Physiological Reviews

Life and Death: Metabolic Rate, Membrane Composition, and Life Span of Animals

A. J. Hulbert, Reinald Pamplona, Rochelle Buffenstein, W. A. Buttemer


Maximum life span differences among animal species exceed life span variation achieved by experimental manipulation by orders of magnitude. The differences in the characteristic maximum life span of species was initially proposed to be due to variation in mass-specific rate of metabolism. This is called the rate-of-living theory of aging and lies at the base of the oxidative-stress theory of aging, currently the most generally accepted explanation of aging. However, the rate-of-living theory of aging while helpful is not completely adequate in explaining the maximum life span. Recently, it has been discovered that the fatty acid composition of cell membranes varies systematically between species, and this underlies the variation in their metabolic rate. When combined with the fact that 1) the products of lipid peroxidation are powerful reactive molecular species, and 2) that fatty acids differ dramatically in their susceptibility to peroxidation, membrane fatty acid composition provides a mechanistic explanation of the variation in maximum life span among animal species. When the connection between metabolic rate and life span was first proposed a century ago, it was not known that membrane composition varies between species. Many of the exceptions to the rate-of-living theory appear explicable when the particular membrane fatty acid composition is considered for each case. Here we review the links between metabolic rate and maximum life span of mammals and birds as well as the linking role of membrane fatty acid composition in determining the maximum life span. The more limited information for ectothermic animals and treatments that extend life span (e.g., caloric restriction) are also reviewed.


Human awareness that species differ in their life span surely predates science. Indeed, the variation in maximum life span potential (MLSP) among different species is substantial, exceeding the variation in longevity achieved by either genetic or physiological manipulation by about four orders of magnitude. This variation was a critical component of an important early analysis of the processes involved in aging and the determination of life span. Almost a century ago, Max Rubner (306) coupled the observation that MLSP of mammals increases with body size with another of his scientific interests, namely, that the mass-specific rate of metabolism of mammals decreases with body size (305). He combined the MLSP and metabolic rates of five mammal species (guinea pigs, cats, dogs, cattle, and horses) to calculate their “lifetime energy potential” (i.e., their mass-specific lifetime resting energy turnover) and found that it was a relatively constant value for these five mammal species. Curiously, he did not include humans in his comparison (which will be discussed later). Two decades later, this perspective was elaborated and extended by Raymond Pearl (284), who proposed that life span differences of Drosophila maintained at different temperatures could be explained by differences in metabolic rate. Pearl (284) gave the rate-of-living theory of aging its name.

Later, Harman (127) proposed the free radical theory of aging, which provided a possible mechanistic link between metabolic rate and aging. Harman (127) proposed that normal oxygen consumption inevitably results in the production of oxygen free radicals which, in turn, damage important biological molecules (nucleic acids, proteins, and lipids, among others), resulting in the process of “aging,” and eventually this accumulated damage results in death of the organism. Over the last half century, the free radical theory has developed into the oxidative-stress theory of aging which takes into account that 1) not all the damaging reactive oxygen molecules (or “species” in the chemical sense hence the abbreviation; ROS) produced by normal metabolism are free radicals, and that organisms have both 2) antioxidant defenses to minimize ROS and 3) mechanisms to repair or remove biological molecules damaged by ROS. Currently, the oxidative-stress theory is the most widely accepted mechanistic explanation of aging and life span, and there is much evidence to support it. In the past, the oxidative-stress and rate-of-living theories have been reconciled by supposing that higher levels of ROS are generated at higher levels of metabolic activity (e.g., Refs. 22, 328). However, this popular concept is flawed, because while it appears true in some situations (e.g., when broadly comparing mammals of different size), it is not true in others (e.g., between different states of mitochondrial respiration, following calorie restriction, or in bird-mammal comparisons).

Furthermore, although there is a broad correlation between body size and MLSP, there are a number of problems associated with presuming a linkage between rate-of-living and MLSP. For example, 1) voluntary exercise and its associated increase in metabolic rate does not shorten life span of rats (147) or humans (200); 2) there is no inverse relationship between life span and mass-specific metabolic rates of individuals in populations of mice (331) or fruit flies (157); 3) although caloric restriction extends life span, it does not do so by reducing metabolic rate (MR) (see later sections); 4) long-living insulin/insulin-like growth factor (IGF) mutant fruit flies do not have a decreased metabolic rate (157); 5) although birds and mammals have similar rates of living, birds are generally much longer living than similar-sized mammals; and 6) within both groups of endotherms, there are significant species differences in MLSP that cannot be explained by metabolic rate differences (see later sections). Collectively these problems show that part of the scientific jigsaw puzzle that explains aging and maximum life span is missing. In this contribution, we review the evidence that membrane fatty acid composition is this missing piece of the puzzle.

A century ago, mass-specific MR (and its physiological correlates, such as heart rate) were the obvious properties to be related to life span. There have been a variety of attempted explanations for the size-dependent differences in metabolic rate over the years, and differences in cellular composition were initially excluded (see Ref. 176). However, recently it has been shown that one aspect of cell composition, namely, membrane fatty acid composition, does vary in a systematic manner with body size in mammals (69, 164) and in birds (37, 162). It has also been shown that membrane fatty acid composition is related to MLSP (266, 271, 272, 275277). These observations, when combined with the long-known differences in susceptibility of different fatty acids to peroxidative damage (148), were seminal in the development of the “homeoviscous-longevity” membrane theory (267) and the later “membrane-pacemaker” theory of aging (155). These modifications of the oxidative-stress theory of aging emphasize the effects of membrane fatty acid composition on lipid peroxidation (and its products) on the process of aging and how this might determine the distinctive maximum life span of a species.

In this contribution, we concentrate on the comparison of the maximal longevity (MLSP) distinctive to different species. We first discuss the mechanisms associated with aging (based on the oxidative-stress theory) and briefly describe the means by which the fatty acid composition of membranes might be important in the processes of aging. We then examine the links between body mass, metabolic rate, and maximum longevity separately for the two classes of endothermic vertebrates, mammals and birds, with an emphasis on how membrane fatty acid composition might account for variation in MLSP among species. We briefly review the limited data on this topic for ectothermic vertebrates and invertebrates. Finally, we consider how changes in membrane composition might explain how some experimental treatments can extend longevity. We restrict ourselves largely to a physiological context, paying limited attention to the genetic manipulations involved in much recent aging research.


The biological aging process is characterized by the progressive decline in the efficiency of physiological functions. The ability to maintain homeostasis is correspondingly attenuated with age, leading eventually to increased risk of developing many diseases (including cancer and neurodegenerative and cardiovascular diseases) and increasing the chance of death. For a long time aging was believed to be restricted to multicellular animals; however, recently detailed measurements have shown it also to occur in the common bacteria Escherichia coli (340). The natural aging process has four characteristics: it is progressive, endogenous, irreversible, and deleterious for the individual (344). The progressive character of aging suggests that causes of aging are present during an organism's whole life span, both at young and old ages. That aging is an endogenous process suggests exogenous factors are not causes of the intrinsic aging process, although they may interact with endogenous causes and either enhance or diminish their effects. The endogenous character of aging suggests that the rate of aging and MLSP of different animal species are predominantly genotypically determined and explain why different animal species can age at substantially different rates in similar environments. Available evidence consistently points to the idea that ROS, mainly those of mitochondrial origin (129, 235), are causally related to the basic aging process (17, 22, 312, 328). Accordingly, mitochondrial ROS production fulfils the four characteristics of aging described above: ROS are endogenously produced by mitochondria under normal physiological conditions, they are produced continuously throughout life and can thus lead to progressive aging changes, and their deleterious effects on biological macromolecules may inflict irreversible damage during aging, especially in postmitotic tissues because cells that are irreversibly damaged or lost cannot be replaced by mitosis. There have been many theories and explanations of aging, and while it has not yet been “proven,” currently the strongest candidate is the oxidative-stress theory (also known as the oxidative-damage theory), and this is the basis of the following description.

A. Mitochondrial Free Radical Production

That oxygen can be toxic, and consequently the beginnings of the oxidative-stress theory of aging, dates back to the late 19th century. Paul Bert (25) described the toxic effects of air at high pressure, as well as oxygen alone at high concentration, and several subsequent studies showed that oxygen toxicity was particularly acute in endotherms. In ectotherms its toxicity increased with ambient temperature, which suggested a relationship with metabolic rate rather than oxygen solubility. These studies demonstrated that three organ systems (lungs, retina, and nervous system) are especially sensitive to high oxygen tensions in whole animals (195); however, a scientific explanation of the cause of these harmful reactions was lacking for many decades.

After Fenton (100) discovered that Fe(II) catalyzed the oxidation of tartaric acid by hydrogen peroxide, it was proposed that hydroxyl radicals, hydrogen peroxide, and superoxide anions undergo a chain reaction that results in a net conversion of hydrogen peroxide into water (119, 120). This became known as the Haber-Weiss reaction and is now believed to be at the core of much ROS generation in cells. However, the connection between oxygen toxicity and ROS generation was not then understood (27). Using electron spin resonance, Commoner et al. (67) published the first direct evidence that free radicals were produced in living cells and found that free radical levels were higher in tissues that were more metabolically active. The free radical hypothesis of oxygen toxicity was first proposed by Gerschman et al. (109), taking into account the strong similarity of the histological alterations produced by hyperoxia and by ionizing radiation (264). Harman (127) postulated that even at normoxia, oxygen utilization causes subtle tissue damage due to production of free radicals. He integrated evidence from the electron spin resonance study, post-Hiroshima studies of radiation damage, Fenton's findings, and contemporary theories that proposed free radical mechanisms for the oxidation of organic compounds and dismutation of hydrogen peroxide by iron salts (358). Harman proposed that physiological iron and other metals would cause ROS to form in the cell via Haber-Weiss chemistry as a by-product of normal redox reactions. The ROS would damage nearby structures like mitochondrial DNA (129). He predicted that administering compounds that are easily oxidized, such as cysteine, would slow down the aging process, a suggestion recently reviewed by Dröge (83). Harman's proposal constitutes the basis of the free radical theory of aging, which is supported by a large body of scientific evidence. The discovery of the superoxide dismutase enzyme in living organisms (226) led to serious consideration of the role of free radicals in biology. If an enzyme that decomposed oxygen radicals inside the body was highly conserved during evolution, oxygen free radicals were likely important in biological systems.

A free radical is any molecule capable of independent existence that contains one or more unpaired electrons (124). ROS is a collective term that includes oxygen radicals as well as nonradical derivatives of oxygen (124). Reduction of univalent oxygen generates three main ROS: superoxide and hydroxyl radicals as well as the nonradical hydrogen peroxide. Among these, the hydroxyl radical is extremely reactive. It can be generated by the combination of superoxide radical and H2O2 in the presence of trace amounts of iron (or copper) during the Fenton-Haber-Weiss reaction. Thus H2O2, although not a free radical itself, can behave as a Trojan horse, diffusing away from sites of its production to generate the hydroxyl and other reactive radicals at other cellular locations, hereby propagating oxidative damage. In many scientific papers, use of the term ROS is restricted to what we might describe as the three primary ROS described above (especially superoxide and hydrogen peroxide); however, there are a number of other important ROS, both radicals and nonradicals (see Table 1). The roles of many of these other ROS in aging have not been fully evaluated. There is also some overlap between ROS and reactive nitrogen species (RNS) such as nitric oxide. In the case of mitochondria, nitric oxide production is much smaller than superoxide production. However, nitric oxide can still be important due to interaction with superoxide and other radicals to produce reactive species like peroxynitrite (76), which can modify many kinds of macromolecules and possibly contribute to various pathologies (65).

View this table:

Important reactive oxygen species and lipid aldehydes and their common abbreviations

ROS can be generated at various sites and under a variety of conditions. These include the membrane NADPH oxidase of polymorphonuclear leukocytes, during ischemia-reperfusion, as well as other enzymatic reactions (e.g., lipoxygenases, cyclooxygenases, peroxidases, and other heme proteins), the enzyme xanthine oxidase, peroxisomes, or the hepatic P-450 microsomal detoxifying system (124). In healthy tissues under normal conditions, most ROS are generated by the mitochondrial respiratory chain (312).

The finding that the percentage of total electron flow in mitochondria directed to radical generation is not constant in different tissues, and under different conditions, suggests that ROS generation is not a constant stoichiometric by-product of mitochondrial respiration (17, 19, 113, 136), as is frequently assumed. Oxygen radical generation at the respiratory chain has been classically attributed to complex III (32). However, in agreement with early information obtained in submitochondrial particles (347), complex I is also an important ROS generator in intact functional heart and brain mitochondria from rats (20, 108, 136, 137), as well as other organs and other species (17, 184). Nowadays the concept that both complex I and complex III produce ROS is widely accepted in the scientific literature and is already part of mainstream biochemical knowledge (e.g., Ref. 253). There is, however, current debate concerning the identity of the ROS generator inside complex I. Some studies have suggested a role for flavin mononucleotide (184, 206, 380), while others favor complex I linked ubisemiquinones (191, 256). However, other studies localize the ROS generator in the electron pathway inside complex I in a place between the ferricyanide reduction site and the rotenone-binding site (108, 139, 186). This finding discards flavin and suggests that the source of ROS might be the complex I FeS clusters situated in that region (139), although a role for complex I-linked ubisemiquinones cannot be ruled out at present. Because all FeS clusters of complex I are situated in the hydrophilic matrix domain of the complex, ROS arising from them will damage targets situated in the mitochondrial matrix. In contrast, complex III ROS generation seems to be directed to the cytosolic side (342).

Apart from the degree of electron flow, other possible physiological mechanisms influencing the rate of mitochondrial ROS generation include 1) a decrease in the relative concentration of the respiratory complex(es) responsible for ROS generation; 2) the degree of electronic reduction of these generators; the more reduced, the higher their rate of ROS production (20, 113, 114, 208, 310); 3) uncoupling proteins (discussed in the next paragraph); 4) as well as chemical modification (e.g., glutathionylation) of mitochondrial ROS generators. The nonenzymatic glutathionylation of complex I increases its superoxide production, and when the mixed disulfides are reduced, superoxide production returns to basal levels (125, 351). Oxidation of the glutathione pool within intact mitochondria to glutathione disulfide also leads to glutathionylation of complex I, which correlates with increased superoxide formation. In this case, most of the superoxide is converted to hydrogen peroxide, which can then diffuse into the cytosol. This mechanism of reversible mitochondrial ROS production suggests how mitochondria might regulate redox signaling and shows that oxidation of the mitochondrial glutathione pool could contribute to the pathological changes that occur to mitochondria during oxidative stress (125, 351).

Mitochondrial superoxide production is very sensitive to the proton-motive force across the inner membrane, so it can be strongly decreased by mild uncoupling. For example, one study (239) has shown that a 10-mV decrease in mitochondrial membrane potential is associated with 70% decrease in superoxide production. The decrease in ROS production by isolated mitochondria between state 4 and state 3 is likely due to the decrease in mitochondrial membrane potential between these two states. Similarly, it has been proposed that an ancestral function of uncoupling proteins (UCP) is to cause mild uncoupling and so diminish mitochondrial superoxide production, hence protecting against oxidative damage at the expense of a small loss of energy (38). For example, UCP3 (found in skeletal muscle mitochondria) is activated by increased superoxide production, and the consequent increased proton conductance will cause a mild decrease in proton-motive force and consequently a diminished superoxide production (88). The role of UCP3 activation during exercise and during the transition from state 4 to state 3 in muscle mitochondria, and its control on ROS production is one possible reason why increased voluntary exercise of mammals is not associated with a decrease in life span. It has been proposed that this negative-feedback loop will protect cells from ROS-induced damage and might represent the ancestral function of all UCPs (38, 249).

Other possible factors that modulate complex I activity are cardiolipin content (280) and S-nitrosation (45). Cardiolipin, a phospholipid of unusual structure localized almost exclusively within the inner mitochondrial membrane, is particularly rich in unsaturated fatty acids. This phospholipid plays an important role in mitochondrial bioenergetics by influencing the activity of key mitochondrial inner membrane proteins, including several anion carriers and electron transport complexes I, III, and IV (143). Mitochondrial cardiolipin molecules are possible targets of oxygen free radical attack, due to their high content of polyunsaturated fatty acids and because of their location in the inner mitochondrial membrane near the site of ROS production. In this regard, it has been recently demonstrated that mitochondrial-mediated ROS generation affects the activity of complex I (280), as well as complexes III and IV (278, 279), via peroxidation of cardiolipin following oxyradical attack to its fatty acid constituents. These findings might explain the decline in respiratory chain complexes observed in mitochondria isolated from aged animals and in pathophysiological conditions that are characterized by an increase in the basal rate of the ROS production.

In addition to cardiolipin, S-nitrosation may also serve as another regulatory system for complex I. Nitric oxide is a pleiotropic signaling molecule, with many of its effects on cell function being elicited at the level of the mitochondrion. One of the many cellular reactions of nitric oxide is S-nitrosation of protein thiols, resulting in the generation of S-nitrosothiols. There are several reasons why S-nitrosation may be an important mitochondrial regulatory mechanism. Mitochondria contain sizable thiol pools, are abundant in transition metals, and have an internal alkaline pH, all of which are known to modulate S-nitrosothiol biochemistry (102). In addition, mitochondria are highly membranous, sequester lipophilic molecules such as nitric oxide, and the formation of the putative S-nitrosating intermediate N2O3 is enhanced within membranes (42). Thus the mitochondrial respiratory chain embedded within the inner membrane would seem to be an ideal target for S-nitrosation. In this scenario, it has been described that mitochondria interact with nitric oxide at several levels, and one particularly well-characterized example is the inhibition of complex I and IV (45, 265).

Although free radicals and derived reactive molecular species can have damaging effects, at moderate concentrations they are often important signaling molecules involved in the regulation of several physiological functions (e.g., control of ventilation, erythropoietin production, immune responses, apoptosis). In such situations, ROS are typically generated by tightly regulated enzymes such as NAD(P)H oxidases (82).

B. Cellular Protection: Antioxidant Defenses

Aerobic life demands antioxidant defenses. An antioxidant is extremely difficult to define, but a broad definition is “any substance that, when present at low concentrations compared with those of an oxidizable substrate, significantly delays or prevents oxidation of that substrate” (124). An antioxidant reacts with and neutralizes an oxidant, or regenerates other molecules capable of reacting with the oxidant. Oxidative damage is a broad term used to cover the attack upon biological molecules by free radicals. Cellular protection against oxidative damage includes both elimination of ROS and repair of damage with antioxidants constituting the first line of this defense.

Direct ROS scavenging antioxidant enzymes include superoxide dismutase (SOD), glutathione peroxidase (GPX), and catalase (CAT). SOD converts the superoxide radical to oxygen and hydrogen peroxide. There are different types of SOD in different cellular compartments: Mn-SOD in the mitochondrial matrix, a Cu/Zn-SOD in the cytosol and the intermembrane space, and a different Cu/Zn-SOD in the extracellular compartment. Although SOD eliminates superoxide radicals, it is not a complete antioxidant defense because it produces H2O2. CAT and GPX are two different but kinetically complementary enzymes that can eliminate H2O2. CAT is one of the most active enzymes known and removes H2O2 by breaking it down directly to H2O and O2. However, it has a low affinity for H2O2; thus it needs high H2O2 concentrations to function effectively. Conversely, GPX enzymes (both the selenium- and non-selenium-dependent forms) are most functional at low hydrogen peroxide concentrations, since they have high affinity and low rates of catalysis. GPX(s) have the same locations as SOD, suggesting that GPX is the main enzyme that deals with H2O2 produced by Cu/Zn-SOD in the cytosol and Mn-SOD in the mitochondria (124).

As well as antioxidant enzymes, there are different kinds of endogenous nonenzymatic scavenger molecules that cooperate to limit cellular oxidative stress. After reacting with ROS, these molecules are oxidized and thus must be reduced to regain their antioxidant capacity. Their low molecular mass allows them to remove ROS at sites that larger antioxidant enzymes cannot access. Glutathione (GSH) and ascorbate are two major nonenzymatic antioxidants of the hydrophilic cellular compartments. The antioxidant activity of GSH resides in the reduced thiol group of its cysteine residue. Glutathione can react directly with ROS or can act as a cosubstrate of GPX enzymes. Oxidized glutathione (GSSG) also plays an important role in the regulation of protein function, as mentioned above for mitochondrial complex I. Glutathione reductase is the enzyme responsible for reducing GSSG back to GSH using NADPH (124).

Another thiol-related redox-active substance is the protein thioredoxin (238). Thioredoxin has a redox-active disulfide/dithiol at the active site that regulates transcription factors (such as nuclear factor κB and AP-1). It is induced by various oxidative stresses and is translocated to the nucleus. Thioredoxin is cytoprotective against oxidative stress by scavenging ROS in cooperation with peroxiredoxin/thioredoxin-dependent peroxidase (238).

Apart from glutathione, ascorbate (vitamin C) is the next most abundant reduced nonenzymatic antioxidant inside cells. It is endogenously synthesized in most vertebrates (although not in human beings, fruit bats, or guinea pigs), and it is maintained at levels as high as 1 mM in tissues. After reacting with ROS, the oxidized form of ascorbate must be reduced by NADPH-, GSH-, or NADH-dependent reductases to regain antioxidant capacity (14, 124).

Tocopherols and carotenoids are the main radical scavenger antioxidants that act in lipophilic environments of cells. The major scavenger inside membranes is d-α-tocopherol (vitamin E). Most membranes are thought to contain approximately 1 tocopherol molecule per 1,000 lipid molecules (369). Vitamin E acts on lipid peroxyl groups inside membrane bilayers, reducing them to hydroperoxides, and thus inhibiting the propagation of the peroxidative chain reaction. It breaks the chain reaction of lipid peroxidation but is itself converted to a radical during the process. Vitamin E also reduces lipid alcoxyl radicals to lipid alcohols. Oxidized vitamin E can be recycled back to its reduced form by ascorbate or ubiquinone (coenzyme Q). Carotenoids quench singlet oxygen and interact with other ROS at physiological tissue oxygen partial pressures. Ubiquinol, the reduced form of coenzyme Q, is an important antioxidant. It is a hydroquinone that is synthesized and present in all cellular membranes (15), and its antioxidant activity is exhibited through scavenging of lipids radicals or reduction of vitamin E radical. Regeneration of coenzyme Q is performed by reductases that use NADPH or NADH as cofactors. Methionine residues in proteins (see sect. iiC) can also be considered as scavengers of oxidants.

C. Cellular Protection: Repair Mechanisms

Despite the plethora of antioxidant defense systems, oxidative damage still occurs in vivo. The next level of cellular protection against oxidative stress is the repair/removal of oxidatively damaged macromolecules. As with antioxidant defenses, space limitations restrict our coverage here. The reader is directed elsewhere (e.g., Ref. 124) for more detailed presentation.

To avoid the gradual accumulation of oxidative DNA base lesions and thus maintain genomic stability, DNA damage is repaired by a number of mechanisms. In general, oxidized DNA bases are removed by base excision mechanisms, and although these correcting mechanisms show high fidelity, DNA lesions still accumulate with age. Such accumulation is most dramatic in mitochondrial DNA (312), and less is known about mitochondrial than nuclear DNA repair (31). DNA polymerase enzymes have proofreading and error-correcting facilities, and when this capacity of mitochondrial DNA polymerase was removed in mice, there was a three- to fivefold increase in mitochondrial mutations and a corresponding decreased life span (185, 357).

Damaged proteins can either be repaired or removed. For example, methionine residues of proteins are among the amino acids most susceptible to oxidation by ROS, and the sensitivity of proteins to oxidative stress increases as a function of their number of methionine residues (337). Oxidation of methionine residues generates methionine sulfoxide in proteins, which deprives them of their function as methyl donors, and may lead to loss of their biological activity. Methionine sulfoxide can be repaired by the enzyme methionine sulfoxide reductase, which uses the reducing power of thioredoxin (248). In this context, it is very illustrative that knocking out methionine sulfoxide reductase A lowers MLSP and increases protein carbonyls and sensitivity to hyperoxia in mice (248). The opposite manipulation, overexpression of methionine sulfoxide reductase A, increases life span and delays the aging process in Drosophila (304). The oxidized form of thioredoxin, produced during the reduction of methionine sulfoxide, can be converted back to reduced thioredoxin by the enzyme thioredoxin reductase. In agreement with the methionine oxidation hypothesis of aging, it has been reported that overexpression of thioredoxin reductase increases longevity in mice (238). However, protein repair mechanisms are generally limited, and most oxidized proteins are removed by proteolysis either via lysosomal degradation or via the proteasome. The proteasome complex recognizes amino acid residues that are exposed following the oxidative rearrangement of secondary and tertiary protein structures (117).

Protection of membranes is achieved by a more complex system that involves three mechanisms: lipid repair, lipid replacement, and scavenging of lipoperoxidation-derived end products. The action of chain-breaking antioxidants can result in the production of lipid hydroperoxides. Some of these are metabolized by GPX antioxidant enzymes, which act on H2O2 and also on free fatty acid hydroperoxides, reducing them to fatty acid alcohols. However, the peroxidized fatty acids must be first released from the membrane lipids. A phospholipid hydroperoxide glutathione peroxidase (PHGPX) has also been described in some mammalian tissues and is capable of acting on peroxidized fatty acid chains still esterified to membrane lipids and is thus a very important antioxidant defense for membranes in situ (167). An important mechanism for removing peroxidized lipids is membrane lipid remodeling. While the enzyme PHGPX is important in removing peroxidized acyl chains from phospholipids, other enzymes are also involved in the continual deacylation/reacylation of phospholipids, and this turnover of membrane acyl chains is very rapid. Phospholipase A2 is an important enzyme for removing acyl chains from phospholipids while acyltransferase and transacylase enzymes are responsible for reacylation of phospholipids (98). When isolated rat liver cells are subjected to oxidative stress, lipid peroxidation is increased; however, there is no change in either the fatty acid composition of or in the rate of acyl turnover of membrane phospholipids. There is, however, a decrease in the polyunsaturated fatty acid (PUFA) content of cellular triacylglycerols. It is suggested that rapid constitutive recycling of membrane phospholipids rather than selective in situ repair is responsible for eliminating peroxidized phospholipids, with triacylglycerols providing a dynamic pool of undamaged PUFA for phospholipid resynthesis (110). Finally, lipid peroxidation-derived end products resulting from membrane lipid oxidative damage can be detoxified by glutathione S-transferases (GST; Ref. 348).

D. Membrane Composition and Lipid Peroxidation

Oxygen radicals can attack many cellular molecules. In addition to protein and DNA modification (reviewed in Ref. 312), damage to membrane lipids is also very relevant for life span determination. The susceptibility of membrane lipids to oxidative damage is related to two traits. The first is that oxygen and many radical species are several times more soluble in lipid membrane bilayers than in the aqueous solution (246). The second property is related to the fact that not all fatty acid chains are equally susceptible to damage. It is this second property that is the key to the link between membrane composition and oxidative damage to membranes. The carbon atoms that are most susceptible to radical attack are the single-bonded carbons between the double-bonded carbons of the acyl chains (124). The hydrogen atoms attached to these carbons are called bis-allylic hydrogens. This means that saturated and monounsaturated fatty acyl chains (SFA and MUFA, respectively) are essentially resistant to peroxidation while PUFA are damaged. Furthermore, the greater the degree of polyunsaturation of PUFA, the more prone it is to peroxidative damage. Holman (148) empirically determined (by measurement of oxygen consumption) the relative susceptibilities of the different acyl chains (see Fig. 1). Docosahexaenoic acid (DHA), the highly polyunsaturated omega-3 PUFA with six double bonds, is extremely susceptible to peroxidative attack and is eight times more prone to peroxidation than linoleic acid (LA), which has only two double bonds. DHA is 320 times more susceptible to peroxidation than the monounsaturated oleic acid (OA) (148). With the combination of the relative susceptibilities of different fatty acids with the fatty acid composition of membrane lipids, it is possible to calculate a peroxidation index (a measure of the susceptibility to peroxidation) for any particular membrane.1 The peroxidation index of a membrane is not the same as its unsaturation index (sometimes also called its “double bond index”), which is a measure of the density of double bonds in the membrane. For example, a membrane bilayer consisting solely of MUFA will have an unsaturation index of 100 and a peroxidation index of 2.5, while a membrane bilayer consisting of 95% SFA and 5% DHA will have an unsaturation index of 30 and a peroxidation index of 40. This means that although the 5% DHA-containing membrane has only 30% the density of double bonds of the monounsaturated bilayer, it is 16 times more susceptible to peroxidative damage.

FIG. 1.

The relative susceptibilities of selected fatty acids to peroxidation. Values are from Homan (148), and all were empirically determined as rates of oxygen consumption. They are expressed relative to the rate for linoleic acid 18:2 n-6 which is arbitrarily given a value of 1. SFA, saturated fatty acids; MUFA, monounsaturated fatty acids; n-6 PUFA, omega-6 polyunsaturated fatty acids; n-3 PUFA, omega-3 polyunsaturated fatty acids. The number for each fatty acid refers to acyl chain length:number of double bonds.

Reactive free radicals remove the bis-allylic hydrogen atoms from PUFA chains, and the carbon from which H· is abstracted now has an unpaired electron (i.e., it is also a free radical). When such C· radicals are generated in the hydrophobic interior of membranes, a likely fate is combination with oxygen dissolved in the membrane. The resulting peroxyl radical is highly reactive: it can attack membrane proteins and can also oxidize adjacent PUFA chains. Thus the initial reaction is repeated and a free radical chain reaction is propagated. Unless quenched by antioxidants, lipid peroxidation is a self-propagating autocatalytic process producing several potent ROS. It can also generate lipid hydroperoxides (124, 335, 336), which are more hydrophilic than unperoxidized fatty acyl chains, and these can thus disrupt the membrane structure, altering fluidity and other functional properties of membranes.

The hydroperoxides and endoperoxides, generated by lipid peroxidation, can undergo fragmentation to produce a broad range of reactive intermediates, such as alkanals, alkenals, hydroxyalkenals, glyoxal, and malondialdehyde (MDA; Ref. 95) (see Fig. 2). These carbonyl compounds (collectively described as “propagators” in Fig. 2) have unique properties contrasted with free radicals. For instance, compared with ROS or RNS, reactive aldehydes have a much longer half-life (i.e., minutes instead of the microseconds-nanoseconds characteristic of most free radicals). Furthermore, the noncharged structure of aldehydes allows them to migrate with relative ease through hydrophobic membranes and hydrophilic cytosolic media, thereby extending the migration distance far from the production site. On the basis of these features alone, these carbonyl compounds can be more destructive than free radicals and may have far-reaching damaging effects on target sites both within and outside membranes. These carbonyl compounds, and possibly their peroxide precursors, react with nucleophilic groups in proteins, resulting in their modification. The modification of amino acids in proteins by products of lipid peroxidation results in the chemical, nonenzymatic formation of a variety of adducts and cross-links collectively named advanced lipoxidation end products (ALEs; Ref. 353). These can be useful indicators of lipoxidative stress in vivo (Fig. 2).

FIG. 2.

Schematic chemical pathway for the formation of advanced lipoxidation end products detected in tissue proteins. All of the compounds are derivatives of lysine and arginine residues in protein. Similar compounds are formed on reaction with other nucleophilic groups in proteins, as well as on DNA and aminophospholipids. For propagators, general structures are shown for reactive aldehydes (A) and lipid peroxidation-specific aldehydes (B). For end products, general structures are shown for advanced lipoxidation end products (A) and advanced glyco- and lipoxidation end products (B).

Lipid peroxidation-derived end products (“enals”) can also react at the exocyclic amino groups of deoxyguanosine, deoxyadenosine, and deoxycytosine to form various alkylated products (218). Some common enals that cause DNA damage (analogously to protein damage) are MDA, acrolein, and 4-hydroxynonenal, among others. The most common adducts arising from enals are exocyclic adducts such as etheno adducts, and malondialdehyde-deoxyguanosine (M1dG). These DNA damage markers are mutagenic and carcinogenic, with powerful effects on signal transduction pathways (217). Furthermore, they 1) are present in the genome of healthy humans, and other animal species, at biologically significant levels (similar or even higher than oxidation markers sensu stricto) (55), 2) are efficient inducers of mutations frequently detected in oncogenes or tumor suppressor genes from human tumors (254), 3) show increased levels in aged animals (55), 4) can be repaired by nucleotide excision repair systems and metabolized by oxidative pathways (262), 5) correlate with alterations in cell cycle control and gene expression in cultured cells (169), and 6) increase nearly 20-fold with a high-PUFA diet (97).

The amino group of aminophospholipids can also react with carbonyl compounds and initiate some of the reactions occurring in proteins, thus expanding the negative biological effects of such nonenzymatic modification (291). In this context, biological processes involving aminophospholipids can potentially be affected by this process. Among these, the following may be highlighted: asymmetrical distribution of aminophospholipids in different membranes, translocation between and lateral diffusion in the membrane, membrane physical properties; biosynthesis and turnover of membrane phospholipids, and activity of membrane-bound proteins that require aminophospholipids for their function (291).

Thus lipid peroxidation should not be perceived solely in a “damage to lipids” scenario, but should also be considered as a significant endogenous source of damage to other cellular macromolecules, such as proteins and DNA (including mutations). In this way, variation in membrane fatty acid composition, by influencing lipid peroxidation, can have significant effects on oxidative damage to many and varied cellular macromolecules. For example, peroxidized cardiolipin in the mitochondrial membrane can inactivate cytochrome oxidase by mechanisms both similar to hydrogen peroxide and also mechanisms unique to organic hydroperoxides (251). Such an effect could be very detrimental if it results in a greater degree of reduction of other respiratory chain intermediates and consequently increases production of superoxide. Figure 3 depicts the mitochondrial processes producing ROS, the effects of ROS on lipid peroxidation, and some of the consequent lipoxidative damage to mitochondrial DNA and proteins.

FIG. 3.

Schematic diagram of mitochondrial processes that are important for aging. The schematic shows that mitochondrial complex I and complex III are the main free radical generators. Several physiological mechanisms influencing the rate of mitochondrial ROS generation include 1) the relative concentration of the respiratory complexes, 2) the degree of electronic reduction of these generators, 3) the uncoupling proteins, 4) the cardiolipin content, and 5) specific chemical modifications. Oxygen radicals attack lipids, carbohydrates, proteins, and DNA. The products of lipid peroxidation include highly reactive molecules that can cause lipoxidative damage to mitochondrial DNA and proteins.

As mentioned above, peroxidation of the PUFA chains of phospholipids generates a complex mixture of short-chain aldehydes. Initially, these aldehydes were believed to produce only “cytotoxic” effects associated with oxidative stress, but evidence is increasing that these compounds can also have specific signaling roles during normal function. For example, the superoxide-initiated activation of proton conductance by UCP3, mentioned previously, has been shown to be mediated by the lipid peroxidation product 4-hydroxy-trans-2-nonenal (HNE) and its homologs (88). The role of ROS and other related molecules in normal physiological regulatory processes is extensively reviewed elsewhere (82).

As with many biochemical processes that can be measured in vitro, it is very difficult to assess the rate of lipid peroxidation in vivo, and consequently to be able to compare rates of lipid peroxidation either between species or between treatments. One method is to measure the exhalation rate of hydrocarbons produced from lipid peroxidation. Ethane is one of the end products of n-3 PUFA peroxidation, while pentane is produced from peroxidation of n-6 PUFA. Both are volatile and excreted from the body via respiration, and the rate of their exhalation has been used to assess the in vivo rate of lipid peroxidation (178).

Unsaturated fatty acids are also responsible for the “fluid” nature of functioning biological membranes. It has been shown that it is the introduction of the first double bond to an acyl chain that is primarily responsible for its fluidity at normal physiological temperatures, and introduction of more and more double bonds to the acyl chain has relatively little additional effect on membrane fluidity (39). Thus substitution of fatty acids with four or six double bonds with those having only two (or sometimes three) double bonds will strongly decrease the susceptibility to lipid peroxidation while maintaining membrane fluidity. This phenomenon may be helpful in longevous animals, and in view of membrane acclimation to low temperature in ectotherms, has been called homeoviscous longevity adaptation (267).

The importance of membrane fatty acid composition for understanding aging (described above) and its addition to the standard scenario describing the oxidative-stress theory of aging is illustrated in Figure 4. It emphasizes that membrane fatty acid composition influences 1) metabolic rate (via effects on the physical properties of membrane bilayers) and 2) the degree of oxidative stress and damage to cellular molecules (via the peroxidative susceptibility of fatty acyl chains).

FIG. 4.

Schematic diagram outlining the modification of the oxidative-stress theory of aging emphasizing the importance of the fatty acid composition of membranes in the determination of maximum life span. Two different examples are presented (low membrane polyunsaturation and high membrane polyunsaturation). The thickness of the arrows in each example represents the relative intensity of the process.

In the following sections we examine the links between metabolic rate and membrane composition in the maximum longevity of mammals and birds. It is not possible to seriously examine these relationships without first taking into account the influence of body size on these processes.


As noted in section i, the relation between both maximum life span and metabolic rate and body size of different mammal species was important in the development of the rate-of-living theory and consequently other later insights into the mechanisms of aging. For a brief account of the history of this subject preceding Rubner's seminal contribution (306), the reader is referred to Speakman (330).

A. Metabolic Rate and Body Size of Mammals

Body size is an important species characteristic that influences almost every aspect of organismal biology (49). As species increase in size, they obviously require more energy, but metabolic rate does not increase in direct proportion to increases in body mass. This is largely because of the constraints of geometry. For every doubling in size, the surface area of an object increases by only 59% and not by 100% (i.e., surface area is related to mass0.67). The basal metabolic rate (BMR) of mammals similarly scales allometrically with body mass, and depending on the particular study, BMR has been reported to only increase by 59–69% with every doubling of body mass (i.e., it is proportional to mass0.67−0.76; Refs. 176, 371). The constraints imposed by such surface limitations are manifest in the following calculation. If a mouse increased in size to that of a horse and its BMR increased in direct proportion to the increase in body mass, the horse-sized mouse would need a surface temperature of ∼100°C to rid itself of the heat produced by its BMR (134). Obviously if mammals, ranging in size from shrews to elephants, are to maintain essentially the same body temperature, then surface constraints (for heat loss, nutrient uptake, waste excretion, etc.) mean BMR cannot be directly proportional to body mass (see Ref. 160 for a more detailed discussion; the reader is also referred to the other papers in May 2005 issue of Journal of Experimental Biology “Scaling functions to body size: theories and facts” for a detailed coverage of the scaling of physiological functions to body size). When BMR data are expressed as mass-specific metabolic rate, the metabolic rate of mammals is related to the −0.24 to the −0.33 power of body mass, depending on the particular compilation of mammalian BMR values. In other words, small mammal species have higher mass-specific BMRs than do larger mammals, such that for every doubling in body mass, there is a 15–20% decrease in mass-specific BMR. The compilation of mammalian mass-specific BMR values presented here shows it to be proportional to mass−0.31 (see Fig. 5A; the data for birds in Figs. 5 and 6 will be discussed in sect. iv, and we will only consider the mammalian data in this section).

FIG. 5.

The relationship between body mass of mammals and birds and mass-specific basal metabolic rate (A), maximum life span (B), and lifetime energy expenditure (C) is shown. Metabolic rate and body mass data for mammals are from White and Seymour (371), and for birds are from the database accumulated by Dr. C. White (personal communication). The maximum life span data are from Carey and Judge (50), and where there was more than one value provided for a particular species, the largest value was used. For mammals, n = 267 species, while for birds, n = 108 species.

FIG. 6.

An examination of the rate-of-living theory for mammals and birds. The relationship between basal metabolic rate and maximum life span for mammals and birds is shown. The data are the same as those plotted in Figure 5, A and B.

Not all body tissues contribute equally to BMR. For example, ∼70% of the BMR of humans is contributed by internal organs that constitute only ∼7% of body mass (317). The variation in BMR among species is thus due to variation in both tissue size and tissue metabolic rate (159). Although the relative importance of different cellular processes that constitute tissue metabolic rate varies between tissue types, an interesting finding over the last couple of decades has been that membrane-associated activities constitute a large component of cellular metabolic activity during rest (302). For example, it has been estimated that the maintenance of two transmembrane ion gradients (namely, the Na+ gradient across the plasma membrane and the H+ gradient across the mitochondrial inner membrane) together are responsible for about half of the energy turnover associated with the BMR of mammals (302). Another interesting finding has been that although cellular metabolic rate decreases with increasing body mass in mammals, the relative contribution of various processes to total metabolic activity does not vary. For example, in a comparison of the respiration rate of isolated hepatocytes from eight mammalian species, ranging in size from mice to horses, although hepatocyte respiration rate decreased by 13% for every doubling of species' body mass (i.e., proportional to mass−0.20), the relative contribution of mitochondrial ATP production, mitochondrial proton leak, and nonmitochondrial processes did not change with body mass (288). When BMR of mammals varies with body mass, all subcellular processes that constitute metabolic activity seem to vary in unison (see Ref. 159).

B. Maximum Life Span, Body Size, and Lifetime Energy Expenditure of Mammals

Although death may not always be due to the decline in physiological function and/or homeostatic imbalance that occur during aging, MLSP is nevertheless an important species characteristic. Reported MLSPs vary more than 40,000-fold across the animal kingdom, and within mammals by at least 2 orders of magnitude. Generally, MLSP of mammals increases with body size, such that the smallest mammals (shrews) live ∼1 yr while the reported maximum life span for elephants is ∼80 yr (50). Thus maximum life span is positively related to body mass in mammals, such that for every doubling of body mass in mammals, there is on average a 16% increase in MLSP (i.e., MLSP is proportional to mass0.22, see Fig. 5B and Ref. 330). As has long been known, both BMR and MLSP are correlated with body mass in mammals, but the MLSP of mammals is not as strongly correlated with body mass as is mammalian BMR. For example, whereas body mass variation can statistically explain 64% of BMR variation, it only explains 35% of the variation in MLSP of mammals (see Fig. 5, A and B).

The core of the rate-of-living theory ascribes species differences in longevity to differential rates of energy expenditure, such that a short maximum life span will be associated with a high mass-specific metabolic rate. When the MLSP data for the 267 mammal species in Figure 5B is plotted against their mass-specific BMR data in Figure 5A, it can be seen that the MLSP of mammals is indeed negatively correlated with BMR (Fig. 6). Although the relationship is statistically significant, it is obvious that there is a huge amount of variation in the relationship. For example, BMR can statistically explain only 26% of the variation in MLSP of mammals, which is less than the predictive power of body size for mammalian MLSP (35%, see Fig. 5B) and also less than the predictive of body size for the BMR of mammals (64%, see Fig. 5A). The variation obvious in Figure 6 is a clear demonstration that the rate-of-living generalization is only a rough predictor of how long a mammal species can maximally live. Its inability to precisely describe the maximum longevity of a mammal suggests other factors are involved in the determination of maximum life span.

The initial version of the rate-of-living theory posited that body mass-related variation in mass-specific BMR and MLSP would mathematically cancel each other, such that different-sized mammals would have a relatively constant mass-specific lifetime energy expenditure (LEE). In other words, it suggested that because the exponents (as a function of body mass) for MLSP and mass-specific BMR have opposite signs and are similar in value, when these two variables are multiplied together, their product (LEE) would show little dependence on body mass. Regardless of the size of the species of mammal, a gram of any animal would be expected to expend approximately the same amount of energy before it dies at MLSP.

When lifetime resting energy expenditure (LEE) is calculated for the mammal data illustrated in Figure 5, A and B, and plotted against body mass (see Fig. 5C), there is a significant negative slope such that, for every doubling of body mass, LEE of mammals decreases by 6% (i.e., it is proportional to mass−0.09). A similar recent compilation of LEE for 240 mammal species similarly showed that there was a small (but statistically significant) negative allometric relationship of LEE with body mass (330). Lifetime resting energy expenditure is thus not constant (as predicted by the rate-of-living theory) but declines with increasing body mass in mammals. Such average trends, however, mask a huge amount of variation. Statistically, body mass only explains 12% of the variation in LEE between mammals, and the average decline with body mass described above (6% for every doubling of mass) is very small compared with the 81-fold range of values across all mammal species.

Of course, mammal species do not spend all their lives resting in a thermoneutral environment, and thus BMR is not a good measure of total lifetime energy metabolism. Physical activity level can account for as little as 20% of the total energy expenditure in sedentary individuals to as much as 50% in active individuals (104). As well, the energetic cost of maintaining body temperature varies with environmental conditions, body size, and insulation. Similarly, foods eaten may variably influence postprandial diet-induced thermogenesis. Field metabolic rate (FMR), which incorporates all of these influences on energy turnover, may be a better measure of lifetime energy requirements and thus test of the rate-of-living theory. FMR has been measured in a wide range of different-sized mammal species by use of the doubly labeled water technique, which can measure the rate of CO2 production in free-living individuals. A recent compilation of such values for 79 mammal species shows that FMR of mammals is proportional to body mass0.73 (252), while another mammalian data set suggests it is proportional to mass0.62 (330). These studies suggest that for every doubling of body size in mammals, there is on average a 17–23% decrease in mass-specific FMR. When lifetime mass-specific FMR was calculated for 49 mammal species (all weighing <4 kg), there was a statistically highly significant negative relationship with species body mass, which is also contrary to the rate-of-living prediction of a relatively constant lifetime energy turnover per gram of animal (330).

A key premise of the combination of the rate-of-living and oxidative-stress theories is that the more energy an organism expends, the more oxidative damage it will accrue, and the quicker it will die. Although it does not fully explain the differences between mammal species, might it be an explanation within the lifetime of an individual mammal, or between individuals within a species? If this explanation was completely adequate, couch potatoes would lead long lives, while those who follow modern trends for promoting good health by exercising regularly would die sooner. In both humans and mice, although voluntary exercise increases metabolic rate, it does not reduce life span (147, 200). Recent studies have revealed that this theory is fraught with exceptions and problems. Regardless of whether data are expressed per gram body mass, per whole animal or per lifetime, a different pattern emerges when intraspecific variation is examined relative to interspecific mammalian comparisons. Intraspecific studies on dogs (333), mice (234, 332), and humans (301) reveal a positive association between maximum life span and mass-specific metabolic rate and a negative relationship between life span and body size. For example, large dog breeds have lower mass-specific metabolic rates than smaller dogs and also show a reduction in maximum life span. However, confounding any definitive conclusions that may be drawn from this finding, other studies focusing within a particular dog breed (107, 283) revealed no clear relationship between body size and life span. The mass-specific metabolic rate of three dog breeds that differ in maximum longevity (from 8.5 to 14 yr) varied in the opposite direction to that predicted by the rate-of-living hypothesis, namely, the smallest and longest-living breed had the highest rate of living (333).

Several intraspecific studies using mice and rats (40, 146, 202, 332, 333) have not observed an inverse relationship between mass-specific metabolic rate and MLSP. Indeed, some of these studies show the opposite of rate-of-living predictions, namely, that mice with high mass-specific metabolic rates tend to live longer than those individuals with low metabolic rates. One study reported that the individual mice with highest BMR had higher levels of mitochondrial proton leak and higher activation of both UCP-3 and adenine nucleotide translocase (333). Uncoupling of mitochondrial activity to promote thermogenesis may also contribute to the combined effects of higher metabolic rate and longevity in smaller dog breeds. Such uncoupled mitochondrial metabolism has been suggested to result in lower generation of ROS but greater generation of heat (36). Clearly, the simplistic concept that metabolic rate and ROS production are directly correlated in a direct linear manner does not appear to be true. Thus insight into the relationship between metabolism and longevity requires a better understanding of mitochondrial function and efficiency.

C. Mitochondrial Free Radical Production and Antioxidant Defenses of Mammals

The decrease in mass-specific BMR with increasing body size in mammals is also associated with a decrease in tissue density of mitochondrial inner membranes (92), the site of mitochondrial ROS production. The in vitro production rates of both superoxide (327) and hydrogen peroxide (326) by isolated liver mitochondria have been measured in mammals ranging in size from mice to cattle. Allometric analysis of these two data sets shows there is a 10% decrease in superoxide production (it is proportional to mass−0.15) and a 18% decrease in hydrogen peroxide production (i.e., proportionality to mass−0.29) per milligram of mitochondrial protein with every doubling of body mass of the mammal species. It is of interest that a separate study of liver mitochondria from mammals ranging from mice to horses (289) reported that the amount of mitochondrial membrane per milligram mitochondrial protein decreases by ∼7% for every doubling of body mass (i.e., proportionality to mass−0.10). Thus the link between maximum life span and mass-specific metabolic rate in mammals of different body size is associated with appropriate changes in production rates of primary ROS (i.e., superoxide and hydrogen peroxide) by isolated mitochondria, at least in the liver.

However, these studies only included mammal species that followed the rate-of-living theory so the results obtained could also be interpreted as a correlate of that phenomenon. That is, small mammals with short MLSP show high mitochondrial ROS production because their rates of mitochondrial oxygen consumption are higher. Many biochemical reactions apart from oxygen consumption occur at an accelerated rate when the metabolic rate is high, and some of them, unrelated to ROS production, could also, in principle, be responsible for the accelerated aging rate of the short-lived mammals.

It may be that it is enhanced antioxidant defenses that are responsible for the longer MLSP of larger mammals. Thus it was initially surprising for investigators to find that the level of endogenous antioxidant defenses (both enzymatic and scavenger antioxidant systems) of long-living mammal species are lower than those of short-living mammals (2, 17, 285, 312). As can be seen from Table 2, for a number of antioxidant defense systems in a number of tissues, there is a significant negative relationship (or no significant correlation) between the endogenous level of the antioxidant system and MLSP of mammals. This suggests that enhanced antioxidant defenses of mammals are not the reason for the extended longevity of some mammal species but rather are more likely a reflection of the degree of oxidative stress experienced by the tissues of such long-living mammals. These results are compatible with both the notion that endogenous antioxidant defenses are already at optimal levels in animals and that their supplementation, while possibly improving general health of a population (and thus increasing mean longevity), generally has no significant influence on MLSP (see Tables 3 and 4).

View this table:

Comparative studies of endogenous levels of tissue antioxidants in vertebrates differing in their maximum life span potentials

View this table:

Effect of increased antioxidant levels (by either dietary or pharmacological manipulation) on the mean and maximum longevity of vertebrate animals

View this table:

Effects of modifying tissue antioxidants in vivo on MLSP: transgenic/knockout mutants

D. Membrane Composition, Lipid Peroxidation, Body Size, and Life Span of Mammals

Initially, attempts to understand the basis of the body-size trends in mammalian BMR discounted differences in cell composition between species as an explanation (e.g., Ref. 176). However, nearly 30 years ago, Gudbjarnason et al. (118) published a little-noticed relationship between the DHA content of cardiac phospholipids and resting heart rate (a direct correlate of mass-specific BMR) of mammal species ranging from mice to whales. To our knowledge, this is the first record of a relationship between the fatty acid composition of cellular membranes and metabolic intensity. Later, it was demonstrated that this relationship was not restricted to heart membranes, but was also manifest in the fatty acid composition of the membranes of other important tissues of mammals (69, 164). For every doubling of body mass of mammal species there is a 12–24% decrease in the relative DHA content of cellular membranes of mammals (164), which is not dissimilar to the ∼19% decrease in their mass-specific BMR. The cellular membranes of small mammal species are more polyunsaturated than those of large mammal species, and in mammals, this body size-related variation in membrane DHA content is the dominant influence on the body size-related variation in membrane composition. A substantial number of membrane-associated processes have also been shown to vary with body size of mammals and have been related to this difference in membrane composition. Indeed, it has been proposed that it is the difference in membrane fatty acid composition that is causal to the differences in BMR, and this proposal has been called the membrane-pacemaker theory of metabolism (refer to Refs. 156, 158160 for detailed explanation). This theory notes that the cellular membranes of animals with high rates of mass-specific metabolism have high degrees of polyunsaturation and proposes that such high membrane polyunsaturation results in physical properties of membrane bilayers that speed up the activity of membrane-associated proteins and consequently the metabolic rate of cells, tissues, and the whole animal. While this theory is supported by a large body of correlational evidence, there is also important experimental evidence (e.g., from “species cross-over” studies) that substantiates the theory (93, 341, 374).

At the same time that the fatty acid composition of membranes from a variety of mammalian tissues was related to variation in BMR (69, 158, 159), Pamplona and colleagues (266, 271, 272, 275277) also measured the membrane fatty acid composition of a number of mammalian tissues and related it to MLSP of the particular species. Thus the novel and unexpected finding that the composition of membrane bilayers varies systematically in mammals provided a link between metabolic rate and maximum life span of mammalian species. In agreement with this, it was demonstrated that in long-lived mammals, the low degree of membrane fatty acid unsaturation was accompanied by a low sensitivity to in vivo and in vitro lipid peroxidation (269, 272, 275277) and a low steady-state level of lipoxidation-derived adducts in both tissue and mitochondrial proteins of skeletal muscle, heart, liver, and brain (267, 268, 274, 277, 292, 307). These findings were consistent with the negative correlation between MLSP and the sensitivity to lipid autoxidation of mammalian kidney and brain homogenates that had been described some years earlier (72).

If the fatty acid composition of membrane phospholipids is known, it is possible to calculate a peroxidation index (PI; see sect. iiD), which is a number that expresses the calculated susceptibility of the particular membrane to lipid peroxidative damage. When this is done for both liver mitochondrial phospholipids and skeletal muscle phospholipids from a range of different-sized mammal species, it can be seen that there is a strong inverse relationship between PI and MLSP of mammals (see Fig. 7). Interestingly, the relationship between liver mitochondrial phospholipid PI and MLSP has a similar slope to that between skeletal muscle PI and MLSP. The liver mitochondrial membrane PI of mammals is proportional to their MLSP−0.40, which means that a 24% decrease in their peroxidative susceptibility is associated with every doubling of maximum life span. For skeletal muscle membranes, the corresponding value is that a 19% decrease in peroxidative susceptibility is associated with every doubling of MLSP in mammals (i.e., muscle PI is proportional to MLSP−0.30).

FIG. 7.

The relationship between maximum life span of mammals and birds and the peroxidation index of skeletal muscle phospholipids (A) and liver mitochondrial phospholipids (B) is shown. The data points for naked mole rats are from Hulbert et al. (161) and are superimposed on graphs from Hulbert (155).

While maximum life span can differ dramatically between mammal species, there can also be significant longevity differences within a species. For example, populations of two wild-derived strains of mice display extended longevity (both average and maximum life span) compared with genetically heterogeneous laboratory mice when kept under identical conditions (233). The longevity of these wild-derived mice strains in captivity exceeds that yet observed for any laboratory-derived mutant mouse strain. The PI of skeletal muscle phospholipids and liver phospholipids of both wild-derived mice strains with the extended longevity is significantly reduced compared with laboratory mice maintained under identical conditions (163). This is of interest because, as the different mice strains were fed the same diet, it shows that the difference in membrane composition is under genetic control and not determined by dietary differences.

We know almost nothing of the ways in which the regulatory mechanisms controlling membrane fatty acid composition differ between species. With the assumption of the dietary availability of PUFAs, the control of membrane fatty acid composition will reside in the synthetic pathways for fatty acyl chains, the pathways of de novo synthesis of phospholipids, and also in the rapid membrane-remodeling pathways that involve enzymatic deacylation/reacylation cycles. Because PUFA cannot be synthesized de novo by higher animals, they are essential components of the diet. They are only required in small amounts, and in nature, because animals eat other organisms (or their products) they are almost always present in the diet. Another source of PUFA will be their biosynthesis by gut microorganisms. It is however possible to vary availability of different types of fats in laboratory diets. When this has been done in rats, membrane fatty acid composition shows a degree of homeostasis that resists dietary alteration (see Ref. 165). For example, if fed a diet devoid of PUFA, mammals will synthesize an unusual PUFA, mead acid (20:3 n-9) and accumulate it, together with more than normal amounts of MUFA in their membranes. However, with extreme manipulation of dietary fat composition, it is possible to effect small changes in membrane fat composition.

A low PUFA content in cellular membranes (and particularly in the inner mitochondrial membrane) will be advantageous in decreasing the sensitivity of the membrane to lipid peroxidation and would consequently also protect other molecules against lipoxidation-derived damage. To clarify whether the low membrane PUFA content can protect mitochondria from lipid oxidation and lipoxidation-derived protein modification, experimental dietary modification of in vivo membrane fatty acid composition in rats has been performed. These studies partially circumvent the homeostatic system controlling membrane composition and demonstrate that altering the fatty acid composition of cellular membranes does indeed protect postmitotic tissues against lipid peroxidation and lipoxidation-derived macromolecular damage (140, 273, 290).

The decrease in the PUFA content of mammalian membranes as body mass increases would be expected to result in a decreased rate of lipid peroxidation in larger mammals. While there is considerable evidence showing that membrane fatty acid composition affects the rate of lipid peroxidation measured in vitro, there is less evidence that it has an effect on in vivo lipid peroxidation. One of the products of n-3 PUFA peroxidation is ethane, and the rate of ethane exhalation has been used as a noninvasive method for assessing in vivo lipid peroxidation. Figure 8 shows the resting ethane exhalation rates obtained from the literature for mice (control mice in Ref. 79), rats (179, 245, 356), dogs (375), humans (178), and horses (375) relative to the BMR for the same five species (317). As can be seen from Figure 8, the calculated ratio of moles ethane exhaled:moles oxygen consumed ranges from 0.12 in mice to 0.06 in horses (with the exception of the human values which will be discussed in the next section). However, measurement of the exhalation rate of hydrocarbons is technically difficult and, because of potential interlaboratory variation, definitive conclusions as to whether in vivo lipid peroxidation does decrease more than BMR in larger mammals awaits measurement of different-sized mammals by the same laboratory.

FIG. 8.

A comparison of the rates of lipid peroxidation in mammals of different body size relative to their metabolic rate is shown. Ethane exhalation is a measure of the in vivo peroxidation rate of n-3 PUFA (178), and values from the literature are for mice (79), rats (179, 245, 356), dogs and horses (375), and humans (178). Oxygen consumption is the basal metabolic rate of each species (317).

The importance of membrane fatty acid composition in the aging process is highlighted in Tables 5 and 6. The studies summarized in Table 5 show that there are many reports of 1) an increase in either PUFA content or PI of membranes with age, 2) an increase in both in vitro and in vivo membrane lipid peroxidation with age, as well as 3) age-related changes in physicochemical membrane properties. In Table 5, there was either an increase, or no change, in PUFA content/PI reported for most tissues, except the brain, which showed a decrease with age in the rat. Although it does not show the compositional changes with age observed in other tissues, it does reveal an elevated lipid peroxidation, and a decrease in membrane fluidity (or increase in microviscosity) with age that are also characteristic of the other tissues. The increased lipid peroxidation with age is also associated with much greater degrees of lipoxidative damage, to both proteins and DNA (Table 6).

View this table:

Effect of aging on lipid parameters in tissues from different species

View this table:

Effect of aging on steady-state levels of lipoxidative protein and DNA modifications in tissues from different species

PUFAs have lower melting points than saturated fatty acids, and consequently, a relative increase in the PUFA content of a membrane would be expected to render the membrane more fluid. Evidence shows that during aging, the membrane fatty acid profile tends to increase toward peroxidizable PUFAs in many cases (e.g., see Table 5). Based solely on the fatty acid composition, membranes from older animals should exhibit greater fluidity than those from younger individuals. Paradoxically, the opposite is often reported; membrane fluidity decreases with age (see Table 5). The reasons for the changes of membrane fluidity can be explained in two ways: 1) a change caused by fatty acid chain composition and 2) a change in the cholesterol content of the membrane. The paradox described above is potentially resolved by considering that cellular components experience increased oxidative stress with the passage of time. Because the PUFAs are more vulnerable to free radical attack, they experience greater lipid oxidative damage, and lipid peroxidation products have been shown to contribute significantly to membrane rigidity. Lipid peroxidation products and peroxidized lipids are likely dominant factors in creating age-related membrane rigidity. The modification of the physical state of the membrane correlates with the loss of its functioning. Lipid peroxidation products and peroxidized lipid, compared with cholesterol, are greater inducers of membrane rigidity. Furthermore, and interestingly, both mitochondrial structure and function are very sensitive to reactive aldehydic compounds from lipid peroxidation (188, 189, 291, 382, 383).

In view of these widespread changes in membrane composition and lipid peroxidation with age, it is of interest that in the senescence-accelerated mouse (SAM) strain, those mice that are SAM-prone (SAM-P mice) have greater levels of the highly polyunsaturated peroxidation-prone fatty acids (both 22:6 n-3 and 20:4 n-6) and lower levels of the less peroxidation-prone PUFA (18:2 n-6) in their membranes, and consequently a greater PI, than SAM-resistant mice (59, 281). SAM-prone mice also show greater degrees of lipid peroxides in their tissues than do SAM-resistant mice (221).

E. Insights From Relatively Long-Living Mammals (Including Humans)

As noted earlier, there is considerable variation in maximum life span among mammals, and MLSP may be a poor indicator of the relative rate of aging in mammals. For most nondomesticated mammalian species, MLSP values are based on a limited number of captive longevity records maintained by zoological collections. The longevity quotient (LQ), or the ratio of observed MLSP to that predicted from body mass by appropriate allometric equations (e.g., see Fig. 5B), may be a better indicator of the relative rate of aging. Mammals with an LQ value significantly >1 live significantly longer than predicted by their body size, and most likely exhibit slow rates of aging. More limited conclusions might be drawn from those species with an LQ of <1. While an LQ of <1 may reflect accelerated aging processes, it might also indicate an inaccurate MLSP value for the particular species due to inappropriate housing conditions and poor nutrition in captive animals, or high predation rates and other ecological constraints on species in the wild. Relatively few mammals have LQ values >2, but there are some notable exceptions. In Figure 9 we have presented the LQ values for three such mammals: bats, naked mole rats, and humans.

FIG. 9.

The longevity quotients of select mammals and birds are shown. Longevity quotient is the ratio of a species actual maximum longevity to the predicted maximum life span for its body mass. The predicted maximum life span for each species was obtained using the appropriate equation from Figure 5B. The value for bats is mean (±SE) for 11 species of Chiroptera, that for albatrosses is mean (±SE) of 11 species of Procellariiformes, and that for fowl is mean (±SE) of 8 species of Galliformes.

Considerable data based on millions of records confirm that humans are extremely long-living mammals, both for our size and in absolute terms (see Fig. 10). We have an MLSP that is four to five times longer than predicted from the allometric relationships (e.g., Fig. 5B; see also Refs. 7, 294). For example, the similar-sized pig is recorded as having a maximum life span of only 27 yr and sheep only 24 yr, compared with >100 years for Homo sapiens (50). Relatively few other mammals attain similarly high longevity quotients, and there have been numerous theories (some involving superior intelligence) to explain our extraordinary longevity. Indeed, Rubner (306) calculated the lifetime energy potential (LEP) of his five nonhuman mammalian species to range from 590 to 1,110 MJ/g, while if he had included his data for humans (∼3,000 MJ/g) in his comparison, the consequent fivefold range in the LEP of mammals would have made his proposal less convincing.

FIG. 10.

A comparison of the distribution of maximum life span potential (MLSP) in 5 vertebrate classes. Values are means ± SD; range and number (n) of MLSP values are also presented for each vertebrate class. All data are from Carey and Judge (50), and only the maximum value given for each species was used in the analysis. In each graph, values on the y-axis represent number of species.

It has been suggested that the bowhead whale may live to more than 200 yr; however, such an estimated MLSP for this species is not based on captive or free-ranging records, but rather has been estimated from bone growth patterns (300). The longevity record for the larger blue whale is only 110 yr (50). The most reliable record of the longest-living human is for the French woman Jeanne Louise Calment, who lived for 122 years and 164 days (February 21, 1875 to August 4, 1997). Analyses of mortality and census data from the United States suggest that human maximum life span may be ∼130 yr. Mean life expectancy of humans has changed dramatically with improved public health conditions in recent times; however, maximum life span of humans is unchanged (80).

Mammals with a similar extended longevity as humans (i.e., a similar LQ; see Fig. 9) include bats (43) and the subterranean naked mole rat (44). It is tantalizing to assess common features that may contribute to their exceptional longevity. Shared physiological, biochemical, and cellular traits in these species may reveal important components of slow aging and therefore give insight into human aging and our maximum life span.

Naked mole rats are mouse sized and are the longest living rodents known, with a recorded MLSP exceeding 28 yr (44). Naked mole rats have a low BMR for their size (255), but the 30% reduction in BMR is not great enough to explain their fivefold extended maximum longevity. In a series of recent studies, several variables that are considered key tenets of the oxidative stress theory of aging have been compared between naked mole rats and similar-sized mice. Surprisingly, a large amount of the data obtained to date provide little support for a diminished level of oxidative damage in these long-living rodents. Although naked mole rats have low BMR, rates of hydrogen peroxide production are similar in heart mitochondria from both species (190a) and in vascular endothelial cells (187); no differences in overall antioxidant activities are evident (3), and levels of accrued oxidative damage to proteins, DNA, and lipids are greater in the longer-living species (4). One important difference between these two similar-sized rodent species appeared to be the resistance of vascular cells and fibroblasts to oxidative insults and heat stress (187). While mice cells readily underwent apoptotic cell death, in naked mole rat blood vessels only the highest doses of H2O2 induced apoptotic cell death. These data suggest increased life span potential is associated with an increased resistance to proapoptotic stimuli and oxidative insults. Cell membrane properties may play an important role in this regard.

When membrane fatty acid composition was measured in tissues from naked mole rats, they were found to have very low levels of DHA in their tissue phospholipids for their body size. Although both mice and naked mole rats have similar levels of total unsaturated fatty acids in their tissue phospholipids, the low DHA levels in naked mole rats result in more peroxidation-resistant membranes. The PI values calculated for both skeletal muscle and liver mitochondria show that the peroxidation susceptibility of the membranes of naked mole rats is what one would predict for such a long-living rodent species (161; see Fig. 7).

Similarly, the low PI of human membrane phospholipids may be an important trait for the relatively long MLSP of Homo sapiens (275). The relationship between human longevity, membrane composition, and dietary DHA in humans may be complex and is still to be fully elucidated. For example, humans with severe coronary atherosclerosis have erythrocyte membranes with more PUFAs and particularly more DHA (28). Centenarians show a reduced susceptibility to lipid peroxidation than other human subjects, yet have higher levels of DHA (295, 329), which is associated with a decreased incidence of metabolic syndrome and cardiovascular disease and superior insulin sensitivity (21). These data suggest that membrane composition may possibly be involved in the extended longevity of humans, yet the precise picture is still emerging. In this respect it may be that the DHA content of different subcellular membranes is important. For example, it may be excluded from mitochondrial membranes of cells to minimize lipid peroxidation, but be necessary in plasma membranes for normal insulin sensitivity. In this respect it is of interest that the molar ratio of ethane exhalation to oxygen consumption for humans is very low compared with the other mammals for which there are data (see Fig. 8). As ethane is one of the end products of the peroxidation of n-3 PUFA (124) and the rate of ethane exhalation is used as an indicator of the in vivo rate of lipid peroxidation (178), this may indicate that relative to our metabolic rate that humans have a low rate of lipid peroxidation. This is also consistent with the low peroxidation susceptibility calculated for our membranes (see Fig. 7) and may be an important factor in why we are such exceptionally long-living mammals.


A. Metabolic Rate and Body Size of Birds

It was 20 years after the publication of the famous “mouse to elephant” curve by Benedict (23), which showed that mammalian BMR was allometrically related to body mass, before the comparable “hummingbird to ostrich” curve was produced (196). The latter study, which demonstrated that BMR of birds was also allometrically related to body mass, responded to the opinion (174) that small birds, mainly songbirds, had relatively higher metabolic rates than predicted from existing allometric relations and, accordingly, produced two separate allometric relations: an elevated relationship for perching birds (passerines) and a lower one for nonpasserine birds. This separate treatment persisted in later allometric comparisons (e.g., Ref. 6), but was challenged because it was biased by overrepresentation of closely related groups. When an updated data set of avian BMR values was corrected for phylogenetic bias (298), there was no statistically significant difference between passerine and nonpasserine BMR values, a result later confirmed (228). There are a number of allometric comparisons of mammalian and avian metabolic rates, but a robust interpretation of these is that the scaling of mass-specific BMR in relation to body mass is indistinguishable between birds and mammals (e.g., in Fig. 5A, mass-specific BMR is proportional to mass−0.34 for birds and to mass−0.31 for mammals), but that birds have BMRs that average ∼1.5 times greater than those of comparably sized mammals (Fig. 5A and see Refs. 228, 330, 372). Therefore, if the rate-of-living theory accurately described vertebrate life span potential, birds would be expected to live only about two-thirds as long as comparably sized mammals.

Birds have evolved endothermy, and its associated high level of BMR, independently to mammals and, as a test of the membrane-pacemaker theory of metabolism, the body-size variation in metabolic rate has been recently examined in bird species ranging in size from finches to emus. As for mammals, membrane-associated processes are significant components of respiration rate of avian hepatocytes and, while the total respiration rate decreases with body mass, the relative contribution of mitochondrial ATP production, mitochondrial proton leak, and nonmitochondrial processes is unchanged (91). As it is in mammals, the proton leak of isolated liver mitochondria is allometrically related to body mass in birds (37). The PUFA content of avian liver mitochondrial membranes is negatively related to body mass, and this was predominantly due to decreased levels of n-6 PUFA as birds increased in body mass (37). This differs from the situation in mammals where it is n-3 PUFA variation that is primarily responsible for body size trends in liver mitochondrial membrane composition (289). The functional significance of this mammal-bird difference will be discussed later. The general conclusion from these studies is that membrane-associated processes are similarly important in metabolic activity of birds, as they are in mammals, and that the high mass-specific BMR of small avian species is associated with more polyunsaturated membranes, as it is for mammals.

B. Maximum Life Span, Body Size, and Lifetime Energy Expenditure of Birds

Estimates of MLSP for many bird species suffer similar constraints to those for mammals. There is a scarcity of longevity data for most free-living species, and even when available, the extent to which extrinsic mortality factors might disguise the actual MLSP for the species is often unknown because in the wild many individuals die from external causes and not from endogenous aging processes. This is sometimes overcome by basing MLSP on survivorship data for captive animals, but this assumes that diets and caging conditions are conducive to realization of that species' maximum life span. Although it is very likely that zookeepers were aware of the relative differences in longevities of different types and sizes of animals (e.g., Ref. 101), the first formal characterization of avian MLSP in relation to size was performed by Lindstedt and Calder (204). Their analysis was prompted by the concept of “physiological time” (141), which noted that because metabolic rate and longevity covaried with body size, physiological time itself might be similarly scaled with body size. Based on the limited data then available, these authors showed that longevity of birds and mammals increased similarly with increased body size, but that birds lived up to four times longer than comparably sized mammals (204). A more recent characterization of bird longevity using a larger set of data (see Fig. 5B) confirms the common scaling of longevity with body size in birds and mammals (i.e., MLSP is proportional to mass0.20 in birds and mass0.22 in mammals), but reveals that birds live, on average, twice as long as similar-sized mammals (Fig. 5B and Ref. 330). Body mass is a better predictor of both BMR and MLSP in birds than it is in mammals. For example, body mass can explain 84% of the variation in BMR in birds, compared with 64% in mammals, and it can explain 48% of MLSP variation in birds, compared with 35% in mammals (see Fig. 5, A and B).

An important prediction of the rate-of-living theory is that the maximum longevity of an organism should be predicted by its mass-specific metabolic rate. Thus when the MLSP values for 108 species of birds (Fig. 5B) are plotted against their respective BMR (Fig. 5A), there is indeed an inverse relationship between the MLSP and BMR of birds (Fig. 6). Although this relationship is statistically significant, the BMR of a bird species is a less reliable predictor of its MLSP than is its body mass. That is, whereas variation in BMR can explain 41% of variation in MLSP of birds (see Fig. 6), body mass can explain 48% of variation in MLSP and 84% of BMR variation (see Fig. 5, B and A). As well, the relationship between BMR and MLSP of birds is different from the relationship for mammals. This supports the conclusion reached following the analysis of the equivalent data set for mammals, namely, while the rate of living does have some explanatory power concerning the maximum longevity of bird species, this power is limited and suggests that other factor(s) are involved in the determination of MLSP of avian species.

As in mammals, the steeper relationship between BMR and body mass of birds than between their MLSP and body mass (Fig. 5, A and B) means there is a negative relationship between calculated LEE of birds and body mass (see Fig. 5C). The slope of the relationship between LEE and body mass means that for every doubling of mass there is, on average, a 9% decrease in LEE. This also is contradictory to a strict interpretation of the rate-of-living theory. The combination of a relatively higher metabolic rate and a longer life span means that birds, on average, will have about a threefold greater LEE than mammals.

As in mammals, there is 1) a very large (32-fold) amount of variation in the calculated LEE among birds and 2) when field metabolic rate instead of BMR was used to calculate LEE, there was a slight negative relationship with body size in birds, but unlike the situation in mammals, it failed to reach statistical significance (330).

C. Insights From Bird-Mammal Comparisons

About 15 years after the disparity in MLSP between birds and mammals was initially highlighted (204), two separate research groups (the Sohal group in the United States and the Barja group in Spain) compared mitochondrial ROS production and antioxidant levels in short-lived rats (MLSP ∼3 yr) and long-lived pigeons (MLSP ∼35 yr). Ironically, another laboratory had examined H2O2 production in rat and pigeon mitochondrial preparations ∼20 yr before this time, but not in a comparative context (33, 211). The 10-fold greater longevity of pigeons was suggested to be mostly due to a lower rate of mitochondrial H2O2 and superoxide formation compared with rats (19, 183). One group suggested a slightly greater level of antioxidant defense (183) while the other found antioxidants to either not differ significantly between these animals or to be inversely related to MLSP (19, 210; Table 3). Importantly for reconciling the ability of birds to have both a higher MLSP and a higher BMR than mammals, the rate of mitochondrial ROS production per unit of oxygen consumed was considerably lower in pigeons than in rats (19). The notion that MLSP is inversely related to rate of mitochondrial ROS production gained further support from the finding that heart mitochondrial H2O2 production in canaries (MLSP ∼24 yr) and budgerigars (MLSP ∼21 yr) was also significantly lower than in mice (MLSP ∼3.5 yr) (137).

In addition to low rates of mitochondrial ROS formation, these bird species were shown to also have a membrane fatty acid composition that is considerably more resistant to peroxidation than equivalent membranes from the mammals and also to have low levels of lipid peroxidation products in their tissues compared with mammals (275277). That the mammal-bird difference in membrane fatty acid composition was an important determinant of this low level of lipid peroxidation in bird tissues was demonstrated by the dramatic reduction in MDA produced by avian mitochondria compared with mammalian mitochondria when both were subjected to an in vitro free radical generation system (276, 277). Measurement of the fatty acid composition of tissues from eight bird species, ranging from 14 g zebra finches to 34 kg emus, shows that, as in mammals, small birds have more polyunsaturated membranes than larger species (37, 162) and also confirms a lower level of membrane polyunsaturation in birds compared with mammals. A common finding is that bird phospholipids (and thus membranes), compared with mammal phospholipids, have a PUFA balance that generally favors n-6 PUFA over n-3 PUFA. This difference results in bird membranes having a lower PI than mammalian membranes. They are thus less susceptible to peroxidative damage.

When the PI of skeletal muscle and liver mitochondrial membranes of bird species are plotted against their MLSP (see Fig. 7), they follow the same relationship as observed for mammals. This is an intriguing finding and suggests that membrane fatty acid composition differences between birds and mammals are part of the explanation for the relatively greater longevity of birds compared with similar-sized mammals. Such bird-mammal differences in membrane composition have also been related to differences in lipoxidative damage to proteins in the tissues of birds and mammals (138, 292).

D. Insights From Comparison Among Birds

As with mammals, considerable insight into the mechanisms of aging will likely emerge from comparison of similar-sized birds that differ dramatically in maximum life span. The variation in MLSP within birds of similar size dwarfs the average twofold difference in the MLSP between birds and mammals (see Fig. 5B). As with mammals, there are particular groups among birds that are either longer living or shorter living than the MLSP predicted from the general avian equation for their mass. As in mammals, these differences can be expressed by their LQ values (i.e., ratio of actual MLSP to predicted MLSP from body mass). Two such families of birds, the Procellariiformes (petrels and albatrosses) and the Galliformes (pheasants and fowl), are highlighted in Figure 9 as examples. Species from both families cover a substantial and overlapping size range. The selected Procellariiformes range from the 45-g Leach's petrel to the ∼12-kg wandering albatross, while the Galliformes range from the 45-g little button quail to the ∼11-kg wild turkey. The average LQ for the Procellariformes (labeled albatrosses in Fig. 9) are 1.6 ± 0.2 (SE, n = 11) while the average LQ for Galliformes (labeled as fowl in Fig. 9) is 0.5 ± 0.12 (SE, n = 8). The two families have an average threefold difference in maximum life spans, but have indistinguishable BMRs. Thus mass-specific BMR does not explain these longevity differences.

The substantial difference in MLSP between these two families might be related to their very different life-history traits. The Galliformes breed early and produce large clutches with a short incubation period, whereas the Procellariiformes commence breeding much later and have a very small clutch size with an extended incubation period (361). Experimental studies have demonstrated that both early breeding and high fecundity reduce longevity of kestrels and goshawks (74, 182), but these traits foster higher rates of annual reproduction and are, therefore, beneficial to species such as Galliformes that have high extrinsic rates of mortality. The tendency of MLSP to increase directly with size in mammals and birds may also be influenced by size-related differences in extrinsic mortality rates (299), due in part to larger animals being less prone to predation and coping better with periods of food deprivation than smaller forms (7, 46).

Regardless of the factors ultimately responsible for MLSP variation, there are two traits that are often associated with long-lived species: reduced rates of mitochondrial free radical production and reduced susceptibility of membranes to lipoxidation. The differential rates of mitochondrial H2O2 formation in pigeons compared with rats are especially pronounced when fatty acids are used as a metabolic substrate (342), which, in turn, may be due to a greater extent of upregulation of uncoupling proteins (UCPs) in the pigeon mitochondria. This interpretation is consistent with observations that fatty acids stimulate UCP2 and UCP3 expression (47, 309) and that avian muscle mitochondria show increased uncoupled respiration following exposure to exogenous superoxides (348). If the different MLSP of birds with different life histories is partly due to variation in mitochondrial free radical production, then comparative studies focusing on UCP expression and function in birds, particularly the potential for superoxides to stimulate UCP2 and UCP3 expression (38), might yield significant insights (70).

The other factor coinciding with an extended MLSP is having peroxidation-resistant membranes (16, 154, 162). This attribute would complement reductions in mitochondrial free radical production, but is essential for protecting against all sources of oxidative stress. In this regard, it is important to recognize that the intensity of immune reactions, which are a potent source of free radicals, increases directly with MLSP in birds (352). Consequently, one would predict that membrane fatty acid composition favoring peroxidation resistance to be a strongly selected attribute in long-lived species. Comparison of the membrane acyl composition in birds with particularly extended MLSP (e.g., petrels, hornbills, parrots) to those of birds with relatively short MLSP (e.g., gallinaceous species) is an obvious test. Preliminary measurements (W. A. Buttemer, H. Battam, and A. J. Hulbert, unpublished results) of skeletal muscles of petrels and albatrosses (taken as bycatch during long-line fishing operations) show them to have membranes with ∼30% more n-6 than n-3 PUFA, and PI values indicative of animals with MLSP >30 yr. In contrast, their adipose tissue triglycerides were dominated by n-3 fatty acids (5 times more n-3 than n-6 PUFA). This dominance of n-3 PUFA in storage fats is not surprising in view of their fish diet, but the relative absence of n-3 PUFA in their membranes highlights the control of membrane composition as a potentially important mechanism in the determination of a species maximum longevity.


In this section we consider the relatively limited information available for nonendothermic species, first the ectothermic vertebrates and then the invertebrates. Maximum longevity is positively correlated with body size in reptiles, amphibians, and fish (29) as it is for mammals and birds. The equation relating the maximum life span of reptiles to their body mass (kg) is very similar to the equation for mammals (for reptiles MLSP = 14.6mass0.23 compared with 11.6mass0.20 for mammals, see Table 6.1 of Ref. 49). In addition, the mass-specific rate of metabolism is also negatively related to body mass in reptiles, amphibians, and fish, with the slope of this allometric relationship being essentially the same as for mammals and birds (134). However, while mammals and birds are endotherms and therefore maintain a relatively constant body temperature throughout their individual lives, this is not the case for reptiles, amphibians, and fish. In the wild, ectothermic vertebrates will have variable, environmentally determined body temperatures throughout their individual lives.

This complication highlights the different manners by which species-specific data for 1) maximum longevity and 2) metabolic rates are obtained. It thus precludes the calculation of lifetime mass-specific metabolism for reptilian, amphibian, or fish species in the same manner as it has been done for mammals and birds. Mass-specific metabolic rate of ectotherms is a laboratory measurement obtained at a body temperature determined by the experimenter. Values for maximum longevity, although relatively species specific, are a more heterogeneous data set, generally obtained under a wide range of conditions and in the case of ectothermic vertebrates under unknown average body temperatures. Maximum longevity values for a particular species are often obtained opportunistically from captive animals or from marked individuals recaptured in the wild and may not always be reliable maximum values.

The distribution of MLSP values for ectothermic vertebrates (reptiles, amphibians, and fish) and endothermic vertebrates (mammals and birds) together with mean and standard deviation of MLSP values for each of the vertebrate classes are presented in Figure 10. This figure does not take into account size differences between the different species but does indicate that there is considerable overlap between the different vertebrate classes in their MLSP. The mean MLSP values range from 8.3 yr for amphibians to 17.5 yr for mammals. While humans have the highest MLSP record for mammals, there are some ectothermic vertebrates (fish and reptile) with longer MLSP values.

The level of resting metabolism in mammals and birds is 5–10 times that of similar-sized ectothermic vertebrates at the same body temperature (see Ref. 152). Thus, if the rate of living is a significant determinant of a species life span, we would expect that the maximum longevities of ectothermic vertebrates should, on average, be an order of magnitude greater than that of similar-sized endothermic vertebrates at the same body temperature, and even greater if their lifetime average body temperature was lower than the body temperatures typical of mammals and birds (which is almost always the case). As can be seen from Figure 10, this is not the case and is thus an additional argument against a determinate role for the rate-of-living theory on maximum life span of vertebrates. Comparative studies of tissue antioxidant defenses that include values for amphibians and fish show them to be similar to those of mammal and bird species (18, 285).

The effect of body temperature on aging of vertebrates was examined by Walford and Liu in a series of elegant studies on fish (see Ref. 378). These studies questioned whether the temperature effect is due to the simple “slowing down” or “speeding up” of general metabolic activity. For example, lowered temperature increased life span but at the same time increased growth rate of fish (368). More intriguingly, these authors later showed that fish kept at a low temperature during the last half of their life lived longer than fish kept at the same low temperature throughout their whole life span and also lived longer than fish kept at the low temperature for only the first half of their life (205). These observations raise questions concerning the mechanisms by which low temperature extends life span in fish and argue against their mediation simply being due to decreased rate of living.

Invertebrate species are also ectothermic. However, unlike most ectothermic vertebrates, their maximum longevity values have generally been determined in the laboratory under conditions of constant specified body temperature. Two invertebrates have been especially important in elucidating the mechanisms determining life span. They are the fruit fly Drosophila melanogaster and the nematode Caenorhabditis elegans. Loeb and Northrop (207) first demonstrated that reducing the temperature increased the life span of Drosophila. This finding was important evidence in the development of the rate-of-living theory of aging when Pearl (284) argued that it was mediated by the general influence of temperature on metabolic rate.

Lowered temperature has been shown to extend life span of the fruit fly (236), the housefly (323), and the milkweed bug (223). All three studies also showed that lowered temperature decreased the rate of oxygen consumption by these insect species, and the calculated lifetime oxygen consumption was relatively similar at the different experimental temperatures. Although this is compatible with the rate-of-living theory, it is not proof of the theory as temperature is known to affect the rate of many processes (as well as other variables such as cellular composition), and it may be that longevity extension is due to one (or more) of these other effects of temperature. Interestingly, repeated short periods of mild heat stress have also been shown to extend life span of Drosophila (135).

In the housefly, the in vivo rate of lipid peroxidation of n-6 PUFA has been shown to increase with age when houseflies are kept at 25°C and to be reduced when houseflies are kept at 18°C (325). Whether this lower n-6 PUFA peroxidation is due to a decrease in the rate of oxygen consumption or change in membrane composition (or both) is unknown. It is of interest that the relative n-6 PUFA content increases with age in houseflies at 25°C and decreases when flies are maintained at 18°C. In both cases, these n-6 PUFA changes are compensated by opposite changes in the content of peroxidation-resistant MUFA (323). In Drosophila, there is no increase in metabolic rate with age (5, 157, 242, 303), and similarly in the blowfly Calliphora stygia, there was no increase in food consumption with age (166). Rather than an increase, some of these studies actually report a drop in metabolic rate at the end of the adult life span of these insects.

Fruit flies that differ in life span have also been used to examine the role of metabolic rate in the determination of life span. There is no association between metabolic rate and life span in five species of Drosophila (293). Similarly, laboratory-produced strains of Drosophila melanogaster that differ substantially in life span do not differ in metabolic rate (5, 242, 364). Furthermore, the long-living lines of Drosophila had essentially the same antioxidant enzyme activities as the short-living fruit-fly lines (242). Thus neither the rate-of-living nor antioxidant defenses appear to explain longevity differences in these Drosophila lines. Similarly, within a Drosophila population, there is no correlation between individual metabolic rate and individual life span (157). Within a population of blowflies, there was also no correlation between individual rates of food consumption and individual life span (166).

The role of metabolic rate in the determination of adult life span of insects has also been examined by manipulation of flight activity. Unlike healthy rats and humans, where the increased metabolism associated with exercise does not shorten life span (e.g., Refs. 147, 200) but often has beneficial effects sometimes even increasing average longevity (e.g., Ref. 145), a number of studies have shown that this is not the case for insects. Both houseflies and fruit flies allowed to fly have much shorter longevities than those prevented from flight (215, 322). In both species, flight activity is associated with greater oxidative damage to some proteins. Interestingly, whereas flight activity is associated with greater mitochondrial H2O2 production in houseflies (376), there is no effect of activity on H2O2 production in fruit flies (215). In the Drosophila, there were changes in fatty acid composition such that the peroxidation index of lipids increased with age and flight activity and was associated with an increased susceptibility to lipid peroxidation (215). The fatty acid composition of phospholipids from the heads of wild-type Drosophila shows them to have short-chain PUFA but lack long-chain PUFA (338, 379). Although there are considerable data for fatty acid composition of invertebrates in general, they are almost never presented in relationship to aging or maximum life span of the particular species.

The social insects have been suggested as particularly good model organisms to investigate the mechanisms of aging for two reasons: 1) “queens” can be extraordinarily long living, and 2) there is sometimes tremendous variation in life span between genetically identical “queens” and “workers” (171). In some ant species, queens can live up to 30 yr and frequently live 10 times longer than workers (144). Similarly, queen honeybees are reported to have a maximum longevity an order of magnitude greater than worker (i.e., nonreproductive female) honeybees (373). Enhanced antioxidant defenses in queens are not a likely explanation, as one study has shown queen ants have a reduced expression of Cu/Zn-SOD compared with shorter-living workers (282) while others have shown that queen honeybees generally have the same (or, in some cases, lower) levels of antioxidant defenses than worker honeybees (68, 370). The mass-specific metabolic rates of worker and queen honeybees are essentially the same (96). Recent measurements show that queen and worker honeybees have different membrane fatty acid composition and that the peroxidation index of phospholipids of from queen honeybees is ∼33% of that of workers (122). If the slope of the relationship between PI and MLSP is mathematically similar in honeybees to that described for mammals and birds (155, see Fig. 7), then this difference is capable of explaining the order of magnitude difference in longevity between queen and worker bees. It may be that the dramatic longevity differences between castes of social insects are mediated by different fatty acid composition of cellular membranes of these insects.

The nematode C. elegans has been a popular model organism in the study of aging. Both its short adult life span and its relatively simple genome have been important aspects of this experimental popularity. However, while C. elegans has given much insight into the genes that influence aging and the determination of life span, its small size has limited its ability to give insight into the role of metabolic rate in aging. Metabolic rate has been measured both by respirometry and calorimetry in C. elegans, and both methods show an unusual pattern of changing metabolism during the nematode's lifetime. Unlike the situation in higher multicellular animals, where metabolic rate is relatively constant throughout adulthood, there is a dramatic and substantial continual decline in metabolic rate throughout the adult life span of C. elegans (34). In addition, the metabolic rate of C. elegans is very labile and dependent on environmental conditions (362). There is no agreement as to whether metabolic rate of control and long-living mutant strains of C. elegans differ (35, 363, 365). To our knowledge, there are no examinations as to whether membrane fatty acid composition differs between strains of C. elegans that differ in longevity. One interesting recent study reports that genetic manipulation to increase the expression of glutathione transferases (enzymes that metabolize 4-HNE, an important and damaging product from peroxidation of n-6 PUFA) increase both the resistance to stress and the longevity of C. elegans(10). This study showed a fascinating and strong positive correlation between glutathione transferase activity for 4-HNE and median life span of lines of C. elegans.


A. Calorie Restriction

Dietary calorie restriction (CR) is a long-known treatment that can extend life span and was first described for rats (224). It has been shown to extend longevity in a wide range of other animal species and is probably the most investigated treatment in the study of aging (for an excellent review, see Ref. 219). In rats and mice, the degree of life span extension is linearly related to both the degree of CR and the time for which CR is imposed (231). CR is proposed to extend life span by both affecting the intrinsic rate of aging as well as suppressing pathogenesis and thus reducing the short-term risk of death (381). There is some argument as to whether CR slows the rate of aging or whether it delays the start of aging but does not slow its rate (see Ref. 220).

In line with the rate-of-living theory of aging, many investigators initially thought that CR extended life span by lowering metabolic rate. While rats subjected to CR had a low BMR per whole animal compared with controls, CR also results in a reduced body size, and it was shown over twenty years ago that the CR in rats does not reduce their mass-specific metabolic rate, expressed relative to either total body mass or lean body mass (225). Indeed, it has recently been shown that CR in rats results in a metabolic rate higher than would be predicted from the altered body composition (319). Similarly CR does not reduce metabolic rate in either mice (99) or Drosophila (157).

A reduction in oxidative damage is a consistent finding following CR and, although the reason for this reduced oxidative damage is not completely understood, it is thought to be the cause of its ability to extend longevity (112, 219, 231). Measurement of mitochondrial in vitro production of superoxide and hydrogen peroxide (generally referred to as ROS production) shows it to be lowered following long-term CR (112), but the results are more equivocal following short-term CR. In mice there is a small effect on isolated liver and muscle mitochondria following 6 mo CR but none after 3 mo (99). In vitro ROS production is reduced in rat heart mitochondria following 1-yr CR but not after 6-wk CR (113). It is reduced in rat liver mitochondria after 6-wk CR (114), while others report no effect after 1-, 6-, or 12-mo CR (123, 296). It is reduced in rat muscle mitochondria after 2-wk and 2-mo CR (26); however, another study reports no change after 2-mo CR (115). All of these studies have measured in vitro ROS production by isolated mitochondria, and it is of interest that a recent report shows no effect of 4-mo CR on cellular ROS production by rat hepatocytes (194).

A reduction in mitochondrial production of superoxide or hydrogen peroxide is not the only means to limit oxidative damage. Two other means are 1) enhanced antioxidant defenses and 2) diminished lipid peroxidation. The effect of CR on antioxidant defenses is mixed. Some studies report that CR enhances antioxidant defenses, whilst others report either no changes or diminished defenses. There does not appear to be a consistent effect of CR on antioxidants, and the reader is referred elsewhere for details (219). As pointed out earlier, unlike oxidative damage to nucleic acids and proteins, oxidative damage to polyunsaturated lipids produces many powerful reactive molecules that can further damage other important biological molecules. Thus lipid peroxidation can be both the manifestation of, as well as the cause of, oxidative damage. In vivo lipid peroxidation is decreased during CR in rats, with one study reporting a decreased pentane exhalation (indicator of n-6 PUFA peroxidation) but no decrease in ethane exhalation (indicator of n-3 PUFA peroxidation) following 12-mo CR (222) while another study reports decreased ethane exhalation following 2-wk CR in rats (121).

A decreased rate of lipid peroxidation can come about via a decrease in the availability of peroxidizable fatty acids, and it has been proposed that such membrane alteration is the basis of the protective effect of CR (382). Laganiere and Yu (188) first showed CR altered the fatty acid composition of liver mitochondrial and microsomal membranes in rats such that they became less susceptible to peroxidative damage. Since this seminal observation, others have reported similar CR-induced changes for other tissues (53, 190, 201, 268, 270, 346) as well as for different classes of liver phospholipids (54, 168). All of these studies involved long-term CR in rats with the shortest period examined being 10-wk CR (54). A study of CR in mice (99) reports similar changes in the fatty acid composition of phospholipids from liver, heart, kidney, and brain, as well as liver and muscle mitochondria, with the changes in this study being manifest following 1 mo of CR (the earliest period sampled). Recently, it has been found that CR treatments as low as 8.5 and 25% result in a decreased PI of membrane lipids and lipoxidative damage to proteins (Pamplona, unpublished results). The effect was especially pronounced when specific components of the diet (protein and methionine) were restricted (9, 313).

A few studies have failed to show significant CR-induced changes in mitochondrial fatty acid composition (192, 270, 296); however, these studies have measured total lipids rather than phospholipids, and thus their values include the fatty acid composition of triglycerides associated with the mitochondrial pellet. In view of the fact that CR will have significant influence on the triglyceride content of cells, and triglycerides generally differ in their fatty acid composition compared with phospholipids, it has been suggested (99) that CR effects on membrane composition may have been masked in these studies.

Rapid and reversible effects of CR have been reported for mice (75, 334) and fruit flies (216). The most rapid responses to changes in energy intake will be hormonal systems (e.g., insulin). In mice, CR-induced changes in membrane fatty acid composition are early effects and are likely mediated by hormones. Insulin, triiodothyronine, and growth hormone are candidates as they are all rapidly and significantly lowered by CR (219). For example, low insulin levels (due to diabetes) are associated with membrane fatty acid changes similar to those observed following CR (126, 131, 227). Insulin and growth hormone have also been shown to change membrane fatty acid composition, both during ad libitum feeding and during CR in rats (311). Insulin has recently been shown to reverse some CR-induced changes in liver mitochondrial function in rats (193). Hypothyroidism decreases the degree of polyunsaturation of n-6 PUFA in membranes (153), which is similar to the changes observed during CR.

B. Antioxidant Defenses

A number of techniques have been used in attempts to influence the maximum life span of animals by manipulation of antioxidant defenses. These include dietary supplementation, pharmacological intervention, and more recently transgenic techniques. The antioxidants examined include both those naturally produced by animals as well as novel antioxidant defenses. The effects of increased antioxidant levels, either by dietary or pharmacological means, on both the mean and maximum longevity of various vertebrate animals are summarized in Table 3. As can be seen from Table 3, there was generally no effect on maximal life span, although in some studies there was an increase in average longevity. This is also consistent with studies in invertebrates (198, 261). The extension of maximum life span by SOD mimetic drugs described a few years ago for C. elegans (230) has been unable to be repeated by others either for C. elegans (170) or Drosophila (214).

Antioxidant defenses have also been modified by transgenic means in a number of studies (see Table 4). While maximum life span has been decreased in a number of studies following the complete removal of particular antioxidant defense (e.g., in homozygous knockouts), the overexpression of antioxidant defenses has essentially had no effect on maximum longevity. It is of interest that heterozygous knockouts (i.e., resulting in a 50% reduction in antioxidant levels) have often had no influence on maximum life span. The results from the transgenic studies (Table 4) are in agreement with the other studies (Table 3). They indicate that 1) while the complete ablation of specific antioxidant defenses can be detrimental and 2) enhanced antioxidant defenses might improve the health of some individuals in a population (and thus increase average longevity), that antioxidant defenses do not appear to be limiting and do not appear to slow down the endogenous aging process. Consequently, they do not appear to be responsible for determining the maximum life span distinctive for individual species.

Overall, studies that experimentally increase tissue antioxidant levels do not increase maximum life span in vertebrates (312). While for most dietary studies the additional antioxidants do not change MLSP, a few show marginal increases in longevity. However, these increases have only been observed in investigations in which the MLSP of the control animals was less than 3 yr, which is a short life span for the examined rodents. In studies in which the maximum life span of the control rodents was greater than 3 yr, there was no effect on MLSP. The most frequent benefit observed was an increase in mean life span. In most reports this effect was due to a decrease in early death from a reduced incidence of disease. Antioxidant supplementation increased mean life span when survival of the controls was suboptimal, whereas when the survival of the control animals was more optimal, the antioxidants were without effect. Antioxidants have not increased maximum life span beyond the values of controls reared under optimal conditions. These results suggest that antioxidants can protect against increased oxidative stress from pathological situations or exogenous damage but do not change aging rate. This is confirmed in models in which transgenic mammals overexpressing different kinds of antioxidant enzymes show increased resistance to oxidative stress, but no increase in maximum longevity.

Similarly, in the case of Drosophila, early studies found increases in maximum longevity after combined SOD plus catalase overexpression (but not individually); however, this was not observed in later studies when longer-lived control strains were used (259). It is possible that overexpression of antioxidant enzymes is more important in short-lived very active insects than in mammals. In summary, enhanced antioxidant defenses seem to increase life span in studies where the control group has short life span (possibly because of suboptimal antioxidant defenses in the control animals). Antioxidant defenses seem to be optimized in vivo such that additional expression is unable to increase protection against aging-related oxidative damage, whereas they are useful when oxidative damage is higher than normal.

C. Genetic Manipulations

Other genetic manipulations (apart from the antioxidant-related genes mentioned above) have also been shown to increase the maximum life span of some species. Discussion of these is limited in the current contribution, but one type deserves special mention. Mutations that diminish signaling in insulin/IGF pathway have been shown to extend life span in C. elegans (103, 172), Drosophila (62, 350), and mice (30, 149).

The role of diminished IGF in extending longevity is interesting and likely explains the opposite relationship between body size and maximum life span observed within a mammalian species compared with between species. As outlined in earlier sections, a long life span is often associated with a large body size among vertebrate species. However, within a species, extended longevity is often associated with a small body size. For example, long-living mice are often small and have low IGF-I levels (234). Similarly, the smaller dog breeds often have longer life spans than large breeds of dog (203), and smaller dog breeds have the same growth hormone secretory capacity but reduced IGF compared with large breeds (89).

To our knowledge, nothing is known as to whether membrane fatty acid composition is altered in any of these mutant strains and thus is a good topic for further investigation. It is known that both insulin and growth hormone affect membrane fatty acid composition in rats (311). The suggestion that diminished insulin secretion may be responsible for mediating some of the effects of CR (see sect. viA) provides a common thread between these two methods of extending life span. In this connection it is also of interest that the longest mean life span achieved by CR chico1 mutant flies (with their diminished insulin/IGF pathway) is the same as that achieved by CR in wild-type fruit flies (63), which suggests that the long-living mutant flies are less sensitive to food concentration and both this genetic mutation and CR may extend life span by the same mechanism.


The fact that maximum life span is a species characteristic suggests that it has a genetic basis, but the specific biochemistry that is being controlled by such genes is not always obvious. The observation that large mammals tend to live longer than small mammals was seminal in the development of theories to explain aging and the determination of life span. The connection between this trend in longevity and the slower metabolic rate of large compared with small mammals identified aerobic metabolism as being central for understanding what determines a species life span. This centrality was initially expressed in the rate-of-living theory that, in turn, led to the development of the free-radical theory and consequently the oxidative-stress theory of aging, which is currently the most accepted explanation of aging. We do not suggest that the rate-of-living theory and the membrane-pacemaker modification of the oxidative-stress theory described here are incompatible. Rather, we suggest that the rate-of-living theory cannot alone explain all of the variation in longevity of animals. However, many of the exceptions that cannot be explained by the rate-of-living theory (and are summarized in sect. i) do appear capable of explanation by knowledge of membrane fatty acid composition in each particular case.

Although antioxidant defenses are an important part of the oxidative-stress theory, it is now obvious that antioxidant defenses are not significantly involved in determining the maximum life span of a given species. Certainly diminution of antioxidant defenses can shorten life span, but many attempts over the years to significantly extend the maximum life span of species by manipulation of antioxidant status have not been successful. Like most other enzymes, enzymatic antioxidant defenses are likely present in excess. If antioxidant defenses were limiting, we might expect more variation in maximum longevity between closely related species than already exists, as changes in the abundance of specific cellular proteins is a relatively easy change for evolutionary processes to select.

Although body size-related variation in metabolic intensity is certainly a factor in explaining some of the intraspecific body size-related variation in longevity, it is inadequate in explaining the longevity differences between vertebrate classes, such as the difference between mammals and birds, or between ectothermic and endothermic vertebrate classes. However, the incompleteness of this explanation does not mean it should be fully discarded.

Mitochondrial oxygen consumption results in the production of superoxide, and variation in resting metabolic rate among mammalian species will result in variations in the rate of mitochondrial superoxide and consequently other ROS. There is not a simple stoichiometry between metabolic rate and mitochondrial ROS production. Some of the variation in MLSP among vertebrates can be related to mitochondrial ROS production.

It is only relatively recently that we have appreciated that the body size-related variation in whole animal metabolic rate is largely a cellular property and that variation in the fatty acid composition of membranes may underlie this difference in metabolic activity of cells. Indeed, it is only recently that we have become aware that cell composition (and specifically membrane composition) varies in a systematic manner between species. This fact was not known when the connection between the speed of metabolism and a species longevity was first recognized. We have suggested here that membrane fatty acid composition may be a fundamental property determining maximum longevity.

When the fact that fatty acids differ dramatically in their susceptibility to peroxidative damage is combined with species variation in membrane composition, the link between body size, metabolic rate, and longevity becomes more apparent. In this contribution, we have discussed evidence that the fatty acid composition of membranes can potentially explain 1) the shorter longevity of small mammals compared with larger mammals, 2) the exceptional longevity of naked mole-rats compared with similar-sized mice, 3) the extended longevity of wild-derived lines of mice compared with laboratory mice, 4) the longer life spans of birds compared with similar-sized mammals, 5) the extended longevity of rodents caused by calorie restriction, 6) the longevity difference between workers and queens in honeybees, and 7) also suggested it may be an explanation for the exceptional longevity of our own species, Homo sapiens. Furthermore, we have also discussed the studies which show that membranes with fatty acid compositions prone to peroxidation are also associated with greater levels of lipoxidative damage to other cellular constituents. All these studies suggest that variation in membrane fatty acid composition may be an important missing piece of the jigsaw puzzle explaining aging and the mechanisms that determine the maximum life span specific for each species. It is a testable hypothesis that awaits further experiments. Since metabolic rate has been called the “fire of life,” then maybe the mitochondrial production of superoxide from this fire can be likened to the “pilot light” and lipid peroxidation (and its reactive products) can be likened to the “furnace” of oxidative stress.


The investigations from our respective laboratories described here have been supported by the following sources: Australian Research Council grants (to A. J. Hulbert and W. A. Buttemer); Grants FIS (98/0752, 00/0753), MEC (BFI2003-01287, BFU2006-14495/BFI), and Generalitat of Catalonia (2001SGR00311, 2005SGR00101) (to R. Pamplona); and National Institutes of Health Grants GM-08168-25 and AG-22891 (to R. Buffenstein).


We especially thank Dr. Craig White who provided us with access to his extensive database on the BMR of birds. We also thank Zoe Hulbert Johnson for assistance. The editor and two anonymous reviewers provided comments that resulted in an improved manuscript.

This work is a combined contribution from the four speakers in the “Life and Death: Metabolic Rate and Life Span” symposium at the XXXV International Congress of Physiological Sciences (San Diego, Mar-Apr 2005).

Address for reprint requests and other correspondence: A. J. Hulbert, School of Biological Sciences, Univ. of Wollongong, Wollongong, NSW 2522, Australia (e-mail:


  • 1 Peroxidation index = 0.025 × (% monoenoics) + 1 × (% dienoics) + 2 × (% trienoics) + 4 × (% tetraenoics) + 6 × (% pentaenoics) + 8 × (% hexaenoics), while unsaturation index = 1 × (% monoenoics) + 2 × (% dienoics) + 3 × (% trienoics) + 4 × (% tetraenoics) + 5 × (% pentaenoics) + 6 × (% hexaenoics).


  1. 1.
  2. 2.
  3. 3.
  4. 4.
  5. 5.
  6. 6.
  7. 7.
  8. 8.
  9. 9.
  10. 10.
  11. 11.
  12. 12.
  13. 13.
  14. 14.
  15. 15.
  16. 16.
  17. 17.
  18. 18.
  19. 19.
  20. 20.
  21. 21.
  22. 22.
  23. 23.
  24. 24.
  25. 25.
  26. 26.
  27. 27.
  28. 28.
  29. 29.
  30. 30.
  31. 31.
  32. 32.
  33. 33.
  34. 34.
  35. 35.
  36. 36.
  37. 37.
  38. 38.
  39. 39.
  40. 40.
  41. 41.
  42. 42.
  43. 43.
  44. 44.
  45. 45.
  46. 46.
  47. 47.
  48. 48.
  49. 49.
  50. 50.
  51. 51.
  52. 52.
  53. 53.
  54. 54.
  55. 55.
  56. 56.
  57. 57.
  58. 58.
  59. 59.
  60. 60.
  61. 61.
  62. 62.
  63. 63.
  64. 64.
  65. 65.
  66. 66.
  67. 67.
  68. 68.
  69. 69.
  70. 70.
  71. 71.
  72. 72.
  73. 73.
  74. 74.
  75. 75.
  76. 76.
  77. 77.
  78. 78.
  79. 79.
  80. 80.
  81. 81.
  82. 82.
  83. 83.
  84. 84.
  85. 85.
  86. 86.
  87. 87.
  88. 88.
  89. 89.
  90. 90.
  91. 91.
  92. 92.
  93. 93.
  94. 94.
  95. 95.
  96. 96.
  97. 97.
  98. 98.
  99. 99.
  100. 100.
  101. 101.
  102. 102.
  103. 103.
  104. 104.
  105. 105.
  106. 106.
  107. 107.
  108. 108.
  109. 109.
  110. 110.
  111. 111.
  112. 112.
  113. 113.
  114. 114.
  115. 115.
  116. 116.
  117. 117.
  118. 118.
  119. 119.
  120. 120.
  121. 121.
  122. 122.
  123. 123.
  124. 124.
  125. 125.
  126. 126.
  127. 127.
  128. 128.
  129. 129.
  130. 130.
  131. 131.
  132. 132.
  133. 133.
  134. 134.
  135. 135.
  136. 136.
  137. 137.
  138. 138.
  139. 139.
  140. 140.
  141. 141.
  142. 142.
  143. 143.
  144. 144.
  145. 145.
  146. 146.
  147. 147.
  148. 148.
  149. 149.
  150. 150.
  151. 151.
  152. 152.
  153. 153.
  154. 154.
  155. 155.
  156. 156.
  157. 157.
  158. 158.
  159. 159.
  160. 160.
  161. 161.
  162. 162.
  163. 163.
  164. 164.
  165. 165.
  166. 166.
  167. 167.
  168. 168.
  169. 169.
  170. 170.
  171. 171.
  172. 172.
  173. 173.
  174. 174.
  175. 175.
  176. 176.
  177. 177.
  178. 178.
  179. 179.
  180. 180.
  181. 181.
  182. 182.
  183. 183.
  184. 184.
  185. 185.
  186. 186.
  187. 187.
  188. 188.
  189. 189.
  190. 190.
  191. 190a.
  192. 191.
  193. 192.
  194. 193.
  195. 194.
  196. 195.
  197. 196.
  198. 197.
  199. 198.
  200. 199.
  201. 200.
  202. 201.
  203. 202.
  204. 203.
  205. 204.
  206. 205.
  207. 206.
  208. 207.
  209. 208.
  210. 209.
  211. 210.
  212. 211.
  213. 212.
  214. 213.
  215. 214.
  216. 215.
  217. 216.
  218. 217.
  219. 218.
  220. 219.
  221. 220.
  222. 221.
  223. 222.
  224. 223.
  225. 224.
  226. 225.
  227. 226.
  228. 227.
  229. 228.
  230. 229.
  231. 230.
  232. 231.
  233. 232.
  234. 233.
  235. 234.
  236. 235.
  237. 236.
  238. 237.
  239. 238.
  240. 239.
  241. 240.
  242. 241.
  243. 242.
  244. 243.
  245. 244.
  246. 245.
  247. 246.
  248. 247.
  249. 248.
  250. 249.
  251. 250.
  252. 251.
  253. 252.
  254. 253.
  255. 254.
  256. 255.
  257. 256.
  258. 257.
  259. 258.
  260. 259.
  261. 260.
  262. 261.
  263. 262.
  264. 263.
  265. 264.
  266. 265.
  267. 266.
  268. 267.
  269. 268.
  270. 269.
  271. 270.
  272. 271.
  273. 272.
  274. 273.
  275. 274.
  276. 275.
  277. 276.
  278. 277.
  279. 278.
  280. 279.
  281. 280.
  282. 281.
  283. 282.
  284. 283.
  285. 284.
  286. 285.
  287. 286.
  288. 287.
  289. 288.
  290. 289.
  291. 290.
  292. 291.
  293. 292.
  294. 293.
  295. 294.
  296. 295.
  297. 296.
  298. 297.
  299. 298.
  300. 299.
  301. 300.
  302. 301.
  303. 302.
  304. 303.
  305. 304.
  306. 305.
  307. 306.
  308. 307.
  309. 308.
  310. 309.
  311. 310.
  312. 311.
  313. 312.
  314. 313.
  315. 314.
  316. 315.
  317. 316.
  318. 317.
  319. 318.
  320. 319.
  321. 320.
  322. 321.
  323. 322.
  324. 323.
  325. 324.
  326. 325.
  327. 326.
  328. 327.
  329. 328.
  330. 329.
  331. 330.
  332. 331.
  333. 332.
  334. 333.
  335. 334.
  336. 335.
  337. 336.
  338. 337.
  339. 338.
  340. 339.
  341. 340.
  342. 341.
  343. 342.
  344. 343.
  345. 344.
  346. 345.
  347. 346.
  348. 347.
  349. 348.
  350. 349.
  351. 350.
  352. 351.
  353. 352.
  354. 353.
  355. 354.
  356. 355.
  357. 356.
  358. 357.