Induction of UCP1 and thermogenesis by a small molecule via AKAP1/PKA modulation

Strategies to increase energy expenditure are an attractive approach to reduce excess fat storage and body weight to improve metabolic health. In mammals, uncoupling protein-1 (UCP1) in brown and beige adipocytes uncouples fatty acid oxidation from ATP generation in mitochondria and promotes energy dissipation as heat. We set out to identify small molecules that enhance UCP1 levels and activity using a high-throughput screen of nearly 12,000 compounds in mouse brown adipocytes. We identified a family of compounds that increase Ucp1 expression and mitochondrial activity (including uncoupled respiration) in mouse brown adipocytes and human brown and white adipocytes. The mechanism of action may be through compound binding to A kinase anchoring protein (AKAP) 1, modulating

and regulation energy balance.
The increasing prevalence of obesity worldwide reflects changes in lifestyle, including a combination of increased food intake and reduced physical activity. Obesity causes complex metabolic, endocrine, and hemodynamic changes that may promote dyslipidemias, cardiovascular disease and type 2 diabetes (1,2). Because obesity develops when energy intake exceeds energy expenditure, increasing the latter is an attractive strategy to reduce body weight and fat storage (3,4).
There has been extensive interest in modulating thermogenesis as a treatment for obesity (5)(6)(7)(8)(9). During adaptive thermogenesis, particularly in response to cold exposure, mammals dissipate energy in brown adipose tissue (BAT) as heat by decreasing coupling between fatty acid oxidation and ATP synthesis as well as increasing mitochondrial biogenesis. Stored energy in fat is converted to heat, so changes in mitochondrial respiration or in the level of uncoupling activity could promote fat utilization. Recently, the existence of BAT in humans has been reappraised and there is good evidence that brown fat depots are active in adults and are capable of energy dissipation (10)(11)(12)(13). Moreover, adipocytes within white adipose tissue (WAT) may be induced to acquire brown adipocyte characteristics in both animals and humans by recruiting precursor cells or by transdifferentiation (6,(14)(15)(16)(17). These beige adipocytes derive from a distinct cellular lineage than brown adipocytes but have a similar multilocular morphology and high respiration rates. Thus, human BAT and browning of WAT may be important regulators of body fat accumulation/utilization and potential antiobesity drug targets.
One key factor in adaptive thermogenesis in BAT is the mitochondrial uncoupling protein-1 (UCP1). UCP1 is responsible for enabling the protein leak in mitochondria that dissipates energy resulting from oxidative metabolism (18,19). The presence of UCP1 in both classical brown adipocytes and beige adipocytes has spurred interest in targeting UCP1 as a means of increasing energy expenditure. Ucp1 expression is induced by stimulation of the sympathetic nervous system during cold exposure through activation of the b-adrenergic receptor. This leads to cAMP production and activation of cAMPdependent protein kinase (PKA), p38 mitogenactivated protein kinase (p38 MAPK), and transcription factors such as peroxisome proliferator-activated receptor gamma (PPARg), PPARg coactivator 1 alpha (PGC1a), and activating transcription factor 2 (ATF2) (18,(20)(21)(22). PKA activation by the second messenger cAMP is critical for the subsequent posttranslational modification of transcription factors that induce a thermogenic gene expression program. The PKA holoenzyme comprises 2 catalytic and 2 regulatory subunits (23). PKA is involved in several signaling pathways and acts in multiple tissues and subcellular compartments. The spatiotemporal organization of PKA activity is facilitated by scaffolding proteins, including Akinase anchoring proteins (AKAPs). AKAPs compartmentalize PKA to specific subcellular locations such as the cellular membrane, the nucleus, or the mitochondria, allowing distinct substrate phosphorylation and specific signal transmission (24)(25)(26). There has been an interest in targeting AKAPs to influence PKA activity (26,27).
While the thermogenic pathway and Ucp1 transcriptional effectors have been relatively well characterized, few small molecules have been identified that target Ucp1 expression or activation (7,(28)(29)(30). In addition, b3-adrenoceptor agonists such as CL316,243 increase thermogenesis in rats and mice, but they have poor efficacy on the human b3-adrenoceptor homolog (9,31). We hypothesized that an unbiased small molecule screen could identify compounds that regulate UCP1 levels or activity by mechanisms other than b3-adrenoceptor activation. We generated a brown adipocyte cell line containing a reporter for Ucp1 expression and screened nearly 12,000 small molecules for induction of reporter expression. We identified a family of compounds that effectively induces endogenous UCP1 levels in mouse brown adipocytes and human white adipocytes. A lead compound from this family promotes mitochondria-related gene expression, and activates PKA and lipolysis. Experiments based on protein stabilization suggest that the compound acts by binding AKAP1, thus modifying the PKA signaling pathway in adipocytes.

Identification of small molecules that induce Ucp1 expression
Transcriptional regulation of Ucp1 has been extensively documented (21,22). The transcriptional activation of Ucp1 is driven by two known regulatory regions: a proximal region next to the promoter and an enhancer element located 2.5 kb upstream of the transcription start site. By cross-species sequence comparisons, we noted that additional evolutionarily conserved sequences are present upstream and downstream of the enhancer region (Fig. 1A). To assess the functional significance of these conserved elements, we cloned different lengths of the mouse Ucp1 promoter (2.3 kb, 3 kb, 5 kb, and 7 kb) upstream of a luciferase reporter gene, and established stable brown adipocyte cell lines with each of these constructs. We tested each cell line for luciferase activity in response to known Ucp1 transcriptional activators: CL316,243 (synthetic b3-adrenergic agonist), forskolin (adenylyl cyclase activator), rosiglitazone (PPARg agonist), and retinoic acids (retinoic acid receptor agonists). The maximal luciferase activity was observed for the 3 kb Ucp1 promoter construct, which includes the known enhancer (Fig. 1B). Longer sequences had diminished activity, suggesting the presence of negative regulatory elements upstream of the enhancer. Based on these pilot studies, we selected the cell line expressing the 3 kb construct for our small molecule screen. We screened 11,712 compounds (final concentration of 10 µM) in duplicate from a combination of libraries (see Experimental procedures). Duplicate samples showed good reproducibility (Fig. S1A), providing confidence in the results, even though the Z score achieved with known Ucp1 inducers such as forskolin were modest (0.193 with forskolin).

Validation of compounds using endogenous Ucp1 expression
For further characterization, we selected compounds showing increased activation over vehicle of >55% for both duplicates (Fig. 1C). This group comprised 97 molecules, of which 30% had known functions, including several compounds with adrenergic agonist activity. To validate the compounds, we treated brown adipocytes with 92 compounds (excluding known adrenergic agonists) and measured endogenous Ucp1 expression by qPCR. Twenty two compounds induced Ucp1 expression >2-fold during two independent experiments (Table S1). Most of these compounds also increased expression of Ppargc1a and Pparg. We selected one compound with unknown function, AST 7062601, for further characterization, with the aims of understanding its mechanism of action.

AST070 and Z160 promote mitochondrial activation in brown and white adipocytes
We further characterized the effects of AST070 and Z160 on adipocyte metabolism by analyzing gene expression and mitochondrial respiration. Treatment of primary ( Fig. 2A) and immortalized brown adipocytes ( Fig. S3) with AST070 or Z160 promoted expression of markers of brown adipocyte identity (32,33), and of genes involved in mitochondrial function and fatty acid oxidation. Larger effects were observed in primary compared to immortalized cells. To investigate whether the compounds affected mitochondrial function, we treated brown adipocytes with AST070 and Z160 before assessing mitochondrial respiration with a Seahorse XF analyzer. Both AST070 and Z160 increased mitochondrial respiration, and more specifically, uncoupled respiration (Fig. 2B). Maximal respiration was also significantly elevated by AST070, and showed the same trend for Z160, suggesting an increase in mitochondrial reserve capacity. To confirm that the increase in uncoupled respiration was due to UCP1, we isolated mitochondria from cells treated with Z160. The GDP-sensitive respiration, which represent UCP1-dependent respiration, was increased similarly to the total uncoupled respiration, demonstrating the UCP1-dependent leak (Fig. 2C).
To validate these findings in a human cellular model, we used an immortalized human brown adipocyte cell line. Z160 induced UCP1 expression nearly 7-fold, and also increased expression of CIDEA (Fig. 3A). Importantly, Z160 also increased mitochondrial respiration (primarily uncoupled respiration), as well as maximal respiration, in the human brown adipocyte cell line (Fig. 3B) Given that white adipocytes have the ability to express brown adipocyte properties with specific metabolic stimuli, we also assessed whether Z160 influences brown adipocyte character in human white adipocytes. Interestingly, Z160 also induced UCP1 expression 7-fold, and greatly increased expression of CIDEA, ACADM, CPT1B and ELOVL3 (Fig. 3C). Similarly to the human brown adipocytes, Z160 also increased mitochondrial respiration (uncoupled), as well as maximal respiration, in the human white adipocyte cell line (Fig. 3D). Together, these results indicate that Z160 activates mitochondrial respiration in mouse brown adipocytes as well as in human brown and white adipocytes.

Global gene expression analysis highlights Z160 effects on energy metabolism
To assess the effects of Z160 on global transcription in mouse brown adipocytes, we performed microarray analysis of RNA isolated from immortalized brown adipocytes treated with either vehicle or Z160. Compared to vehicletreated cells, 581 and 504 probes were upregulated and down-regulated, respectively, by at least 1.5-fold (Table S2). Consistent with our qPCR results, Ucp1 was increased 2.5-fold in response to Z160, placing it in the top 20 upregulated genes. We performed functional annotation of the genes up-regulated or downregulated >1.5-fold by Z160 using the DAVID functional annotation tool (34). Z160 treatment increased expression of genes in mitochondrial categories (Fig. 4A), and down-regulated expression of genes that are distinct from mitochondrial function (Fig. 4B). In addition to Ucp1, up-regulated genes included five belonging to mitochondrial complex I (Ndufb2, Ndufb4, Ndufb5, Ndufb9, Ndufab1), one associated with complex II (Sdhd) as well as cytochrome C (Cycs), two subunits of complex IV (Cox6a2 and Cox7b), and two of complex V (Atp5a1, Atp5e). We confirmed results of the microarray using qPCR for representative genes from each electron transport chain complex (Fig.  4C). A slight enhancement in mitochondrial complex protein abundance was also observed in isolated mitochondria by Western blot (Fig. 4D).

Z160 stimulates thermogenesis in the mouse
We assessed the ability of Z160 to stimulate thermogenesis in vivo in C57BL/J mice. A single subcutaneous injection of the drug led to an increase in body temperature by 0.8ºC 24 h later (37.5ºC vs. 38.3ºC, p < 0.05), consistent with activated thermogenesis (Fig. 5A). Additionally, BAT from treated mice had elevated Ucp1 mRNA and protein levels, and enhanced expression of several genes implicated in mitochondrial function and lipolysis (Fig. 5B, C). No liver toxicity was observed as assessed by circulating aspartate aminotransferase (AST) levels (Fig. 5D). Plasma glucose levels were not affected by Z160 (172.8 ± 26.0 mg/dl vs. 157.6 ± 26.2 mg/dl, for vehicle and Z160-treated mice, respectively).

Z160 activates PKA
To understand the mechanism by which Z160 and related compounds enhance mitochondrial respiration, we treated brown adipocytes with Z160 in the presence of several known antagonists in the adrenergic receptor signaling pathway. Treatment with vehicle or non-selective a-(tolazoline, Tola) and b-(propranolol, Prop) adrenergic receptor antagonists did not prevent the induction of Ucp1 mRNA by Z160 (Fig. 6A). The same result was obtained with b3-adrenergic receptor antagonists (SR59230A, SR). By contrast, treatment with antagonists of PKA (H-89) or p38 MAPK (SB202190, SB) blocked the effect of Z160, suggesting that Z160 requires PKA activity to exert its effect on Ucp1 expression. Accordingly, the Z160 enhancement of Cidea and Elovl3 expression was also blunted by treatment with H-89 (Fig. 6B).
Based on these findings, we tested whether the AST070 or Z160 compounds influence PKA activity using a solid phase ELISA. First, we measured PKA activity in lysates from immortalized brown adipocytes after 30 min treatment with 50 µM AST070 or Z160. Treatment with either compound significantly increased PKA activity in an expected range (35,36) (Fig. 6C). Next, we treated lysate with different Z160 concentrations. Z160 concentration as low as 25 µM showed a significant increase in PKA activity (Fig. 6D). Importantly, 25 µM Z160 also stimulated PKA activation in mouse BAT or liver extracts (Fig.  6E). To confirm the results, we treated cell lysates for 10 min with PKA antagonist H-89 (500 µM) followed by 30 min incubation with Z160 (50 µM). Pre-incubation with the PKA antagonist reduced the PKA activity and negated the Z160 response ( Fig.6F).

Z160 promotes p38 MAPK phosphorylation and lipolysis
The demonstration that Z160 influences PKA activity prompted us to investigate its effects on pathways downstream of PKA activation. In adipocytes, PKA indirectly activates p38 MAPK and promotes lipolysis (37,38). Thus, the Z160-induced PKA activation was further characterized by first analyzing p38 MAPK phosphorylation. As depicted in Fig. 7A, Z160 caused p38 MAPK phosphorylation to a similar level as CL316,243. Additionally, Z160 induced lipolysis in brown adipocytes (Fig. 7B), and increased expression of lipolytic enzymes (adipose triglyceride lipase, Atgl; hormonesensitive lipase, Hsl) and FGF21 (a lipolytic mediator) at the mRNA level in both immortalized (Fig. 7C) and primary brown adipocytes (Fig. 7D).

Z160 modifies AKAP protein conformation and the mitochondrial PKA/AKAP interaction in brown adipocytes
Our data showing an effect of Z160 on PKA activity when added to cellular extracts (e.g., Fig. 6D, E) suggested an effect of the compound directly on PKA or functionally associated proteins. The PKA tetramer consists of two catalytic and two regulatory subunits. In the mouse, these are catalytic subunits C-a and C-b, and regulatory subunits RI and RII, each with a and b isoforms (RI-a, RI-b, RII-a, RII-b). PKA is also bound to a family of anchoring proteins, A-kinase anchoring proteins (AKAPs), which allow the compartmentalization of cAMP signaling. To investigate whether Z160 binds PKA complex proteins, we applied two techniques that rely on ligand-induced alterations in protein stability: Cellular Thermal Shift Assay (CETSA) and Drug Affinity Responsive Target Stability (DARTS). CETSA is based on the principle that ligand binding will change the temperature at which a protein starts to unfold and aggregate (39), while DARTS relies on the protection of a protein from proteolysis upon specific binding to a small molecule (40).
We examined the potential interaction of Z160 with PKA subunits and AKAPs. First, we identified which PKA subunits and AKAP isoforms are normally present in BAT. Robust protein levels were observed in BAT for PKA Ca and RII-b subunits, AKAP1, and AKAP6 (also known as mAKAP) (Fig. 8A). To perform the CETSA assays, we incubated cultured brown adipocytes with 10 µM Z160 for 8 h, then analyzed the thermal shift by Western blot. The PKA subunits expressed in BAT (C-a and RII-b) did not show differences in denaturation in the presence of Z160 (Fig. 8B). However, AKAP1 showed altered protein stability in the presence of Z160: Z160 increased the heat stability of AKAP1 (compare vehicle and Z160 heated at 55.8-63.3ºC), suggesting that Z160 may interact with this protein (Fig. 8B). By contrast, Z160 did not influence the heat stability of AKAP6, p38 MAPK, or cGMP-dependent kinase 1 (PKG-1).
To provide further evidence for an interaction between Z160 and AKAP1, we performed DARTS assays. Lysates from brown fat were treated with different Z160 concentrations for 1 h followed by pronase digestion and immunoblotting. Representative data shown in Fig. 8C indicate that the presence of Z160 altered pronase susceptibility of AKAP1 (compare with and without Z160 at a pronase concentration of 1:4000). The protease susceptibility of other proteins tested was not influenced by the presence of Z160. The effect of Z160 on AKAP1 was confirmed using mouse brown adipocyte extracts (Fig. S4A). Importantly, the effect of AST070 or Z160 on AKAP1 was not observed with CL316,243, a chemically unrelated agent known to stimulate Ucp1 expression (Fig. S4B).
AKAP1 is known to localize PKA to the surface of mitochondria and to relay cAMP signaling for mitochondrial functions (41,42). We wondered whether treatment with our compounds could alter the binding of AKAP1 to mitochondria. To answer this question, we treated brown adipocytes with Z160 and isolated mitochondria at different times. In the presence Z160, AKAP1 protein levels were increased in the mitochondrial fraction after 6-8 h, suggesting that it may influence the localization of AKAP1 (Fig. 8D). In parallel, PKA subunits showed increased mitochondrial association after 6-8 h of Z160 treatment. By contrast, the levels of CYTC and GAPDH in the mitochondrial fraction were not altered by the compound. Total AKAP1 protein levels were not altered by Z160, suggesting that the increased AKAP1 levels on mitochondria following treatment is due to AKAP1 translocation rather than an increase in protein synthesis (Fig. 8E). Finally, we assessed whether Z160 enhances interaction between AKAP1 and PKA subunits, a mechanism that may promote the localization of PKA at mitochondria. Brown adipocytes were treated with Z160 or AST070 for 7 h and cell lysates were precipitated by anti-AKAP1 antibody and immunoblotted with anti-PKA C-a and RII-b (Fig. 8E). The signal for PKA subunit precipitation by AKAP1 was slightly enhanced after treatment, especially with Z160, suggesting that the compounds increase the interaction between AKAP1 and the PKA subunits.
To further assess the requirement of AKAP1 to mediate the effects of Z160, we inactivated AKAP1 in an immortalized brown adipocyte cell line (AKAP1wt) using CRISPR/Cas9 gene editing. One clone (AKAP1mut) was selected based on very low AKAP1 protein levels (Fig. 9A). Unlike AKAP1wt cells, treatment of AKAP1mut cells did not respond to Z160 with increased Ucp1 expression (Fig. 9B). Additionally, the Z160-induced gene expression was blunted or reduced in AKAP1mut for Hoxa5, Cidea, Elovl3, Acox1, and Cpt1b. However, Z160 increased Ppargc1a expression in both AKAP1wt and AKAP1mut cells, suggesting that Z160 may exert some effects that are AKAP1-dependent, and others that are AKAP1-independent. Finally, we treated brown adipocytes with Z160 before assessing mitochondrial respiration. Z160 increased mitochondrial and uncoupled respiration in AKAP1wt but not in AKAP1mut cells (Fig. 9C). In contrast, Z160-induced maximal respiration was significantly elevated in AKAP1mut cells.
To confirm that the increase in uncoupled respiration was due to UCP1, we isolated mitochondria from cells treated with Z160. Uncoupled and UCP1-dependent respiration were significantly increased by Z160 in AKAP1wt but not in AKAP1mut mitochondria (Fig. 9D).

Discussion
In the present study, we identified a new compound family with the ability to increase mitochondrial respiration in mouse brown adipocytes and human white adipocytes. Furthermore, in vivo administration led to increased mitochondrial activity and thermogenesis. The compounds influenced the expression of many genes involved in mitochondrial and fat oxidation processes (e.g., Ppargc1a, Cpt1b, Elovl3, Cidea, and Acox1), indicating that Z160 may influence energy utilization through means that are UCP1independent. Indeed, we identified PKA activation as a mechanism of action of these compounds, with particular effects on AKAP proteins and subcellular localization of PKA (Fig.  10). Our results highlight AKAP1 as a potential novel target for increasing cellular energy expenditure.
PKA is a key node in the b-adrenergic signaling pathway, controlling many cellular processes, including gene expression, lipolysis and lipogenesis. It is thought that upon binding of cAMP to the two regulatory subunits of the PKA heterotetramer, the catalytic subunits of PKA are released and activated. Our findings are consistent with an effect of Z160 and related compounds on PKA activation, as they caused increased p38 MAPK phosphorylation, enhanced lipolysis, and induction of Ucp1 expression. A proportion of PKA within the cell is associated with mitochondria (43), and PKA can increase electron transport chain activity by direct phosphorylation of complex I and IV (44,45). AKAP1 and its splice variants localize to mitochondria and interact with type I and type II regulatory subunits of PKA. Our findings show that Z160 and AST070 increase the amount of AKAP1 in the mitochondrial fraction and increase PKA/AKAP protein interaction. Possible mechanisms include the compounds promoting the interaction of AKAP at the mitochondrial surface with PKA, or increased localization of PKA/AKAP complexes to the mitochondria. AKAP1 and the RII-b subunit of PKA colocalize in mitochondria of white adipocytes (46). Our tissue distribution analysis indicates that AKAP1 and the RII-b are both present at higher levels in BAT than WAT, consistent with a role in mitochondria-rich cells.
Our studies showed altered protein conformation of AKAP1 in the presence of Z160, suggesting a direct interaction between Z160 and AKAP1. Although it might be expected that binding of a compound to a protein will enhance stability, in the DARTS assays we observed increased protease sensitivity of AKAP1 in the presence of Z160. There are several possible explanations. For example, Z160 binding could disrupt the tridimensional structure of the AKAPs and make specific domains more accessible to protease digestion. The actual binding domain may still be stabilized by Z160, but not detectable with the specific antibody employed. It is also possible that Z160 binds another protein associated with the PKA/AKAP complex, and thereby modifies the protein complex to stimulate local PKA activity. As noted, our AKAP1mut brown adipocyte cell line attenuated several effects of Z160, including induction of several genes (Cpt1b, Elovl3, Cidea, Acox1, and Ucp1), as well as uncoupled respiration. However, AKAP1 was not required for Z160 effects on Ppargc1a expression, nor total respiration. prevented Z160-induced Ucp1 mRNA induction dramatically, but did not affect other Z160targeted genes to the same extent, nor maximal respiration. This suggests that AKAP1 is required for some, but not all, effects of Z160. Nevertheless, the two techniques used here (CETSA and DARTS) identified AKAP1 (and ruled out PKA subunits) as the best target. Future studies to identify Z160 targets using highthroughput approaches may shed additional light on this compound's mode of action.
The induction of thermogenesis is an appealing approach to influence energy expenditure and reduce obesity. Pharmacological agents that increase metabolic rate by increasing mitochondrial uncoupling (such as 2,4dinitrophenol) have been dismissed for lack of tissue specificity and severe side effects (47). Targeting proteins downstream of PKA, such as p38 MAPK and PPARg, has limitations due to pleiotropic effects (48)(49)(50). Therefore, there is a real need to identify new mechanisms to modulate this pathway (9). The AKAP proteins, which control the spatiotemporal activation of PKA in a tissue-dependent manner, may offer an alternative therapeutic target for promoting energy expenditure. AKAP1 is the most abundant AKAP in adipose tissue (51), and plays a role in mitochondrial-related PKA activity, as well as in lipoprotein lipase expression (25,41). Furthermore, AKAP1 appears to be an amenable target to increase mitochondrial activity and lipolysis in both brown and white adipocytes, and makes it an attractive candidate for therapeutic applications related to obesity. Further studies will investigate whether the compounds can have a long-term effect on diet-induced obesity in mice. However, we acknowledge that modulation of AKAP1 and mitochondrial uncoupling may have unpredictable and detrimental effects in tissues other than brown adipose tissue, such as heart. Thus, further development for therapeutic applications would require evaluation of effects in other organs, and may require targeted delivery to brown adipose tissue.
In summary, we have identified new compounds with the ability to promote uncoupled respiration in mouse brown adipocytes and human white adipocytes, and to promote thermogenesis in the mouse. These compounds influence a broad range of gene expression related to mitochondrial respiration, and appear to exert some actions independently of UCP1. These include previously unexploited mechanisms of action that involve modulation of PKA/AKAP interactions and/or localization to mitochondria, which may overcome some limitations in existing compounds.

Experimental procedures Cell culture
Primary brown adipocytes were isolated as described previously (52). An established mouse brown adipocyte cell line was obtained from Dr. Bruce Spiegelman (Dana-Farber Cancer Institute, Boston, USA) (53). To differentiate primary and immortalized brown adipocytes, cells were grown until confluency in DMEM containing 10% FBS, 25 mM glucose 1 mM pyruvate, 2 mM glutamine, 20 nM insulin, 1 nM T3, and antibiotics (53). Once confluent (day 0), differentiation was induced by supplementing the medium with 0.5 mM isobutylmethylxanthine (IBMX), 0.5 µM dexamethasone, and 0.125 mM indomethacin, for 2 days. Differentiation was continued in the original medium (without IBMX, dexamethasone and indomethacin) for 7-10 days.
Human immortalized brown and white preadipocytes were obtained from Dr. Yu-Hua Tseng (Joslin Diabetes Center, Harvard Medical School, USA). The generation of the two cell lines from human neck adipose tissue biopsies and immortalization was performed as previously described (54)

Generation of brown adipocyte UCP1reporter cell lines
Different lengths of mouse Ucp1 promoter were cloned into pGL3-basic vector (Promega) by PCR, using the NheI and XhoI sites. We PCR amplified Ucp1 promoter fragments of 2.3 kb (cttgatgtgtggagctgagtagc), 3 kb (gtgccgtcactaacagtactg, 5 kb (ctgcagactcctgacacagct), and 7 kb (ggaaagtggttcagtttgattagaagg) plus 94 bases downstream of the transcription start site (reverse primer: ctaggtagtgccagtgcagag). To perform chemical library screening in mature brown adipocytes, we created stable cell lines in the immortalized mouse brown adipocytes (53). For this purpose, we introduced a neomycin cassette from pcDNA3.1/V5-His vector (Invitrogen) into the SalI site of pGL3-basic plasmid. All constructs were verified by sequencing. Stable cell lines were selected with 500 µg/ml G418.

Luciferase assay for pilot studies prior to drug screen
Stable brown adipocyte cell lines carrying Ucp1 promoter-luciferase constructs were seeded in 96-well plates and differentiated for 7-10 days. Cells were treated with either 10 nM CL316,243, 10 µM forskolin, 1 µM rosiglitazone, or 1 µM cis-and trans-retinoic acid, overnight. Luciferase activity was assayed using LAR II or Bright-Glo luciferase assay system (Promega). For the LAR II assays, passive lysis was performed on the cells. Luminescence was measured with a GloMax Luminometer. We did not notice differences in response between LAR II and Bright-Glo and the latter was used for the drug screen.

Small molecule library screen
The screen was conducted at the UCLA Molecular Screening Shared Resource using automated instruments. Immortalized brown adipocytes were differentiated in T225 flasks and plated in Matrix 384-well plates (26,000 cells per well) with white flat bottom (ThermoScientific) using trypsin and collagenase type II (Sigma C6885). The small molecule libraries screened were a BioMol library (204 compounds), an FDA-approved drug library (1120 compounds), a Microsource spectrum collection (2000 compounds), and a druggable compound set (8000 compounds). Molecules were delivered at 10 µM final concentration in DMSO. The screen was performed in duplicate on different days. After 18 h, luciferase activity was measured with Bright-Glo luciferase assay system and an LJL instrument. Data were normalized to the basal response (100% activity) in the presence of DMSO. Following the primary screen, 92 molecules were selected for validation and used to treat brown adipocyte cells plated in a 96-well plate overnight. RNA was extracted with an SV 96 Total RNA Isolation System (Promega). Compounds used for follow-up studies were identified using search tools available from Molport (molport.com), and ordered from the same company.

Cellular bioenergetics
Cellular respiration was measured using a Seahorse XF24 or XF96 analyzer (Agilent), as previously published (55). Immortalized brown adipocytes were differentiated in 6-well plates and replated in the XF24 plates at a density of 50,000 cells per well using trypsin and collagenase type II. Cells were treated with vehicle (DMSO) or compounds for 18-24 h. Oxygen consumption rates were obtained before and after the sequential injection of 0.75 µM oligomycin, 0.5 µM FCCP, and 0.75 µM rotenone/myxothiazol. Results were normalized to total protein. For human white adipocytes, cells were cultured as described above and differentiated in XF96 microplates. DMSO or Z160 (10 µM) was added at day 17 and for 4 days. Oxygen consumption rates were measured before and after the sequential injection of 1 µM oligomycin, 1 µM FCCP, and 0.5 µM rotenone/antimycin A. For these cells, results were normalized to cell count determined by nuclei fluorescent staining with Hoescht staining imaged on Cytation 5 Imaging Reader and analysis with Gen5 software (BioTek). Respiration from isolated mitochondria were performed with 10 mM succinate as described previously (56,57). Oxygen consumption rates were obtained before and after the sequential injections of 2.5 µM oligomycin and 1.5 µg/ml antimycin A. UCP1-dependent respiration was measured by injecting 1 mM GDP and represents the GDP-sensitive respiration. For both mouse and human cell lines, mitochondrial respiration was calculated by subtracting the nonmitochondrial respiration present after the last injection (rotenone/myxothiazol or rotenone/antimycin A). Uncoupled respiration corresponds to the respiration difference between oligomycin and the last injection. Maximal respiration was determined after FCCP injection.

Gene expression analysis
Mouse RNA levels were measured by qPCR as described previously (56). Data were normalized to B2m and Tbp reference genes. Primers are listed in Table S3. Human RNA was extracted using an RNA Mini Plus kit (Zymo) and gene expression determined by Taqman assay. The 1-step qPCR was run on a QuantStudio 12K Flex Real Time PCR System (ThermoFisher) using the following protocol: 50ºC for 5 minutes, 95ºC for 20 seconds, 40 cycles of 95ºC for 15 seconds and 60ºC for 60 seconds. Data were normalized to PPIA and PSMB2 reference genes. Taqman probes (ThermoFisher Scientific) are listed in Table S3. For global gene expression analysis, RNA isolated from brown adipocytes treated with compounds as indicated (four biological replicates) was hybridized to Illumina mouse Ref 8 V2.0 bead chips at the University of California, Los Angeles Neuroscience Genomics Core as previously described (56). Data were processed with GenomeStudio V2011.1 using the quantile normalization, background subtraction, and a present call of P < 0.05. Immunoblot analysis. Cells and tissues were lysed in 10 mM Tris pH 7.5, 10 mM NaCl, 1 mM EDTA and 0.5% Triton X-100, supplemented with complete mini EDTA-free protease (Roche Diagnostics) and phosphatase (Cocktail 2 and 3, Sigma) inhibitors, followed by 10 second sonication. For analysis of mitochondrial proteins, mitochondria were isolated from cells by dual centrifugation, as described (57). Protein lysates were separated by SDS-PAGE and transferred to a nitrocellulose membrane. Transfer was confirmed by Ponceau staining (P7170, Sigma). After blocking in 5% milk, 0.1% Tween-20 in Tris-buffered saline (TBS), primary antibody was incubated overnight at 4ºC in 5% bovine serum albumin and 0. Peroxidase goat anti-rabbit (sc-2030, Santa Cruz Biotechnology, Inc) or rabbit antimouse (A9044, lot 115K4811, Sigma) secondary antibody was used at a 1:10,000 dilution for 1 h at room temperature in 5% milk and 0.1% Tween-20 in TBS. Immunoreactive bands were revealed with ECL prime (Amersham) and visualized with a Bio-Rad Gel-doc imager. Quantification using representative bands from the Ponceau staining was performed with ImageJ (58).

Co-immunoprecipitation
Immortalized brown adipocyte cells were differentiated for 10 days and treated with 10 µM AST070 or Z160 for 7 h. Cells were lysed in 150 mM NaCl, 50 mM Tris pH 7.5, 1% Nonidet P40, 0.5% Na-deoxycholate containing protease and phosphatase inhibitors. Cell lysates were incubated overnight at 4ºC with 2 µl of anti-AKAP1 antibody. Twenty microliter of Protein A/G PLUS agarose beads (Santa Cruz Biotechnologies) were added for 2 h at 4ºC. After an initial wash (500 mM NaCl, 50 mM Tris pH 7.5, 0.1% Nonidet P40, 0.05% Na-deoxycholate) and a final wash (10 mM Tris pH 7.5, 0.1% Nonidet P40, 0.05% Na-deoxycholate), proteins were eluted from the beads with 1x loading buffer and 2% b-mercaptoethanol, boiled for 10 min and analyzed by Western blot. To avoid detecting the IgG heavy chain, TidyBlot Western Blot reagent:HRP (Bio-Rad) at 1:100 was used as secondary antibody to reveal the PKA subunits from the co-immunoprecipitation blot.

Assessment of PKA activity
PKA kinase activity was measured with an ELISA that utilizes a synthetic peptide as substrate for PKA and a polyclonal antibody that recognizes the phosphorylated form of the substrate (ab139435, Abcam). PKA activity was measured in lysates from cultured brown adipocytes, mouse BAT, or liver. Cells and tissues were lysed in 20 mM MOPS, 5 mM EGTA, 2 mM EDTA, and 0.1% Triton X-100 supplemented with protease and phosphatase inhibitor as described above.
Protein concentration was determined by Bradford assay and 10 µg of cells, 0.25 µg of BAT extract, or 0.5 µg of liver extract were assessed for PKA activation according to the manufacturer's instruction, with minor changes. Briefly, vehicle or Z160 was incubated with brown adipocyte cells overnight or with protein lysates for 30 min, followed by one wash before the primary antibody, 60 min incubation with the primary antibody, 30 min incubation with the secondary antibody, and by 10 min washing. Absorbance was measured after the substrate was added for 20-60 min, depending the intensity of the signal.

Lipolysis assay
Lipolysis was assessed in cultured cells using medium collected over 3 h with the Adipolysis assay kit (AB100, Millipore), according to the manufacturer's protocol.

Cellular thermal shift assay (CETSA)
CETSA assays were performed as described (59). Briefly, differentiated brown adipocytes (one 10 cm dish per treatment) were treated with vehicle or compounds, trypsinized, counted, and resuspended in approximately 450 µl PBS containing protease inhibitors (volumes were adjusted to have the same number of cells in each treatment). For each sample, 18 µl was distributed into each of 7 PCR tubes. Samples were heated using a thermocycler with a temperature gradient (iCycler, Biorad) for 3 min, followed by 3 min at room temperature, and then snap-frozen in liquid N2. After two freeze-thaw cycles, samples were centrifuged at 20,000 g for 15 min at 4ºC, and the supernatants were transferred to another set of tubes. Proteins were analyzed by immunoblotting.
Drug affinity responsive target stability (DARTS) DARTS assays were performed as described (40). Briefly, BAT or cells within a 10 cm dish were lysed with 600 µl M-PER buffer containing protease and phosphatase inhibitors. Debris was pelleted by centrifugation at 18,000 g for 10 min at 4ºC, and the lysates were harvested and supplemented with TNC buffer (50 mM Tris-HCl, 50 mM NaCl, 10 mM CaCl 2). The lysates were split into two tubes and treated with either vehicle or 20 µM Z160 for 1 h at room temperature. For experiment with BAT, 100 µg of lysate was incubated with different Z160 concentrations. Different concentrations of pronase (from 1:100 to 1:10,000 dilution from a 1.25 mg/ml pronase stock, #10165921001, Roche) were incubated with the samples for 30 min at room temperature and the reaction was stopped with SDS loading buffer followed immediately by heating at 70ºC for 10 min. Samples were analyzed by immunoblotting.

Animal experiments
All mouse studies were conducted in accordance with and approved by the Institutional Animal Research Committee of the University of California, Los Angeles. C57BL/6J male mice were obtained from the Jackson Laboratory. For the drug injection, Z160 (diluted in 100 µl DMSO) was injected subcutaneously, near BAT, at 1.5 mg/kg body weight and compared with vehicle alone. Plasma and tissues were obtained after 20 h. Body temperature was obtained with a rectal probe (BAT-12, Physitemp). Aspartate aminotransferase (AST) activity was determined from plasma according to the manufacturer's protocol (MAK055, Sigma).

Statistical analyses
Statistical analyses were performed by unpaired 2-tailed Student's t test. A value of p < 0.05 was considered significant.