The Comparative Biology of Mitochondrial Function and the Rate of Aging

Abstract

The mitochondrial hypothesis of aging evolved from the rate-of-living theory. That theory posited that the rate of aging was largely determined by the rate of energy expenditure. The mechanistic link between energy expenditure and aging was hypothesized to be oxidative stress. As both energy expenditure and reactive oxygen species (ROS) centered on the mitochondria that organelle became a central focus of aging research. Until about the turn of the 21st century available evidence largely supported the efficiency of mitochondrial function as a key contributor to aging. However as methods for investigating mitochondrial oxidant production and tissue level oxidative damage improved, evidentiary support for the theory weakened. Recently, direct disruption of mitochondrial function has been shown not to shorten life or health as expected, but in many cases in multiple laboratory species disrupted mitochondrial function has lengthened life, sometimes without apparent tradeoffs. Does this mean that mitochondrial function plays no role in aging as had been posited for many years? One key consideration is that experiments under laboratory conditions can be misleading about physiological processes that occur in the uncertain conditions of nature. Before we discard the mitochondrial hypothesis of aging, more field experiments targeted at that hypothesis need to be performed. Fortunately, emerging technology is making such experiment more possible than ever before.

Introduction: The mitochondrial hypothesis of aging

Cooling poikilothermic species such as flies within their tolerance limits reduces their metabolic rate and also increases their longevity (Miquel et al. 1976). Warming them has the opposite effect. These observations led to the notion that aging and longevity were largely, if not exclusively, determined by the rate of energy expenditure (Pearl 1928). That notion was also consistent with the pattern of interspecific differences in mammalian longevity as larger species typically have lower basal mass-specific metabolic rates and live longer than smaller species (Rubner 1908; Sacher 1959; Speakman 2005). A potential mechanistic link between metabolism and longevity was forged by Harman who, extrapolating from radiation chemistry, hypothesized that free radicals, due to the direct utilization of oxygen during metabolism, were the tissue-damaging agents largely responsible for the progressive degeneration we call aging (Harman 1956). The large armamentarium of antioxidant defenses, including a broad range of scavengers of superoxide (${\mathrm{O}}_2^{-}$) and hydroxyl (^•OH) radicals, deployed by all species gave additional weight to Harman’s hypothesis.

With the discovery that mitochondria were the chief sources of both cellular energy and oxygen radicals, they became a central focus of research on mechanisms of aging and the rate of living theory gradually transformed into the mitochondrial or more generally, the oxidative stress theory of aging (Bokov et al. 2004). Specifically, the oxidative stress theory of aging posits that aging is a consequence of the gradual accumulation of oxidative damage to tissue over time due to the imbalance between oxidant production largely by mitochondria and antioxidant protection (Sohal 1986). In fact, reduced energy flux became a chief suspect for the mechanism underlying the life-extending impact of dietary restriction (DR) in laboratory animals (Weindruch and Walford 1988). The later finding that after a several month period of adjustment DR did not actually reduce mass-specific metabolic rate (Duffy et al. 1989; McCarter and Palmer 1992) led to a long-running debate about whether whole body metabolism or only metabolism in certain key tissues was regulating the rate of aging (Sohal and Weindruch 1996).

As the bioenergetics of mitochondria became better understood, and improved methods for measuring, reactive oxygen species (ROS) production and oxidative damage to macromolecules in cells and tissues developed, it emerged that in laboratory mice both oxidative damage to macromolecules and ROS production progressively increased with age. Notably, mitochondrial DNA which is near the site of most ROS production was particularly susceptible to damage during aging, with large mtDNA deletions accumulating in cells over time (Kowald and Kirkwood 2018). Furthermore, the rate of increase in both were markedly reduced by DR (Sohal and Weindruch 1996). In addition, most long-lived mutant worms and mice were found to be resistant to oxidative stress (Martin et al. 1996; Liang et al. 2003) suggesting they possessed enhanced antioxidant defenses and/or particularly efficient damage repair mechanisms. Even the comparative biology of aging suggested a key role for oxidative stress in species differences in longevity. Longer-lived species regularly exhibited lower rates of ROS production than shorter-lived species (Lambert et al. 2007; Ungvari et al. 2011; Barja 2014) and in some studies lower levels of oxidative tissue damage (Barja and Herrero 2000). By the end of the 20th century, the mitochondrial hypothesis of aging was well entrenched and seemed to explain both cellular mechanisms of damage and broad interspecific longevity patterns.

Methods for investigating the mitochondrial hypothesis

As should be evident from the above discussion, there are multiple observational and experimental approaches to investigate the relationship between mitochondrial function and aging or longevity, not all of them equally informative. For instance, either comparing levels of oxidant production or tissue antioxidant concentration among species provides weak evidence without also assessing actual oxidative tissue damage. Tissue damage is where the rubber meets the road, so to speak, because it is the damage that is presumed to compromise physiological function (Monaghan et al. 2009; Selman et al. 2012). In fact, measuring oxidative tissue damage alone as a function of age provides stronger relevant evidence that measuring either mitochondrial oxidant production or antioxidant levels singly or together. Although it will not be discussed further, a third team of players in the oxidative stress scenario besides oxidant production and antioxidant activity is the macromolecules themselves, which can be more or less susceptible to oxidative damage depending on their composition. For instance, as membrane lipids are composed of more or less saturated lipids they become less or more susceptible to peroxidation, respectively (Hulbert et al. 2007). Similarly in proteins certain amino acids are more susceptible to oxidation than others, a feature that could potentially impact longevity. Accordingly, some bioinformatic analyses of species of varying longevity have reported that the amino acid composition of mitochondrial proteins in particular is highly correlated with species longevity. Specifically, longer-lived species have been reported to be relatively depleted in cysteine (Moosmann and Behl 2008) or methionine (Aledo et al. 2011) than shorter-lived species. These two amino acids are particularly susceptible to oxidation. Finally, some types of oxidative lesions are capable of being repaired. For instance, methionine residues are sensitive to oxidation sulfoxides but these lesions may be enzymatically repaired by methionine sulfoxide reductase (Stadtman et al. 2005). Similarly, a host of oxidation lesions to DNA may be removed and replaced typically by base excision repair (Cooke et al. 2003). So assessing the activity of repair pathways may also be informative. Once again though, the key issue is actual unrepaired tissue damage.

Techniques for quantifying oxidative damage to macromolecules are steadily improving which should allow more refined assays in the future. Each of the assays has its intricacies, however. To take one example, oxidative DNA damage, often measured by assessing 8-oxo-2-deoxyguanosine (oxo8dG) concentration, is highly sensitive to the method of DNA extraction which itself can generate oxidized DNA product. Thus DNA extraction with sodium iodide rather than traditional phenol method reduces nuclear oxo8dG by nearly 100-fold indicating the scale of this common experimental artifact (Hamilton et al. 2001). Similarly, the MDA-TBARS assay for lipid peroxidation is sensitive to sample preparation method as well as being nonspecific compared with a more technically challenging but more accurate and precise metric of isoprostane concentration (Halliwell and Lee 2010; Ho et al. 2013). These important subtleties are often lost on researchers tangentially involved in the oxidative stress field.

One experimental approach—experiments are the test of all knowledge as physicist Richard Feynman once said—is to manipulate longevity such as with DR or genetic mutations such as several dwarfing mutations in mice and assess how that manipulation impacts the balance of oxidant production and antioxidant concentrations or preferably actual oxidative damage to tissues. Unfortunately, although this approach can uncover evidence consistent with the oxidative stress theory, it cannot directly test the theory, as many other cellular and molecular factors are also likely to change with altered lifespan. A more focused and informative approach is to directly manipulate oxidative stress itself, either by altering ROS production or antioxidant defenses or repair processes and making predictions about how those manipulations are expected to affect both oxidative damage to tissues and aging or longevity. ROS production is difficult (but not impossible) to manipulate in a targeted fashion (see below). However, antioxidant defenses can be easily modulated by knockdown or knockout of specific antioxidant genes or by overexpressing them. The impact of these manipulations needs to be validated by noting whether oxidative damage to tissues changes in the predicted manner, but having done so, provides solid evidence for or against the theory.

Initial studies which overexpressed antioxidants in fruit flies provided some compelling supporting evidence. Specifically, ubiquitous overexpression of cytoplasmic superoxide dismutase (SOD1) (Sun and Tower 1999) or even expression of SOD1 in motor neurons alone increased fly longevity and resistance to experimentally imposed oxidative stress (Parkes et al. 1998). Overexpression of the mitochondrial version of SOD (SOD2) also extended fly longevity (Sun et al. 2002). In the hands of other researchers, neither overexpression of SOD1 (which catalyzes the dismutation of superoxide to hydrogen peroxide, another ROS) nor the overexpression of catalase (which like the glutathione peroxidases catalyze the conversion of hydrogen peroxide to oxygen and water) alone affected fly longevity; however overexpressing them both together did extend fly life. It also delayed the age-related loss of physical performance and reduced oxidative damage to proteins (Orr and Sohal 1994).

Challenges to the mitochondrial hypothesis

This simple and satisfying picture, where all evidence seemingly supported the oxidative stress hypothesis of aging, began to come apart early in the 2000s. First, laboratory mice genetically engineered to have reduced expression of SOD2 were found to accumulate higher levels of DNA damage and more cancer as expected, but lived no longer than controls (Van Remmen et al. 2003). Six additional studies that genetically reduced the expression of a range of cellular antioxidants also failed to affect mouse longevity with a single exception—knocking out SOD1, which did shorten life as predicted. Conversely, overexpressing either SODs, catalase, glutathione peroxidase 4, or combinations of these antioxidants indeed increased cellular resistance to oxidative stress as expected, but failed to increase mouse longevity (Huang et al. 2000; Perez et al. 2009). Thus, what was shown to occur at the cellular or tissue level did not necessarily translate to the organismal level. Another blow to the oxidative stress hypothesis was the report that the naked mole-rat (Heterocephalus glaber), which lives roughly 10-fold longer than a similar size laboratory mouse, exhibited significantly higher levels of oxidative damage to a broad range of tissues compared with mice (Andziak et al. 2006).

It would be too strong a statement to claim that all emerging evidence contradicted the mitochondrial hypothesis. There continues to appear evidence consistent with the hypothesis as well. For instance, targeted overexpression of catalase to the mitochondria, where catalase is normally not found, did extend mouse life and health (Schriner et al. 2005). However, as more and more evidence accumulated empirical support of the theory became increasingly uneven.

Inconsistent support for oxidative stress as a contributor to organismal dysfunction was not confined to longevity studies. Oxidative stress has also been hypothesized to play a role as a mediator of other aspects of life history evolution such as reproductive and growth patterns (Costantini 2018). For instance, oxidative stress has been hypothesized to be the mechanistic mediator of the somatic cost of reproduction which of course also impacts longevity. However the results here are also uneven. Specifically, some studies found that increasing reproductive energy expenditure during mammalian pregnancy and lactation was accompanied by increased oxidative damage (Sainz et al. 2000; Fletcher et al. 2013) which was consistent with the theory. However others found either no effect or actual reductions in oxidative damage associated with reproductive energy expenditure (Garratt et al. 2011; Ołdakowski et al. 2012; Schmidt et al. 2014). Thus the role of oxidative stress as a modulator of life history tradeoff also became more questionable.

Aging, longevity, and specific mitochondrial function

The oxidative stress theory itself does not distinguish between the contributions of mitochondrial function, antioxidant defenses, repair or replacement of oxidative lesions, and tissue resistance to overall oxidative stress. As oxidant production itself has been claimed by some to correlate more closely with longevity than other potential contributors to oxidative stress (Barja 2002), a reasonable question to ask is how mitochondrial efficiency itself, defined as the rate of ROS production relative to oxygen consumption or ATP production, affects aging. This question has been previously examined by interspecies comparisons (Barja 1998; Lambert et al. 2007) and manipulation of longevity by DR (Barja 2002), however it has also been evaluated experimentally in model laboratory species.

A particularly thorough examination of mitochondrial function and longevity was performed by Dillin et al. (2002), who inhibited by RNAi various subunits of all the protein complexes in the electron transport chain in both developing and adult C. elegans. They found that inhibiting—from the time of hatching—subunits of complex I (nuo-2), complex III (cyc-1), complex IV (cco-1), and complex V (atp-3) significantly extended mean life from 32% to 87%. Interestingly, these RNAi gene knockdowns extended life even in already long-lived daf-2 (insulin/IGF receptor) mutant and in a daf-16 (FOXO) mutant, which nullifies the longevity effect of daf-2 under normal circumstances. Thus the longevity effect of these genetic mitochondria inhibitors clearly work through a different molecular network than insulin/IGF signaling which has been the focus of most worm longevity studies. In addition to long life, these worms displayed a 40–80% reduction in ATP production, smaller size, slower development, and slower movement rate as well as a reduced rate of food consumption compared with wild-type worms. Treatment with antimycin A, a complex III inhibitor which also typically increases ROS production, extended life and had similar phenotypic effects as RNAi inhibition. ROS production is also typically increased by inhibition of complexes I and III so it might be expected that inhibiting these complexes would shorten life relative to the other complexes, but in fact the longevity effect was roughly the same whether inhibiting complexes I and III or IV and V. Thus ROS production seemingly plays no significant part in these results.

These results at first glance look suspiciously like a “refrigerator effect,” that is a simple slowing of metabolism which has been known to extend life in invertebrates for a century. However, this does not seem to be the case because knocking down the same genes in adults, rather than in developing worms, also reduces ATP production about the same amount, but does not lengthen life, even though adult life lasts many times longer than the developmental period. In sum the results seem inconsistent with the ROS production or oxidative stress hypotheses of longevity, although unfortunately neither ROS production nor oxidative damage to any macromolecules were reported.

In another C. elegans study, Rea et al. (2007) used a novel RNAi dilution approach to further explore the apparent paradox that disrupting mitochondrial function could lengthen life. By using a lifelong RNAi dilution series, these researchers could modulate the degree of suppression of five genes, three of which were the same nuclear-encoded mitochondrial genes as in Dillin et al. (2002) showed extended life, plus two others. They found in all five genes that as normal expression levels were progressively suppressed, multiple respiratory chain complexes were affected, not just the one in which the suppressed subunit played a role. This is not surprising given that mitochondrial “supercomplexes” are now well established (Cogliati et al. 2013). They also found that with progressive gene suppression, worm longevity progressively increased to a point and then decreased again at lower levels of suppression. As longevity increased, larval development slowed, adult size and egg production were reduced, and larval viability declined. Significantly, degree of protein oxidation did not parallel these life history changes. These authors discounted the effect of both total metabolism and ROS production as mechanisms for these findings and instead hypothesize that disruption of normal somatic cell cycle progression during development is responsible for the observed phenotypes.

Similar experiments in fruit flies provided some strikingly similar results although there were some major differences as well (Copeland et al. 2009). In this case, genetically encoded RNAis were used to knockdown more than 50 nuclear encoded respiratory chain subunits. Most of these were lethal or semi-lethal but at least one in each respiratory chain complex extended life. A subset of these was chosen for more thorough examination using at inducible RNAi system. Induced RNAi knockdown of five genes, two subunits of complex I and one each of complexes III, IV, and V, resulted in extended mean longevity by 8–19% in females. Males showed less consistent longevity effects. Note that only one worm sex (hermaphrodites) has been investigated. The longevity effect seen in these flies was much smaller than observed in worms (see above).

This difference between worm and fly results could be due to some unknown differences between the species, although it could also be due to differing degrees of gene product knockdown by the RNAi. In the fly study RNAi only knocked down mRNA levels by between 6% and 47%, depending on the gene. No equivalent measurements were made in the worm study, but C. elegans is well-known to be an excellent responder to RNAi in food with mRNA levels often being reduced by nearly 100% (Rea et al. 2007). Another clear difference between the studies was that none of the five long-lived flies displayed a reduction in ATP levels and in fact one showed higher ATP concentration than controls despite the fact that a subunit of complex I was suppressed. Also, unlike the worm results, when the RNAi effects were induced only in adulthood, two of the five lines still lived longer although the other three did not. Finally, RNAi suppression of one complex I gene in neurons only in adult worms extended life as well. Unfortunately for placing things in context, no measurements of ROS production or oxidative tissue damage were included in this study, although both RNAi induction of both complex I and complex III subunit suppression exhibited increased resistance to exogenous oxidative stress. The other two complexes showed no similar effect. Despite some apparent differences between worms and flies, the key finding in all these studies is that regardless of some differences, some degree of disruption of normal mitochondrial function can extend life in these two invertebrate species.

But could the same be true of mammals with their high energy demands compared with invertebrates? Yes, as it turns out.

The C. elegans clk-1 gene encodes an enzyme required for ubiquinone biosynthesis (Miyadera et al. 2001). Ubiquinone is a cofactor in redox reactions which can serve as a membrane antioxidant but is most notable for its role in mitochondrial respiration where it is involved in shuttling electrons between complex I and complexes II and III. Mutations in clk-1 extend worm longevity by ∼30–50% and pleiotropically slow development, egg-laying, and a range of behaviors (Lakowski and Hekimi 1996) without dramatically reducing metabolic rate (Braeckman et al. 2001). In some studies, clk mutants exhibit low levels of ROS production relative to controls (Kayser et al. 2004) but in others no change or increased ROS production has been reported (Yang et al. 2009). This longevity pathway appears distinct from the mitochondrial respiratory chain inhibition that fails to lengthen life in adult worms (Dillin et al. 2002).

The mouse version of this gene, mclk1, is embryonically lethal when knocked out, but mice heterozygous for the allele are born at the expected frequency. In three different genetic backgrounds, mclk1± mice lived 15–30% longer than controls and showed no reduction in fertility (Liu et al. 2005). DNA damage as measured by COMET assay was also reduced in the liver of these mice.

In sum, apparent genetic disruptions of the normal mitochondrial electron transport chain have been shown to lengthen life in C. elegans, fruit flies, and mice, the three most common model laboratory animals. I say “apparent” because the disrupted genes are encoded in the nucleus and some such as the clk genes are known to have nuclear functions in addition to their mitochondrial roles (Monaghan et al. 2015) and so others may as well.

Is the mitochondrial hypothesis of aging dead?

Given the findings outlined above, in which apparent disruption of mitochondrial function can lengthen life in multiple animal models, a fair question to raise is whether mitochondrial function indeed plays the critical role in organismal health, longevity, and life history evolution that it has been hypothesized to do for decades. In other words, is the mitochondrial hypothesis of aging, a more narrow form of the oxidative stress hypothesis, dead or at least on life support? The current evidence supporting the mitochondrial hypothesis looks increasingly weak and uneven, but is there any reason to harbor a bit of skepticism about the nature of that evidence? Indeed, I think there is.

The first issue has to do with the animals in which the overwhelming majority of the evidence has been generated. Animals in common use in the laboratory have all undergone considerable selection for both known and unknown traits associated with laboratory life. Most so-called wild-type worms than have been maintained in the laboratory for decades, for instance, are shorter-lived than recently collected worms (Gems and Riddle 2000). In addition, it is known that the traditional two-week life cycle of fruit flies in the laboratory has reduced lifespan by about 50% and altered larval competitive ability (Sgrò and Partridge 2000; Linnen et al. 2001). Similarly, laboratory mice compared with wild mice reared in the laboratory are much larger, more rapidly developing, more fecund, and live roughly 20% shorter lives (Miller et al. 2002). Less obvious differences is that most laboratory mice have lost the ability to synthesize pituitary melatonin (Goto et al. 1989) and have longer telomeres than wild mice (Hemann and Greider 2000). On top of inadvertent selection, most laboratory species have been purposely inbred resulting in animals that are homozygous at all genetic loci and are therefore subject to inbreeding depression in a variety of traits (Phelan and Austad 1994). The life history implications of these changes associated by laboratory selection are unclear but some phenomena such as the life-extending impact of DR seemingly disappears in mice recently derived from the wild (Harper et al. 2006).

The second issue has to do the environment in which animals are reared, maintained, and in which experiments are done. The typical laboratory environment is constant and apparently benign. I say “apparently” because what we may think is benign may not be. For instance, laboratory mice are typically maintained at temperatures that are mildly stressful which impacts body composition, immune function, endocrine, and heat shock responses among other things (Messmer et al. 2014; Eng et al. 2015). Visible light has recently been found to reduce worm lifespan (De Magalhaes Filho et al. 2018). The life-extending effects of daf-2 mutations in worms can be life shortening under modestly more realistic conditions such as rearing in sand or soil rather than on agar (Van Voorhies et al. 2005). Laboratory environments lack variation in many variables and environmental complexity is known to have dramatic effect on some normal aging changes as well as some diseases (Berardi et al. 2007; Wood et al. 2010). Laboratory environments also lack unpredictable climatic changes, predators, competitors, erratic food availability, many pathogens, and a host of other features to which evolution has adapted species in nature.

The point of enumerating these potential shortcomings of traditional laboratory practices is not to suggest that they are inevitably misleading, but that it is always preferable if possible to perform experiments under conditions as similar to the conditions under which animals evolved as possible. In the absence of experiments under such conditions, it is best to reserve judgment on the generality of findings based solely on laboratory studies. Certainly there are manifold challenges in critically evaluating hypotheses such as the mitochondrial hypotheses of aging or life history tradeoffs under field conditions (Speakman et al. 2015). However as technical advances are made in assessing mitochondrial function in vivo such as various NMR and optical techniques and it assessing oxidative damage in smaller and smaller samples, we should ultimately be able to determine whether mitochondrial function is as important to the biology of aging and life history tradeoffs as had been originally suspected.

Funding

This work was supported by the National Institutes of Health/National Institute on Aging [P30 AG050886 and R01 AG057434].