into the mitochondria; 2) a mitochondrial depolarizing agent, 2-4-dinitrophenol (DNP), which is inactive when linked to MitoPhotoDNP; and 3) a nitrobenzyl photocleavable linker, which cages the active DNP until exposed to ultraviolet (UV) light. Combining MitoPhotoDNP with a fluorescent mitochondrial membrane potential marker, tetramethyl rhodamine methyl ester (TMRM), allowed for spatially controlled depolarization and simultaneous assessment of the mitochondrial membrane potential in freshly isolated adult mouse soleus fibers (169). Exposing only the interior of the muscle fiber to low level UV light resulted in depolarization of not only the mitochondria in the interior but also uniform depolarization of the mitochondria near the cell periphery. The depolarization of the mitochondrial membrane potential in the UV irradiated and adjacent regions occurred as fast as we could image the muscle fibers (∼300 ms) implying that mitochondria within the two regions were functionally connected through electrical conduction. These results (169) suggested that the physically connected mitochondrial reticulum in skeletal muscle was capable of rapid distribution of electrical energy through conduction of the mitochondrial membrane potential just as Skulachev (324) had hypothesized five decades earlier. A follow-up study using a similar approach in cardiomyocytes yielded comparable results showing that mitochondria in the paravascular, paranuclear, and intrafibrillar regions were all connected through intermitochondrial junctions and that these networks also appeared capable of rapid distribution of the mitochondrial membrane potential (167).

Electrical connectivity of mitochondrial networks offers significant advantages for cellular energy distribution compared with simple or facilitated metabolite diffusion, particularly regarding speed (365). However, just as with man-made networks like the electrical power grid or internet, connectivity also creates vulnerability [e.g., lightning strikes, hackers, viruses, etc.), Skulachev proposed in 2004 (333)] that a mitochondrial thread-grain transition, i.e., mitochondrial fission, provided a mechanism to reduce the spread of dysfunction throughout a connected mitochondrial reticulum. However, the mitochondrial structural transitions that were shown in cell culture in that study took 45 min to 8 h to complete. For cells with high-energy turnover such as skeletal and cardiac muscle, significant dysfunction for that length of time could be devastating. We investigated the protection mechanisms within the mitochondrial reticulum of heart and skeletal muscle cell by using MitoPhotoDNP to create dysfunction (depolarization) in the center of the cell, but instead of looking at the immediate (within 1 s) response to assess electrical connectivity as we had previously (169), we focused on the secondary events occurring in response to persistent dysfunction (167). Within 5–10 s after the initial shared depolarization of the membrane potential between the irradiated and adjacent regions of the mitochondrial reticulum, the two regions became electrically distinct. The irradiated region continued toward full depolarization while the adjacent region of the mitochondrial reticulum quickly recovered its membrane potential to baseline levels. Thus, in addition to suggesting that the initial shared depolarization was not simply the result of diffusion of DNP or TMRM, these results demonstrated that the muscle mitochondrial reticulum has a reactive protection mechanism that is similar to a circuit breaker or fuse in an electrical circuit (FIGURE 5, A–F). After sharing the brunt of the initial dysfunction throughout the network (FIGURE 4E), the damaged region is quickly separated allowing the undamaged portion to return to supporting cellular energy demands (FIGURE 5F). The rapid nature of the electrical separation mechanism may explain why localized microflow of mitochondrial uncoupler, FCCP, did not previously reveal connectivity of mitochondrial networks when imaged at 5- to 7-s intervals (366). Despite occurring much faster than the thread-grain transition shown by earlier by Skulachev (333), we also hypothesized that the physical mechanism for the rapid electrical separation of the mitochondrial network occurred through mitochondrial fission. Because depolarization resulted in the loss of TMRM signal, we were unable to observe the physical response of mitochondria in the damaged region in our original experiments. Thus we combined the use of MitoPhotoDNP with the transgenic photoswitchable mitochondrial fluorophore MitoDendra2 (347), to perform an experiment where photoactivation not only depolarizes mitochondria but also changes their fluorescence from green to red allowing separate structural tracking of the depolarized and nondepolarized mitochondria. Depolarization of mitochondria in the interior of the cell resulted in a physical separation of damaged mitochondria from the healthy portion of the network and a morphological change from elongated to more spherical mitochondria starting after ∼30 s (167) (FIGURE 5, B and C). However, the use of fission inhibitors, mdivi-1 (367) and dynasore (368), did not prevent the physical separation and condensation of damaged mitochondria. Closer inspection of the remodeling of the damaged mitochondria revealed that there were actually fewer and larger mitochondria upon completion of the restructuring, which is the opposite of mitochondrial fission. Due to the 2-D nature of the images, we were unable to determine whether mitochondria were merging (fusion) or simply condensation or retraction of irregular, elongated structures was occurring. Since mitochondria do not have significant levels of structural proteins or macromolecules, the shape of mitochondria is likely dictated by the cell cytoskeleton molding the mitochondria into the complex shapes observed (167). We speculated that these “tether points” could be released on damaged mitochondria resulting in the formation of their intrinsic spherical shape and also removing them from the reticulum network (167).

FIGURE 5.

Muscle mitochondrial network protection system. A: schematic of the putative electrical and physical disconnection processes upon localized damage to the muscle mitochondrial reticulum. B and C: 3-dimensional renderings of the connected (green) and disconnected (red) regions of the muscle mitochondrial reticulum upon localized depolarization in a MitoDendra2 mouse muscle fiber and cardiomyocyte, respectively. D: raw membrane potential voltage image from a mouse soleus muscle fiber before depolarization. E: post/preratiometric image of mitochondrial membrane potential voltage immediately after localized depolarization in the middle of cell from D. Dark signal indicates decreased voltage. F: post/preratiometric image of mitochondrial membrane potential voltage 55 s after localized uncoupling showing continued depolarization in the irradiated region and repolarization back to baseline in the adjacent regions. All images adapted from Ref. 167).

Subsequently, a fission-independent mitochondrial shape transition has been described in other cell types (369). This calcium- and Miro1-dependent mitochondrial shape transition is likely similar to that observed upon regional depolarization within the muscle mitochondrial reticulum, and thus, the Miro-Trak-kinesin complex, which can link to microtubules may provide a mechanistic basis for mitochondrial retraction. Other proteins that have been implicated in the association of mitochondria with the myofibril structures include desmin (intermediate fibers) and plectin (370–372), but the dynamic regulation of this process is unknown. Accordingly, the response to localized damage within the mitochondrial reticulum occurs in two sequential steps. Because mitochondrial fission does not seem to be the physical mechanism for either step, it appears likely that modulation of the intermitochondrial junctions is occurring where first, the junction is functionally closed, and second, the adjacent mitochondria physically separate. The ability to quickly open and close intermitochondrial junctions would provide protection against transient local dysfunction without the cost of physically remodeling the network. Indeed, transient depolarization and repolarization of individual or groups of mitochondria in muscle cells have been widely shown (373–375). Moreover, O’Rourke and others (376–381) have reported that synchronized, cell-wide oscillations in membrane potential and other energetic intermediates can occur in cardiomyocytes along with regions of heterogeneous membrane potential. These perturbations in membrane potential have been found to increase in likelihood during periods of oxidative stress such as hypoxia (381), nutrient deprivation (380), or diabetes (370). This propensity of mitochondrial networks to depolarize in a synchronized manner has been termed mitochondrial criticality and is often tested by monitoring the response of adjacent mitochondria to localized laser-induced depolarization. However, despite the similar design to the MitoPhotoDNP experiments described above, the cell-wide oscillations in membrane potential in these experiments generally occur after a delay and in a wave-like manner attributed to diffusion of reactive oxygen species or calcium (370, 376–380). While it is possible that exposure of cells to MitoPhotoDNP increases the criticality so much that the delay in depolarization is no longer measurable at the speed of confocal microscopy, it may be more likely that the at least order of magnitude faster response to localized depolarization by MitoPhotoDNP may be assessing a different phenomenon than that observed after photodamaging mitochondria. Implementation of faster imaging strategies such as structured illumination microscopy (382) for future MitoPhotoDNP experiments may resolve some of these issues. Notwithstanding the technical differences in methodology, both approaches demonstrate that clusters of many adjacent mitochondria can respond synchronously to localized perturbations and that the connectivity among functionally coupled mitochondria can be dynamic in nature dependent on cellular conditions (167, 379). However, further investigation is warranted to elucidate the precise molecular and physical mechanisms by which acute mitochondrial network connectivity is regulated in striated muscles.

Recently, development of improved analytical platforms has allowed for more rigorous quantitative comparisons of high resolution, 3-D muscle mitochondrial structures among different cell types in both humans and mice (303, 304, 383) as Weibel had called for 30 yr earlier (302). Bleck et al. (303) adapted the connectomics approach used for assessing neuronal connections (384) to develop an analytical framework geared toward high-throughput assessment of the 3-D interactions among mitochondria and other organelles within the cell. This subcellular connectomics approach was validated by assessing mitochondria in adult mouse glycolytic, oxidative, and cardiac muscle fibers (FIGURE 4H) and showed that mitochondrial structure appears to be tailored to support cellular goals at the network, individual mitochondrial, and mitochondrial interaction levels. Furthermore, quantitative assessment of organelle interactions suggested mitochondria can be specialized for different purposes even within the same network as mitochondria connected to lipid droplets were structured to better facilitate energy distribution throughout the cell whereas nonlipid droplet mitochondria had higher surface area-to-volume ratios and more interactions with the SR suggesting an increased capacity for calcium cycling. Picard’s group (383) assessed muscle mitochondrial structures from human patients and found that people with mitochondrial DNA diseases exhibit morphological changes in mitochondria compared with controls. They also reported that human muscle mitochondrial structures are largely different than those found in rodent muscles as the human muscle were not very connected and had many thin mitochondrial nanotunnels. However, they did not control for fiber type differences, which greatly affect mitochondrial network and individual mitochondrial structures (303), and stress associated with the biopsy procedure (385, 386) could explain the prevalence of nanotunnels (387). Indeed, the Subramaniam laboratory, in collaboration with Ferrucci’s group, recently reported human muscle mitochondrial structures (388) (FIGURE 4I), which were very similar to the mouse muscle mitochondrial structures previously visualized by the same FIB-SEM system (169) as well as the rat diaphragm structures from 40 yr earlier (299). In the study by Caffrey et al. (388), the muscles were fixed immediately after a carefully vetted biopsy procedure to ensure maintenance of normal muscle structure. In fact, mitochondrial nanotunnels, a marker of cell stress (387), were seldom observed and two types of mitochondrial networks were shown, which appear to represent oxidative and glycolytic muscle types. Thus muscle mitochondrial networks in humans and rodents appear remarkably similar in well-preserved specimens when controlled for fiber type differences. It is possible, however, that muscles with highly specialized functional demands due to anatomical location (e.g., extraocular) or in specific mammalian species may arrive at slightly different optimal mitochondrial configurations, although this remains to be tested.

It is also important to note that the classical use of isolated mitochondria from striated muscle generally creates 0.8- to 1-μm spheres of mitochondria that do not correlate with the complex structures observed in 3-D living cell superresolution optical studies or fixed electron microscopy outlined above. This again is consistent with the lack of structural proteins in the mitochondria, that they revert to simple spheres consistent with the hydrophobic effect, and the notion that the complex mitochondrial shapes in the intact cell are the result of interactions with the cellular cytoskeleton. The small, spherical isolated mitochondria imply that what is being studied in isolated mitochondria are fragments of the large structures. Thus they were likely once open to the isolation medium and then resealed to generated the “tight” membrane structures. However, it is likely that the fidelity of isolated mitochondria to the intact mitochondria is low and extrapolations to the complex intact system are difficult. As the composition of mitochondrial regions within the muscle cell may also be different, concerns arise with regard to the selection of regions by the isolation process. Since mitochondrial isolation always yields a just a fraction of total mitochondria, it begs the question as to whether all of the cytoskeleton-linked regions are retained similarly by the “bulk” tissue isolation procedures. In other words, do different regions contribute differentially to the isolation? While isolated mitochondria may be useful for first principles of function, their relevance to the function and control of mitochondria in the intact muscle cell is limited.

3.2.3.4. comparison of metabolite diffusion and mitochondrial reticulum conduction.

Can the capacity of the mitochondrial reticulum to electrically distribute potential energy be estimated relative to metabolite diffusion? TABLE 4 presents estimates of the driving forces for diffusion or mitochondrial conduction for a 10% gradient of metabolites or voltage. The 10% value was selected as this is likely the measurement threshold for biochemical assays in intact cells. The ATP flux associated with these 10% gradients was calculated assuming 1 ATP per ATP, ADP, Pi, PCr, or Cr arriving from metabolite diffusion while the equivalent charge associated with 4H⁺ was required to generate 1 ATP. Excellent data on the diffusion coefficients for most of the metabolites involved in energy metabolism is readily available, and the 3-D structure of the mitochondrial and cytosolic volume permits normalization of the space available for diffusion or mitochondrial conduction. Data on passive conductance of membrane structures on the scale of mitochondria are provided from early neuroscience studies. The transfer accessible cross section per cell was estimated by measuring the mitochondrial volume for mitochondrial cross section while all of the nonmitochondrial volume was assumed available for metabolite diffusion.

Table 4.

The driving forces for diffusion and conduction within striated muscle cells

Metabolite	Concentration,a mmol/L Cell Water)	Effective Diffusion Coefficient,c µm2/s (37°C)	Transfer Accessible Cross Section, per cell µm2	10% Driving Force Flux, ATP mM/s	Estimated Maximum Oxidative Capacity, ATP mM/s
Heart
ATP	9	510 (389)	380f	1.21	16.6
PCr	18	640 (389)	274	4.20
ADP	0.010	510	380	.0013
Cr	5	640d	274	1.17
Pi	0.7	860e	380	0.16
Mito Ret (K+Cl–)b	280 mM 180 mV	2,470b [93 µ2/s/mV]b	106	652 [1,105]j,k
Slow-twitch muscle
ATP	5 (262)	510	706g	0.36	1.3
PCr	16	640	600	1.70
ADP	0.011	510	706	0.0008
Cr	7	640	600	0.75
Pi	6	860	706	0.73
Mito Ret	280 mM 180 mV	2,470 [93 µ2/s/mV]b	106	652 [1,105]
Fast-twitch Muscle
ATP	8 (262)	510	7,854h	0.05	0.2
PCr	32	640	7,697	0.27
ADP	0.008	510	7,854	0.00005
Cr	7	640	7,697	0.06
Pi	0.8	860	7,854	0.009
Mito Ret	280 mM 180 mV	2470 [93 µ2/s/mV]b	157	440 [746]

^a Estimated using the following conversions for the in vivo heart summarized by Aliev et al (390) depending on the reporting parameters: 214 g dry mass/kg wet weight, 2.7 mL cell water/g dry mass. ^bIt is assumed that the major conductance of the mitochondrial reticulum is due to K⁺-Cl⁻; this likely represents an underestimate of the total conductance. The K⁺ conductance, using the data from Hodgkin and Keynes (391) for K⁺ diffusion and mobility corrected for temperature from 18° to 37° C (1.9 × D₁₈= D₃₇) (392). It is assumed that Cl⁻ has a similar conductivity to K⁺ (392). J = –D(ΔC/Δx) J = –Cu(ΔC/Δx); u, mobility; C, concentration. ^cDiffusion coefficients (D) were corrected for temperature using the following relationship determined in solution 1.6 × D₂₅ = D₃₇ (389). ^dAssuming same as PCr (267). ^eAssuming 90% of free water diffusion (389). ^fCytosolic volume participating in process. ATP, ADP, and Pi assumed to be in rapid exchange with mitochondrial volume due to transporters. PCr and Cr diffusion assumed to be limited to cytosolic space. Mitochondrial reticulum is the mitochondrial volume assuming the mitochondria are coupled electrically. Mouse heart data used for this calculation: heart cell diameter is 22 µm with cross-sectional area of 380 µm², 28% mitochondrial volume so diffusional surface area for cytosol metabolites is 274 µm² while the mitochondrial conductive cross section is 106 µm². ^gAssumed slow twitch fiber has diameter of 30 µm with 15% mito content. ^hAssumed fast twitch fiber has diameter of 100 µm with 2% mito content. ⁱThe flux driven by a 10% gradient in metabolites/potential well within most physiological measures. ^jCalculated using 18 mV driving force (10% of the assumed 180 mV mitochondrial membrane potential). ^kCharge (H⁺) to ATP ratio estimated as 4:1. ^lBased on specific power calculations above. Fast- and slow-twitch rates derived from human tibialis anterior maximal OxPhos flux (393) assuming 75% slow-twitch and 25% fast-twitch fiber type composition (394–396).

These estimates suggest that the conductive pathway in the mitochondrion far exceeds the capacity of any metabolite diffusion pathway, suggesting that the conductive pathway, simply based on these physical constraints, would dominate the energy distribution process. The mitochondrial conductive pathway scales with mitochondrial content increasing from gracilis to soleus to the heart cell. It is important to note that the calculated high flux through the mitochondrial reticulum with very small driving forces implies that the mitochondrial reticulum can explain the ability of the tissue to effectively maintain its potential energy in the face of large changes in energy demand in the cytosol. It is also important to note that membrane potential “communication” through the reticulum is reciprocal permitting the regions utilizing the membrane potential to generate a modest ΔΨm decrease throughout the network. Indeed, if the membrane potential generating systems in complex I, III, and IV were sufficiently sensitive to ΔΨm, on the order of the flux dependence on the ΔΨm estimated above, then this retrograde signal could also increase the generation of ΔΨm providing the rapid feedback mechanism to generate a metabolic homeostasis with modest changes in free energy as observed in vivo. The evaluation of the impact of ΔΨm alone on complex I, III, IV, as well as V needs to be quantitatively evaluated, preferably in intact systems to evaluate this signaling hypothesis.

3.2.3.5. distribution of oxidative phosphorylation complexes within the mitochondrial reticulum.

Within the mitochondrial reticulum, especially in skeletal muscle, the paravascular mitochondrial regions are close to the oxygen delivery system yet physically separated from the myosin ATPase activity deep in the muscle cell. While the mitochondrial segments deep in the muscle are closely associated with the ATPase rich regions of the cells, Glancy et al. (169) showed that the distribution of the OxPhos complexes is different in these two regions in oxidative skeletal muscle. In the paravascular regions, the OxPhos complexes that generate membrane potential are in relatively higher abundance than in the intrafibrillar regions while the membrane potential utilizing and ATP producing complex V is in relatively higher abundance in the intrafibrillar regions. An example of this is shown in FIGURE 6A where direct immunohistochemistry shows the relative upregulation of complex IV in the periphery of the cells and complex V in the intrafibrillar region. This distribution of OxPhos complexes is then overlayed on a super-resolution microscopy TMRM image of a mouse soleus fiber (FIGURE 6B) illustrating the tendency for the generation of membrane potential to be in the periphery of the cell with the utilization of membrane potential enzymatically favored in the intrafibrillar region of the cell. This distribution is consistent with the model that ADP and ATP do not have to diffuse all the way to the paravascular mitochondria to utilize this potential energy for OxPhos (FIGURE 7). The potential energy is generated peripherally and delivered deep in the muscle via the reticulum network for ATP generation. In contrast, the heart cell with minimal paravascular pools, and much smaller cells, does not reveal a regionally specific distribution of OxPhos complexes (167) (FIGURE 6C) suggesting a more homogenous generation and utilization of membrane potential in this design.

FIGURE 6.

Distribution of oxidative phosphorylation complexes in striated muscle cells. A: 3-dimensional rendering of an oxidative skeletal muscle cell cross-section immunostained for complexes IV and V and nuclei (blue). Complex IV and V are present throughout the cell, but complex IV is relatively higher in abundance (green) near the periphery and complex V is relatively higher in abundance (red) in the interior. Adapted from Ref. 169. B: overlay of the relative complex IV (green) and V (red) distributions on a super-resolution microscopy rendering of a soleus muscle fiber demonstrating the grid-like physical properties of the mitochondrial reticulum. Adapted from (365). C: complex IV (green, left) and complex V (red, middle) distributions in cardiomyocyte cross sections showing homogenous complex IV/V ratios (right) throughout the heart cell. Adapted from Ref. 167.

FIGURE 7.

Capacity for energy distribution between mitochondria and myofibrils. Top left: 3-dimensional (3-D) rendering of the diffusive energy distribution capacity within the paravascular and intrafibrillar regions of the muscle mitochondrial reticulum. Mitochondria are colored based on the minumum distance to the nearest myofibril. Bottom left: 3-D rendering of the conductive energy distribution capacity within the paravascular and intrafibrillar regions of the muscle mitochondrial reticulum. Mitochondria are colored based on the minimum distance to the nearest myofibril of the closest mitochondria within a connected network. Top right: raw 2-dimensioanl oxidative muscle electron micrograph overlayed with tracings of several paravascular mitochondria. Bottom right: 3-D rendering of the mitochondria from the image above showing the wire-like projections from the paravascular space into the intrafibrillar regions of the muscle cell. All images adapted from Ref. 169.

Using conventional ionic conductances and reticulum scales from TABLE 4, we estimate that the membrane potential gradient from periphery to muscle core could be very small to meet metabolic needs. However, how the ion current is maintained with respect to the classical recycling of protons occurring over these long distances remains unclear. Patel et al. (365) modeled the current flows in an typical I-band mitochondrial geometry in skeletal muscle and found that the low concentration of protons could not carry the required current. Instead, the more dominant cellular cations, Na⁺ and K+, were most likely carrying the current. They demonstrated that with the current density that osmotic and pH challenges would rapidly develop in these small structures without a system to exchange protons for cations across the inner membrane, a process that was first proposed by Mitchell (397) to minimize the pH gradient contribution to the mitochondrial protonmotive force. These authors proposed the hypothesis presented in FIGURE 8 where the well-known cation:proton exchange mechanism is operating in opposite directions in the paravascular and intrafibrillar regions dissipating both the pH and cation gradients without impacting membrane potential, permitting the constant current down the I-band mitochondria without generating osmotic or pH gradients along its length. Other than the demonstration of high cation:proton exchange in muscle mitochondria, the uncertainty of the proton identity of the cation:proton exchanger in skeletal muscle, and the small predicted gradients in cations and pH of the model have made the experimental testing of this hypothesis difficult.

FIGURE 8.

Proposed hypothesis for current generation along mitochondrial tubules. In the paravascular region of a mitochondrion (at right), protons are pumped out by the electron transport chain and then recycled back across the inner mitochondrial membrane through a high abundance sodium or potassium/proton exchanger. In the interior region of the mitochondrion, protons are brought into the mitochondrion through complex V and sent back out through a sodium or potassium/proton exchanger. Current thus proceeds from the interior to the paravascular regions of the mitochondrion through the high concentration sodium or potassium ions. Adapted from Ref. 365.

3.2.3.6. development of the muscle mitochondrial reticulum.

Although mitochondria in adult muscle fibers are arranged in highly organized, stable configurations, this is not the case in early developmental muscle cells such as myoblasts, myotubes, and neonatal cardiomyocytes where mitochondria are highly dynamic and found throughout the cytosol (344, 348). The transition to mature muscle fibers results in mitochondria being surrounded by the contractile apparatus and slowing mitochondrial dynamics; however, the connected mitochondrial network is not yet formed at this stage (346, 398, 399). Thus the muscle mitochondrial reticulum forms after birth, although how this occurs as well as the importance of mitochondrial network formation to muscle development is not well understood. In 1981, Skulachev (398) found that the mitochondrial network took 2 mo to form in the rat diaphragm starting as many individual, disconnected mitochondria aligned longitudinally between the myofibrils at birth before transitioning to the grid-like network found in oxidative fibers. Unfortunately, there has been little investigation into the development of the muscle mitochondrial reticulum since that time and many questions remain. The diaphragm is a very specialized oxidative muscle required for breathing, and thus, may not develop along a similar time course or pattern to glycolytic or locomotor muscles such as found in the legs. Moreover, development is species specific, so additional studies are needed to better understand a particular species of interest such as mice or humans. Recently, the Kim et al. (135) used a light microscopy approach to show that in mouse glycolytic muscle, mitochondrial network development was complete about three weeks after birth and that the mature, perpendicularly oriented mitochondrial networks were preceded by grid-like networks at 2 wk of age and disconnected, parallel networks at birth. However, it is still unclear how this formation pattern differs from comparable oxidative muscle mitochondrial networks as well as the role of individual mitochondrial shape and location changes during development. Higher resolution studies using 3-D electron microscopy may offer insight in this regard.

In the heart, Gong and colleagues (354) reported that the developmental transition from many tortuous, worm-like mitochondria between myofibrils in the perinatal period to the regular, compact shapes observed in mature cardiomyocytes occurred through a complete turnover of all of the mitochondria in the cell. Blocking mitophagy through transgenic manipulation of PINK1-Mfn2-Parkin interactions was sufficient to prevent this turnover and maintained mitochondrial structure and cardiac metabolism in the perinatal state. Whether this postnatal mitochondrial turnover occurs in skeletal muscle remains to be seen, although mitophagy-related proteins have been shown to be upregulated in the perinatal stage relative to mature glycolytic skeletal muscle (135). Additionally, it is currently not well understood what signals drive the formation of specific mitochondrial networks, which can take very different configurations depending on muscle fiber type (303). Determination of mitochondrial network formation is likely related to the transcription factors involved in defining muscle fiber type (400–404); however, metabolic fiber type can be separated from contractile fiber type (e.g., fast oxidative fibers) suggesting that these decision pathways are intertwined but divergent. Untangling these separate but related pathways would go a long way toward our understanding of how muscles with specific contractile and metabolic capacities are built and how these processes may be altered in disease.

3.2.3.7. maintenance of muscle mitochondrial networks.

Once a mitochondrial network is formed, preservation of the network becomes crucial for proper support of muscle function (340, 341). Proteins and structures are constantly built up and broken down in an effort to maintain healthy mitochondria. Mitochondrial fission and fusion are likely part of this balancing act as discussed above. Other mechanisms involved in the maintenance process are mitochondrial biogenesis to increase mitochondrial content and mitophagy to remove damaged or dysfunctional mitochondria (405). There are multiple fluorescent reporters for mitophagy that have been used to visualize the sparse events present within the mitochondrial reticulum of normal muscle (406, 407) though visualization of increasing mitochondrial content as it happens has been challenging. Moreover, mitochondrial protein half-lives are highly variable within cardiac and skeletal muscle (408, 409) suggesting that the bulk changes in mitochondrial content achieved through biogenesis and mitophagy are not the prevailing mechanisms of mitochondrial turnover under normal conditions in mature muscle. Instead, import of newly synthesized proteins and degradation of individual proteins through proteases or other means appear to be the primary methods for maintaining mitochondrial composition. Whether these maintenance mechanisms vary among regions of the muscle mitochondrial reticulum or between fiber types is currently not well understood, though the use of transgenic fluorophores such as MitoTimer (410–412) may offer some insight when coupled with sufficiently high-resolution imaging techniques.

3.2.3.8. mitochondrial structural support for sites of atp utilization in muscle.

As discussed above, mitochondria in the interior of mature muscle can be found both between the myofibrils with groups of adjacent mitochondria often spanning many sarcomeres in length as well as wrapped perpendicularly around the sarcomeres on each side of the z-disk with the proportion of mitochondria in each location depending on fiber type (169, 299, 303, 323). Both oxidative and glycolytic muscle have many mitochondria wrapped around the I-band in close proximity to the terminal cisternae of the SR as well as the lateral ends of the sarcomere thereby providing tight localized coupling of calcium and high energy phosphates between structures in this region. Conversely, glycolytic muscle fibers have relatively few mitochondria arranged along the length of a sarcomere resulting in longer calcium diffusion distances between mitochondria and the longitudinal SR as well as longer ATP diffusion distances between mitochondria and the myosin ATPases near the center of the sarcomere. This problem is overcome in cardiac and oxidative skeletal muscles by placing more mitochondria longitudinally between the myofibrils thus shortening the local calcium and ATP diffusion distances. However, this increased mitochondrial volume comes at the cost of a lower myofibril cross-sectional area as a longitudinal row of mitochondria was placed in between myofibrils instead of more myofibrils. The additional use of longitudinal mitochondria in oxidative and cardiac muscles also comes at the expense of longitudinal SR, thereby lowering the calcium cycling and, thus, contraction frequency of these muscles (413). As a result, cardiac and oxidative muscles are configured such that contractile power is sacrificed in exchange for a greater ability to sustain contractions whereas glycolytic muscle is built for greater power at the expense of endurance. During periods of adaptation due to altered functional demands, the cross-sectional area of the muscle can increase by adding additional sarcomeres in parallel. The specific mechanisms of how the supporting mitochondrial network senses and responds to this change in increased contractile apparatus remain unknown, although there is likely a relationship to the known regulators of mitochondrial biogenesis (405). Conversely, endurance exercise training can lead to increased longitudinal mitochondrial content without an increase in muscle cross-sectional area, resulting in an effective loss in myofibrillar volume. Again, the mechanisms regulating how the contractile apparatus senses and adapts to changes in mitochondrial volume have yet to be elucidated. However, it is likely that muscle adaptation uses many of the integrated developmental processes that occur during muscle maturation but in a slightly different manner analogous to remodeling a house versus building one from scratch.

Yet, if placing mitochondria longitudinally between the myofibril matrix (414) results in the effective displacement of myofibrils, why does wrapping mitochondria perpendicularly around the ends of each sarcomere not result in the same effect? While the I-band mitochondria are thinner than longitudinally arranged mitochondria (169), the lower displacement effect of the I-band mitochondria may be due to a lower sarcomere cross-sectional area at the I-band and Z-disk compared with the center of the A-band. In addition, the intracellular shear associated with contraction is lowest in the I-band and Z-disk region potentially reducing mechanical damage to the mitochondria. TEM images of muscle (415) as well as light microscopy images of isolated myofibrils (416, 417) both suggest that the sarcomere may be thicker in the middle than at the ends. Being thinner in the I-band region would allow more room for mitochondria as well as SR to wrap around the sarcomere without decreasing the effective myofibril cross-sectional area available to generate force within the cell. Indeed, the junctional SR which wraps perpendicularly near the ends of the sarcomere as part of the triad is well known to have a greater diameter than the longitudinal SR which wraps the sarcomere in a mesh-like fashion (418, 419). By building sarcomeres in a slightly nonlinear fashion where the ends are thinner than the middle, more efficient packing of the myofibrils and the organelles required to support contraction would be permitted. More in-depth analysis of sarcomere 3-D structures, particularly in relation to the proximity of mitochondria and SR, would provide more concrete insight into this idea as well as the related hypothesis that the linear arrangement between actin and myosin is not uniform within a single sarcomere.

3.2.3.9. future directions for the muscle mitochondrial reticulum.

There remain a large number of open questions relating to the connectivity of the muscle mitochondrial reticulum, particularly regarding how the connections are formed and regulated, how they vary within different physiological (e.g., male vs. female) or pathological (e.g., obesity/diabetes) environments, and the functional consequences in live animals. One of the most controversial aspects is the functionality of intermitochondrial junctions. Contact sites between both the inner and outer membranes of adjacent mitochondria appear as electron dense regions on electron micrographs and have been observed for decades (169, 299, 326, 349, 363, 420–422) (FIGURE 9A) although the precise function and molecular makeup of these intermitochondrial junctions are currently unclear. Organelle-organelle contact sites are well known to facilitate the transfer of molecules and signals among many different organelle types (423, 424). However, the existence of functional contact sites between adjacent mitochondria has been questioned partly because molecular transfer would be required across four membranes. However, many types of adjacent bacteria have been known to form connections across their four membranes, e.g., microplasmodesmata (425, 426), so possibly this is a conserved process in these organelles of bacterial origin. Contact sites between mitochondria and endoplasmic reticulum (ER) have been shown to facilitate lipid transfer from the ER membrane to the mitochondrial inner membrane (427) suggesting that the highly permeable outer mitochondrial membrane is not necessarily a barrier to molecular exchange between organelles. Further, calcium cycling is also regulated by ER-mitochondrial contact sites (428) showing that rapid ion uptake into the matrix can occur more expeditiously due to organelle contact sites despite having to cross “extra” membranes. Recently, Picard et al. (420) reported that the cristae of adjacent mitochondria appeared to coordinate their alignment across the majority of, but not all, intermitochondrial junctions suggesting this structural arrangement was consistent with electrical conduction of the mitochondrial membrane potential at these sites. Additionally, we demonstrated that while the distance between the respective outer mitochondrial membranes is generally 8–10 nm at intermitochondrial junctions, there are also an abundance of direct contacts (<1 nm spacing) between outer membranes at these junctions (167), which may be the sites at which exchange occurs (FIGURE 9B). Moreover, Bleck et al. (303) found that despite only ∼6% of over 5,000 mitochondrial 3-D structures analyzed being longer than 5 µm in glycolytic and oxidative muscles, functional connectivity in both fiber types was generally observed across distances of 10 µm or more suggesting that conduction of the membrane potential must be occurring by a mechanism beyond propagation within a single mitochondrion and most likely through intermitochondrial junctions.

FIGURE 9.

Intermitochondrial junctions couple adjacent mitochondrial structures. A: electron tomogram of a mouse left ventricle showing intermitochondrial junctions (yellow arrows) between adjacent mitochondria. Adapted from Ref. 167. B: sequential images of the electron tomogram series from the boxed region in A showing direct physical contact between adjacent mitochondria within intermitochondrial junctions. Adapted from Ref. 167. C: schematic of the intermitochondrial junction-based coupling of the paravascular mitochondria and transitional mitochondria into the intrafibrillar region of the muscle cell. D: prevalence of paravascular mitochondrial coupling (purple dots) and transitional mitochondria (green) linking the paravascular and intrafibrillar regions of the mitochondrial reticulum. Adapted from Ref. 169.

Considering the specialized case of paravascular and paranuclear mitochondria in striated muscle, we have proposed that the mitochondria membrane potential is coupled within the paranuclear/paravascular pools via intermitochondrial junctions that are then coupled to the I-band mitochondria emanating from these pools. FIGURE 9C presents this schematically with the blue paravascular/paranuclear mitochondria coupled via intermitochondrial junctions and then coupled to transitional mitochondria that emerge from these pools along the I-band as a single, contiguous mitochondrion that runs deep in the muscle. In FIGURE 9D, the prevalence of coupled mitochondria (purple marks) and mitochondria coupled with contiguous matrix into the intrafibrillar region are shown as green/purple marks. The vast majority of paravascular mitochondria are coupled via intermitochondrial junctions while the interface between the paravascular mitochondria pool and intrafibrillar region is lined by transitional mitochondria coupled to both the paravascular and intrafibrillar regions of the muscle cell.

It is important to note that the intermitochondrial junction coupling hypothesis discussed here is based on a strong correlation between structural and functional measurements. However, direct, functional measurements of the conductivity across intermitochondrial junctions have yet to be made. It is clear that greater experimental precision is needed to fully elucidate the role of intermitochondrial junctions in muscular energy distribution and will likely require the physical isolation of these junctions for functional testing and/or technical advances allowing direct measurements of ionic currents with subcellular precision under the microscope. Performing localized depolarization experiments such as with MitoPhotoDNP (167, 169, 303) on muscles under a superresolution microscope with individual mitochondrial resolution may offer more exact knowledge on how mitochondrial membrane potential conduction is occurring within the muscle mitochondrial reticulum as well as offer better insight into the role of intermitochondrial junctions within the power grid protection system (167). Additionally, while there are techniques available to isolate and assess the specific nature of contacts between different organelles, there is currently a lack of methods to probe junctions between the same type of organelle, which presents a great challenge in determining the molecular identity of the putative components associated with intermitochondrial junctions. Proteomic screens of cells under conditions with more or less intermitochondrial junctions may be helpful in narrowing down potential candidates, although the lack of junctions in in vitro muscle cells make high-throughput candidate testing difficult in these models. A better option may be a nonmammalian model such as Drosophila, which has well characterized flight muscles in addition to an array of available transgenic approaches though care must be taken when attempting to translate results back to mammalian systems.

Another controversy surrounding the electrical connectivity of the muscle mitochondrial reticulum is the use of MitoPhotoDNP to cause localized depolarization and all the potential confounding variables that could arise. The use of a photoactivatable uncaging molecule implicitly requires increased light exposure on the cells of interest, and photodamage is a well-known potential side effect that can also cause a redistribution of electrical potential energy (376–380). Simple light exposure controls with and without MitoPhotoDNP suggested that the UV power required to uncage DNP has a negligible effect on muscle fibers (169), but it is possible that addition of MitoPhotoDNP could alter the sensitivity of the cells to photodamage. This possibility has been tested in two ways. First, we purposely induced localized photodamage in muscle fibers by increasing the UV power eightfold, which led to asynchronous, repetitive, transient membrane potential depolarizations (373) only in the damaged region with no apparent depolarization to the adjacent mitochondria (167). Thus the response to photodamage is different than to uncaging MitoPhotoDNP. Second, during the normal photoactivation step in the presence of MitoPhotoDNP, we irradiated two separate regions of the same cell, one with a 355-nm laser that can uncage MitoPhotoDNP, and the other with a 405-nm laser which does not uncage MitoPhotoDNP. Upon photoactivation of muscle fibers from MitoDendra2 mice, the fluorescence of mitochondria in both regions switched from green to red with equal intensity. However, only the mitochondria in the 355-nm exposed region underwent the physical remodeling of the mitochondrial network that occurs after localized depolarization, suggesting that this response was specific to depolarization and not due to photodamage (167).

It is also possible that diffusion of TMRM or uncaged DNP could have resulted in the apparent shared depolarization immediately after photoactivation of MitoPhotoDNP. Diffusion of TMRM toward or away from a region would result in an increase or decrease in fluorescence in that region, respectively. However, in both the irradiated and adjacent regions, the TMRM signal goes down in the mitochondria and up in the cytosol, which cannot be explained by TMRM diffusion (167, 169, 303). Rapid diffusion of the uncaged DNP could depolarize mitochondria in adjacent, nonirradiated regions. However, after the initial shared depolarization, the nonirradiated region repolarizes back to baseline levels suggesting that DNP is not present in this region 5–10 s after depolarization (167). Another possibility is that uncaged DNP diffuses to the adjacent regions and is only concentrated enough to induce depolarization transiently before continuing to diffuse away. To test this possibility, MitoPhotoDNP experiments were performed in cardiac, oxidative, and glycolytic muscle fibers that have parallel, grid-like, and perpendicular oriented mitochondrial networks, respectively (167, 303). In all cell types, irradiation was performed separately along both the perpendicular and longitudinal axes to test the directionality of electrical coupling. For each cell type, the functional connectivity matched the orientation of the physical networks (FIGURE 10). Cardiac fibers had greater connectivity along the parallel axis, glycolytic fibers had greater connectivity along the perpendicular axis, and oxidative fibers had similar connectivity along both axes (167, 303). Diffusion of molecules including oxygen, calcium, ATP, and even photons is well known to occur more expeditiously along the longitudinal axis in muscle fibers (429–432). Thus transient DNP diffusion could only potentially explain the results in the longitudinally aligned mitochondrial networks of cardiac fibers but not in either skeletal muscle fiber type, which suggests that DNP diffusion is likely not a factor in the MitoPhotoDNP experiments. Nonetheless, additional experimental evidence using different methods would greatly add support to the current data on the electrical connectivity of the muscle mitochondrial reticulum. A cell-permeable, photoactivatable ADP molecule would be helpful for creating transient, localized mitochondrial energy demands and in overcoming the many issues associated with mitochondrial uncouplers. Additionally, recent development of optogenetic mitochondrial transgenes that insert a photoactivatable ion channel into the inner mitochondrial membrane (433, 434) may offer an even cleaner approach to induce spatially and temporally controlled depolarization without any issues related to rapid probe diffusion.

FIGURE 10.

Structural and functional connectivity of striated muscle mitochondrial networks. Top left: structural and functional connectivity of cardiac muscle mitochondrial reticulum is primarily parallel to the axis of contraction. Top right: Structural and functional connectivity of oxidative muscle mitochondrial reticulum occurs both parallel and perpendicular to the axis of contraction. Bottom right: structural and functional connectivity of glycolytic muscle mitochondrial reticulum is primarily perpendicular to the axis of contraction. Bottom left: ratiometric mitochondrial membrane potential voltage maps in response to localized depolarization along each axis in cardiac, oxidative, and glycolytic muscles as in FIGURE 5. Adapted from Refs. 167, 303).

What is the benefit of the speed of electrical conduction of the mitochondrial membrane potential if the upstream (e.g., fuel and NADH) and downstream (e.g., ATP) molecules in the energy conversion pathway still rely on slower movement by diffusion? With regard to carbon substrates, the amount of ATP generated per pyruvate or lactate molecule is roughly 14 ATP/mol assuming complete rotation of the citric acid cycle and minimal proton leak, so the diffusion time could be much slower than ATP itself. Little information on the concentration gradients for carbon substrates within muscle cells is available (435). Additionally, while rapid distribution of a key intermediate in the OxPhos pathway would help maintain homogeneity of the diffusion driving forces throughout the cell, the membrane potential is also a critical driving force for many other mitochondrial processes besides the formation of ATP by ATP synthase. Indeed, the membrane potential is utilized for substrate transport into mitochondria (436), ATP transport out of mitochondria (437, 438), and ion exchange across the inner mitochondrial membrane (439, 440). Specifically, entry of calcium and sodium into mitochondria, both of which have been shown to activate the mitochondrial energy conversion process through different mechanisms (68, 69, 441–443), is dependent on the membrane potential. Consequently, the entire mitochondrial energy conversion process from fuel transportation (436, 444), to substrate dehydrogenases (445), to the electron transport chain (68, 446), to ATP transport out of mitochondria (437, 438) relies either directly or indirectly on utilization of the membrane potential. As such, maintaining a consistent membrane potential in the face of energetic perturbations is critical to mounting a rapid response to meet the new demands. In fact, maintenance of the membrane potential appears to be so critical that cells will run ATP synthase in reverse and hydrolyze ATP to maintain the membrane potential as a protective measure upon the onset of ischemia (447–451). Thus maintenance of the membrane potential by rapid conduction through the mitochondrial reticulum would play a role throughout the energy conversion pathway bringing together the power of the membrane potential generating complexes I, III and IV to provide a much larger benefit than simple activation of the ATP synthase enzyme.

4. SUMMARY OF THE METABOLIC SUPPORT DESIGN OF THE MUSCLE CELL

Force generation by the striated muscle cell is an energetically expensive activity, which requires significant investment of valuable cellular real estate in supporting energy conversion and distribution processes. The demonstration of a metabolic homeostasis, as first described by Hill (34), illustrates the balance between energy conversion and utilization by contraction and contraction control processes. The homeostatic nature of the free energy in the cell is likely to maintain the driving force for contraction as well as generating ion gradients such sarcoplasmic reticulum calcium during increases in work. Energy homeostasis also implies that large changes in metabolite concentration gradients driving metabolite diffusion through the muscle do not occur. How is this balance between energy conversion and work performed? Much of the sustained power of muscles is supported by OxPhos. Despite a nearly 10-fold difference in sustained work demands of different muscles, the protein programming of mitochondria is nearly identical. That is no change in mitochondrial internal design despite large differences in cellular demand. Thus the content and architecture of mitochondria are the major variables for tuning mitochondrial ATP production capacity to match muscle action. Indeed, as seen in this review, the content and structure of mitochondria in different muscle types are variable. In addition to ATP production, energy must be distributed throughout an ATPase rich cytosol. Facilitated diffusion through myoglobin and creatine kinase has long been considered primary energy distribution mechanism in muscle cells. However, the relatively modest functional consequences and adaptations observed with the loss of these systems, together with computational investigations and the observation that most metabolite concentrations are held constant, have brought newfound skepticism concerning the importance of myoglobin and creatine kinase driven facilitated diffusion under normal physiological conditions. Recent experimental support for the hypothesis that electrical energy could be conducted through mitochondrial reticulum networks has now suggested an alternative, and likely much faster, mechanism for cellular energy distribution that can function with minimal free energy or metabolite gradients. Each muscle type has adapted the mitochondrial structure to apparently meet the remarkable energy conversion and distribution challenge in this cellular machine.

The constantly beating heart is a highly aerobic tissue, precisely matching the delivery of oxygen to the OxPhos machinery. However, the actual mechanisms associated with this balance of flow with workload are not understood. With alteration in workload, the heart is capable of maintaining both the free energy of ATP as well as the PCr content implying a matching of ATP production with ATP utilization. Again, the cellular mechanisms that balance ATP production with ATP utilization are also poorly understood. Both of these “homeostatic” processes suggest that the energy transfer systems within the heart cell are extremely efficient in delivering appropriate potential energy across the heart cell, which is essentially a near solid mass of ATPase activity. Facilitated diffusion of metabolites and oxygen was early idea to explain the rapid distribution of potential energy in the cell; however, recent genetic manipulations along with biophysical measures have suggested that these facilitated diffusion processes do not dominate the energy distributions processes in the heart. The stable network structure of mitochondria within the cardiomyocyte may play a larger role in the capacity for cellular energy distribution, although the precise mechanisms that regulate delivery of potential energy throughout the mitochondrial reticulum have yet to be elucidated. Physical coupling of the many adjacent mitochondria linked together across the paravascular, paranuclear, and intrafibrillar regions appears to provide a structural pathway for maintaining energy homeostasis within connected networks. Electrical coupling across the structural intermitochondrial junctions would allow for rapid distribution of the membrane potential, a key regulator of many mitochondrial processes, throughout the cell, although detailed information on how and when electrical conduction may occur remains to be uncovered.

Despite the larger and more intermittent demands for force production in skeletal muscle relative to the heart, oxygen delivery to the OxPhos machinery does not to appear to be limiting across the aerobic scope. While oxygen partial pressures in skeletal muscle do fall with an increase in workload, unlike in the heart, the oxygen gradient from the capillary to the mitochondrion appears to remain stable demonstrating an alternate homeostatic solution. However, the mechanisms of how a muscle contraction-induced increase in oxygen flux occurs in the absence of a change in driving force are still unclear. Again, in contrast to the heart, the structural design of the skeletal muscle prioritizes maximal power over perfect maintenance of energy homeostasis. As a result, the onset of muscle contraction is marked by a slight fall in ATP free energy which is minimized by the fiber-type specific efficiency of the cellular energy distribution system within the skeletal muscle cell. While the protein composition of the OxPhos machinery per mitochondrial volume is similar across the glycolytic to oxidative metabolic spectrum of muscle fibers, the size and configuration of the mitochondrial reticulum are largely dependent on muscle fiber type. Glycolytic muscle fibers have smaller mitochondrial networks comprised largely of thinner mitochondria aligned perpendicularly to the muscle contraction axis likely as a means to minimally displace the contractile apparatus. Oxidative fibers have severalfold larger mitochondrial networks made of thicker mitochondria aligned both parallel and perpendicular to the contractile apparatus demonstrating a design priority for more sustainable force production compared with the glycolytic muscle though not to the degree of the heart. However, the mechanisms that guide design of the cellular energy distribution system across striated muscle types remain to be resolved.

GRANTS

This work was supported by funding from the Division of Intramural Research of the National Heart, Lung, and Blood Institute and the Intramural Research Program of the National Institute of Arthritis and Musculoskeletal and Skin Disease.

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the authors.

AUTHOR CONTRIBUTIONS

B.G. and R.S.B. conceived and designed research; B.G. and R.S.B. analyzed data; B.G. and R.S.B. interpreted results of experiments; B.G. and R.S.B. prepared figures; B.G. and R.S.B. drafted manuscript; B.G. and R.S.B. edited and revised manuscript; B.G. and R.S.B. approved final version of manuscript.