Adaptive thermogenesis with weight loss in humans


  • M.J. Müller,

    Corresponding author
    1. Institut für Humanernährung und Lebensmittelkunde, Christian-Albrechts-Universität zu Kiel, Düsternbrooker Weg 17, 24105 Kiel, Germany
    • Institut für Humanernährung und Lebensmittelkunde, Christian-Albrechts-Universität zu Kiel, Düsternbrooker Weg 17, D 24105 Kiel, Germany
    Search for more papers by this author
  • A. Bosy-Westphal

    1. Institut für Ernährungsmedizin (180c), Universität Hohenheim, 70593 Stuttgart, Germany
    Search for more papers by this author

  • Disclosure: The authors do not have any conflict of interests.

  • Funding agencies: Our own research data presented here have been funded by generous grants from German Research Foundation (DFG Mü 8-1, DFG Bo 3296,1-1) and the Federal Ministry of Education and Research (BMBF catch up fat, BMBF Competence Network Obesity, BMBF Reference Centre for Body Composition, Kiel).


Adaptive thermogenesis (AT) with weight loss refers to underfeeding-associated fall in resting and non-resting energy expenditure (REE, non-REE); this is independent of body weight and body composition. In humans, the existence of AT was inconsistently shown and its clinical significance has been questioned.


Discrepant findings are mainly due to different definitions of AT, the use of various and nonstandardized study protocols, and the limits of accuracy of methods to assess energy expenditure. With controlled underfeeding, AT takes more than 2 wk to develop. AT accounts to an average of 0.5 MJ (or 120 kcal) with a considerable between subject variance.

Design and Methods:

Low-sympathetic nervous system activity, 3,5,3′-tri-iodothyronine (T3) and leptin are likely to add to AT; however, the kinetic changes of their plasma levels with underfeeding differ from the time course of AT and controlled intervention studies substituting and titrating these hormones are rare in humans. AT in response to underfeeding is independent of thermogenesis in response to either diet or cold. Although fat-free mass (FFM) and, thus, liver, and skeletal muscle are considered as major sites of AT, cold-induced nonshivering thermogenesis relates to the metabolism of brown adipose tissue (BAT). In humans, diet-induced thermogenesis is related to postprandial substrate metabolism of FFM with a questionable role of BAT. Obviously, the REE component of AT differs from and its non-REE component with respect to organ contribution as well as mechanisms. Thus, AT cannot be considered as unique.


AT should be characterized based on individual components of daily energy expenditure, detailed body composition analyses, and mathematical modeling. The biological basis of AT as well as the influences of age, sex, obesity, stress, and inflammation remain to be established in humans.


Cellular bioenergetics are considered as a suitable target for antiobesity therapy. The capacity of energy production, heat release, and metabolic adaptations to either starvation, diet, cold, stress, and inflammation differ between subjects. Although the metabolic basis of these adaptations as well as the intraindividual and interindividual variances between them still remain to be characterized, modulation of adaptive thermogenesis (AT) to affect energy balance is challenging; its main regulators, that is, either sympathetic nervous system (SNS) activity, 3,5,3′-tri-iodothyronine (T3), and/or leptin may provide a suitable basis for pharmacotherapies. However, before going so far, recent findings on cellular and molecular aspects of energy metabolism should be discussed in the context of human physiology. The intent of this article is to provide a concept of changes in energy expenditure that occur in response to the environment and, thus, to add to future research on bioenergetics in humans. Our major focus is on AT in response to underfeeding. It should be kept in mind that today more than 60 years have passed by since Ancel Keys had performed his landmark study on human starvation (1). Most of our ideas about metabolic adaptation to food in humans are still based on that study. It is only recently that modern approaches of integrative physiology have addressed changes in human energy expenditure in response to the environment.

Classical Findings and Definition of AT

There is a decline in metabolic rate associated with reductions in food intake and body weight. In the classical Minnesota experiments of Ancel Keys and coworkers, 32 healthy male volunteers underwent 24 wk of semistarvation (i.e., a 55% reduction of energy intake from a mean of 14.6 MJ/d to 6.6 MJ/d) and decreased body weight by 16.8 kg or 24%; at months 3 and 6 of semistarvation, basal metabolism fell from 6.6 MJ/d to 4.0 MJ/d (i.e., by 39%; 1,2). Decreased energy expenditure was to a great part explained by the loss in oxygen-consuming lean and fat tissues. At the end of the Minnesota semistarvation study, resting energy expenditure (REE) expressed per kg body weight had decreased by 19.5%. When expressed per unit of metabolically active tissue (i.e., the sum of oxygen consuming cells within the body) or fat-free mass (FFM), the decline in REE was 15.5%. AT refers to changes in basal metabolism independent of changes in body mass or mass of its metabolically active component, that is, AT is due to the decrease in the average and specific metabolic activity of body cells (2).

In the Minnesota study, about 35% of the fall in basal metabolic rate (0.8 MJ or about 180 kcal) was independent of changes in FFM and therefore ascribed to AT (1). AT may take weeks to develop (3, 4), thus, it is not considered as an immediate adaptation. In the Minnesota experiment, four measurements of basal metabolism were performed within 24 wk with the first documented after 4 wk of semistarvation (1). At that time, AT was found in all volunteers and seemed independent of the magnitude of weight loss. A maximum adaptation was reached after a 10% weight loss or after 12-20 wk. After 24 wk of semistarvation, there was no further adaptation of basal metabolic rate.

Current Controversies on AT

Current controversies on AT relate to its definition as well as its existence. AT is not unequivocally defined and, recently, its original meaning has been extended to any changes in thermogenesis that occur in response to the environment. Some authors referred AT to a mass-independent adaptive suppression of thermogenesis in the resting (or nonactivity-related) compartment of energy expenditure which includes REE1 as well as the thermic effect of food (5,6; see Figure 1 for the different components of daily energy expenditure). Other authors defined AT as the regulated heat production or facultative (or regulatory) thermogenesis. This is activated in excess of requirements and defined as heat production in response to cold (i.e., cold-induced shivering as well as nonshivering thermogenesis) and diet, that is, energy dissipation in response to over- and underfeeding in mitochondria of brown fat and skeletal muscle (7). Facultative thermogenesis can be turned off and on within minutes and may be manipulated therapeutically in obese patients (7); it adds to immediate metabolic adaptations, for example, after a meal (4). Finally, AT has also been related to all components of daily energy expenditure (i.e., resting plus non-REE = REE + thermic effect of food + spontaneous activity + exercise) contributing to so-called metabolic efficiency (i.e., an attenuation of energy balance to restore body weight and body components by affecting energy cost for weight loss or weight gain; 8). The latter idea also implies that during underfeeding, the underlying system of AT aims at reducing the overall rate of energy metabolism to conserve body energy by sparing both, lean and fat mass. In fact, the Minnesota semistarvation data not only showed a fall in REE (see above), but also reduced diet-induced thermogenesis (DIT) as well as costs of daily activities by about 55 and 71%, respectively (1).

Figure 1.

Different components of energy expenditure and thermogenesis. DIT refers to the postprandial increase in energy expenditure which is explained by the cost of processing macronutrients (obligatory DIT) plus a facultative component, that is, the postprandial increase in energy expenditure in excess of energy needs of processing (specific dynamic action). (AT) is defined as the reduction of metabolic rate adjusted for body mass and body composition in response to food and environment. AT has a REE- and a non-REE-component (modified according to Ref. 6).

Some authors refer AT to mass and activity of brown adipose tissue (BAT), which is the site of nonshivering thermogenesis (9, 10). Other authors have subdivided AT into three subtypes: cold-induced shivering thermogenesis (i.e., a function of skeletal muscle), cold-induced nonshivering thermogenesis, and DIT (6).

In weight-reduced obese patients, AT has been defined as the difference between predicted (i.e., based on age, sex, height, and weight) and measured REE and reached about 25% (11). Alternatively, AT in response to underfeeding was defined as the decrease in REE in relation to the loss of FFM compared to what would be expected on the basis of the cross-sectional relationship between REE and FFM (12). In contrast to its original meaning (i.e., Ancel Keys definition based on his semistarvation experiments), AT had been referred to metabolic adaptations in response to both, underfeeding and overfeeding; the large variance in weight gain between individuals in response to a fixed excess of calories by overeating has also been taken as evidence for the existence of AT (6). AT has never been defined in quantitative terms.

In humans, the existence of AT has not been shown unequivocally. In the obese patient, the existence as well as the impact of AT on either weight loss or weight regain after weight loss was modest or none (13-15). In healthy lean subjects, only small (i.e., −6%) adaptive reductions in 24 h energy expenditure were found despite large reductions in energy intake (−55%) during a period of 4 wk (16). Accordingly, using continuous whole body calorimetry to measure changes in energy expenditure in response to 12-d controlled underfeeding in lean men, weight loss was 3.2 kg but REE decreased by 0.42 MJ (or about 100 kcal or 5.7%) only (17). Most of this “adaptation” was brought about by a loss in body mass. The authors of that study took their results as evidence for the inability of the body to prevent weight loss by metabolic adaptation. In lean and obese Caucasians, 48 h of total starvation resulted in small changes in mean sleeping metabolic rate only (i.e., −0.4 MJ or about 92 kcal/d; 18).

In contrast, REE increased (by 0.66 MJ or 160 kcal/d) rather than decreased in healthy lean volunteers totally starved for 4 d resulting in a mean decrease in weight (from 64.2 to 61.5 kg) and RQ, that is, respiratory quotient (from 0.83 to 0.71; 19). These data were in line with an earlier study, where REE was elevated in healthy male lean subjects following 48 h total starvation (20). Increased REE had been explained by an increase in energetically expensive processes, for example, gluconeogenesis or futile substrate cycles of lipids. More confusing, there was no change in REE per kg body weight in five obese patients undergoing a 21 d fast (21). All these human data cast some doubts about the existence of AT with underfeeding. This brings us back to Ancel Keys again who considered the weight loss independent effect of starvation on REE as real but “slight” (see p 291 in Ref. 1).

Contrary to the previous findings, other authors consistently demonstrated weight loss-induced decreases in energy expenditure in lean and obese adults. In the CALERIE study, at a 25% reduction of energy intake in healthy normal- and overweight adults, 24-h energy expenditure and sleeping metabolic rate were about 0.36 MJ (or 87 kcal; −7%) lower than predicted after 3-6 mo of caloric restriction (22). This adaptation occurred within the first 3 mo of underfeeding; despite an ongoing weight loss there was no further effect after 6 mo. However, no AT was observed in a parallel group on a 3.7 MJ/d (900 kcal/d very low-calorie diet at weight maintenance (22). In another clinical study on calorie-restricted obese patients, REE decreased by 0.4 and 0.7 MJ (or 95 and 167 kcal)/d after 5 and 10 kg weight loss, respectively (23); there were no sex differences in metabolic adaptation. These data argued for a role of AT in unsuccessful weight loss interventions and reduced body weight maintenance but were contrary to the idea of other authors that AT can compensate for up to 25% of a given energy deficit while on a restricted diet (24).

In accordance with the clinical data, an adaptive reduction in energy expenditure was seen in normal weight and healthy subjects who were chronically underfed as part of the Biosphere 2 experiment; the adaptation was observed with long-term under-nutrition (at an intake of 7.5 MJ or 1,800 kcal/d) and in response to a substantial weight loss (i.e., 9.1 kg; 25). Controlled underfeeding of normal and overweight subjects at a stable weight at −10% of habitual energy intake reduced 24-h energy expenditure (adjusted for changes in body weight and body composition) by −10 to −15% (26, 27). In that study, 24-h energy expenditure decreased and activity-related energy expenditure (as estimated from the difference between 24-h energy expenditure and REE taking into account the thermic effect of food) was most affected by weight loss. By contrast, REE was either decreased or unchanged with a mean AT of −54 ± 98 and −137 ± 305 kcal, for normal and overweight subjects, respectively. Concomitantly, the thermic effect of food was unchanged. In a subsequent publication on the same protocol, the authors also reported a small but sustained reduction in REE in subjects with sustained weight loss (28). These data were against clinical observations in overweight and obese women suggesting a transient rather than a sustained reduction in energy expenditure in response to energy restriction (29, 30). Anyhow, the sustained metabolic adaptations to a weight loss-stabilized state refer to non-REE (or the non-REE-component of AT) and do not support the existence of AT in its original meaning (28).

Mathematical modeling adds to the question about existence of AT. Using the original body composition and energy expenditure data of eight different studies on normal, overweight and obese subjects, Kevin Hall has recently modeled weight loss in response to underfeeding and weight loss maintenance (31). In his model, some degree of metabolic adaptation was required to explain weight changes. In addition, modeling the original Keys data, a maximum amplitude of AT was associated with dynamic changes in weight with an estimated time constant of 7 d (32). Perturbations in energy expenditure persisted as long as energy intake was different from baseline. Quantitative modeling using baseline data on body composition and metabolism may be used for future definitions of AT, which is the greater than expected change in measured REE in response to underfeeding.

Taken together, the definition of AT varies between studies and its existence is not consistently documented in humans. Between subject variance is high and reproducibility of AT is unknown. Because most controlled feeding studies have been performed in males, the role of sex is unclear. Treating obese patients, AT may or may not become evident. The variance in data may be due to different study populations, various definitions of AT, the degree of weight loss, its dynamics (which are again linked to the degree, duration, as well as control of negative energy balance), as well as uncontrolled conditions of daily clinical practice (e.g., some weight fluctuations due to varying long-term adherence of patients to treatment strategies). Alternatively, AT may be a genetic trait and is thus seen as a metabolic phenotype in subgroups of subjects only. However, before addressing this putative phenotype, one should keep in mind that current estimates of AT are within or near to the error of methods to assess energy expenditure which is around 3-7% (33). Thus, it is likely that methodological problems add to inconsistent results on AT.

AT in response to different weight loss strategies

To gain further insight and to explain discrepant findings on AT observed in experimental and clinical studies, we have compared data of three different protocols. First, a 12-wk diet-induced weight loss protocol in 85 overweight and obese patients. Second, 6 mo weight loss after bariatric surgery in 17 severely obese patients and third, 3 weeks of controlled underfeeding under metabolic ward conditions in eight normal weight men. The latter group received all meals at the metabolic ward of our institute (thus, macronutrient intake and composition were controlled) with limited physical activities (i.e., at a physical activity level, PAL, of 1.5). In our studies, identical methods have been used, that is, indirect calorimetry, air displacement plethysmography, and magnetic resonance (MR) technologies have been applied for assessment of REE and body composition analysis (BCA); regression analysis of REE against body composition data were performed. In that studies, AT was defined by unexplained REE (i.e., weight loss-associated changes in REE which could not be explained by changes of FFM or components of FFM assuming constant specific metabolic rates of individual organs as assessed by magnetic resonance imaging (MRI). Details of the study protocols and methods have been published recently (34-37).

Initial body weight and also weight loss differed between the three study groups (Figure 2). Comparing the results, average REE decreased with weight loss, this effect remained significant after adjustment for FFM (Figure 3). After further adjustment for decreases in fat mass, the weight loss-associated fall in REE remained significant with 3 wk controlled underfeeding only (Figure 3). AT was not apparent in each subject. Defining AT as a fall of adjusted REE of −0.5MJ or about 120 kcal (36) its prevalence was 50% during dietary weight loss, 29% in severely obese patients undergoing bariatric surgery, and 40% in response to 3 wk controlled underfeeding (Figure 4). REE was highly variable in the clinical studies as well as under controlled feeding conditions. AT was independent of the degree of weight loss and it's dynamics; the week-to-week changes in body weight had no effect on AT. This is contrary to previous data based on multiregression analyses with AT as dependent variable (12). These data suggested that the degree of weight change as well as its rate in g/day are significant predictors of AT. Our three protocols differed in that patients had reached a stable body weight at the end of the two clinical studies, whereas in the underfeeding study subjects were still in a dynamic phase of weight loss. It is possible that AT had already had waned in some patients 6 mo after dietary weight reduction and/or after bariatric surgery (Figure 4).

Figure 2.

Mean body weight before and after weight loss in three different study protocols: 12 wk of dietary weight loss in 85 healthy obese subjects (on a low-calorie formula diet), 6 mo after bariatric surgery in 17 severely obese patients, and after 3 wk of controlled caloric restriction (−50% of energy requirement) in eight normal weight young men. For details of the different protocols see Refs. 34-37 and text.

Figure 3.

Resting energy expenditure (REE) measured by indirect calorimetry (A) and adjusted for fat-free mass (FFM) (B) and FFM + fat mass (FM) (C). Data are taken from three different study protocols: 12 wk of dietary weight loss in 85 healthy obese subjects (on a low-calorie formula diet), before and 6 mo after bariatric surgery in 17 severely obese patients, and before and after 3 wk of controlled caloric restriction (−50% of energy requirement) in eight normal weight young men. For details of the different protocols see Refs. 34-37 and text.

Figure 4.

Prevalence and magnitude of adaptive thermogenesis (defined as a decrease in REE adjusted for FFM and FM −0.17 MJ/d) during weight reduction. Data are taken from three different study protocols: 12 wk of dietary weight loss in 85 healthy obese subjects (on a low-calorie formula diet), before and 6 mo after bariatric surgery in 17 severely obese patients and before and after 3 wk of controlled caloric restriction (−50% of energy requirement) in eight normal weight young men. For details of the different protocols see Refs. 34-37 and text.

During 3 wk of controlled caloric restriction, REE adjusted for FFM and FM decreased from 7.4 ± 0.4 to 7.0 ± 0.5 MJ/d (or from 1,770 to 1,674 kcal/d, i.e., −5.1%; P 0.05). Concomitantly, glucose-induced thermogenesis (GIT) decreased from 6.5 to 4.1% of ingested energy (−36.9%; P = 0.07) with no correlation between AT and GIT (r = −0.075; P = 0.800). These data suggest that AT and DIT are regulated independently. With underfeeding, energy expenditure related to physical activity also decreased (i.e., by about 20%); however, in our controlled feeding protocol, PAL was restricted to a maximum of 1.5, thus, this may underestimate the effect of underfeeding on physical activity.

Implications of AT

As Ancel Keys stated “adaptation must be purposeful” (see p 338 of Ref. 11). Because the rate of weight loss decreased with time, the decline in REE in response to deficient energy intake and weight loss and, thus, “sparing of energy” was seen as “purposeful” (3). Weight loss is not linear and depends in part on metabolic adaptation; at a low level of food intake, this was calculated to be equivalent to maintenance of 2.5 MJ (or about 600 kcal) of body energy per day and may help the organism to survive (2). In fact, using a simulation model, AT reduced the underfeeding-induced loss in body energy stores by about 50% (38). More recent mathematical models on weight loss have, thus, included AT as a function of the magnitude of underfeeding (39) or as a fixed parameter that is independent of the degree of weight loss (40).

AT might reflect an autoregulatory adjustment in body weight related to weight maintenance after weight loss. In contrast, AT, if sustained, may encourage weight regain after diet-induced weight loss and, thus, accelerated fat gain during refeeding (5,8,28). However, if REE is appropriate for the subject's new body composition in the reduced weight state, then, it will not predispose to weight regain as long as food intake is matched to energy expenditure (30). Although AT does not represent a perfect adaptation, it seems to be substantial; thus, the concept of AT is still attractive in human physiology as well as in clinical nutrition. Interindividual variances in AT are assumed to add to small imbalances in energy balance, which—if persistent—result in considerable between subject variances in weigh change (12).

Components of Daily Energy Expenditure and AT

Faced with the recent developments in molecular and cellular biology, AT is within the focus of research on cellular bioenergetics (6, 7). However, we feel that physiology of metabolic changes have to be addressed first followed by an integration of that data into cellular and molecular biology. Mechanisms of adaptations in response to underfeeding may refer to both, a reduced stimulus of energy expenditure (e.g., brought about by underfeeding-induced decreases in hormone secretion) and/or a reduced capacity of thermogenic effector mechanism (i.e., cellular energy metabolism). The present evidence suggests that both mechanisms may add to explain AT.

REE, DIT, and energy expenditure due to physical activity (including exercise and nonexercise activities) are the major components of daily energy expenditure, with varying contributions explaining about 55-75, 10-15, and 15-30%, respectively (6,10; Figure 1). In addition, growth, reproduction, and lactation add further components during specific life periods (Figure 1). Energy is also dissipated in response to cold, smoking, stress, inflammation, and/or the thermic effect of certain components like caffeine, capsaicin, or drugs. In general, major changes in energy expenditure to match changes in food intake would arise from physical activity and cold (10) rather than from REE (see Figure 1).

From a metabolic point of view, the components of daily energy expenditure have an obligatory and a facultative component. Obligatory energy expenditure refers to basic cellular and organ function related to the resting state, postprandial metabolism, and physical activity obtained during a physiological and weight stable situation. In the postprandial state, heat is generated during digestion, absorption, processing, and storage of food energy. The obligatory component of the postprandial increase in energy expenditure can be calculated on theoretical grounds (41). In addition, excess energy is dissipated as heat (=facultative component); this is, mainly due to increased β-adrenergically mediated SNS activity. The facultative component of energy expenditure is variable and may, thus, change in response to excess calories, macronutrient composition, cold exposure, or other thermogenic compounds (6, 7).

The full-adaptive response of energy expenditure to changes in energy balance is equivalent to 24-h energy expenditure (13). AT thus refers to both, an REE- as well as a non-REE-component (Figure 1). The non-RE component of AT resembles facultative thermogenesis in response to cold, exercise, and diet. In contrast, the REE component of AT in response to underfeeding relates to obligatory energy expenditure in the resting state. Although contradictory to its original meaning, changes in obligatory energy needs (e.g., the fall in REE observed during underfeeding and starvation) occurs in the nonsteady state and in potentially life-threatening situations. During the nonsteady state in response to underfeeding, energy consuming processes (e.g., protein turnover and protein synthesis, see below) are reduced. Theoretically, underfeeding reduced protein turnover which accounts to about 50% of the mass-independent fall in REE and thus obligatory energy needs (12). After weight stabilization, a new steady state is reached; then, metabolism and, thus, obligatory energy expenditure reestablishes.

Body Composition as Determinant of the REE- and the non-REE-Components of AT

FFM is the major determinant of REE, but FFM is not an “entity” (1). Because individual tissues greatly differ in their specific metabolic rates, weight loss-associated changes in the proportion between different tissues may also add to the weight loss-associated fall in REE (42). For example, loss of muscle tissue (which has a relatively low metabolic rate of about 50 kJ or about 12 kcal/kg × d) has a limited effect, but loosing liver mass or weight of other high-metabolic rate organs (e.g., heart and kidneys which have specific metabolic rates of more than 840 kJ or about 200 kcal/kg × d) has a major impact on the starvation-induced fall in REE. Thus, strictly spoken, AT refers to weight loss-associated changes in REE which are independent of changes in both, body mass and body composition.

Thus, to accurately assess AT, precise and accurate BCA is needed. Because standard methods (like densitometry, DXA, or deuterium dilution) have a poor validity under nonsteady-state conditions (e.g., during early weight loss; 34), this adds to variances in existence and degree of AT reported in the literature. However, in the long-term semistarvation protocol of the Minnesota study, it is likely that a steady state was reached and, thus, underwater weighing, which has been used to assess body composition, may have given valid results. By contrast, this may not apply to many clinical studies on weight loosing obese patients with varying energy intakes and physical activities, some weight fluctuations and observation periods lasting up to few weeks only.

Components of daily energy expenditure differ with respect to site. Although REE results from oxygen consumption of all organs, exercise, and nonexercise-related energy expenditure are mainly related to skeletal muscle and heart. By contrast, postprandial metabolism mainly relates to oxygen consumption in the gastrointestinal tract and the liver with no role for BAT in humans (10). BAT is the site of cold-induced nonshivering thermogenesis with cold-induced shivering thermogenesis being explained by skeletal muscle (Figure 1).

Weight loss associated changes in mass of individual tissues and organs together with changes in their individual metabolic rates determine metabolic adaptation and, thus, AT. Autopsy data suggest that with severe starvation, there is a considerable loss in individual organ masses (except for brain; 43). In addition to changes in masses, experimental data in rats suggest starvation-induced falls in specific metabolic rates of brain, heart, kidneys, skeletal muscle, and gut; by contrast, liver metabolism increased in response to fasting (44). In that study, starvation-induced fall in skeletal muscle and gut oxygen consumption accounted for about 2/3 of AT. Using state of the art MR technologies in humans (MRI, quantitative magnetic resonance (QMR)), weight loss is characterized not only by losses in lean and fat mass but also by considerable changes in the composition of lean as well as fat mass (36,37,45,46). Overweight and obese women undergoing a low-calorie diet for about 12 weeks lost on average 9.5 kg body weight, 8.0 kg fat mass, and 1.5 kg FFM, respectively; changes in fat mass and FFM corresponded to 84 and 16% of weight lost. Regarding high metabolically active components of FFM, masses of heart, liver, and kidney decreased by 8, 4, and 6%, whereas skeletal muscle (i.e., a low-metabolic rate organ) decreased by 3% only (36). Fifty percent of the weight-loss associated fall in REE was explained by losses in FFM and fat mass; in addition, changes in the composition of FFM increased explained variance in REE to 60% (36). These first detailed body composition data suggested that weight loss is associated with a loss of FFM as well as changes in the composition of FFM with a disproportional loss in high-metabolic rate organs. It then also becomes clear that taking into account the composition of FFM reduces the degree of AT.

Our data are limited to organ mass and could not take into account the specific metabolic rates of individual organs in humans. However, modeling based on greater cross-sectional data sets suggested that specific metabolic rates of individual organs were nearly constant with age, sex, and body weight (47-49). This is contrary to experimental data obtained in starved rats (see above, 44). It should be mentioned that use of advanced technologies in human body composition research increases the explained variance in changes in REE in response to energy balance suggesting that specific metabolic changes (i.e., energy expenditure per body or organ mass or tissue or cells) with weight loss become smaller and of minor importance; this questions the existence of AT. If this comes true, scientists should be open-minded to give up a widely beloved concept.

Contrary to lean mass determining REE, BAT provides a defence mechanism against cold. Under thermoneutral conditions (4), BAT has no effect on REE and DIT in humans (10). In animals, a role for BAT in DIT has been proposed (50) but this has not been confirmed in humans (10, 51). Rodents have a high-thermogenic capacity of BAT which can rise to approximately twice the REE in response to cold (4). Heat production in BAT arises from stimulation of the adrenergic nervous system, thyroid hormones, and an interaction of free fatty acids with uncoupling protein 1 (UCP12, i.e., uncoupling ATP synthesis from respiration). Without adrenergic stimulation, there is neither uncoupling nor functional BAT-activity (4). After genetic ablation of UCP1, mice did not show adrenergic thermogenesis but had a normal REE and, thus, the animals were viable (4). These data provide evidence for the idea that regulation of REE differs from cold-induced thermogenesis (Figure 1). This is also against the idea that BAT contributes to AT during underfeeding.

Most ideas about BAT are based on animal studies (9). Recent evidence suggested a role of BAT in humans. PET-CT data obtained in humans have shown a prevalence of BAT in the order of 5-10% (52-54). However, the reproducibility was low (with only one in eight patients with BAT having positive scans at an additional PET-CT-investigation) and the prevalence of BAT increased with the number of reinvestigations and true prevalence was estimated to be 64% (54). Although lack of standardization (i.e., ambient temperature was not standardized in the studies cited) may add to the variance of interindividual and intraindividual variance of BAT, there was an association between BAT and either body weight or metabolism, both, interindividually and intraindividually, suggesting the functional impact of BAT (52-54). However, preliminary data of a recent clinical study showed, that in obese patients BAT tended to increase rather than to decrease after a 12% weight loss; weight loss was not associated with changes in BAT activity with a high correlation between baseline BAT and BAT after weight loss (55). Under thermoneutral conditions, REE was similar in BAT positive and BAT negative subjects and between group differences in energy expenditure became evident after cold exposure only (56). Altogether these data suggest that under thermoneutral conditions and during underfeeding, BAT does not contribute to REE but refers to the non-REE component of AT in response to cold.

Hormonal Control of Energy Expenditure and AT

The hypothalamus and the brain stem are key structures involved in the involuntary control of food intake and thermogenesis in response to changes in energy balance and energy stores (57, 58). The control of food intake and energy expenditure are interconnected by anabolic and catabolic mediators. The production of these mediator molecules is modulated by short-term and long-term signals from the periphery. These signals inform brain cells about energy stores and energy fluxes. SNS activity, thyroid hormones, and leptin serve as major effectors regulating metabolic adaptation to weight changes. Most of these ideas are based on animal data.

In humans, SNS activity, leptin, and thyroid hormones have been suggested as major determinants of AT. With starvation and reduction in carbohydrate intake, there is a fall in plasma concentrations of both, 3,5,3′ triiodo-L-thyronine (T3; 59,60) and leptin (61, 62) with variable changes in the plasma levels and urinary excretion of adrenaline and noradrenaline (17, 60). Urinary catecholamine excretion was related to energy intake (63) but alterations in the SNS control of thermogenesis may occur without whole body changes in SNS activity or changes in plasma or urinary concentrations of catecholamines (64).

The fall in thyroid hormone levels has been explained by starvation-induced alterations in the activity of the hypothalamic-pituitary-thyroid axis (59, 64) as well as in peripheral thyroid hormone metabolism (65). The latter finding is in part explained by reduced cellular uptake of thyroid hormones (66). The short-term starvation-induced fall in bioactive thyroid hormone concentrations had been related to protein sparing in skeletal muscle due to lower rates of hepatic gluconeogenesis as part of adaptation and protection limiting tissue catabolism (59). There is a close association between REE and plasma T3 levels in mixed populations of under- and normal weight subjects (67). In addition, in vitro data suggested rapid and direct effects of T3 and it's metabolite 3,5-diiodo-L-thyronine on cellular oxygen consumption in skeletal muscle and liver (68-70). These two organs are the major sites of thyroid hormone action on REE (71). Thyroid hormones (and also thyroid hormone mimetics) also stimulate UCP1 activity in BAT (6,65,72). Within BAT, type 2 iodothyronine deiodinase regulates cellular thyroid hormone activity and thus thermogenesis.

In euthyroid subjects, thyroid hormones may affect changes in REE in response to underfeeding (and, thus, presumably AT) as well as thermogenesis in response to overfeeding and cold. In our own controlled 3 wk underfeeding study (see above), there was a tendency toward a reduction in serum triiodothyronine levels (−0.2 ± 0.4 ng/l; P = 0.056) that correlated with the decrease in REE adjusted for FFM and FM (r = −0.56; P < 0.05). These data suggest that T3 adds to regulate AT, which was contrary to the negative results of the CALERIE study (23). The difference between the two studies is most probably explained by the study protocol. In the CALERIE study, caloric restriction was about 25% compared to 50%; in addition, the duration of the protocol was 6 mo in the CALERIE study compared to 3 wk in our study. Thus, a weight stable situation had been reached in the former study, whereas subjects were still in the dynamic phase of weight loss after 3 wk of controlled underfeeding.

Dulloo and coworkers have postulated that the metabolic response to undernutrition and weight loss is determined by autoregulatory feedback systems based on signals from adipose tissue (5, 8). This system involved both, suppression of energy expenditure as well as partitioning of weight loss, that is, the pattern of lean and fat tissue lost (which is on average 75% fat and 25% FFM in a weight-loosing obese subject). That idea suggests that besides the food energy deficit, the greater the reduction in body fat, the greater the reduction in the REE component of AT. Because leptin signals energy stores to the hypothalamus, it may serve to explain the association between fat mass and AT.

Human and animal data suggest that, with starvation, there is a rapid decline in leptin secretion by white adipocytes as well as in plasma leptin concentrations which precede significant depletions of fat mass (57,61,62). Leptin is considered central in energy homeostasis. In laboratory animals, it inhibits food intake and increases oxygen consumption (57, 58). A putative thermic effect has been explained by leptin-stimulated SNS activity to its target tissues such as BAT (57, 58). In pair-fed mice, the weight-reducing effect of leptin is dependent on UCP1-activity and thus the presence of BAT (4). In addition, a direct thermic effect of leptin has been described in skeletal muscle (73). Low leptin levels may also reduce the hypothalamic-pituitary-thyroid axis, and thus may add to a low T3- and, thus, hypometabolic state as seen in starvation (65). A thermic effect of leptin has been questioned because in that animal experiments energy expenditure was expressed per kg bw, which is misleading when one compares energy expenditure of animals with different body weights (i.e., normal weight vs. obese animals; 51,74). In fact, comparing energy expenditure expressed per animal no thermic effect of leptin was seen (51).

In humans, the starvation-induced fall in plasma leptin levels has been proposed to add to reduce energy expenditure; leptin may serve as a signal for energy conservation. This idea is supported by animal data showing that leptin replacement blunted the endocrine and metabolic response during fasting (57). However, a thermic effect of leptin is not well defined in humans. A “hypo-leptinemic” or “leptin insufficient” state as seen in response to underfeeding is associated with low energy expenditure but longitudinal data suggested that the magnitude of changes in plasma leptin do not correlate with REE in a linear relationship (75-78). In the CALERIE study, 6 mo of caloric restriction decreased energy expenditure (i.e., −0.5MJ or 125 kcal/d) as well as 24-h leptin secretion with concomitant falls in urinary norepinephrine and plasma T3 concentrations (79); in that study, metabolic adaptation was correlated with changes in leptin secretion.

There might be an effect of sex, that is, because women have a higher fat mass and higher plasma leptin levels than men, the weight loss-induced fall in leptin resulted in low leptin levels in women but “hypo-leptinemic” levels were seen in men only (80). There is also evidence for the idea that leptin is effective below a certain threshold only and control of energy balance is “asymmetric,” that is, a thermic effect of leptin becomes evident with weight loss below normal weight only (58, 80). In fact, plasma leptin levels positively correlated with REE adjusted for lean mass in the underweight but not in normal or overweight women (82). If we assume that leptin has a thermogenic effect, sex differences may add to variance in AT cited above. It should be mentioned that in the classical Minnesota experiments, Keys had investigated men only (1). This is also true for our controlled underfeeding study (34,37). By contrast, in the CALERIE study a mixed population has been investigated with a greater proportion of women compared with men (22). In the controlled feeding experiments performed by Rosenbaum and Leibel about 50% males and 50% females were included (26-28,60). By contrast, in the clinical studies on obese patients, the proportion of females exceeded males (e.g., Refs. 11,29,35).

There was no thermic effect to exogenously supplied leptin either in patients with leptin deficiency (83) or in obese patients (84) or in healthy lean man after 72 h of starvation (85). Some evidence for a role of leptin in AT came from observational studies in three severely obese patients with congenital leptin deficiency undergoing leptin replacement and thus weight loss (86). These patients were compared with a group of obese patients with normal or elevated plasma leptin levels who underwent a low calorie diet. This protocol allowed to compare the effect of weight loss on energy expenditure at reduced leptin levels (as observed in the control group) with maintained plasma leptin (as seen in the leptin-replaced group). With similar weight losses of about 15 kg, AT was 1.1 MJ (or about 260 kcal) in the control group (which was a very high AT compared to the results of the studies cited above) and 0.6 MJ (or 144 kcal) in the leptin-replaced group. The authors of that study concluded that preventing the weight loss-induced fall in plasma leptin levels may curb AT. However, this study protocol was uncontrolled.

Other authors have used controlled experimental conditions in weight-reduced weight stable humans (87): A 5-wk treatment with leptin reversed the weight loss-associated fall in thyroid hormones and SNS activity and returned 24-h energy expenditure to normal values. However, in that protocol, leptin had no effect on REE, and its effect was due to energy expended for physical activity and skeletal muscle energetics (defined as mechanical efficiency of skeletal muscle while bicycling to generate 10 W of power; 87). The authors concluded that the major effect of leptin is on the non-REE component of AT to defend fat mass and is thus acting at reduced fat stores only.

Comparing the kinetics of starvation-induced hormonal changes with the kinetic decrease in REE, endocrine adaptations were seen in early starvation (i.e., within the first day) and preceded AT which became evident after more than 2 wk of underfeeding only (1,5,57,59). In the case of T3, the time lag between “signal” and “effect” reflects it's biological half life. In addition, local (i.e., cellular) concentrations of hormones or even the intracellular generation of active metabolites (in the case of iodothyronines) may differ from their plasma levels which may not accurately reflect the tissue state. However, during underfeeding, detailed data on kinetic changes of hormones and their metabolic responses or even modelling of the possible associations between determinants and effects are missing in humans.

There is no clear evidence for a thermic effect of additional hormones involved in the metabolic adaptation in response to underfeeding. With weight loss, there is a long-term persistence of the hormonal adaptations including insulin, glucagon, peptideYY, cholecystokinine, amylin, ghrelin, gastric-inhibitory polypeptide, and pancreatic polypeptide (88). However, in humans none of these hormones has an effect on energy expenditure. The results of a single study suggested a thermogenic effect of glucagon, which appeared at supraphysiological plasma levels at concomitant insulin deficiency only (89). This effect has been explained by glucagon's effects of glucose and protein metabolism. Experimental data further suggest that glucagon acutely increased energy expenditure (90). In addition, chronic glucagon stimulation in nonphysiologic doses promotes thermogenic activity of BAT (90) suggesting a role of glucagon in cold-induced thermogenesis. This is in line with the idea that glucagon is involved in stress response interacting with SNS activity (91). Taken together, the transient increase in plasma glucagon at a concomitant fall in insulin secretion in response to starvation would increase rather than decrease REE arguing against a role of pancreatic hormones in AT.

Environmental Factors Affecting Energy Expenditure

Pollutants such as organochlorines accumulate in subcutaneous and visceral fat tissue. Their plasma concentrations increase with weight loss. The weight loss associated increase in plasma organochlorines was associated with the decrease in REE (11). Based on a stepwise regression analysis, 47% of the difference between predicted and measured changes in sleeping energy expenditure were explained by changes in plasma organochlorine concentrations with an additional effect of changes in plasma leptin levels of 20% (92). There was further evidence that changes in plasma pollutants are negatively related to changes in T3 as well as to oxidative capacity in skeletal muscle (93). In addition to pollutants, hypoxic stress was shown to be related to low REE which has been explained by a low SNS activity (11). All these data refer to obese and (in the case of hypoxia to) severely obese patients.

Cellular Mechanisms and Metabolic Pathways of AT

Ancel Keys had proposed three possible means to explain AT (1). First, simple fuel exhaustion. Second, decrease in oxidative enzymes. And third, reduction in circulation. In the Minnesota study, a reduction in cardiac work (i.e., by 50%) and the fall in body temperature (by about −0.7°C) were considered as main mechanisms of REE reduction (2). In 1950, Keys did not know “to which is cause and which is effect” (see p 281 in Ref. 1) as we still do not know in 2012. Biochemically, protein turnover, the maintenance of ionic gradients across all membranes, gluconeogenesis, ureagenesis, and turnover of glucose, glycerol, and fatty acids by so-called futile cycles are the major contributors of REE. The REE component of AT is explained by reductions in protein turnover, adaptations in glucose and urea productions, metabolic cycles, and/or decreases in Na+, K+, and Ca++ leak back across membrane channels and, thus, reduced energy costs. Quantitative measurements performed during controlled underfeeding in humans suggested that about 50% of AT is due to decreases in protein turnover and substrate cycling, leaving about 50% for Na+, K+-ATPase, and other energy consuming reactions (12, 16). In fact, semistarvation reduced the skeletal muscle concentration of Na+, K+-ATPase by 20-48% (94, 95).

By contrast, the “energetic contribution” of underfeeding-induced changes in glucose and fat metabolism is limited and not likely to exceed 10% of REE (96). In prolonged starvation, the rates of hepatic and renal gluconeogenesis remain stable but ureagenesis decreases (21,97,98). Quantitative estimates of substrate cycling suggested that triglyceride-fatty acid cycling (TG/FA-cycling) is the most important substrate cycle, it contributes to REE with a total energy cost of 144 kcal per mol TG recycled (95,97,98). However, although TG/FA-cycling increased by 74% in humans starved for 84 h, this corresponds to 2.0 versus 1.3% (=fed) of REE only (95). In another human study, healthy nonobese adults were starved for 4 d; during that time, starvation produced a threefold increase in glycerol turnover while the energy cost associated with TG/FA-cycling increased from 0.36 to 6.29% of REE (99). The difference between the two numbers (2% in Ref. 98 and 6.29% in Ref. 99) might be explained by different methods (i.e., an isotope dilution techniques was used in the former study, whereas a nonisotopic approach was applied in the latter study). Anyhow, the energy expended for that cycle is a small proportion of REE. It should be mentioned that relatively comprehensive computational models of human metabolism that account for these various metabolic flux changes (as well as organ mass changes) and their energy costs still required a model of AT to explain the changes in energy expenditure observed in various underfeeding studies (32).

Conclusions and Areas of Future Research

  • 1Adaptations in energy expenditure in response to different environments are not unique. AT in response to underfeeding differs from AT in response to cold as well as from DIT. These data argue in favor of different components of AT. In its original meaning, AT is referred to underfeeding only. However, adaptive responses of energy expenditure to environment and to changes in energy balance refer to all components of daily energy expenditure. Thus, AT has a REE-component and a non-REE-component.
  • 2The REE-component of AT corresponds to changes in REE independent of changes in body weight and body composition. A new definition of AT is based on quantitative mathematical modeling, that is, AT is the greater than expected change in REE with underfeeding, where the expected change in REE is calculated using a quantitative model that includes body weight and body composition.
  • 3To induce AT, a controlled underfeeding protocol should last more than 2 wk. Whether AT is a defined phenotype and is affected by age, sex, in the obese or in chronically and severely ill patients remains to be established.
  • 4Faced with its definition, detailed and accurate in vivo BCA is necessary to characterize AT. Because standard techniques (e.g., assessment of body density or DXA measurements) have limitations during nonsteady-state conditions (i.e., during short-term negative energy balance) (34), this adds to problems to assess AT. In the nonsteady state, MR technologies seem to be most promising for in vivo BCA; the recent QMR technology has the potential to accurately assess small changes in fat mass and body water (34). Within FFM, the masses and the metabolism of high-metabolic rate organs (brain, liver, heart, and kidneys) can be addressed by MRI (36, 42); this should be the specific focus of research.
  • 5Because most of the fasting-induced neuroendocrine abnormalities were reversed by physiological replacement of leptin, there is evidence that leptin adds to the non-REE-component of AT. However, a direct effect of leptin on AT is not documented in humans. Because the thermic effect of leptin is related to BAT specific UCP1, its contribution to underfeeding-induced adaptations of REE is unlikely. By contrast, during underfeeding, a role of low SNS activity and/or low T3 (or other iodo-thyronines) in the REE-component of AT is suggestive. The hormonal regulation of the REE- and the non-REE-component of AT as well as its cellular mechanisms remains to be clarified. At present, these hormones do not provide a suitable basis for antiobesity treatment in humans.

A Personal Note Added

Recent advances in cellular bioenergetics pursue to the idea of pharmaceutical modulation of metabolic adaptation as a measure of obesity treatment (6, 7). This is questioned in the face of our present and still limited knowledge about human physiology of energy metabolism. It is evident that animal data on AT end up in some speculations which remain to be proven by controlled studies in humans (10, 13). The idea of immediate translation of basic research into cures (e.g., into antiobesity therapy) is hyped up. Personally, we feel that, Tom Südhof is right who faced with the recent advances in genomics and molecular biology, there is strong need of “solid descriptive science … that cannot claim to immediately understand functions or provide cures, but which forms the basis of everything we do” (100).


All research protocols have been approved by the local ethical committee.

  1. 1

    REE was previously mentioned as basal metabolic rate, that is, energy expended in an awake, unstressed, resting, not actively food digesting subject = 12–14 h after the last meal, at an environmental temperature of about 22°C; during deep sleep energy expenditure decreases by about 8–10%.

  2. 2

    A 32-KD protein that is present in BAT only.