Non-shivering thermogenesis is most extensively studied in rodents. The documented level of NST in rodents amounts to more than 150% percent of basal metabolic rate (BMR; Davis et al., 1960). In adult humans the NST under mild cold exposure shows a large individual variation and varies from 0 to 30% (Dauncey, 1981; Warwick and Busby, 1990; van Ooijen et al., 2001; Claessens-van Ooijen et al., 2006; Celi et al., 2010; Wijers et al., 2010). For a more detailed overview of these studies see van Marken Lichtenbelt and Schrauwen (2011). The maximal level of NST of humans is not known. Rodents, who are fully dependent on BAT for NST, clearly demonstrate that the capacity for NST and BAT activity depends on the level of acclimatization to cold (Golozoubova et al., 2001; Meyer et al., 2011). Heat production in BAT is by mitochondrial uncoupling and is executed by uncoupling protein-1 (UCP1). This is a BAT specific inner-membrane mitochondrial protein. Under “normal” circumstances F (0)F(1)-ATPase uses most of the energy stored in the proton gradient to produce ATP, which is the energy intermediate in the organism. UCP1 causes a reflux of protons into the mitochondrial matrix, bypassing the ATPase. Instead of ATP, heat is dissipated by the uncoupling itself but also by the other parts of the phosphorylation process (Nicholls and Locke, 1984; Klingenberg and Huang, 1999; Cannon and Nedergaard, 2004). In other words the overall cell efficiency to produce ATP drops by mitochondrial uncoupling. UCP1 is a member of a mitochondrial carrier protein family (like UCP2, UCP3). UCP1, however, is the only protein from this family that is shown to mediate NST (Nedergaard et al., 2001) in BAT, at least in rodents.

In humans the classical studies of Davis clearly show that (parallel to rodent studies) during regular cold exposure shivering decreases, while the level of CIT increases (Davis et al., 1960). This indicates that indeed NST increased in adult humans. The anatomical/histological studies by Huttunen et al. (1981) demonstrate high amounts of BAT in Finnish outdoor workers (Huttunen et al., 1981). Combining the results of these studies in humans at least suggest that with increasing NST capacity the amount of BAT increases. Contrary to rodents there are studies that indicate that in humans skeletal muscle NST by cold-induced mitochondrial uncoupling is involved (Wijers et al., 2008, 2011). It is interesting to note that with cold acclimatization, not only the capacity for NST increases, but in parallel the BMR also rises.

In conclusion, human NST can amount up to 30% of BMR, but it is possible that by means of cold acclimatization higher levels are possible. There is a large individual variation, which can in part be explained by different levels of cold acclimatization. These levels of cold acclimatization may in turn be linked to body composition (see below).

Adult human brown fat was anatomically described (Cramer, 1920; Heaton, 1972; Tanuma et al., 1975, 1976) many years ago, and significant amounts of BAT have for long been demonstrated in adults living in cold environments from samples obtained by necropsy (Huttunen et al., 1981). More recently, unrelated pursuits within nuclear medicine, using FDG-PET/CT scans, indicated that BAT is present and active in adult patients (Hany et al., 2002; Cohade et al., 2003; Yeung et al., 2003; Nedergaard et al., 2007), followed by experimental studies on cold activated BAT (Saito et al., 2009; van Marken Lichtenbelt et al., 2009; Virtanen et al., 2009). With a PET scan performed 45–60 min after intravenous injection of ¹⁸F-2-fluoro-2-deoxyglucose (FDG), glucose uptake can be visualized. FDG is taken up via members of the sodium-independent glucose transporter family (such as GLUT1, GLUT3, and GLUT4). Therefore it provides a measure of the volume of brown fat and a measure of glucose uptake in BAT, especially so with the dynamic PET/CT scans. From these scans, however, it cannot be determined how pure (brown adipocyte density relative to white adipocytes) the BAT tissue is. Alternatively BAT perfusion has recently been measured using PET/CT in humans using labeled water ([¹⁵O]H₂O; Orava et al., 2011).

Animal and cell line studies have revealed much information on the embryonic development of BAT and skeletal muscle (Seale et al., 2008). Brown adipocytes from distinct depots and skeletal muscle cells share common precursors. Subscapular and perirenal BAT in rodents consists of almost 100% brown adipocytes. In contrast, brown-like adipocytes emerging in white adipose tissue (“BRITE-brown-in-white adipocytes”) originate from other precursors, that are shared with white adipocytes (Petrovic et al., 2010). This tissue thus consists of a mixture of brown and white adipocytes. It is also possible that BRITE cells come from transdifferentiation of white adipocytes (Cinti, 2009). A very recent cell culture study provided the first evidence of inducible brown adipogenesis in humans (Lee et al., 2011b).

To address the specific origin of human brown fat cells definite anatomical, histological, and molecular measurements are needed. Some information is already available. To some extend there is a shift in the BAT depots from newborns to adults (Figure 1). Metabolic studies related to brown fat or NST in newborns are very scarce (but see for instance; Karlberg et al., 1965). However, older histological studies show that there is a large interscapular depot in newborns, resembling that of rodents, but probably more mixed with white adipocytes (Merklin, 1974). The most pure BAT depots in newborns are located posterior cervical, perirenal, and suprailiac. Most of the interscapular and suprailiac depots disappear during growth. Also the perirenal BAT is much more abundant in newborns than in adults. In older children at least from the age of 9 year and in adult humans the most important sites are supraclavicular, axillary, and paravertebral (Hong et al., 2011; van Marken Lichtenbelt and Schrauwen, 2011). Although these sites are also reported in newborns it is possible that in adults much of these tissues consist of later recruited brown(like) adipocytes. It therefore is possible that these sites in fact are BRITE cell depots. Indeed, all histological observations of BAT in humans show that brown adipocytes are mixed with white adipocytes (Heaton, 1972; Cypess et al., 2009; van Marken Lichtenbelt et al., 2009; Zingaretti et al., 2009). Even in the major supraclavicular depots the tissue consists of a mixture of brown and white cells (Virtanen et al., 2009). The cell composition of human BAT and whether the distinction between BAT and BRITE has functional consequences remains to be investigated. Nevertheless, it is likely that the density of the brown fat (like) adipocytes determines the mass-specific metabolic capacity of the tissue.

Rodents fully depend on BAT with respect to NST (Cannon and Nedergaard, 2004). The quantitative contribution of BAT to NST in humans still needs to be elucidated. Up to this day, dedicated studies on human cold activated BAT combined with measurements on energy expenditure are restricted to five studies (Saito et al., 2009; van Marken Lichtenbelt et al., 2009; Orava et al., 2011; Vijgen et al., 2011; Yoneshiro et al., 2011). Two of these studies show that subjects with BAT have higher CIT (Vijgen et al., 2011; Yoneshiro et al., 2011). Only one study shows that there is a correlation between BAT activity (i.e., FDG uptake) and NST (Yoneshiro et al., 2011). One study shows an association between BAT perfusion (by [¹⁵O]H₂O PET/CT) and energy expenditure in the cold (Orava et al., 2011). However in this study no distinction has been made between CIT and RMR. Another study shows that there is a relation between BAT activity and RMR (van Marken Lichtenbelt et al., 2009). It is disputed whether the RMR measurements (at 22°C) in this study were really under thermoneutral conditions, because 28°C is reported to be thermoneutral (Nedergaard et al., 2011). However, the thermoneutral conditions were established in naked or very sparsely clothed humans (Erikson et al., 1956; Dauncey, 1981), while during the BAT tests the subjects wore more clothes (0.49 clo). Therefore, indeed RMR was measured. Higher levels of BAT could indicate higher levels of cold acclimatization, which in turn are related to elevated levels of RMR. It should be noted, however, that the issue of measurements conditions during thermoneutral conditions and during mild cold does deserve special attention (see below).

In summary, the available studies at best provide a relation between BAT FDG-activity and NST, but from these data the contribution of BAT cannot be quantified. Prudent calculations based on maximal BAT oxygen consumption in rodents combined with allometric considerations, and based on PET/CT scans reveal a contribution of activated BAT of 3–5% of BMR (Virtanen et al., 2009; van Marken Lichtenbelt and Schrauwen, 2011). These numbers are based on 50 g of BAT (based on FDG-PET/CT). From a physiological and practical point of view a 5% contribution is significant. Moreover, the individual variation is high, meaning that higher contributions are likely. Based on PET/CT the amount of BAT ranges from 0 to 130 g (van Marken Lichtenbelt et al., 2009; Orava et al., 2011). For the calculations several assumptions were made that must be verified. For actual quantification of human BAT energy expenditure more information is needed about the actual amount of BAT, the brown fat cell density of the tissue, and the mass-specific (or cell specific) BAT energy expenditure. Moreover it is possible that BRITE cells from less distinct depots have an additional contribution to NST. They have not yet been mapped in humans, because their FDG uptake is below the detection limit of PET/CT. However these cells may be abundant in many white adipose tissue, and therefore can have a significant contribution.

Another important aspect not yet studied is to what extent BAT needs to be activated to have a significant effect on energy expenditure. May be maximal activation of BAT needs uncomfortable low temperatures. It is likely that temperatures just below the thermoneutral zone do not activate BAT maximally, but are acceptable from a comfort point of view. Studies are needed to find out to what extend sub maximally activated BAT increases thermogenesis. It is likely that those conditions will be more readily accepted in our daily living environment such as offices and dwellings.

Diet induced thermogenesis can be subdivided in an obligatory part and a facultative part. Obligatory DIT encompasses the heat generated by digestion, absorption, and processing of the food. The facultative DIT is the regulated heat production used to dissipate food energy. This part is affected by for instance glucose tolerance, insulin sensitivity, body fat (distribution), and the sympathetic nervous system (SNS) activity (Valensi et al., 1998; Camastra et al., 1999). The SNS activity in its turn is elevated due to feeding (Young and Landsberg, 1977). Especially carbohydrate intake (Schwartz et al., 1999) and a very low protein diets increase the facultative thermogenesis (Stock, 1999). On the other hand the activity of the SNS is reduced in response to starvation (Young and Landsberg, 1977). The SNS-response to feeding significantly contributes to DIT (Ravussin and Swinburn, 1992).

The SNS is concerned in both the facultative part of DIT and NST. Where NST is a means of protecting the body against cold, the facultative part of DIT is a mechanism for enriching nutrient poor diets by disposing of the excess non-essential energy (Stock, 1999). Thus, facultative DIT increases as the diet becomes unbalanced (e.g., low protein content). On the other hand the increase in DIT after a high protein diet results from the high metabolic costs of metabolizing protein and is a form of obligatory thermogenesis not involving the SNS. One of the first studies on facultative DIT was on pigs showing that animals with a rather extreme low protein diet had a much higher DIT than those with a high protein weaning diet (Miller and Payne, 1962). Several studies followed that indicted that the SNS was involved. Stirling and Stock (1968) showed that both CIT and DIT involve SNS activation of heat production.

In humans such experiments are very rare. One study compared the variations in thermogenesis in subjects exposed to both mild cold (3 days) and overfeeding (3 days, 160% of normal energy intake; Wijers et al., 2007). The changes in energy expenditure during mild cold exposure and overfeeding were significantly correlated. These results thus reveal that both overfeeding-induced and mild CIT are individual specific and may share common regulating mechanisms, confirming results from animal studies (Feldmann et al., 2009). Indeed, there was a significant correlation between fasting norepinephrine plasma concentrations and energy expenditure in both situations. Although fasting catecholamine levels may not be the best indicator, this result is in line with SNS involvement.

It was not until the publication of Rothwel and Stock (1979) that the link between DIT and BAT was established in rodents, although this link has recently been disputed (Kozak, 2010).

Are there any indications for a relation between the food intake or diet composition and BAT activity in humans? Currently there are no dedicated studies available. From studies from Nuclear Medicine carried out to reduce BAT FDG uptake for cancer research, it is evident that the diet composition may affect FDG uptake in BAT. A high fat low carbohydrate diet reduced BAT uptake compared to a fasted condition (Williams and Kolodny, 2008). This was explained by the fact that fatty acids are used as main fuel and secondly by the Randle effect (fatty acids loading inhibits glucose metabolism). This study however, does not provide evidence for nutritional activation of BAT in humans. It does show that nutritional manipulation can affect BAT activity, in this case reduced glucose (or FDG) uptake. However, there are no elaborate studies that show that food intake can increase BAT activity.

Indirectly, there is some evidence of involvement of BAT in DIT in humans. That is, insulin, which normally is increased after eating, increases human BAT metabolism (Orava et al., 2011). Glucose uptake was measured with FDG-PET/CT, while plasma insulin concentration was elevated by using euglycemic hyperinsulinemic clamp technique (1 mU/kg/min). It turned out that Insulin-stimulated glucose uptake rate in BAT was higher than in white adipose tissue, similar to skeletal muscle, but less than cold stimulated glucose uptake in BAT. This indicates that food intake, which elevates insulin levels, may activate BAT as well. Interestingly, while during cold exposure BAT both glucose uptake and blood perfusion are elevated, insulin induced glucose uptake was not accompanied with an increase in blood perfusion. Probably, the glucose is transported into the brown adipocyte without an increase in thermogenesis. If certain diets increase both glucose uptake and/or perfusion in BAT awaits investigation.

In conclusion, there is no dietary intervention study in humans available that shows a relation between food intake, dietary composition, or overfeeding and BAT activity. Such studies simply still need to be performed.

Most human BAT studies use body fat (by dual X-ray absorptiometry, DXA) or BMI as an indication of obesity. Because BAT activation by cold depends on the insulation properties of the body, preferably more information about body fat distribution then DXA gives should be obtained. Especially the amount of subcutaneous body fat and skeletal muscle in the limbs is needed. DXA provides to some extend a possibility to map body fat, but with scanning techniques such as magneto resonance imaging (MRI) or CT more details on subcutaneous body fat can be determined. DXA is the most widely used technique for determining whole body fat percentage, but it is not very accurate on an individual level. It is used because of convenience, not because of accuracy. In fact DXA has a substantial error up to 6.7% compared to the gold standard (four-compartment model; Wong et al., 2002; van Marken Lichtenbelt et al., 2004; Minderico et al., 2008). This means that the DXA results are comparable to methods such as skinfold thickness and bioimpedance analyses. A serious problem is that the DXA companies do not provide details about their assumptions and calculations and comparisons of different instruments indicate a large inter-instrument variability. More attention should be given to the errors involved and that a significant part of the variation in metabolic studies potentially can be ascribed to the variation in body composition (DXA) results. Most accurate method is the three-compartment model combining underwater weighing with deuterium dilution (or the four-compartment model including a DXA bone scan; useful in subjects with low bone mineral densities). Unfortunately these techniques do not provide information about the body fat distribution. Alternatively imaging techniques such as CT and MRI are very useful because they do provide details on body fat distribution and subcutaneous fat. CT has better resolution than DXA but provides greater ionizing radiation exposure. MRI provides detailed information but needs a longer imaging time and specialized equipment for acquisition and analysis (Silver et al., 2010).

Metabolic studies on BAT activity using (mild) cold exposure need to be carried out correctly, preferably in a standardized fashion. Both the cold and the thermal neutral conditions have to be selected carefully. Wrong environmental temperatures can result in wrong conclusion as has been shown recently with rodents that are in general raised and housed at temperatures below their thermoneutral zone. These conditions resulted in the wrong conclusions that UCP1 knock out mice fail to become obese (Feldmann et al., 2009). The mice shivered in order to keep warm. Feldmann et al. (2009) elegantly demonstrated that when these mice were kept at thermoneutrality, UCP1 ablation in itself did induce obesity.

Cold exposure experiments in humans should be carried out with the temperature conditions adjusted to the individual. This holds for both the cold situation (e.g., maximal non-shivering conditions without shivering; Vijgen et al., 2011) and for the thermoneutral condition (Kingma et al., 2011). Both conditions are important because the level of NST is calculated by the difference in energy expenditure of each condition. The importance of individual attuned environmental conditions in human metabolic studies has very often been neglected.

Many factors influence the thermoneutral zone, such as body composition (degree of obesity), fat distribution, clothing, energy expenditure, age, and gender (Kingma et al., 2011). These factors have the potential to introduce bias in study results and therefore need to be taken into consideration in studies on energy metabolism, obesity, medical conditions, thermal comfort, or vigilance. It is needed to go back to the basics of human thermoregulation. For instance, when baseline conditions are carried out at too low temperatures some NST may already occur. The NST calculated after cold exposure would then result in too low values. For more examples and the determination of the thermoneutral zone see Kingma et al. (2011).

The same applies for the experiments under mild cold conditions. The goal is to establish maximal BAT activity and NST under non-shivering conditions. During shivering much of the FDG label will be taken up by the skeletal muscle and the relation between CIT and BAT activity will be obscured by shivering thermogenesis due to competition in FDG uptake in the tissues. The few dedicated human BAT studies use a standard, more, or less fixed protocol under mild cold conditions (around 16°C) sometimes with additional ice cooling of the feet. It is disputed whether ice (or cold pressor test) is an effective way of stimulating BAT, because it works systemically (Nedergaard et al., 2011). BAT, however, is normally selectively stimulated via sympathetic nerve fibers that are activated by the coldsensitive centers in the brain.

In our recent studies we cool subjects until the start of shivering, which is monitored by EMG and questionnaires (Vijgen et al., 2011). Then we raise the temperature slightly, so that the shivering just stops. In this way we want to achieve maximal NST. It is assumed that the NST level will not increase when shivering starts, although this is hard to prove. Our current trials also indicate that the rate of cooling may affect the level of NST with a lower level of NST during fast cooling due to an earlier onset of shivering. Finally, we also obtained indications that for activating BAT it matters which parts of the body skin are exposed to the cooling medium.

In conclusion, several studies indicate differences in BAT activity and thermogenesis between individuals and between groups (lean and obese), but for more subtle and physiological relevant differences more care should be taken in the study design and the attuning of environmental conditions.