Obesity represents a major risk factor for the development of several of our most common medical conditions, including Type 2 diabetes, dyslipidaemia, non-alcoholic fatty liver, cardiovascular disease and even some cancers. Although increased fat mass is the main feature of obesity, not all fat depots are created equal. Adipocytes found in white adipose tissue contain a single large lipid droplet and play well-known roles in energy storage. By contrast, brown adipose tissue is specialized for thermogenic energy expenditure. Owing to its significant capacity to dissipate energy and regulate triacylglycerol (triglyceride) and glucose metabolism, and its demonstrated presence in adult humans, brown fat could be a potential target for the treatment of obesity and metabolic syndrome. Undoubtedly, fundamental knowledge about the formation of brown fat and regulation of its activity is imperatively needed to make such therapeutics possible. In the present review, we integrate the recent advancements on the regulation of brown fat formation and activity by developmental and hormonal signals in relation to its metabolic function.
BAT (brown adipose tissue), owing to its inherent capacity to convert excess food energy into thermal energy, has become a focus of biomedical research as a means to counter the spread of obesity [1,2]. This pathological form of overweight is associated with increased risk for numerous metabolic disorders, such as diabetes, cardiovascular complications, atherosclerosis and other dyslipidaemias . Owing to its involvement in many aspects of metabolic control, BAT could potentially help to ameliorate these metabolic perturbations . Excess energy from food is stored in the form of triacylglycerols (triglycerides) within numerous WAT (white adipose tissue) depots throughout the body . This organ possesses a high degree of plasticity, and a total body fat content ranging from less than 10% to more than 50% of total body weight has been observed . Sedentary lifestyles and the general availability of processed energy-dense food in modern societies have led to an alarming epidemic of metabolic diseases that represents a serious challenge for medical care systems worldwide.
Brown adipocytes are equipped with abundant mitochondria, the so-called powerhouses of the cell, and uniquely express UCP1 (uncoupling protein-1). UCP1 dissipates the proton gradient of the inner mitochondrial membrane that is formed as a result of oxidative phosphorylation of nutrients. This process, also known as thermogenesis, generates heat instead of ATP . In mouse models, the absence of UCP1 has been associated with increased body mass due to loss of brown fat-mediated thermogenesis [7,8]. The transcriptional control of brown adipocyte formation has been analysed extensively and has been reviewed in great detail elsewhere [9–14]. In brief, the transcription factors PPARγ (peroxisome-proliferator-activated receptor γ)  and the C/EBPs (CCAAT/enhancer-binding proteins) (specifically C/EBPα/β/δ)  have been established as essential parts of the transcriptional cascade that precedes the formation of mature fully differentiated adipocytes. It is generally accepted that these transcription factors direct the differentiation of both brown and white adipocytes. The transcriptional events specifying a brown adipogenic programme are regulated by the nuclear co-activator PGC1α (PPARγ co-activator 1α)  and the transcription factor PRDM16 (PR-domain-containing 16), which physically interacts with C/EBPβ [18,19]. Interestingly, recent evidence suggests that modulators of PPARγ-binding activity may specify brown adipogenic lineage transcriptional activity . The NAD-dependent deacetylase SIRT1 (sirtuin 1) could promote browning of white fat by deacetylating PPARγ. This modification is essential for recruitment of PRDM16 to PPARγ and initiation of the brown-fat-specific programme . Transcription factor EBF2 (early B-cell factor 2) acts to recruit PPARγ to genes that are selective for brown adipogenic processes, thereby also affecting the commitment of myogenic and white adipogenic progenitors to the brown adipose lineage . A critical component of these cascades is the activation of de novo mitochondrial biogenesis and the expression of functional components that are essential to thermogenesis, such as UCP1.
In recent years, along with the widely publicized re-discovery of metabolically active brown adipose tissue in adult humans [22–25], the common view that brown adipocytes represent a mere variation of the more prevalent white adipocytes has been challenged profoundly. In humans, the presence of BAT throughout life has been well documented for some decades, although an age-related decline in total BAT mass is evident . Expression of the key brown fat marker UCP1 was subsequently also observed in human BAT, further supporting the notion that human adults are in possession of functional brown fat depots . The first evidence for the role of the sympathetic nervous system and its neurotransmitter NE (norepinephrine/noradrenaline) in the process of brown fat activation in adult humans was gained from patients with phaeochromocytoma, a cancer that is accompanied by high levels of circulating NE and excessive formation of brown adipocytes in fat depots in the vicinity of, but also at distal locations from, the tumour [28,29].
A key finding of the more recent renaissance of BAT has been the observation that two developmentally distinct types of brown adipocytes exist in mammals: the classical, or cBAT (constitutive BAT) that arises during embryogenesis ; and the beige  or brite (brown-in-white)  adipose tissue that is recruited postnatally within WAT  or skeletal muscle , and which has therefore also been referred to as rBAT (recruitable BAT) . It is now clear that the developmental origin of cBAT is distinct from white fat and rBAT, placing its ancestry closer to skeletal muscle . Furthermore, the formation of rBAT appears to occur through two different pathways, depending on the fat depot: either transdifferentiation of mature white to brown adipocytes (and vice versa) , or through induction of differentiation of brown adipogenic progenitors [38–40].
These findings highlight an important theme of brown adipocyte biology which is characterized by a high degree of plasticity in brown adipose tissue formation. Secreted and circulating factors act as autocrine, paracrine and/or endocrine signals to regulate these processes on the levels of (brown) adipogenic lineage specification during embryogenesis, the differentiation of brown adipocytes, and the regulation of the thermogenic activity of mature brown adipocytes (summarized in Figure 1). The aim of the present review is to summarize the current state of the field, focusing on the brown adipogenic effects of secreted physiological signals that act locally or in circulation, and the role of BAT in systemic energy metabolism.
Distinct stages of brown adipocyte formation are regulated by instructive signals acting through endocrine, paracrine or autocrine mechanisms
EMBRYONIC DEVELOPMENT OF BAT
It is generally accepted that adipocytes, both brown and white, arise from progenitor cells of the embryonic mesoderm, the only exception being the minor fat depots of the skull, which arise from stem cells of the ectodermal neural crest . Adipocytes thus share a developmental origin with other mesodermal tissues, such as skeletal muscle, bone and other connective tissues [3,42]. In the mouse, clearly distinguishable pads of cBAT are first recognizable on or after day 15.5 of embryonic development (E15.5) which then rapidly expand until birth (Figure 2) . During this stage, developing cBAT appears to arise from a highly proliferative mass of fibroblast-like cells in the interscapular region [44,45]. Interestingly, lipid accumulation can only be observed in a small subset of the brown adipocytes during the early embryonic stages of cBAT formation, but increases markedly after birth (Figure 2). Since definite markers that label progenitors exclusively giving rise to BAT, but not WAT or other mesodermal tissues such as skeletal muscle, are currently lacking, most research in this area has so far been confined to analysis of generally adipogenic progenitors that can give rise to brown and white fat, regardless of their developmental origin. Accordingly, it is now established that adipogenic progenitor cells in brown and white fat depots are located within the mural cell compartment, also known as pericytes, anatomically located within the vasculature of the adipose tissue [46,47]. Whether this is also true for adipogenic cells within other tissues remains to be determined, although they appear to share surface marker characteristics with adipose tissue-resident progenitors, such as Sca (stem cell antigen) 1, CD29, CD34 and PDGFRα (platelet-derived growth factor receptor α) [39,48–50]. Despite their endothelium-associated location [46,47], their developmental ancestry from endothelial progenitors is still not entirely resolved, as lineage tracing methods have yielded inconclusive data, which may be due to differences in the methodological approaches [47,51].
Transcriptional control of adipogenic progenitor cell commitment has also been linked to the Zfps (zinc finger proteins) Zfp423 and Zfp521. Zfp423 acts as a positive regulator of progenitor cell commitment to an adipogenic lineage, at least in part by regulating expression of the master regulator of adipogenesis, PPARγ, and by amplifying the adipogenic effects of the BMP (bone morphogenetic protein) signalling pathway . Interestingly, Zfp423 acts as a definite marker for committed adipogenic cells isolated from different adipose tissue depots that reside within the adipose tissue vasculature, further corroborating the importance of this location as the niche for adipogenic progenitor cells . Zfp521 has recently been identified as a negative regulator of adipogenesis that appears to act upstream of its paralogue, Zfp423 . At the molecular level, Zfp521 inhibits commitment to the adipogenic lineage by binding and sequestering the pro-adipogenic factor Ebf1 .
BAT emerges during late stages of embryogenesis
Among the two distinct types of brown fat, cBAT has been studied in greater detail with regard to its developmental ancestry. These studies collectively indicate a close relationship of cBAT to skeletal muscle rather than WAT [11,30,32,56–58]. Accordingly, it was first observed that progenitors isolated from cBAT, but not WAT, express a myogenic gene expression signature . Specifically, cBAT progenitors express high mRNA levels of myogenic transcription factors such as myf5, myoD and myogenin. Expression levels of these genes decline as the cells differentiate into mature brown adipocytes. Subsequently, it was demonstrated that the transcription factor PRDM16 is a key regulator of a developmental switch between the myogenic and brown adipogenic cell fate lineages [18,30]. Overexpression of PRDM16 in a myoblast cell line led to a change of cell fate and preferential differentiation into UCP1-expressing brown adipocytes, whereas RNAi-mediated down-regulation of PRDM16 in brown pre-adipocytes resulted in the formation of mature myotubes. Similarly, formation of cBAT was attenuated in PRDM16-deficient mice, concomitant with reduced expression of common brown fat markers and increased expression of myogenic transcription factors . How this mechanism is involved in the early cell fate decision processes of embryonic development of cBAT remains to be investigated in molecular detail. Interestingly, this study also demonstrated the presence of a common lineage of progenitor cells that expresses the myogenic transcription factor Myf5 and that can give rise to interscapular cBAT as well as skeletal muscle . This notion is further supported by an independent study where Pax7-driven lineage tracing was used to demonstrate a common progenitor cell population in early embryonic stages that diverges into two distinct lineages, i.e. uniquely myogenic or brown adipogenic lineages, at about E11.5 . This is also consistent with earlier reports that are of a somewhat serendipitous nature, generally referring to an unspecified form of adipose tissue rather than distinguishing between brown or white fat in any detail. In a mouse model lacking myogenin, a general defect in skeletal muscle formation was observed. Interestingly, the knockout mice displayed increased amounts of adipose tissue in the interscapular region . Similarly, in animals with a genetic deletion of both myf5 and myoD, excessive amounts of adipose tissue in place of skeletal muscle were observed . Although the underlying mechanisms remain unclear, the inability of the common myo-adipogenic progenitor to undergo myogenic differentiation may promote those progenitor cells to follow a brown adipogenic lineage.
More recently, it was reported that some depots of WAT partially derive from the Myf5+ lineage . This study shows that progenitors derived from the interscapular and retroperitoneal white fat depots contain significant proportions (approximately 50%) of white adipocytes that derive from the Myf5+ lineage, whereas other, more classical depots of white fat, such as inguinal/subcutaneous and epididymal WAT, do not contain adipogenic cells of the Myf5+ lineage . This observation indicates a high degree of heterogeneity among the different populations of adipogenic cells and strongly suggests the presence of yet-to-be-identified local signals that act to specify individual cell fate. In rodents, the major depots of rBAT-containing WAT develop after birth . Initial studies have established that a distinct developmental origin of these two types of fat is likely, as postnatal white adipocytes do not arise from a brown adipocyte-like progenitor . However, not much is known about developmental distinction between rBAT and the surrounding white fat or skeletal muscle. While cBAT arises from the Myf5-expressing progenitors, it appears clear that muscle-resident rBAT derives from a Myf5− lineage [30,39,48]. Despite their anatomical localization within skeletal muscle, these cells are clearly non-myogenic [48,50,63], possess a high adipogenic potential and require inductive signals to differentiate into brown adipocytes . Less is known about the existence of progenitors for rBAT within WAT. Transdifferentiation of a phenotypically indistinguishable white adipocyte is likely to be a potential mechanism [37,64–66], although some evidence indicates that differentiation of progenitor cells into rBAT following inductive cues occurs . Interestingly, depot-specific differences have been reported on whether transdifferentiation or direct proliferative expansion and differentiation of progenitors into rBAT takes place . Given the high degree of plasticity within adipose tissue depots, it seems likely that both processes may exist in parallel to some extent. Importantly, both processes are initiated by inductive cues either through central pathways or from locally secreted or endocrine factors.
Recent evidence has established the intriguing idea that a subset of white adipocytes may have brown adipogenic potential, suggesting that distinct subpopulations of beige/brite adipocytes and even pre-adipocytes occur in WATs . A recent study by Wu et al.  analysed clonal adipogenic progenitor cell lines and demonstrated that a subpopulation of cells possesses the potential to differentiate into beige adipocytes. The beige cells express a low level of UCP1, but it can be greatly increased by forskolin. Whether this process is based on lineage determination steps that occur during embryogenesis remains to be determined. The authors further identify the cell-surface markers CD137 and TMEM26 (transmembrane protein 26), among other markers, that may help to define and prospectively isolate this subpopulation from different adipose tissues.
MORPHOGENS: DEVELOPMENTAL FATE REGULATION OF BROWN ADIPOSE PROGENITORS
Morphogens regulate the developmental patterning during embryogenesis. Concentration gradients of certain morphogens and other secreted signals may also be involved in the lineage commitment of brown adipogenic progenitors during various developmental stages . Accordingly, morphogenic signals of (i) the Wnt (named after the Wingless and INT proteins in Drosophila), (ii) the BMP, (iii) the FGF (fibroblast growth factor) and (iv) the Hedgehog signalling pathways have been implicated in the formation of brown adipocytes and may do so at early embryonic stages when multipotent populations of mesodermal progenitors occur (Figure 1) [68–73]. During the early stages of embryogenesis, a subset of cells within the central dermomyotome expresses the homeobox transcription factor En1 (engrailed-1), and their developmental fate is regulated by activation of Wnt signalling. Using developmental lineage tracing, it was shown that the cells of the early embryonic dermomyotome are En1+ and give rise to cBAT in addition to skeletal muscle and dermis, indicating that a tripotent population of En1+ progenitor exists during early stages of embryogenesis. In addition, activation of the Wnt tranducer β-catenin was required to induce expression of dermo-1 and subsequent dermal differentiation of a subset of En1+ cells located close to the surface ectoderm . Whether the Wnt signalling cascade also plays a functional role in the fate specification of the residual En1+ lineage into myogenic compared with brown adipogenic cells in the developing embryo is not known. However, the En1 study suggests that activation of Wnt signalling may inhibit the development of myogenic progenitors, thereby also reducing the frequency of brown adipogenic progenitors. This is also consistent with the notion that Wnt signalling negatively affects (brown) adipogenesis postnatally [70,74].
Emerging evidence has implicated BMP signalling in the control of brown (and white) adipogenesis [67,73,75]. The ligands of this signalling pathway, the so-called BMPs, were initially identified due to their ability to induce bone formation [76,77]. Since then, it has become clear that BMPs function as morphogens that direct the development of multiple organ systems during embryogenesis . Although previous findings have implicated a general role for BMPs in adipogenic processes, their role in early embryonic lineage commitment of different adipogenic cell populations is less clear. Several studies have demonstrated that BMP2 and BMP4 are involved in the commitment of multipotent progenitors to the white adipogenic lineage [75,79–81]. However, a recent study has also implicated BMP4 in the browning of mature white adipocytes , suggesting that individual BMPs may have distinct windows of activity to regulate brown adipocyte physiology. Importantly, it was recognized that BMP7 induces commitment of multipotent mesenchymal progenitor cells to the brown adipogenic lineage and drives differentiation of brown pre-adipocytes as compared with other BMPs that are rather white adipogenic . Animals completely lacking BMP7 die shortly after birth due to kidney failure . Interestingly, one of the major defects observed in this mouse model is a marked reduction in cBAT mass and function, indicating that BMP7 is a critical promoter of embryonic cBAT development . The BMP signalling cascade, and particularly its downstream signalling through the type 1A BMP receptor BMPR1A, is a critical regulator of cBAT development. Ablation of BMPR1A in the Myf5+ lineage resulted in a marked paucity of cBAT due to impaired expansion of brown adipogenic progenitors during embryonic development, further highlighting the critical importance of BMP signalling for brown fat development . Further studies are needed to establish the molecular mechanisms involved in the BMP-mediated regulation of progenitor cell commitment to the cBAT lineage.
Like BMPs, other members of the TGFβ (transforming growth factor β) family have also been implicated in the regulation of BAT. Circulating levels of TGFβ, for instance, correlated with obesity, and deletion of Smad3 promoted the browning of WAT, suggesting that some signalling of the TGFβ family may negatively affect brown adipocyte development . Emerging evidence further supports a differential role for some members of the TGFβ family. Administration of a soluble form of the type IIB activin receptor, where the extracellular domain of this receptor was coupled to a human IgG Fc domain (ActRIIB–Fc), rescued the negative effects of a high-fat diet on cBAT and additionally resulted in marked recruitment of rBAT in visceral WAT . Mechanistically, the authors demonstrate that high-affinity ligands of ActRIIB, the GDFs (growth and differentiation factors) 3, 8 and 11, as well as activins A and B, are negative regulators of brown adipogenesis and are depleted following administration of ActRIIB–Fc. Low-affinity ligands, on the other hand, such as BMP7, GDF5 and GDF7, were pro-brown adipogenic and were less affected by ActRIIB, thus permitting the emergence of rBAT within WAT by promoting expression of PGC1α and UCP1 . Taken together, these data support the notion that members of the TGFβ superfamily have very distinct regulatory effects on brown adipogenesis and modulation of the profile of these circulating ligands may prove a feasible tool to affect brown adipogenesis.
Among the FGF proteins, FGF16 and FGF19 have been implicated in the regulation of embryonic cBAT development. It was found that FGF16 is predominantly expressed in developing cBAT in rat embryos, where a high proliferation rate was observed that was regulated by FGF16, acting through its receptor FGFR4 (FGF receptor 4) . Interestingly, exposure to cold decreased FGF16 expression in adult cBAT, suggesting that the role of FGF16 in brown adipogenesis is restricted to embryonic developmental processes. Similarly, increased expression of FGF19 resulted in increased energy expenditure and decreased body fat content that could be related to increased cBAT mass, suggesting that this FGF might also be involved in the development of cBAT . Of note, FGF19, like FGF21, can also induce thermogenic activity of adult cBAT , an effect that will be discussed in more detail later in the present review.
Lastly, morphogens of the hedgehog signalling pathway have been linked to white versus brown adipocyte lineage determination. While activation of the hedgehog signalling cascade leads to activation of brown adipose metabolism , a genome-wide screening connects hedgehog activity to the lineage determination of brown versus white adipogenic lineages . It remains to be determined whether this could also affect the progenitor composition in WAT and whether brown adipogenic processes can be altered by modulation of hedgehog signalling. Many other secreted proteins and circulating factors are likely to be involved in the embryonic processes of cBAT formation. Oftentimes, these signalling proteins are involved in several stages of brown adipogenesis, as is the case for BMP7, and a clear distinction between these processes is not always possible. External signals that regulate the differentiation/maturation of brown adipocytes will be discussed in more detail later in the present review. In summary, much remains to be determined with regards to the mechanisms that determine an adipogenic fate for mesenchymal progenitor cells. Additionally, the identity of putative rBAT progenitor cells in WAT and skeletal muscle warrants further investigation, as this type of BAT could well be an important contributor of brown adipocytes in adult humans.
REGULATION OF BAT ACTIVITY: CROSS-TALK WITH OTHER ORGANS BY SECRETED FACTORS
Numerous studies have implicated increased formation and/or activity of both types of BAT in increased resistance to diet-induced obesity. Although activation of cBAT activity may be one mechanism to achieve this goal, it is the inducible formation of rBAT in WAT that has garnered extensive scientific interest in recent years. Genetic manipulation of the transcription factors that control brown adipocyte differentiation is of only limited relevance for the treatment of human obesity. The discovery and analysis of brown-adipogenic signals, either present in the circulation or acting as locally secreted paracrine factors, has therefore become a key aspect of brown fat research. The identification of such factors is of particular interest as the first generation of agents to target BAT as a treatment for obesity, the ADRB3 (β3 adrenergic receptor) agonists, have not fulfilled their initial promise in clinical trials . This section will therefore address recent advances and discoveries of novel physiological factors that may affect brown fat formation and activity, their underlying molecular mechanisms, and their physiological sources. An emerging theme is the notion that signals coming from multiple organs and tissue types are integrated to control the total mass and thermogenic activity of BAT (summarized in Figure 3). Each organ produces a specific signal (or a set of signals) that contributes to the overall regulation of BAT activity. A greater understanding of the cross-talk between these individual sources of brown adipogenic and/or thermogenic factors will help to develop new therapeutic approaches that may prove effective for treating human obesity.
Brown adipocyte physiology is controlled by cross-talk mechanisms through secreted factors from multiple anatomical sites
Neuronal control of BAT, both on the central and the peripheral levels, has been well recognized as the key mechanism regulating the thermogenic activity of mature BAT. The SNS (sympathetic nervous system) is a critical controller of thermogenesis by acting on ADRB3, the adrenergic receptor subtype most abundantly expressed in adipocytes [89–91]. By SNS-mediated activation, the heat output of cBAT can be quickly adapted in response to rapid changes in ambient temperature [92,93]. At the same time, sympathetic input to WAT, for instance by cold exposure or by chemical agonism of ADRB3, results in the browning of WAT [37,94–97], a process that apparently replaces the more short-term measure of cBAT activation . Several neuropeptides and hormones have been linked to the thermogenic activity of BAT. BDNF (brain-derived neurotrophic factor) has recently been implicated in hypothalamic control of energy expenditure through BAT. Administration of BDNF in different areas of the hypothalamus, namely the ventromedial and the paraventricular nuclei, resulted in increased energy expenditure that could at least in part be attributed to increased BAT activity [98,99]. More recently, it was also discovered that BDNF acts on the browning of WAT. Animals housed in a so-called ‘enriched environment’ that provides stimulatory cues to different brain regions displayed greater browning of white fat and resistance to diet-induced obesity. This effect was dependent on BDNF . Interestingly, and considering the specific role of BMP signalling in the developmental aspects of brown adipogenesis [39,67,73,101], it was recently recognized that BMPs may also act on a central level to regulate energy homoeostasis. In this line, we were able to demonstrate that systemic administration of BMP7 resulted in activation of systemic energy expenditure and weight loss . Importantly, this involved the CNS (central nervous system): intracerebroventricular administration of BMP7 resulted in an mTOR (mammalian target of rapamycin)-dependent inhibition of food intake, altogether establishing the importance of BMP7 for systemic energy balance. Likewise, hypothalamic BMP8b acts as a positive regulator of thermogenesis in BAT. Expression of BMP8b was responsive to metabolic cues such as fasting/re-feeding regimens, that down- and up-regulated BMP8b mRNA levels respectively. Loss of BMP8b markedly reduced activation of AMPK (AMP-activated protein kinase) signalling, thereby also reducing thermogenesis and overall metabolic rate. Central administration of BMP8b had the opposite effects, resulting in increased sympathetic tone to and activation of cBAT, altogether providing compelling evidence that, in addition to BMP7, BMP8b also affects the thermogenic capacity of BAT . In the same line, orexin regulates thermogenesis in BAT by inducing sympathetic activation of cBAT [104–107]. Importantly, it appears that orexin may also act in the periphery : loss of this neuropeptide resulted in decreased proliferation of brown pre-adipocytes and enhanced expression of the brown adipogenic transcription factors PRDM16  and PGC1α . Other neuropeptides and hormones play an important role in the regulation of brown fat activity. Their specific activities and roles in different brain areas have been reviewed in more detail elsewhere [93,108].
As summarized in Figure 3, several organs are known to secrete factors that can be linked to brown fat formation and thermogenic activity. Although not much is known about the molecular details, BAT itself is a source of several secreted signals, such as BMP7 (Y.-H. Tseng and T.J. Schulz, unpublished work), BMP8b , FGF21  and IL (interleukin)-6 , whose effects on brown adipocyte physiology will be discussed later in the present review. Although BAT is also a source of typical adipokines, such as leptin or adiponectin, the contribution from different types of brown adipocytes to the systemic circulating levels of these proteins is unclear, but probably much smaller than from the more prevalent WAT.
One of the best-established mechanisms to control brown fat activity is through thyroid hormones, most notably T3, which is the most active thyroid hormone. Its precursor, T4, is taken up by brown adipocytes and converted into the active T3 by DIO2 (type 2 deiodinase) following sympathetic stimulation . Production and release of T3 is under tight control by the CNS, specifically the so-called hypothalamic/pituitary/thyroid axis, and a close synergistic interaction of T3 with adrenergic activation of brown fat has been closely analysed . Along with T3, the pancreatic hormone insulin is a major regulator of brown adipocyte formation and function. The role of the insulin signalling cascade in brown adipogenesis is well established and loss of the downstream signalling components has detrimental effects on BAT function  that can also lead to a diabetic phenotype . Although T3 and insulin are classical and well-known mediators of brown adipogenesis, they have pleiotropic effects on multiple organ systems in the body. Therefore the identification of novel circulating factors that act more specifically on brown fat is highly desirable.
FGF21 is produced and secreted by the liver, and also by cBAT after sympathetic activation [109,115,116]. Interestingly, this circulating factor is involved in multiple metabolic processes and appears to have beneficial effects on whole-body metabolic balance [117,118]. Accordingly, it was also found that FGF21, by acting through the transcription factor PGC1α, induces browning of white fat in an autocrine/paracrine manner . Taken together, these observations indicate that FGF21 derived from cBAT may also act as a brown adipose-derived adipokine that regulates the recruitment of brown adipocytes in white fat as an adaptive mechanism to prolonged metabolic stresses, such as cold exposure. Moreover, FGF21 secreted from brown adipocytes may exert direct metabolic effects, thereby providing a novel mode of communication of BAT with other peripheral tissues that may not necessarily involve the CNS. Cardiac NPs (natriuretic peptides), in both the atrial (ANP) and ventricular (BNP) forms, can induce browning of WAT, as signified by elevated expression of BAT markers, mitochondrial biogenesis and oxygen consumption in cultured adipocytes treated with either peptide. In mice, cold exposure resulted in increased circulating levels of NPs, and direct infusion of BNP induced energy expenditure of brown and white fat. Loss of the NP clearance receptor, NPRC, similarly resulted in decreased fat pad size of cBAT and WAT in mice, whereas expression of thermogenic genes was increased, suggesting that fat pad weight loss was due to increased oxidation of stored lipids . Secreted from skeletal muscle, the myokine irisin has been identified as an important mediator of white fat browning . Interestingly, irisin secretion was induced by exercise as well as overexpression of PGC1α, which mimics the physiological effects of exercise. Administration of irisin protected against diet-induced obesity and the development of glucose intolerance, altogether establishing the myokine irisin as a skeletal muscle-derived contributor to systemic control of energy expenditure and thermoregulation. An autocrine regulatory mechanism of browning was recently identified in white fat depots. Specifically, two studies identified COX2 (cyclo-oxygenase 2) as a critical regulator of this process within white fat. COX2 is a key enzyme in prostaglandin biosynthesis, and, accordingly, treatment of mice with prostaglandins resulted in increased formation of rBAT [122,123].
In summary, emerging evidence suggests that multiple cell types produce a wide variety of different signals that act as positive regulators of thermoregulation. These involve central as well as peripheral mechanisms and are integrated by brown adipocytes to control not only temperature homoeostasis, but also metabolic regulation. In this line, we were recently able to demonstrate that different depots of brown adipocytes are subject to systemically active cross-talk mechanisms that control overall thermoregulation. Specifically, loss of cBAT by genetic ablation of BMPR1A in the Myf5+ lineage or surgical denervation resulted in increased browning of white fat. Thus these findings highlight the existence of a sophisticated regulatory mechanism in the body that involves cBAT-to-brain and brain-to-WAT circuits to control BAT activity .
THE ROLE OF BAT-MEDIATED THERMOGENESIS IN ANTI-OBESITY
Although the main function of BAT is thermoregulation, an important aspect is the associated role as a negative regulator of obesity. Two major signals for non-shivering thermogenesis by BAT have been described: cold exposure and adaptive/facultative DIT (diet-induced thermogenesis), the latter representing the portion of DIT that is not related to the essential energy-consuming processes of food metabolization [124,125]. The mechanisms regulating this type of DIT in humans still warrant further investigation, but evidence suggests that BAT-dependent DIT may act through enhanced adrenergic activation of BAT and depends on nutrient type [126–128]. In rodents, it was initially recognized that the major mediator of non-shivering thermogenesis is NE, which is secreted from the SNS upon exposure to cold [129,130], thereby activating the adrenergic system. Later on, this discovery led to the development of specific agonists of the major NE receptor in adipose tissues, ADRB3 [131,132]. Unfortunately, after initially promising studies performed in rodent models, it was soon recognized that this effect was negligible in adult humans, which was either due to reduced ligand-binding ability or bioavailability to human ADRB3 [133,134]. It is generally accepted that the thermogenic activity of UCP1 is absolutely required for normal brown fat function. Surprisingly, studies conducted in mouse models with partial or complete UCP1 deficiency have yielded somewhat inconclusive results: in some models, deletion of UCP1 is accompanied not only by increased cold sensitivity, but also obesity [135,136]. Other studies, however, were able to corroborate cold-sensitivity, but found no effect of UCP1 depletion on obesity [7,135,137]. This surprising lack of obesity in UCP1-deficient mice has been explained by a chronic activation of metabolism in mice housed under standard conditions: at ambient temperatures of approximately 21°C, mice are under thermal stress, meaning that their basic metabolic rate is elevated in order to maintain normal body temperatures. This effect alone is sufficient to correct for the obesogenic effect of UCP1 ablation. In contrast, when these mice were maintained at thermoneutrality, i.e. an ambient temperature where increased basic metabolic rate is no longer required to maintain body temperature, loss of UCP1 resulted in a striking obesity phenotype . These findings suggest that regulation of systemic thermogenesis is not exclusively controlled by the heat-generating activity of BAT. For instance, it was described that SNS-controlled non-shivering thermogenesis by skeletal muscle may play a significant role in overall thermoregulation . More importantly, the systemic control of these processes is ensured by a tight web of control mechanisms that involves different tissue types and that are likely to be integrated by central mechanisms involving regulation of SNS output . In summary, the role of BAT-mediated thermogenesis in counteracting obesity appears well-established, at least in rodents. The discovery that systemic compensatory mechanisms are in place that regulate the overall mass of metabolically active BAT and its metabolic output warrants further investigation to determine whether pharmacological approaches may be used to affect the delicate balance of this system.
THE ROLE OF BAT IN METABOLIC DISEASE
While the natural focus of brown fat research lies on its capabilities as an anti-obesity target, emerging evidence suggests that brown adipocytes may be critical regulators of general metabolic homoeostasis, affecting a number of metabolic complications. Specifically, it has been well-documented that some mouse strains are more sensitive to diet-induced obesity and insulin resistance. This has, at least in part, been attributed to elevated recruitment of brown adipocytes in skeletal muscle and white fat [33,34]. Interestingly, the determination of rBAT in different strains of mice is determined by genetic components , and it is currently unclear to what extent circulating and/or locally secreted factors are involved in the strain-specific differences. A study from our own laboratory suggests that cell-autonomous properties play an important role: pre-adipocytes isolated from muscle and white fat of obesity-prone mice, specifically the mouse strain C57Bl/6, had lower brown adipogenic potential compared with cells isolated from the relatively obesity-resistant strains, 129-S1 and A/J , although these observations do not rule out autocrine mechanisms.
As discussed above, loss of insulin signalling in BAT resulted in impaired insulin secretion from pancreatic β-cells, thereby affecting glucose homoeostasis . These findings strongly suggest the existence of a yet unidentified mode of interaction between brown adipocytes and insulin-producing pancreatic β-cells that remains to be characterized. These findings were extended in a model of increased BAT mass by transplantation of cBAT pads into the visceral cavity of mice. This resulted in a marked improvement of glucose homoeostasis that is, surprisingly, dependent on the cytokine IL-6 and secretion of FGF21 from the transplanted fat pad , whose beneficial metabolic properties are well established . When cBAT was transplanted from IL-6−/− mice, the beneficial effects on glucose metabolism were no longer present. An additional function of BAT may be the regulation of lipoprotein homoeostasis, and specifically the clearance of TRLs (triacylglycerol-rich lipoproteins). A recent study demonstrates that after exposure to cold, BAT is involved in the rapid clearance of TRLs that are used as fuel for thermogenesis. This process was dependent on the activities of LPL (lipoprotein lipase) and the transmembrane receptor CD36, which is involved in fatty acid uptake from the blood stream . Taken together, these findings may have uncovered an additional mechanism by which BAT may be used to improve metabolic health, apart from its well-known function in obesity control.
HUMAN BROWN FAT
Brown fat was once thought to have a functional role only in rodents and human infants; however, recent studies using 18F-labelled 2-deoxy-glucose PET (positron emission tomography) (18FDG-PET) in combination with CT (computed tomography) clearly demonstrated the presence of metabolically active brown fat in adult humans [22–25,141–144]. These findings rekindled research on human brown fat. Like rodent brown fat, human BAT also has extensive neuronal and vascular connections and abundant mitochondria. Interestingly, human BAT is an amalgam of brown and white adipocytes. At the physiological level, human brown fat displays differential rates of glucose uptake in response to cold compared with insulin stimulation . Cold-activated human brown fat takes up more glucose per gram of tissue than insulin-stimulated skeletal muscle, suggesting that human brown fat could contribute to glucose homoeostasis. With regard to detection of human brown fat, although 18FDG-PET/CT has been proven to be a useful tool for monitoring BAT in humans, it still has limitations in terms of sensitivity and reflection of BAT functionality. In particular, increased FDG uptake may not be a direct measurement of brown fat activity. Thus several imaging methods have been used to monitor brown fat mass and activity in humans. Using dynamic 15O PET imaging, oxidative metabolism of brown fat was found to be highly variable in individuals with detectable BAT determined by 18FDG-PET . PET studies using 11C-acetate and 18F-fluoro-thiaheptadecanoic acid (a fatty acid tracer) have been applied in human volunteers to determine BAT oxidative capacity and non-esterified fatty acid uptake respectively, and these studies confirmed that brown fat serves as an important tissue for non-shivering thermogenesis in humans . In addition to PET, chemical-shift water-fat MRI (magnetic resonance imaging) has been used to characterize human brown fat due to its unique cellular structure, fat content and vascularization .
More recently, a new focus in human brown fat research is to determine whether humans possess both classical and recruited types of brown fat cells as seen in rodents. Similar to rodents, human infants indeed retain classical brown adipocytes in the interscapular region , but these brown fat cells disappear with age [26,150]. Using molecular markers originally identified in rodent cells, supraclavicular adipose depots in adult humans and some children are found to contain a significant portion of recruited beige cells [40,151]. Interestingly, deep neck fat from adult humans displays molecular signatures of classical brown fat and possesses high levels of oxygen consumption rate . These findings suggest that the properties of human fat vary from location to location, hence detailed mapping of human brown fat from different anatomical sites in conjunction with functional characterization are needed to better understand human brown fat.
Certainly, increasing energy expenditure by targeting brown fat has become an attractive way to combat obesity. New knowledge gained from both the human and rodent systems has advanced our understanding on brown adipose biology and provided important insight to the role of brown fat in metabolic regulation of physiological and pathological conditions. These efforts have opened up exciting new avenues for developing new anti-obesity therapies that may also help to treat obesity-related complications. However, although the findings to date suggest promising opportunities for developing anti-obesity treatments through increasing brown fat-mediated energy expenditure, the safety issue needs to be considered. In particular, the cardiovascular effect of any potential drug, especially in subjects with chronic cardiovascular diseases, needs to be closely evaluated.
β3 adrenergic receptor
brown adipose tissue
brain-derived neurotrophic factor
bone morphogenetic protein
central nervous system
early B-cell factor
18F-labelled 2-deoxy-glucose positron emission tomography
fibroblast growth factor
growth and differentiation factor
positron emission tomography
PPARγ co-activator 1α
peroxisome-proliferator-activated receptor γ
sympathetic nervous system
transforming growth factor β
white adipose tissue
zinc finger protein
This work was supported, in part, by the National Institutes of Health [grant number R01 DK077097 (to Y.-H.T.)], the Joslin Diabetes Center's Diabetes Research Center [grant number P30 DK036836 (from the National Institute of Diabetes and Digestive and Kidney Disease)], the American Diabetes Association and the Harvard Stem Cell Institute (to Y.-H.T.). T.J.S. was supported by the Mary K. Iacocca Foundation, the German Research Foundation [grant numbers SCHU 2445/1-1 and SCHU 2445/2-1] and the European Research Council [grant number ERC-StG-311082].