Oxygen and the Upper Limits to Animal Design and Performance

A biochemist might rationalize the use of mitochondrial oxidative phosphorylation by animals that achieve the highest body-mass-specific metabolic rates (⁠ /M_b) on the basis of the high yield of ATP per mole of substrate oxidized, the high rates of ATP synthesis possible and the exquisite control mechanisms that allow oscillations of up to 2–3 orders of magnitude between basal and maximal rates. However, the oxidation of 1 mole of glucose or C₁₆ fatty acid requires the consumption of 6 or 23 moles of O₂, respectively. Biochemical pathways have not evolved and do not function in animals in the absence of structural and functional context. Anatomists and physiologists would therefore superimpose upon the biochemist’s conceptual framework the idea that the steady-state fluxes required for aerobic metabolism are achievable as a consequence of support systems and processes that supply O₂ and metabolic fuels to metabolically active organs. By adopting an integrative approach to the study of the energetics of exercise in animals, it has been possible to gain insights into how the design and function at the level of intracellular pathways of aerobic energy metabolism are related to the design and function of gas and fuel transport systems at higher levels of organization. Such an integrative approach is adopted here to analyze the design of organisms that achieve the highest mass-specific aerobic metabolic rates in the Animal Kingdom.

The inverse relationship between body mass and resting mass-specific metabolic rate is well known among comparative physiologists (Schmidt-Nielsen, 1984). This inverse relationship also holds for maximal aerobic body-mass-specific metabolic rates (⁠ /M_b) (Taylor et al. 1989). Because of their small size and the energetic demands of flight, small hummingbirds display the highest known mass-specific aerobic metabolic rates among vertebrates (Suarez, 1992). Flying insects surpass hummingbirds and probably achieve the highest /M_b of all animals (Sacktor, 1976). In my laboratory, hummingbirds and honeybees have served as useful models in studies of animal design and performance. The flight muscles of hummingbirds (Grinyer and George, 1969) and insects (Casey et al. 1992) consist of one fiber type. These account for more than 90% of whole-body O₂ consumption during steady-state exercise (Rothe and Nachtigall, 1989; Suarez, 1992). Such features make possible the estimation of O₂ and fuel flux rates into working muscles, as well as rates of substrate catabolism and enzyme function during flight.

Muscles are biological machines that convert chemical energy into mechanical work. At the contraction frequencies that characterize flight, ATP is hydrolyzed primarily by actomyosin ATPase, Ca²⁺-ATPase and Na⁺/K⁺-ATPase (Homsher and Kean, 1978). In asynchronous insect flight muscles, such as those found in bees, an extensive sarcoplasmic reticulum is lacking, and Ca²⁺-ATPase is thought to make a relatively minor contribution to ATP hydrolysis (Casey et al. 1992). Metabolic fuels are oxidized and O₂ is consumed by the muscle mitochondria as ATP is resynthesized at the same rate at which it is hydrolyzed. Unlike the locomotory muscles in larger mammals (Weber, 1997), hummingbird and insect flight muscles are both characterized by a high degree of reliance upon fuels originating from exogenous sources. Their highly aerobic metabolic organization requires that rates of mitochondrial O₂ consumption be matched by flux rates through the lungs, cardiovascular system, capillary walls in the case of hummingbirds (Fig. 1) and the tracheal system in the case of insects.

When a 4 g hummingbird hovers at a wingbeat frequency of 80 Hz, its /M_b increases approximately 12-fold above basal resting values to approximately 40 ml g⁻¹ h⁻¹ (Bartholomew and Lighton, 1986; Lasiewski, 1963), which is roughly 10 times higher than rates achieved by human athletes exercising at (e.g. Blomstrand et al. 1986). By flying hummingbirds in helium–O₂ mixtures less dense than air, it has been possible to induce short-term metabolic rates even higher than this (Chai et al. 1997). During ‘normal’ hovering, a steady-state activity hummingbirds engage in for the purpose of foraging on floral nectar, flight muscle O₂ consumption rates of approximately 2 ml g⁻¹ min⁻¹ have been estimated, corresponding to rates of ATP turnover of close to 500 μmol g⁻¹ min⁻¹ (Suarez, 1992; Suarez et al. 1990, 1991). Worker honeybees, hovering at wingbeat frequencies of 250 Hz, are even more impressive as their flight muscles consume 6 ml O₂ g⁻¹ min⁻¹ (Suarez et al. 1996). A single flying honeybee turns over (i.e. hydrolyzes and resynthesizes) 1.39×10¹⁵ molecules of ATP per wingbeat cycle!

Among vertebrates, skeletal muscle mitochondria are thought to be O₂-limited during exercise at ⁠, and the control of O₂ flux rates is distributed or shared among all the convective and diffusive steps involved in O₂ transport from the external environment (Di Prampero, 1985; Wagner, 1996). The abandonment of the idea of single rate-limiting steps in favour of the concept of shared control of O₂ flux is supported both by theoretical calculations and by evidence demonstrating changes in resulting from perturbations experimentally induced at various steps. Given shared control of O₂ fluxes, it is not surprising to find up-regulation of O₂ flux capacities at multiple steps at all levels of organization in hummingbirds. Lung O₂-diffusing capacities, estimated by morphometric techniques, are estimated to be 8.5 times those of mammals of similar body mass (Dubach, 1981). Hummingbird hearts are twice as large as would be predicted on the basis of allometric relationships (Schmidt-Nielsen, 1984), while heart rates of approximately 1300 min⁻¹ have been measured (Lasiewski, 1964). Hematocrits, O₂ capacities and unloading efficiencies are high (Johansen et al. 1987), and pectoral muscle capillary volume densities are 2–6 times higher than in mammalian hindlimb muscles (Mathieu-Costello et al. 1992; Suarez et al. 1991).

Among the steps involved in the transport of O₂, the carrier-free zone between the red cell surface and the muscle membrane is thought to be a region of particularly high impedance (Honig et al. 1997). Thus, although intravascular is high, intracellular (i.e. muscle fiber) is thought to be much lower on the bases of theoretical calculations and estimates using various techniques. These have included microcryospectrophotometry (Gayeski et al. 1987) and ¹H-NMR spectroscopy (Richardson et al. 1995) to estimate the degree of O₂ saturation of myoglobin, as well as in vivo infrared spectroscopy to determine cytochrome redox state (Duhaylongsod et al. 1993). Hummingbird flight muscles consist of type II fibers with unusually small cross-sectional areas (Grinyer and George, 1969); each fiber in cross section is surrounded by an average of five capillaries (Mathieu-Costello et al. 1992) (Fig. 2). These result in large capillary surface areas relative to muscle fiber surface areas and, along with small diffusion distances between the sarcolemma and muscle mitochondria, result in significantly enhanced capacities for O₂ flux. In contrast with mammals, hummingbirds hovering under hyperoxic conditions do not increase values beyond those measured under normoxic conditions (Chai et al. 1996). This supports the suggestion of higher capacities for O₂ flux in hummingbirds than in mammals.

Myoglobin, which is thought to function as an important carrier of oxygen at high intracellular O₂ flux rates, occurs at high concentrations in hummingbird flight muscles (Johansen et al. 1987). Mitochondrial volume densities are approximately 35%, which is among the highest known for vertebrate locomotory muscles (Suarez, 1992; Suarez et al. 1991). A substantially greater fraction of these mitochondria are localized under the sarcolemma than in mammalian skeletal muscles, further lessening intracellular diffusion distances for O₂. It has been suggested that this clustering of mitochondria adjacent to capillaries may result in further enhancement of capacities for O₂ flux (Mainwood and Rakusan, 1982). The mitochondrial cristae, where electron transport, O₂ consumption by cytochrome c oxidase and ATP synthesis via oxidative phosphorylation take place, occur at a surface density (surface area per unit volume) of 58 m² cm⁻³ (Suarez et al. 1991) (Fig. 3). This is between 1.5 and three times greater than the range of cristae surface densities found in mammalian skeletal muscles. Mitochondrial oxidative capacities in hummingbird flight muscles are therefore enhanced through more mitochondria and more inner membrane per mitochondrion.

In a historic deviation from the traditional concept of single rate-limiting steps in biochemical systems, metabolic control theory proposed, and metabolic control analysis has demonstrated, that the control of carbon flux is shared by multiple enzyme-catalyzed and membrane-carrier-mediated steps in pathways (Kacser and Burns, 1973; Fell, 1992). In recent years, much progress has been made in putting metabolic control theory into practice in the study of the regulation of muscle energy metabolism. In a landmark study, Kashiwaya et al. (1994) showed that the control of glucose utilization by isolated, perfused rat hearts is shared by the glucose transporter and various enzyme-catalyzed reactions in glycolysis. The relative degree of control exerted by several steps was shown to change by providing insulin or an alternative source of acetylCoA. In chemically skinned rat soleus, control analysis has been used to quantify the degree of control over O₂ consumption exerted by actomyosin ATPase as well as the mitochondrial reactions and transport processes involved in oxidative phosphorylation. Consistent with the idea that rates of ATP hydrolysis determine rates of ATP synthesis, a flux control coefficient of 0.5 was estimated for actomyosin ATPase, while the rest of the control coefficients (adding up to a value of approximately 1) were found to be distributed among the mitochondrial reactions (Wisniewski et al. 1995). In vivo, muscles are characterized by their ability to undergo dramatic changes in metabolic flux rates with little change in the concentrations of pathway intermediates (Fell and Thomas, 1995; Thomas and Fell, 1996). Theoretical calculations and experimental manipulation of the level of expression of genes encoding single enzymes (e.g. Schaaff et al. 1989) indicate that changing the enzyme activities at single steps cannot account for such large changes in flux, leading to the proposal of multi-site modulation of metabolism (Fell and Thomas, 1995; Thomas and Fell, 1996). Given such distribution of control, one would expect that evolutionary up-regulation of flux capacities would be achieved not by the up-regulation of single (supposedly ‘rate-limiting’) enzymes or transporters, but rather through a concerted up-regulation of many reactions and transport steps in biochemical pathways. Consistent with this expectation, coordinated regulation of the expression of genes coding for glycolytic enzymes has, in fact, been observed (e.g. Robin et al. 1984; Webster, 1987).

V_max values (equal to [E]×k_cat, where [E] is enzyme concentration and k_cat is catalytic efficiency) serve as appropriate measures of flux capacities through pathways (Newsholme and Crabtree, 1986; Newsholme et al. 1980; Suarez, 1996). It has been recognized for decades that tissues capable of sustaining high rates of metabolic flux possess high enzyme V_max values throughout the relevant pathways. Accordingly, hummingbird (Suarez et al. 1986, 1990) and insect (Suarez et al. 1996) flight muscles possess high V_max values for many of the enzymes involved in the pathways of energy metabolism. An important issue from a mechanistic perspective is how homologous enzymes function in muscles across species: at what fractional velocities (v/V_max) do they work in vivo during exercise? From the perspective of evolutionary design, one may ask how capacities are related to maximum physiological loads; are these closely matched or do pathways possess ‘too much’ enzyme?

Enzymes catalyzing near-equilibrium reactions are typically found at V_max values 2–3 orders of magnitude higher than net forward flux rates (v) in vivo (Gitomer and Veech, 1988; Veech et al. 1969). However, it has been known at least from the time of Haldane’s work on the subject (Haldane, 1930) that net forward flux at near-equilibrium reactions is possible only if V_max values greatly exceed v. Honeybee flight muscles possess V_max values for the enzyme phosphoglucoisomerase (PGI) that are approximately 20-fold greater than the rate of glycolysis during flight (Suarez et al. 1996). We have recently shown (Staples and Suarez, 1997) that these V_max values are very close (within 5%) to what can be predicted using the Haldane equation, given the intracellular conditions in which the enzyme functions, its affinities for substrate and product, and the net glycolytic flux required for flight. These findings provide evidence in support of the biological relevance of enzyme kinetic parameters as well as the near-equilibrium nature of the PGI reaction during flight in honeybee flight muscles. Although it is claimed that such close matches between capacities and maximum physiological loads are unlikely to be the outcome of natural selection (Dudley and Gans, 1991; Garland and Huey, 1987), honeybees possess, in Diamond’s words (Diamond, 1991), ‘enough, but not too much’ PGI.

Using data available from the literature as well as our own, we examined the relationships between flux rates (v) in vivo and V_max in various muscle types at the hexokinase, glycogen phosphorylase and phosphofructokinase steps, all nonequilibrium reactions in glycolysis (Suarez et al. 1997). The pattern that emerged reveals that, in species where maximum flux rates are relatively low, enzymes operate at very low v/V_max, whereas in muscles that achieve the highest flux rates, close matches between v and V_max are observed. This pattern implies, first, that homologous enzymes operate at higher fractional velocities in high-flux muscles and, second, that excess capacities are large at the low-flux end of the range but become smaller or even non-existent in species whose muscles sustain the highest flux rates.

If the high /M_b during flight in hummingbirds and insects can be fully explained by the high mitochondrial volume densities and cristae surface densities found in their flight muscles, rates of O₂ consumption per unit cristae surface area should be similar to those estimated in mammalian skeletal muscles during exercise at ⁠. However, contrary to this expectation, hummingbirds and insects display higher rates of O₂ consumption per unit cristae surface area than mammals (Suarez, 1992). This is consistent with the suggestion of higher capacities for O₂ delivery. Also supporting this suggestion is the observation that cytochrome c oxidase, the terminal (O₂-consuming) enzyme in mitochondrial electron transport, operates closer to V_max in the muscles of flying honeybees than in the skeletal muscles of mammals exercising at (Suarez et al. 1996). In more recent work, we have determined that cytochrome turnover rates (i.e. electron flux rate per cytochrome molecule) in flying honeybees are higher than in mammalian skeletal and cardiac muscles during maximal aerobic exercise (R. K. Suarez, J. F. Staples and J. R. B. Lighton, unpublished observations).

Thus, a pattern common to glycolytic and mitochondrial enzymes has started to emerge from our comparisons of capacities with maximum loads (i.e. flux rates) at the biochemical level.

The distribution of the control of O₂ and carbon fluxes among multiple steps at all levels of biological organization and the evolutionary enhancement of flux capacities at most, if not all, of these steps in highly aerobic animals necessarily leads to the expectation that there should be many factors that might constrain the up-regulation of capacities and ultimately set upper limits to design and performance.

It has been proposed that the upper limits to the design of O₂ transport systems may set the ultimate upper limits to maximal /M_b among vertebrate homeotherms (Schmidt-Nielsen, 1984). It is possible that mass-specific cardiac outputs cannot increase further, given the already large heart mass relative to body mass and heart rates measured in shrews and hummingbirds. Further increases in hematocrit may be constrained by consequent increases in blood viscosity.

Because biological structures occupy space, the need to accommodate components that are larger or more numerous within a given volume might also set limits to the extent to which functional up-regulation is possible. Capillaries occupy approximately 10% of flight muscle volume in hummingbirds (Mathieu-Costello et al. 1992; Suarez et al. 1991). Hummingbird flight muscles possess 25 km of capillary length and 3500 cm² of capillary surface area per milliliter of mitochondrial volume compared with 10 km ml⁻¹ and 1400 cm² ml⁻¹, respectively, in mammals (Hoppeler et al. 1997). Among vertebrates and insects possessing synchronous flight muscles, volume densities of sarcoplasmic reticulum increase with average contraction frequencies (Josephson and Young, 1987; Rome et al. 1996). At the higher contraction frequencies that characterize the flight of hummingbirds and such insects, a greater fraction of the ATP synthesized by mitochondria would be used by the sarcoplasmic reticulum Ca²⁺-ATPase. This may contribute towards the low energetic efficiencies of such muscles and set a limit to how far evolution can go in the direction of higher frequencies and metabolic rates. Mitochondrial volume densities in hummingbird flight muscles are amongst the highest known for vertebrate locomotory muscles (Suarez, 1992; Suarez et al. 1991). Higher mitochondrial volume densities would result in less space for myofibrils and the other cytosolic components required for contractile function. Thus, higher mitochondrial volume densities are only found in modified muscles that do little or no mechanical work (e.g. brain and eye heater organs in billfish and tunas) (Block, 1991). Similarly, mitochondria occupy approximately 45% of cell volume in shrew cardiac muscles (Weibel, 1985). This is not much higher than those in hummingbird flight muscles and similar to the volume densities found in the flight muscles of bees (Casey et al. 1992) and flies (Suarez, 1992). Within the mitochondria, cristae surface densities in hummingbird and insect flight muscles are close to the upper limit imposed by the need for matrix space to accommodate Krebs cycle enzymes (Srere, 1985; Suarez, 1992). It has been estimated that the protein components involved in electron transport and oxidative phosphorylation occupy more than 40% of cristae surface area in liver mitochondria (Schwerzmann et al. 1986). Because respiratory chain enzymes interact like ships randomly colliding in a sea of phospholipid, higher concentrations make possible higher collision frequencies and higher capacities for electron transport (Schneider et al. 1980). Such higher enzyme packing densities and higher flux capacities are found in muscle mitochondria (Schwerzmann et al. 1989). However, membrane protein packing densities probably cannot be increased without limit (Fig. 4) without adverse consequences, e.g. molecular traffic jams and excessively high energetic cost of protein turnover.

At a given rate of dietary intake, increased allocation of material (e.g. proteins and fats) towards hypertrophy of certain organs and the enhancement of a specific function will result in less material being available for other organs and functions. Rates of feeding and digestion can, of course, be increased. However, even dietary intake rates cannot be increased without limit (Diamond and Hammond, 1992; Hammond and Diamond, 1997). The extent to which material costs might constrain the up-regulation of /M_b has not been considered. It has been argued that large organs are energetically expensive to build and maintain. Among birds, active individuals displaying higher ‘field metabolic rates’ possess larger metabolically active organs and display higher basal metabolic rates than individuals with lower rates of metabolism (Daan et al. 1990). Flightless rails on islands without predators are thought to have lost the ability to fly in their evolution from mainland ancestors (Trewick, 1997). Smaller, less aerobic flight muscles are presumed to result in energetic savings. In juvenile ‘grower’ chickens, artificial selection for higher rates of muscle growth has resulted in greater pectoral muscle mass and more rapid synthesis of the enzyme glycogen phosphorylase (Flannery et al. 1992). Protein synthesis costs 5 ATP equivalents per peptide bond (Hawkins, 1991). On this basis, it can be calculated that both larger muscle mass and higher rates of synthesis contribute towards a 10-fold greater cost of glycogen phosphorylase synthesis than in ‘layer’ chickens of the same age (Suarez, 1996).

Animals are so well integrated in their design that, by studying them within the confines of traditional disciplines (e.g. biochemistry, organismal physiology, cell biology), we hardly do better than the six blind men groping at various parts of the elephant. Fig. 1, a schematic diagram of a hummingbird modified from Weibel’s cartoon of a frog (Weibel, 1985), shows the pathway of O₂ from the external environment, through the lungs and cardiovascular system and into the muscle mitochondria. The cycle frequency of the flight muscles determines the rate of ATP hydrolysis which, in turn, determines the rate of mitochondrial ATP synthesis. These determine the steady-state rate of O₂ flux through the system. In the same way that a clock will not work if its parts do not fit properly, integrated animal function (e.g. the hovering flight of the hummingbird) would be precluded by improperly matched functional capacities. Even the hummingbird small intestine is up-regulated in its capacity for glucose transport (Diamond et al. 1986) to meet the great demands of its fuel-hungry metabolic machinery.

There are, no doubt, many different factors that contribute to establishing the ultimate upper limits to /M_b among vertebrates and insects. Similarly, multiple factors probably contribute to set the lower limit to adult body mass among vertebrate homeotherms. Because /M_b values at rest and during maximal exercise, along with associated structures and functional capacities, increase with decreasing body mass, the factors that set the upper limits to /M_b must also be involved in setting the lower limit to vertebrate homeotherm body mass. It is significant that the smallest vertebrate homeotherms, including Thai bumblebee bats, Etruscan shrews and Cuban bee hummingbirds, all weigh approximately 2 g as adults. One would imagine that smaller birds and mammals would have evolved, if this were possible. Sphingid moths display metabolic rates similar to those of hummingbirds of the same body mass (Bartholomew and Casey, 1978). However, species of sphingid moths exist which weigh much less than 2 g and achieve higher body-mass-specific rates of metabolism than hummingbirds. Many species of flying insects, possessing either synchronous or asynchronous muscles, exceed the apparent limits to vertebrate homeotherm size and metabolic rate during flight. It is thought that these metabolic rates, probably the highest in the Animal Kingdom, are made possible via a higher-capacity tracheal system for O₂ transport (Weis-Fogh, 1964) along with enhanced capacities for fuel catabolism (Sacktor, 1976).

The arguments presented by various authors and summarized here are mainly qualitative. Clearly, we have barely scratched the surface, and much remains to be done to clarify the relationships between capacities and loads in animals. Detailed studies to explore design limits and constraints in nature’s most metabolically active animals have barely begun. It is likely that control analyses at the organismal (e.g. Wagner, 1996) and biochemical (e.g. Kashiwaya et al. 1994) levels, as well as continued efforts towards integration, will play important roles and lead to further understanding in the years to come.

I thank Drs P. W. Hochachka, J. R. B. Lighton, O. Mathieu-Costello, J. Harrison, J. F. Staples, B. Joos and C. L. Gass for the collaborative efforts that made possible the research on which this paper is based. Discussions with them, as well as with R. K. Josephson and J. M. Diamond, are also gratefully acknowledged. I thank O. Mathieu-Costello for providing the electron micrographs for Figs 2 and 3, and Springer-Verlag for granting permission to use the cartoon shown in Fig. 4. My work is supported by the U.S. National Science Foundation (IBN 9507407).