Glycerol kinase stimulates uncoupling protein 1 expression by regulating fatty acid metabolism in beige adipocytes

Browning of adipose tissue is induced by specific stimuli such as cold exposure and consists of up-regulation of thermogenesis in white adipose tissue. Recently, it has emerged as an attractive target for managing obesity in humans. Here, we performed a comprehensive analysis to identify genes associated with browning in murine adipose tissue. We focused on glycerol kinase (GYK) because its mRNA expression pattern is highly correlated with that of uncoupling protein 1 (UCP1), which regulates the thermogenic capacity of adipocytes. Cold exposure-induced Ucp1 up-regulation in inguinal white adipose tissue (iWAT) was partially abolished by Gyk knockdown (KD) in vivo. Consistently, the Gyk KD inhibited Ucp1 expression induced by treatment with the β-adrenergic receptors (βAR) agonist isoproterenol (Iso) in vitro and resulted in impaired uncoupled respiration. Gyk KD also suppressed Iso- and adenylate cyclase activator-induced transcriptional activation and phosphorylation of the cAMP response element-binding protein (CREB). However, we did not observe these effects with a cAMP analog. Therefore Gyk KD related to Iso-induced cAMP products. In Iso-treated Gyk KD adipocytes, stearoyl-CoA desaturase 1 (SCD1) was up-regulated, and monounsaturated fatty acids such as palmitoleic acid (POA) accumulated. Moreover, a SCD1 inhibitor treatment recovered the Gyk KD-induced Ucp1 down-regulation and POA treatment down-regulated Iso-activated Ucp1. Our findings suggest that Gyk stimulates Ucp1 expression via a mechanism that partially depends on the βAR-cAMP-CREB pathway and Gyk-mediated regulation of fatty acid metabolism.

Adipose tissues are classified as white adipose tissue (WAT) 2 and brown adipose tissue (BAT) according to their functions. WAT mainly consists of white adipocytes and stores extra energy in the form of triglycerides (TG) (1). BAT mostly comprises brown adipocytes and decomposes TG and dissipates thermal energy. It plays an important role in the regulation of thermogenic homeostasis (2). Nonshivering thermogenesis liberates heat via mitochondrial uncoupling protein 1 (UCP1) in the inner mitochondrial membrane. Mitochondrial ATP synthesis is driven by a proton gradient produced by the respiratory electron transport chain. In contrast, UCP1 uses the proton gradient to generate heat rather than synthesize ATP (3).
Recent studies revealed that brown-like "beige" adipocytes appear in rodent WAT under cold exposure or the ingestion of certain dietary factors (4,5). Beige adipocytes expend energy via UCP1-mediated thermogenesis (6). Like other subcutaneous WAT, inguinal WAT (iWAT) browning is more readily induced than visceral WAT such as epididymal WAT (eWAT) (7). Several previous studies indicated that regulatory mechanisms of Ucp1 expression in each cell do not completely correspond (8). Moreover, the BAT in adult humans is mainly composed of beige adipocytes (9). Thus, elucidation of the browning mechanism in WAT may lead to effective new obesity prevention and treatment approaches.
Glycerol kinase (GYK) phosphorylates glycerol to glycerol 3-phosphate. This is an important step for the metabolism of glycerol that acts as the backbone of TG (10). Previous studies showed that deficiency of Gyk causes an X-linked reces-sive disease in humans, which shows hyperglycerolemia associated with congenital adrenal hypoplasia and developmental delay (11). In addition, Gyk KO mice show growth retardation, altered fat metabolism, and neonatal death (12). Although Gyk is more highly expressed in the liver and kidney than the adipose tissue (13), Gyk activity and expression in BAT has been reported to be up-regulated in parallel by cold exposure or prolonged norepinephrine infusion (14,15). These reports indicated that Gyk might be related to thermogenic function in BAT. However, the physiological functions of Gyk in WAT browning and their underlying mechanisms are unclear.
Here, we examined the effects of Gyk on Ucp1 expression during adipocyte browning in vitro and in vivo. Gyk positively regulated Ucp1 by controlling the ␤AR-cAMP-CREB signaling pathway via altering FA metabolism in adipocytes.

Ucp1 and Gyk had similar mRNA expression patterns in adipose tissues and adipocytes
First, we obtained comprehensive gene expression pattern in adipose tissue of mice during cold exposure (4°C for 0 to 384 h) using microarray. Twenty-nine genes that might be related to regulate thermogenesis in BAT were selected (Table 1). We then analyzed their expression pattern in iWAT via clustering analysis (see "Experimental procedures"). As shown in Fig. 1A, Gyk was located in the nearest cluster as Ucp1 and their expression levels were highly correlated (Pearson's correlation coefficient ϭ 0.951). Esrra located in a cluster next to the cluster of Ucp1 and Gyk is reported to be essential for high levels of mitochondrial biogenesis and oxidative capacity, characteristic of BAT, and thus for energy production in thermogenesis (16). On the other hand, the role of Gyk in BAT thermogenesis is unclear, and its function in iWAT has not been reported yet. We confirmed the gene expression levels of Ucp1 and Gyk in iWAT by real-time PCR. During cold exposure, mRNA expression levels of both Ucp1 and Gyk were rapidly up-regulated by 48 h (Fig. 1B). Therefore, we focused on the physiological role of Gyk in iWAT.
In iWAT, cold-induced Gyk and Ucp1 up-regulation were observed in the adipocyte fraction but not the stromal vascular fraction (Fig. 1C). We used primary preadipocytes derived from iWAT to examine Gyk expression in this tissue. As shown in Fig. 1D, Ucp1 and Gyk were significantly up-regulated after the induction of adipocyte differentiation. Furthermore, Ucp1 and Gyk were significantly up-regulated in adipocytes treated with isoproterenol (Iso), a nonselective ␤-adrenergic receptor (␤AR) agonist (Fig. 1E).

Gyk knockdown suppressed ␤AR-induced Ucp1 up-regulation in iWAT
To investigate the role of Gyk in adipocyte browning in vivo, we performed iWAT-targeting short hairpin RNA (shRNA) injection against Gyk using adeno-associated virus (AAV) as a vector. The shRNA against LacZ served as the control. Two weeks after AAV infection, the mice were individually housed either at room temperature (23°C) or low temperature (10°C) for 24 h. Cold stimulation reduced the weight of iWAT, whereas Gyk KD did not cause significant variations in body or tissue weight (Table 2). Plasma TG levels were decreased both by cold exposure and Gyk KD. However, the effect of Gyk KD on plasma TG levels was abolished under low temperature ( Table  2). These findings suggested that cold stress-activated ␤AR signaling may be involved in Gyk KD-regulated TG metabolism in vivo. As shown in Fig. 2A, fluorescence from ZsGreen encoded in the AAV vector was observed only in iWAT. The shRNA injection successfully down-regulated Gyk in iWAT (Fig. 2B). No AAV infection-mediated Gyk down-regulation was confirmed in the BAT, eWAT, or liver ( Fig. S1), suggesting AAV infection-mediated Gyk down-regulation occurred exclusively in iWAT. Browning marker genes such as Ucp1, Cidea, Pgc1a, and Pparg were also down-regulated by Gyk KD in iWAT (Fig.  2B). Consistent with mRNA expression, UCP1 protein in iWAT was increased by cold exposure; however, this up-regulation was attenuated by Gyk KD (Fig. 2C). A histochemical analysis showed that cold exposure increased the number of small multilocular adipocytes in the iWAT of control mice, whereas this response was relatively attenuated in the iWAT of Gyk KD mice (Fig. 2D). Low-temperature enhancement of the UCP1 immunohistochemical (IHC) staining intensity in iWAT was decreased by Gyk down-regulation (Fig. 2D). On the other hand, Ucp1 gene expression and its protein expression levels in BAT were not affected by shGyk transfection (Fig. S1, B and C). H&E and IHC staining disclosed that shGyk transfection had no effect in BAT (Fig. S1D). To examine the effect of iWATspecific Gyk KD on cold-induced thermogenesis, we performed a cold tolerance test. Although there was no significant difference in rectal temperature changes during cold exposure, the rectal temperature in Gyk KD mice tended to be lower than that in control mice (Fig. 2E).

Gyk knockdown suppressed ␤AR-induced Ucp1 up-regulation in adipocytes
Gyk was down-regulated in immortalized primary iWAT cells infected with an adenovirus encoding Gyk-specific shRNA. Approximately 60 and 50% knockdown of Gyk mRNA expression occurred with and without Iso treatment, respectively (Fig. 3A). Ucp1 was up-regulated by Iso in the control (shRNA for LacZ)-transfected cells. Gyk KD had no effect on Ucp1 expression in the absence of Iso treatment. Nevertheless, Ucp1 up-regulation by Iso treatment was suppressed by Gyk KD (Fig. 3A). Cidea, a browning marker gene, was also up-regulated by Iso treatment and this effect was suppressed by Gyk KD (Fig. 3A). Luciferase reporter assays revealed that Ucp1 promoter and cAMP-response element (CRE) transcriptional activities were up-regulated by Iso treatment, and these up-reg- Glycerol kinase regulates Ucp1 expression ulations were suppressed by Gyk KD (Fig. 3B). Next, phosphorylation of protein kinase A (PKA)-target protein were measured by Western blotting. In the presence of Iso, adipocytes with Gyk KD had lower hormone-sensitive lipase (HSL) and CRE-binding protein (CREB) phosphorylation levels than shLacZ-induced adipocytes (Fig. 3C). A chromatin immunoprecipitation (ChIP) assay was performed to investigate whether CREB recruitment was reduced in the Ucp1 promoter region of Gyk KD adipocytes. Four CREs (CRE1-4) with the same core sequence were identified in the mouse Ucp1 5Ј flanking region (17). However, CRE1 has no regulatory function and the impact of CRE3 is only marginally significant (18). Only CRE2 and CRE4 were reported to be necessary for Ucp1 regulation by CREB (18). Therefore, we investigated these two regions. In the presence of Iso, CREB recruitment increased in CRE2 and CRE4 and decreased in response to Gyk KD (Fig. 3D). These findings indicated that Ucp1 was down-regulated by Gyk KD via inhibition of the ␤AR signal. To investigate whether Gyk KD affect adipocytes energy metabolism, we performed a mitochondrial stress test. Gyk KD lowered the basal mitochondrial respiration rate. Proton leak also decreased without any apparent change in maximal respiration or ATP production, suggesting that Gyk KD decreased the uncoupling rate (Fig. 3, E and F). These findings indicated that Gyk KD down-regulated Ucp1 during Iso treatment in beige adipocytes by attenuating Iso-mediated ␤AR signaling activation.

Glycerol kinase regulates Ucp1 expression
Forskolin also enhanced CREB phosphorylation and Ucp1 promoter and CRE transcriptional activities but these were suppressed in Gyk KD cells (Fig. 4, D and E). In contrast, in the case of 8-Br-cAMP, the Gyk KD-induced suppressions were not observed (Fig. 4, D and E). In control adipocytes, Iso elevated intracellular cAMP levels, whereas Gyk KD attenuated them (Fig. 4F). cAMP, synthesized from ATP by AC, is catalytically degraded into AMP by phosphodiesterases (PDE). Although Gyk KD showed no effect on AC protein levels, it significantly increased PDE activity (Fig. 4, G and H), suggesting that the decrease in cAMP levels in Gyk KD adipocytes is mediated by increased PDE activity, at least partially. These results indicated that ␤AR-stimulated cAMP accumulation in adipocytes was down-regulated by Gyk KD, leading to the suppression of Ucp1 up-regulation.

Gyk knockdown down-regulated Ucp1 expression by altering fatty acid metabolism in adipocytes
Gyk catalyzes the conversion of glycerol to glycerol 3-phosphate, which is a substrate for TG synthesis. We hypothesized that glycerol and fatty acids (FAs) used for TG biosynthesis were accumulated in Gyk KD adipocytes. Glycerol and free FA levels were significantly increased in the supernatant of Isotreated Gyk KD adipocytes compared with Iso-treated control adipocytes (Fig. 5A). Intracellular free FAs were detected by LC-MS to confirm the effects of Gyk on FA metabolism in Gyk KD cells. Iso treatment increased the monounsaturated FAs (palmitoleic acid (POA) and oleic acid (OA)), whereas these enhancements were further increased in Gyk KD cells (Fig. 5B). We hypothesized that monounsaturated FA accumulation down-regulates Ucp1 in Gyk KD cells. As shown in Fig. 5C, stearoyl-CoA desaturase (SCD) catalyzes the conversion of monounsaturated FAs, such as OA and POA, from saturated FAs, such as stearic acid (SA) and palmitic acid (PA). We measured the expression levels of four murine Scd isoforms in Gyk KD cells in response to Iso treatment. As shown in Fig. 5D, Scd1 and Scd2 were highly expressed in primary iWAT cells and Scd1 was especially up-regulated in Iso-treated Gyk KD cells. We investigated whether Scd1 links Gyk-mediated regulation of Ucp1 expression. Under Iso treatment, Gyk KD suppressed Ucp1 expression but this inhibition was abolished by SCD1 inhibitor treatment (Fig. 5E). We evaluated SCD1 activity by calculating the ratio of saturated FA (PA ϩ SA) to monounsaturated FA (POA ϩ OA). It was markedly reduced in the Gyk KD group but the SCD1 inhibitor treatment rescued it (Fig. 5F). Thus, the influences of Gyk on Ucp1 expression may depend partially on SCD1-mediated intracellular monounsaturated FA accumulation.

DGAT and ATGL inhibitors regulated Ucp1 expression
To further investigate the relationship between intracellular FA accumulation and Ucp1 expression, we inhibited diglyceride acyltransferase (DGAT) and adipose triglyceride lipase (ATGL), which convert diacylglycerol to TG and regulate the first step in TG hydrolysis, respectively. No significant variations were detected in the absence of Iso. However, in the presence of Iso, DGAT inhibitor down-regulated Ucp1 expression, whereas ATGL inhibitor up-regulated it (Fig. 6A). With Iso addition, the Ucp1 promoter and CRE transcriptional activities were also decreased or increased by treating using DGAT or ATGL inhibitor, respectively (Fig.  6B). As shown in Fig. 6C, PA, POA, and OA levels were

Glycerol kinase regulates Ucp1 expression
dramatically augmented by DGAT inhibitor treatment but considerably decreased by ATGL inhibitor treatment. Thus, intracellular FAs may modulate ␤AR signaling-regulated Ucp1 mRNA expression.

Palmitoleic acid treatment suppressed ␤AR agonist-induced Ucp1 expression
As shown in Fig. 7A, Iso-induced Ucp1 up-regulation was down-regulated by only POA treatment. In the presence of Iso, POA reduced Ucp1 promoter and CRE transcriptional activities in a dose-dependent manner (Fig. 7B). Similar to the results of Gyk KD (Fig. 4, B and C), POA treatment inhibited forskolininduced Ucp1 promoter and CRE transcriptional activation but failed to inhibit 8-Br-cAMP-induced their transcriptional activation (Fig. 7C). In addition, PDE activity was increased by POA treatment (Fig. 7D). To confirm whether POA treatment influenced FA accumulation, we measured the intracellular FAs. POA application induced intracellular POA accumulation (Fig. 7E). Thus, Gyk KD-inhibited Ucp1 upregulation is the result of ␤AR signaling attenuation by increased PDE activity via a mechanism partially dependent on enhanced POA accumulation.

Discussion
Browning in white adipocytes is an inducible process that is a potential target for obesity treatment or prevention.
Ucp1 is a classical thermogenic gene that has been widely used to evaluate adipocyte browning. Here, we searched for genes related to iWAT browning using transcriptome analysis and identified Gyk as a candidate gene. Gyk and Ucp1 up-regulation were observed in adipocyte fraction in iWAT during cold exposure and Iso-treated primary iWAT cells (Fig. 1, C and D). Gyk KD suppressed cold exposure-or Isoinduced Ucp1 expression in vivo and in vitro (Figs. 2B and 3A) via ␤AR signaling, by suppressing the increase in cAMP production. Because Gyk-catalyzed glycerol 3-phosphate formation, which is a substrate of TG synthesis, we hypothesized that FAs metabolism might be related to Gyk KDinduced suppression of Ucp1 up-regulation. Actually, the expression of Scd1 was enhanced in Iso-treated Gyk KD cells accompanied with an increase in monounsaturated FAs accumulation. Furthermore, POA addition suppressed Ucp1 up-regulation via ␤AR signaling. As shown in Fig. 8, Gyk regulates Ucp1 expression, via POA accumulation-mediated suppression of the ␤AR-cAMP-CREB pathway.
Although Gyk has been reported to be highly expressed in the liver and kidney (13,19), a recent study reported that Gyk expression levels were induced by cold exposure in BAT (20,21). Lasar et al. (22) reported that Gyk is partially related to peroxisome proliferator-activated receptor ␥ function in BAT. These reports indicated that Gyk is implicated in BAT

Glycerol kinase regulates Ucp1 expression
function but do not elucidate its detailed mechanism. Moreover, the physiological function of Gyk in iWAT browning remains unknown. To the best of our knowledge, this study is the first to indicate the role of Gyk in iWAT browning. The molecular signatures of adult human brown adipocytes resemble those of mouse beige adipocytes rather than brown ones (23). Therefore, functional analysis of mouse beige adipocytes would be more informative than that of brown adipocytes. Our findings could help uncover the role of Gyk in human BAT.
Thiazolidinediones administered for type 2 diabetes induce Gyk expression to promote TG synthesis and inhibit glycerol and FA secretion by adipocytes (24). Thus, Gyk upregulation increases TG synthesis and decreases intracellular glycerol and FAs. Intracellular FAs are thermogenesis substrates via UCP1. In Gyk KD adipocytes, intracellular FAs were increased, hence, Gyk KD had been considered acting positively for thermogenesis. Although the FA contents were enhanced, Gyk KD, DGAT inhibitor, and POA all downregulated Ucp1 expression. Previous studies (25,26) showed that lipogenesis is up-regulated in BAT during cold exposure. In mice with BAT-specific hypothyroidism, cold-induced lipogenesis in BAT was inhibited, and this inhibition led to the reduction of thermogenic activity in BAT (26). These reports indicated that not only lipolysis but also lipogenesis in BAT during cold exposure are important for thermogenic activity in BAT. In this study, we showed in the presence of ␤AR stimuli, Gyk plays an important role to reg-ulate intracellular FAs levels and composition. Therefore Gyk seems to be important regulator for the balance between lipolysis and lipogenesis in beige adipocytes.
The Gyk KD adipocytes that were not treated with Iso presented with comparatively lower fatty acid release levels (Fig. 5A). Concerning this point, we revealed that the Gpat1 expression level was significantly higher in Gyk KD adipocytes than control adipocytes (data not shown). Glycerol-3-phosphate acyltransferase (GPAT) catalyzes the conversion of glycerol 3-phosphate and long-chain acyl-CoA to lysophosphatidic acid. It was reported that GPAT is the ratelimiting enzyme in de novo glycerolipid biosynthesis (27). GPAT1 activity may be positively correlated with the rate of fatty acid storage in the form of TG in human adipose tissue (28). Moreover, Gpat1 overexpression has been reported to increase cellular triacylglycerol accumulation (29). An increase in Gpat1 mRNA expression may not accurately reflect a corresponding increase in GPAT1 enzyme activity. However, the capacity for fatty acid acylation might be higher in Gyk KD adipocytes not treated with Iso than that in control adipocytes.
Scd1 is expressed in various tissues including the liver and adipose tissue. Enser (30) reported that obesity mouse models exhibiting a high level of Scd1 expression was present with liver steatosis and insulin resistance. In contrast, Scd1-deficient mice were resistant to diet-induced obesity possibly related to enhanced energy expenditure (31,32). In this study, we showed a Scd1 up-regulation-mediated increase in POA levels sup-

Glycerol kinase regulates Ucp1 expression
pressed Ucp1 up-regulation in beige adipocytes, suggesting that Scd1 up-regulation negatively regulates energy expenditure. This mechanism might contribute to anti-obese phenotype in Scd1-deficient mice.
Saturated FAs have been reported to enhance various cellular stresses, such as oxidative stress and endoplasmic reticulum stress and inflammation, in adipose tissue (33,34). We hypothesized that Gyk KD could reduce the intracellular concentration of saturated FAs by up-regulating Scd1 to attenuate saturated FA-induced cellular stresses. However, as the result, accumulated POA may negatively regulate ␤AR-stimulated Ucp1 up-regulation.
Here, we proposed that POA functionally participates in Ucp1 regulation. POA may be an important signaling molecule produced mainly by WAT (35). POA functions as a WAT-derived lipokine implicated in insulin sensitivity in skeletal muscle, hepatic lipogenesis (36), and lipid metabolism in WAT (37). Although the effects of POA on the regulation of energy metabolism are largely unknown, Souza et al. (38) reported that daily administration of 300 mg kg Ϫ1 POA for 10 days successfully attenuated insulin resistance, liver inflammation, and damage caused by high-fat diets. Although several benefits of POA in mice were mentioned, POA may have several disadvantages in humans. Plasma POA levels were highly correlated with nonalcoholic fatty liver disease (39) and heart failure (40). As the data on the effects of POA on human health are inconsistent with those for animal and cell culture models, further human POA trials are warranted (41). Our results showed that the increase in intracellular POA levels is involved in the suppression of Ucp1 expression. Under Iso-treated conditions, POA cellular contents were increased to 4.3 (Fig. 5B), 6.6 ( Fig. 6C), and 2.9 fmol/cell (Fig. 7D) by Gyk KD, DGAT inhibitor treatment, and POA treatment, respectively. Ucp1 suppression levels were 44, 53, and only 16%, respectively, suggesting that cellular POA contents are positively correlated with Ucp1 suppression levels. Although it is difficult to estimate accurate POA amounts required for Ucp1 suppression, approximately 2.9 fmol/cells or more POA levels may be needed to suppress Ucp1 expression.
The present study disclosed that Gyk is an important gene in the white adipocyte browning process. We proposed that Scd1regulated FA composition was codetermined with the role of Gyk in Ucp1 expression. The latter partially depends on the ␤AR-cAMP-CREB pathway in adipocytes (Fig. 8). Our findings provide insights into the potential mechanisms by which beige adipocyte thermogenesis is controlled via the regulation of Gyk expression. These modes of action could be exploited as a potential strategy to modulate Ucp1 expression and body temperature control.

Animal experiments
Male C57BL/6J mice were purchased from Japan SLC (Shizuoka, Japan). They were kept in individual cages at 23 Ϯ 1°C and under a 12-h light/dark cycle. They had access to food and water ad libitum and fed a commercial chow diet (MF; Oriental

Glycerol kinase regulates Ucp1 expression
Yeast Co. Ltd., Tokyo, Japan). The study was approved by the Animal Care Committee of Kyoto University.

RNA sample preparation for microarray analysis
Male 6-week-old C57BL/6J mice were subjected to cold exposure (4°C) for 0, 1,2,4,8,12,24,48,192, or 384 h. Then their iWAT was harvested and its RNA was extracted with the RNeasy lipid tissue mini kit (Qiagen, Hilden, Germany) according to the manufacturer's instructions. The extracted total RNA was used to synthesize fluorescently labeled cDNA.

Microarray analysis
Microarray analysis was carried out using Agilent SurePrint G3 Mouse GE 8 ϫ 60K microarrays (ϳ60,000 probes including 39,430 Entrez Gene RNAs and 16,251 lincRNAs). All experiments were performed in triplicate (three microarray experiments or three different mouse RNA samples per time point). Microarray data quality control was performed by Agilent feature extraction (Agilent Technologies, Santa Clara, CA, USA) and GeneSpring software (Agilent Technologies). The probes for each gene symbol were collapsed into one representative probe with maximum variance (24,163 genes). Raw signal intensities were normalized for each microarray by quantile normalization in the "limma" package (42). Expression values were then log-transformed (base 2). The batch effect was removed with the "Combat" function in the "sva" package (43).

Clustering analysis
A clustering analysis was run on the selected genes using the "pvclust" package (44). Pearson's correlation coefficients were selected on the basis of similarity between gene expressions and the average linkage for the clustering method.

Recombinant AAV vectors
The DNA sequences corresponding to the shRNA sequences of Gyk were 5Ј-ACCCTCCATGCCTGAAACA-3Ј and 5Ј-ACC-ACTTTCTGGAGACTGAGTT-3Ј. They were annealed and ligated into the shRNA expression vector pAAV-U6-ZsGreen1 (TaKaRa Bio, Kusatsu, Shiga, Japan). Recombinant AAV6 ex-pressing the Gyk shRNA were generated according to the manufacturer's protocol. A recombinant AAV6 expressing a LacZ shRNA was generated as a negative control. The AAV particles were purified with the AAVpro purification kit (TaKaRa Bio). The virus titers (viral genomes (vg) ml Ϫl ) were determined by qPCR.

Administration of AAV vectors
Male 10-week-old C57BL/6J mice were anesthetized with isoflurane and injected into their bilateral iWAT with 40 l of AAV6 vector solutions (4 ϫ 10 12 vg ml Ϫl ) per pad. After 2 weeks, the mice were subjected to cold treatment (10°C) for 24 h and sampled for analysis.

RNA preparation and quantification of gene expression
Total RNA samples were isolated with a commercially available reagent (Sepasol Super-I, Nacalai Tesque). Aliquots of total RNA were reverse-transcribed with Moloney murine leukemia virus reverse transcriptase (Promega, Madison, WI) according to the manufacturer's instructions. To quantify mRNA expression, real-time RT-PCR was performed in a LightCycler system (Roche Diagnostics) with SYBR Green fluorescence signals as previously described (45). The oligonucleotide primers were designed in a PCR primer selection program available from the GenBank TM database Virtual Genomic Center. They are shown in Table S1 mRNA expression levels were normalized to those of 36B4 mRNA.

Plasma characteristics
Plasma glucose, TG, and free FA levels were enzymatically determined with glucose CII, triglyceride E, and NEFA C test kits (Wako Pure Chemicals Industries Ltd., Osaka, Japan), respectively.

Histological analysis
Tissues were excised from each mouse and fixed in 10% (v/v) paraformaldehyde/phosphate-buffered saline. After ethanol dehydration, the fixed samples were embedded in paraffin, cut into 5-m sections with a microtome, and

Cold tolerance test
Two weeks after administration of AAV vectors, rectal temperature of mice was measured at 0, 1, 2, and 4 h after 10°C cold exposure using a thermometer probe (T&D Corp., Nagano, Japan). The area under the curve was calculated using the trapezoidal rule.

Cell culture
Primary mouse iWAT preadipocytes were immortalized by transfection with Simian Virus 40 large T antigen (kind gift from Prof. S. Kajimura, University of California, San Francisco, CA). Successfully transfected clones were screened for puromycin resistance. Mouse 10T1/2 preadipocytes were purchased from the American Type Culture Collection (Manassas, VA). All cell lines were maintained in a humidified 5% CO 2 atmosphere at 37°C in basic medium (Dulbecco's modified Eagle's medium-high glucose supplemented with 10% (v/v) fetal bovine serum, 10,000 units ml Ϫ1 of penicillin, and 10,000 g ml Ϫ1 of streptomycin). The cell cultures were raised to 95-97% confluence. The iWAT preadipocytes cells were differentiated with basic medium plus 2 g ml Ϫ1 of dexamethasone, 5 g ml Ϫ1 of insulin, 0.5 mM 3-isobutyl-1-methylxanthine, 125 M indomethacin, 1 nM T 3 , and 0.5 M rosiglitazone for 48 h. The medium was replaced every 2 days with growth medium (basic medium supplemented with 5 g ml Ϫ1 of insulin and 1 nM T 3 ). To knockdown Gyk, adenovirus (multiplicity of infection ϭ 300) was inducted into cells at the 6-day differentiation stage (47). After 8 days differentiation, the cells were subjected to 0.

Luciferase reporter assays
Luciferase reporter assays were performed as previously reported (48). Briefly, pUcp1-pro-Luc or pCRE-Luc (kindly provided from Prof. Y. Kamei, Kyoto Prefectural University) were transfected into 10T1/2 cells growing on 100-mm tissue culture dishes. Four hours after transfection, the cells were seeded in 96-well-tissue plates and incubated with 0.5 M Iso, 1-5 M FSK, and 100 -500 M 8-Br-cAMP for 4 h. The cells were then lysed for the luciferase assay in a dual-luciferase reporter gene assay system (Promega) in accordance with the manufacturer's protocol.

Chromatin immunoprecipitation (ChIP) assay
The ChIP assay was performed according to the manufacturer's protocol (Upstate Cell Signaling Solutions, Lake Placid, NY), with some modifications. The cells were fixed in 1% (v/v) formaldehyde, quenched with 125 mM glycine, collected, resuspended in 1% (v/v) SDS lysis buffer, and sonicated to shear their DNA into 100 -1,000-bp fragments. The supernatant was collected and immunoprecipitated overnight with 5 g of CREB antibody (Cell Signaling Technology) or with 1 g of rabbit IgG isotype (Novus Biological, Littleton, CO) as a mock control along with Magna ChIP TM protein AϩG magnetic beads (EMD Millipore, Burlington, MA) at 4°C in a rotary shaker followed by reverse crosslinkage and protease K digestion. The eluted DNA was purified with a MinElute PCR purification kit (Qiagen, Hilden, Germany) and analyzed by real-time PCR. Primer sequences are listed in supporting Table S1.

cAMP content quantification
The cAMP content was enzymatically determined with a cAMP-Glo TM max assay kit (Promega) according to the manufacturer's instructions.

Phosphodiesterase (PDE) activity assay
The cells were lysed at 4°C using the RIPA buffer (50 mM Tris-HCl, 150 mM NaCl, 1% (v/v) Nonidet P-40, 0.5% (w/v) deoxycholate, 1 mM EDTA, 0.1% (w/v) SDS, pH 7.4) mixed with protease and phosphatase inhibitors. The lysates were collected and centrifuged 5 min at 10,000 ϫ g at 4°C to remove cellular debris and the supernatant was used for the PDE activity assay using a colorimetric PDE Activity Assay Kit (Abcam, Cambridge, UK). Briefly, the supernatant was purified by gel filtration using desalting resin and desalting columns. The PDE activity was measured by the method based on the sequential hydrolysis of cyclic nucleotides by PDE and 5Ј nucleosidase. The released phosphate by enzymatic cleavage is directly pro-Glycerol kinase regulates Ucp1 expression portional to PDE activity and quantified using a modified Malachite Green reagent.

Oxygen consumption rate measurement
The oxygen consumption rate (OCR) reflects the mitochondrial respiration rate and was determined with the XF24 extracellular flux analyzer (Seahorse Bioscience, North Billerica, MA), according to the manufacturer's instructions. Briefly, differentiated primary mouse iWAT adipocytes were seeded in 24-well-plates (Seahorse Bioscience) and the OCRs were measured and recorded with a sensor cartridge and Seahorse XF-24 software (Seahorse Bioscience). Basal respiration was first measured and then the following were sequentially injected: 500 nM oligomycin (a mitochondrial ATP-synthase inhibitor), 1 M carbonyl cyanide p-trifluoromethoxyphenylhydrazone (FCCP; mitochondrial oxidative phosphorylation uncoupler), and a mixture of 15 M antimycin A (a mitochondrial complex III inhibitor) and 15 M rotenone (a mitochondrial complex I inhibitor). The OCR values related to basal respiration, ATP production, proton leak, maximum respiration, and coupling and uncoupling rates were calculated as follows: 1) basal respiration ϭ last point before oligomycin injection Ϫ minimum value after antimycin A and rotenone; 2) ATP production ϭ last point before oligomycin injection Ϫ minimum value after oligomycin; 3) proton leak ϭ minimum value after oligomycin Ϫ minimum value after antimycin A and rotenone; and 4) maximum respiration ϭ maximum value after FCCP Ϫ minimum value after antimycin A and rotenone.

Statistical analysis
Data are presented as mean Ϯ S.E. Student's t test or analysis of variance followed by the Tukey-Kramer test were used to identify statistically significant differences among treat-ment means. Differences were considered significant at p Ͻ 0.05.

Data availability
All data for this publication are included in this published article and its supporting information.