Age-related changes in the metabolism and body composition of three dog breeds and their relationship to life expectancy

Summary

We measured body composition and resting metabolic rates (RMR) of three dog breeds (Papillons, mean body mass 3.0 kg (n = 35), Labrador retrievers, mean body mass 29.8 kg (n = 35) and Great Danes, mean body mass 62.8 kg (n = 35)) that varied between 0.6 and 14.3 years of age. In Papillons, lean body mass (LBM) increased with age but fat mass (FBM) was constant; in Labradors, both LBM and FBM were constant with age, and in Great Danes, FBM increased with age but LBM was constant. FBM averaged 14.8% and 15.7% of body mass in Papillons and Labradors, respectively. Great Danes were leaner and averaged only 10.5% FBM. Pooling the data for all individuals, the RMR was significantly and positively associated with LBM and FBM and negatively associated with age. Once these factors had been taken into account there was still a significant breed effect on RMR, which was significantly lower in Labradors than in the other two breeds. Using the predictive multiple regression equation for RMR and the temporal trends in body composition, we modelled the expenditure of energy (at rest) over the first 8 years of life, and over the entire lifespan for each breed. Over the first 8 years of life the average expenditure of energy per kg LBM were 0.985, 0.675 and 0.662 GJ for Papillons, Labradors and Great Danes, respectively. This energy expenditure was almost 60% greater for the smallest compared with the largest breed. On average, however, the life expectancy for the smallest breed was a further 6 years (i.e. 14 years in total), whereas for the largest breed it was only another 6 months (i.e. 8.5 years in total). Total lifetime expenditure of energy at rest per kg LBM averaged 1.584, 0.918 and 0.691 GJ for Papillons, Labradors and Great Danes, respectively. In Labradors, total daily energy expenditure, measured by the doubly labelled water method in eight animals, was only 16% greater than the observed RMR. High energy expenditure in dogs appears positively linked to increased life expectancy, contrary to the finding across mammal species and within exotherms, yet resembling observations in other intraspecific studies. These contrasting correlations suggest that metabolism is affecting life expectancy in different ways at these different levels of enquiry.

Introduction

Among the oldest of the current theories of why we age is the ‘rate of living’ theory (Rubner, 1908; Pearl, 1928), which suggests that increases in metabolism of individuals shortens their lifespan. Comparisons of the metabolic rates and lifespans of different mammalian species provide support for this hypothesis. Rubner (1908) noted that the mass-specific rate of metabolism decreases as mammals become larger, concomitant with an increase in their lifespans. More refined measures of mass-specific metabolic rate across a wide range of mammals (Brody, 1945; Kleiber, 1961) revealed that the interspecific scaling exponent for metabolic rate was around −0.27, whereas the scaling of lifespan in mammals was around +0.29 (Sacher, 1977). The product of the two traits, expressing the lifespan expenditure of energy per gram of tissue, is consequently virtually independent of mass across a broad spectrum of mammals (Calder, 1984), a so-called life history invariant (Charnov, 1993). The ‘rate of living’ theory was further strengthened when Harman (1956) proposed a mechanism whereby oxygen consumption might be linked to aging and lifespan. Harman (1956) suggested that free-radicals, produced as a by-product of oxidative phosphorylation, damage macromolecules, leading to physiological attrition (aging) and ultimate failure (death). Conceptually the ‘metabolism – free-radical – aging’ hypothesis is extremely appealing (Arking, 1998; Beckman & Ames, 1998; Sohal, 2002; Sohal et al., 2002) and amenable to experimental testing (Golden et al., 2002), although it is not universally accepted (see debate in Jacobs, 2003a,b; Pak et al., 2003a,b).

More recently, however, the growing consensus around the idea has started to collapse. It was noted that there are many exceptions to the fixed ‘amount of living’ estimates derived from scaling relationships in mammals. Most conspicuously, birds (Aves) combine high rates of metabolism with long lifespans (Lindstedt & Calder, 1976; Holmes & Austad, 1995a,b; Pamplona et al., 1996; Holmes et al., 2001) whereas other groups such as marsupials (Marsupialia) combine low metabolic rates with short lives (Austad & Fischer, 1991). Even among Eutheria there is considerable variability (Austad, 1997, 2000), with bats (Chiroptera) showing exceptional longevity and high lifetime expenditures of energy compared with rodents (Rodentia) (Austad & Fischer, 1991). Primates, like bats, also exhibit high lifetime energy expenditure (Austad & Fischer, 1992). These exceptions have been considered to be so extreme that they make the whole notion of the’rate of living’ theory untenable.

It is worth noting, however, that in its original formulation the ‘rate of living’ theory is based on comparisons of energy expenditure between individuals within a species (Pearl, 1928) rather than comparisons between species, genera or higher taxonomic levels. Put simply, the theory suggests that if an individual were to expend more energy it would increase the rate of damage and attrition and die earlier (Pearl, 1928). The demonstration that there appears to be a fixed ‘amount of living’ across a wide range of body masses using data from different species (Rubner, 1908) does not therefore support the original idea, but suggests only that it may have more general applicability. The existence of grade shifts in the relationships linking metabolism to lifespan in different classes and orders (such as observed in the birds, marsupials and bats – see References above) or insignificant relationships across small groups of species (Promislow & Haselkorn, 2002) do not therefore refute the original theory. They merely point to limitations in its generality. Indeed, grade shifts are a common feature of many biological systems.

Allometric relationships between biological variables and body mass are generally fitted to models of the form y = aM^b, where a is the scaling coefficient, b is the scaling exponent, y is a biological trait of interest and M is body mass. The form of this relationship is such that if both sides of the equation are converted to logs the resultant equation log_e(y) = log_e(a) + b · log_e(M) has a linear form where ‘a’ is the intercept on the y-axis and b is the gradient of the relationship. For example, many interspecific relationships linking metabolism to body mass, where the animals under consideration exhibit a large range of body sizes, have scaling exponents (b) between 0.5 and 0.8 (Gillooly et al., 2001, 2002; White & Seymour, 2003). This relationship is so robust that it has been suggested that it should be considered a fundamental biological law (West et al., 1997, 1999; West, 1999). Yet between different taxonomic groups there are clear grade shifts (differences in the y intercept ‘a’) such that, even within endotherms at a given body mass, the metabolic rates of birds are almost 50% greater than those of equivalent sized mammals (Kleiber, 1961; Reynolds & Lee, 1996). Yet no one would suggest that these grade shifts refute the whole notion of a link between metabolism and body mass, in the manner that their presence in the relationship between metabolism and lifespan has been considered to refute the ‘rate of living’ hypothesis.

In fact, within individual species, support for the ‘rate of living’ idea is very strong – particularly in exothermic species. There have been many demonstrations that manipulations of ambient temperature, which has large effects on the metabolic rates of exotherms, results in alterations in lifespan in the appropriate direction, (e.g. in houseflies) (Ragland & Sohal, 1975). More subtle experimental manipulations that alter activity levels (Yan & Sohal, 2000) or the energy demands of activity (Wolf & Schmid-Hempel, 1989) are also consistent with the theory. Moreover, some studies have suggested that some mutant strains of Drosophila (Trout & Kaplan, 1970) and Caenorhabditis elegans only exhibit elevated longevity by virtue of their lowered metabolism (Van Voorhies & Ward, 1999; Van Voorhies, 2001), although this is an issue of hot debate (Braeckman et al., 2002a,b; Van Voorhies, 2002a,b).

In endotherms, however, the situation is far less clear. Some manipulations are consistent with the original hypothesis – animals forced to work harder die sooner (Daan et al., 1996). Other experimental manipulations of metabolism, however, have failed to provide such support (Holloszy & Smith, 1986). Caloric restriction, which is the experimental manipulation that most consistently generates an increase in lifespan (Masoro, 1993; Masoro & Austad, 1996; Sohal & Weindruch, 1996; Weindruch & Sohal, 1997), appears to involve either no change or even an increase in metabolic intensity (metabolism per gram of tissue) in mammals (McCarter & Palmer, 1992; Ramsey et al., 2000). Moreover, there have been several recent studies of dwarf strains of mice (that presumably have higher metabolic intensities) and these generally also have greater lifespans than the parent strains (Brown-Borg et al., 1996; Coshigano et al., 2000; Flurkey et al., 2001, 2002; Hsieh et al., 2002a). Speakman et al. (2000) observed a positive relationship between metabolic intensity (independent of body mass) and lifespan in individual mice of a single outbred strain (MF1) and Miller et al. (2002) observed a similar inverse link between body size and longevity in a different heterogenous strain of mice.

Across dog breeds, smaller dogs (with presumably higher mass-specific metabolic rates) live longer than large dogs (Li et al., 1996; Miller, 1999; Flurkey et al., 2001). Dog breeds provide a particularly valuable endothermic model to test the ‘rate of living’ and other aging hypotheses. Although domestication of the dog occurred around 6000 years ago, it is only within the last 300–400 years that the enormous diversification of breeds has occurred, leading to unparalleled variations in body mass, metabolism and lifespan within a single species. Mass, for example, varies from 1.4 kg in the Chihuahua to over 100 kg in the St Bernard. Populations of dogs are large, and because many owners insure their pets, the records of dates of birth, death and hence lifespan (Patronek et al., 1997; Michell, 1999; Egenvall et al., 2000; Proschowsky et al., 2003) are probably better than for any other species apart from humans. We have less information, however, on the metabolic rates of dogs, and, although some estimates have been made (Rubner, 1883; Kunde & Steinhaus, 1926; Singer et al., 1993; Scantlebury et al., 2000, 2001), single time-point estimates of metabolic rate for given groups (like breeds) miss out on potential complexities that may occur because of age-related variations in metabolism (Promislow & Haselkorn, 2002). Here, we examine the age-related changes in the metabolic rates of three breeds of dog (Papillons, Labrador retrievers and Great Danes) of different body sizes, and compare these with breed differences and age-related changes in body composition, as well as differences in lifespan of the three breeds.

Results

Body composition

Age-related changes in body composition were different between the three species. In Papillons, fat body mass (FBM) was highly variable between individuals and did not have a significant association with age (Fig. 1a: F_1,33 = 0.68, P = 0.42) but there was a significant increase in lean body mass (LBM) as animals became older (Fig. 1b: F_1,34 = 5.14, P = 0.03, r² = 0.134). Despite the absence of a significant trend in FBM with age there was a strong positive association between LBM and FBM across individuals (Fig. 1c: F_1,33 = 14.3, P = 0.001, r² = 0.302). In Labradors, neither LBM nor FBM were significantly associated with age (F_1,33 = 1.29, P = 0.64; F_1,33 = 0.36, P = 0.551, respectively, Fig. 1d,e), but LBM and FBM were significantly and positively associated with each other (Fig. 1f: F_1,33 = 5.92, P = 0.021, r² = 0.152). In Great Danes, LBM was not significantly related to age (Fig. 1g: F_1,33 = 2.01, P = 0.65) but FBM was (Fig. 1h: F_1,33 = 8.32, P = 0.007, r² = 0.201) with older Great Danes being significantly fatter. On average, young (1–2 years old) Great Danes had about 5 kg of fat but by the age of 7 years this had increased to around 8 kg, with some individuals carrying as much as 12 kg. As with the other breeds, LBM and FBM were also significantly and positively associated with each other (Fig. 1f: F_1,33 = 48.3, P < 0.001, r² = 0.594).

Figure 1. Plots of the relationships between FBM and age, LBM and age and between LBM and FBM in dogs of three breeds (Papillon, Labrador and Great Dane) (n = 35 in all breeds). In all cases the variables are normalized by log transformation.

Body fatness averaged 14.8% (SD = ±2.6) in the Papillons and was similar at 15.7% (SD = ±3.0) in the Labradors, neither of which varied significantly with age. Great Danes were on average leaner with only 10.4% (SD = ±1.4) body fat, but this was significantly related to age (F_1,33 = 8.75, P = 0.006, r² = 0.21), increasing from 8% in young to 12% in older dogs.

Resting metabolic rate (RMR)

We pooled the RMR data across all individuals of all three breeds (n = 105). We investigated the effects of lean and fat body masses, age, breed, sex and castrate status on resting metabolism using a generalized linear model, with all the factors except breed, sex and castrate status entered as covariates. None of the two-way interactions were significant, and neither were the main factors, sex and castrate status. Breed was a significant factor (F_1,104 = 3.58, P = 0.032) and analysis of the residuals suggested that Labradors had significantly lower RMRs than either Papillons or Great Danes (Fishers pairwise comparisons, P < 0.05) once all the other effects had been accounted for. Papillons and Great Danes did not differ from each other significantly. We therefore generated a dummy variable (1 for Labradors and 2 for the other two breeds) and entered this with the other traits into a multiple regression analysis to generate a predictive equation for metabolism as a function of breed, age and body mass. The resultant equation:

• log_e(RMR kJ/day) = 6.19 + 0.681 log_e(LBM kg) + 0.220 log_e(FBM kg) − 0.162 log_e(age years) + 0.162 breed(1)

explained 96.3% of the observed variation in resting metabolic rate of these dogs. The significance of the respective factors was (LBM, t = 7.2, P < 0.001; FBM, t = 2.06, P = 0.042; age, t =−5.12, P < 0.001; breed, t = 2.69, P = 0.008; d.f. = 104). Plots of the major LBM and FBM effects are illustrated in Fig. 2. We removed these effects and then plotted metabolic rate as a function of age (Fig. 3), which illustrates the clear negative effect of age on resting metabolism in these dogs.

Figure 2. Relationships between RMR and (a) LBM and (b) FBM in data pooled across 105 dogs from three breeds (Papillon, Labrador and Great Dane; n = 35 for each breed).

Figure 3. The relationship between RMR, with the effects of both LBM and FBM removed (residual RMR), and age. Data are for 105 dogs drawn from three separate breeds (Papillon – closed circles, Labrador – open circles and Great Dane – closed squares; n = 35 for each breed).

Daily energy expenditure

We used the doubly labelled water (DLW) technique to measure the total daily energy expenditure in eight Labrador dogs. These were different individuals from those involved in the RMR measurements. The average body mass of the dogs during the DLW measurement period was 28.7 kg, which was slightly lighter than the animals for which we measured RMR (29.9 kg), but not significantly so (t = −1.22, P = 0.36, d.f. = 36). On average, using the two-pool model for calculation of daily energy demands, the energy expenditure was 6642 kJ day⁻¹ (SD =±2124), which was 1.16 times greater than the mean RMR (5694 kJ day⁻¹). On average then RMR amounted to around 86% of the total daily energy expenditure in these dogs.

Lifetime metabolic rate

We obtained estimated mean life expectancies of two of the three species from the literature, which suggested that in the UK the median lifespans are: Labrador, 12.6 years and Great Dane, 8.4 years (Michell, 1999). Using web-based resources primarily generated by breeders of Papillons, we estimated the equivalent median lifespan of Papillons is 13–15 years and use here a figure of 14 years. Although this source of information is not ideal it is unlikely to be erroneous by more than 1 year. Using Eq. (1) in combination with the information on body composition changes as a function of age (Fig. 1), we reconstructed the accumulated total expenditure of energy over the first 8 years of life (Table 1). This calculation suggests that by the age of 8 years the average Papillon would have expended 2.48 GJ, the average Labrador 16.92 GJ and the average Great Dane 36.9 GJ of energy when at rest. All of this energy expenditure does not occur in lean tissue, as indicated by the significant effect of fat mass in Eq. (1). However, because the intercept of the equation is not zero we could not model the absolute contributions that the different body compartments make to the total. Assuming most energy expenditure occurs in the lean tissue, the expenditures of energy amount to 0.985 GJ kg⁻¹ LBM in the Papillon, with Labradors expending 0.675 and Great Danes 0.662 GJ kg⁻¹ LBM.

Table 1.  Calculated average resting expenditures of energy over the first 8 years of life for three dog breeds. In all cases the trends in body composition as a function of age were predicted from the fitted relationships (Fig. 1). The RMR is then predicted as a function of age using Eq. (1) (see text) and the total resting expenditure of energy summed over the 8 years of life is shown in bold type

Age (years)	log LBM	log FBM	RMR (kJ day^−1)	RMR (kJ year^−1)	LBM (kg)	kJ/year/ kg LBM
Papillons
1	0.86	−0.83	1006.6	367 403	2.36	155 658
2	0.90	−0.83	926.9	338 324	2.47	137 194
3	0.93	−0.83	883.1	322 341	2.53	127 406
4	0.95	−0.83	853.6	311 545	2.58	120 920
5	0.96	−0.83	831.2	303 384	2.61	116 104
6	0.97	−0.83	813.4	296 876	2.64	112 311
7	0.98	−0.83	798.6	291 481	2.67	109 201
8	0.99	−0.83	786.0	286 887	2.69	106 577
				2 518 242		985 371
Labradors
1	3.22	1.49	7 144	2 607 403	25.1	104 012
2	3.22	1.49	6 385	2 330 460	25.1	92 965
3	3.22	1.49	5 979	2 182 306	25.1	87 055
4	3.22	1.49	5 707	2 082 934	25.1	83 091
5	3.22	1.49	5 504	2 008 983	25.1	80 141
6	3.22	1.49	5 344	1 950 508	25.1	77 808
7	3.22	1.49	5 212	1 902 405	25.1	75 889
8	3.22	1.49	5 101	1 861 692	25.1	74 265
				16 926 692		675 226
Great Danes
1	4.02	1.57	14 710	5 369 312	55.7	96 347
2	4.02	1.71	13 565	4 951 403	55.7	88 848
3	4.02	1.80	12 938	4 722 188	55.7	84 735
4	4.02	1.86	12 510	4 566 016	55.7	81 932
5	4.02	1.90	12 188	4 448 455	55.7	79 823
6	4.02	1.94	11 931	4 354 640	55.7	78 140
7	4.02	1.97	11 717	4 276 877	55.7	76 744
8	4.02	2.00	11 536	4 210 629	55.7	75 555
				36 899 520		662 124

Given the mean lifespans of the animals, we also calculated the total average lifetime expenditure of energy (at rest) per kg LBM. This revealed that by the end of an average Papillon's life, the total expenditure of energy on resting metabolism per kg LBM would have amounted to 1.584 GJ, whereas for the average Labrador it would be 0.918 GJ and for the average Great Dane 0.691 GJ. If the estimated ratio between resting and daily energy demands measured here for Labradors pertains to all breeds under common housing conditions, then actual total energy fluxes would be about 16% higher at 1.84, 1.06 and 0.80 GJ kg⁻¹ LBM, for Papillons, Labradors and Great Danes, respectively. Figure 4 summarizes the mean body masses, RMRs, metabolic intensities, maximum lifespans, along with the 8 years and lifetime expenditures of energy for each of the three breeds.

Figure 4. Bar charts summarizing for each of the three dog breeds their mean body masses (kg), RMRs (RMR – kJ/day), metabolic intensities (the resting energy expenditure expressed per kg of LBM – kJ/kg, LBM/day), maximum lifespans (years), energy expended over the first 8 years of life per kg LBM (EE8 – MJ/kg LBM/8 years) and the lifetime expenditure of energy per kg LBM (EElife – MJ/kg LBM/life).

Discussion

Body composition

The largest breed in our study (Great Dane) progressively gained body fat as the animals became older. This phenomenon has been frequently observed in humans, although there do appear to be some race-related differences in the extent to which it occurs. Similar trends are also apparent in some smaller mammals such as mice and rats (Miller et al., 2002). The cause of the change in body fatness with age is generally presumed to be a result of sustained energy imbalance either because levels of activity or RMR progressively decline with age and hence total expenditure is reduced, or because there is a progressive increase in energy intake. We did not monitor activity levels in the dogs, but all the dogs in this study were kept in a standard housing unit where levels of exercise are regulated closely by staff who walked the dogs daily. This procedure may have minimized any age-related changes in activity levels. Moreover, the daily energy demands monitored by DLW in one of the species (Labradors) suggested that exercise energy expenditures were at most 14% of the daily energy budget, so scope for alteration of this component of the budget was relatively small.

By contrast, resting metabolism comprised about 86% of the total metabolic rate in Labradors and thus age-related modulations in this component of the budget might be more significant. Because there was an age-related reduction in metabolic rate in our dogs, this could have contributed to the age-related increase in body fatness. However, similar patterns of age-related decline in metabolism were apparent not only in the Great Dane, but also in the other two breeds we studied, and in these two species there was no evidence of a progressive expansion in body fatness. This suggests either that age-related modulations in RMR are not an important feature of the age-related accumulation of body fat, or that different breeds respond to the changes in resting metabolism differently – with some species compensating their energy intake to match the changed metabolic demands, whereas others do not, with a consequent elevation in fat storage. Differences in the age-related responses of dog breeds may provide a valuable model to assess the signalling factors that associate expenditure and intake.

Very elderly humans tend to lose body weight, which reflects both a decline in fat and in lean body tissue. Similar phenomena have been observed in old rodents (Black et al., 2003). In all three of the breeds we studied there was no evidence of any age-related decline in either lean or fat tissue, and indeed in the smallest breed (Papillon) there was an age-related accumulation of lean tissue. The causes of the changes in body composition in the elderly are complex (Roberts & Dallal, 1998; Roberts, 2000) but in addition to biological factors related to regulation of body mass, and sensory effects on satiety, there appears to be a strong influence of social factors – such as isolation and depression, which would not be significant factors in these dog populations, perhaps explaining why no age-related decline in lean or fat tissue was observed in our study. In addition, the phenomenon in rodents appears to occur only very late in life (Black et al., 2003), and there were no very old dogs in our sample.

Resting metabolic rate

We observed a significant decline in resting metabolism with age in all three breeds. The effect was such that the metabolic rates of the oldest dogs were on average only half those of dogs that were 1–2 years old. An age-related decline in resting metabolism has been reported many times for humans (Poehlman, 1993; Poehlman et al., 1993a,b; Van Pelt et al., 1997, 2001; Hughes et al., 1998; Neuhauser-Berthold et al., 2000), although such declines are not universally observed (e.g. Promislow & Haselkorn, 2002; Sukhotin et al., 2002). The cause of the age-related decline in metabolic rate is not entirely clear. However, an important contributory factor is generally presumed to be the age-related decline in lean body tissue and organ mass (Hughes et al., 1998). In our case we can discount this possibility because there was no evidence in any of the three breeds that such changes in LBM were apparent (in fact the opposite in Papillons), yet the decline in age-related metabolism was observed in all three breeds, pointing to a more complex causality.

Several other factors appear to contribute to the age-related decline in RMR. Pumping of ions is known to decline as we get older (Poehlman, 1993; Poehlman et al., 1993a,b) as is the rate of protein synthesis (Short & Nair, 2001; Volpi et al., 2001; Dorrens & Rennie, 2003). Whether these are causal or correlated effects, however, remains unclear. The problem is that we are as yet still uncertain about the precise functional nature of all the energy that is expended on RMR. In particular, the decline of resting metabolism (per kg body mass) with body size poses a lasting enigma. There is a constraint on the maximum rate of metabolism imposed by heat dissipation capacity, which is a function of the surface-to-volume ratio. This constraint may prevent larger animals from having high RMRs (see, for example, Kleiber, 1961), because the scope for further increases in metabolism would be restricted. However, it is not immediately clear why the RMRs of smaller animals, within the thermoneutral zone, are so high, because the constraint permits them to have elevated metabolism, but does not oblige them to do so.

An interesting possibility then is that the size-related differences in RMR under thermoneutral conditions reflects at least in part the energy that is devoted at rest to defence and repair mechanisms. The age-related decline in metabolic rate may then be a consequence of a systematic shutdown of the defence and repair mechanisms with age. Because many of these systems are enzymatic in nature and thus require protein synthesis, this interpretation would be consistent with the observed age-related decline in protein synthesis rates. Moreover, many studies have documented a decline in oxidative defence and repair mechanisms with age. The decline in metabolic rates with age may then be a consequence of the declining force of selection as we age. Older animals may have lower metabolic rates because they invest progressively less in defence and repair mechanisms. More puzzling in this respect then are the animals in which age-related declines in metabolism have not been reported (Promislow & Haselkorn, 2002; Sukhotin et al., 2002).

It is perhaps significant that those species in which declines have not been observed are exothermic post-mitotic species for which the energy devoted to protection and repair may be a relatively minor component of the energy budget. The suggestion that individual differences in resting metabolism within a species reflect the energy devoted to defence and repair processes is also consistent with the disposable soma theory of aging, which hinges on the selective allocation of resources between the processes of somatic repair/protection and reproduction. If this trade-off is significant we would expect to be able to quantify the costs of somatic protection and repair.

Metabolism and lifespan

The suggestion that a portion of RMR reflects the cost of defence and repair mechanisms and that differences between individuals within a species in RMR reflect different allocations to defence and repair leads to some interesting predictions. In particular, we would expect lifespan at the individual level within species to be positively associated with variations in resting metabolism, thus apparently directly contradicting the ‘rate of living’ theory. Yet this appears to be exactly what happens in the different dog breeds exemplified here by the three breeds we have studied. By the age of 8 years, the smallest breed has expended almost twice the amount of energy per gram of lean tissue compared with the largest breed, and at that stage can expect to live a further 6 years compared with an additional life expectancy averaging only 6 months in the largest breed. Furthermore, it has been repeatedly observed that dwarf strains of mice have increased lifespans (Brown-Borg et al., 1996; Flurkey et al., 2002; Hsieh et al., 2002a,2002b) and that larger mice have shorter lives (Miller et al., 2002). Although we still await measurements of metabolic rates in these dwarf strains, their small size and favourable surface-to-volume ratios would certainly permit them to have greater metabolic intensities (expenditure per gram of lean tissue). Finally, Speakman et al. (2000) found that metabolic rates of individual mice of a single strain were positively linked to lifespan.

There are, however, other possible interpretations of these data. Rather than the costs of metabolism reflecting the costs of defence and repair, which are permitted to be elevated in smaller strains, the high metabolic rates of smaller breeds may reflect the capacity to generate heat when faced with a metabolic challenge such as cold exposure. During cold exposure the surface-to-volume ratio that permits high resting rates in smaller animals starts to require high rates of expenditure to balance heat loss – a function of surface area and the temperature-driving gradient. Because smaller animals need to be perpetually prepared for periods of cold exposure, they may need a relatively high RMR at thermoneutral temperatures (Speakman, 2000).

Much of the thermoregulatory energy expenditure comes from the action of uncoupling proteins that decouple proton transport from ATP generation, thereby directly generating heat. It has been suggested that oxygen free-radical production in mitochondria depends critically on the longevity of ubisemiquinone in the Q-cycle at complex III (Brand, 2000), which may depend on the extent to which mitochondria are uncoupled (Brand, 2000). RMR in smaller animals may be associated with more uncoupled mitochondria than in larger animals by virtue of their need to be ready for the energy demands concomitant with cold exposure. From this they gain the additional benefit of lowered oxygen free-radical production and lowered damage to lipids and DNA and thus increased lifespan. This latter interpretation is consistent with the likelihood that proton leakage contributes a large part to the total resting metabolism (Porter & Brand, 1993; Rolfe & Brand, 1996), although we do not know the contribution made by defence and repair mechanisms. Moreover, the decline in uncoupling with age would result in a lowered age-specific metabolic rate and would be consistent with the age-related decline in thermoregulatory capacity. Finally, the apparent fundamental differences between exo- and endothermic animals come sharply into focus with this interpretation because exotherms by definition do not generate endogenous heat in the cold.

These interpretations suggest that there are fundamentally different processes underpinning the relationships between metabolic rate and lifespan in exo- and endotherms and also between endotherms at different levels of comparison. The differences in the nature of metabolic rate differences between a small dog like the Papillon weighing 3 kg and a large dog like the Great Dane weighing 65 kg may be radically different from the differences between a capucin monkey (Cebus capucinus) and a human, although the weight differences are around the same order (3.5 kg for the Capucin and 70 kg for the human) and the actual measured metabolic rates may be similar. The key questions revolve around the differences in what the metabolism is being used for. This may help explain the initially perplexing observation that the small dog lives longer than the large dog, whereas the lifespan of the capucin monkey is substantially shorter than that of humans.

Finally, we should not discount the possibility that the positive linkage between RMR and lifespan in different dog breeds, in dwarf strains of mice and within single mouse strains, contrasted with the negative association observed within many exothermic species (and in comparisons across different species of mammal), may simply reflect the fact that the ‘energetics – free-radical – damage – aging’ theory is largely incorrect. For example, it has been pointed out that the differences in body size in mice reflects differences in expression of Insulin-like Growth Factor-1 (IGF-1: Flurkey et al., 2001; Miller et al., 2002), differences that may also underpin the differences in body sizes of dogs (Flurkey et al., 2001). Moreover, expression of IGF-1 has been strongly implicated in the aging process (Gems & Partridge, 2001; Carter et al., 2002a,b; Shimokawa et al., 2002; Bartke et al., 2003). Hence aging and body size may be linked only because of the shared effects of IGF-1 and not causally associated for any other reason.

Experimental procedures

Study animals

Measurements were made on three breeds of dog (Papillons, Labradors and Great Danes) kept at the Waltham Centre for Pet Nutrition, Leicestershire, UK. We studied 35 individuals of each breed. The Papillons ranged in body weight from 1.9 to 3.98 kg and averaged 3.02 kg (SD = ±0.46). The Labradors ranged in body weight from 23.9 to 41.4 kg and averaged 29.8 kg (SD = ±4.2), whereas the Great Danes ranged in body weight from 48.2 to 81.0 kg and averaged 62.8 kg (SD = ±8.7). The animals were of mixed sex and castrate status. The animal ages ranged from 0.6 to 13.8 years for the Papillons, 2.0 to 14.3 years for the Labradors and 1.4 to 10.2 years for the Great Danes. The animals were kept in housing facilities that have been described elsewhere (Scantlebury et al., 2001) and were fed proprietary dog foods at levels appropriate for their breed and size, given ad libitum water and were excercised daily.

Body composition

Measurements of LBM and FBM were made using a DXA scanner (Lunar Hologic pencil beam scanner QDR-1000W). Dogs were anaesthetized and then placed on their sides under the scanner. Scans took about 20 min to complete, after which the anaesthetic was reversed and the dogs returned to their living quarters. We have previously validated use of the DXA scanner using the exact same protocols against whole body carcass composition in both dogs and cats (Speakman et al., 2001a), showing that it provides an accurate measure of the lean and fat tissue contents of dogs over the range from at least 2 to 35 kg. We did not validate the machine for dogs the size of Great Danes, but the very strong linearity in the relationships at the lower mass range leads us to be confident that the estimates for the Great Dane are not biased. Scans were analysed using the standard analysis version 6.20A software, which was also used in the validation.

Resting metabolic rate

We made measurements of RMR using a mask-based respirometry system (Europa Instruments). Dogs were first introduced to the mask and then over a period of several sessions were trained to sit quietly with it secured over their face for a period of about 30 min. We took two measurements for every individual and then analysed the data only from the second measurement period when the dogs had become familiar with and unconcerned by the equipment. Measurements were made at room temperature (22 °C), which is within the thermoneutral zone for all these breeds. Excurrent air from the mask was dried and then routed via a mass flow controller to a gas analyser, which monitored the oxygen and CO₂ content of the expired gases at 30-s intervals and generated from the resultant estimates of oxygen consumption and respiratory quotient (RQ), an estimate of the energy expenditure of the animal. We used data from the lowest 5 min of continuous measurement.

Daily energy expenditure

Daily energy expenditure was measured using the DLW method (Speakman, 1997) in eight Labradors. The animals were injected via a forelimb vein with an isotonic mix of labelled water containing both ¹⁸O and ²H, following collection of a blood sample to establish background isotope enrichments. A blood sample was collected from the contralateral forearm at 6 h post-injection, established previously (Speakman et al., 2001b) as the time required for the injected isotopes to reach complete isotopic equilibrium in the body water of dogs of this size. We collected blood at 24-h intervals for a further 4 days. Blood samples were distilled using the pipette method (Nagy, 1983). Mass spectrometric analysis of ²H enrichment was performed using H₂ gas, produced from distilled water after reaction with LiAlH₄ (Ward et al., 2000). Reactions were performed inside 10-mL Vacutainers (Beckton Dickinson Ltd) as detailed in Krol & Speakman (1999). For analysis of ¹⁸O enrichment, distilled water was equilibrated with CO₂ gas using the small sample equilibration technique (Speakman et al., 1990). Preweighed Vacutainers were injected with 10 µL of distilled water and re-weighed (±0.0001 g) to correct for differences in the amount of water added. Subsequently the Vacutainers with the samples were injected with 0.5 mL CO₂ with a known oxygen isotopic enrichment, and left to equilibrate at 60 °C for 16 h.

²H : ¹H and ¹⁸O : ¹⁶O ratios were measured using dual inlet gas source isotope ratio mass spectrometry (Optima, Micromass IRMS), with isotopically characterized gases of H₂ and CO₂ (CP grade gases, BOC Ltd) in the reference channels. Reference gases were characterized every 3 months relative to SMOW (standard mean ocean water) and SLAP (standard light Antarctic precipitate) supplied by the International Atomic Energy Agency. Analysis of standards distributed to other laboratories showed good comparability with their mean results. Each batch of samples was run with triplicates of three laboratory standards to correct for day-to-day variation in performance of the mass spectrometers. All isotope enrichments were measured in delta (per mil, ) relative to the working standards and converted to p.p.m. using the established ratios for these reference materials. Measures of isotope enrichment were based on independent analysis of two subsamples of the water distilled from the blood samples.

Because our previous validation study in Labradors (Speakman et al., 2001b) had indicated the two-pool model was superior, we used this model to estimate CO₂ production. The error in individual estimates was determined using the iterative procedures outlined in Speakman (1995). Conversion to daily energy expenditure (DEE) was made using the measured average RQ of 0.8. All calculations were made using the Natureware DLW software (available at www.abdn.ac.uk/zoology/jrs.htm).

Acknowledgment

We are grateful to Peter Thomson for technical assistance with the isotope analysis for the doubly labelled water study.