Sci. Aging Knowl. Environ., 7 August 2002
Mitochondrial Bioenergetics, Aging, and Aging-Related Disease
The author is at the Buck Institute for Age Research, Novato, CA 94945, USA. E-mail: firstname.lastname@example.org://sageke.sciencemag.org/cgi/content/full/sageke;2002/31/pe12
Key Words: mitochondria bioenergetics antioxidants reactive oxygen species neurodegeneration neurodegenerative disease
This Perspective was written by a "classical" mitochondrial bioenergeticist new to the field of aging, who is attempting to come to terms with the extensive literature on mitochondrial theories of aging and aging-related disease. Clearly it is not possible to be comprehensive; instead, I shall discuss some of the bioenergetic consequences, for good or bad, of the various alterations in mitochondrial function that have been hypothesized, or actually demonstrated, to be associated with human aging and aging-related disease. More than 30 years of investigation by many groups into the bioenergetic behavior of "normal" mitochondria (isolated typically from tissues of the adult rat) have provided a foundation on which to base the more complex and speculative in vivo studies.
In our laboratory, we use a simple electrical analogy to describe the proton circuit linking the mitochondrial respiratory chain with the adenosine triphosphate (ATP) synthase (Fig. 1). Embedded in the inner membrane are three proton pumps (complexes I, III, and IV) driven by the sequential "downhill" flow of electrons from reduced nicotinamide adenine dinucleotide (NADH) to oxygen. Each pump extrudes protons from the matrix, generating a membrane potential (or, more correctly, a protonmotive force, p, which takes account of the generally small pH gradient built up across the inner mitochondrial membrane). The three pumps contribute in parallel to the proton current, and the circuit is completed by the potential-driven reentry of protons through the ATP synthase, forcing this ATP-hydrolyzing, proton-extruding pump to run backward and synthesize ATP. A small but important proton leak provides an alternative reentry path. The proton current is governed by Ohm's law (current = voltage/resistance); the "resistance" to proton reentry is governed by the rate of ATP synthesis and the state of the proton leak. Electron flux through the respiratory chain, and hence the rate of respiration, is proportional to the proton current [see (1) for review]. Single electrons can "leak" to oxygen from complexes I and III, forming the toxic superoxide anion (O2-), a reactive oxygen species (ROS).
In addition to ATP generation, mitochondria control a complex network of fundamental bioenergetic interactions (1), including (i) generation and detoxification of ROS; (ii) maintenance of the antioxidant glutathione in a reduced state; (iii) cytoplasmic calcium homeostasis, particularly under conditions of cellular calcium loading; (iv) transport of metabolites between cytoplasm and matrix; and (v) both programmed and necrotic cell death (Fig. 2). It is not therefore surprising that the mitochondrion has come to center stage in so many current investigations into aging and aging-related disease (see "The Two Faces of Oxygen").
The mitochondrial membrane potential, , or p is the "voltage" term (see Fig. 1) that occupies the central position in this network and accounts for much of its interconnectivity (Fig. 2). Intricate control mechanisms integrate these various parameters. For example, increasing ATP (cellular energy) demand lowers , which results in an increase in electron flux through the respiratory chain and, contrary to what is sometimes believed, a decrease in the production of ROS and in proton leakage. Thus any realistic discussion of mitochondrial bioenergetics in the aging context has to take a holistic approach in considering the likely consequences of disturbing any component in this network. Metabolic control analysis (2) provides a rigorous but underexploited treatment of such manipulations and will be summarized briefly below.
Why Are Even Slight Restrictions in Maximal Respiratory Chain Capacity So Dangerous?
Here in California, we are threatened with electricity blackouts each summer as peak demand threatens to exceed supply capacity. The failure of even one power plant can tip the balance and lead to an energy deficit, a voltage drop or "brown-out," and computer crashes. In contrast to most other tissues, which have a relatively constant ATP turnover, brain and muscle cells have a high and variable energy demand; thus their mitochondria must possess sufficient spare ATP-generating capacity to cope with peak demands such as vigorous exercise or rapid neuronal firing. It is notable that many mitochondrial mutations manifest as diseases of the brain and/or muscle even though the mutation is equally distributed throughout all tissues of the body (3). The consequences of partial respiratory chain restriction have been most thoroughly investigated in the brain, where it is established, for example, that partial inhibition of complex I by 1-methyl-4-phenylpyridinium (MPP+), which accumulates selectively in dopaminergic neurons, induces experimental Parkinsonism (see Andersen review) (4). More significantly, low systemic administration of the complex I inhibitor rotenone is selectively toxic to the dopaminergic neurons of the substantia nigra, even though the inhibitor acts uniformly throughout the brain (5). The implication is that these neurons possess the least safety margin of ATP-generating capacity. Similarly, systemic administration of complex II inhibitors such as malonate or 3-nitropropionate reproduces some of the pathology of Huntington's disease (6). It is not known why systemic inhibition of the two complexes selectively generates different pathologies, although it almost certainly reflects differing spare ATP-generating capacities.
An appreciation of metabolic control analysis is helpful in order to understand the functional consequences of partial respiratory chain inhibition under conditions of variable ATP demand (2). The traditional concept of a single rate-limiting step in a multistep pathway is incorrect, because each step exerts some control over the overall flux through the pathway (2, 7). The flux control coefficient for an individual step in the pathway is equal to the fractional change in flux through the entire pathway in response to a small fractional change in activity of the step itself. The sum of all the control coefficients for the steps in a pathway equals 1. In mitochondria, flux control coefficients have been calculated for metabolite supply, for each respiratory chain complex, for transport and phosphorylation of ATP, and for the inherent inner membrane proton leak. The analysis could (and perhaps should) be extended to each step shown in Fig. 2.
Flux control coefficients are not constant but vary with the metabolic status of the mitochondria (7). Expressed most simply, under conditions of low cellular ATP demand, most of the control of respiration is exerted by the inherent inner membrane proton leak (Fig. 3). As cellular ATP use increases, the control exerted by the proton leak decreases and ATP demand becomes the dominant factor controlling respiration rate. Finally, under high cellular energy demands, control is shared by ATP demand and the respiratory chain (Fig. 3). In practical terms, this means that manipulation of the proton leak would have the most effect on resting cells, whereas respiratory chain capacity is of great importance under conditions of maximal energy demand. Partial inhibition of a respiratory chain complex increases its flux control coefficient. As is intuitive, the effects of this will be greatest under conditions of high respiration and ATP demand.
As well as changing the pattern of control, increasing the demand for ATP causes a modest decrease in , which in turn is believed to decrease ROS generation (see below). Changes in the activity of one or more respiratory chain components not only exert control over respiratory rates but also perturb the steady-state levels of intermediates in the respiratory chain. For example, the ubiquinone/ubiquinol (UQ/UQH2) pool connects complexes I (NADH dehydrogenase) and II (succinate dehydrogenase) with complex III (ubiquinone-cytochrome c reductase). In a steady state, the flux of electrons into the UQ pool must exactly balance the flux out of the pool. A partial inhibition of the downstream complex III will lead to an increased reduction state of the UQ pool until a new steady state is achieved, in which electron entry from complexes I and II is restricted and/or electron flow into complex III is enhanced. Similarly, p is the link between the upstream respiratory chain (complexes I through IV) and the downstream proton leak and ATP synthase. Partial inhibition of the respiratory chain will decrease p until the proton current used for the leak and ATP synthesis exactly balances that generated by the respiratory chain. The consequences of these effects are apparent in models of oxidative stress. For example, partial inhibition of the electron transport chain downstream of a potential site for superoxide generation (for example, complexes I and III) can increase the extent of reduction of the intermediate that is "leaking" electrons to oxygen and thus produce ROS. The converse situation, decreased ROS production, has been hypothesized to be a consequence of an increased proton leak. This will now be discussed.
Could a Controlled Increase in the Proton Conductance Be Neuroprotective?
There are two sites where electrons can "leak" from the respiratory chain to molecular oxygen to generate the superoxide anion O2- (8). In complex III, UQH2 loses two electrons sequentially at the outer UQ binding site. The loss of the second electron is delayed when is high, and the increased dwell time of the semiquinone radical at this site increases the chances of an electron being transferred to oxygen to generate O2-, which is then released into the intermembrane space. Less is known about the second site for O2- generation, at complex I (9), except that it is upstream of the rotenone binding site. It is likely that the complex I site releases O2- into the matrix. On the basis of these findings, researchers have proposed that it might be beneficial to artificially lower , by enhancing proton conductance, to just below the threshold needed for O2- production (10). A subtle lowering of the membrane potential should not adversely affect other vital functions of the mitochondrion, such as ATP production. It should be noted, however, that -dependent production of O2- by isolated mitochondria may be dependent on depletion of their reduced glutathione pool (11).
All mitochondria possess some inherent proton leak across their inner membranes. As discussed above, this leak is most apparent at high p and accounts for the slow respiration of mitochondria in the absence of ATP synthesis. Twenty-five years ago, we described the first mitochondrial protein whose function is to increase the proton leak (and thus decrease ATP synthesis) (12). Because the coupling between the respiratory chain and ATP synthase via the proton circuit is dependent on the proton impermeability of the inner membrane, the evolution of a short-circuiting "uncoupling protein" (UCP1) was surprising. However, UCP1 expression was shown to be confined to a specialized tissue, brown fat, whose function is to oxidize triglycerides at high rates, uncontrolled by cellular ATP demands, in order to generate heat during arousal from hibernation. The protein was also shown to be inducible by cold and tightly regulated, functioning only in the presence of fatty acids liberated as an energy source by hormone-activated lipolysis. Much more recently, gene homology studies have identified several novel candidate uncoupling proteins. These include UCP2, UCP3, UCP4, and UCP5 (BMCP). The best-characterized proteins in this group are UCP2 and UCP3, whereas UCP4 and BMCP (UCP5) show more tenuous homology with UCP1 (13). It has long been apparent that uncoupling proteins have the potential to counter obesity, and this accounts for much of the current interest in these proteins.
There is a lively ongoing debate as to which (if any) of these proteins actually functions in vivo to increase proton conductance; one problem is that artifactual overexpression of any membrane protein can increase proton conductance in a nonspecific manner by simply disturbing the bilayer. A second problem is that the levels of expression of the novel UCPs are generally much less than that of UCP1 in brown fat, where the induced protein can account for almost 10% of the inner membrane protein. Finally, it is unclear how the conductance of the novel UCPs would be controlled, although there is an intriguing proposal that UCP2 and UCP3 might be activated by (and conduct) O2- (14), providing a mechanism for the expulsion of this ROS from the matrix as well as for lowering the membrane potential and decreasing ROS generation itself.
It is important to consider the bioenergetic consequences of enhancing proton conductance before pursuing the hypothesis that this phenomenon might be useful for reducing aging-related oxidative stress. Titrating isolated mitochondria with a synthetic protonophore while monitoring (or p) and rate of respiration yields relationships between respiration and similar to those shown in Fig. 3. Upon viewing such a graph, the basic problem is immediately apparent: The mitochondrial respiratory chain is so efficient that it can deliver an increased proton current (depicted as ATP synthesis in Fig. 3) with little decrease in until maximal respiratory chain activity is attained. This is compounded by the ability of key enzymes in the citric acid cycle to increase their supply of NADH under conditions of enhanced energy demand. In summary, one can decrease by increasing proton leak, but at a great cost in terms of energy demands on the respiratory chain. The question that must thus be addressed is whether the supposed decrease in O2- production outweighs the inevitable decrease in the reserve capacity of the mitochondria to generate ATP in times of peak energy demand. Bear in mind that the tissues that are most susceptible to failure in aging and aging-related disease are muscle and brain, and that these are precisely the organs that are required to cope with irregular peaks in ATP demand. Physiologically enhanced proton conductance might well turn out to be neuroprotective, but it would clearly require careful region-selective regulation and control, and cannot be considered as a global constitutive mechanism.
Why Does a Decrease in Glutathione Pools Enhance Oxidative Stress and Neurodegeneration?
Glutathione (GSH) is a tripeptide composed of glutamic acid, glycine, and cysteine and is a powerful antioxidant that helps to neutralize the effects of ROS on cellular components. The oxidized form of GSH is referred to as GSSG. The GSH/GSSG pools in the mitochondrial matrix and cytoplasm provide the main defenses against oxidative stress, both by maintaining protein thiols in the reduced state and by detoxifying H2O2 via glutathione peroxidase (15). Whilst it might seem consequently intuitive that partial glutathione depletion is deleterious to the cell, the underlying reasons are rather complex; after all, the redox potential of, for example, a 50% reduced pool of NAD+/NADH is -320 mV regardless of whether the total pool size is 10 mM or 5 mM. Why then is the glutathione pool functionally so sensitive to depletion? The answer appears to be a fascinating combination of thermodynamics and kinetics. First, the glutathione pool is continuously turning over and is very far from thermodynamic equilibrium. Otherwise, the redox potential of the matrix GSH/GSSG pool, which appears to be normally around -250 mV (16), would be closer to that of the matrix NADP+/NADPH pool (-380 mV) as GSH is regenerated by the NADP-linked glutathione reductase enzyme (Fig. 2). The kinetic (rather than equilibrium) nature of the glutathione redox state is also apparent from the ease with which the pool can become oxidized by substrates for glutathione peroxidase, such as t-butylhydroperoxide or H2O2 itself.
One clue to the susceptibility of the cell to glutathione depletion is the nature of its redox reaction: Two GSH molecules are oxidized to give one molecule of GSSG. In thermodynamic terms, this means that, unlike the NADP+/NADPH pool discussed above, the midpoint potential (that is, the redox potential when concentrations of GSH and GSSG are equal) is dependent on the absolute concentrations of GSH and GSSG. In order to maintain a constant redox potential as the pool size is decreased, the ratio of GSH to GSSG must increase. Combined with the absolute decrease in pool size, this means that the concentration of GSSG in the mitochondrial matrix or cytoplasm must decrease dramatically to maintain the same redox potential [for a quantitative treatment, see chapter 5 of (1)]. Because the glutathione pool is being continuously turned over, such a decrease in GSSG concentration could pose a kinetic problem for glutathione reductase. The consequence of a decreased glutathione pool, such as has been observed or induced in aging-related models (17), is to make the redox potential of the pool more oxidized and less able to maintain protein thiols in the reduced form.
How Do Mitochondria Influence Glutamate Excitotoxicity?
Glutamate is used as the neurotransmitter at the majority of central nervous system synapses. The postsynaptic receptors for glutamate include alpha-amino-3-hydroxy-5-methylisoxasole-4-propionic acid (AMPA) receptors, which are used in normal synaptic transmission, and NMDA (N-methyl-D-aspartate) receptors, which are glutamate-gated ion channels that have a particular role in learning and memory. In order to use such a universal amino acid as a transmitter, it is essential to maintain precise compartmentation; if this fails, glutamate becomes a neurotoxin, killing neurons in a process termed glutamate excitotoxicity.
Mitochondrial function, oxidative stress, and glutamate excitotoxicity are intimately interconnected (Fig. 2). Depriving a region of the brain of oxygen and glucose during stroke leads to an immediate failure of mitochondrial ATP synthesis, rapid depletion of ATP, collapse of the Na+ electrochemical gradient across the plasma membrane, and reversal of the plasma membrane glutamate transporters, which results in a massive efflux of glutamate from the cell (18). Diffusion of glutamate to the region surrounding the infarct (the penumbra) results in the pathological activation of NMDA receptors and the accumulation of intracellular Na+ and Ca2+. Because Na+ entry into the cell collapses the Na+ electrochemical gradient across the plasma membrane, the driving force for the plasma membrane glutamate transporters to remove glutamate from the extracellular space is lacking. This might explain why NMDA receptor-mediated neuronal damage seems to proliferate after the restoration of blood flow. There is a "catch 22" situation in which the removal of glutamate requires a Na+ gradient, and the presence of glutamate collapses that gradient.
When the intracellular Ca2+ concentration exceeds 0.5 µM, the activity of the mitochondrial Ca2+ channel, which is responsible for uptake of the cation into the mitochondria, exceeds that of the mitochondrial Na+/Ca2+ exchange efflux pathway. Isolated mitochondria are rather perfect Ca2+ buffers, particularly in the presence of phosphate. They are able to decrease external Ca2+ concentrations to 0.5 µM until enormous quantities of the cation, up to 1 M, have been accumulated in the matrix (19). Subsequently, mitochondria undergo a nonselective permeabilization of the inner membrane, termed the mitochondrial permeability transition (MPT), release their Ca2+ into the cytoplasm, depolarize, and hydrolyze rather than synthesize ATP.
Within intact neurons, and in the context of glutamate excitotoxicity, activation of the MPT as a consequence of mitochondrial Ca2+ overload must be considered as a mechanism of neuronal cell death. A massive MPT induces rapid cell necrosis, whereas an MPT limited to a subpopulation of mitochondria within a single cell could release sufficient cytochrome c to initiate apoptosis, while the remaining population maintained cellular ATP concentrations. It is possible that MPT induction occurs in neuronal cultures chronically exposed to glutamate. There are differing reports as to the ability of MPT inhibitors such as cyclosporin A or bongkrekic acid to delay the failure of cytoplasmic Ca2+ homeostasis that precedes necrotic cell death. However, these are imperfect tools that lack selectivity in cellular systems. Furthermore, transient glutamate exposure can induce a delayed cell death hours after the glutamate is withdrawn, even though the mitochondria have long before expelled their Ca2+.
A second mechanism by which glutamate can cause cell necrosis is by simple bioenergetic overload. Chronic NMDA receptor activation imposes a double pressure on the neuron. First, ATP-requiring plasma membrane pumps are activated to expel the Ca2+ and, more particularly, the Na+ that entered via the NMDA receptor. A fully activated Na+/K+-ATPase might require nearly all of the maximal ATP-generating capacity of the mitochondria. Compounding this, the ability of the mitochondrion to generate ATP during Ca2+ accumulation is decreased, because the two processes compete for the proton current. If ATP demand exceeds supply, then the ATP/ADP ratio will collapse. An ATP-depleted cell might not recover even after termination of NMDA receptor activation, as, paradoxically, neurons require ATP to generate ATP (due to the two ATP-requiring reactions of glycolysis). Pyruvate (or lactate) might be able to rescue these ATP-depleted cells by directly supplying substrate to the mitochondria. It is apparent (and readily demonstrable in vitro) that adversely affecting the balance between NMDA receptor and mitochondrial respiratory chain activities greatly facilitates glutamate excitotoxicity, consistent with the energy-linked excitotoxicity hypothesis proposed several years ago by Henneberry (18). This observation also suggests a direct mechanism for the neurodegenerative consequences of having only partial respiratory chain activity, as discussed above.
The third process associated with glutamate excitoxicity is oxidative stress, although many aspects of this connection remain to be clarified. In the experimental context, it is important to distinguish between the absolute generation of ROS and the proportion of ROS that can be trapped by the oxidation of reporter molecules. For example, the reporter dihydroethidine, which is oxidized relatively selectively by O2- to fluorescent ethidium, will compete with superoxide dismutases, nitric oxide-mediated generation of the ROS peroxynitrite, and actual O2--mediated oxidation of proteins and lipids; thus the reporter will capture only a small percentage of the total ROS generated. Similarly, indicators responsive to intracellular H2O2 compete with glutathione peroxidase (catalase activity, which converts to H2O2 to water and oxygen, is low in neurons). Does an increased signal by a ROS reporter molecule, therefore, represent increased ROS generation or decreased trapping by competing processes? Ca2+ loading of mitochondria is generally assumed to increase the generation of ROS, although actually demonstrating this in a physiologically relevant experiment is not easy. Antioxidants generally are not effective in prolonging the survival time of neuronal cultures after exposure to glutamate, although in the late stages of necrosis, where mitochondria depolarize and cytoplasmic Ca2+ increases, a large increase in ROS generation (or overflow?) can be observed.
How Does One Relate Acute Glutamate Excitotoxicity to Slow Chronic Neurodegeneration?
With cultured neurons, in vitro glutamate excitotoxicity results in cell death in minutes or hours. In contrast, slow, insidious neurodegenerative disorders may take years or decades to destroy sufficient cells to produce symptoms. It is likely that this latter process does not occur uniformly but rather as a result of a series of downward quantal steps, either in individual neurons or in neuronal populations, that is perhaps initiated by sporadic local glutamate-induced bioenergetic crises in which ATP demand exceeds supply. This would clearly be exacerbated in individuals with compromised ATP generation, be it by mitochondrial mutation, environmental factors, or the natural aging process. Regardless of the mechanism, it is apparent that a more detailed quantitative understanding of the bioenergetic network shown in Fig. 2 is vital for our understanding of some of the most prevalent aging-related neurodegenerative disorders.
August 7, 2002
THIS ARTICLE HAS BEEN CITED BY OTHER ARTICLES:
Science of Aging Knowledge Environment. ISSN 1539-6150