Urolithin A exerts antiobesity effects through enhancing adipose tissue thermogenesis in mice

Obesity leads to multiple health problems, including diabetes, fatty liver, and even cancer. Here, we report that urolithin A (UA), a gut-microflora–derived metabolite of pomegranate ellagitannins (ETs), prevents diet-induced obesity and metabolic dysfunctions in mice without causing adverse effects. UA treatment increases energy expenditure (EE) by enhancing thermogenesis in brown adipose tissue (BAT) and inducing browning of white adipose tissue (WAT). Mechanistically, UA-mediated increased thermogenesis is caused by an elevation of triiodothyronine (T3) levels in BAT and inguinal fat depots. This is also confirmed in UA-treated white and brown adipocytes. Consistent with this mechanism, UA loses its beneficial effects on activation of BAT, browning of white fat, body weight control, and glucose homeostasis when thyroid hormone (TH) production is blocked by its inhibitor, propylthiouracil (PTU). Conversely, administration of exogenous tetraiodothyronine (T4) to PTU-treated mice restores UA-induced activation of BAT and browning of white fat and its preventive role on high-fat diet (HFD)-induced weight gain. Together, these results suggest that UA is a potent antiobesity agent with potential for human clinical applications.


Introduction
Natural products are a rich resource for the identification of bioactive compounds with beneficial pharmacological activities [1,2]. Pomegranate fruit, red berries, and walnuts are highly valued as "healthy" food and have been consumed by humans from hunter-gatherer days until the present [3,4]. Among other bioactive compounds, ellagitannins (ETs) have been shown to facilitate many of the beneficial effects of pomegranate. ETs exhibit antioxidant and antiinflammatory functions [5][6][7][8][9][10]. Once ingested, ETs are quickly metabolized in the intestinal tract to ellagic acid (EA) [8,11,12]. Successive lactone hydrolysis and decarboxylation, as well Together, these results suggest that UA could efficiently prevent both HFD-induced and genetic obesity in mice.
Next, to further compare the preventive effect of UA on body weight control with an FDAapproved drug for obesity management, orlistat [32], 3 groups of male C57BL/6 mice kept on an HFD were treated with either 30 mg/kg/day of UA or 15 mg/kg/day of orlistat or vehicle for a testing period of 6 weeks. Compared to vehicle controls, both UA-treated mice and orlistattreated mice showed significantly lower body weight after 2 weeks of treatment (S3A Fig). No significant difference in food intake was observed between the UA-treated group and the orlistat-treated group (S3B Fig), suggesting that the preventive effect of UA on HFD-induced body weight gain is similar to that of orlistat.
Given the safety of the compound, treated C57BL/6 mice tolerated UA well during the 10 weeks of compound administration with no apparent side effects. Plasma activities of the liver transaminases alanine aminotransferase (ALT) (−22%) and aspartate aminotransferase (AST) (−26.8%) were lower in UA-treated mice than that of controls (S4A and

UA prevents HFD-induced insulin resistance and glucose intolerance in mice
Next, we assessed the effects of UA on glucose metabolism. Plasma glucose concentrations (−16.2% and −29.4% after 5 and 10 weeks of treatment, respectively) and insulin (−16.7% and −37.1% after 5 and 10 weeks of treatment, respectively) were lower in UA-treated mice than those in controls (Fig 1F and 1G). Based on levels of fasting glucose and insulin, the calculated homeostasis model assessment-estimated insulin resistance (HOMA-IR) index was lower after 5 weeks (−16.2%) and 10 weeks (−29.4%) in UA-treated mice than that of the vehicle group ( Fig 1H). Insulin sensitivity and glucose tolerance were further evaluated using the insulin tolerance test (ITT) and glucose tolerance test (GTT), showing improved insulin sensitivity ( Fig  1I and 1J) and glucose tolerance (Fig 1K and 1L) in UA-treated mice over controls. Finally, insulin sensitivity was assessed by hyperinsulinemic-euglycemic clamp assay. Higher levels of glucose infusion rate (GIR) were found in the UA-treated mice than controls during the clamp ( Fig 1M). This increase in GIR was caused by elevated whole-body glucose uptake as well as the decreased endogenous hepatic glucose production during the clamp (Fig 1N and 1O). Consistently, UA-treated mice showed higher levels of insulin-stimulated glucose uptake in gastrocnemius muscle, BAT, and WAT than controls during the clamp (Fig 1P). Note that similar levels of glucose were observed between control and UA-treated mice under normal diet feeding (S1E Fig), consistent with their comparable body weight. Altogether, these results showed that UA protects mice from HFD-induced hyperglycemia and insulin resistance.

UA prevents HFD-induced systemic inflammation and liver steatosis
Besides insulin resistance and glucose intolerance, systemic inflammation and fatty liver are 2 common metabolic manifestations that often accompany obesity [33]. To detect whether UA attenuates HFD-induced inflammation in mice, levels of plasma tumor necrosis factor-α (TNF-α) and interleukin 6 (IL-6) were measured. Lower concentrations of TNF-α and IL-6 were shown in UA-treated mice than in controls (S5A and S5B Fig)

UA reverses HFD-induced obesity and dysfunctional glucose homeostasis in mice
To determine whether UA, in addition to its preventive action on obesity and hyperglycemia, also has therapeutic potential, we used obese mice that were kept on an HFD for 8 weeks prior to UA or vehicle treatment. UA treatment was initiated when the obese mice had a total body weight of approximately 35 g and continued for 10 weeks (  S6P Fig, respectively). Together, these results showed that UA reverses HFD-induced obesity and dysfunctional glucose homeostasis in mice, possessing therapeutic potential for human obesity treatment.

UA increases EE in mice
To explore the cause of lower body mass in UA-treated mice, EE was measured by indirect calorimetry. We found significantly higher oxygen consumption (Fig 2A) as well as EE (Fig 2B-2D) in UA-treated mice than that of controls during both the light and dark phases, while locomotor activity was slightly higher but not statistically different between UA-treated and control mice during both light (p = 0.06) and dark phases ( Fig 2E). The respiratory exchange ratio (RER), a parameter reflecting the relative use of carbohydrate versus lipid as source of energy, was markedly lower in UA-treated mice than in controls (Fig 2F), suggesting that UA could shift energy source towards lipid. We noticed that UA-treated mice exhibited about 0.7˚C higher body temperature than that of controls at ambient temperature ( Fig 2G). This temperature difference was also confirmed by an infrared camera (Fig 2H). In line with this, UA-treated mice maintained their body temperature better than controls (Fig 2I) when placed at 4˚C. Higher oxygen consumption was shown both in the steady state ( Fig 2J) as well as in the norepinephrine (NE)-stimulated non-steady state (Fig 2K) of UA-treated mice than that of controls at 4˚C, indicating that UA was able to activate nonshivering thermogenesis upon cold stress. Collectively, these results show that UA enhances EE by promoting adipose tissue thermogenesis in mice, which may lead to the observed lower body weight in UA-treated mice. However, the opposite may also be true: UA treatment could cause a lean phenotype first, which could subsequently lead to increased thermogenesis because in general, a lean phenotype is associated with increased thermogenic adipocyte function. To test it, thermogenesis was measured in mice received UA or vehicle for 7 days (S7 Fig), when the body weight was similar between 2 groups of mice (control, 21.85 ± 0.54 g versus UA, 21.82 ± 0.85 g). A significantly higher oxygen consumption was shown in UA-treated mice than in controls (S7 Fig), suggesting that functional thermogenesis happens earlier than body weight changes.

UA promotes BAT thermogenesis and induces inguinal white fat browning in an adipocyte-autonomous manner
To further study the thermogenic effect of UA, tissue weights of BAT, inguinal (iWAT), and epididymal WAT (eWAT) were measured. All 3 adipose depots weighed less (BAT, −37.4%; iWAT, −57.2%; eWAT, −46.1%) in UA-treated mice than in controls ( Fig 3A). Histological analysis revealed smaller adipocyte sizes in BAT, iWAT, and eWAT of UA-treated mice compared with controls ( Fig 3B and S8 Fig), suggesting that UA treatment causes resistance to HFD-induced adipose tissue hypertrophy. Immunohistochemistry staining of Uncoupling protein 1 (UCP-1) was higher in BAT and iWAT of UA-treated mice than of controls ( Fig  3C). BAT (2.5-fold), iWAT (2.1-fold), and eWAT (1.4-fold) exhibited an increased amount of mitochondrial DNA (mtDNA) (Fig 3D). Furthermore, mRNA expression levels of mitochondrial thermogenic genes were significantly higher in the 3 fat depots of UA-treated mice than those of controls ( Fig 3E). Finally, BAT and iWAT of UA-treated mice had higher protein levels of UCP-1 than untreated controls (Fig 3F). Taken together, these data indicate that the low weight gain in UA-treated mice is due to the increased mitochondrial biogenesis and activated thermogenic program in BAT and iWAT.

PLOS BIOLOGY
Urolithin A, a novel antiobesity agent To further assess whether UA increases the "browning propensity" and BAT activation in an adipocyte-autonomous manner, preadipocytes from inguinal fat and interscapular brown fat were isolated and differentiated in the presence or absence of UA. After differentiation, higher levels of UCP-1 protein (shown by immunostaining and western blotting) (Fig 3G and  3H) and elevated mRNA levels of typical browning markers were observed in UA-treated adipocytes than those of controls ( Fig 3I). Adipocyte thermogenic capacity is mitochondrial uncoupling, as determined by respiration. Therefore, mitochondrial oxygen consumption rate (OCR) was measured by a Seahorse analyzer. The results showed that the basal, uncoupled, and maximal OCRs were significantly higher in UA-treated brown adipocytes, as well as white adipocytes, than those of controls ( Fig 3J). Together, these results demonstrate that UA promotes browning of inguinal white fat and activates BAT in a cell-autonomous manner.

The β-adrenergic receptor is not a major contributor for UA-stimulated thermogenesis
One of the well-established mechanisms of nonshivering thermogenesis involves activation of the β-adrenergic receptors (β-ARs) [34][35][36]. We therefore tested whether β-ARs mediate UAinduced activation of BAT and browning of WAT. Mice were placed in a thermoneutral environment (30˚C) (Fig 4A) where stimulation of β-ARs is blunted and nonshivering thermogenesis is not required [37]. At thermoneutrality, UA-treated mice (8 weeks) exhibited lower body weight (−17.2%) than controls (Fig 4B), despite this difference in body weight being smaller (6.066 g in thermoneutrality versus 7.865 g at room temperature) than that observed at room temperature. Calorie intake was similar between the 2 groups of mice studied (Fig 4C). The fat mass was lower in UA-treated mice than that of controls ( Fig 4D). This could contribute to the lower body weight of UA-treated mice. In accordance with the results acquired at room temperature, oxygen consumption was 35% and 26% higher in the UA group than in controls during the light and dark cycles, respectively ( Fig 4E). UA-treated mice had a higher body temperature (+0.52˚C) than that of controls at thermoneutrality (Fig 4F and 4G). Similar to mice kept at ambient temperature, mRNA levels of thermogenic genes were higher in the iWAT and BAT of UA-treated mice than in controls at thermoneutrality (Fig 4H and 4I). Furthermore, when NE was injected into mice at thermoneutrality, significantly higher O 2 consumption was shown in UA-treated mice than in controls (Fig 4J), suggesting β-ARs are not necessary for the action of UA. To further confirm this, we employed a chemical approach to denervate the sympathetic nervous system (SNS) of BAT by microinjection of a specific neurotoxin 6-hydroxydopamine (6-OHDA) into brown fat of mice. The result showed that mRNA expressions of thermogenic markers were higher in BAT of UA-treated mice than that of controls ( Fig 4K). Together, these results showed that β-AR activation is not necessary for UAstimulated thermogenesis.

Adipose tissue TH signaling is involved in the UA-stimulated nonshivering thermogenesis
To find any clues to the potential underlying mechanism implicated in UA-mediated thermogenesis, we analyzed the available gene expression in UA-treated prostate cancer cells (LNCaP) from the Gene Expression Omnibus (GEO) data set (GSE65527). Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analysis of the data set showed that the TH signaling pathway, which plays key roles in the regulation of thermogenesis and body temperature in humans and animals [38][39][40], is highly up-regulated after UA treatment ( Fig  5A). To test whether TH signaling is involved in UA-mediated thermogenesis in mice, the circulating levels of triiodothyronine (T3) and tetraiodothyronine (T4) were assayed. There was no significant difference in levels of circulating T4 or T3 between UA-treated mice and vehicle controls (Fig 5B and 5C). However, the activated form of TH is T3, which is mainly converted from inactive T4 in its target tissue [41]. Adipose tissue is one of its target tissues [42]. When the 2 hormones were measured in adipose tissue, higher concentrations of T3 were detected in the BAT (1.76-fold) and iWAT (1.56-fold) of the UA-treated mice than in controls (Fig 5D  and 5E); a similar pattern of T3 was also observed in ob/ob mice (S2D Fig), whereas levels of T4 tended to decrease in the UA-treated group (Fig 5F and 5G), suggesting that UA treatment enhances the production of active T3 in adipose tissue, probably from converting the inactive T4. Consistent with this, we noticed higher mRNA levels of Deiodinase 2 (Dio2) (iWAT, 2.49-fold and BAT, 19.1-fold), a gene encoding the rate-limiting enzyme D2 that is responsible for the conversion of T4 to T3, and increased enzyme activities of D2 (iWAT, 1.94-fold and BAT, 1.57-fold) in the UA group than in controls (  The underlying data for this figure can be found in S1 Data. BAT, brown adipose tissue; Cox7a, Cytochrome c oxidase subunit 7a1; DEXA, dual-energy X-ray absorptiometry; Dio2, Deiodinase 2; HFD, high-fat diet; iWAT, inguinal white adipose tissue; NE, norepinephrine; Pgc1-α, Peroxisome proliferator-activated receptor gamma coactivator 1α; UA, urolithin A; Ucp1, Uncoupling protein 1. https://doi.org/10.1371/journal.pbio.3000688.g004 significantly higher in the BAT and iWAT of UA-treated mice than in control mice (Fig 5L  and 5M). Consistently, in primary cultured white adipocytes and brown adipocytes, significantly higher expression levels of TH-responsive genes were shown in UA-treated cells than those of controls (Fig 5N and 5O). Together, these results indicate that TH is involved in the UA-mediated thermogenesis in mice.

PLOS BIOLOGY
To further explore TH signaling in the UA-mediated thermogenesis in vitro, primary cultured white adipocytes and brown adipocytes were differentiated to adipocytes for 6 days in the standard differentiation medium for beige or brown adipocytes without UA treatment. Cells were then incubated with either 100 nM T3 or 100 nM T4 in the presence or absence of UA. Higher levels of UCP-1 and DIO2 expression were observed in UA -treated cells incubated with T4 than in control cells (Fig 5P and 5Q). These results suggest that UA converts inactive T4 to activate T3, which induces thermogenesis in vitro.

The beneficial metabolic effect of UA depends on TH
To corroborate the role of TH in UA-mediated thermogenesis and the prevention of dietinduced weight gain in vivo, a hypothyroidism mouse model was created (S9 Fig) by treating the mice with propylthiouracil (PTU), an inhibitor of TH production [43], for 6 weeks ( Fig  6A). PTU treatment completely blunted the preventive effect of UA on body weight gain ( Fig  6B). Food intake was similar in both groups (Fig 6C). Total fat, BAT, and iWAT weights were similar between UA-treated and control groups after PTU treatment (Fig 6D-6F). The beneficial effect of UA on glucose homeostasis also disappeared after PTU treatment, showing similar levels of glucose and courses of GTT and ITT (Fig 6G and 6H). Comparable body temperatures in UA-treated and control mice ( Fig 6I) suggested that the thermogenic effect of UA is lost upon PTU administration. Consistent with this observation, the mRNA expression levels of thermogenic genes in iWAT and BAT were not statistically different in UA-treated mice and vehicle controls after PTU administration (Fig 6J and 6K). A similar amount of Pgc1α and UCP-1 protein were shown in the BAT of both groups after PTU treatment ( Fig  6L). In iWAT of PTU-treated mice, Pgc1α and UCP-1 protein were undetectable in both groups of mice ( Fig 6M). Together, these results showed that inhibition of TH production abolishes the beneficial metabolic effect of UA in mice.
Because UA treatment increases the expression and activity of D2 (Fig 5H-5K), we reasoned that the main function of UA in TH production is to convert inactive T4 to active T3, which subsequently activated thermogenesis. To test this, exogenous inactive T4 instead of active T3 was given to PTU-treated mice. UA-treated and control mice were initially treated with PTU for 2 weeks and subsequently received PTU plus T4 (1 μg/100 g/day) for a period of 4 weeks (Fig 7A). Already, after 3 weeks of T4 treatment, body weights of UA-treated mice were 18.33% lower than those of controls ( Fig 7B) despite similar food intake ( Fig 7C). In line with lower body weight, the total fat mass and BAT and iWAT mass decreased in UA-treated mice by 25.8%, 24.9%, and 40%, respectively (Fig 7D-7F). Levels of plasma glucose were 25.2% lower in UA-treated mice than in controls after 3 weeks of T4 supplementation (Fig 7G). Increased body temperature by 0.8˚C (Fig 7H) was accompanied by elevated transcript levels of thermogenic genes implicated in the activation of BAT and browning of iWAT of UAtreated versus control mice after 4 weeks of T4 supplementation (Fig 7I and 7J). Finally, T4 treatment resulted in higher UCP-1 and PGC-1α protein levels in BAT and iWAT of UAtreated than in control mice (Fig 7K and 7L). Together, these results demonstrate that UAmediated thermogenesis depends on TH. Inhibition of TH production prevents, whereas exogenous T4 supply restores, the beneficial effects of UA on body mass control and glucose homeostasis.

Discussion
The prevalence of obesity has nearly tripled worldwide during the past 30 years [44][45][46][47]. Tremendous efforts have been made to develop effective antiobesity strategies. Unfortunately, strategies based on diet, exercise, and behavioral adaptation modifications have turned out to be insufficient for the long-term prevention of obesity in the majority of the population [48][49][50][51], and pharmacotherapy of obesity has been repeatedly found to be unsafe [52]. Therefore, novel, effective, and safe antiobesity interventions are urgently needed. In this study, we showed that UA, a gut-microflora-derived metabolite of natural products, prevents and also reverses HFD-induced obesity by promoting thermogenesis in BAT and iWAT. Besides the antiobesity effect, UA causes a series of beneficial metabolic effects, including improved insulin sensitivity and reduced systemic inflammation and fatty liver, which are at least partially due to weight loss. The mechanism of UA-stimulated thermogenesis is summarized in Fig 7M, showing that UA enhances adipose tissue production of T3, which subsequently activates transcription and translation of the thermogenic genes PGC1a and UCP-1, leading to higher EE and lower body weight.
In regard to the effect of UA on body weight control, we showed that the lower body weight (lower fat mass but similar lean mass) is due neither to the food intake nor to intestine fat

PLOS BIOLOGY
absorption. This raises the question of how UA treatment decreases fat accumulation. In principle, several factors may contribute to it: i) decreased adipose tissue lipid synthesis. Indeed, it is reported that UA decreases adipogenic protein and gene expression in primary cultured human adipocytes and murine 3T3-L1 adipocytes, leading to less TG accumulation [29,30]; ii) increased lipid catabolism and subsequent utilization; and iii) decreased lipid uptake by the intestine. Our data support that lower lipid synthesis and higher lipid utilization together contribute to UA-mediated decreased fat accumulation in mice. The decreased lipid synthesis is

PLOS BIOLOGY
Urolithin A, a novel antiobesity agent demonstrated by lower mRNA and protein levels of TG synthesis markers in the iWAT of UA-treated mice than in controls (S10 Fig), as well as in primary cultured white and brown adipocytes (S11 Fig). The increased lipid utilization is demonstrated in UA-treated mice by i) a lower RER, reflecting that lipids instead of carbohydrates are used as the main energy substrate; and ii) more EE is directly shown by pronounced higher levels of heat production and elevated core body temperature in UA-treated mice than in controls, which could have a profound contribution to energy consumption and therefore was explored further in this study.
Thermogenesis occurs in several tissues, including BAT, WAT, and skeletal muscle [53]. Here, several pieces of evidence collectively support that UA treatment enhances heat production by increasing the activity of BAT and inducing browning of iWAT: i) morphologically, BAT and iWAT of UA-treated mice resisted HFD-induced hypertrophy, maintaining a relatively normal appearance; ii) by immunohistochemistry, positive staining of UCP-1, a typical marker of BAT and browning of WAT, is higher in BAT and iWAT of UA-treated mice than in controls; iii) more mitochondria were shown by mtDNA copy number in BAT and iWAT of UA-treated mice; iv) mRNA levels of thermogenic markers and of UCP-1 protein were higher in the BAT and iWAT of UA-treated mice; and v) finally, infrared thermography revealed that UA-treated mice had a significantly higher interscapular BAT temperature than controls at room temperature. Together, these results show that UA directly enhances BAT thermogenesis and induces browning of iWAT, both of which contribute to higher levels of EE.
It is well established that thermogenesis in BAT and iWAT is regulated by the sympathetic nervous system through NE-mediated β-ARs and TH [54][55][56][57][58][59]. Our data rule out the involvement of β-ARs in the UA-mediated thermogenesis based on i) at thermoneutrality, at which stimulation of β-ARs is blunted, the effects of UA on body weight control and thermogenesis remained; and ii) when SNS of BAT was denervated, the effect of UA on tissue levels of DIO2 was maintained. Although there exist other thermogenic mechanisms, such as the proteasome system (S12A Fig), calcium (S12B Fig) [60], creatine cycling (S12C Fig) [61], and the glycolytic pathway (S12D-S12F Fig) [62], results derived from preliminary experiments in our study do not support a direct contribution of these pathways in the UA-mediated thermogenesis (S12 Fig). Instead, our data emphasize the importance of TH in the UA-mediated thermogenesis in BAT and iWAT. First, the mRNA expression and activities of the enzyme D2, which is responsible for the conversion of T4 to T3, were higher in BAT and iWAT of UA-treated mice than in controls, with a consequently higher concentration of T3 in the corresponding tissues. Consistently, in the presence of inactivated T4, higher levels of Dio2 and Ucp1 were shown in UAtreated primary cultured white and brown adipocytes than in controls (Fig 5P and 5Q). More directly, blocking the production of TH in mice abrogated the effects of UA on body weight control and nonshivering thermogenesis in BAT and iWAT. Conversely, restoration of TH production rescued the beneficial effects of UA. Together, these evidences strongly support that TH is indispensable for the beneficial effects of UA in mice.
It is worth noting that mice only respond to UA under an HFD, but not under a chow diet. This diet-dependent antiobesity effect is widely present in compound-treated normal mice [63,64], as well as in genetically modified mouse models [65]. The underlying mechanisms of these phenomena might be different and need further investigation. Mechanistically, the potential mechanism underlying the different effect of UA on mice body weight under an HFD and a chow diet might be due to tissue levels of T3, which is relatively low and may not reach concentrations at which it could significantly activate the thermogenic program in BAT and iWAT under normal diet feeding. Instead, higher levels of T3 were shown in UA-treated HFD mice than in normal-diet-feeding mice (Fig 5D and 5E (S12D-S12F Fig) [62], results derived from prelimin HFD increases tissue levels of T3 [66]. In fact, several similar lines of dose-dependent mechanism were reported on obesity research; for example, Growth differentiation factor 15 (GDF15) is widely accepted as a secreted protein that reduces food intake in mice [67]. Studies showed that the effect of GDF15 on food intake is dose-dependent [68]. More robust dose-dependent thermogenic evidence was derived from succinate study in mice, showing that 1.5% but not 1% of succinate could efficiently activate BAT thermogenesis, resulting in lower body weight [69]. Therefore, it is tempting to speculate that tissue levels of T3 may contribute to the different effects of UA under HFDs and normal diet feeding.
Safety is the top consideration for evaluating the potential clinical applications of antiobesity agents [70][71][72][73][74]. UA-mediated elevation of T3 concentrations in BAT and iWAT provokes us to consider the safety of UA application. Our data showed no obvious side effects and even an improvement for HFD-induced liver damage within the tested dose (30 mg/kg/day) and period of time (10 weeks). In rats, up to 5% of UA in the diet showed no adverse or toxic effects over a period of 90 days [75]. In addition, 50 mg/kg/day of UA treatment in mice for 8 months showed no adverse effects but an improvement of muscle grip strength and spontaneous exercise [27]. More importantly, the safety evaluation of UA in humans showed no side effects but improved mitochondrial function of skeletal muscle within the tested dose (a maximum of 1,000 mg/day) and period (56 days) [28, 76,77]. In our study, we also measured the effect of UA on muscle functions in mice. A greater exercise capacity (grip experience: 1.86-fold, grip strength: 1.57-fold, and swimming time: 1.72-fold) in UA-treated mice than in controls was shown in S13 Fig. Together, results derived from both humans and rodents consistently proved no toxicity but a beneficial effect of UA on skeletal muscle functions. Also, a recent publication showed that UA significantly enhances gut barrier function and attenuates inflammatory bowel disease in preclinical models [20]. Thus, based on these findings, it is speculated that UA does not have the side effects such as gastrointestinal disorders [78] and myopathy [79] caused by orlistat. Despite the effects of UA on body mass and lipid metabolism not being reported in the human study [28], the potential clinical application of UA seems promising based on its current safety assessment as well as the beneficial effects in humans and mice.
In summary, the data presented in this study showed that UA has beneficial effects on body weight control and glucose homeostasis without causing any adverse effects in mice. Our work establishes a key role of UA in the regulation of lipid metabolism and offers great promise for identifying new therapies to combat obesity and its related metabolic disorders.

Ethical approval
All animal studies were approved by and performed according to the guidelines of the ethics committee of the Northwest A&F University (Yangling, Shaanxi, China) (approval number: 20171208-010).

Animals
Six-week-old male C57BL/6 mice purchased from the animal center of Xi'An Jiao Tong University (Xi'an, China) and 4-week-old male leptin-deficient ob/ob mice purchased from Tengxin Biotechnology Co. Ltd. (Chong Qing, China) were acclimated for 2 weeks in the animal facility of Northwest A&F University. Unless otherwise specified, C57BL/6 mice were raised under an HFD (composed of 26.2% protein, 34.9% fat, D12492; Xie Tong Corp., Nanjing, China), and ob/ob mice were raised under normal diet (1010009, Xie Tong Corp.). At the age of 8 weeks, C57BL/6 mice were started on receiving UA (30 mg kg −1 d −1 , Jinan Feiteng Pharmaceutical Technology Co., Jinan, China) or orlistat (15 mg kg −1 d −1 , Cayman Chemical Co., Ann Arbor, MI, USA) or vehicle (0.1% Tween 80) by gavage. At the age of 6 weeks, ob/ob mice were started on receiving UA or vehicle. All mice were raised with free access to food and water in a 12-hour light/dark animal facility. For BAT denervation, 6-OHDA was injected at 4 different sites of each BAT under anesthesia (1.5 μL per site of 10 mg/mL dissolved in 1% ascorbic acid). The mice were allowed to recover for a minimum of 10 days before further experiments. For inhibition of TH production, UA-and vehicle-treated mice were orally given PTU (8.5 mg/100 ml in the drinking water, Chengdu Jia Ye Biotechnology Co., Ltd., # 51-52-5; Nanjing, China), an inhibitor of TH synthesis, for a period of 6-week HFD feeding. For TH restoration experiments, UA-treated and control mice were initially treated with PTU for 2 weeks and subsequently received PTU plus T4 (1 μg/100 g in 0.2% bovine serum albumin [BSA]-PBS, daily IP injection, Shanghai Baoman Biotechnology Co., Shanghai, China) for a period of 4 weeks.

GTTs and ITTs
For the GTT, overnight-fasted (approximately 14 h) mice received D-glucose (2 g/kg body weight) by IP injection. For ITT, 5-h fasted mice (from 6 AM to 11 AM) received recombinant human insulin at the concentration of 0.75 units/kg body weight. Levels of glucose were measured at indicated time points (0, 30, 60, 90, and 120 min) after injection.

Hyperinsulinemic-euglycemic clamps assay
Hyperinsulinemic-euglycemic clamps were performed as previously described [80]. Briefly, an indwelling silastic catheter was implanted in the right jugular vein of anesthetized mice 7 d before the experiments. Mice with less than 10% weight loss were used for the subsequent studies. After overnight fasting (16 hours), conscious mice were infused with [3-3 H]-glucose (HPLC purified; Perkin-Elmer Life Sciences, Waltham, MA, USA) at a rate of 0.05 μCi/min for 120 min for basal glucose turnover measurement. Then, clamps were performed for 140 min with a 3-min primed infusion of insulin (6.0 mU/[kg min]), followed by a continuous infusion of insulin (2.5 mU/[kg min]) and [3-3 H]-glucose (0.1 μCi/min) and a variable infusion of 20% dextrose to maintain euglycemia (approximately 120 mg/dL). A bolus of 2-deoxy-D-[1-14 C] glucose (10 μCi) was administrated to assess insulin-stimulated glucose uptake at t = 85 min. At the end of the clamps, mice were killed, and WAT, BAT, and gastrocnemius muscles were taken for subsequent analyses.

Body composition
Lean and fat masses were evaluated using DEXA as described [81].

Body temperature
The body temperature was recorded by a rectal probe (BAT-12, Physitemp, Clifton, NJ, USA).

Calorimetry
This was determined using Oxymax/CLAMS as described [81]. Briefly, mice were acclimated to the metabolic chamber for 48 h before the formal experiment. Oxygen consumption (VO 2 ), EE, and RER were measured with an Oxymax/CLAMS (Columbus Instruments, Columbus, OH, USA). An ANCOVA correction of EE with body composition as covariants was performed. For the evaluation of the effect of UA on whole-body thermogenic adipocyte activity, oxygen consumption was also measured acutely for mice receiving 1 mg/kg NE (Sigma-Aldrich, St. Louis, MO, USA) stimulation at 4˚C and thermoneutrality (30˚C) as described [82].

Cellular respiration detection
Cellular OCRs were determined using a Seahorse XFe96 Extracellular Flux Analyzer (Seahorse Bioscience, Billerica, MA, USA). Cells from iWAT and BAT were seeded at 20,000 cells/well. Differentiation was induced as described above [69,83], and then the cells were analyzed. Before the measurements, cells were washed twice with assay medium and incubated in 175 μL of assay medium for 1 hour in an incubator without CO 2 at 37˚C. Injection solutions were prepared as follows: oligomycin (2 μM final concentration), FCCP (1 μM final concentration), and a cocktail containing rotenone (1 μM) and antimycin A (1 μM).

Cold and thermoneutral challenges
For cold stress, mice were transferred from 21˚C to 4˚C with free access to water and food. The rectal temperature was measured at the indicated time points. For the thermoneutrality challenge, mice were placed at 30˚C with free access to water and food. Body weight, food intake, oxygen consumption, and body temperature were measured.

Indirect assessment of thermogenesis
This was evaluated by infrared thermography as described [84]. Representative infrared images of mice were measured using a thermal imaging camera (FLIR ONE PRO; FLIR systems, Wilsonville, OR, USA) and analyzed by FLIR Tools software.

Proteasomal activity of BAT
This was assayed with a proteasome activity fluorometric assay kit (BioVision, Milpitas, CA, USA) as described [62].

Primary adipocyte
Primary adipocytes were cultured as described [85][86][87]. Briefly, iWAT and BAT were taken from 5-h fasted mice. The collected adipose tissue was minced and digested in 0.1% Type I collagenase (Sigma-Aldrich) containing 2% BSA (Sigma-Aldrich) at room temperature for 1 h. The cell suspension was filtered and centrifuged at 2,000 × g for 10 min. Cells were then resuspended and cultured in DMEM with 10% FBS (ZETA Life, Menlo Park, CA, USA) at 37˚C. Adipocyte differentiation was induced with medium containing 10% FBS, 0.25 mM IBMX, 0.5 μM dexamethasone, 1 μM rosiglitazone, 850 nM insulin, 1 nM T3, and 0.125 mM indomethacin for 48 h. Then, the medium was replaced by one containing 10% FBS, 850 nM insulin, 1 nM T3, and 20 μM UA or DMSO. DMSO was used as the vehicle for different treatments.

Histology
Portions of liver, BAT, iWAT, and eWAT were fixed in 4% paraformaldehyde. HE staining was performed as described [88]. Immunohistochemistry and immunofluorescence staining of UCP-1 were performed according to the manufacturer's instructions. In primary cultured adipocytes, immunofluorescence staining of MyHC (R&D Systems, 1:100; Minneapolis, MN, USA) and MyOD (Santa Cruz Biotechnology, 1:100; Dallas, TX, USA) were performed according to the manufacturer's instructions. The images were acquired using BioTEX Software. mtDNA copy number mtDNA copy number was determined by quantitative Real-time PCR as described [89]. Briefly, total DNA was isolated from the cells using DNeasy Blood & Tissue Kit (Qiagen, 69506; Hilden, Germany) according to the manufacturer's instructions. The mtDNA copy number was calculated from the ratio of Co1 (mitochondrial-encoded gene)/Ndufv1 (nuclearencoded gene).

Deiodinase type 2 activity
The activity of type 2 deiodinase was determined in duplicate in iWAT and BAT homogenates (100 μg protein per 100 μL reaction) as described [90,91].

T3 and T4 measurements
Levels of T3 and T4 in the circulation, adipose tissue, and cultured adipocytes were quantified by a multiplexed ELISA kit (# EIAT3C; Thermo Fisher Scientific, Waltham, MA, USA) with a Luminex 200 analyzer following the manufacturer's instructions.

IL6 and TNFα measurements
Circulating levels of IL-6 and TNF-α were quantified using a multiplexed ELISA kit (Shanghai Enzyme Biotechnology Co., Shanghai, China).

Real-time qPCR and western blots
qPCR was performed as described [88]. The sequence of primers used is listed in S1 Table. Western blots were performed as described [88]. The information about antibodies used for immunoblotting is listed in S2 Table.

Skeletal muscle function analysis
Kondziela's inverted screen test was performed as described [92,93]. Briefly, each mouse was put on the center of the wire mesh and rotated to an inverted position. The time when the mouse falls off was measured and scored as the following: 1, falling between 1 and 10 s; 2, between 11 and 25 s; 3, between 26 and 60 s; 4, between 60 and 90 s; 5, after 90 s. The forelimb grip strength test was performed in mice by using a grip strength meter (Columbus Instruments). The antifatigue experiment was performed as described [94]. Briefly, mice were placed in a swimming pool filled with fresh water at 25 ± 1˚C. A tin wire (5% of body weight) was loaded on the root of the mouse's tail. The swimming period was regarded as the time spent by the mouse floating in the water, struggling, until exhausted (failed to rise to the surface of the water within 10 s).