Brown Adipose Tissue—A Translational Perspective

Abstract

Brown adipose tissue (BAT) displays the unique capacity to generate heat through uncoupled oxidative phosphorylation that makes it a very attractive therapeutic target for cardiometabolic diseases. Here, we review BAT cellular metabolism, its regulation by the central nervous and endocrine systems and circulating metabolites, the plausible roles of this tissue in human thermoregulation, energy balance, and cardiometabolic disorders, and the current knowledge on its pharmacological stimulation in humans. The current definition and measurement of BAT in human studies relies almost exclusively on BAT glucose uptake from positron emission tomography with ¹⁸F-fluorodeoxiglucose, which can be dissociated from BAT thermogenic activity, as for example in insulin-resistant states. The most important energy substrate for BAT thermogenesis is its intracellular fatty acid content mobilized from sympathetic stimulation of intracellular triglyceride lipolysis. This lipolytic BAT response is intertwined with that of white adipose (WAT) and other metabolic tissues, and cannot be independently stimulated with the drugs tested thus far. BAT is an interesting and biologically plausible target that has yet to be fully and selectively activated to increase the body’s thermogenic response and shift energy balance. The field of human BAT research is in need of methods able to directly, specifically, and reliably measure BAT thermogenic capacity while also tracking the related thermogenic responses in WAT and other tissues. Until this is achieved, uncertainty will remain about the role played by this fascinating tissue in human cardiometabolic diseases.

Graphical Abstract

Graphical Abstract.

Essential Points.

• The current standard definition of brown adipose tissue (BAT) in humans is based on glucose metabolism measured with ¹⁸F-fluoro-deoxyglucose–positron emission tomography/computed tomography (¹⁸FDG-PET/CT).

• BAT is a thermogenic organ that is effectively recruited on acute and chronic cold exposure.

• BAT primary source of energy for thermogenesis is its own triglyceride (TG) content, with glucose and amino acids contributing to rapid intracellular TG repletion.

• The activation of BAT thermogenesis is primarily from sympathetic nervous system (SNS) outflow activated by cold exposure; modulation of BAT thermogenesis by local metabolites and systemic hormones such as glucocorticoids, thyroid, and sex hormones is possible, but still unascertained in humans.

• BAT glucose uptake is often reduced in conditions associated with insulin resistance, without a concomitant reduction of BAT thermogenic activity.

• Because BAT mass in vivo is currently defined from glucose uptake, BAT thermogenic capacity is likely underestimated in insulin-resistant states and its contribution to energy expenditure awaits new methods allowing more specific quantification.

• No drug tested thus far selectively activates BAT in humans; all drugs tested also activate white adipose tissue and/or cardiovascular responses that also contribute to whole-body energy expenditure.

Brown adipose tissue (BAT) displays the unique capacity to generate heat through uncoupled oxidative phosphorylation. Its thermogenic potential confers on small mammals, in which it is relatively abundant, the ability to survive in the cold without relying on shivering to generate heat. This outstanding thermogenic property makes BAT a very attractive therapeutic target for obesity and its cardiometabolic complications, although its presence in humans has been contested for years. First described in marmots by Gessner in 1551 (1) and identified early by Aherne and Hull in newborn infants (2) and then by Heaton in human necropsies (3), metabolically active BAT was demonstrated in vivo in adults in 2003 through positron emission tomography (PET) with the glucose analogue tracer ¹⁸F-fluorodeoxiglucose (¹⁸FDG) (4, 5). Three back-to-back papers in The New England Journal of Medicine in 2009 (6–8) then fanned the flames of BAT investigation not only in humans, but also in preclinical models. Since this evidence for metabolically active BAT in adult human was published, there has been an exponential accumulation of knowledge on the possible physiological and pathophysiological roles of this fascinating tissue.

We offer herein a review of the current state of knowledge on BAT, focusing on investigations in humans while offering a translational perspective on the pathophysiological roles of BAT, beige, and white adipose tissues (WAT) as an integrated thermogenic organ. We first overview the brown adipocyte cellular metabolism and then discuss the current functional definition of BAT and the tools employed to this effect. We review the control of BAT by the nervous and endocrine systems and local and circulating metabolites. A discussion on the plausible roles of BAT in human physiology, energy balance, and in cardiometabolic disorders follows. We finish with a review of the attempts at pharmacological activation of BAT in humans and by offering our perspective on gaps and future directions of clinical BAT investigation.

Cellular Biology of Brown Adipose Tissue

Transcriptional and Epigenetic Regulation of Thermogenic Adipocytes

Thermogenic, or brown adipocytes, display a typical molecular signature, including relatively high expression levels of uncoupling protein 1 (UCP1), cell death activator (CIDEA), and peroxisome proliferator–activated receptor gamma coactivator 1-alpha (PGC1A) in PRD1-BF1-RIZ1 homologous domain–containing 16 (PRDM16) positive cells (9–12), and histopathological features, including multiple small lipid vacuoles and rich mitochondrial content (3, 13–16), easily distinguishable from WAT. Thermogenic adipocytes develop from heterogeneous stem cells of mesodermal origin that variously express homeobox protein engrailed 1 (En1), myogenic factor 5 (Myf5), paired box protein 7 (Pax7), Pax3, and paired-related homeobox transcription factor 1 (Prx1) (17). Peroxisome proliferator–activated receptor gamma (PPAR-γ) and CCAAT/enhancer proteins (C/EBPs) are necessary, but not sufficient for brown adipocyte differentiation (18–20). Early B-cell factor-2 (EBF2) recruits PPAR-γ to selective gene-binding sites that promote brown adipogenesis and thermogenic programming of myoblasts and preadipocytes (21). PRDM16 is essential in Myf5+ cell commitment to brown adipogenesis (22, 23) and thermogenesis programming of adipocytes from this lineage (20, 24, 25). Its absence in mice does not however compromise classical BAT early development, but does impair WAT browning (ie, beige adipocytes) and thermogenic activity on chronic cold exposure (26) and impairs the maintenance of an effective interscapular BAT (iBAT) thermogenic phenotype throughout the life of the mouse (27). Indeed, thermogenic adipocytes distinct from the classic brown adipocyte (ie, PRDM16 negative) have been identified (28). Bone morphogenic protein 7 (BMP7) is another critical factor for brown adipogenesis and thermogenic programing (29). In addition to these prothermogenic transcriptional regulators, transcriptional brakes have been described (30) such as zinc finger protein 423 expressed by WAT that prevent the conversion of white adipocyte precursors to thermogenic adipocytes (31–33).

Epigenetic modifications such as DNA methylation are important regulators of adipose tissue function (34). DNA demethylation has generally been associated with stimulation of the adipocyte thermogenic phenotype. For example, alpha-ketoglutarate–stimulated demethylation under the control of adenosine 5′-monophosphate (AMP)-activated protein kinase (AMPK) increases PRDM16 expression and brown adipogenesis and thermogenesis (35). Lysine-specific demethylase-1 promotes brown adipocyte thermogenesis by repressing adipose tissue hydroxysteroid 11-β-dehydrogenase isozyme 1 and therefore reducing local corticosterone levels (36). Another histone, H3 lysine 9 demethylase, JMJD1A, upregulates β-adrenergic receptor and PPAR-γ and stimulates adipose browning (37). Histone deacetylation also regulates adipocyte thermogenesis. In mice, histone deacetylase 11 (HDAC11) suppresses iBAT and iWAT thermogenic programming and adaptive thermogenesis during cold exposure (38). Sirtuins (eg, SIRT1, SIRT5), NAD+-dependent histone deacetylases, are also involved in the regulation of brown adipocyte thermogenesis (39, 40). SIRT1 increases the expression of PRDM16 and brown adipocyte differentiation through deacetylation (40). SIRT1 furthermore reduces brown adipocyte apoptosis and endoplasmic reticulum (ER) stress during high-fat diet (HFD) conditions in mice (41). In addition, several microRNAs (miRNAs) are produced by white and brown adipocytes (42), and some have been shown to inhibit or stimulate brown adipocyte differentiation (summarized in Alcalá et al) (43), further supporting epigenetic regulation of brown fat recruitment. Epigenetic mechanisms have also been evoked in the regulation of the sympathetic output signal driving BAT thermogenesis. For example, hypothalamic miR-33, induced by ER stress, is implicated in increasing the sympathetic tone necessary for cold- and high-fat diet-induced thermogenesis in mice, through the reduction of γ-aminobutyric acid (GABA) A receptor–related gene expression (44).

The characterization of the heterogeneous origins and the transcriptional and posttranscriptional control mechanisms of brown adipocyte development and thermogenic programing is of outstanding importance for the development of future therapeutic avenues to exploit the unique BAT thermogenic properties. Readers are referred to more complete recent reviews on this important topic (45–47). At the moment, however, this knowledge is not directly applicable for the in vivo characterization or modulation of BAT function in humans.

Mitochondrial Function and Energy Uncoupling

The mitochondria of classical brown adipocytes have the unique capacity for large inner membrane proton conductance that diffuses the proton gradient independent from adenosine 5′-triphosphate (ATP) production, leading to heat production as the main product of the mitochondrial respiration of these cells. The presence and activation of UCP1, described more than 40 years ago (48), is responsible for this unique feature. There is still debate about the exact mechanism by which UCP1 exerts this profound mitochondrial uncoupling (49–51), but the basic function and regulation of UCP1 is felt to be the same between rodent models and humans. UCP1 is downregulated by ATP or other purine nucleotides binding (52, 53), and UCP1 binding to radiolabeled guanine nucleotide diphosphate (GDP) has been used as a functional marker of UCP1 activation in vitro (54, 55). “Unmasking” of these GDP binding sites occurs in BAT mitochondria after adrenergic or cold stimulation, independent of change in UCP1 protein expression (54, 55). A direct competition for purine nucleotide binding sites by long-chain fatty acids generated by intracellular triglyceride (TG) lipolysis, leading to induction of a protonophoric conformation of UCP1, has been proposed (56). The association between long-chain fatty acids with UCP1 may alternatively form protonable carboxylate groups in the mitochondrial matrix (57). Another possibility, the protonophoretic model, stipulates UCP1-independent intramitochondrial transport of protonated long-chain fatty acids, followed by deprotonation in the mitochondrial matrix and UCP1-dependent export or long-chain fatty acids (58, 59). Finally, the shuttling model suggests simultaneous transport of a long-chain fatty acid and a proton with the inability of UCP1 to release the long-chain fatty acid (60). Using magnetic nuclear resonance and functional mutagenesis, the binding of a long-chain fatty acid to UCP1 was shown to be necessary for UCP1-mediated proton flux (61). Whatever the precise molecular mechanism, long-chain fatty acids are thus generally considered to be the most likely activation signal of UCP1.

Intracellular TG lipolysis in brown adipocytes is the likely source of long-chain fatty acids for the activation of UCP1-mediated thermogenesis. This was supported in vivo by the demonstration of the inhibition of BAT thermogenesis using nicotinic acid–mediated suppression of intracellular lipolysis in rats and humans (62, 63). However, BAT-specific knockout (KO) of adipose tissue TG lipase (ATGL) or its activating protein CGI-58 demonstrated that BAT metabolic activity can also be fueled by nonesterifed fatty acids (NEFAs) from WAT lipolysis or from intravascular lipolysis of TG-rich lipoproteins (TRLs) (64, 65). It is also possible that another downstream metabolite of intracellular lipolysis or norepinephrine signaling in BAT may activate UCP1-mediated uncoupled respiration (51). For example, mice with adipose tissue–specific KO of adipose alpha/beta-hydrolase domain 6, normally hydrolyse 2-monoacylglycerol (2-MAG), display greater cold tolerance, enhanced WAT NEFA/TG cycling, and iBAT thermogenesis on cold exposure through 2-MAG–mediated activation of PPAR-α compared to control mice (66). Mitochondrial reactive oxygen species (ROS) during respiration also contributes to UCP1 activation in the brown adipocyte (67). Other lipids such as peroxisome-derived plasmalogens (ether phospholipids) may also be important for cold-induced adipose mitochondrial mass and thermogenesis (68).

Early studies by Golozoubova and colleagues (69) showed that only shivering thermogenesis could compensate the absence of UCP1 for cold-induced thermogenesis. However, a series of alternative nonshivering thermogenic mechanisms were proposed following the demonstration that UCP1 KO mice gradually adapted to cold display–increased iWAT energy expenditure (70), a mechanism dependent on the presence of leptin (71). Creatine has been shown to drive a futile cycle leading to energy expenditure in beige and brown adipocytes able to compensate for the absence of UCP1 in mice (72). Creatine availability and transport to adipocytes is key to this thermogenic mechanism (73). However, creatine supplementation failed to increase cold-induced thermogenesis and BAT ¹⁸FDG accumulation and volume of activity in vivo in healthy individuals on a diet characterized by low creatine intake (a vegan diet) (74). More studies are needed to understand the implication of cellular creatine availability for BAT thermogenesis in vivo in humans.

Creatine kinase B, which is activated by cAMP in brown adipocytes, activates a creatine-mediated futile cycle liberating very large amounts of adenosine 5′-diphosphate (ADP) to accelerate basal and β-adrenergic–stimulated cellular respiration in a UCP1-independent fashion; the absence of creatine kinase B in BAT or in all adipose tissues leads to increased weight gain and higher glucose levels in mice (75). The protein catalyzing phosphocreatinine hydrolysis process in BAT mitochondria was recently shown to be tissue-nonspecific alkaline phosphatase (TNAP), which is potently induced by cold exposure (76). Genetic ablation of TNAP in adipocytes leads to reduced whole-body resting energy expenditure and obesity in mice (76).

One important ATP-requiring process in thermogenically activated brown adipocytes is lipogenesis, which theoretically uses 24% of the energy generated from the metabolism of the glucose molecules to drive this process (77). ATP production is thus necessary to sustain lipogenesis in adipocytes (78). Glycerol-3-phosphate synthesis from glycolysis and glyceroneogenesis, and esterification reactions needed for TG synthesis, are also ATP-requiring processes (77). The in vivo energy cost of TG deposition is 1.5 to 2 times higher when both fatty acids and glucose are available compared to fatty acids alone (79). Because very rapid rates of TG synthesis and glucose uptake occur in thermogenically active BAT (80), important utilization of ATP is thus expected in brown adipocytes.

It is therefore highly likely that a substantial fraction of the thermogenic adipocyte energy expenditure is driven by UCP1-independent processes. Recently, single-nuclei RNA-sequencing analyses of mouse WAT revealed heterogeneity of thermogenic adipocytes in response to cold or beta-3 adrenergic stimulation: One population displays the classic beige thermogenic program with increased mitochondrial oxidative genes, whereas another is characterized by increased expression of genes involved in NEFA/TG cycling (81). Intercellular exchange of energy metabolites between adipocytes within BAT is therefore likely (discussed in subsequent sections).

In summary (Fig. 1), UCP1-mediated mitochondrial uncoupling is the primary driver of the remarkable thermogenic activity of BAT in rodents and humans. While a phosphocreatine/creatine futile cycle may contribute to further reduction of the ATP/ADP ratio of brown adipocytes, ATP production is nevertheless essential to sustain BAT TG synthesis that in turn constitutes the primary source of fatty acids driving UCP1-mediated thermogenesis. A high rate of intracellular NEFA/TG cycling is an important energy-requiring process that characterizes this tissue. It would be very interesting to apply imaging methods able to tract changes in cellular ATP production (eg, ³¹P-magnetic resonance spectroscopy [MRS]) and thermogenesis (¹¹C-acetate or ¹⁵O-O₂ PET) to simultaneously track these metabolic rates and determine the relative contribution of uncoupled vs coupled respiration in activated human BAT. To the best of our knowledge, no group has performed such a study thus far.

Figure 1.

Brown adipocyte energy metabolism. Long-chain fatty acids (FA-CoA) activate uncoupling protein 1 (UCP1) and are the major energy source of the brown adipocyte thermogenesis. The main source of these FA-CoA is intracellular triglyceride (TG) lipolysis, but circulating nonesterified fatty acids (NEFA) and triglyceride-rich lipoproteins (TRL), through lipoprotein lipase (LPL)-mediated lipolysis, also contribute fatty acids to drive thermogenesis. Glucose, branched-chain amino acids (BCAA), glutamate, and other sources of energy contribute mainly to drive anaplerosis and cataplerotic processes such as de novo lipogenesis (DNL) and glycerol synthesis that are essential to replete intracellular triglycerides and to sustain the very high rate of TG/nonesterified fatty acid cycling necessary for brown adipose thermogenesis. In addition to UCP1, a phosphocreatine/creatine (PCr/Cr) futile cycle contributes to reduce the ATP/ADP ratio and drive mitochondrial thermogenesis. ADP, adenosine 5′-diphosphate; ATP, adenosine 5′-triphosphate; BCAA, branched-chain amino acids; Cr, creatine; DNL, de novo lipogenesis; FA-CoA, long-chain fatty acyl coenzyme A; LPL, lipoprotein lipase; NEFA, nonesterified fatty acids; TG, triglycerides; PCr, phosphocreatine; TCA, tricarboxylic acid cycle; TRL, triglyceride-rich lipoproteins; UCP1, uncoupling protein 1.

Energy Substrate Metabolism

As discussed in a subsequent section, glucose metabolism has been the first and still is the main in vivo process by which BAT is identified and defined in vivo. BAT glucose uptake is stimulated by norepinephrine and occurs concurrently with activation of BAT thermogenesis. We have extensively reviewed BAT glucose metabolism in previous works (80, 82). The major conclusions of these reviews are still currently relevant. From a pathophysiological standpoint, 2 key points must be emphasized. First, glucose is mostly used for lactate and TG synthesis in BAT (83–85), not to drive thermogenesis. Glucose contributes to the synthesis of fatty acids esterified into TGs that are simultaneously hydrolyzed to drive brown adipocyte thermogenesis (86). Glucose availability increases insulin-mediated TG synthesis and fatty acid storage in adipocytes in vitro (87–89), but also in vivo (90). Carbohydrate-response element-binding protein (ChERBP), which is activated by cold exposure and by increased glucose availability (91), and which controls the genes that drive de novo lipogenesis, plays an important role in BAT TG accumulation in mice (92). Rats adapted to a carbohydrate-poor, protein-rich diet display reduced BAT de novo lipogenesis (93, 94). The number of acyl-chain double bonds and methylene-interrupted double bonds is lower in BAT vs WAT, suggesting higher levels of saturated fatty acids from de novo lipogenesis in BAT (95). Therefore, glucose metabolism likely contributes to maintain BAT TG content, its primary energy supply for cold-induced thermogenesis. Second, BAT glucose uptake is dependent of insulin, not only of norepinephrine-induced thermogenic activation, and insulin resistance (IR) and low BAT glucose uptake may occur without reduction in acute cold-induced BAT thermogenesis (discussed in previous and subsequent sections). To the best of our knowledge, in vivo BAT glucose metabolism under insulin vs noradrenergic stimulation has not been directly measured in animal models or humans with methods able to determine oxidative vs nonoxidative glucose metabolism (ie, ¹¹C-glucose PET). Therefore, the metabolic fate of glucose in BAT under these 2 different stimulations is unknown.

Intracellular TG lipolysis is critical for the acute stimulation of BAT thermogenesis in vitro (96) and in vivo in rats (62) and humans (63). BAT TG content is reduced (97–107) and BAT glycerol release is enhanced (84) during acute cold exposure in humans. From the glycerol release, it has been estimated that approximately 65 nmol/g/min of NEFA are released in this condition (84). This is likely an underestimation given the presence of high levels of glycerol kinase in human BAT (84, 108), that allows the recycling of glycerol produced during lipolysis for TG resynthesis. BAT TG content reaches a nadir within 35 minutes and plateaus thereafter in young healthy men on cold exposure (105). This demonstrates rapid and active TG replenishment during BAT metabolic activation. Currently, no one has measured the rate of this BAT TG/NEFA cycling in vivo in humans during BAT metabolic activation. Studies are ongoing in our laboratory to quantify this intracellular BAT TG/NEFA cycling during acute cold exposure in humans using the combination of ¹¹C-acetate, ¹¹C-palmitate, ¹⁸FDG-PET and magnetic resonance imaging (MRI) methods (Clinicaltrials.gov No. NCT05092945).

In rodents, fatty acids in circulation are also an important source of substrates to drive BAT thermogenesis. In mice, genetic KO of genes essential for BAT intracellular TG lipolysis leads to upregulation of the utilization of circulating fatty acids to drive thermogenesis (64, 65). The absence of BAT intracellular lipid droplets in BAT-specific DGAT1 + DGAT2 KO mice does not prevent BAT thermogenesis and results in increase glucose and circulating fatty acid utilization to drive adaptive thermogenesis and resistance to diet-induced glucose intolerance (109). Fatty acid transport protein, fatty acid binding protein, and CD36 are expressed in brown adipocytes and are important for thermogenesis (110–112). In mice, activated BAT has the capacity to clear most circulating TRL lipids (112–114). Lipoprotein lipase (LPL) expression is increased during cold exposure specifically in BAT, but not in WAT in humans (115). Serum angiopoietin-like 4 (ANGPTL4), which inhibits LPL activity, is increased on cold exposure in humans (116). In mice, BAT Angptl4 expression is however reduced, whereas that of WAT is increased during cold exposure, providing a potential mechanism to shuttle TRL TG content to BAT via LPL-mediated lipolysis to provide fatty acids to drive thermogenesis (117). In addition to their TG content, TRL particles can also be cleared by BAT in rodents (112–114). ¹⁸F-labeled BODIPY-TG-chylomicron-like particle uptake by BAT is increased in mice on acute, but not chronic, cold stimulation (118). TRL particles are most likely taken up by BAT endothelial cells through endocytosis; lysosomal acid lipase then hydrolyses TRL-TG, which stimulates endothelial cell beta oxidation and activates hypoxia-inducible factor-1 alpha-dependent proliferation of endothelial cells and adipocyte precursors (119).

However, most studies in humans showed no substantial change or even increase in plasma TG levels during acute cold exposure (63, 101, 103, 120, 121). Four-week cold acclimation that increases BAT thermogenic activity up to 2.6-fold on acute cold exposure also does not lead to a reduction in fasting (122) or even postprandial levels of total, chylomicron or very low-density lipoprotein (VLDL) TG levels (104). In a later study, we directly measured BAT uptake of dietary fatty acids, which are transported to tissues largely as chylomicron-TG, using the oral ¹⁸F-FTHA PET method (123). During cold exposure, BAT takes up dietary fatty acids at greater rates than subcutaneous WAT and resting skeletal muscles, but at lower rates than the heart and the liver (104). BAT dietary fatty acid uptake is also unchanged in the face of a 2.6-fold increase in BAT oxidative metabolism after cold acclimation and accounts for only 0.3% of total body utilization of dietary fatty acids.

Activated BAT uses succinate (124, 125) and branched-chain amino acids (126) at high enough rates to provide a systemic metabolic sink for these metabolites in mice. BAT also uses glutamate at greater rates during acute cold exposure in humans, but at much lower rates than circulating glucose (84). Circulating valine, an anaplerotic substrate (127), is reduced during cold exposure in humans, but only in individuals displaying positive BAT ¹⁸FDG uptake (126). However, skeletal muscles are the most important site for branched-chain amino acid metabolism in mice and humans (128). Therefore, metabolic activation of skeletal muscles is a much more likely explanation for cold-induced changes in circulating branched-chain amino acid levels in humans (129).

One possible metabolic fate of branched-chain amino acids (valine, isoleucine), glutamate, and succinate is the anaplerotic/cataplerotic pathways, that is, glyceroneogenesis and de novo lipogenesis (127). Glyceroneogenesis (ie, the production of glycerol-3-phosphate from pyruvate, lactate, and amino acids) is essential for TG synthesis in mice (130). The absence of PEPCK, the rate-limiting enzyme for glyceroneogenesis, causes a marked reduction in WAT and BAT TG content (83, 131). Glyceroneogenesis is stimulated in adipocytes by norepinephrine and by PPAR-γ stimulation (130, 132–135). It is also activated in vivo in rats by cold exposition (136) even on a carbohydrate-free, protein-rich diet (94), suggesting that amino acids can substitute for carbohydrates to sustain BAT TG synthesis. UCP-1–driven KO of the branched-chain amino acids transporter SCL25A44 results in considerable reduction in BAT thermogenesis (126). However, in mice exposed to mild cold, branched-chain amino acids account for only approximately 6% of Krebs cycle carbon flux (128) and are therefore not important contributors for BAT cataplerotic pathways.

In summary (see Fig. 1), fatty acids are the most important energy substrate to drive BAT thermogenesis. Its source is primarily the brown adipocytes’ TG content on acute activation, although circulating NEFA and TG may likely increasingly contribute to thermogenesis with sustained BAT activation. Intracellular brown adipocyte TG/NEFA cycling is very rapidly activated with ongoing and rapid replenishment of intracellular TG from glucose and other anaplerotic/cataplerotic sources such as branched-chain amino acids. More studies are needed to characterize this BAT TG/NEFA cycling in vivo on metabolic activation.

In Vivo Methods to Define Brown Adipose Tissue

PET coupled with computed tomography (PET/CT) and MRI-based methods have been thus far the most important modalities for the in vivo investigation of BAT in humans. Thermographic and other optic methods were proposed very early to study BAT activity in vivo (137, 138). These methods may be more informative in preclinical studies because of the capacity to use fluorescent labels, the much smaller size of the animals, and the more superficial iBAT depot (typically within 5 mm of the skin surface in rodents, whereas supraclavicular BAT in adult humans is > 5-10 mm deep, varying widely depending on the thickness of the subcutaneous adipose tissue depot). However, they are very limited by their low penetration depth and by the nonspecific signal mixing superficial vasculature of the skin, subcutaneous fat, and muscle in addition to any BAT-mediated signal. Other emerging methods such as contrast-enhanced echography have been proposed, but have not been largely used in human investigations. Excellent reviews exist of these various methods (139–143). We restrict our discussion herein to some of the PET/CT and MRI-based methods that have provided major insights into BAT metabolic function.

From an integrative physiology and clinical point of view, BAT is best defined by its function as a thermogenic adipose tissue (80). It is the revelation of highly metabolically active adipose tissues by ¹⁸FDG PET/CT that convinced the broad scientific community of the existence and possible physiological relevance of BAT in adult humans (6–8, 144–146). The mere presence of molecular and histopathological features of adipose tissue browning however does not necessarily translate into in vivo significant thermogenic capacity, as shown by the absence of detectable in vivo thermogenic activity despite robust browning of WAT after prolonged cold exposure in rodents (147, 148). The molecular signature of supraclavicular BAT depots in humans is more similar to that of “beige” than brown adipocytes of rodents (149). Ex vivo mitochondrial respiration of brown adipocytes appears increased in mice acclimated at room temperature vs in humans (150). Despite these differences, Ucp1 content is similar between human and mouse BAT (150) and the typical supraclavicular adipose tissue depots in humans display increased in vivo thermogenic activity on acute cold exposure (101), as does the iBAT in rodents (147, 148).

There have been attempts at developing noninvasive methods to identify BAT using PET tracers targeting specific molecular ligands (151) or outer mitochondrial membrane proteins (ie, translocator protein [TSPO]) (152–155) instead of its metabolic function. However, with the exception of one small human study (n = 3 participants) that reported relatively selective ¹¹C-PBR28 (a TSPO radioligand) uptake in BAT vs WAT at room temperature (155), no such method has been used in humans. MRI with fat fraction and/or T2* mapping to quantify mitochondrial content from the signal of iron-containing heme has also been used, but with limited capacity to differentiate BAT from WAT (reviewed in [139–143]). We therefore favor the definition of BAT as an adipose tissue displaying substantial thermogenesis in vivo.

The next question is how best to detect and measure this thermogenic adipose tissue. Currently, the combination of high glucose uptake using intravenous (IV) ¹⁸FDG administration with PET in a tissue displaying the anatomical characteristics and high fat content compatible with adipose tissue, determined by CT or MRI, is still the standard definition of BAT (80, 156). This glucose-using adipose tissue is scattered in multiple small depots in the supraclavicular, paravertebral, pericardial, and suprarenal regions (157). Measured using this method, BAT volume in human adults spans 2 orders of magnitude from a few to hundreds of milliliters (82). This huge variability depends largely on environmental exposure to cold (102, 158–160), but also on biological factors such as age, sex, visceral adiposity, IR/diabetes, cardiometabolic risk, and circadian rhythm (8, 146, 157, 161–169) and the use of some drugs (eg, β-adrenergic agonists and antagonists) (121, 170–172).

Critically, BAT glucose uptake is not a measure of thermogenesis. First, a large fraction of BAT glucose uptake is metabolized into lactate or glycerol ex vivo (83), which was confirmed in vivo during acute cold exposure in humans (84). Second, glucose uptake in BAT does not need the activation of thermogenesis in rodents (173, 174) and can be dissociated from in vivo thermogenic activity in humans (103). Third, insulin administration increases BAT glucose uptake, but not blood flow, suggesting dissociation between glucose uptake and thermogenesis under varying degree of insulin stimulation (175). BAT glucose uptake is reduced in genetic variants of IR (176) and in conditions that induce IR such as prolonged fasting (177), glucocorticoid treatment (178), chronic ephedrine administration (179), and fructose overfeeding (Richard, Blondin, Carpentier et al. Unpublished). In the latter randomized, controlled, crossover study, 2-week high-fructose, but not high-glucose, feeding led to a significant reduction of cold-induced BAT glucose uptake without a change in thermogenesis and before any significant change in systemic IR. This demonstrates the exquisite sensitivity of BAT glucose uptake to dietary and potentially other lifestyle changes, drugs, and health conditions leading to deterioration of cardiometabolic health, without necessarily altering BAT thermogenic capacity.

BAT thermogenesis can be assessed directly using the ¹⁵O-O₂ (180, 181) or the ¹¹C-acetate (101) PET methods. We used the ¹¹C-acetate combined with ¹⁸FDG PET method for 3-dimensional mapping of supraclavicular BAT thermogenic activity in vivo and found a large degree of heterogeneity of response to acute cold stimulation vs ¹⁸FDG activity. Unfortunately, these methods rely on PET dynamic scanning as the tissue metabolism of oxygen and acetate is very fast. Although ¹¹C-acetate can be modeled relatively simply using monoexponential function fitting of the rapid BAT tissue signal decline to assess oxidative metabolism and tissue peak activity to assess blood flow (101), this method does not assess nonoxidative metabolism of ¹¹C-acetate. Multicompartmental modeling of ¹¹C-acetate (182) offers more specific assessment of these parameters and has the added advantage of assessing acetate retention into tissue metabolism (ie, anaplerotic/cataplerotic pathways) (127). More studies using this novel method are needed to determine BAT nonoxidative metabolism of ¹¹C-acetate in humans. Ideally, the addition of another method to measure tissue interstitial volume (eg, contrast-enhanced MRI) is necessary for the optimal precision of ¹¹C-acetate PET determination of BAT oxidative metabolism (183). Unless one has access to a PET scanner with an extended field of view including the neck, thorax, and abdomen (184), these methods are also not amenable to image all of the BAT-containing fat depots and therefore cannot measure the entire BAT volume and thermogenic capacity. The very short radioactive half-lives of these tracers also necessitate a cyclotron and radiochemistry facilities at the site of scanning, making these methods inapplicable widely. Finally, ¹⁵O and ¹¹C produce lower-resolution PET images than ¹⁸F (full width of half maximum of 2.48, 0.92, and 0.54 mm, respectively) (185), primarily because of the energy of their emitted positrons, which determines their diffusion range.

Alternative methods have been proposed to assess BAT in vivo in humans. Imaging of sympathetic nerve activity by exploiting the presynaptic reuptake of norepinephrine demonstrated by the 1970 Nobel Prize in Physiology or Medicine winner Ulf von Euler is very attractive and has been attempted with different PET tracers including ¹¹C-epinephrine, ¹⁸F-fluoro-norepinephrine, ¹¹C-metahydroxyephedrine (¹¹C-mHED), ¹⁸F-fluoro-propoxy-benzylguanidine, and ¹⁸F-fluoro-dopamine (186–189). Sympathetic activation is the endogenous driver of BAT thermogenesis, but is not necessarily directly proportionally to the ensuing thermogenic response. Furthermore, this method would not allow detection of the BAT thermogenic response to exogenous stimulation (ie, β-adrenergic agonists).

The most promising current approach to measure total thermogenic adipose tissue mass is through the detection of the fat fraction shift that occurs during activation of BAT thermogenesis. BAT TG content is hydrolyzed and mobilized within 1 to 3 hours through sympathetically stimulated intracellular lipolysis. This response can be seen using CT radiodensity or the MRI Dixon method or proton MRS that demonstrates a shift of BAT water-to-fat ratio, not observed in WAT or in shivering muscles (97–107). Disappearance of intracellular BAT TG, as opposed to glucose uptake, is not necessarily affected by age and type 2 diabetes (T2D) status at equivalent cold exposure (103). In vivo inhibition of intracellular TG lipolysis using nicotinic acid suppresses this BAT shift of water-to-fat ratio and inhibits BAT thermogenesis in rats and humans (62, 63). Three-dimensional mapping of this shift is possible using the MRI Dixon method (100, 190, 191). In healthy young men during acute cold exposure, the supraclavicular BAT fat fraction declines in voxels displaying 60% to 100% fat fraction at baseline, whereas it increases in voxels displaying below 30% fat fraction (192). Therefore, a cold-induced shift in BAT water-to-fat ratio measured with this method is quite heterogeneous. Furthermore, in vitro experiments have shown that up to 50% of fatty acids hydrolyzed by BAT could be released in the extracellular media (193) and subsequently oxidized or reesterified elsewhere. Therefore, more studies are needed to understand in situ BAT TG and fatty acid metabolism to interpret appropriately the dynamic BAT changes in water-to-fat ratio.

A promising and very versatile MRI-based modality for BAT imaging is deuterium metabolic imaging. This MRS technique allows the noninvasive imaging of tissue metabolism of deuterium-labeled tracers to study rapidly proliferative cells (194), hormone (195), or energy substrate metabolism (196). A clear added value of this method over that offered by PET is the possibility of specifically tracing the appearance of downstream metabolites (eg, deuterated lactate and glutamate from deuterated glucose) in tissues. BAT glucose uptake and metabolism into lactate and glutamate was reported in cold-acclimatized rats in one study using deuterium metabolic imaging, showing promising results (197). One human study is ongoing (Clinicaltrials.gov No. NCT04060745) using this modality after oral ingestion of deuterated glucose. Because deuterium-labeled fatty acids are available and safe for use in humans, this method appears very promising for future in vivo investigations of BAT fatty acid uptake and esterification into TG.

There is clearly a gap between the in vivo definition of BAT, resting solely on imaging, and the presence of BAT using histopathological methods. Because BAT thermogenesis is the feature that defines BAT in vivo, BAT thermogenesis has to be activated during imaging, that is, some stimulus needs to be applied. In turn, this stimulus needs to be controlled to be able to compare experimental groups or treatments. In our opinion, standardized cold stimulus applied to a large proportion of the body surface area that results in the same rate of heat loss (ie, same differential temperature in and out of the cold-water perfusion system) and therefore the same increase in whole-body energy expenditure is the best condition to measure BAT thermogenic response in humans.

In summary, the current standard definition of BAT still rests on ¹⁸FDG-PET/CT, as no other method has yet gained wide recognition and applicability in humans (Fig. 2). This definition is based on BAT glucose metabolism, not thermogenesis. This has profound implications for the interpretation of most of the current human data on BAT, as will be discussed in the next sections.

Figure 2.

Definition of brown adipose tissue through metabolic imaging. First, computed tomography (CT) or magnetic resonance imaging (MRI) is necessary for anatomic definition and quantification of tissue fat content. Second, metabolic function of the fat tissue needs to be measured. The standard procedure for the latter is positron emission tomography (PET) with intravenous administration of ¹⁸F-fluoro-deoxyglucose (¹⁸FDG) that measures glucose uptake. Other experimental approaches can provide measurement of other important characteristics of brown adipose tissue such as oxygen utilization or carbon dioxide production (thermogenesis), fatty acid uptake and/or oxidation, intracellular triglyceride (TG) mobilization, mitochondrial content, or sympathetic activity. CT, computed tomography; DFA, dietary fatty acids; ¹⁸FDG, ¹⁸F-fluoro-deoxyglucose; FSF, fat signal fraction; ¹⁸FTHA, ¹⁸F-fluoro-thia-heptadecanoic acid; MRI, magnetic resonance imaging; MRS, magnetic resonance spectroscopy; NEFA, nonesterified fatty acids; PET, positron emission tomography; TG, triglycerides; TSPO, translocator protein.

Regulation of Brown Adipose Tissue Activity and Capacity

Central Nervous System Regulation of Brown Adipose Tissue Activity and Capacity

Two models have been proposed to describe the homeostatic control of body temperature: a feed-forward or feedback model. The first refers to a preemptive increase in thermogenesis in response to skin cooling, resulting in the stimulation of thermoeffectors before any changes in body temperature occurs (198). The second model refers to body temperature being regulated by “independent thermoeffector loops, each having its own afferent and efferent branches,” with each thermoeffector being triggered by a unique combination of superficial (skin) and deep (core) temperature (199). Although BAT appears to express temperature-sensitive receptors, they likely modulate the thermogenic activity rather than serve as the primary triggering signal.

BAT thermogenic capacity (brown cell differentiation, proliferation, mitochondrial biogenesis, and increased UCP1 protein content) and thermogenic activity (uncoupling activity and activation of intracellular TG lipolysis) are under a dynamic brain control of BAT’s sympathetic output innervation (200–204). Different sympathetic activating stimuli lead to specific patterns of fat depot stimulation, with cold predominantly stimulating BAT and WAT (205) and hypoglycemia predominantly stimulating WAT (206). It is noteworthy that the brain control of BAT and WAT has essentially been studied in laboratory rodents, mainly in rats, hamsters, and mice, in which classic brown adipocytes and recruitable beige fat cells typically develop in iBAT and inguinal WAT (ingWAT), respectively (207, 208). In both iBAT and ingWAT, Ucp1-expressing adipocytes are under major sympathetic nervous system (SNS) control (201, 209, 210) and, accordingly, these cells are surrounded by a very high density of SNS nerve-ending varicosities, which, in contrast, are only sparsely present in Ucp1-deprived white adipocytes (211, 212). An anabolic role for parasympathetic efferent vagal signal, with insulin-mediated increased in glucose and fatty acid uptake and stimulation of leptin synthesis, has been suggested in WAT in rodents (213). However, others failed to find immunohistochemical evidence for parasympathetic innervation of adipose tissues (214). Furthermore, iBAT and most other BAT depots in rodents do not display histological evidence of cholinergic postganglionic parasympathetic nerves (215). Therefore, the consensus is that the parasympathetic nervous system plays no substantial role in adipose tissues (210). As for cortical involvement in thermoregulation, studies performed in rodents suggest it is limited to behavioral thermoregulatory responses to the perception and discrimination of cutaneous temperature, without triggering of the thermoeffector responses, including BAT metabolism (198).

The SNS efferent pathways, which innervate iBAT and ingWAT distinctly, have recently further been delineated in studies carried out in mice (211, 212, 216). Those studies have elegantly demonstrated that murine iBAT is innervated by SNS preganglionic neurons emerging from the intermediolateral column (IML) of the spinal cord at the levels of the thoracic vertebrae T2 to T8, which synapse with postganglionic neurons found in the stellate and sympathetic chain ganglia T2 to T5 (212, 216). For its part, ingWAT is innervated in mice by SNS preganglionic neurons leaving the IML from T7 to the first lumbar vertebra (L1), which connect with postganglionic neurons emerging from the thoracic/lumbar chain ganglia T12, T13, and L1 (211).

The brain autonomic centers governing the SNS-mediated activity of brown/beige adipocytes are essentially located in the hypothalamus and brainstem (200–204), which are known to participate in most homeostatic regulations. The hypothalamic structures involved in the brain control of SNS-mediated function of UCP1-expressing adipocytes include the preoptic area (POA), dorsomedial hypothalamus (DMH), dorsal hypothalamic area (DHyA), arcuate nucleus (ARC), paraventricular hypothalamus (PVH), lateral hypothalamus (LH), and ventromedial hypothalamus (VMH). Neurons from those structures connect to the SNS outflow via the spinal cord IML column either directly or via premotor brainstem nuclei such as the raphe nuclei, which include the raphe pallidus (RPa) (200–204).

Most areas implicated in the SNS control of iBAT and ingWAT, namely the POA, RPa, PVH, DMH, LH, and brainstem nuclei, receive sensory inputs from the fat depots, pointing toward an SNS outflow-sensory feedback loop that exists both in iBAT and ingWAT depots to regulate thermogenesis and NEFA mobilization, respectively (217–221). The sensory innervation of iBAT, whose removal induces the atrophy of the tissue (222), significantly contributes to the control of thermogenesis by signaling the brain about iBAT’s thermal status (222), blood flow (219), or intracellular lipolytic activity (209). In WAT (including ingWAT), sensory nerves participate in the control of lipolysis by sensing lipolytic products (223). Local neuronal sensing of WAT lipolysis and/or change in temperature may contribute to activating BAT sympathetic activity through afferent nerve signaling to the brain (223, 224). Of note, iBAT appears under a weaker sensory control than WAT (218, 219).

The brain centers and circuits governing the SNS outflow to BAT and WAT are largely determined by 2 major processes, namely, body temperature and body energy homeostasis, whose brain regulations are fairly distinct.

Body temperature regulation entangles an extensive LPB-POA-DMH/DHyA-RPa-IML-SNS outflow pathway responsive to cold and heat stimuli that controls the activity of both iBAT and ingWAT to ensure body temperature stability (thermoregulatory thermogenesis) (202). The activation of this pathway by cold exposure increases the thermogenic capacity in both fat depots. Of note, the thermogenic activity of ingWAT, at least in mice (147), appears limited, even in iBAT-denervated cold-exposed animals in which iBAT becomes functionally inoperative and in which ingWAT capacity is enhanced (225). The current models propose that the POA, through excitatory glutamatergic neurons and inhibitory GABAergic, controls the activity of sympathoexcitatory neurons found in the DMH that project to the RPa to ultimately influence the activity of the brown adipocytes (226). Concretely, skin cooling leads to the stimulation of excitatory LPB glutamatergic neurons, which project to the POA, to trigger the concurrent stimulation of median preoptic nucleus excitatory glutamatergic neurons and inhibition of the medial preoptic area inhibitory GABAergic neurons (226). This results in stimulation of the DMH sympathoexcitatory neurons, which in turn excite the RPa neurons innervating SNS-mediated brown and beige fat depots (226).

BAT and WAT thermogenesis are involved not only in temperature regulation but also likely participate in energy homeostasis, which is acknowledged at least in small mammals. Food restriction (energy shortage) and overfeeding (energy surfeit) have been reported to respectively reduce and stimulate thermogenesis, thereby altering energy expenditure to achieve the stability of energy stores (1). The observation that iBAT (227) and ingWAT (228) polysynaptically connect to hypothalamic and brainstem centers implicated in the regulation of energy balance and the direct demonstrations that those centers govern the activity of iBAT and ingWAT tend to further support a genuine role for brown and beige adipocytes in energy homeostasis regulation. The main brain structures involved in such regulation include the ARC, PVH, DMH, and VMH (200, 204, 229). Those nuclei accommodate neurons that control energy intake and energy expenditure, while responding to homeostatic signals informing about energy balance status, which include variations in the leptin, insulin, and ghrelin levels (230–232). Those neurons are arranged in circuits or systems assembled to regulate energy reserves (233).

One key energy homeostasis regulator, which appears to be particularly important in the control of iBAT and ingWAT activities, is the brain melanocortin system (BMS) (234–237). This system essentially consists of distinct neuropeptidergic ARC neurons that either express proopiomelanocortin (POMC) or agouti-related peptide (AgRP) and neuropeptide Y (NPY) as well as widely distributed brain neurons that express the melanocortin receptors 3 (MC3R) and MC4R. MC4R is viewed as the main melanocortin receptor involved in energy homeostasis (238–241). It is for instance expressed in the PVH (242) toward which POMC and AgRP/NPY neurons project, and which constitutes with the ARC a major duet in energy homeostasis regulation (243). POMC neurons release α-melanocyte–stimulating hormone (α-MSH), a peptidergic fragment emerging from the POMC cleavage acting as a physiological activator of the MC4R, whose action is in turn opposed by AgRP, a neuropeptide regarded as an inverse agonist on the MC4R (244). The catabolic role of POMC and MC4R activation as well as the anabolic action of AgRP have been demonstrated in various laboratory rodent models (245). In humans, the loss of function of the MC4R has been reported as being the most common cause of monogenic obesity (246) and conversely gain of function of MC4R, driven by MC4R variants that allow for an enhanced β-arrestin recruitment and the activation of MAPK pathway, protects against weight gain (247). The genuine role of the BMS in energy homeostasis is further supported by the sensitivity this system has toward homeostatic signals such as leptin level variations, which affect POMC neurons either directly or indirectly via ARC interneurons (232, 243, 248, 249).

There is a bulk of neuroanatomical and functional evidence to emphasize the major role of the MC4R, hence the BMS, in the control of both iBAT and ingWAT metabolism (200, 204, 209, 239). Transneuronal tract tracing studies have been instrumental in establishing the (poly)synaptic connections between the brain MC4R-expressing neurons and the SNS outflow to iBAT (250) and ingWAT (217). In the Siberian hamster, the PVH represents the hypothalamic region with the highest absolute number of MC4R neurons connected to the SNS outflow to iBAT (84% of some 1200 PVH neurons connected to iBAT express Mc4r) (250), supporting the role of the ARC-PVH MC4R neuronal axis in iBAT homeostatic thermogenesis. Once challenged (251), the role of PVH MC4R neurons in the control of iBAT thermogenesis and energy homeostasis now seems confirmed. Reinstatement of the MC4Rs selectively in the PVH (through cells expressing single-minded homolog 1 neurons—Sim 1) rescues the thermogenic effect of the MC3R/MC4R agonist melanotan 2 in Mc4R-null mice (252) and, moreover, the chemogenetic activation of PVN MC4R neurons stimulates iBAT thermogenesis (253). Those findings support our own data showing that the activation of the PVH MC4R with melanotan 2 stimulates iBAT thermogenesis (250).

The full circuit linking the ARC-PVH MC4R neuronal axis to the SNS outflow to iBAT and ingWAT, however, remains to be fully disentangled. The PVH MC4R neurons governing the SNS outflow to iBAT have been reported to be glutamatergic (252) but do not seem to be oxytocinergic (254) despite the evidence that Mc4r is expressed in oxytocin-positive cells in rats (255) and even in humans (256). PVH cells positive for oxytocin not only send direct projection to the IML, but have also been shown to be (poly)synaptically connected to iBAT (257). The PVH MC4R neurons project directly to the IML and to the brainstem, in numerous areas known to accommodate premotor neurons that connect to the SNS outflow to iBAT thermogenic cells (253). Meanwhile, the role of PVH MC4R neurons in the control of WAT Ucp1-expressing cells has yet to be deciphered. The thermogenic role of the brown adipocyte islands found in WAT depots such as ingWAT remains controversial (147, 225).

The ARC and PVH are not the only hypothalamic nuclei acting on brown and beige adipocyte to modulate energy homeostasis. The VMH and the DMH have also been described to play a role in the control of both iBAT and ingWAT and in the regulation of energy homeostasis (200). The role of the VMH in iBAT thermogenesis has been alleged for years but somewhat questioned because of the inability of viral tracing methods to establish an iBAT-VMH connection (227). The demonstration of the importance of the steroidogenic factor 1 neurons, which are specifically expressed in the VMH, in the control of iBAT thermogenesis further supports the functional link between the VMH and iBAT (258). The inability of viral tracing to link iBAT to the VMH remains mysterious. Similar to that of the VMH, the role of the DMH in thermoregulation (259) and energy homeostasis (260, 261) has been known for years. As outlined earlier, the DMH is a key node in the LPB-POA-DMH/DHyA-RPa-IML-SNS outflow pathway controlling iBAT and regulating body temperature (202). DMH has been reported as a key site causing WAT browning (221, 262). DMH expresses the anabolic peptide NPY (also coexpressed in ARC AgRP neurons) (263, 264), the knockdown of which induces Ucp1 expression in iBAT and ingWAT through the SNS activation and prevents diet-induced obesity (265).

In summary, both thermoregulatory and energy homeostasis pathways control the SNS outflow to iBAT and ingWAT to induce proliferation of Ucp1-expressing cells and other adipose metabolic functions (Fig. 3). The thermoregulatory system in large part consists of an LPB-POA-DMH/DHyA-RPa-IML-SNS efferent pathway that responds to thermal sensitive skin neurons and informs the POA temperature control center via the LPB. On the other hand, the energy homeostasis system involves several hypothalamic and brainstem nuclei responding to homeostatic signals informing about the energy reserve status, such as metabolic hormones (eg, leptin, insulin, ghrelin) levels. This system is best portrayed by the pathway connecting the ARC/PVH axis to BAT and WAT via the SNS outflow through BMS/MC4R neurons relaying in the brainstem and IML. Additional regulatory nerve signals from WAT and BAT lipolysis, blood flow, and/or temperature and visceral cues (hypoxia, intestinal lipids) are integrated in numerous central nervous system (CNS) areas to modulate the SNS outflow.

Figure 3.

Central nervous system (CNS) regulation of sympathetic outflow to brown and white adipose tissues. The classic thermoregulatory sympathetic nervous system afferent signal from skin thermal receptors to efferent signal to brown and white adipose tissues is displayed in black. Thermosensitive neurons connect to the spinal dorsal horn (DH) neurons that in turn connect to neurons of the lateral parabrachial nucleus (PBN). Glutamatergic neurons from this structure then connect to the preoptic area that in turn connect to the dorsomedial hypothalamus/dorsal hypothalamic area (DMH/DHyA). The sympathetic outflow signal is relayed directly to the spinal intermediolateral (IML) column neurons, or indirectly via the raphe pallidus nucleus (RPa). In the mouse, the sympathetic efferent signal to the interscapular brown fat transits through the stellate and T2 to T5 sympathetic ganglia, whereas that of inguinal white adipose depots transits through the T12 to L1 sympathetic ganglia. Afferent nerve signals from sensing of local lipolysis, blood flow, and/or temperature in white and brown adipose tissues and from visceral cues (eg, intestinal fat, arterial blood hypoxia) can be relayed to several CNS sympathetic regulatory areas (depicted in blue to red color phasing) such as the nucleus tractus solitarius (NTS), ventromedial hypothalamus (VMH), and lateral hypothalamus (LH), but also directly to primary sympathetic outflow nuclei such as the POA, DMH, and RPa. Likewise, peripheral cues of energy balance such as insulin and leptin can be detected by sympathetic regulatory areas (in blue to red color phasing) such as the arcuate nucleus (ARC), periventricular hypothalamus (PVH), LH, and VMH, but also in primary sympathetic outflow areas such as the DMH. ARC, arcuate nucleus; CNS, central nervous system; DH, dorsal horn; DMH/DHyA, dorsomedial hypothalamus/dorsal hypothalamic area; IML, intermediolateral column; LH, lateral hypothalamus; NTS, nucleus tractus solitarius; PBN, parabrachial nucleus; POA, preoptic area; RPa, raphe pallidus; VMH, ventromedial hypothalamus.

Sympathetic Noradrenergic and Purinergic Stimulation and Signaling in Brown Adipose Tissue

Noradrenergic stimulation of BAT results in cAMP–protein kinase A (PKA) activation of lipolysis and a cAMP-voltage–dependent calcium channel-mediated calcium influx, downregulated by outward potassium current (266). In addition to stimulation of intracellular TG lipolysis, PKA also stimulates p38-MAPK, which leads to PPAR-dependent stimulation of the thermogenic program of brown adipocytes (267). Sustained cAMP production leads to ADRs G_i recruitment and extracellular signal-regulated kinase (ERK) activation, a minor pathway for the stimulation of intracellular TG lipolysis in the brown adipocyte, and also leads to the activation of beta-arrestin and ADR desensitization (267). Furthermore, activation of phosphoinositide 3-kinase (PI3K) and ERK1/2 stimulates brown adipogenesis (268, 269).

The 3 types of β-adrenergic receptors (ADRB1, ADRB2, and ADRB3) are expressed in brown adipocytes, and their activation increases cAMP levels (270–272). ADRB3 is most abundant in BAT and WAT of rodents; it is however much less abundant than ADRB1 and ADRB2 in humans (273, 274). Furthermore, ADRB1 and ADRB2 display more affinity than ADRB3 for noradrenaline (273). It should be noted however that the effect of KO of any of the 3 β-adrenergic receptors on brown adipocyte thermogenesis can be compensated by the others in mice (272). Brown preadipocytes in mice express only Adrb1, with a subsequent switch to the Adrb3 phenotype of differentiated brown adipocytes (275). More recently, ADRBs have been shown to stimulate de novo lipogenesis and TG deposition in brown adipocytes through the mammalian target of rapamycin complex 2 (mTORC2)-Akt signaling pathway (276–278). mTORC1 is also required for cold-induced BAT expansion and mitochondrial biogenesis in mice (279, 280). Although browning of WAT is stimulated by the cold-induced β-adrenergic response, beige adipocyte lineages (ie, glycolytic beige fat) responsive to cold-induced alternative pathways have been identified and shown to play a role in adaptive thermogenesis (281). Most noradrenaline-induced intracellular TG lipolysis results from ADRB1/2/3 stimulation in rodents, although small residual stimulation still occurs in ADRB1/2/3 triple-KO mice (282). α-1 ADRs are also expressed in BAT, may display greater affinity for noradrenaline, and may lead to increased BAT thermogenesis through PI3K-mediated activation of protein kinase C (270, 283). α-2 ADRs are expressed in presynaptic neurons to inhibit noradrenaline release and in brown adipocytes, where they are known to inhibit thermogenesis through reduction of cAMP levels (270).

Adenosine is a purinergic cotransmitter of noradrenaline in sympathetic neurons. Adenosine receptor A_2A and, to a greater extent, A₁, are expressed and functional in adipose tissues. The former activates cAMP via Gs, whereas the latter does the opposite via Gi-coupled signaling, the net effect of adenosine resulting in the ex vivo inhibition of adipocyte lipolysis given the predominance of the A₁ receptor (274, 284). The in vivo concentration of adenosine in human WAT determined by microdialysis is considered high enough to inhibit lipolysis (285). Early studies in brown adipocytes of hamsters and rats showed considerable inhibition of lipolysis and thermogenesis by adenosine (286, 287). A_2A is, however, the predominant adenosine receptor in mouse and human BAT, leading to low-dose adenosine-mediated stimulation of BAT lipolysis and thermogenesis (288). Adenosine is produced in BAT during BAT sympathetic stimulation and loss of the A_2A receptor blunts BAT thermogenic activation (288), while adenosine administration increases BAT blood flow in vivo in humans (289). In humans, BAT A_2A receptor expression is reduced in obesity and correlates with BAT ¹⁸FDG uptake (290).

In summary (Fig. 4), BAT thermogenesis is mostly driven by the sympathetic output signal mediated by ADRB3 in mice and ADRB1/ADRB2 in humans. However, the α-ADR and adenosine receptors may also contribute to modulate the biological responses of this tissue. The in vivo control of BAT metabolic activity is the result of the interaction between the sympathetic output signal to BAT, the type and level of ADR expression, other concomitant sympathetic signaling processes, plus postsignaling modulation of the effect of this signal. The complexity and redundancy of the endogenous sympathetic regulation of thermogenic adipocytes may explain the relative inefficacy of selective β-adrenergic agonists to activate BAT thermogenesis in vivo compared to cold exposure (121) and the continuing controversy that exists with regard to the optimal pharmacological approach to stimulate BAT in vivo (see subsequent sections).

Figure 4.

Sympathetic regulation of brown adipocyte thermogenesis and triglyceride/nonesterified fatty acid (TG/NEFA) cycling. The β-adrenergic receptors (β-3 in mice, β-2, and β-1 in humans) drive the cAMP-mediated signal that stimulates intracellular triglyceride signaling to activate UCP1 and brown adipose thermogenesis. cAMP signal also leads to the activation of the p38 MAPK and mTORC1 pathways to increase brown adipogenesis and mitochondrial genesis to increase the thermogenic capacity of brown adipocytes. Sustained elevation of cAMP leads to the recruitment of Gi-mediated ERK signaling that also stimulates intracellular lipolysis, β-arrestin and β-adrenergic receptors desensitization. Alpha-1 adrenergic receptors lead to PI3K/DAC, ERK, and PKC activation, participating in brown thermogenesis. The latter also activate the AKT-mTORC2 pathway, leading to increase glucose uptake, de novo lipogenesis, and triglyceride repletion. Alpha-2 receptors on presynaptic neurons downregulate noradrenaline secretion, whereas those on the brown adipocytes reduce cAMP levels and brown thermogenesis. Adenosine, cosecreted with noradrenaline by sympathetic neurons, activates adenosine receptors to amplify the cAMP-driven thermogenic response. A_2A, adenosine receptor 2A; cAMP, cyclic adenosine 3’,5’-monophosphate; DAG, diacylglycerol; ERK, extracellular signal-regulated kinase; FA-CoA, fatty acyl-coenzyme A; mTOR, mechanistic target of rapamycin; NEFA, nonesterified fatty acids; p38 MAPK, phospho-38 mitogen-activated protein kinase; PI3K, phosphoinositide 3-kinase; PKA, protein kinase A; PKC, protein kinase C; TG, triglycerides.

Metabolic Regulation of Brown Adipose Tissue Activity and Capacity

Nucleosides and nucleotides may regulate BAT lipolytic and metabolic activity. As discussed in the preceding section, adenosine produced in situ in BAT during sympathetic nerve stimulation may contribute to the modulation of BAT lipolysis and metabolic activity through adenosine purinergic G-coupled receptors (286–288). ATP and ADP also signal through other purinergic receptors (P2Xs and P2Ys) that control lipolysis and several other metabolic processes in adipocytes (291). Recently, it has been demonstrated that FABP4, which is secreted by adipocytes during stimulation of lipolysis, forms a complex in circulation with adenosine kinase (ADK) and nucleoside diphosphate kinase (NDPK) that attenuates their activity and reduces extracellular ATP levels (292). This new hormonal complex called Fabkin was demonstrated to reduce glucose-stimulated insulin secretion in rodent and human β cells through the P2Y₁ receptor. ATP is known to activate BAT UCP1 expression and to induce WAT browning through P2 receptors (291). Whether FABP4 secreted by WAT during excessive intracellular TG lipolytic activity, as in obesity and T2D (293, 294), exerts a negative feedback on BAT metabolic activity is an intriguing possibility that deserves future studies. However, FABP4 KO mice display reduced cold-induced thermogenesis through the reduction of intracellular brown adipocyte conversion of thyroxine to triiodothyronine (T3) (111).

Lactate is the most abundant product from the very high glucose metabolic rate of brown adipocytes (83, 84) and can exert local effects through membrane receptor–mediated mechanisms and changes in cellular redox state (see [295] for review). Lactate is known to inhibit intracellular adipocyte TG lipolysis through the activation of GPR81-mediated reduction in adenylate cyclase (296, 297). Lactate can also stimulate preadipocyte TG accumulation and differentiation into white adipocytes (298) and can feed glyceroneogenesis for TG synthesis (127, 130). Therefore, lactate is a potential autocrine/paracrine metabolic break for brown adipocyte TG mobilization and may simultaneously contribute to rapid BAT TG replenishment after cold exposure. However, the in vivo contribution of glucose through glycolysis (and therefore lactate) to glycerol synthesis and TG deposition in adipose tissues was shown to be small in rats (135). Furthermore, lactate has been proposed in contrast to act as a mediator of WAT browning through stimulation of the expression of UCP1 by the increase in NADH, H+/NAD+ ratio (299), likely triggered by (67) and for the purpose of quenching (300) the increase in ROS. Lactate transmembrane transport in beige adipocytes is also important for normal glucose metabolism and may serve as an oxidative substrate for brown adipocytes (301). Consequently, lactate has been proposed as a possible mediator of stress- and exercise-induced WAT browning (302). The physiological relevance of these potentially divergent roles of lactate for the regulation of thermogenic adipose tissues remains to be established in humans.

Several other metabolites have the potential to regulate BAT metabolic activity. Succinate, produced by ischemic tissues, inflammatory cells, and the gut microbiota, inhibits cAMP-mediated coupling in adipocytes via GPR91 (303). Acetate and propionate, also in part derived from the gut microbiota, inhibit cAMP-mediated coupling via GPR43 (304), whereas β-hydroxybutyrate, produced by the liver, does the same via HM74a (GPR109A) (305). Some (306, 307) but not all studies (308) have shown that these metabolites may circulate at higher concentrations in individuals with metabolic diseases. As for lactate, however, the effect of acetate on thermogenic adipocytes is probably exerted locally, not systemically, as acetate can be produced locally by an adipocyte subpopulation that may inhibit BAT thermogenesis in a paracrine fashion (309, 310). On the other hand, β-hydroxybutyrate, as lactate, has also been proposed as a WAT browning mediator (299). Systemic administration of succinate in mice has also been shown to induce brown adipose thermogenesis via succinate dehydrogenase-mediated oxidation–derived ROS production (124).

To summarize (Fig. 5), autocrine/paracrine control of thermogenic adipocyte activity and capacity from ATP, acetate, and lactate is very likely but their physiological relevance in vivo in humans has not yet been unraveled. Possible endocrine roles for β-hydroxybutyrate and FABP4 are emergent and exciting, but need more study in humans. BAT imaging using ¹¹C-lactate or ¹¹C-β-hydroxybutyrate would be possible, although to our knowledge it has never been reported.

Figure 5.

Regulation of brown adipocyte thermogenesis by metabolites. Extracellular adenosine 5′-triphosphate (ATP) can promote brown adipocyte thermogenesis through purinergic 2 receptors (P2X, P2Y). FABP4, secreted by white adipocytes during intracellular lipolysis, can form a complex with adenosine kinase and nucleotide diphosphate kinase to reduce ATP levels and therefore limit this effect. Lactate, produced from glycolysis in brown adipocytes or during skeletal muscle exertion, can inhibit brown adipocyte thermogenesis via the activation of GPR81, but could also promote adipose browning through alteration of cellular redox state and the production of reactive oxygen species (ROS). Acetate, propionate, and succinate from the gut microbiota or ischemic or inflammatory cells, and β-hydroxybutyrate from excessive hepatic lipid oxidation, can also reduce brown adipose thermogenesis via the activation of G protein-coupled receptors. ADK, adenosine kinase; ATP, adenosine 5′-triphosphate; cAMP, cyclic adenosine 3’,5’-monophosphate; FABP4, fatty acid binding protein 4; FA-CoA, fatty acyl-coenzyme A; GPR, G protein–coupled receptor; NDPK, nucleoside diphosphate kinase; NEFA, nonesterified fatty acids; P2X, purinergic receptor 2X; P2Y, purinergic receptor 2Y; ROS, reactive oxygen species; TG, triglycerides.

Hormonal Regulation of Brown Adipose Tissue Activity and Capacity

The hormonal regulation of BAT function is a vast topic that cannot be exhaustively discussed in this review. We herein limit our discussion to the hormones known to play major roles in the regulation of energy metabolism in humans.

Insulin regulates adipose tissue development and function and is the major inhibitory control of adipose tissue intracellular TG lipolysis (293, 311, 312). BAT-specific insulin receptor + insulin-like growth factor-1 receptor double-KO mice display reduced BAT mass and cold-induced thermogenesis, with weight gain and IR (313). Acute inhibition of intracellular TG lipolysis using nicotinic acid blunts BAT thermogenesis in rats and in humans (62, 63) and the main energy source of acute cold-induced BAT thermogenesis is its own TG content (80). Accordingly, insulin, which profoundly suppresses adipose tissue intracellular TG lipolysis (293), would be expected to inhibit cold-induced thermogenesis. In support for this view, chronic insulin treatment in mice indeed leads to a reduction of ex vivo respiration and UCP1 and PGC1A expression in iBAT (314). Genetically mediated reduction in insulin levels (Ins1^+/–) in mice on an HFD also results in the increased expression of oxidative proteins without an increase in UCP1 in inguinal fat, and increased visceral and BAT UCP1 expression associated with increased energy expenditure in this model (315). However, insulin signaling does not acutely and directly regulate adipocyte respiration in vitro (89, 316). Again, in mice, insulin release stimulated by WAT lipolysis (317) has been shown to drive BAT glucose, NEFA, and TRL-TG uptake on cold or β-3 adrenergic stimulation (318). In the latter study, BAT insulin signaling stimulated BAT energy substrate uptake on acute, but not chronic, BAT activation (318). In humans, acute IV insulin administration increases BAT glucose uptake, but does not change BAT blood flow, suggesting no change in BAT thermogenesis (175, 319). Furthermore, insulin secretion is not changed on acute cold stimulation in humans (320). BAT oxygen consumption increases postprandially, at a time when insulin level peaks, but this is more likely the result of postprandial sympathetic activation (321). Recent work has shown that human subcutaneous adipocyte populations display a differential response to insulin in vivo (322), but whether this heterogeneity of insulin response occurs in BAT in vivo is unknown. In summary, insulin likely regulates BAT glucose uptake and fuel selection, but unlikely directly affects BAT thermogenesis. To our knowledge, no study has thus far reported the effect of insulin on BAT thermogenesis or lipid metabolism in vivo in humans during cold-induced BAT activation. Any effect of insulin on BAT energy metabolism in humans would likely be dwarfed by the fact that BAT accounts for only less than 1% of whole-body NEFA and lipoprotein-derived fatty acid metabolism, even after prolonged cold acclimation that increases BAT oxidative metabolism (80, 101, 104).

Glucagon administration increases whole-body oxygen consumption in rodents (323) and humans (324–326). Ex vivo, glucagon has long been demonstrated to activate BAT lipolysis and thermogenesis in rats and mice (327–329). However, mice lacking the glucagon receptor in BAT displayed no change in thermogenic or metabolic responses to glucagon administration in vivo (323). Furthermore, glucagon does not activate adipose tissue lipolysis in vivo in humans at physiological or low supraphysiological concentrations (330, 331). Finally, the in vivo thermogenic response to glucagon administration is not accompanied by increased BAT ¹⁸FDG uptake in humans (326). However, glucagon’s effect on BAT thermogenesis per se has never been determined. Thus, although glucagon’s thermogenic effect in animals and humans is unlikely mediated by BAT, more studies are needed to fully rule out this possibility.

The glucagon-like peptide 1 (GLP1) receptor is not expressed functionally in WAT or BAT (332) (see also (333) for a detailed review). GLP1 stimulates BAT and WAT browning via CNS sympathetic activation in rodents (332, 334, 335), but does not appear necessary for cold-induced BAT thermogenesis (336). Treatment with GLP1 agonists in humans induces weight loss without increasing energy expenditure (337–340). Although one study found a substantial increase in cold-induced BAT glucose uptake after 12-week treatment with the GLP1 receptor agonist exenatide in healthy individuals, this was not associated with a change in energy expenditure (341). This effect on BAT glucose uptake may have been mediated by improved BAT insulin sensitivity (IS; see discussion in subsequent sections). Thus, stimulation of BAT or WAT thermogenesis does not appear implicated in the antiobesity effects of GLP1 receptor agonists.

Glucose-dependent insulinotropic polypeptide (GIP) is another incretin hormone that is currently attracting considerable attention with the recent demonstration of greater weight loss and improvement of glucose control with dual GIP/GLP1 vs GLP1 receptor agonist treatment (342, 343). Intriguingly, both stimulation and inhibition of GIP receptor signaling drive weight loss and metabolic benefits (344). GIP receptor KO mice display resistance to high-fat feeding–induced obesity with both increased WAT fat storage and intracellular lipolysis (ie, increased WAT TG/NEFA cycling) (345). In humans, GIP administration stimulates postprandial WAT blood flow, intracellular fatty acid reesterification rate, uptake of fatty acids, and whole-body glucose utilization (346–349). Further evidence supporting the effect of GIP on adipose tissue metabolic function comes from post hoc analyses of clinical trials of dual GIP/GLP1 receptor agonist treatment revealing reduction in TG levels and liver inflammation and increase in adiponectin levels (350, 351). The GIP receptor is expressed in the stromal and adipocyte fractions of WAT and BAT (352–354), although it is mostly expressed in pericytes and mesothelial cells in WAT (355). GIP stimulates human omental adipose–derived mesenchymal stem cell beige adipogenesis in vitro (356). Interestingly, both reduced and enhanced GIP signaling lead to increased thermogenic gene expression in mouse BAT adipocytes (353). In the latter study, Myf5-driven Cre expression was used to generate Gipr^BAT–/– mice, found to display reduced oral lipid tolerance, but normal weight on an HFD and reduced BAT mass and β-3 adrenergic agonist–induced energy expenditure under chronic cold exposure condition. In a recent large cross-sectional human study, GIP levels were associated with the preferential selection of glucose as an energy substrate during fasting (357). However, acute GIP infusion does not increase energy expenditure in humans (349, 358) and postprandial energy expenditure was inversely correlated with GIP levels in a study in Japanese men (359). To our knowledge, the effect of GIP on BAT metabolic function has not been assessed in vivo in humans.

The hypothalamic-pituitary-adrenocortical axis is a very important physiological regulator of adipose tissue development and function (360, 361). The chronic activation of this axis leads to Cushing syndrome, which is characterized by the development of obesity, IR, and a high risk of developing T2D. Brown adipocytes express glucocorticoid receptors (362). In vitro, glucocorticoids, which are essential to inducing adipocyte differentiation, were found to increase human brown adipocyte thermogenesis, whereas glucocorticoids decreased murine brown adipocyte thermogenesis (363). Glucocorticoids are important for the maintenance of BAT TG and glycogen content (364). Chronic corticosterone treatment elicits BAT whitening and inhibits BAT metabolic activation in rodents (365–367), whereas treatment with a selective glucocorticoid receptor antagonist has the opposite effects (368). Whether these effects on BAT thermogenesis contribute to glucocorticoid-induced obesity is unclear, however, as UCP1 KO mice do not show a different glucocorticoid-induced obesity phenotype (369). BAT glucocorticoid receptor KO mice display normal BAT development and cold adaptation and thermogenesis, thereby casting doubt on the importance of a direct role for corticosterone or cortisol in BAT development and function (370). Nevertheless, BAT endogenous sympathetic stimulation and TG-derived fatty acid uptake display a circadian rhythm that is canceled by the flattening of corticosterone secretion through constant exogenous corticosterone delivery in mice (371). However, acute administration of prednisolone to healthy individuals increases cold-induced BAT ¹⁸FDG uptake (363) and supraclavicular temperature (363, 372). Remarkably, 1-week prednisolone treatment (15 mg/day) significantly reduced cold-induced BAT ¹⁸FDG uptake in a randomized, crossover, controlled study performed in healthy participants (178).

Although adrenocorticotropin (ACTH), through action via the melanocortin 2 receptor, stimulates adipocyte lipolysis in mice, it does not in humans given the interspecies difference in melanocortin receptor expression (373). Therefore, the ACTH-mediated increase in brown adipocyte thermogenesis and in vivo ¹⁸FDG uptake in BAT in mice (365) is of unclear physiological significance in humans.

The mineralocorticoid receptor is also expressed in brown adipocytes in which, as for the glucocorticoid receptor, its activation leads to the inhibition of the thermogenic phenotype of these cells (374, 375). Treatment of mice with spironolactone, a mineralocorticoid receptor antagonist, leads to WAT browning and prevention of HFD-induced weight gain and dysmetabolism (376). A 2-week randomized, crossover study of spironolactone (100 mg/day) vs placebo in healthy individuals demonstrated increased ¹⁸FDG BAT uptake in the former condition (377), supporting a substantial inhibitory effect of the mineralocorticoid receptor on BAT metabolic function in vivo.

The hypothalamus-pituitary-thyroid axis is another endocrine system that regulates energy metabolism in humans. It has long been demonstrated that thyroid hormones stimulate BAT TG accumulation and WAT TG mobilization in rodents (364). Hypothyroidism, however, also increases BAT weight, mitochondrial content, and ex vivo respiratory rates in rats, but blunts its TG mobilization on acute cold exposure (378). Thyroid hormones upregulate adrenergic receptor signaling through adenylate cyclase activity (379–381) so that BAT biological responses to noradrenaline and cold-induced nonshivering thermogenesis are severely blunted in the hypothyroid state in rodents (364). Indeed, noradrenaline-stimulated brown adipocyte expression of UCP1 requires the presence of T3 (382, 383). Increased circulating T3 also participates in the increase of BAT thermogenesis during a high-calorie diet, cold acclimation, or chronic noradrenaline administration in rodents (384–388). Conversion of thyroxine (T4) to T3 in situ in brown adipocytes by thyroxine 5’-deiodinase is also increased during direct adrenergic stimulation, cold exposure, and after dietary intake (389–391), is the prominent source of T3 receptor occupancy (392), and participates in the cold- or noradrenaline-induced stimulation of lipolysis and UCP1 expression and activity in BAT (393, 394). This evidence shows that the endogenous thyroid axis participates in a feed-forward stimulation of BAT thermogenesis during cold exposure and high-calorie–induced adrenergic activation. However, BAT thermogenesis paradoxically decreases in hyperthyroid animals during chronic cold acclimation because thyroid hormone–stimulated whole-body thermogenesis is driven more importantly by other organs (395), and because local BAT thyroxine 5’-deiodinase, the primary source of BAT T3, is downregulated by high circulating thyroid hormones (396). Furthermore, hypothyroidism in rodents is well known to increase BAT sympathetic nerve outflow signal because these animals are more cold sensitive (397), leading to a paradoxical increase in BAT mass, lipogenesis, thyroxine 5’-deiodinase activity, and adenylate cyclase expression (398, 399). Hyperthyroidism also leads to inhibition of AMPK in the hypothalamic ventromedial nucleus that activates the sympathetic output signal to activate BAT thermogenesis and browning of WAT (400–402). Thyrotropin (TSH) is also reported to regulate brown adipocyte browning, offering another possible mechanism by which severe primary hypothyroidism may also affect energy metabolism in vivo (403, 404).

In humans, acute cold exposure slightly increases serum free T4 and reduces serum TSH levels (63, 103, 405–407), an effect also observed after acute administration of mirabegron, a β-3 adrenergic receptor agonist (121). However, these acute changes in circulating T4 and TSH levels are of unclear clinical significance. Higher serum T3 levels were found in cold-induced ¹⁸FDG BAT–positive individuals in one study (406) and with lower BAT fat fraction on MRI in another (408). A positive association between serum free T4 levels and cold-induced BAT glucose uptake was found in 24 euthyroid and healthy participants (409). However, another large cross-sectional study did not find any association between serum thyroid hormone levels and cold-induced BAT glucose uptake in 106 healthy individuals (407). Hyperthyroidism was associated with increased cold-induced BAT ¹⁸FDG uptake, which was reversed on normalization of thyroid status; BAT perfusion was however not changed in this study (410), suggesting no increase in BAT thermogenesis. In another study, BAT ¹⁸FDG uptake at room temperature was not detectable in 9 patients with hyperthyroidism (411). A third study also showed a significant increase in BAT ¹⁸FDG uptake in the hyperthyroid vs normothyroid state in 19 patients, but circulating T3 levels did not correlate with BAT glucose metabolism; hyperthyroidism was more closely associated with increased skeletal muscle glucose uptake in that study (412). A fourth prospective study did not find increased BAT ¹⁸FDG uptake, but did find lower BAT fat fraction in the hyperthyroid vs euthyroid state of patients treated for Graves disease (413). In a large retrospective study in 124 patients who underwent total thyroidectomy for thyroid cancer, 6 individuals had positive BAT ¹⁸FDG uptake, leading the authors to suggest that hypothyroidism activates BAT metabolism (414). However, this rate of spontaneously positive BAT ¹⁸FDG uptake is similar to that observed in cancer patient populations undergoing PET scanning (157). A prospective study in 10 patients undergoing total thyroidectomy for thyroid cancer showed higher cold-induced BAT ¹⁸FDG uptake after treatment with levothyroxine to induce mild hyperthyroidism vs when the participants were hypothyroid after surgery (415). Similar findings were found in 4 out of 6 patients in similar conditions in another study (416). However, in a large retrospective analysis of 4852 patients who had an ¹⁸FDG PET scan, TSH levels were higher in BAT+ individuals (417). Cold-induced BAT glucose uptake was also increased after IV administration of TSH-releasing hormone in another study (418). In summary, thyroid hormones participate in the SNS stimulatory effect on BAT recruitment directly in BAT and indirectly in the CNS. In humans, hyperthyroidism generally leads to increased cold-induced BAT glucose uptake. In addition to their effects on BAT, thyroid hormones however exert profound effects on tissue sympathetic outflow, liver glucose production, IR, whole-body thermogenesis and cold tolerance, and feeding behavior (400, 419, 420), potentially confounding the human investigation results based largely on measurement of BAT glucose metabolism alone (see subsequent sections). The role of BAT thermogenesis in the well-described dysregulation of whole-body thermogenesis of patients with hypothyroidism or hyperthyroidism has not been specifically investigated.

Sex hormones also influence adipose tissue and energy metabolism in humans. Sex hormones modulate regional adipose tissue fatty acid storage (421, 422) and also BAT function (see (361) for a detailed review). The reduction of circulating estradiol levels with menopause or the administration of gonadotrophin-releasing hormone agonists is associated with the development of central obesity and dysmetabolism associated with decreased energy expenditure (423–426). Rodents reduce their energy expenditure, gain fat mass, and lose BAT function with estrogen signaling disruption or after ovariectomy, which is prevented by estrogen replacement (427–433). Murine (434, 435) and human (436) brown adipocytes express estrogen receptor α, which stimulates mitochondrial biogenesis and oxidative capacity of these cells (437, 438). Estrogen also stimulates BAT function through a CNS-mediated action as intracerebral administration of estrogen increases BAT thermogenesis in mice (439). Furthermore, mice lacking estrogen receptor α in steroidogenic factor-1 expressing hypothalamic neurons display lower BAT UCP1 levels and larger lipid droplets (440). Follicle-stimulating hormone, which is elevated with estrogen deficiency, has been shown to downregulate BAT function in vivo in mice (441), suggesting another mechanism for estrogen deficiency–mediated BAT dysfunction in rodents. There are currently no published data on the effect of estrogen or estrogen deficiency on in vivo BAT function in humans. One study using infrared thermography showed some correlation between circulating estradiol levels and cold- and meal-induced increase in supraclavicular skin temperature in healthy men and women (442), but this observation is inconclusive given the cross-sectional design and technical limitations of this method to study BAT function (discussed earlier). The group of Ed Melanson at the University of Colorado is currently investigating the effect of estrogen deficiency in premenopausal and postmenopausal women on in vivo BAT thermogenesis, using the ¹¹C-acetate PET method, in collaboration with our group (Clinicaltrials.gov No. NCT02927392).

In rodents, pregnancy is associated with BAT whitening and reduced function that has been attributed at least in part to high circulating progesterone levels (443). In vitro treatment of brown adipocytes with progesterone has resulted in varying outcomes including stimulation (438, 444, 445), inhibition (438, 443, 446), or no change (445) in brown adipocyte thermogenic activity. These divergent responses are at least partially explained by a heterogeneous concentration-dependent response, with stimulation at low, but not at high, progesterone concentrations (445). However, replacement of progesterone in ovariectomized rats does not change BAT adaptation or energy balance during cold exposure (447). It is possible that progesterone effects on BAT function may be mediated by its binding to the glucocorticoid receptor in brown adipocytes or elsewhere throughout the hypothalamic-pituitary-adrenal axis (see earlier discussion) (361). Correlation between circulating progesterone levels and basal supraclavicular temperature using infrared thermography was found in healthy men and women (442), but the physiological relevance of this observation is unclear.

In contrast to estrogen, brown adipocyte treatment with testosterone reduces UCP1 and estrogen receptor α expression, mitochondrial biogenesis, and intracellular TG lipolysis (434, 438, 444, 445). However, in vivo treatment with testosterone is associated with a loss of fat mass in rodents, associated with increased mitochondrial biogenesis in muscle, but not BAT (448), and does not alter cold-induced BAT cellular composition and whole-body cold adaptation (449). Interestingly, testosterone may stimulate the in situ conversion of cortisone to cortisol in adipose tissues, contributing to the glucocorticoid-induced BAT whitening phenotype (see (361) for a detailed discussion). One study suggested lower BAT thermogenesis in women with hyperandrogenism (polycystic ovary syndrome) based on lower supraclavicular temperature using infrared thermography (450). Studies using more specific methods are needed to ascertain the effect of testosterone on in vivo BAT function in humans.

In rodents, leptin deficiency is associated with hypothermia and reduced energy expenditure that is corrected on leptin replacement (451). Direct effects of leptin to increase brown adipocyte glucose uptake and lipolysis have been reported (452), but leptin’s effect on thermogenesis has been primarily attributed to its signaling in the hypothalamus, increasing SNS outflow and stimulating BAT UCP1 expression (204, 232, 453–457). Leptin signaling through an ARC/PVH-brain–derived neurotrophic factor neuronal circuit to enhance the sympathetic outflow toward BAT and WAT to regulate energy homeostasis has been recently elegantly demonstrated (458). However, although leptin regulation of SNS outflow remains accepted, the initial interpretation that reduced energy expenditure contributes to the obese phenotype of leptin deficiency in mice has recently been questioned on the absence of evidence for BAT atrophy with leptin deficiency on a stable C57BL/6 genetic background (459). In humans, leptin deficiency is not associated with hypothermia (460) and leptin administration increased energy expenditure in some (461, 462) but not in all studies (463–466). Leptin appears to blunt the reduction of energy expenditure associated with weight loss in leptin-deficient (467) but not in obese individuals not deficient in leptin (468). The net effect of leptin on whole-body energy expenditure thus likely depends on the endogenous leptin secretion and resistance status, on chronic caloric balance, and on the response of thyroid and gonadal hormones. Leptin levels generally go down during acute cold exposure in humans (469, 470) and have been associated with reduced BAT ¹⁸FDG uptake in some (470, 471) but not all studies (469, 472). To our knowledge, no study has reported the effect of leptin treatment on BAT metabolism in humans.

Fibroblast growth factor 21 (FGF21) was very early evoked as a possible BAT-activating hormone. FGF21 is produced by the liver on activation of PPAR-α during fasting (473), but also after glucose (474), fructose (475), alcohol intake (476), or cold exposure (477). FGF21 can activate brown and beige adipocyte thermogenesis in rodents (478, 479) and activates the ERK pathway to stimulate fatty acid and glucose uptake in BAT and WAT depots (480). Disruption of hepatic insulin signaling leads to reduced FGF21 production owing to activation of hepatic FoxO1, leading to impaired BAT gene expression, BAT glucose uptake, and whole-body thermogenesis in mice (481). In humans, the circulating levels of FGF21 correlate with BAT ¹⁸FDG uptake during cold exposure in some (472, 482) but not all studies (469). However, the metabolic improvements conferred by FGF21 do not appear to require the presence of UCP1 in mice (483, 484). There is actually evidence for a WAT-liver-BAT activation pathway in mice: Liver fatty acid oxidation and FGF21 secretion appear to both be dependent on WAT ATGL activation during fasting and β-3 adrenergic receptor agonist administration via hepatic PPAR-α activation; the absence of hepatic PPAR-α partially prevents upregulation of BAT UCP1 and elongation of very long-chain fatty acid protein 3 (ELOVL3) expression on β-3 adrenergic receptor agonist administration and is associated with reduced cold tolerance https://doi.org/10.1016/j.celrep.2022.110910). Administration of the long-acting FGF21 agonist PF-05231023 exerts pharmacological reduction of circulating TGs, without improvement in blood glucose control and with modest and variable reduction in body weight in patients without and with T2D (485, 486). This TG lowering in mice appears to result from reduced NEFA mobilization from WAT, leading to reduced VLDL-TG production by the liver, and from increased LPL-mediated catabolism of TRL in WAT and BAT (487). Although FGF21 exerts beneficial effects on WAT and BAT lipid metabolism leading to TG lowering, there is currently little evidence that FGF21 confers metabolic benefits through specific activation of BAT thermogenesis in animal models or in humans.

Other gastrointestinal hormones have been shown to influence BAT metabolism in vivo. For example, a cholecystokinin-mediated afferent vagal signal activates SNS-mediated BAT thermogenesis in rats (488). Adipose tissues express the secretin receptor that activates the canonical cAMP-PKA pathway, activating intracellular lipolysis (274). Secretin can directly activate BAT lipolysis and metabolism in mice and human brown adipocytes, which in turn may increase satiation via neural BAT afferent-brain signal (489). Postprandial increase in secretin level is associated with postprandial BAT oxidative metabolism as determined by the ¹⁵O-O₂ PET method, and secretin infusion raises BAT ¹⁸FDG uptake, but not blood flow in humans (489, 490).

Finally, myokines may stimulate BAT metabolic function. IL-6 is an immunomodulatory, insulin sensitizing and WAT lipolysis-promoting cytokine produced by the immune cells and muscle in response to exercise (491–493). IL-6 has been shown to play a role in exercise and cold-induced WAT browning in mice (494). However, physical activity is not associated with subcutaneous WAT UCP1 expression (495) and endurance athletes do not display increased BAT metabolic activity (496).

In summary, the in vivo effect on BAT thermogenesis has not been characterized in humans for almost all hormones that otherwise display significant effects on whole-body energy metabolism or even BAT glucose uptake (Fig. 6). Insulin and secretin were shown, using BAT blood flow as a surrogate, not to affect BAT thermogenesis while stimulating glucose uptake. Several studies have shown increases in BAT glucose uptake with thyroid hormone administration, but the multiple systemic effects of these hormones, including on IS and thermogenesis in other organs, confound any conclusion about BAT thermogenesis. There is evidence that glucocorticoids and mineralocorticoids reduce BAT glucose metabolism in humans, but whether this is secondary to reduced BAT IS or whether BAT thermogenesis is truly reduced is unknown at the moment. Studies are currently being performed to document the effect of estrogen deficiency on BAT thermogenesis in women.

Figure 6.

Hormonal regulation of brown adipose tissue thermogenesis (T°) and metabolism in vivo. AT, adipose tissue; BAT, brown adipose tissue; EE, energy expenditure; FGF21, fibroblast growth factor 21; FSH, follicle-stimulating hormone; GIP, glucose-dependent insulinotropic polypeptide; GLP1, glucagon-like peptide 1; IS, insulin sensitivity; SNS, sympathetic nervous system; T°, thermogenesis; T3, 3,5,3′-triiodothyronine; TEE, total energy expenditure (whole body); TG, triglycerides; TSH, thyrotropin.

Physiological Roles of Brown Adipose Tissue

Brown Adipose Tissue and Cold-induced Thermogenic Responses

Our functional definition of BAT refers to adipose tissue displaying significant thermogenesis in vivo, given its primary function as a heat-generating thermoregulatory organ on cold exposure. The lower critical temperature (ie, the lower bound of the “thermoneutral zone”) is approximately 23 °C in humans and declines from 33 to 25°C according to higher body weight in mice (497). In the mouse, the thermoneutral point (ie, the temperature below which energy expenditure increases and above which body temperature increases) also displays a circadian variation between 29 and 33°C (498). In rodents and humans, BAT is highly responsive to acute, chronic, and repeated cold exposure (62, 102, 104, 158, 499–501). In mice originally housed at thermoneutrality, at which there is little demand for thermoregulatory heat production, total protein content and Ucp1 messenger RNA (mRNA) in BAT increases rapidly over the first day of exposure to 4 °C (502). However, Ucp1 mRNA rapidly plateaus whereas total UCP1 protein content only begins to increase over the ensuing days, increasing more than 5-fold after 3 weeks of being housed at 4 °C (502). This progressive increase in UCP1 protein content parallels the increased capacity for nonshivering thermogenesis (NST) in cold-acclimated gerbils (503), mice (503), and rats (504, 505). The increased capacity for NST is demonstrated by the significantly greater thermogenic response to noradrenaline injection (506) and the lower ambient temperature needed for eliciting an increase in muscle electromyography (shivering and voluntary contractions) relative to the cold-induced rise in thermogenesis (lower critical temperature) (503). Whereas metabolic rate and muscle electromyography increase concomitantly with decreasing ambient temperatures in warm-acclimated mice, the increase in muscle activity occurs at colder ambient temperatures than the cold-stimulated increase in thermogenesis in cold-acclimated mice. Using radioactive microspheres combined with arteriovenous oxygen gradient to measure cardiac output and tissue blood flow distribution across various tissues, Foster and Frydman (507) estimated that BAT thermogenesis accounts for 60% of the thermogenic response to noradrenaline injection in cold-acclimated rats, similar to what has been estimated in mice living at 23 °C (508). However, under cold stimulation, using the same techniques, BAT was estimated to account for 29% to 37% of cold-induced thermogenesis in warm-acclimated rats and for 61% to 72% in rats living at 6 °C for 4 weeks (505). The contribution of skeletal muscles to thermogenesis was estimated to be limited to 8% to 10% of cold-induced thermogenesis in the cold-acclimated rats. This is consistent with earlier work demonstrating that muscle electromyographic activity (shivering and voluntary movement) decreases progressively to baseline levels seen at 30 °C in unanesthetized and unrestrained rats after 29 days at 6 °C (504). It is important to note that the acclimation temperature is an important determinant influencing the thermogenic capacity of BAT, as progressively lower temperatures elicit greater increases in BAT mass, total protein content, cytochrome oxidase activity (a marker of mitochondrial content), total UCP1 protein content, and GDP binding (a marker of UCP1 activity) (509). Thus, increased BAT thermogenic capacity substitutes for shivering during cold-induced thermogenesis; however, once acclimated to cold, acute exposure to colder temperatures nevertheless results in the resumption of a shivering-dominant thermogenic phenotype (504).

Studies including models of surgical or inducible BAT ablation or denervation as well as UCP1 deletion (UCP1-KO mice) have been invaluable for characterizing the relative importance of BAT and UCP1 in the cold-induced thermogenic response. Surgically denervating 1 of the 2 lobes that form the iBAT depot of adult rats results in TG accumulation (whitening of depot) and suppression of adipocyte proliferation in the denervated lobe compared to the contralateral innervated lobe that can be restored with norepinephrine infusion (510). Surgical denervation also inhibits the cold-induced increase in mitochondrial content, UCP1 content and capacity to dissipate proton conductance (511–513), and BAT glucose uptake (514). These findings demonstrate that intact sympathetic innervation is essential for the cold-induced proliferation and differentiation of brown adipocytes and to increasing its thermogenic and energy substrate utilization capacity. Despite these changes in BAT recruitment and thermogenic capacity resulting from iBAT denervation, these animals do not however present any apparent cold intolerance or diminished whole-body thermogenesis, likely because of thermogenic compensatory responses in other organs. Indeed, a characteristic feature resulting from the bilateral denervation or ablation of iBAT in cold-exposed rodents is the compensatory recruitment of UCP1-dependent and -independent mechanisms of heat production in other adipose tissue depots and other organs. In rodents chronically exposed to cold, both BAT ablation and denervation result in the ectopic recruitment of Ucp1-expressing cells in WAT depots, including inguinal, anterior subcutaneous, and suprascapular WAT (225, 515). The increase in Ucp1-expressing cells in ingWAT, the most inducible WAT depot, however, does not lead to an increase in ingWAT thermogenesis, glucose uptake, or fatty acid uptake in vivo (147, 515). Interestingly, 7-day cold exposure in rats leads to increased fatty acid oxidation and thermogenesis in BAT, but rather reduced fatty acid oxidation and increased lipolysis, glyceroneogenesis, and TG synthesis in ingWAT (516). These studies suggest that ingWAT browning is associated with increased TG/NEFA cycling, but may not directly drive the cold-induced thermogenesis compensatory response for the loss of BAT in rodents.

Mice lacking UCP1 are also able to acclimate to cold, provided that the ambient temperature is lowered progressively (517, 518). The absence of UCP1 in mice results in the recruitment of alternative UCP1-independent thermogenic mechanisms, such as increased calcium cycling in beige or white adipocytes (70, 519), substrate cycling in white adipocytes (70), or skeletal muscle calcium cycling (520). The absence of UCP1 also results in sustained elevation in electromyographic activity, which some have compared to muscle training (69), suggesting partial compensation from shivering muscles. This adaptive capacity is modulated by the genetic background of the mice; for example, UCP1-deficient mice on a 129/SvImJ background display a significantly greater acute sensitivity to the cold (4 °C) than mice on a C57/BL6J background, whereas UCP1-deficient mice on a mixed 129/SvImJ and C57/BL6J background display a similar sensitivity to wild-type 129/SvImJ and C57/BL6J mice (70, 517). When exposed to progressively colder temperatures between 28 and 4 °C, all these models, regardless of genetic background, are able to increase their heat production in proportion to the cold stress. The mitochondria of iBAT in these UCP1-null mice also display significant alterations in mitochondrial morphology, considerable reduction in the protein abundance of electron transport chain subunits, greater sensitivity to ROS-induced dysfunction and activation of the immune response (521). Very recently, a novel UCP1 cysteine-253-null mouse was generated to induce an inactive conformation of UCP1 (522). UCP1 cysteine-253-null mice display lower whole body thermogenesis at 4 °C, but not at 28 or 22 °C compared to wild-type mice. This new model is free of the profound respiratory chain protein defects of the UCP1-null mice and of any accumulation of excess fat mass on an HFD. Furthermore, only males display glucose intolerance and liver and systemic inflammation with an HFD, which are corrected on estrogen administration (522).

In humans, our understanding of the role of BAT in the cold-induced thermogenic response remains relatively limited. Like other endotherms, humans exposed to the cold rely on somatic and autonomic thermoregulatory processes, such as shivering and BAT thermogenesis to counteract the heat lost to the environment. However, in contrast to rodents, their large skeletal muscle mass serves as both a large thermal reservoir that can affect heat loss to the environment (523) while also serving as a critical source of heat production. For example, even during mild cold exposure (decrease of skin temperature by 5 °C without change in core temperature), heat is not generated entirely by BAT activation but also by muscle tension or shivering of deep and superficial muscles located in the trunk (405). As cold exposure intensifies and induces a maximal reported thermogenic response of approximately 5 times the resting metabolic rate (524), it is likely that heat produced by shivering dominates over that of BAT. Accurately quantifying the heat generated by BAT and its contribution to total body heat production is particularly challenging. Early estimates were obtained by (1) quantifying changes in the body cooling rate, shivering threshold, and/or energy expenditure after pharmacologically blocking shivering (using meperidine, buspirone, magnesium sulfate, or dexmedetomidine) (525–534); (2) quantifying the increase in energy expenditure following the administration of noradrenaline or ephedrine (535, 536); and (3) quantifying the change in tissue blood flow using the ¹³³Xe-clearance method (536). Unfortunately, these various methods do not provide direct measures of BAT metabolism or heat production. More recently using, ¹⁵O-O₂ and ¹⁵O-H₂O with PET/CT to estimate tissue oxygen consumption, BAT thermogenesis could be assessed directly in vivo (180, 181, 321, 537). During mild cold exposure (< 2-fold increase in whole-body energy expenditure), oxygen consumption in BAT was shown to increase by approximately 40%, from 0.7 to 1.0 at baseline to 1.2 to 1.4 mL of O₂·100 g of tissue⁻¹ min⁻¹ in the cold (180, 181, 321). From these measurements, Richard et al (538) estimated that 50 to 150 g of BAT can account for approximately 0.6% of total body oxygen consumption in cold-exposed women and men. While these estimates suggest that the role of BAT is negligible in cold-exposed humans, these measurements were performed only under very mild cold conditions and for a short duration. Consequently, maximal BAT oxidative metabolism was certainly not achieved. In another study, intracellular lipolysis was suppressed using nicotinic acid to inhibit BAT thermogenesis during mild cold exposure (63). Remarkably, this resulted in an approximately 75% increase in shivering intensity that compensated for the loss of BAT thermogenic activity, as core temperature and whole-body cold-induced thermogenic responses were the same both in control and nicotinic acid treatment conditions.

While the role of BAT in cold-induced thermogenesis in rodents has been well characterized through the use of surgical ablations, denervation, or genetic deletion of UCP1, such approaches are not possible in humans. However, some insight could be gained by interrogating cold-stimulated BAT function in individuals in whom BAT is potentially absent through genetic lipodystrophies, or in individuals with UCP1 gene variants (539, 540). To our knowledge, only a single study has examined cold-stimulated BAT metabolism in individuals presenting with possible genetic BAT dysfunction. In individuals with familial partial lipodystrophy, brown adipocytes were shown to be larger in size with larger lipid droplets and low levels of UCP1 and were associated with lower levels of BAT ¹⁸FDG uptake in response to intermittent immersion of the feet in cold water (541).

We have demonstrated unequivocally that BAT volume determined from ¹⁸FDG uptake and its thermogenic activity from ¹¹C-acetate PET increase in response to repeated intermittent exposure to the cold (102). In some individuals, this increase in ¹⁸FDG BAT volume can be limited to the supraclavicular and axillary depots, whereas in others it can be distributed across several depots that were previously not recruited (102, 501). Given its much smaller relative mass than in rodents, it is unlikely that BAT in humans can serve the same preponderant thermoregulatory function. However, its anatomical distribution and its recruitment pattern in response to repeated cold exposure may shed some light on its actual thermogenic function in humans and provide an explanation as to how it might influence shivering thermogenesis in some individuals but not in others. For instance, one particular depot that can be recruited, especially in individuals with relatively high amounts of BAT, are the paravertebral depots that lie close to the spinal cord and paraspinal sympathetic ganglia. These neural structures harbor both cold- and warm-sensitive neurons, in proportions of as much as 17:1, respectively (542, 543). In the 1960s, Kurt Brück and Wolf Wünnenberg published a series of remarkable papers that demonstrated the effect of locally heating the cervical or thoracic spinal cord of newborn and cold-acclimatized guinea pigs (544). The authors found that locally applying heat to these neural structures could rapidly suppress shivering and posited that iBAT and cervical BAT thermogenesis could provide the necessary heat to reduce shivering. Similar experiments were performed in dogs, which also demonstrated suppression of muscle shivering with local heating of the spinal cord (545). Whether this is also the case in humans remains unclear, but could serve as a plausible mechanism by which the presence of large volumes of BAT, particularly at the base of the neck and in the paravertebral depots, could suppress shivering. Similarly, it could be argued that supraclavicular, axillary, and cervical BAT may provide the necessary heat to supply warm blood to the brain, given their proximity to large blood vessels. The perirenal adipose depot, which demonstrates remarkable plasticity, heterogeneity, and unique browning characteristics (546), may also support thermogenesis to protect adrenal, renal, and liver function under extreme cold stresses.

Because of the large difference in the critical temperature for cold-induced thermogenesis between rodents and humans, attention has been paid to housing temperature (498, 547–550) in an attempt to “humanize” mouse models by ensuring comparable chronic cold exposure conditions (149). Housing at subthermoneutral vs thermoneutral temperatures has demonstrated profound differential effects on experimental outcomes in rodents (551), leading to the recommendation by many to conduct rodent experiments at thermoneutrality to ensure thermal conditions usually experienced by humans. For instance, mice living at thermoneutrality on an HFD (45% of calories from fat) for 25 weeks develop iBAT that is morphologically comparable and share a similar molecular signature to that of human supraclavicular BAT (148). The different anatomical BAT depots in humans nevertheless display variable molecular signatures and varying overlap with that of iBAT in rodents (552, 553). The vascularization of rodents vs human BAT is also strikingly different. The iBAT depot in mice connects to the inner vertebral venous plexus and to the large Sulzer vein that can then supply warm blood to the trunk (554). An analogous venous drainage is present in the iBAT found in infants (2, 555), but not in the supraclavicular BAT depots. Finally, humans have a body mass that is 3 orders of magnitude greater than mice (~75 000 g in humans vs ~25 g in mice) and a body surface area to mass ratio that is 1 order of magnitude smaller (eg, ~178 cm²: 75 kg = 2.4 cm²/kg vs 0.7 cm²: 0.025 kg = 28 cm²/kg, in humans vs mice respectively). Humans thus have a much greater capacity to produce heat, a much larger thermal mass, and a much lower propensity to lose heat to the environment. These attributes make humans better equipped to defend their core body temperature in cold environments compared to mice.

Five prospective intervention studies have examined the effects of daily cold exposure under controlled laboratory conditions (acclimation) on BAT volume (using ¹⁸FDG PET) in humans (102, 158, 160, 501, 556). Only 2 studies also reported BAT thermogenesis (using ¹¹C-acetate PET) under the same conditions (102, 104) in humans. Whether exposed to 14 to 16 °C air, 6 hours per day for 10 days (158, 160, 556), 19 °C air, 10 hours per night for 1 month (501), or 10 °C cold water perfused through a liquid-conditioned garment, 2 hours per day for 4 weeks (102) the volume of adipose tissue taking up ¹⁸FDG increased by only 40% to 45% and the rate of ¹⁸FDG uptake increased by 10% to 31%. BAT oxidative metabolism and blood flow, however, increased 2.7- and 1.3-fold, respectively (102, 104). Interestingly, this increase in thermogenic capacity did not result in lower shivering intensity (102, 104). This surprising result may be explained by decreased skeletal muscle proton leak with repeated cold exposure, potentially reducing muscle NST (500). This suggests that the contribution of nonshivering thermogenesis had shifted from skeletal muscles toward BAT following cold acclimation. Cross-sectional studies that have examined the effects of cold acclimatization (seasonal acclimatization or cold-water swimming) also showed significant differences in BAT volume/mass (determined by necropsy or with ¹⁸FDG PET) in response to repeated exposure to the cold (14, 557, 558). Importantly, in contrast to studies in rodents, no study has demonstrated a complete suppression of shivering thermogenesis in humans following either cold acclimation or cold acclimatization (natural acclimation). This may be due to the intermittent nature of the cold exposure or the relatively short duration of the interventions (days to weeks) or may reflect the relative importance of skeletal muscles for generating heat in large mammals with a large muscle mass. Nevertheless, these results and the fact that relatively short and mild cold exposure leads to a 2- to 3-fold increase in BAT thermogenic activity and capacity (102) and that pharmacological inhibition of acute cold-induced BAT thermogenesis leads to increased shivering intensity (63) suggest a physiologically significant role for BAT in the cold-induced thermogenic response in humans (Fig. 7).

Figure 7.

The known physiological roles of brown adipose tissue in humans. Brown adipose tissue (BAT) contributes to nonshivering thermogenesis (NST) during cold exposure in humans. First, daily short and mild cold exposure bouts repeated during a few weeks have been shown to increase BAT volume and thermogenic capacity. Second, inhibition of BAT thermogenic activity during acute cold exposure leads to increased skeletal muscle shivering activity without change in whole-body energy expenditure. Third, prolonged cold exposure leads to reduced skeletal muscle uncoupled respiration, suggesting reduced muscle and increased BAT NST in this condition. BAT also contributes to diet-induced thermogenesis (DIT) in humans. However, the uncertainty about total thermogenically active BAT volume, which is based on BAT ¹⁸F-fluorodeoxyglucose uptake, makes imprecise the true contribution of BAT vs muscles and other organs to NST and diet-induced thermogenesis. Despite its high uptake rates of glucose, glutamate, nonesterified fatty acids (NEFA), and triglyceride-rich lipoproteins (TRL), the very small current estimates of BAT volume translate into very low (< 1%) BAT clearance rates of these blood substrates in humans. Likewise, the very small relative mass of BAT in humans and the fact that all batokines are also produced by other organs makes any purported endocrine function of BAT undetermined at the present time. BAT, brown adipose tissue; NEFA, nonesterified fatty acids; NST, nonshivering thermogenesis; TRL, triglyceride-rich lipoproteins.

Brown Adipose Tissue and Systemic Substrate Utilization

We have previously reviewed extensively the role played by BAT in systemic substrate utilization and clearance in humans (80). The major conclusion that BAT accounts for a small fraction (< 1%) of circulating glucose and fatty acid utilization and clearance during acute cold exposure remains valid (see Fig. 7). These very low values are driven by the low total body BAT volume based on the ¹⁸FDG PET method during acute cold exposure in humans. Obviously, this conclusion will remain for glucose utilization and clearance, but greater fractional uptake values may be expected for other substrates (fatty acids, branched-chain amino acids, glutamate) if larger thermogenically active BAT volumes are eventually revealed using alternative approaches (discussed earlier). Because BAT volume and metabolic activity can potentially be recruited at much higher levels than during acute cold exposure (see earlier sections), it is not excluded that BAT may eventually serve as a substantial systemic sink for energy substrates if potent and chronic metabolic activation of this tissue can be achieved in humans (see the subsequent section on pharmacological activation). However, the volume of BAT would need to be at least 1 order of magnitude above the typical 150 g of BAT currently typically determined using ¹⁸FDG PET (80).

Brown Adipose Tissue and Postprandial Thermogenesis

The classical studies by Rothwell and Stock and Bray et al have shown increased sympathetic activation, BAT hypertrophy, and increased energy expenditure during the administration of high-calorie diets in rats (559–562). This diet-induced thermogenic response is associated with increased BAT respiration and mitochondrion proton conductance (562, 563) and is blunted in the absence of UCP1 (564). The role of BAT in diet-induced thermogenesis is however not universally accepted, mainly on the basis that a clear increase of the contribution of BAT energy expenditure has not been demonstrated in vivo in rodents with excess caloric intake (565).

In humans, carbohydrate-rich meals, in particular, activate the SNS response (566). BAT ¹⁸FDG uptake increases postprandially, but this increase is lower than cold-induced BAT ¹⁸FDG uptake and does not correlate with the increase in energy expenditure after the meal (567). Using ¹⁵O-O₂ PET dynamic scanning, Din and colleagues (321) found postprandial BAT oxidative metabolism to be increased at a similar level to that induced by acute cold exposure in healthy participants. Postprandial increase in energy expenditure has been shown to be higher in those with higher BAT glucose uptake (568). A more recent study, however, did not find an association between maximally stimulated BAT ¹⁸FDG uptake using combined acute cold exposure and treatment with mirabegron and diet-induced thermogenesis in healthy individuals (569). As discussed in the preceding sections, the effect of this postprandial BAT activation on whole-body energy balance and energy substrate clearance is unclear, but likely to be modest at best (see Fig. 7). Also, as discussed in earlier, several pancreatic, gastrointestinal, and hepatic hormones, such as insulin, GLP1, GIP, secretin, cholecystokinin, and FGF21, are known to modulate BAT metabolism in experimental models. Postprandial secretin levels have been associated with postprandial increase in BAT thermogenesis, and secretin administration increases BAT ¹⁸FDG glucose uptake in humans (489). The role played by the other hormones in this postprandial increase of BAT oxidative metabolism is currently unknown.

Brown Adipose Tissue as a Paracrine and Endocrine Organ

BAT can produce a wide variety of peptides and metabolites, coined “batokines,” that are known to exert biological effects through autocrine, paracrine, and potentially endocrine routes (570, 571). Deshmukh et al (572) found that the secretome of human BAT includes at least 471 proteins, 101 of which were not found in the secretome of human subcutaneous abdominal WAT. Adiponectin, fibrillin-1, granulins, hepatoma-derived growth factor, but not leptin, are among the potential batokines (572). Many of these proteins stimulate brown adipogenesis, mitogenesis, angiogenesis, neurogenesis, and brown adipocyte energy metabolism and can modulate the innate immune response. Among the proteins preferentially secreted by the human brown adipocytes, mammalian ependymin-related protein 1 (EPDR1) has been shown to stimulate brown differentiation and noradrenaline-mediated thermogenic response in brown adipocytes in vitro (572). However, EPDR1 did not stimulate BAT energy metabolism in vivo in mice (572).

BMP4 and BMP7, proteins of the transforming growth factor (TGF)-β superfamily, are among the best studied of these batokines. BMP4 and BMP7 are secreted by mature adipocytes and stimulate early commitment of brown preadipocyte lineage and, possibly later stages of brown adipogenesis (29, 573–577). Another member of the TGF-β superfamily, BMP8b, is secreted by mature brown adipocytes and stimulates the production of angiogenic and neurogenic factors (vascular endothelial growth factor A [VEGFA] and neuregulin-4) for angiogenesis and neurite outgrowth (578, 579). In WAT, brown adipogenesis occurs in close proximity to dense sympathetic neurites, critically dependent on the expression of PRDM16 in adipocytes early in life (580, 581). This coordinated vascular and neural development is likely critical for BAT metabolic activity in vivo in humans (189). Indeed, brown adipocytes stimulate BAT vascularization via paracrine regulation. VEGFA is secreted by human BAT, but not WAT, on noradrenaline stimulation (572). VEGFA ablation in BAT leads to the whitening phenotype seen during the development of obesity (582), and its overexpression in BAT stimulates its vascularization and thermogenic capacity (583, 584). The development of beige adipocytes is also stimulated by angiogenic factors and coupled with angiogenesis in adipose tissues (585).

Batokines can also recruit immune cells contributing to its development and function. For example, CXCL14 is secreted by thermogenically active brown adipocytes leading to the recruitment of M2 macrophages in BAT. These macrophages may in turn contribute to brown adipogenesis, angiogenesis, and neurogenesis (586). Although early reports found evidence for immune cell-mediated production of noradrenaline for the regulation of BAT metabolic activity (587, 588), these findings were subsequently contested (589).

Batokines that appear to improve in vivo energy metabolism at least in mice also include Slit homolog 2 protein (Slit2) (590) and IL-6 (591). In addition to metabolites discussed previously (lactate, acetate), brown adipocytes secrete other metabolites such as 3-methyl-2-oxovaleric acid and 5-oxoproline that activate cAMP-PKA-p38 MAPK, and beta-hydroxyisobutyric acid that activates mTOR, leading to activation of mitochondrial oxidative metabolism in muscle cells in vitro and in vivo in mice in addition to inducing white adipocyte browning; in humans, plasma levels of these metabolites correlate inversely with body mass index (BMI) and directly with WAT RNA expression of UCP1 and CPT1b (592). FGF21 was reported to be produced in mice brown adipocytes on thermogenic activation (593), but was not detected in the human brown adipocyte secretome (572). In contrast, myostatin secretion by BAT is increased at thermoneutrality and leads to a reduction in muscle mitochondrial function and exercise capacity in mice (594). 12,13-Dihydroxy-9Z-octadecenoic acid (12,13-diHOME), a bioactive lipid secreted by BAT, has been shown to promote fatty acid uptake in BAT and skeletal muscles and to improve cardiac function in mice (595–597). miRNAs produced by BAT (ie, miR-99b) may also stimulate liver FGF21 secretion to exert beneficial metabolic effects, at least in mice (reviewed in [598]). Readers are also referred to a recent review by Scheele and Wolfrum for a detailed discussion on batokines (570).

Because most of the batokines are not exclusively produced by brown adipocytes and because of the relatively small size of BAT compared to the other organs producing these potential hormones, the role of BAT as an endocrine organ is very doubtful in humans (see Fig. 7). Although autocrine/paracrine effects of many of these batokines are likely, whether they play important endocrine roles remains to be demonstrated.

Brown Adipose Tissue in Cardiometabolic Diseases

Mechanisms of Obesity-induced Impairment of Brown Adipose Tissue Activity and Capacity

Obesity and obesogenic conditions have profound effects on BAT development and function. Obesity and adipocyte hypertrophy lead to adipose tissue infiltration by proinflammatory macrophages and other immune cells, reduced regulatory T cells, lower rates of adipogenesis, and adipocyte apoptosis and senescence (599). The inflammatory response seen in BAT, however, is of a lower magnitude than that observed in WAT (600, 601). The presence of BAT-specific regulatory T cells allows the adaptive thermogenic response in mice and prevents the BAT proinflammatory phenotype (602). iBAT expansion is delayed compared to that of WAT in response to an HFD in mice; iBAT immune trafficking gene expression is preceded by muscle development gene expression and followed by lipid metabolic and immune response gene expression (603). M1 macrophage programming with HFD represses BAT UCP1 expression and in vivo adaptive thermogenesis at least in part via the production of tumor necrosis factor α (TNF-α) in mice (604). Interleukin-4–dependent alternative macrophage activation programming was shown, however, to support the cold-induced thermogenic response (587). Apoptosis of white and brown adipocytes can also be induced by TNF-α (605, 606). Local excessive production of nitric oxide via the activation of nitric oxide synthase 2 can also inhibit brown adipocyte respiration in addition to inducing inflammation and impairing insulin signaling (607). Activation of the nuclear factor κB (NF-κB) pathway in adipocytes also leads to the downregulation of ADRB3 receptors and catecholamine resistance in mice (608, 609).

Toll-like receptors (TLRs) and nucleotide-oligomerization domain-containing proteins (NOD) are pattern-recognition receptors playing an important role in the initiation of the innate immune response. An HFD induces iBAT TLR2, TLR4, NOD1, and NOD2 gene expression, the stimulation of which activates NF-κB and MAPK pathways, triggering proinflammatory cytokine/chemokine production (610). This leads to the reduction of UCP1 expression and brown adipocyte respiration in vitro (610). Activation of these pattern-recognition receptors during adipogenesis inhibits brown adipose commitment likely through proinflammatory cytokine-mediated reduction of PPAR-γ transactivation (611–613).

ER stress occurs from proteotoxicity, lipotoxicity, and/or glucotoxicity and affects protein synthesis and quality control, calcium homeostasis, cellular energy metabolism, oxidative stress, inflammation, and autophagy (614). ER stress with unfolded protein response is seen in the BAT of HFD-fed mice (41, 601). Although excess ER stress may lead to apoptosis, unfolded protein response without caspase activation may increase brown preadipocyte differentiation (615, 616).

Mast cells are proinflammatory immune cells that are more prevalent in WAT of mice housed at thermoneutrality (617) and in WAT of obese mice (618) and humans (619). Mast cells’ serotonin production was shown to inhibit brown adipocyte development and function in mice (617, 620, 621). This effect is likely mediated in situ as CNS serotonin stimulates the sympathetic outflow signal to activate BAT and WAT browning (622), although a recent study suggests that systemic serotonin can also inhibit BAT sympathetic signaling via the stimulation of GABA signaling in the hypothalamus (623).

Obesity is associated with WAT resistance to noradrenaline-induced lipolysis in vitro (624) and in vivo in mice (609, 625) and in humans (626). Inflammation inhibits catecholamine signaling through cAMP, with downregulation of ADRB3 (609), leading to reduced brown adipocyte proliferation and recruitment (586). This is likely caused by IKK-ε-induced production of phosphodiesterase 3B (PDE3B) (608) and tribbles pseudokinase 1 (TRIB1)-mediated degradation of the ADRB3 transcriptional activator C/EBPA (609). Switching of M2 to M1 macrophages was initially shown to reduce local tissue macrophage catecholamine production (587). However, catecholamine production by macrophages was not found in subsequent investigations by other groups (589). The activity of monoamine oxidase (MAOA) is increased in adipose tissue macrophages in aging mice (627) through reduced growth differentiation factor-3 (GDF3) expression in macrophages (628, 629). The activity of MAOA in adipose tissues also increases with aging in humans, but this is attributable to increased expression directly in adipocytes (630). Interestingly, humans do not express MAOA in immune cells as mice do according to consortium-based gene expression atlases (631), although the presence of the catecholamine degradation enzymes in some human macrophages of the nervous system has been reported (632). Therefore, macrophages associated with sympathetic nerves in adipose tissues can potentially take up and degrade catecholamines, reducing the browning process in WAT (632).

To summarize, multiple overlapping and possibly synergistic mechanisms lead to brown adipocyte inflammation, IR, reduced catecholamine signaling, apoptosis, and impaired recruitment, potentially leading to impaired thermogenic activity and capacity during the development of obesity and cardiometabolic diseases (Fig. 8). Although these mechanisms have been well characterized in animal models, their role in human BAT metabolism in vivo is still poorly understood.

Figure 8.

Proposed pathophysiological mechanisms of reduced brown adipose tissue glucose metabolism and thermogenic activity and capacity. In situ reduction of regulatory T cells, macrophage M1 activation, and infiltration by other immune cells such as mastocytes increase local production of proinflammatory cytokines that reduce brown adipocyte recruitment. Proinflammatory cytokines, nitric oxide, and serotonin produced by these cells have also all been shown to activate intracellular brown adipocyte inflammatory signaling pathways, including the nuclear factor-kappa B (NF-kB) and the mitogen-activated protein kinase (MAPK)/extracellular signal-regulated kinase (ERK) pathways. In turn, these pathways activate the intracellular degradation of cyclic adenosine 3’,5’-monophosphate (cAMP) that impairs beta adrenergic signaling and shuts down the thermogenic program. They also lead to adipocyte insulin resistance and reduced glucose uptake. Energy surfeit contributes to increase brown adipocyte triglyceride (TG) deposition, leading to whitening of brown adipose tissues. The ensuing lipotoxicity and glucotoxicity activate toll-like receptors (TLRs) and endoplasmic reticulum (ER) stress, also increasing intracellular inflammation. Aging has been shown to increase monoamine oxidase (MAOA) expression and activity in brown adipocytes, leading to local degradation of noradrenaline and reduced beta adrenergic signaling. β1/2/3, beta adrenergic receptors 1, 2, and 3; cAMP, cyclic adenosine 3’,5’-monophosphate; ER, endoplasmic reticulum; ERK, extracellular signal-regulated kinase; MAOA, monoamine oxidase; MAPK, mitogen-activated protein kinase; NF-κB, nuclear factor-kappa B; TG, triglycerides; TLRs, toll-like receptors.

In Vivo Brown Adipose Tissue Activity in Cardiometabolic Disorders

Obesity and obesogenic conditions lead to the reduction of in vivo BAT glucose uptake in animal models and humans. For example, male Zucker diabetic fatty rats display lower BAT ¹⁸FDG uptake with whitening of BAT; caloric restriction partially restores these defects (633). In humans, several cross-sectional studies have demonstrated a reduction in BAT ¹⁸FDG uptake in with obesity and IR states (7, 8, 157, 319, 417). In a recent large cross-sectional study, 856 ¹⁸FDG PET BAT-positive individuals were found to have lower visceral fat mass, liver fat content, and prevalence of T2D vs 846 propensity score–matched ¹⁸FDG PET BAT-negative individuals, independent of total fat mass (169). In a later study, the level of BAT ¹⁸FDG uptake was also significantly and inversely associated with these dysmetabolic features. Insulin stimulates BAT ¹⁸FDG uptake but not blood flow (175), and this insulin stimulation of BAT glucose metabolism is blunted in individuals with obesity (319). BAT ¹⁸FDG uptake also increases after weight loss in patients with obesity in some (634) but not all studies (319). In a 12-week exenatide-treatment, single-arm study, nonobese healthy men lost 1.5 kg of body weight and had reduced plasma TG levels with a significant increase in cold-induced BAT ¹⁸FDG uptake (341). However, 28 days of treatment with pioglitazone, a powerful IS drug, reduced cold-induced BAT ¹⁸FDG uptake in healthy lean individuals (635). In a later study, plasma insulin levels were not affected by pioglitazone treatment, suggesting no effect on whole-body IS. Also, the degree of habitual physical activity or energy intake are not associated with cold-induced BAT ¹⁸FDG uptake in young, healthy individuals (636, 637). Furthermore, exercise training that increases insulin-stimulated muscle and subcutaneous WAT ¹⁸FDG uptake does not increase insulin-stimulated BAT ¹⁸FDG uptake in healthy men (638) and endurance athletes even display lower cold-induced BAT ¹⁸FDG uptake vs sedentary but healthy men (496). Thus, cold-stimulated BAT glucose uptake is likely maximal in lean, healthy individuals, but can probably improve with treatment of obesity or IR in people with dysmetabolism.

Several lines of evidence suggest that obesity-associated reduction of BAT glucose uptake is not indicative of reduction of in vivo BAT thermogenesis. iBAT of mice on an HFD display increased ER stress, ROS generation, and inflammation, but with higher mitochondrial respiration and UCP1 (601). In rats, iBAT insulin signaling and lipid metabolism gene expression is increased with an HFD (639). An HFD in mice increases BAT CPT2 and CPT1b protein expression while deteriorating whole-body IS (640). In humans, we found a profound reduction of cold-induced BAT ¹⁸FDG uptake in the face of normal BAT oxidative metabolism, based on ¹¹C-acetate PET, and normal BAT TG mobilization, based on CT radiodensity shift, in patients with T2D (103). As discussed in previously, we recently showed that a hypercaloric (+25% of daily caloric intake) high-fructose feeding protocol that eventually leads to the development of visceral obesity, systemic IR, hepatic steatosis, and postprandial dyslipidemia (641, 642) reduces cold-induced BAT ¹⁸FDG uptake, but not BAT oxidative or BAT TG mobilization (Richard, Blondin, Carpentier et al. Unpublished). Remarkably, this change occurred after only 2 weeks of high-fructose feeding and before any substantial increase in body fat mass, systemic IR, or circulating TGs.

Nevertheless, other metabolic dysfunctions of BAT may occur in obesity. In rodents, streptozotocin-induced diabetes and diet-induced obesity result in lower ¹⁸F-labelled BODIPY-TG-chylomicron-like particle uptake by BAT during acute cold stimulation (118). Supraclavicular BAT uptake of (S,S)-¹¹C-O-methylreboxetine (¹¹C-MRB), a selective ligand for the noradrenaline transporter, is reduced in obese individuals, suggesting reduced sympathetic innervation (643). Supraclavicular fat NEFA uptake and blood flow (using IV ¹⁸FTHA and ¹⁵O-H₂O PET methods) are lower at room temperature and during acute cold exposure in obese individuals compared to age-matched nonobese controls (644, 645). Cold-induced BAT NEFA uptake and thermogenesis (using the ¹¹C-acetate PET method), however, were not different in men with T2D vs healthy controls of similar BMI in another study (103). Supraclavicular fat NEFA uptake also significantly increases with weight loss 6 months after bariatric surgery (644). Although BAT NEFA uptake and blood flow were associated with BAT thermogenic activity using the ¹⁵O-O₂ PET method in healthy individuals in 1 study (n = 7) (180), there exist no such published data in individuals with cardiometabolic disorders.

Cold-induced reduction in BAT fat fraction was significantly associated with whole-body IS determined by euglycemic-hyperinsulinemic clamp in a small cross-sectional human study (99). Supraclavicular BAT TG content using either CT radiodensity or MRI fat fraction is higher in participants with obesity or T2D and is associated with dysmetabolic features such as increased adiposity, visceral fat mass, circulating TGs, and whole-body IR (97, 103, 644–646). BAT TG content also significantly decreases with weight loss and improvement in systemic IS after bariatric surgery (644). The metabolic source and consequence of this TG accumulation in BAT with obesity is unknown, although a recent study in mice suggested an important role for de novo lipogenesis in BAT thermogenic involution associated with BAT TG accumulation during acclimation to thermoneutrality (92).

Thus, obesity and obesogenic conditions lead to a whitening phenotype of BAT in vivo that is associated with reduced BAT blood flow and glucose and NEFA uptake. The reduction of BAT glucose uptake appears earlier than other BAT dysmetabolic features during the development of obesity and IR. However, there are very limited assessments of in vivo BAT thermogenic activity per se in these conditions. The relationship between the whitening phenotype of BAT and in vivo thermogenic activity and capacity of BAT is poorly understood at the moment. More studies are needed to determine the relation between BAT TG/NEFA cycling and metabolism and its thermogenic activity in vivo in humans.

Role of Brown Adipose Tissue in the Pathogenesis of Obesity and Cardiometabolic Diseases

Despite the profound effect of BAT thermogenesis on energy balance and metabolic health in rodent models (647), any impairment of BAT thermogenic activity is unlikely to play a predominant role in the development of obesity in humans. Several lines of evidence demonstrate the predominant contribution of disordered regulation of food intake vs energy expenditure in the pathogenesis of human obesity. First, genetic disorders causing severe obesity and most gene polymorphisms associated with obesity in humans affect primarily the regulation of food intake (648). Second, weight loss is more readily achieved with restriction of energy intake vs physical activity (649). Third, there is a larger contribution of increased energy intake vs reduced expenditure in the metabolic adaptation to weight loss (650). Fourth, bariatric surgery and the most effective pharmacotherapies such as GLP1 receptor agonists and coagonists exert their weight-loss–promoting effects through restriction of energy intake, not increased expenditure (651, 652).

Nevertheless, a modest effect of BAT thermogenesis on long-term body weight regulation is plausible, as suggested by the association between UCP1 genetic variants and differential weight gain in some (539, 653–657) but not all (658–660) human studies. As discussed earlier, UCP1-independent thermogenic mechanisms may also contribute to BAT energy expenditure, contributing to these discrepant observations. BAT volume determination is currently based on the assessment of BAT glucose metabolism, not thermogenesis. Because BAT glucose uptake is specifically reduced in IR states (as discussed earlier), thermogenically active BAT volume is likely severely underestimated in most individuals using the current standard ¹⁸FDG PET method, given the very high prevalence of IR even in otherwise healthy people. We have previously calculated BAT potential energy expenditure in humans (80) based on (1) the reported rate of BAT thermogenesis at room temperature and during mild cold exposure using the ¹⁵O-O₂ PET method (180); (2) the reported possible range of BAT mass determined by the current ¹⁸FDG method (higher range at ~150 g) or by the reported BAT-containing adipose mass range (550-2550 g) (661). Using these assumptions, we could estimate BAT energy expenditure to range between 7 to 123 kcal/day at room temperature and between 12 to 211 kcal/day during mild cold exposure (80). In addition, postprandial BAT thermogenesis was reported at up to approximately 13 kcal/day (321). It should be noted that these figures do not take into account the possibility that WAT browning may increase in energy expenditure through uncoupled respiration or, more likely, through increased TG/NEFA cycling. We previously determined that WAT TG/NEFA cycling accounts for up to one-third of the increase in energy expenditure observed during mild cold exposure in humans (121). The obesity epidemics in North America result from an average sustained weight gain of approximately 0.5 to 0.7 kg per year in young adults (662, 663). An average weight loss of approximately 2 kg sustained over 10 years is associated with a 18% to 34% reduction in incident T2D (664). Based on an energy density of 8840 kcal/kg (665), these figures represent a daily energy imbalance ranging from 12 to 48 kcal. Therefore, chronic activation of BAT thermogenesis has the potential to curb weight gain and incident T2D over prolonged periods. The magnitude of BAT-attributable energy expenditure needs to be directly quantified in humans, particularly in the context of obesity.

Because of its high metabolic rate and glucose uptake, BAT metabolic activation has been proposed as a therapeutic target for T2D and IR states. Early studies using cold acclimation in T2D found significant reduction in plasma glucose levels and marked improvement in whole-body IS associated with increased BAT ¹⁸FDG uptake (556). However, cold exposure increases shivering and glucose uptake of several large groups of postural skeletal muscles that account for approximately 50% of systemic glucose production vs approximately 1% for BAT during cold exposure (405). A more recent study confirmed the importance of the muscle shivering response in cold-acclimation–induced whole-body IS in participants with T2D (666).

A recently published large cross-sectional study has provided evidence for a relationship between BAT metabolism and cardiovascular disease. Becher et al (168) retrospectively categorized 52 487 patients as BAT-positive or BAT-negative based on ¹⁸FDG PET scans and showed that the presence of BAT was associated with a lower prevalence of T2D, dyslipidemia, coronary artery disease, cerebrovascular disease, congestive heart failure, and hypertension independently of sex, age, BMI, and outdoor temperature. These associations of the presence of BAT with a lower prevalence of cardiovascular disease remained with adjustment for the presence of T2D. However, it was not possible to adjust for IR per se, which may confound the association between the absence of ¹⁸FDG uptake in BAT and higher risk of cardiovascular disease. Another smaller study (5 men/26 women) demonstrated that cold-induced BAT glucose uptake and blood flow were directly associated with brachial flow–mediated dilatation and inversely associated with carotid intima media thickness after 5 years of follow-up (667).

In summary, there is strong preclinical evidence to suggest an obesity-induced reduction in brown adipocyte thermogenic activity and capacity. In humans, BAT ¹⁸FDG uptake is very sensitive to obesity, and may be an early marker of the development of dysmetabolic features during hypercaloric high-fructose feeding. BAT TG content also increases with obesity, but whether in vivo BAT thermogenic activity and capacity are reduced in obese humans is currently not resolved. More studies are needed to address this question.

Pharmacological Activation of Brown Adipose Tissue

Although cold exposure and acclimation is a very effective way to increase BAT thermogenic activity in humans, these interventions may be cumbersome, are unacceptable to many people, and do not specifically target the activation of BAT thermogenesis. Furthermore, cardiovascular safety concerns have been raised regarding intense cold exposure in patients at high risk of cardiovascular events (668, 669). Safe, effective, and specific pharmacological activation of BAT will be key to prove the relevance of this potential target for the treatment of cardiometabolic disorders. In this section, we review the currently existing evidence for drug-induced metabolic activation of BAT. These include sympathomimetic drugs and PPAR-γ, transient receptor potential vanilloide 1 (TRVP1), and G protein–coupled bile acid receptor Gpbar1 (TGR5) agonists.

Sympathicomimetic drugs known to increase energy expenditure such as phentermine (670), sibutramine (671), norpseudoephedrine (672), fenfluramine, and dexfenfluramine (673) have been shown to induce small to moderate weight loss in individuals suffering from obesity. The weight loss–promoting effects of phentermine and norpseudoephedrine have been shown to wane after a few weeks, suggesting a tachyphylactic response to treatment, and therefore limiting their indication to short-term treatment only (673). In the only long-term controlled cardiovascular outcome study with these drugs, sibutramine was shown to increase nonfatal cardiovascular events (671), probably because of chronotropic stimulation of the heart, an effect shared by all β-ADRs, including β-3 ADR agonists (discussed subsequently).

Acute administration of ephedrine was shown to increase (674) or not change (675) BAT ¹⁸FDG uptake in healthy individuals. Prolonged administration of ephedrine for 28 days in healthy men led to reduced total and visceral fat mass, but no change in resting energy expenditure and even to a reduction in BAT ¹⁸FDG uptake after acute administration of ephedrine at the end of the treatment period (179). The increases in systolic blood pressure and blood glucose after acute ephedrine administration were blunted after prolonged ephedrine treatment, but the authors did not report plasma insulin levels to help the interpretation of these findings.

More recently, acute oral administration of mirabegron (200 mg), a more selective β-3 ADR agonist, was shown to increase BAT ¹⁸FDG uptake in healthy men, associated with significant increase in whole-body energy expenditure and heart rate (171). Acute administration of mirabegron 200 mg was also shown to increase supraclavicular temperature and reduce BAT fat fraction, although less markedly than cold exposure (676). Acute administration of the 50-mg dose of mirabegron, in contrast to the 200-mg dose, did not increase BAT glucose uptake or resting energy expenditure in healthy men (677). In an open-label, 4-week treatment trial with mirabegron (100 mg/day) in 14 healthy women, O’Mara et al (172) showed increased BAT ¹⁸FDG uptake and volume with increases in resting energy expenditure, IS, secretion, and disposition index based on a frequently sampled IV glucose tolerance test, and increase in high-density lipoprotein cholesterol and adiponectin. There was no change in homeostatic assessment model of insulin resistance, body weight, or body composition in a later study. Interestingly, 4-week mirabegron treatment also resulted in lower WAT ¹⁸FDG uptake in that same study (172). This is compatible with blunting of the metabolic response of adipose tissues during chronic treatment with β-3 adrenergic agonists demonstrated in rodents (678–680). Another open-label study showed that 12-week mirabegron treatment (50 mg/day) of healthy individuals led to increased WAT expression of UCP1, TMEM26, CIDEA proteins, and phosphorylation of hormone-sensitive lipase on serine 660, suggesting considerable browning of subcutaneous WAT with prolonged treatment with mirabegron (681). Another study by the same group extended these findings by showing that pioglitazone 30 mg/day for 12 weeks in overweight or obese participants also resulted in WAT browning, but that combined treatment with mirabegron + pioglitazone actually reduced WAT browning marker levels compared to monotherapy (682). None of the treatment arms led to increased cold-induced BAT ¹⁸FDG uptake.

Again, the demonstration of increased BAT glucose uptake does not necessarily imply activation of BAT thermogenesis. Even mice lacking UCP1 increase their metabolic rate by 50% following injection of a β-3 adrenergic agonist (647), supporting additional mechanisms than BAT thermogenesis for the increase in whole-body energy expenditure elicited by these drugs. We subsequently demonstrated that acute administration of mirabegron 200 mg indeed increases BAT glucose uptake, whole-body energy expenditure (+17%), and heart rate, but not BAT thermogenesis (assessed using ¹¹C-acetate PET) or BAT CT radiodensity (a marker of reduced BAT TG content during cold exposure) in healthy men (121). In that same study, acute administration of mirabegron 50 mg, the appropriate therapeutic dose to activate the β-3 adrenergic receptor in humans (https://www.accessdata.fda.gov/drugsatfda_docs/nda/2012/202611Orig1s000SumR.pdf), did not increase BAT glucose uptake or thermogenesis, but nevertheless increased resting energy expenditure by 12% and increased the heart rate of the participants. In contrast, acute cold exposure in the same participants led to further increase in BAT glucose uptake and a significant increase in BAT thermogenesis and reduction in TG content, together with a much higher increase in whole-body energy expenditure (+55%). Similarly to Baskin et al (677), we also found a significant increase in plasma NEFA levels and apparition rate after acute administration of mirabegron 200 mg, associated with increased whole-body TG/NEFA cycling (121). This strongly suggests that mirabegron 200 mg, a supra-therapeutic dose for activation of the β-3 adrenergic receptor, also activates WAT lipolysis and TG/NEFA cycling probably by activating β-2 adrenergic receptors. We indeed found that ADRB2 is the predominantly expressed β-adrenergic receptor and stimulates lipolysis and thermogenesis in human BAT (121). Another group found predominant ADRB1 expression with low ADRB3 levels in immortalized human brown adipocytes (683). Interestingly in mice, acute β-3 adrenergic stimulation leads to increased energy expenditure, NEFA production, insulin secretion, and satiation that is dependent on the expression of the β-3 adrenergic receptor both in WAT and BAT, not only in BAT (684, 685). Therefore, the in vivo metabolic response to β-adrenergic stimulation is driven indissociably by WAT and BAT in mice as well as in humans. Furthermore, the increase in heart rate and blood pressure observed with the use of mirabegron is cause for concern about the risk of cardiovascular events over the long term.

PPAR-γ activation leads to the recruitment of thermogenic brown adipocytes in WAT in vitro (28, 686) and has long been known to increase adipocyte browning in rodents in vivo (687–691). However, pioglitazone treatment for 28 days in humans did not induce WAT browning and even led to reduced cold-induced BAT glucose uptake in 1 study (635). Another study found a significant increase in subcutaneous WAT browning markers after 12 weeks of treatment with pioglitazone 30 mg per day in overweight or obese individuals, but again without an increase in cold-induced BAT glucose uptake (682). The clinical significance of WAT browning for the marked antidiabetic effects of PPAR-γ agonists is unclear at the moment, whereas the metabolic activation of BAT through PPAR-γ agonists is unlikely in humans.

Green tea extracts such as catechin and caffeine have been shown to increase energy expenditure and fat oxidation and promote adipose tissue loss in animal and human studies (692, 693). These compounds inhibit catechol-O-methyl transferase and cAMP-degrading phosphodiesterases, leading to increased noradrenaline signaling and adipose tissue thermogenesis (694). Caffeine exposure of adipose stem cells in vitro increases the expression of UCP1 protein and other markers of adipocyte browning and increases mitochondrial respiration and proton leak; it also leads to increased skin temperature in the supraclavicular region after acute ingestion of caffeine in young healthy individuals (695). Twelve-week daily ingestion of a catechin-rich beverage led to an increase in supraclavicular hemoglobin concentration assessed by near-infrared time-resolved spectroscopy, a surrogate marker of BAT vascularization, with a reduction in extracellular muscle TG content using proton-MRS in healthy women (696). Ingestion of a beverage including catechin and caffeine acutely increased resting energy expenditure in healthy men with positive ¹⁸FDG BAT uptake, but not in those with negative uptake (697). Furthermore, daily consumption of a beverage with catechin + caffeine for 5 weeks increased cold-induced energy expenditure in healthy participants (697). More studies are needed to document specifically BAT thermogenesis in response to catechol-O-methyl transferase and cAMP-degrading phosphodiesterase inhibitors.

Menthol (2-isopropyl-5-methyl-cyclohexanol) activates the cold-sensing transient receptor potential melastatin 8 (TRPM8), a calcium channel receptor in sensitive neurons of the skin and gastrointestinal system (698, 699). TRPM8 is also present on the cell membrane of white and brown adipocytes, where its activation increases UCP1 expression (670–702). TRPM8 KO mice also show increased brown adipocyte intracellular TGs and reduced UCP1 expression, together with reduced expression and circadian oscillation of clock genes in brown adipocytes (703). Oral treatment with menthol increases core temperature through activation of BAT and an increase in locomotor activity in mice, an effect abolished in TRPM8 KO animals (700). Menthol treatment also induces WAT browning and prevents HFD-induced obesity in mice (702). Treatment with icilin, another TRPM8 agonist, also prevents diet-induced obesity in mice by increasing thermogenesis (704). It is possible that at least part of these antiobesity properties of TRPM8 agonists is mediated through stimulation of glucagon secretion in mice (705). Acute topical administration of L-menthol increased energy expenditure marginally together with reduction in skin blood flow compared to oral L-menthol or placebo in healthy men and women (706). We are not aware of any study on the effect of menthol or other TRPM8 agonists on BAT metabolism or thermogenesis in vivo in humans.

Capsinoids are potential BAT-activating agents (694, 707, 708). Chronic capsaicin administration was shown to promote weight loss in humans and animal models in many but not all studies of this topic (see (709) for review). Capsaicin activates TRPV1, a cell membrane cation channel originally described as a sensory nerve noxious heat receptor (710). Capsaicin treatment prevents HFD-induced obesity in mice potentially through TRPV1-dependent stimulation of SIRT1, deacetylation of PPAR-γ and PRDM16, and stimulation of BAT thermogenesis (40) and ingWAT browning (711). Capsaicin administration also increases the postprandial adrenergic response and energy expenditure in lean but not in obese individuals (712, 713). Cold-induced thermogenesis was shown to be higher after 6 weeks of capsaicin supplementation in healthy participants (159). Higher resting metabolic rate was reported after acute administration of capsaicin in ¹⁸FDG-BAT–positive individuals (714, 715). Six-week capsaicin supplementation in healthy individuals increased supraclavicular BAT ¹⁸FDG uptake (716) and supraclavicular hemoglobin concentration assessed by near-infrared time-resolved spectroscopy (716, 717), taken as a surrogate for BAT vascularization and thermogenic activity. In contrast, acute capsaicin administration did not increase BAT ¹⁸FDG uptake even in healthy individuals displaying a cold-induced increase in BAT ¹⁸FDG uptake (715). Experiments in rats also suggest that activation of TRPV1 in the nucleus tractus solitarius or IV administration of the TRPV1 agonist dihydrocapsaicin can reduce efferent adrenergic stimulation and BAT thermogenesis (718). Furthermore, in vivo capsaicin’s effects on energy metabolism are pleiotropic. Activation of TRPV1 may reduce appetite and increase satiety through reduced ghrelin secretion, increased GLP1 secretion, and/or increased vagal afferent signaling (709, 713). It is also well known to promote skin vasodilation, which can increase heat loss (709), leading to indirect activation of the cold-adaptive response. In addition to capsaicin, TRPV1 can be activated by linoleic and arachidonic acid metabolites, such as 9- and 13-hydroxyoctadecadienoic acid, 20-hydroxyeicosatetraenoic acid, and endocannabinoids (719–722). Recently, TRPV1-positive perivascular smooth muscle cells from the periaortic region and iBAT were found to contribute to cold-induced brown adipocyte recruitment in addition to platelet-derived growth factor receptor A (PDGFRA)-positive smooth muscle cells (723, 724). Whether capsinoids or endogenous fatty acid metabolites could potentially increase TRPV1+ smooth muscle cell–derived brown adipocyte recruitment is currently unknown. The physiological regulation of TRPV1 is very complex and not enough understood to make TRPV1 activation with capsinoids a good drug target for the selective activation of BAT in humans. Although capsaicin appears to stimulate energy expenditure, there is no experimental evidence currently supporting a direct stimulation of BAT thermogenesis in vivo in humans.

Fish oil supplements containing DHA and/or eicosapentaenoic acid (EPA) reduce plasma TGs and may exert limited cardiovascular benefits (725), although combined DHA + EPA did not show any benefit in a recent large, randomized controlled trial (726). In contrast, treatment with ethyl EPA (icosapent ethyl) was shown to confer important cardiovascular benefits in high-risk participants in a large, randomized controlled trial (727). Despite some early reports of increased expression of iBAT UCP1 and thermogenic gene expression, increased BAT mass and induction of WAT browning with fish oil supplementation in rodents (728–732), a more recent report did not reproduce these findings under thermoneutral conditions (733). Another recent study showed a similar antiobesity effect and stimulation of energy expenditure with fish oil supplementation in wild-type or UCP1 KO mice (734), suggesting that activation of BAT thermogenesis was not responsible for the effects of this treatment. Fish oil supplements in humans increased WAT browning in one report (735) but not in another (733). An association was reported between BAT ¹⁸FDG uptake with fasting plasma DHA and EPA in humans (736). However, no study has thus far directly addressed the effect of fish oil, DHA, or EPA on in vivo BAT thermogenesis in humans.

Finally, 2-day treatment with chenodeoxycholic acid, a bile acid that activates BAT type 2 iodothyronine deiodinase via activation of the TGR5 (737), has been shown to increase BAT ¹⁸FDG uptake at room temperature in 12 young, healthy women, with an increase in basal energy expenditure (738). In that same publication, there was a significant TGR5-dependent increase in ex vivo uncoupled respiration of human brown adipocytes with chenodeoxycholic acid treatment.

To summarize, there is currently no direct in vivo evidence that BAT thermogenesis can be selectively activated by any of the tested classes of drugs, from sympathicomimetics, including the β-3 adrenergic agonist mirabegron, to PPAR-γ agonists, or other drugs known to induce some metabolic effects in humans (Table 1). Regarding mirabegron, its demonstrated stimulation of BAT glucose uptake occurs at a dose that likely stimulates the β-1/β-2 adrenergic receptors that are predominant in human BAT. Using this drug, it appears that any stimulation of BAT metabolism may be inseparable from activation of WAT TG/NEFA cycling and some cardiac chronotropic effects. The few small studies showing an increase in WAT browning markers did not provide convincing evidence for a role of this effect on any potential metabolic benefit observed.

Table 1.

Effect of pharmacological interventions on brown adipose tissue metabolism in humans

Target	Agent	Reference	Dose/duration	BAT glucose uptake	BAT T°	∆ Body wt	REE	CV effects
Sympathomimetic	Ephedrine	(673)	Acute 2.5 mg/kg	↑	?	–	–	↑ BP
	Ephedrine	(674)	Acute 1 mg/kg	↔	?	–	↑	↑HR, BP
	Ephedrine	(179)	28 d 1.5 mg/kg	↓	?	↓	↔	↑HR, BP
β-adrenergic agonist	Mirabegron	(171)	Acute 200 mg	↑	?	–	↑	↑HR, BP
	Mirabegron	(121)	Acute 50 mg	↔	↔	–	↑	↑HR
	Mirabegron	(121)	Acute 200 mg	↑	↔	–	↑	↑HR, BP
	Mirabegron	(172)	4 wk 100 mg/d	↑	?	↔	↑	↑HR, BP
PPAR-γ	Pioglitazone	(634)	28 d 45 mg/d	↓	?	↑	↔	↔
	Pioglitazone	(681)	12 wk 30 mg/d	↔	?	↔	–	↑HR
TRPV1	Capsaicin	(715)	6 wk	↑	?	↔	–	↔
	Capsaicin	(714)	Acute 12 mg	↔	?	–	↑	-
TGR5	Chenodeoxycholic acid	(737)	2 d	↑	?	–	↑	-

Abbreviations: BP, blood pressure; CV, cardiovascular; HR, heart rate; PPAR, peroxisome proliferator–activated receptor; REE, resting energy expenditure; Tº, thermogenesis; TGR5, G protein–coupled bile acid receptor Gpbar1; TRPV1, transient receptor potential cation channel subfamily V member 1.

Knowledge Gaps and Perspectives for Human Investigations

Knowledge of the cellular biology and regulation of BAT metabolic responses and of the many overlapping mechanisms leading to its dysfunction in cardiometabolic diseases has grown exponentially over the past 15 years. The remarkable thermogenic activity of BAT caused by the unique mitochondrial uncoupling capacity of UCP1 and fueled by the rapid utilization of the brown adipocyte’s own TG content is similar in rodents and humans. In rodents, BAT thermogenic capacity is such that activation of this tissue can profoundly shift the caloric balance and significantly affect the development of obesity and IR. However, our current understanding of in vivo BAT metabolism in human rests heavily on the ¹⁸FDG PET method. This overreliance on BAT glucose metabolism has several important implications. First and foremost, BAT glucose metabolism is not a reliable marker of BAT energy expenditure because it has been shown to be dissociated from BAT thermogenic activity in several instances in preclinical and clinical studies. In the rare studies that have assessed both processes in the same participants and conditions, BAT glucose uptake is demonstrably lower, but BAT thermogenesis is not in obese, IR individuals. Emerging evidence suggests that reduction of BAT glucose uptake occurs very early during overfeeding conditions known to ultimately lead to the development of systemic IR. Second, the measurement of BAT volume, an important determinant of BAT thermogenic capacity, also relies exclusively on BAT glucose uptake in humans. This may obviously lead to an important underestimation of BAT volume in patients with IR. As a result, it is quite possible that the lower BAT metabolic activity and volume reported in association with obesity, T2D, visceral fat, and cardiovascular diseases is simply confounded by some degree of BAT IR. Reduced BAT glucose uptake may thus be an early marker, but not a cause, of cardiometabolic diseases. Our current incapacity to accurately measure BAT thermogenic capacity makes it impossible to quantify the role of BAT as a possible determinant of chronic energy balance regulation in humans.

BAT thermogenic activity and capacity can expand several-fold during chronic cold acclimation in humans. Furthermore, the reciprocal relationship observed between skeletal muscle shivering activity or metabolic uncoupling and BAT thermogenic activity during cold exposure also suggests some physiological role for BAT in NST in humans. Whether this BAT NST is enough to significantly affect whole-body thermogenesis is still unclear given the uncertainties discussed in the preceding paragraph on the total volume of BAT, and therefore on its thermogenic capacity. One intriguing avenue is the possibility that BAT may exert local thermogenic effects on adjacent neural structures and organs to substantially affect the body’s response to cold.

BAT produces a number of metabolites, cytokines, and hormones that likely exert autocrine and paracrine modulation of its metabolic response during chronic adaptation to cold, change in caloric balance, and other environmental conditions. However, the relatively small size of BAT and its nonexclusive secretion of most of these “batokines” make it unlikely that BAT plays a physiologically major role through the endocrine route. With regard to the control of BAT glucose metabolism, insulin stimulates BAT glucose uptake, but not thermogenesis. Evidence exists for a stimulatory role of thyroid hormones, secretin, and mineralocorticoid antagonists, and an inhibitory role of prolonged administration of glucocorticoids on BAT glucose uptake. Whether BAT thermogenesis is also modulated in these instances is unclear given that all these interventions may have other systemic effects, including changes in IS and SNS activity. BAT metabolism can be clearly stimulated by the β-3 adrenergic agonist mirabegron, but at doses that can activate the β-1 and β-2 adrenergic receptors and not at doses selective to β-3 agonism. Furthermore, this metabolic activation of BAT is accompanied by metabolic activation of WAT and the cardiovascular system in such a way that the stimulation of whole-body thermogenesis observed with β-adrenergic agonists cannot be specifically attributed to BAT. There is evidence from preclinical studies and human investigations for the integration and interdependence of WAT and BAT metabolic responses that need to be further investigated. Specific activation of BAT thermogenesis is not feasible using the pharmacological agents that have been tested thus far.

One important priority for the field of BAT research is the development of methods able to measure accurately and specifically BAT thermogenic capacity (ie, total volume of thermogenic activity). Large field-of-view PET allowing full neck, thoracic, and abdominal dynamic scanning with ¹¹C-acetate or ¹⁵O-O2 is one feasible approach, but these scanners and tracers are still not generally available. Three-dimensional mapping of the TG content shift in adipose tissues on cold stimulation is another interesting avenue, but this is technically challenging given the scattered anatomy and heterogeneous response of adipose tissue depots and the important effect of motion artifacts on this mapping. Furthermore, this method is limited by the current incapacity to quantify BAT TG/NEFA cycling in vivo. Compounding these issues is the fact that proper assessment of BAT thermogenesis also requires standardized and reproducible stimulation protocols and measurement of several other physiological responses (eg, shivering) that require complex experimental settings. At best, these methods can be applied only in small, proof-of-concept human mechanistic studies by a few groups of investigators around the globe. At present, BAT thermogenesis is thus not ready for testing in large cohorts or intervention studies that will be required to assess its role relative to the numerous other factors contributing to the development of cardiometabolic diseases. Only once this is achieved will we be able to design and execute studies to test whether activating BAT thermogenesis may affect cardiometabolic outcomes and ultimately change clinical practice (738, 739).

Abbreviations

12,13-diHOME

12,13-dihydroxy-9Z-octadecenoic acid

A_2A

adenosine receptor 2A

ACTH

adrenocorticotropins

ADK

adenosine kinase

ADP

adenosine 5′-diphosphate

ADR

adrenergic receptor

AgRP

agouti-related peptide

AMPK

adenosine 5′-monophosphate-activated protein kinase

ANGPTL4

angiopoietin-like 4

ARC

arcuate nucleus

AT

adipose tissue

ATGL

adipose tissue triglyceride lipase

ATP

adenosine 5′-triphosphate

BAT

brown adipose tissue

BMI

body mass index;

BMP4

bone morphogenic protein 4

BMP7

bone morphogenic protein 7

BMP8b

bone morphogenic protein 8b

BMS

brain melanocortin system

cAMP

cyclic adenosine 3’,5’-monophosphate

C/EBPs

CCAAT/enhancer proteins

ChERBP

carbohydrate-response element-binding protein

CIDEA

cell death activator

¹¹C-mHED

¹¹C-metahydroxyephedrine

CNS

central nervous system

CPT1b

carnitine palmitoyltransferase 1b

CPT2

carnitine palmitoyltransferase 2

CT

computed tomography

DAG

diacylglycerol

DGAT

diacylglycerol acyltransferase

DH

dorsal horn

DHA

docosahexaenoic acid

DMH/DHyA

dorsomedial hypothalamus, dorsomedial hypothalamus/dorsal hypothalamic area

EBF2

Early B-cell factor-2

EE

energy expenditure

ELOVL3

elongation of very long chain fatty acids protein 3

En1

homeobox protein engrailed 1

EPA

eicosapentaenoic acid

EPDR1

ependymin-related protein 1

ER

endoplasmic reticulum

ERK

extracellular signal-regulated kinase

FABP4

fatty acid binding protein 4

FA-CoA

fatty acyl-coenzyme A

¹⁸FDG

¹⁸F-fluorodeoxyglucose

FGF21

fibroblast growth factor 21

FoxO1

forkhead box protein O1

FSF

fat signal fraction

¹⁸FTHA

¹⁸F-fluoro-thia-heptadecanoic acid

GABA

γ-aminobutyric acid

GDF3

growth differentiation factor 3

GDP

guanine nucleotide diphosphate

GIP

glucose-dependent insulinotropic polypeptide

GLP1

glucagon-like peptide 1

GPR

G protein–coupled receptor

HDAC

histone deacetylase

HFD

high-fat diet

iBAT

interscapular brown adipose tissue

IS

insulin sensitivity

IL-6

interleukin 6

IML

intermediolateral

ingWAT

inguinal white adipose tissue

IR

insulin resistance;

IV

intravenous

iWAT

inguinal white adipose tissue

KO

knockout;

LH

lateral hypothalamus

LPB

lateral parabrachial nucleus

LPL

lipoprotein lipase

2-MAG

2-monoacylglycerol

MAOA

monoamine oxidase

MAPK

mitogen-activated protein kinase

MCR3

melanocortin receptor 3

MCR4

melanocortin receptor 4

miRNA

microRNA;

MRI

magnetic resonance imaging

mRNA

messenger RNA;

MRS

magnetic resonance spectroscopy

mTOR

mammalian target of rapamycin

Myf5

myogenic factor 5

NAD+/NADH

nicotinamide adenine dinucleotide, oxidized/reduced form

NDPK

nucleoside diphosphate kinase

NEFA

nonesterified fatty acids

NF-κB

nuclear factor κB

NOD

nucleotide-oligomerization domain-containing proteins

NPY

neuropeptide Y

NST

nonshivering thermogenesis

NTS

nucleus tractus solitarius

P2X

purinergic receptor 2X

P2Y

purinergic receptor 2Y

p38 MAPK

phospho-38 mitogen-activated protein kinase

Pax3

paired box protein 3

Pax7

paired box protein 7

PDE3E

phosphodiesterase 3B

PEPCK

phosphoenolpyruvate carboxykinase

PET

positron emission tomography

PDGFRA

platelet-derived growth factor receptor A

PGC1A

peroxisome proliferator-activated receptor gamma coactivator 1-alpha

PI3K

phosphoinositide 3-kinase

PKA

protein kinase A

PKC

protein kinase C

POA

preoptic area

POMC

proopiomelanocortin

PPAR-γ

peroxisome proliferator–activated receptor gamma

PRDM16

PRD1-BF1-RIZ1 homologous domain-containing 16

PRV

pseudorabies virus

Prx1

paired-related homeobox transcription factor 1

PVH

paraventricular hypothalamus

ROS

reactive oxygen species

RPa

raphe pallidus

Sim-1

single-minded homolog 1

SIRT1

sirtuin 1

SIRT5

sirtuin 5

SNS

sympathetic nervous system

T3

3,5,3′-triiodothyronine

TEE

total energy expenditure

TMEM26

transmembrane protein 26

TNAP

tissue-nonspecific alkaline phosphatase

TRPV1

transient receptor potential cation channel subfamily V member 1

TSPO

translocator protein

WAT

white adipose tissue

T2D

type 2 diabetes

TCA

tricarboxylic acid cycle

TGs

triglycerides

TGF

transforming growth factor

TGR5

G protein–coupled bile acid receptor Gpbar1

TLR

toll-like receptor

TNF-α

tumor necrosis factor alpha

TRIB1

tribbles pseudokinase 1

TRL

triglyceride-rich lipoprotein

TRPM8

transient receptor potential melastatin 8

TRPV1

transient receptor potential vanilloide 1

TSH

thyrotropin

UCP1

uncoupling protein 1

VEGFA

vascular endothelial growth factor A

VLDL

very low-density lipoprotein

VMH

ventromedial hypothalamus.

Financial Support

This work was supported by a Canadian Institutes of Health Research grant (No. 299962).

Disclosures

A.C.C. has received research funding from Eli Lilly (2019-ongoing) and Novo Nordisk (2021-ongoing) and consultation fees from HLS Therapeutics, Janssen Inc, Novartis Pharmaceuticals Canada Inc, and Novo Nordisk Canada Inc. None of these commercial relationships are directly relevant to the present review. D.P.B. is the GlaxoSmithKline Chair in Diabetes of the Université de Sherbrooke, created via a donation of GlaxoSmithKline to the university. GlaxoSmithKline is not involved in the research activities of the chair. F.H. and D.R. have nothing to disclose.