Gluttony and Thermogenesis Revisited{

Michael J Stock1

1Department of Physiology, St George's Hospital Medical School, University of London, London SW17 0RE, UK

The evolutionary and biological signi®cance of adaptive, homeostatic forms of heat production (thermogenesis) is reviewed. After summarizing the role and selective value of thermogenesis in body temperature regulation (shivering and non-shivering thermogenesis) and the febrile response to infection (fever), the review concentrates on diet- induced thermogenesis (DIT). Animal studies indicate that DIT evolved mainly to deal with nutrient-de®cient or unbalanced diets, and re-analysis of twelve overfeeding studies carried out between 1967 and 1999 suggests the same may be so for humans, particularly when dietary protein concentration is varied. This implies that the role of DIT in the regulation of energy balance is secondary to its function in regulating the metabolic supply of essential nutrients. However, individual differences in DIT are much more marked when high- or low-protein diets are overfed, and this could provide a very sensitive method for discriminating between those who are, in metabolic terms, resistant and those who are susceptible to obesity.

Keywords: diet-induced thermogenesis; non-shivering thermogenesis; energy balance; overfeeding; brown adipose tissue; protein; evolution

Introduction Cafeteria feeding, DIT and obesity

The title of this review echoes that of two papers The relevance of any study of BAT thermogenesis to (Gluttony 1 and Gluttony 2)1,2 published in 1967 that the regulation of energy balance and obesity depends described the effects of overfeeding humans low- and crucially on being able to demonstrate that thermo- high-protein diets. The second of these, dealing with genesis makes an important quantitative contribution the effects of overfeeding on thermogenesis, was my to the regulation of energy balance, and that defective ®rst scienti®c paper and the start of an interest in thermogenesis results in obesity. Until the early thermogenesis that still remains, more than 30 years 1980s, there was very little interest in DIT. Thermo- later. A personal high point in the past 30 years was genesis rarely, if ever, appeared in textbooks, reviews the 1979 Nature paper with Nancy Rothwell3 that or conferences on energy balance regulation and identi®ed brown adipose tissue (BAT) as the likely obesity. Only a few research groups were actively source of diet-induced thermogenesis (DIT). It would investigating the topic, and even these were mainly seem that this paper, and others linking impaired BAT interested in non-shivering thermogenesis (NST) and thermogenesis with obesity,4,5 had a major effect in thermoregulation. Thus, the use of the cafeteria diet in stimulating scienti®c interest in BAT. As shown in the 1979 Nature paper to induce voluntary hyperpha- Figure 1, the number of BAT papers being published gia and stimulate DIT in rats represented an important rose rapidly from approximate 50 per year prior to advance and provided the raison d'eÃtre for invoking 1979 to 150 ± 200 per year thereafter, with our Nature BAT as the effector of this form of thermogenesis. paper accumulating over 1000 citations. While one Table 1 shows the energy balance results from a cannot deny feeling pleased by having helped spark typical cafeteria-feeding experiment, and illustrates this upsurge of interest in brown fat research, I have the remarkable capacity for DIT in young adult rats. always felt that the fascination with BAT de¯ected It can be seen that in spite of a 73% increase in attention away from our important demonstration that voluntary energy intake in the cafeteria-fed rats, this DIT could exert a greater impact on the regulation of had very little effect on the rate of body energy gain energy balance than had hitherto been suspected. because there was an almost equivalent increase in energy expenditure, with 90% of the excess dietary energy being dissipated as heat. These, and many similar experiments, provided unequivocal quantitative evidence of the importance of DIT in the regula- { Review based on the EASO Wasserman Prize Lecture delivered tion of energy balance. at the 9th European Congress of Obesity, Milan, June 1999. It still seems remarkable that simply feeding a Correspondence: Professor MJ Stock, Department of Physiology, St George's Hospital Medical School, Tooting, varied and palatable cafeteria diet to induce voluntary London, SW17 0RE, UK. E-mail: [email protected] hyperphagia can produce more than a 50% increase in Gluttony and thermogenesis revisited MJ Stock 1106 pressing a lever which dispenses food for itself and for its obese partner in an adjacent cage.9 A much more convincing demonstration of reduced DIT contributing to obesity comes more or less by chance from a study by Trayhurn et al in cafeteria-fed lean and genetically-obese (ob=ob) mice.10 Energy balance data from this study are shown in Table 2, where the effects of cafeteria feeding on voluntary hyperphagia and energetic ef®ciency in the lean mice are seen to be very similar to those shown for rats in Table 1 Ð i.e. hyperphagia stimulates energy expenditure rather than increasing body energy stores, and there is a signi®cant decrease in energetic ef®ciency. This contrasts with the response to cafeteria feeding in Figure 1 Number of papers dealing with brown adipose tissue the ob=ob mice where there is only a modest increase appearing each year since 1966, as revealed by a Medline in energy expenditure, a large increase in energy search. deposition and an increase, rather than a decrease, in energetic ef®ciency. This effect of genotype on the response to voluntary Table 1 Energy balance and thermogenesis in rats hyperphagia is quite remarkable and was the main Chow diet Cafeteria diet focus of that paper, but what is equally remarkable is Intake (kJ=d) 208 360*** the comparison of the cafeteria-fed lean mice with the Body gain (kJ=d) 33 50 chow-fed ob=ob mice. These two columns have been Expenditure (kJ=d) 175 310*** highlighted in Table 2 since they show that Trayhurn Net ef®ciency (%) 34 20*** 10 % Excess intake expended 89 et al had, perhaps without realizing it, conducted a pair-feeding experiment in which the lean mice had *** P < 0.01 vs Chow. voluntarily consumed the same amount of energy as Adapted from Rothwell and Stock.6 the obese mice Ð that is, this was pair-feeding that was free of experimental artifacts resulting from metabolic rate, and perhaps even more remarkable restriction of energy intake and disruption of normal that the phenomenon had gone undetected, ignored or meal patterns. This comparison shows that despite the discounted for so long. Likewise, the impact of same energy intake, the obese mice stored nearly three defective DIT in the aetiology of rodent obesity had times more body energy than the lean mice and also been ignored, even though there had been several illustrates the enormous impact that increased meta- pair-feeding studies on both genetic7 and experimen- bolic ef®ciency (that is decreased DIT) can have by tal8 models to show that obesity could develop in the itself on the development of obesity. This emphasizes absence of the hyperphagia Ð i.e. as a result of the fact that hyperphagia is not necessary for obesity reduced DIT and the consequent increase in energetic to develop in the ob=ob mice, and that the primary ef®ciency. However, one criticism of these sorts of lesion is not in the control of intake, but in the control experiments is that controlling the hyperphagia of the of energy expenditure. Since this is due entirely to the obese animal by restricting its intake to the level of absence of leptin in the ob=ob mutant, it should the normal, lean control can by itself increase ener- reinforce the concept of leptin as a thermogenic getic ef®ciency, due to the restricted animal eating a hormone. One suspects that the predilection of most smaller number of larger meals. However, this experi- workers to label leptin as a satiety hormone is mainly mental artefact can be avoided by performing yoke- due to the fact that it is very easy to measure food feeding experiments, where the amount and frequency intake, whereas it is quite dif®cult and demanding to of food eaten is determined by the control animal measure metabolic ef®ciency.

Table 2 Cafeteria feeding in genetically lean and obese (ob=ob) mice

Lean Obese

Chow Cafeteria Chow Cafeteria

Intake (kJ=d) 65 109*** 104 154* Gain (kJ=d) 9 11 31 58*** Expenditure (kJ=d) 56 98*** 73 96** Net ef®ciency (%) 24 13*** 44 50*

* P < 0.05, **P < 0.01, ***P < 0.001 vs chow. Adapted from Trayhurn et al.10 Gluttony and thermogenesis revisited MJ Stock 1107 The ease or dif®culty with which one can mea- Evolution of thermogenesis sure things can often in¯uence the development of scienti®c ideas. In the same way that most of the research on the role of leptin in body-weight reg- One has to step back from DIT and obesity and ulation has concentrated on food intake (even mole- consider thermogenesis in a much broader biological cular biologists can measure food intake), so did the perspective to assess its evolutionary signi®cance and study of thermogenesis and obesity tend to concen- value in natural selection. To begin with, it should be trate on measuring BAT function. Following the noted that thermogenesis has very primitive origins, identi®cation of the BAT mitochondrial proton con- and is not restricted to homeotherms. Table 3 lists ductance pathway11 and its unique uncoupling pro- examples of thermogenic mechanisms in lower life tein (UCP),12 it became relatively easy to measure forms ranging from bacteria to ®sh, and including changes in BAT mitochondrial function and thermo- plants, with the increased expression of UCP in cold- genic activity using a variety of techniques such as exposed potatoes being the botanical equivalent of GDP-binding and UCP-radioimmunoassays. These non-shivering thermogenesis (a hot potato!). This and other more straightforward biochemical mea- shows that thermogenesis is, in evolutionary terms, surements explain the enthusiasm for studying BAT not at all new, and so no-one should be surprised that in relation to energy metabolism and obesity, and the phenomenon has persisted and operates even in why there was such a large increase in the number the most highly evolved biological organisms Ð e.g. of BAT papers appearing in the early 1980s (see humans. Figure 1). However, acceptance of the importance In homeotherms, by de®nition, the defence of body of DIT and variations in energetic ef®ciency, on temperature is essential for survival, and the key to which all these BAT studies depend if they are to their successful occupation of a wide range of habitats be relevant to obesity, was much less enthusiastic that include some very hostile temperature environ- and in certain quarters there was quite vociferous ments. With the exception of aquatic mammals that and hostile repudiation of the idea that DIT was use their thick layers of insulating fat (blubber) to important, or even existed.13 reduce heat loss, most animals rely on increased heat In some ways, one could perhaps understand why production (NST) to protect themselves against the many biologists ®nd it dif®cult to accept that cold or reverse the hypothermia induced by cold, as in mechanisms exist that convert precious food hibernators during arousal. This is not the place to energy directly to heat Ð i.e. mechanisms that describe and review NST in detail, but it is important `deliberately' waste energy. There is widespread to emphasize that practically every mammalian spe- belief, almost a dogma, among biologists that cies has the capacity for NST and, at least at one assumes that animals have evolved in a world crucial stage in life (as neonates) are almost entirely where food is always limited, and that natural dependent on NST for thermal protection. While this selection would favour those that could conserve also includes human neonates, it is generally consid- the most energy by having the most ef®cient meta- ered that NST does not operate in adults. The reason bolic processes. Thus, individuals with the thrifty for this is that adults rarely, if ever, expose themselves genotype would survive, while those with the to the cold for long enough to allow physiological spendthrift genotype would become extinct. How- thermoregulatory adaptations to be recruited. Cold is ever, this does not explain why the most thrifty an unpleasant sensation, and people modify their species or strains are relatively rare (for example behaviour (for example clothing, shelter, heating, Spiny mice, Pima Indians) and suggests that there etc.) to keep warm, although they will resort to must be some evolutionary survival value in having shivering to deal with acute, unavoidable exposure thermogenic mechanisms, even though they waste to the cold. However, the fact that people have no energy. Obviously, such mechanisms would have to need to adapt to the cold because they avoid chronic be adaptive rather than obligatory, since they would cold-exposure, does not mean that adult humans are need to be switched off when food supplies are completely incapable of NST. Nevertheless, it limited or when body-energy reserves have to be requires somewhat heroic chronic cold-chamber accli- increased, as in pregnancy or prior to hibernation or mation experiments to show this. An example of one migration. such study14 (Figure 2) shows that shivering activity

Table 3 Thermogenesis in lower life forms

Example Mechanism

Energy `spilling' in bacteria (E. coli; K. aerogenes) Carboxylation=decarboxylation `futile' cycle (?) Volatilizing insect attractant in spadix of Arum maculatum Cyanide-resistant non-translocating electron chain Cold-exposed potatoes (Solanum tuberosum) Mitochondrial uncoupling protein (StUCP) Bumble bee ¯ight muscle (Bombus terrestris) Fructose phosphate `futile' cycle Bill®sh brain and eye heater (marlin, sword®sh, sail®sh, etc) Fast SR Ca2-ATPase of superior rectus eye muscle (?) Gluttony and thermogenesis revisited MJ Stock 1108 envelope. Given these possible bene®ts, it has been argued by several authorities (for example Kluger16) that fever has an overall adaptive value, and could explain why it occurs not only in humans and other homeotherms, but also in poikilotherms, such as lizards, ®sh, leeches and even insects. As with NST and DIT, the febrile response to infection in mammals involves sympathetic activation of BAT thermogenesis.17 However, the similarities between these different physiological phenomena go further than this since the anorexia and increased heat production following central injection of leptin has been shown Figure 2 Effects of one month's cold-acclimation in humans on to involve the increased expression of IL-1 and is 14 metabolic rate and shivering activity. Adapted from Davis. blocked by IL-1 receptor antagonists.18 Thus, it appears that the regulation of energy balance by declined very rapidly over the ®rst week of cold leptin utilises the same, or similar mechanisms acclimation, but the cold-induced elevation in thermo- involved in the febrile response to infection. It also genesis (metabolic rate) was sustained throughout the emphasizes the point made earlier about recognizing one-month study, thereby implicating substitution by leptin's status as a thermogenic hormone. NST of the initial shivering thermogenesis. Apart In addition to thermal protection against cold and from this, there is also evidence to show that the the febrile response to infection, the other main histological appearance of BAT in cadavers of out- evolutionary advantage of thermogenesis is to provide door workers (Finnish lumberjacks) shows evidence a mechanism for enriching nutrient-poor diets by of greater activity than BAT taken from indoor work- disposing of the excess non-essential energy. This ers, and this difference was greatest in those outdoor effect, of increased DIT on unbalanced diets, had workers dying during the winter months.15 been recognized since the 1920s, and perhaps was In addition to cold-induced thermogenesis, another best described by Kleiber19 who stated that `a diet is aspect of thermoregulation that helps explain the de®cient in any nutrient whose addition decreases the evolutionary persistence of thermogenesis is fever. calorigenic effect of that diet'. Kleiber went on to cite Fever is produced by a combination of reduced heat the increased combustion of fat and carbohydrate in loss (for example vasoconstriction) and increased heat response to low-protein diets as an example of production (both shivering and NST) in order to `homeostatic waste', and what follows is a more produce an elevation in the `set-point' temperature. detailed description and discussion of the activation Even though fever is an unwanted, unpleasant and of this homeostatic waste (i.e. DIT) by low-protein often dangerous response to infection, there are poss- diets. ible bene®cial effects. These include increased neu- trophil migration, interferon activity and polymorpho- nuclear leucocyte bactericidal activity, as well as decreased bacterial growth due to hypoferraemia and Low-protein diets and DIT decreased synthesis of the bacterial outer protective Kleiber was able to draw on a variety of sources to support his concept of homeostatic waste, but perhaps one of the most convincing is the study by Hamilton20 who measured the effect of varying dietary protein concentration on the heat increment of feeding in rats (`heat increment' is an alternative term for DIT). Hamilton's results are summarized in Figure 3, and show that a diet providing approximately 20% protein

Table 4 Energy intake for weight maintenance

To t a l i n t a k e (M J =40 d)

Low-protein diet 199 High protein diet 41 Difference 158 Fat Equivalent (kg) 4.1

Adapted from Miller and Payne.22 Figure 3 Effects of varying dietary protein concentration (% LP and HP diets were 2% and 25% (%ME). metabolisable energy) on the heat increment of feeding in rats. Both LP and HP pigs weighed 4.5 kg at beginning and end of 40 d Lines drawn by hand to link maximum values to minimum value. experiment. Adapted from Hamilton.20 39 MJ=kg used to calculate fat equivalent. Gluttony and thermogenesis revisited MJ Stock 1109 produces the highest ef®ciency of energy utilization, Table 5 Effect of normal vs low-protein cafeteria diet on DIT and with the heat increment of feeding increasing (and BAT thermogenic activity ef®ciency decreasing) as the protein concentration Normal Low decreases. The increase in the heat increment (DIT) protein protein as the diet becomes unbalanced and de®cient with Dietary protein (%) 23 7 respect to protein supports Kleiber's view about Energetic ef®ciency (%) 45 33*** nutrient de®ciencies and homeostatic waste, while BAT GDP-binding (nmol=depot) 2.8 5** BAT aGP shuttle (activity=depot) 236 361*** the increase on high-protein diets is attributable to Plasma triiodothyronine 70 110** the high metabolic cost of metabolizing protein. This (ng=100 ml) effect of protein on heat production was ®rst described ** P < 0.01, ***P < 0.001 vs normal protein. 21 nearly 100 years ago by Rubner, is usually referred GDP-binding mitochondrial guanosine diphosphate binding. to as `Speci®c Dynamic Action' of protein, and is a aGP mitochondrial a-glycerophosphate dehydrogenase form of obligatory, rather than adaptive or facultive activity. Adapted from Rothwell et al.24 DIT. The extent to which low-protein diets could affect energetic ef®ciency was not fully recognized until entire body weight Ð that is as much fat as there was Miller and Payne22 used two weanling pigs to com- pig! pare the effects of restricting protein intake on the It was the effect of low-protein diets on energetic energy cost of weight maintenance with restricting ef®ciency, such as those described by Hamilton20 and energy intake. In this rather bizarre experiment, one Miller and Payne,22 that led Stirling and Stock23 to pig was allowed to eat ad libitum a diet with a protein investigate the metabolic origins of diet-induced ther- concentration so low that however much it ate it could mogenesis in rats fed low-protein diets. As shown in only take in suf®cient protein to meet its maintenance Figure 4, two possible thermogenic mechanisms were requirement, which meant that growth was imposs- investigated, and there was evidence that both had ible. By contrast, the high-protein pig was fed a been activated in the low-protein rats exhibiting DIT. standard, high-protein weaning diet that would nor- Thus, in addition to the 30-fold increase in the activity mally produce rapid growth if fed ad libitum. How- of the hepatic a-glycerophosphate shuttle (a thermo- ever, this pig's food intake was restricted such that the genic `futile' cycle), the low-protein rats also exhib- animal could only just maintain weight Ð i.e. growth ited a marked, 6-fold increase in the thermogenic was limited by energy. As a result of this dietary response to noradrenaline. This is very similar to the manipulation, the low-protein pig was found to increased thermogenic responsiveness seen in cold- require almost 5-times more energy to maintain the adapted rats exhibiting NST, indicating that both NST same body weight as the high-protein pig. It was not and DIT involve sympathetic activation of heat pro- possible to carry out a proper energy balance, but it is duction. Thus, as early as 1968, there was evidence to quite obvious from the results shown in Table 4 that if suggest that DIT and NST involved similar effector the low-protein pig had not converted most of the mechanisms, but it was not until more than a decade extra energy it consumed to heat, it would have later (1979) that this idea gained general acceptance deposited an amount of fat almost equivalent to its when DIT was linked with sympathetic activation of BAT thermogenesis.3 Establishing this link between DIT and BAT thermogenesis prompted a re-investigation of the effects of low-protein diets, only this time a comparison was made between rats overeating low- and high-protein cafeteria diets.24 As shown in Table 5, allowing rats ad libitum access to a variety of palatable, but low- protein cafeteria food items had the expected effect on energetic ef®ciency, and increased BAT mitochondrial uncoupling (GDP-binding) and the activity of the a-glycerophosphate shuttle in BAT, the latter possibly

Table 6 Thermogenesis and BAT in marmosets and rats

Marmoset Rat

Dietary protein (%) 6 20 Body weight (g) 250 300 NA response (VO2 %increase) 80 45 Figure 4 The thermogenic response (increase in oxygen con- BAT (%BWt) 0.34 0.27 sumption, VO2) to noradrenaline in rats fed a low-protein (LP) BAT GDP-binding (pmol=mgP) 65 45 diet to stimulate DIT (HP high-protein control diet). The hepatic mitochondrial a-glycerophosphate dehydrogenase activity in the NA noradrenaline. LP and HP rats was 150 and 4 nmole=mgN=h, respectively. GDP-binding mitochondrial guanosine diphosphate binding. Adapted from Stirling and Stock.23 Adapted from Rothwell and Stock.27 Gluttony and thermogenesis revisited MJ Stock 1110 re¯ecting the increase in circulating triiodothyronine intake, body weight and body composition,1 while levels. the second paper (Gluttony 2) described the effects of In addition to these experimental demonstrations of overfeeding on energy expenditure.2 It was concluded DIT responding to nutrient imbalances, there are that there was clear evidence for increased thermo- several examples that can be drawn from nature. genesis in the volunteers overeating the low-protein Two are of particular interest because they involve diet, but due to the limited or non-existent techniques BAT thermogenesis in two species that live in the available at the time (1960s) for measuring accurately tropics Ð i.e. one can dissociate brown fat activity changes in body composition and energy expenditure completely from any thermoregulatory requirement. (for example human calorimeters, double-labelled The ®rst example is the fruit bat, that eats a diet water, DEXA, etc) there has been considerable resis- containing only 2 ± 5% of energy as protein. The main- tance to accepting this conclusion. Moreover, most of tenance energy intake of these bats is approxim- the human overfeeding studies that followed appar- ately 1300 kJ=kg0.75 per day,25 which is over 3-times ently failed to detect a signi®cant effect on DIT. greater than the mammalian inter-speci®c mean for However, a re-analysis of the original (Gluttony 1) maintenance (420 kJ=kg0.75=d).19 Thus, like Miller and subsequent overfeeding experiments sheds new and Payne's low-protein pig, it seems that the fruit light on the phenomenon, and provides an explanation bat has to eat exceptionally large amounts of energy for the discrepancies between studies carried out by simply to obtain suf®cient protein to meet its require- different research groups. The re-analysis that follows ments. The need to dissipate this excess, non-essential supports the conclusion that humans exhibit homeo- energy probably explains why this tropical animal has static wasting of energy when fed unbalanced diets, such well-developed BAT depots.26 Interestingly, the and may also, just like cafeteria-fed rats, increase DIT BAT depots are metabolically active after feeding, when overfed normal diets. during the day when the bats are roosting and temp- Table 7 summarises results from twelve overfeed- eratures are highest, and become inactive and ®ll with ing studies, starting with the original Gluttony study1 triglyceride at night, when the bat feeds. The other and ending more than 30 years later with that of tropical example is the marmoset, which is mainly Levine et al28 published in 1999. These experiments fructivorous and, as shown in Table 6, also eats a low- employed a variety of methods of varying degrees of protein diet. Compared to a rat of a similar size (Table accuracy and sophistication for measuring changes in 6), the marmoset shows a greater thermogenic energy balance and body composition. Given this response to noradrenaline and has more BAT, and diversity, it was necessary to ®nd some simple, more active BAT than the rat. common method for determining whether there has been any change in energetic ef®ciency and DIT. The method chosen was to calculate the cost of body Human over-feeding experiments weight gain by dividing the total excess energy consumed by the body weight gain. The excess intake was calculated by assuming that the subjects were in Investigations into the protein : energy requirements energy balance on their baseline intake, and does for weight maintenance, such as the pig experiment not account for any effect of weight gain on main- described above,22 led Miller to undertake a human tenance requirements. This estimate of the cost of gain over-feeding study using low- and high-protein diets. in each experiment can then be compared with the The results were described in two papers, the ®rst of cost that would be predicted if there had been no which (Gluttony 1) described the changes in food change in energetic ef®ciency during the overeating

Table 7 Twelve human overfeeding experiments

Overfeeding study P:F:CHO(%ME) Excess (MJ=d) Weight gain (kg=week) Cost of gain* (MJ=kg)

Gluttony1 (1967)1 1 Low-protein 3=46=51 5.0 0.39 112 2 High-protein 15=44=40 5.0 0.93 38 Gluttony 3 (1970)32 3 Nibbling=Gorging 13=52=35 5.7 0.97 44 4 Starch=Sucrose 12=49=39 7.7 1.16 51 5 Low-sodium 6=47=47 5.5 0.67 75 6 Goldman et al (1975)36 Unknown 6.4 1.47 32 7 Norgan and Durnin (1980)37 12=43=45 6.0 1.00 44 8 Webb and Annis (1983)38 Variable 4.2 0.56 57 9 Forbes et al (1986)39 12=42=46 6.0 1.48 28 10 Diaz et al (1992)31 13=42=45 6.2 1.28 34 11 Deriaz et al (1993)40 15=35=50 3.5 0.57 47 12 Levine et al (1999)28 20=40=40 4.2 0.59 60

* Predicted cost would be 30 MJ=kg if gain was 60% fat, or 45 MJ=kg if 100% fat. P : F : CHO protein, fat and carbohydrate as % metabolisable energy intake. Gluttony and thermogenesis revisited MJ Stock 1111 period. The theoretical, or predicted cost of gain will, The difference between the observed cost and the of course, depend on the composition of that weight theoretical cost in the low-protein group gives an gain Ð the relative proportions of fat and fat-free estimate for DIT (67 MJ=kg) that accounts for over mass (FFM). However, because this is unknown or 60% of the excess energy consumed, and should have of dubious accuracy in some of the trials, it cannot be produced an increase in daily energy expenditure of relied upon, and calculations based on two possible 3.4 MJ=d (800 kcal=d). In the paper describing various scenarios have been used instead. energy expenditure measurements (resting, exercising, The ®rst calculation makes the somewhat extreme post-prandial, etc) on the subjects in these Gluttony and unlikely assumption that the gain in body weight experiments,2 no distinction was made between the was entirely composed of fat, at a cost of 45 MJ=kg results obtained on the two different diets. However, it gained. The value of 45 MJ includes 39 MJ for the is interesting to note that 24-h energy expenditure was energy content of the fat, plus an additional 6 MJ to measured in four of the overeating subjects and found allow for the energy cost of depositing that amount of to approximate to energy intake on that day Ð i.e. fat in adipose tissue. This additional 6 MJ is, in itself, these overeating subjects were in energy balance, and an overestimate since it is based on the cost of de novo it turns out that three out of the four were on the low- lipogenesis,29 whereas lipogenesis is normally negli- protein diet. This may have been a fortuitous coin- gible in humans, particularly when overeating high-fat cidence since it should be emphasized that these were (35 ± 50%) diets. The second, more realistic estimate estimates of energy expenditure made on just one day assumes that the gain is made up of 60% fat and 40% out of twenty-one overeating days, and involved FFM at a cost of 30 MJ=kg gained. Apart from the fat making intermittent measurements throughout the gain, this includes an allowance of 10 MJ=kg FFM day and night. that is based on a cost of 52 MJ=kg protein deposited30 and a FFM protein concentration of 20%.31 If one overlooks the differences in methodology and accu- Unpublished gluttony experiments racy in assessing body composition, fat accounted for In the three years following completion of the over- an average of 61Æ 3% of the weight gained in feeding experiments described above, another series Experiments 6 ± 12 listed in Table 7, and so an was undertaken using student volunteers as subjects energy cost of gain of 30 MJ=kg is much more during their summer vacations. After completing the realistic than the 45 MJ=kg for a gain of 100% fat. ®nal experiment, all three were combined in a paper Even so, because it could be argued that the estimated that went through several drafts, but, unfortunately costs shown in Table 7 (last column) depend on the and for a variety of reasons, was never submitted for accuracy of calculating the excess energy intake, a publication. This author's carbon copy of this type- `worse case' value of 45 MJ=kg will be taken to written manuscript remained ®led and forgotten from determine whether there was any evidence for DIT 1970 until just over a year ago when it was redis- in these studies. Having taken this `worse-case' covered while moving of®ce. It is clearly too late now, option, it has to be emphasized that anything above 30 years later, to publish this paper, but the data have 45 MJ=kg means that the remaining excess energy had been analysed and presented in Table 7 in the same to have been dissipated as heat, since it is impossible way as the other published overfeeding studies. For to dispose of that energy in any other way Ð i.e. there convenience, the unpublished manuscript has been is no store in the body with a greater energy density given a reference number.32 than that of fat. The ®rst of these unpublished experiments (Expt 3, Table 7) involved overfeeding a normal diet, but with half the subjects dividing their daily intake into 14 meals (nibblers) and the other half dividing theirs into Gluttony experiments two very large meals (gorgers) in order to study the Experiments 1 and 2 listed in Table 7 show the results effect of meal frequency on DIT. The cost of gain was for the low- and high-protein overfeeding studies in slightly, but not signi®cantly greater for nibblers the ®rst Gluttony paper,1 but it should be emphasized compared to the gorgers (46 vs 42 MJ=kg respect- that the high-protein diet (15% protein) was in fact a ively) and so Table 7 shows the combined average normal diet, and is only high relative to the 3% low- value (44 MJ=kg) which falls just short of the max- protein diet. The cost of gain (112 MJ=kg) in the imum if the weight gain was 100% fat, though well overfed low-protein volunteers was 2.5 times greater above what one would expect if the weight gain were than the theoretical maximum cost, whereas the cost 60% fat. The next experiment (Expt 4, Table 7) also of gain on the high-protein diet (38 MJ=kg) was below involved overfeeding a normal diet, but this time a the predicted maximum, but still above the 30 MJ=kg comparison was made between diets in which the value if the gain was 60% fat. Thus, although the carbohydrate source was principally either starch or evidence for DIT on the high-protein diet is debatable, sucrose. The cost of weight gain was greater than there can be no doubt that there had been a large 45 MJ=kg on both diets (starch 47, sucrose decrease in energetic ef®ciency in the subjects over- 55 MJ=kg), and because these were not signi®cantly eating the low-protein diet. different, the combined average value of 50 MJ=kg is Gluttony and thermogenesis revisited MJ Stock 1112 the observed minus the predicted cost of gain (see above). However, it is considerably less than the estimate of DIT if, as was more likely, the weight gain was less than 100% fat.

Overfeeding studies by other groups One of the few overfeeding studies often claimed to show increased DIT and resistance to weight gain was that carried out on prisoners in Vermont USA by Sims et al.40 Surprisingly, however, a detailed description of this series of experiments35 does not include data to allow the cost of weight gain to be calculated, and the evidence for increased DIT relies mainly on showing a Figure 5 Energy intake and energy expenditure (derived from much greater energy requirement for weight mainte- 24 h heart rate) of subjects on the low-sodium overfeeding nance in experimentally-obese compared to spontane- experiment (Expt 5 in Table 7). Adapted from Stock.33 ously-obese subjects. However, another study was undertaken36 to allow measurements of energy expen- shown in Table 7. The last of these unpublished diture to be carried out, and calculation of the cost of experiments (Expt 5, Table 7) was designed to test weight gain from that experiment gives a value of the idea of homeostatic waste by using a diet that was 32 MJ=kg (Expt 6, Table 7) that is practically identical unbalanced with respect to sodium rather than protein. to that predicted for a gain comprising 60% fat. The The intake of sodium was restricted to 24 mg=MJ, observed gain was actually 66% fat (equivalent to which meant that the total sodium intake was, on 34 MJ=kg), and so there is no evidence for DIT in this average, less than 400 mg=d. As can be seen from study, which is consistent with the investigators' failure Table 7, the very high cost of weight gain (75 MJ=kg) to ®nd any unexplained increase in energy expenditure. supports the concept of unbalanced diets causing high The study by Norgan and Durnin37 involved over- levels of DIT. Subtracting the theoretical cost of feeding six young men a normal diet for six weeks 45 MJ=d from the observed cost of gain gives an (Experiment 7, Table 7). Using the same method to estimated DIT of 2.9 MJ=d (690 kcal=d) that would calculate excess intake as in other experiments listed account for just over 50% of the excess energy in Table 7 gives an estimated cost of gain of consumed. 44 MJ=kg, indicating that there had been no increase Thus, two out of these three unpublished overfeed- in DIT if the fat gain had been 100% fat. However, the ing studies show a cost of gain exceeding the bench- reported gain was 62% fat (equivalent to a theoretical mark value for DIT of 45 MJ=kg, with one falling just cost of 32 MJ=kg), suggesting that there had been a short of this, but well above the lower, more realistic compensatory increase in energy expenditure. This value of 30 MJ=kg. Having said earlier that the data increase could account for approximately 30% of the from these studies remain unpublished, this is not extra energy consumed, and is consistent with the strictly true since measurements of energy expendi- authors' estimate of a gain in body energy equivalent ture were made on the low-sodium overfeeding to 60 ± 70% of the excess energy. experiment and the results published in the author's The overfeeding study by Webb and Annis38 used PhD thesis.33 These measurements were carried out by three groups of four people (2 male, 2 female, with measuring 24-h heart rate with one of the miniatur- one of each being lean and one overweight) on three ized, body-borne and unobtrusive (i.e. socially-accep- different diets, but for simplicity Table 7 shows the table) heart rate recorders that ®rst became available averages obtained by combining all the data. The in the late 1960s. By calibrating each subject once a inter-individual and inter-diet variations are discussed week with simultaneous measurements of heart rate later, but the average cost of gain (57 MJ=kg) shown and energy expenditure at rest and during a variety of here provides clear evidence of an adaptive increase in activities, it was possible to calculate daily energy DIT. Perhaps because these subjects were older (aver- expenditure from the 24-hour heart rate.34 Figure 5 age 46 y) and fatter (half were overweight) than those shows the results for energy expenditure estimated in in other studies, the estimated fat content of the this way during each week of the low-sodium experi- weight gained was quite high (73%), but even this ment, and compares this with the energy intake data. gives a much lower predicted cost of gain (36 MJ=kg) As can be seen, the rise and fall in energy expenditure than the observed cost. The difference between the follows that for energy intake from the ®rst baseline observed and predicted cost suggests that up to 40% week, through three weeks of over-feeding and back of the excess energy consumed had been dissipated as to baseline intake at the end. Compared to the ®rst heat. Compared to this experiment, the overfeeding baseline week, energy expenditure increased on aver- experiments of Forbes et al39 and Diaz et al31 (Expts 9 age by 2.9 MJ=d over the overfeeding period, which is & 10, Table 7) are unequivocal examples of over- exactly the same as the estimate of DIT based upon feeding failing to produce any change in metabolic Gluttony and thermogenesis revisited MJ Stock 1113 ef®ciency, with the observed cost of weight gain being An interesting aspect of the study by Levine et al is well below the 45 MJ=kg criterion for DIT, and the suggestion28 that 60% of the increased heat practically identical to the predicted cost if the gain production is due to `Non-Exercise Activity Thermo- had been 60% fat. genesis' (NEAT). While not denying that NEAT, or in One of the most ambitious overfeeding trials under- ordinary language, ®dgeting, is a component of taken was that by Bouchard and colleagues who energy expenditure that has been ignored in nearly overfed 12 sets of identical twins (i.e. 24 subjects) every study of habitual human energy expenditure, for 6 days per week for 14 weeks. Information on the there is a certain reluctance in accepting that this excess energy consumed and weight gained by the could amount to as much as 1.4 MJ=d (330 kcal=d).28 individual subjects is given in the paper by Deriaz et Other reservations about the magnitude of NEAT al,40 and used to calculate the mean cost of gain concern the way in which it was estimated. Firstly, shown in Table 7 (Expt 11). Using the benchmark NEAT was calculated by difference Ð that is the value of 45 MJ=kg, it can be seen that there must have difference between the change in total energy been a small increase in DIT (observed expenditure (double-labelled water method) minus cost 47 MJ=kg). However, 66% of the weight the change in resting energy expenditure and the gained in this study was fat, with a predicted cost of thermic effect of food. The problem with this, as 33 MJ=kg, which suggests a substantial effect of over- with all calculations `by difference', is that all the feeding on metabolic ef®ciency that could result in errors in the measured parameters accumulate in the approximately 30% of the excess energy intake being difference. The second reservation concerns the dissipated as heat. authors' reliance on the fact that, although they con- The last experiment listed in Table 7 (Expt 12) is strained volitional activity in their subjects to constant the most recent,28 and perhaps one of the most low levels, they assumed that the energy cost (or interesting. This study involved very detailed and ef®ciency) of this activity was unaffected. They sup- thorough measurements of body composition, total ported this assumption by making measurements of energy expenditure and various components of energy exercise ef®ciency, but, however, did this when their expenditure in 16 subjects overfed for eight weeks. It subjects were in the postabsorptive state (Michael also produced some of the best evidence for a Jensen, personal communication). This ignores the decrease in energetic ef®ciency during overfeeding, effect of exercise on post-prandial thermogenesis with an observed energy cost of gain of 60 MJ=kg Ð which, as shown in Figure 6, results in a potentiation well in excess of the 45 MJ=kg if the gain had been of the thermic effect of feeding Ð an effect that 100% fat. In fact, only 50% of the gain was fat, with a becomes even greater as meal size increases (e.g. predicted cost of 28 MJ=kg. The difference between during overeating). However, in spite of these reser- the observed cost and this predicted cost is equivalent vations about the magnitude of NEAT, there can be to an increase in metabolic rate of 2.7 MJ=day little doubt that this overfeeding study by Levine et (645 kcal=d), and close to the 2.3 MJ=d (550 kcal=d) al28 demonstrates an adaptive response to overeating increase measured by the double-labelled water that reduces weight gain to well below what would be method. With a difference of only 0.4 MJ=d predicted if metabolic ef®ciency remained unaltered. (95 kcal=d) in the two values for increased heat Moreover, the double-labelled water measurements production, one feels much more con®dent about show, unequivocally, that the increase in the cost of using the `cost-of-gain' approach for estimating DIT gain is due to an increase in heat production. in other experiments, particularly given accurate body composition data.

Homeostatic waste in humans

The survey of overfeeding experiments (published and unpublished) shown in Table 7 presents a somewhat confusing picture. Using the value of 45 MJ=kg for the cost of gain as the benchmark above which one can claim unequivocal evidence for DIT, only 6 of the 12 studies meet this criterion. However, if one assumes that 60% of the weight gain was fat (i.e. the mean value in Experiments 6 ± 12 in Table 7), then 9 out of the 12 studies exceed the predicted cost (30 MJ=kg) by at least 5 MJ=kg, with only Experiments 6, 9 and 10 failing to show evidence of DIT. This gives a somewhat different perspective on the commonly-held Figure 6 Effect of exercise and meal size on postprandial view that the DIT response to overeating is negligible thermogenesis. Adapted from Miller et al.2 in humans, but still does not explain why the Gluttony and thermogenesis revisited MJ Stock 1114 entirely consistent with the difference in protein requirements for maintenance and growth, respectively. Only half of the experiments shown in Table 7 and Figure 7 provide unequivocal evidence of DIT (that is with cost of gain exceeding 45 MJ=kg), but of course these are mean values for 6 ± 24 subjects in any particular experiment. The extent to which this hides inter-individual variations becomes all too apparent when the individual responses to overfeeding are plotted, as in Figure 8. The differences due to the level of dietary protein, as noted above, can be seen by comparing the higher costs of gain in Experiments 1, Figure 7 Relationship between dietary protein (%metabolisable 5, 8 and 12 (high- and low-protein) with the lower energy) and cost of weight gain in humans. Numbers refer to values seen in most individuals on the other experi- experiments listed in Table 7, but note that values for the three diets studied in Expt 8 have been displayed in the ®gure (Table 7 ments when the diet was balanced with respect to only shows overall mean). Lines drawn by hand to link max- protein. However, it is also obvious that just over 40% imum values to minimum value. This ®gure should be compared of subjects show gains that exceed the 45 MJ=kg with that (Figure 3) shown for rats. threshold. Thus, even in those experiments where the average cost of gain was below or close to this value (for example Experiments 6, 7 and 9 ± 11), there responses are so variable in these different studies. were one or two individuals with a cost of gain greater However, a closer inspection of Table 7 provides a than 45 MJ=kg. Since the Laws of Thermodynamics possible explanation. have to apply to individuals as well as to groups, these The experiments in Table 7 that show the highest subjects must also have responded to overfeeding with cost of gain tend to be those that are unbalanced in an increase in DIT. Of course, if one uses the criterion some way Ð for example Experiment No 1 (low- of 30 MJ=kg, then Figure 8 shows that the vast protein), No 5 (low-sodium), No 8 (includes a high, majority (84%) must have exhibited an increase in 20% protein diet) and No 12 (also 20% protein). The DIT. common feature is the protein concentration, because An intriguing aspect of the individual responses even the low-sodium diet is seen to have been a low- shown in Figure 8 is the greater variation seen in protein diet (6%). This suggests that a similar relation- subjects on the unbalanced diets. This impression is ship might exist between protein and DIT in humans supported by comparing the coef®cients of variation for the mean cost of gain which was, for example, 45, as that demonstrated by Hamilton20 (see Figure 3), and the plot of the results from the human feeding 67 and 58% in Experiments 1, 5 and 12, compared to experiments shown in Figure 7 con®rms this. There is 22, 16 and 18% in Experiments 2, 7 and 9. The a remarkable similarity between the human data and difference in the coef®cient of variation between the rat data, and the only obvious difference is that the Experiments 1 and 2 (45 vs 22%) cannot be ascribed minimum value (optimal for ef®ciency) occurs at 20% to different methodologies because they were carried out simultaneously as part of the same study.1 More- protein in rats compared to 12% for humans. How- 41 ever, since the humans were all adults and the rats over, in the commentary that follows this review, were young and still growing, this difference is Dulloo points out that ®ve of the subjects in the Gluttony experiments switched after 4 weeks of overeating to the other diet. As well as con®rming the larger inter-individual variation in cost of gain seen when these subjects were overfed the 3% protein diet, Dulloo goes on to show that these individual differences are still apparent, but much less obvious, when overeating the 15% protein diet. Presumably it is these small inter-individual differences when consuming a normal diet that help explain why some people are more susceptible to obesity than others, and Dulloo suggests that feeding low-protein diets should make it easier to discriminate between these `easy gainers' those `hard gainers' who are more resistant to obesity. In other words, Figure 8 and Dulloo's analysis41 suggests that overfeeding unbalanced diets can be used to amplify individual genetic Figure 8 Cost of weight gain for individual subjects in all the differences in energetic ef®ciency and susceptibility overfeeding studies shown in Table 7. to obesity. Gluttony and thermogenesis revisited MJ Stock 1115 Fat and thermogenesis on the stoichiometry of glucose conversion to lipid,29 which suggests that when fat stores are replete, lipogenesis in vivo becomes very inef®cient. The predicted, theoretical cost of weight gain in all the Obviously, short-term, aggressive carbohydrate above comparisons included, in addition to the energy overfeeding studies, such as those described above, content of the fat deposited, an allowance of 6 MJ=kg cannot be equated with normal dietary practices, but to cover the cost of depositing that fat in adipose one wonders whether the 100% variation in the tissue. However, this cost only applies if there is de apparent cost of lipogenesis in these two studies (24 novo lipogenesis29 and, as argued above and by many vs 12 MJ=kg) is due to the different experimental others before, due to the high fat content of most protocols, or is typical of the normal population. If modern diets, net lipogenesis in humans is rare. the latter, it might indicate metabolic differences, However, this has not been the case during most of possibly with a genetic basis, in the susceptibility to evolution, or even today in people eating diets very weight gain in later life. low in fat, and the capacity to synthesize fat from Given all the evidence from the animal and human carbohydrate has not disappeared. Nevertheless, in studies reviewed above, a much more cogent case can individuals accustomed to eating high-fat diets, it now be made for the evolutionary survival of DIT as a requires several days of fairly aggressive overfeeding mechanism to ensure an adequate supply of essential with a high-carbohydrate diet as used by Acheson et nutrients while avoiding the risks to survival of al,42 in order to demonstrate net lipogenesis. A recent excessive fat gains. However, on an evolutionary study by Aarsland et al43 used this approach to time-scale, our genes have not had time to deal with determine the contribution of hepatic lipogenesis to the recent increase in the availability of high-fat foods whole-body fat synthesis, but the experiment also resulting from increased agricultural productivity. revealed some interesting effects on energy metabo- Thus, given adequate protein and other essential lism. For this experiment, the subjects stayed in a nutrients, together with high-fat, energy-dense foods, clinical research centre while they were overfed a it is not surprising that the incidence of obesity in liquid high-carbohydrate diet for 4 d by continuous most countries is rising so rapidly. Nevertheless, even enteral (nasogastric) infusion, plus an intravenous if the modern diet provides the worst possible con- infusion of glucose. By Day 4 of the study, the ditions for activation of DIT, and attenuates the indi- subjects were synthesizing 168 g fat per day, but vidual, genetic differences in metabolic ef®ciency, only 2% of this could be accounted for by hepatic this does not mean that these have been obliterated lipogenesis, and the rest was assumed to be synthe- completely. For example, although epidemiological sized by adipose tissue. studies show a correlation between fat intake and The balance between hepatic and adipose tissue indices of obesity (for example body mass index; lipogenesis is an interesting observation, but for those BMI), it has been argued that the variability in this more interested in energy metabolism, the protocol relationship shows that obesity is not an inevitable adopted produced some useful information, although consequence of eating a high-fat diet, and one can ®nd this does require a certain amount of detective work to many normal and under-weight individuals who are estimate the excess energy intake and the increase in habitual high-fat consumers.44 This has prompted energy expenditure (i.e. DIT). Because of various Blundell and colleagues to investigate habitual high- uncertainties in the description of the methods, it is fat and low-fat consumers. possible to arrive at two estimates for the excess In one study,45 this group found evidence to suggest energy intake, and two for the increase in energy that these habitual high- and low-fat consumers may expenditure. However, by adopting the principle of represent distinct behaviour phenotypes with differing always using the lower, more conservative value, the subjective and behavioural responses to dietary chal- estimated excess energy consumed on Day 4 was lenges. However, they also found some interesting 11 MJ=d, and this was associated with an increase in physiological differences in another study involving energy expenditure of 4 MJ=d (60% above baseline). young, lean males habitually consuming a diet pro- Since there was very limited scope for activity in viding either 44% (HF) or 32% (LF) of energy from subjects being continuously infused enterally and fat. In spite of eating approximately 40% more parenterally, and assuming no increase in NEAT, energy, the HF subjects had a similar BMI and this means that DIT accounted for 36% of the %body fat as LF subjects. This suggests that the HF excess intake on Day 4, or 91% of the excess after subjects were less ef®cient than LF subjects, and their deducting the energy that went into the 168 g of signi®cantly higher resting metabolism rate and heart synthesized fat. Another way of looking at this is to rate46 is consistent with this. Without more detailed, calculate the energy cost of lipogenesis from the long-term measurements of energy balance, it is not increase in heat production and net fat synthesis. possible to estimate the extent to which increases in This calculation produces a cost of 24 MJ=kg, and energy expenditure, possibly DIT, help offset the compares with a value of 12 MJ=kg obtained in an greater gain in body weight and fat one would earlier carbohydrate over feeding study.42 Both values expect in HF subjects habitually consuming more differ considerably from the 6 MJ=kg estimate based energy than LF individuals. However, it seems Gluttony and thermogenesis revisited MJ Stock 1116 highly likely that this could involve some form of magnifying and identifying acquired or genetic differ- lipostatic feedback control on metabolism because, in ences in the capacity for DIT and, hence, the propen- another study,47 HF subjects were found to have sity to obesity. signi®cantly higher plasma leptin levels than LF subjects, even when corrected for any differences in BMI and %body fat. Clearly, there is much more to be learnt about the behavioural, physiological and (pos- Acknowledgements sibly) genetic differences between lean HF and LF subjects, and one would particularly like to see how The author apologies for the somewhat egocentric they respond to being overfed a low-protein diet. bias in this review, but uses the excuse that it originated from the lecture he gave when awarded the 1999 Wasserman prize by the European Associa- tion for the Study of Obesity. However, this does not Concluding comments excuse him from acknowledging all those students, colleagues and collaborators that have made the past 35 years of research so stimulating and enjoyable Ð While not trying to de¯ect attention or enthusiasm both scienti®cally (in the lab) and sociologically (in away from the recent rapid advances being made in the pub). Two to mention in particular: Derek Miller, our understanding of the molecular and genetic basis who ®rst ®red, and then fanned my enthusiasm for to obesity, one of the main aims of this review was to research, and Nancy Rothwell, who left (thank God) emphasize that all these exciting advances have to be after 10 of the most frenetic, productive and exhilar- seen in the context of the regulation of energy ating (exhausting) of those 35 years. balance Ð that is intake and expenditure. While most of the researchers helping make these discov- eries are alert to the need to consider potential effects References on the physiological control of food intake when 1 Miller DS, Mumford P. Gluttony 1. An experimental study of interpreting their results, this is generally not so for overeating on high protein diets. Am J Clin Nutr 1967; 20: 1212 ± 1222. the physiological controls operating via DIT on 2 Miller DS, Mumford P, Stock MJ. Gluttony 2. Thermogenesis energy expenditure. in overeating man. Am J Clin Nutr 1967; 20: 1223 ± 1229. One way in which it is hoped that this review may 3 Rothwell NJ, Stock MJ. A role for brown adipose tissue in have helped to raise the status of DIT was by adopting diet-induced thermogenesis. Nature 1979; 281: 31 ± 35. a broader biological perspective, rather than simply 4 Himms-Hagen J, Desautels M. A mitochondrial defect in brown adipose tissue of obese (ob=ob) mouse: Reduced bind- considering DIT in terms of obesity. After a brief ing of purine nucleotides and a failure to respond to cold by an consideration of the role of thermogenesis in thermo- increase in binding. Biochem Biophys Res Comm 1978; 83: regulation and fever, it was argued that DIT evolved 628 ± 634. as a mechanism for enriching nutrient-poor diets by 5 Thurlby PL, Trayhurn P. Regional blood ¯ow in genetically disposing of the excess non-essential energy. The obese, ob=ob mice. P¯ugers Archiv 1980; 385: 193 ± 201. 6 Rothwell NJ, Stock MJ. Effects of feeding a palatable `cafe- disposal of this excess, non-essential energy, once teria' diet on energy balance in young and adult lean ( =?) the animal's fat stores are replete, helps prevent Zucker rats. Br J Nutr 1982; 47: 461 ± 471. obesity which is a hazard to survival in the wild due 7 Thurlby PL, Trayhurn P. The role for thermoregulatory to restricted mobility and vulnerability to predation. thermogenesis in the development of obesity in genetically Thus, DIT evolved as a mechanism for regulating obese (ob=ob) mice pair fed with lean siblings. Br J Nutr 1979; 42: 377 ± 385. nutrient balance with only a secondary role in the 8 Han PW. Hypothalamic obesity in rats without hyperphagia. regulation of energy balance. However, in a world Trans NY Acad Sci 1967; 30: 229 ± 243. where food is now abundant, this secondary role now 9 Cox JE, Powley TL. Development of obesity in diabetic mice assumes greater signi®cance. pair-fed with lean siblings. J Comp Physiol Psychol 1977; 91: Given this broader, nutritional perspective, it was 347 ± 358. 10 Trayhurn P, Jones PM, McGuckin MM, Goodbody AE. possible to explain why most of the human over- Effects of overfeeding on energy balance and brown fat feeding studies using normal, balanced diets failed to thermogenesis in obese (ob=ob) mice. Nature 1982; 295: show very obvious increases in DIT. Even then, closer 323 ± 325. inspection and re-analysis of these studies showed that 11 Nicholls DG. Hamster brown adipose tissue mitochondria: the approximately 40% of subjects must have exhibited control of respiration and the proton electrochemical potential gradient by possible physiological effectors of the proton an increase in DIT to some degree in response to over- conductance of the inner membrane. Eur J Biochem 1974; feeding. Perhaps of greater signi®cance, was the large, 49: 573 ± 583. inter-individual variation observed within and 12 Ricquier D, Kader JC. Mitochondrial protein alteration in between experiments, which were most noticeable active brown fat: a SDS : polyacrylamide gel electrophoretic when the diets were unbalanced with respect to study. Biochem Biophys Res Comm 1976; 73: 577 ± 583. 13 Hervey GR, Tobin G. Brown adipose tissue and diet-induced protein. Thus, although it is currently fashionable to thermogenesis. Nature 1981; 289: 699. test people's responses to high-fat diets, it is suggested 14 Davis TRA. Chamber cold acclimatization in man. J Appl that low-protein diets would be more effective in Physiol 1961; 16: 1011 ± 1015. Gluttony and thermogenesis revisited MJ Stock 1117 15 Huttunen P, Hirvonen J, Kinnula U. The occurrence of brown 33 Stock MJ. The Control of Energy Balance. PhD Thesis 1970; adipose tissue in outdoor workers. Eur J Appl Physiol 1981; University of London 1970. 46: 339 ± 345. 34 Goldsmith R, Miller DS, Mumford P, Stock MJ. The use of 16 Kluger MJ. Fever revisited. Pediatrics 1992; 90: 846 ± 850. long-term measurements of heart rate to assess energy expen- 17 Dascombe MJ, Rothwell NJ, Sagay BO, Stock MJ. Pyrogenic diture. J Physiol 1996; 189: 35 ± 36P. and thermogenic effects of interleukin 1b in the rat. Am J 35 Sims EAH, Danforth Jnr E, Horton ES, Bray GA, Glennon Physiol 1989; 256: E7 ± E11. JA, Salans LB. Endocrine and metabolic effects of experi- 18 Luheshi GN, Gardener JD, Rushforth DA, Loudon AS, mental obesity in man. Rec Progr Hormone Res 1973; 29: Rothwell NJ. Leptin actions on food intake and body temp- 457 ± 496. erature are mediated by IL-1. Proc Nat Acad Sci 1999; 96: 36 Goldman RF, Haisman MF, Bynum G, Horton ES, Sims EAH. 7047 ± 7052. Experimental obesity in man: Metabolic rate in relation to 19 Kleiber M. The Fire of Life. Wiley: New York, 1961. dietary intake. In: Bray GA (ed). Obesity in Perspective, 20 Hamilton JS. Heat increments of diets balanced and unbal- DHEW Publication No. (NIH)75-708. US Government Print- anced with respect to protein. J Nutr 1939; 17: 583 ± 599. ing Of®ce: Washington, DC, 1975; pp 165 ± 186. 21 Rubner M. Die Gesetze des Energieverbrauchs bei der Ernar- 37 Norgan NG, Durnin JVGA. The effect of 6 weeks of over- hung. Deuticke: Leipzig, 1902. feeding on the body weight, body composition and energy 22 Miller DS, Payne PR. Weight-maintenance and food intake. metabolism of young men. Am J Clin Nutr 1980; 33: 978 ± J Nutr 1962; 78: 255 ± 262. 988. 23 Stirling JL, Stock MJ. Metabolic origins of thermogenesis 38 Webb P, Annis JF. Adaptation to overeating in lean and induced by diet. Nature 1968; 220: 801 ± 802. overweight men and women. Human Nutr Clin Nutr 1983; 24 Rothwell NJ, Stock MJ, Tyzbir RS. Energy balance and 37: 117 ± 131. mitochondrial function in liver and brown fat of rats fed 39 Forbes GB, Brown MR, Welle SL, Lipinski BA. Deliberate `cafeteria' diets of varying protein content. J Nutr 1982; overfeeding in women and men: energy cost and composition 112: 1663 ± 1672. of the weight gain. Br J Nutr 1986; 56: 1±9. 25 Delorme M, Thomas DW. Nitrogen and energy requirements 40 Deriaz O, Tremblay A, Bouchard C. Non-linear weight gain of the short-tailed fruit bat (Carollia perspicillata): fruit bats with long-term overfeeding in man. Obesity Res 1993; 1: are not nitrogen constrained. J Comp Physiol B 1996; 166: 179 ± 185. 427 ± 434. 41 Dulloo AG, Jacquet J. Low-protein overfeeding: a tool to 26 Okon EE. Histological changes of the interscapular brown unmask susceptibility to obesity in humans. Int J Obes 1999; adipose tissue in Eidolon helvum in relation to diurnal activ- 23: Ð Commentary that follows this review. ities of the bats. In: Proceedings of the 5th International Bat 42 Acheson KJ, Schutz Y, Bessard T, Anantharaman K, Flatt J-P, Research Conference. Texas Technical University Press: Lub- Jequier E. Glycogen storage capacity and de novo lipogenesis bock, 1980; pp 91 ± 93. during massive carbohydrate overfeeding in man. Am J Clin 27 Rothwell NJ, Stock MJ. Thermogenic capacity and brown Nutr 1988; 48: 240 ± 247. adipose tissue activity in the common marmoset. J Comp 43 Aarsland A, Chinkes D, Wolfe RR. Hepatic and whole-body Biochem Physiol 1985; 81A: 683 ± 686. fat synthesis in humans during carbohydrate overfeeding. Am J 28 Levine JA, Eberhardt NL, Jensen MD. Role of nonexercise Clin Nutr 1997; 65: 1774 ± 1782. activity thermogenesis in resistance to fat gain in humans. 44 Macdiarmid JI, Cade JE, Blundell JE. High and low fat Science 1999; 283: 212 ± 214. consumers, their macronutrient intake and body mass index: 29 Ferrannini E. The theoretic basis of indirect calorimetry: a further analysis of the national diet and nutrition survey of review. Metab Clin Exper 1988; 37: 287 ± 301. British Adults. Eur J Clin Nutr 1996; 50: 505 ± 512. 30 Pullar JD, Webster AJF. The energy cost of fat and protein 45 Cooling J, Blundell J. Are high-fat and low-fat consumers deposition in the rat. Brit J Nutr 1977; 37: 355 ± 363. distinct phenotypes? Differences in the subjective and beha- 31 Diaz EO, Prentice AM, Goldberg GR, Murgatroyd PR, vioural response to energy and nutrient challenges. Eur J Clin Coward WA. Metabolic response to experimental overfeeding Nutr 1998; 52: 193 ± 201. in lean and overweight healthy volunteers. Am J Clin Nutr 46 Cooling J, Blundell J. Differences in energy expenditure and 1992; 56: 641 ± 655. substrate oxidation between habitual high fat and low fat 32 Evans E, Miller DS, Mumford PM, Stock MJ. Gluttony III: consumers. Int J Obes 1998; 22: 612 ± 618. Further studies on overeating. Unpublished manuscript Ða 47 Cooling J, Barth J, Blundell J. The high-fat phenotype: Is limited number of photocopies can be made available to leptin involved in the adaptive response to a high fat (high interested readers. energy) diet? Int J Obes 1998; 22: 1132 ± 1135.