Salsalate Activates Brown Adipose Tissue in Mice

1544 Diabetes Volume 64, May 2015

Andrea D. van Dam,1,2 Kimberly J. Nahon,1,2 Sander Kooijman,1,2 Susan M. van den Berg,3 Anish A. Kanhai,1,2 Takuya Kikuchi,1,2 Mattijs M. Heemskerk,2,4 Vanessa van Harmelen,2,4 Marc Lombès,5 Anita M. van den Hoek,6 Menno P.J. de Winther,3 Esther Lutgens,3,7 Bruno Guigas,8,9 Patrick C.N. Rensen,1,2 and Mariëtte R. Boon1,2

Diabetes 2015;64:1544–1554 | DOI: 10.2337/db14-1125

Salsalate improves glucose intolerance and dyslipid- Salsalate is a nonsteroidal anti-inflammatory drug belong- emia in type 2 diabetes patients, but the mechanism is ing to the salicylate class of drugs. Salicylates are originally still unknown. The aim of the current study was to derived from plants, in which they function as part of the unravel the molecular mechanisms involved in these immune system to combat infections. Nowadays, synthetic beneficial metabolic effects of salsalate by treating mice compounds that break down to salicylates in vivo, including with salsalate during and after development of high-fat aspirin and salsalate, have largely replaced salicylate (1). In – diet induced obesity. We found that salsalate attenu- humans, salicylates have strong anti-inflammatory effects, – ated and reversed high-fat diet induced weight gain, and have therefore been applied in the clinic for several in particular fat mass accumulation, improved glucose decades to treat pain and inflammation caused by rheuma- tolerance, and lowered plasma triglyceride levels. toid arthritis (2,3). Mechanistically, salsalate selectively promoted the up- Studies have shown that salicylates exhibit beneficial take of fatty acids from glycerol tri[3H]oleate-labeled metabolic effects as well. A recent trial (4) has demon- lipoprotein-like emulsion particles by brown adipose tis-

METABOLISM strated that salsalate lowers the levels of HbA ,fasting sue (BAT), decreased the intracellular lipid content in 1c BAT, and increased rectal temperature, all pointing to blood glucose, and circulating triglycerides (TGs) in type more active BAT. The treatment of differentiated T37i 2 diabetes patients. Furthermore, salsalate increases energy brown adipocytes with salsalate increased uncoupled expenditure (EE) in human subjects (5). In contrast to respiration. Moreover, salsalate upregulated Ucp1 ex- aspirin, salsalate is not associated with an increased risk pression and enhanced glycerol release, a dual effect of gastrointestinal bleeding and therefore is relatively safe that was abolished by the inhibition of cAMP-dependent for long-term clinical experience. Thus, it is considered protein kinase (PKA). In conclusion, salsalate activates a promising antidiabetic drug (3). The mechanisms un- BAT, presumably by directly activating brown adipo- derlying the beneficial metabolic effects of salsalate re- cytes via the PKA pathway, suggesting a novel mecha- main largely unknown, partly because its receptor has nism that may explain its beneficial metabolic effects in not yet been identified (6). Salicylates have been shown type 2 diabetes patients. to activate AMPK in liver, muscle, and white adipose

1Department of Endocrinology and Metabolic Diseases, Leiden University Medical 9Department of Parasitology, Leiden University Medical Center, Leiden, the Center, Leiden, the Netherlands Netherlands 2 Einthoven Laboratory for Experimental Vascular Medicine, Leiden, the Corresponding author: Andrea D. van Dam, [email protected]. Netherlands Received 23 July 2014 and accepted 24 November 2014. 3Department of Medical Biochemistry, Academic Medical Center, Amsterdam, the Netherlands This article contains Supplementary Data online at http://diabetes 4Department of Human Genetics, Leiden University Medical Center, Leiden, the .diabetesjournals.org/lookup/suppl/doi:10.2337/db14-1125/-/DC1. Netherlands A.D.v.D. and K.J.N. contributed equally to this study. 5Institut National de la Santé et de la Recherche Médicale, Unité 693, Le Kremlin- P.C.N.R. and M.R.B. contributed equally to this study. Bicêtre, France fi 6Department of Metabolic Health Research, Netherlands Organisation for Applied M.R.B. is currently af liated with the Department of Human Biology, Maastricht Scientific Research, Leiden, the Netherlands University, Maastricht, the Netherlands. 7Institute for Cardiovascular Prevention, Ludwig Maximilian’s University Munich, © 2015 by the American Diabetes Association. Readers may use this article as Munich, Germany long as the work is properly cited, the use is educational and not for profit, and 8Department of Molecular Cell Biology, Leiden University Medical Center, Leiden, the work is not altered. the Netherlands diabetes.diabetesjournals.org van Dam and Associates 1545 tissue (WAT), suggesting a role for this energy-sensing ki- Glucose was measured using an enzymatic kit from nase in the mechanism of action of the drug (1). However, Instruchemie (Delfzijl, the Netherlands), and insulin was salicylate still improves glucose metabolism in mice lacking measured by ELISA (Crystal Chem Inc., Downers Grove, IL). b the regulatory AMPK- 1 subunit, suggesting involvement Intravenous Glucose Tolerance Test of other pathways as well (1). Mice were fasted for 6 h, a baseline blood sample was The objective of the current study was to investigate obtained, and mice were intravenously injected with 10 the mechanisms underlying the beneficial effects of mL/g body weight glucose dissolved in PBS (200 mg/mL). salsalate on lipid and glucose metabolism by treating Additional blood samples were taken at t = 5, 15, 30, 60, APOE*3-Leiden.CETP (E3L.CETP) transgenic mice, a well- 90, and 120 min. Capillaries were placed on ice and centri- established model for human-like lipoprotein metabolism fuged, and glucose levels were measured as described (7–9), with salsalate mixed through the high-fat diet above. (HFD). We found that salsalate both prevented and reduced HFD-induced weight gain by lowering fat mass ac- In Vivo Clearance of Radiolabeled Lipoprotein-Like cumulation, and improved glucose and lipid metabolism. Emulsion Particles Mechanistic studies showed that these effects were ac- Lipoprotein-like TG-rich emulsion particles (80 nm) labeled 3 3 companied by increased activity of brown adipose tissue with glycerol tri[ H]oleate (triolein, [ H]TO) were prepared (BAT). Taken together, our data indicate that BAT may and characterized as described previously (13). Mice were contribute to the beneficial effects of salsalate on lipid fasted for 6 h (from 7:00 A.M. to 1:00 P.M.) and injected with m and glucose metabolism, and corroborate previous find- 200 L of emulsion particles (1.0 mg TG per mouse) via the ings that targeting BAT may be a valuable strategy for tail vein (t =0).After15min,micewerekilledbycervical correcting metabolic derangements. dislocation and perfused with ice-cold PBS through the heart. Thereafter, organs were harvested and weighed, and RESEARCH DESIGN AND METHODS the uptake of [3H]TO-derived radioactivity was quantified Mice, Diet, and Salsalate Treatment and expressed per gram of wet tissue weight. – E3L.CETP mice were obtained as previously described (7 Rectal Temperature Measurement 9). To assess the effect of salsalate on progression of obe- Rectal temperature was measured between 3:00 and 4:00 sity, dyslipidemia, and hyperglycemia, 10-week-old male P.M. using a rectal probe attached to a digital thermometer E3L.CETP mice were randomized to receive an HFD (BAT-12 Microprobe Thermometer; Physitemp, Clifton, NJ). (45% kcal lard fat; Research Diets) without or with 0.5% (weight for weight [w/w]) salsalate (2-carboxyphenyl salic- Histology ylate; TCI Europe N.V.) for 12 weeks. To assess the effect of Interscapular BAT (iBAT) and gonadal WAT (gWAT) were fi salsalate on the regression of obesity and associated met- removed, xed in 4% paraformaldehyde, dehydrated in fi abolic disorders, 10-week-old, male, diet-induced (12 weeks 70% EtOH, and embedded in paraf n. Hematoxylin-eosin on HFD) obese E3L.CETP mice received salsalate (0.5% w/w) (H-E) staining was done using standard protocols. The fi for 4 weeks. To investigate the effect of salsalate indepen- area of intracellular lipid vacuoles in BAT was quanti ed dent of the E3L.CETP background in a progression set- using ImageJ (National Institutes of Health). ting, 10-week-old, male, wild-type (WT) mice (C57BL/6J Quantification of TG Content in BAT background; Charles River Laboratories) were randomized Lipids were extracted from BAT following a protocol to receive an HFD without or with salsalate for 4 weeks. modified from Bligh and Dyer (14). BAT samples (~50 mg) Mice were individually housed at 21 or 28°C (WT mice). were homogenized in 10 mL ice-cold CH OH/mg tissue. To mask the bitter taste of salsalate, anise (3.33% w/w) 3 Lipids were extracted into an organic phase by the addi- was added to the diet of both groups in all studies (10). tion of 1,800 mLCHOH:CHCl (1:3 volume for volume) Mouse experiments were performed in accordance with the 3 3 to 45 mL homogenate and subsequent centrifugation. The Institute for Laboratory Animal Research Guide for the lower organic phase was evaporated, lipids were resus- Care and Use of Laboratory Animals and have received pended in 2% Triton X-100, and TG content was assayed approval from the University Ethical Review Board (Leiden (see above). BAT lipids were reported per milligram of University Medical Center, Leiden, the Netherlands). protein (BCA Protein Assay Kit; Pierce). Body Weight and Body Composition Measurements Isolation of Stromal Vascular Fraction and Flow Body weight was measured with a scale, and body Cytometry composition was measured using an EchoMRI-100 ana- The gWAT was removed, rinsed in PBS, and minced. lyzer (EchoMRI, Houston, TX). Tissues were digested in a collagenase mixture (DMEM Determination of Plasma Parameters with 20 mmol/L HEPES, Collagenase XI, and Collagenase Upon randomization and at 4-week intervals during I; Sigma) for 45 min at 37°C and passed through a 70-mm treatment, blood samples were collected after a 6-h fast nylon mesh. The suspension was centrifuged (6 min, and assayed for levels of TGs, total cholesterol (TC), and 1,250 revolutions per minute), and the pelleted stromal free fatty acids (FFA), as described previously (11,12). vascular fraction (SVF) was resuspended in FACS buffer. 1546 Salsalate Activates Brown Adipose Tissue Diabetes Volume 64, May 2015

Fc blocking (CD16/32 antibody) was performed prior to a normalized to b2-microglobulin (b2m) and 36b4 mRNA 30 min staining with fluorescently labeled primary anti- content and expressed as a fold change compared with con- bodies for CD45, CD11b, Ly6G, F4/80, CD11c, CD206, trol mice using the DDCT method. The primer sequences Gr-1, CD19, CD3, CD4, CD8, CD25, and FoxP3 (BioLegend, used are listed in Supplementary Table 2. fl e-Bioscience, or BD Biosciences). SVF was analyzed by ow Statistical Analysis cytometry with a BD FacsCANTO II flow cytometer and All data are expressed as the mean 6 SEM, unless stated FlowJo software. otherwise. Groups were compared with a two-tailed un- Analysis of DNA Content paired Student test and were considered statistically sig- DNA content in gWAT samples was quantified as pre- nificant if P , 0.05. viously described (15). RESULTS Culture and Differentiation of White and Brown Salsalate Prevents HFD-Induced Obesity Adipocytes To investigate the effect of salsalate on the development 3T3-L1 (American Type Culture Collection, Manassas, VA) of obesity, male E3L.CETP mice were fed an HFD with or and T37i cells (16,17) were cultured and differentiated without salsalate for 12 weeks (i.e., progression study). as described previously. Differentiated cells were treated Salsalate prevented the HFD-induced increase in body with salsalate (SML0070; Sigma), sodium salicylate (S2007; mass seen in the control group (280%, 1.6 6 2.6 vs. Sigma), AICAR (ab120358; Abcam), norepinephrine (A7257; 7.9 6 5.5 g, P , 0.01; Fig. 1A), which was due to a de- Sigma), CL316243 (Tocris Bioscience, Bristol, U.K.), or 2 6 6 vehicle (DMSO) for 15 min or 8 h. H89 (H-5239, LC creased gain in fat mass ( 60%, 3.1 2.4 vs. 7.8 4.2, P , 0.05; Fig. 1B) rather than lean mass (Fig. 1C). Al- Laboratories) was added 1 h before initiation of salsalate though salsalate persistently prevented mice from gaining treatment. Supernatant was collected for the determina- fat mass throughout the treatment period, salsalate de- tion of glycerol concentration (Instruchemie), and cells creased food intake only during the first 3 days (Fig. 1D). were harvested for RNA or protein analysis as described below. Salsalate Prevents HFD-Induced Deterioration of Glucose and TG Metabolism Oxygen Consumption Measurements Next, we assessed whether salsalate protects mice from A Seahorse Bioscience XF96 extracellular flux analyzer developing HFD-induced glucose intolerance and dyslip- (Seahorse Bioscience, North Billerica, MA) was used to idemia. Although salsalate did not affect fasting glucose measure the oxygen consumption rate (OCR) in differen- levels (Fig. 2A), it reduced insulin levels after 4 weeks tiated T37i cells. On day 8 of differentiation, cells were (241%, P , 0.05), 8 weeks (247%, P =0.07),and12 trypsinized and seeded in a 96-well Seahorse Bioscience weeks (267%, P , 0.05; Fig. 2B)oftreatment.Inaddition, assay plate. The next day, the ATP synthase inhibitor salsalate improved glucose tolerance (Fig. 2C), as evidenced oligomycin, vehicle, salsalate, and CL316243 were pre- by a reduction of the area under the curve (AUC) of glucose loaded in the reagent delivery chambers and pneumati- concentrations during an intravenous glucose tolerance cally injected into the wells. Oligomycin was injected to test (AUC 225%, P , 0.05; Fig. 2D). Salsalate did not a final concentration of 1.5 mmol/L, and the OCR was affect plasma TC levels (Fig. 2E) throughout the treatment measured three times after mixing. Then, vehicle (0.1% period but decreased plasma TG levels at 8 weeks (243%, DMSO), salsalate, and/or CL316243 were injected, and P , 0.01) and 12 weeks (247%, P , 0.05; Fig. 2F). the OCR was measured six times in 30 min. All OCR mea- To elucidate which metabolic organs were involved in surements were normalized to cell count using Cyquant the TG-lowering effect of salsalate, the tissue-specific (Invitrogen). uptake of fatty acids (FAs) derived from intravenously Western Blot Analysis injected [3H]TO-labeled lipoprotein-like emulsion par- Pieces of snap-frozen mouse tissue (~50 mg) or T37i cells ticles was determined. Salsalate tended to increase the (grown in 3.8 cm2 wells) were lysed, protein was isolated, uptake of [3H]TO-derived activity by dorsocervical BAT and Western blots were performed as previously de- (dcBAT), although significance was not reached, probably scribed (11). Primary antibodies and dilutions are listed due to the substantial interindividual variation (Fig. 2G). in Supplementary Table 1. Protein content was corrected Salsalate Reverses HFD-Induced Obesity and for a control mix on each blot and for the housekeeping Improves Glucose and TG Metabolism protein tubulin. We investigated the potential of salsalate to reverse diet- RNA Purification and Quantitative RT-PCR induced obesity in E3L.CETP mice (regression study). In RNA was extracted from snap-frozen mouse tissue (~25 this setting, salsalate lowered body weight (Fig. 3A) mg) or T37i cells (grown in 3.8 cm2 wells) using Tripure mainly due to a reduction in fat mass (Fig. 3B). Further- RNA Isolation reagent (Roche). Total RNA (1–2 mg) more, salsalate reduced fasting glucose (230%, P , 0.01; was reverse transcribed using Moloney Murine Leukemia Fig. 3C) and TG levels (246%, P , 0.05; Fig. 3D). Virus Reverse Transcriptase (Sigma) for quantitative RT- Salsalate increased the uptake of [3H]TO-derived activity PCR (qRT-PCR) to produce cDNA. mRNA expression was predominantly by iBAT (156%, P , 0.01) and dcBAT diabetes.diabetesjournals.org van Dam and Associates 1547

Figure 1—Salsalate prevents obesity and fat mass accumulation in E3L.CETP mice fed an HFD. A–D: The 10-week-old male E3L.CETP mice were fed an HFD without (open circles) or with (closed circles) salsalate for 12 weeks. Body weight (A), fat mass (B), and lean mass (C) were monitored throughout the experiment. D: Food intake was measured daily in the ﬁrst 2 weeks and in the 6th week of the experiment. Values represent the mean 6 SEM (n =9–10). *P < 0.05, **P < 0.01, ***P < 0.001 vs. control.

(102%, P , 0.05), and to some extent by the liver (38%, This was probably achieved via an insulin-independent P , 0.01) (Fig. 3E). These data suggest that salsalate acti- mechanism, as no differences in glucose levels were found vates BAT, and this idea is further supported by the obser- between the treatment groups upon an insulin challenge vation that rectal temperature was raised (0.5°C, P , 0.05; (Supplementary Fig. 2E). Of note, AMPK phosphorylation Fig. 3F) upon salsalate treatment. Moreover, salsalate re- in muscle was enhanced upon salsalate treatment (43%, duced iBAT weight (250%, P , 0.01; Fig. 3G), indicating P , 0.05), as was phosphorylation of its downstream reduced fat accumulation within the tissue. Indeed, lipid target ACC (acetyl-CoA carboxylase) (75%, P , 0.05; Sup- droplet content was lower (Fig. 3H–J), pointing to higher plementary Fig. 2F). intracellular lipolysis (18). In WT mice, salsalate also reduced plasma TG levels (Supplementary Fig. 2G). This was not due to lowered in- Salsalate Prevents Body Weight Gain and Activates testinal TG absorption, since salsalate-treated mice did not BAT Irrespective of the Transgenic Background have significantly lower plasma TG levels in response to an To exclude that the effects of salsalate were specifictothe oral olive oil gavage (Supplementary Fig. 2H), and FFA E3L.CETP transgenic model, male WT mice were fed an content in feces collected in the third week of treatment HFD with or without salsalate for 4 weeks. We confirmed was unaltered (Supplementary Fig. 2I). Neither VLDL-TG that salsalate treatment had similar effects on body weight, (Supplementary Fig. 2J) nor VLDL[35S]apolipoprotein B body composition, and food intake (Supplementary Fig. production (Supplementary Fig. 2K) were affected upon 4 1A–D). The prevention of adiposity despite equal food in- weeks of salsalate treatment, further supporting a role for take suggests that either physical activity or EE is en- BAT in the TG-lowering effect of salsalate. Indeed, compa- hanced. Indeed, salsalate increased EE during the dark rable to the effects seen in E3L.CETP mice, salsalate re- phase (10%, P , 0.05; Supplementary Fig. 1E). This was duced iBAT weight (242%, P , 0.001; Fig. 4A), lowered especially due to a higher glucose oxidation (29%, P , lipid content (229%, P , 0.05; Fig. 4B and C), and tended 0.001; Supplementary Fig. 1F), whereas fat oxidation did to increase the uptake of [3H]TO-derived activity by the not differ (Supplementary Fig. 1G). Accordingly, respira- BAT depots (Supplementary Fig. 3C). tory exchange ratio was higher during the dark phase (Sup- Under thermoneutral conditions, salsalate still pre- plementary Fig. 1H). Physical activity was not affected vented the development of HFD-induced obesity (Supple- (Supplementary Fig. 1I), suggesting that the energetic mentary Fig. 3A and B). However, thermoneutrality ablated loss is mainly explained by increased EE. the salsalate-induced tendency toward an increased TG Comparable to our findings in E3L.CETP mice, salsalate uptake by BAT (Supplementary Fig. 3C) and prevented reduced plasma insulin levels (Supplementary Fig. 2A) and the reduction in plasma TG levels seen at room tempera- improved glucose tolerance (Supplementary Fig. 2B–D). ture (Supplementary Fig. 3D). Thus, noradrenergic input 1548 Salsalate Activates Brown Adipose Tissue Diabetes Volume 64, May 2015

(Fig. 4D). Salsalate tended to increase the PKA-mediated lipolysis-stimulating phosphorylation of hormone-sensitive lipase (HSL) on the Ser563 residue (77%, P = 0.06) but not AMPK-mediated phosphorylation of HSL on the Ser565 residue (19) (Fig. 4D), suggesting increased PKA signaling in BAT. Despite the fact that phosphorylation of ACC, the downstream target of AMPK, was increased, we did not observe signiﬁcant differences in AMPK expression or phosphorylation (Fig. 4D).

Salsalate Reduces WAT Cell Size and Lowers Inflammation in WT Mice Fed an HFD Since salsalate massively reduced fat mass, we assessed the phenotype of WAT in more detail. In line with the decreased total fat mass (Figs. 1B and 3B, and Supple- mentary Fig. 1B), salsalate reduced the weight of the gWAT fat pad (249%, P , 0.001; Fig. 5A). Histological analysis showed that salsalate reduced adipocyte size (244%, P , 0.01; Fig. 5B and C), while total cell number, as assessed by DNA content, did not differ (data not shown). Since salsalate is an anti-inflammatory compound, we investigated the immune cell composition of gWAT by flow cytometry. We did not observe a change in the percentage of CD45+ cells within the SVF, nor in relative monocyte, macrophage, granulocyte, T-cell, and B-cell content (data not shown). However, we found fewer proinflammatory (CD11c+) M1 macrophages and more anti-inflammatory (CD206+) M2 macrophages within the F4/80+ fraction (Fig. 5D and E). Accordingly, salsalate decreased F4/80 and Mcp1 mRNA expression in gWAT (Fig. 5F), confirming fewer proinflammatory macrophages and less recruitment of proinflammatory macrophages (20,21) in the gWAT of salsalate-treated mice. Thus, besides preventing fat accumulation in WAT, salsalate pre- Figure 2—Salsalate prevents HFD-induced deterioration of glucose vented HFD-induced skewing of macrophages toward and TG metabolism in E3L.CETP mice fed an HFD. A–G: The 10- fl week-old male E3L.CETP mice were fed an HFD without (open a proin ammatory phenotype. bars/circles) or with (closed bars/circles) salsalate for 12 weeks. Next, we studied whether the diminished adipocyte Blood samples from mice fasted for 6 h were collected by tail size after salsalate treatment could be the result of vein bleeding at different time points, and plasma glucose (A), in- increased TG lipolysis. Phosphorylation levels of HSL on sulin (B), TC (E), and TG (F) levels were determined. After 4 weeks, mice fasted for 6 h were injected intravenously with glucose, and the Ser563 residue (phosphorylatedbyPKA;Fig.5G)in additional blood samples were taken at 5, 15, 30, 60, 90, and 120 gWAT did not differ, nor did plasma FFA levels (data not min after injection (C and D). A clearance experiment was performed shown). Moreover, in vitro stimulation of 3T3-L1 cells – 3 in which 6 h fasted mice were intravenously injected with [ H]TO- with salsalate did not influence Ser563-HSL phosphory- labeled lipoprotein-like emulsion particles. G: After 15 min, mice were killed, and uptake of [3H]TO-derived activity was determined lation (Supplementary Fig. 4A) and even repressed the in the organs. Values represent means 6 SEM (n =9–10). *P < glycerol concentration in the supernatant (Supplemen- 0.05, **P < 0.01 vs. control. ivGTT, intravenous glucose tolerance tary Fig. 4B). Of the oxidation genes we measured in test; sWAT, subcutaneous WAT; vWAT, visceral WAT. gWAT, only Acc2 was upregulated upon salsalate treatment, suggesting slightly enhanced oxidation in gWAT (Supplementary Fig. 5A). Neither the phosphorylation (i.e., slight cold sensing at room temperature) seems nec- statenortheexpressionofAMPKwasaltered(Fig.5G) essary to bring about salsalate-induced TG uptake by BAT. in gWAT. We also investigated markers of lipogenesis in To further elucidate the mechanisms by which salsalate gWAT, but only found an upregulation of Acc1 expres- activates BAT and regulates intracellular lipolysis, we sion (Supplementary Fig. 5B). Collectively, these data measured mRNA and (phosphorylated) protein levels in indicate that it is unlikely that lipolysis, oxidation, or the BAT of mice housed at 21°C upon treatment. Salsalate lipogenesis in WAT is the primary mechanism responsi- did not affect either Ucp1 mRNA expression (data not ble for the prevention of fat mass accumulation after shown) or uncoupling protein-1 (UCP1) protein content salsalate treatment. diabetes.diabetesjournals.org van Dam and Associates 1549

Figure 3—Salsalate reverses HFD-induced obesity and deterioration of glucose and TG metabolism, and enhances lipoprotein-TG–derived fatty acid uptake by liver and BAT in E3L.CETP mice fed an HFD. A–I: The 10-week-old male E3L.CETP mice were fed an HFD for 12 weeks to render them obese and subsequently were fed an HFD without (open bars/circles) or with (closed bars/circles) salsalate for 4 weeks. Body weight (A) and fat mass (B) were monitored throughout the treatment period. After the treatment period, blood samples from 6 h– fasted mice were collected by tail vein bleeding, and plasma glucose (C) and TG (D) levels were determined. A clearance experiment was performed in which 6 h–fasted mice were intravenously injected with [3H]TO-labeled lipoprotein-like emulsion particles. E: After 15 min, mice were killed, and the uptake of [3H]TO-derived activity was determined in the organs. F: Rectal temperature was determined after 4 weeks of treatment. G: After mice were killed, iBAT was collected and weighed. H: Lipids were extracted from BAT, and TG content was measured. H-E staining of BAT sections was performed, and the relative content of lipid vacuoles in BAT tissue sections was quantiﬁed (I); representative pictures are shown (J). Values represent means 6 SEM (n =9–10). *P < 0.05, **P < 0.01, ***P < 0.001 vs. control. sWAT, subcutaneous WAT; vWAT, visceral WAT.

Salsalate Directly Activates Brown Adipocytes In Vitro this ﬁnding. Besides Ucp1, salsalate upregulated several Since our collective data suggested that the improvement in other genes involved in BAT function (Fig. 6D). TG metabolism by salsalate could be due to activation of The majority, but not all, of orally administered salsalate BAT, we investigated whether salsalate directly activates is metabolized to salicylate in vivo, resulting in plasma brown adipocytes and which intracellular mechanisms are concentrations of 1–3 mmol/L (1). Like salsalate, salicylate involved. Strikingly, the treatment of differentiated T37i (1 mmol/L) enhanced Ucp1 expression in T37i brown adi- adipocytes with salsalate increased uncoupled respiration pocytes (53%, P , 0.01, Supplementary Fig. 6C). Salicylate (Fig. 6A). In line with this, increasing concentrations of (1 and 3 mmol/L) also enhanced the expression of Ppara salsalate resulted in a dose-dependent increase in Ucp1 ex- (up to 304%, P , 0.01) and its target gene (22) Elovl3 (up pression (up to 227%, P , 0.001; Fig. 6B)aswellasglycerol to 224%, P , 0.05; Supplementary Fig. 6C), while the release (up to 92%, P , 0.001; Fig. 6C). Accordingly, higher dosage of salicylate (3 mmol/L) enhanced glycerol salsalate decreased lipid content in the cells, as shown by release (80%, P , 0.01; Supplementary Fig. 6D). Nile red staining (218%, P , 0.05; Supplementary Fig. 6A). Since salsalate is known to directly activate AMPK (1), we Decreased de novo lipogenesis found upon treatment with investigated this in brown adipocytes. However, salsalate did salsalate (Supplementary Fig. 6B) could also contribute to not increase the phosphorylation of AMPK (Fig. 6E)orACC 1550 Salsalate Activates Brown Adipose Tissue Diabetes Volume 64, May 2015

enhanced phosphorylation of the 30-kDa PKA substrate in brown adipocytes after 8 h of stimulation (69%, P , 0.05) and increased phosphorylation of HSL on Ser563, the PKA phosphorylation site of this lipolytic enzyme (108%, P , 0.05; Fig. 6F), while phosphorylation of HSL on Ser565, the AMPK phosphorylation site, was unaffected. To investigate the involvement of the PKA pathway for BAT activation by salsalate, we stimulated brown adipocytes with salsalate in the presence of the PKA inhibitor H89 (25 mmol/L). H89 blunted the salsalate-induced Ser563-HSL phosphorylation (Fig. 6G), upregulation of Ucp1 expression (Fig. 6H), and glycerol release (Fig. 6I), indicating that the PKA pathway is required for brown adipocyte activation by salsalate. These data are consistent with our observation that salsalate only increased TG uptake by BAT in vivo when mice were housed at room temperature (i.e., when adrenergic signaling is present). Since PKA is a downstream target of b-adrenergic signaling, we investigated whether salsalate acts in synergism with b-adrenergic stimulation in T37i brown adipocytes. Although salsalate in combination with the b3-agonist CL316243 enhanced uncoupled respiration compared with CL316243 alone, no synergistic effect was found (Fig. 6J). Furthermore, salsalate and norepinephrine showed an ad- ditive effect on Ucp1 expression (Supplementary Fig. 6E). This suggests that increased b-adrenergic signaling is not required for the effects of salsalate, but since the inhibition of PKA abolishes the effects of salsalate, basal b-adrenergic signaling is required. Taken together, both our in vitro and in vivo data point to an interplay between b-adrenergic and salsalate-induced signaling.

DISCUSSION Previous studies (3,4) in humans and animals have shown that salsalate has beneficial metabolic effects by improving glucose tolerance and lowering plasma TG levels. The mechanisms responsible for these changes, however, remained unclear. In the current study, we show that long- term treatment of E3L.CETP and WT mice with salsalate Figure 4—Salsalate activates BAT in WT mice fed an HFD. The 10- recapitulates these beneficial metabolic effects, and in week-old male C57BL/6J mice were fed an HFD without (open bars) or with (closed bars) salsalate for 4 weeks. A: After mice were killed, addition prevents and reduces body weight gain, mainly iBAT was collected and weighed. H-E staining of BAT sections was by lowering fat mass accumulation. Moreover, we provide performed, and the relative content of lipid vacuoles in BAT tissue evidence from both in vivo and in vitro studies that sections was quantified (B); representative pictures are shown (C). D salsalate activates BAT, an important player in energy : Protein content was determined by Western blot. Values repre- fi sent means 6 SEM (n =6–8). *P < 0.05, ***P < 0.001 vs. control. p, metabolism. To the best of our knowledge, this is the rst phosphorylated. study reporting that activation of BAT may, at least partly, underlie the beneficial metabolic effects of salsalate. The finding that salsalate prevented body weight gain and fat mass accumulation accompanied by a reduced (data not shown) after stimulation for 15 min, 1 h, or 8 h white adipocyte size has been reported before in rats (25). (data not shown). We therefore searched for another route Lower fat accumulation may be the consequence of a lower by which salsalate may activate BAT. food intake or higher EE. Although we noticed that food As both AMPK and PKA regulate lipolysis in BAT by intake was transiently reduced upon salsalate treatment in phosphorylation of HSL (23,24), and our in vivo data in- mice, the preventive effect on fat mass accumulation per- dicated PKA-mediated phosphorylation of HSL on the sisted throughout the studies. We found increased EE in Ser563 residue, we investigated whether salsalate acti- our study, which is in line with studies in humans (2,5) vates the PKA-HSL route in brown adipocytes. Salsalate showing that salsalate treatment increases EE by 18%, diabetes.diabetesjournals.org van Dam and Associates 1551

Figure 5—Salsalate reduces white adipocyte cell size and lowers inflammation in the WAT of WT mice fed an HFD. The 10-week-old male WT mice were fed an HFD without (open bars) or with (closed bars) salsalate for 4 weeks. A: gWAT was collected and weighed. H-E staining of gWAT sections was performed, and the relative cell size of white adipocytes in gWAT tissue sections was quantified (B); representative pictures are shown (C). The percentage of macrophages (%CD11b+Ly6G2F4/80+) of CD45+ cells in SVF of gWAT was determined by flow cytometry. Within the macrophage fraction, proinflammatory M1 macrophages (CD11c+)(D) and anti-inflammatory M2 macrophages (CD206+)(E) were measured. mRNA expression of inflammatory markers (F) in gWAT was determined by qRT-PCR. G: Protein content was determined by Western blot. Values represent means 6 SEM (n =6–8). *P < 0.05, **P < 0.01, ***P < 0.001 vs. control. p, phosphorylated. suggesting that the long-term inhibiting effects of salsalate the possibility that salsalate improves glucose tolerance on body weight gain are caused by increased resting EE. through BAT activation, we did show that the TG-lowering In our study, salsalate improved glucose tolerance via effect of salsalate was a result of the higher uptake of an insulin-independent mechanism, and lowered plasma lipoprotein-TG–derived FA by BAT, as intestinal TG ab- glucose and TG levels, which is in line with findings in sorption and hepatic VLDL production were unaffected in humans (2,3). Recently, it has been shown (26) that sa- our experimental settings. We provide strong evidence licylate stimulates phosphorylation of AMPK in the short that BAT was activated upon salsalate treatment, since term and simultaneously increases glucose transport, in- we found reduced intracellular lipid content and higher dependent of insulin, into muscle in rat skeletal muscles rectal temperature in vivo; higher uncoupled respiration, ex vivo. Correspondingly, in our study, AMPK phosphor- and upregulation of the expression of thermogenic ylation in muscle was enhanced upon salsalate treatment, markers Ucp1 and Elovl3; and increased glycerol release as was phosphorylation of its downstream target ACC. upon salsalate treatment in vitro. Taken together, our data suggest that salsalate improves Regarding the mechanism through which salsalate glucose tolerance by enhancing glucose oxidation and glu- activates BAT, we found slightly increased phosphorylation cose uptake via an insulin-independent pathway. BAT, of ACC, the downstream target of AMPK, in BAT in vivo, a metabolically active tissue that contributes to energy but not of AMPK itself. In T37i brown adipocytes in vitro, metabolism by uncoupling the electron transport chain the phosphorylation statuses of AMPK and ACC were also from ATP synthesis through UCP1, also has a key role unaffected. Although a recent study by Hawley et al. (1) in systemic glucose homeostasis by enhancing glucose showed that salicylate activates AMPK in the liver, muscle, clearance (27,28). Although we do not have evidence for and adipose tissue, they also observed that the potency of 1552 Salsalate Activates Brown Adipose Tissue Diabetes Volume 64, May 2015

Figure 6—Salsalate directly activates T37i brown adipocytes. T37i cells were cultured and differentiated into mature brown adipocytes. Cells were treated with salsalate (300 mmol/L) after the addition of oligomycin and OCR was directly measured with an extracellular ﬂux analyzer. A: Values represent means 6 SEM of 11–13 independent wells. Cells were treated with increasing concentrations of salsalate for 8 h, and Ucp1 mRNA expression (B) and glycerol release (C) were determined. D: Cells were treated with salsalate (300 mmol/L) for 8 h, and mRNA expression was determined. E: Cells were treated with salsalate (300 mmol/L) or AICAR for 15 min, and protein content was determined. F: Cells were treated with salsalate (300 mmol/L) for 8 h, and protein content was determined. Cells were treated with salsalate (300 mmol/L) in the presence of H89 for 8 h; and Ser563-HSL phosphorylation (G), Ucp1 mRNA expression (H), and glycerol release (I) were determined. Cells were treated with salsalate (300 mmol/L) and CL316243 (10 mmol/L) after the addition of oligomycin, and OCR was directly measured with an extracellular ﬂux analyzer. J: Values represent means 6 SEM of 11–13 independent wells. Protein content was determined by Western blot, and mRNA expression was determined by qRT-PCR. Values represent means 6 SD of three to six independent sets of RNA, supernatant, or protein. *P < 0.05, **P < 0.01, ***P < 0.001 vs. control or CL316243; #P < 0.05, ##P < 0.01 vs. salsalate. OM, oligomycin; p, phosphorylated. diabetes.diabetesjournals.org van Dam and Associates 1553

Figure 7—Proposed mechanism by which salsalate activates BAT. Salsalate improves intracellular lipolytic capacity of the brown adipocyte by increasing PKA-mediated HSL phosphorylation, thereby enhancing FA release from TG stored in lipid droplets. This leads to the activation of UCP1, which uncouples ATP synthesis. Moreover, salsalate results in enhanced expression of Ppara, Ppargc1a, and Ucp1, altogether resulting in increased activity and uncoupling of ATP synthesis in BAT. A reduction in intracellular lipid content and upregulation of Cd36 in BAT enhances the uptake of TG-derived FA from the plasma by BAT. This eventually results in the lowering of circulating levels of lipids and a reduction in fat mass. CPT1, carnitine palmitoyl-transferase 1.

salicylate to improve fasting glucose and insulin levels, glu- humans, salsalate lowers the inflammatory state, with cose tolerance, and insulin resistance were retained in mice a 34% reduction in circulating levels of C-reactive protein lacking the b1 subunit of AMPK, demonstrating that other and a decline in nuclear factor-kB activity in WAT (31,32). pathways are more important than the AMPK-mediated This reduction in inflammation might contribute to the im- beneficial metabolic effects of salicylates. proved glucose metabolism upon administration of salsalate. Our data point more profoundly toward activation of Since it is becoming increasingly clear that BAT activa- the PKA-HSL pathway in BAT. We found that salsalate tion and subsequent elevation of EE can lower body fat increased Ser563-HSL phosphorylation in BAT in vivo and mass in human adults (33), drugsthattargetBATareof in brown adipocytes in vitro, which points to an increase of great interest in the combat against obesity. Although some PKA-mediated lipolysis. Indeed, salsalate treatment en- human studies (4,34) suggested that salsalate does not af- hanced glycerol release in vitro and reduced the intracellular fect body weight, the effects on fat mass have not been lipid content in BAT in vivo. The enhanced intracellular reported. As mice have a relatively larger amounts of release of FA results in an increased availability of the BAT compared with humans (35,36), activation of BAT substrate for oxidation that can also induce allosteric by salsalate in humans might translate into improved fat activation of UCP1 (29), both of which result in enhanced distribution, and glucose and lipid metabolism without uncoupling, which we also demonstrated in brown adipo- substantially affecting total body weight. cytes in vitro. Importantly, the fact that PKA inhibition In conclusion, we show that salsalate exerts beneficial blunted the salsalate-induced Ser563-HSL phosphorylation, metabolic effects by directly activating BAT through Ucp1 expression, and glycerol release in vitro supports a ne- modulation of the PKA pathway in BAT (Fig. 7). Further cessity for this pathway in brown adipocyte activation. studies are warranted to investigate whether salsalate acti- Besides improving metabolism, salsalate is an effective vates BAT in humans, thereby preventing obesity and as- anti-inflammatory agent in the clinic. With the current sociated disorders. knowledge of a profound link among obesity, inflammation, and type 2 diabetes, we also investigated the effects Acknowledgments. The authors thank Annika Tanke, Ellemiek de Wit, Isabel of salsalate on the inflammatory state of WAT. In gWAT, Mol, Hetty Sips, Trea Streefland,andChrisvanderBent(allfromLeidenUniversity salsalate lowered Mcp1 expression, an attraction factor of fl Medical Center, Leiden, the Netherlands) for their valuable technical assistance. proin ammatory M1 macrophages, and prevented HFD- Funding. This work was supported by a research grant from the Rembrandt induced skewing of macrophages toward this phenotype. Institute of Cardiovascular Science to M.P.J.d.W., E.L., and P.C.N.R., and by Previous research (30) showed that high concentrations of a personal grant of the Board of Directors of Leiden University Medical Center salicylates inhibit nuclear factor-kB activity in vitro, a key to M.R.B. P.C.N.R. is an Established Investigator of the Dutch Heart Foundation transcription factor that regulates inflammation.Inobese (2009T038). 1554 Salsalate Activates Brown Adipose Tissue Diabetes Volume 64, May 2015

Duality of Interest. No potential conflicts of interest relevant to this article 16. Zennaro MC, Le Menuet D, Viengchareun S, Walker F, Ricquier D, Lombès were reported. M. Hibernoma development in transgenic mice identifies brown adipose tissue as Author Contributions. A.D.v.D. helped to conceptualize the project, a novel target of aldosterone action. J Clin Invest 1998;101:1254–1260 perform the experiments, analyze the data, and draft the manuscript; contributed 17. van den Berghe N, Ouwens DM, Maassen JA, van Mackelenbergh MG, Sips to the discussion; and critically reviewed the manuscript. K.J.N. helped to per- HC, Krans HM. Activation of the Ras/mitogen-activated protein kinase signaling form the experiments, analyze the data, and draft the manuscript. S.K., S.M.v.d.B., pathway alone is not sufficient to induce glucose uptake in 3T3-L1 adipocytes. A.A.K., T.K., M.M.H., and V.v.H. helped to perform the experiments, analyze the Mol Cell Biol 1994;14:2372–2377 data, and/or critically review the manuscript. M.L. provided the T37i cell line and 18. Ortega-Molina A, Efeyan A, Lopez-Guadamillas E, et al. Pten positively helped to critically review the manuscript. M.P.J.d.W. and E.L. contributed to the regulates brown adipose function, energy expenditure, and longevity. Cell Metab discussion and helped to critically review the manuscript. A.M.v.d.H. and B.G. 2012;15:382–394 helped to conceptualize the project, contributed to the discussion, and helped to 19. Watt MJ, Holmes AG, Pinnamaneni SK, et al. Regulation of HSL serine critically review the manuscript. P.C.N.R. and M.R.B. helped to conceptualize the phosphorylation in skeletal muscle and adipose tissue. Am J Physiol Endocrinol project and perform the experiments, supervised the project, contributed to the Metab 2006;290:E500–E508 discussion, and edited the manuscript. A.D.v.D., K.J.N., P.C.N.R., and M.R.B. are 20. Kanda H, Tateya S, Tamori Y, et al. MCP-1 contributes to macrophage the guarantors of this work and, as such, had full access to all the data in the study infiltration into adipose tissue, insulin resistance, and hepatic steatosis in obesity. and take responsibility for the integrity of the data and the accuracy of the data J Clin Invest 2006;116:1494–1505 analysis. 21. Weisberg SP, McCann D, Desai M, Rosenbaum M, Leibel RL, Ferrante AW Jr. Obesity is associated with macrophage accumulation in adipose tissue. J Clin References Invest 2003;112:1796–1808 1. Hawley SA, Fullerton MD, Ross FA, et al. The ancient drug salicylate directly 22. Jakobsson A, Jörgensen JA, Jacobsson A. Differential regulation of fatty – activates AMP-activated protein kinase. Science 2012;336:918 922 acid elongation enzymes in brown adipocytes implies a unique role for Elovl3 fi 2. Gold ne AB, Silver R, Aldhahi W, et al. Use of salsalate to target in- during increased fatty acid oxidation. Am J Physiol Endocrinol Metab 2005;289: fl ammation in the treatment of insulin resistance and type 2 diabetes. Clin Transl E517–E526 – Sci 2008;1:36 43 23. Geerling JJ, Boon MR, van der Zon GC, et al. Metformin lowers plasma 3. Rumore MM, Kim KS. Potential role of salicylates in type 2 diabetes. Ann triglycerides by promoting VLDL-triglyceride clearance by brown adipose tissue in – Pharmacother 2010;44:1207 1221 mice. Diabetes 2014;63:880–891 fi 4. Gold ne AB, Fonseca V, Jablonski KA, Pyle L, Staten MA, Shoelson SE; 24. Cannon B, Nedergaard J. Brown adipose tissue: function and physiological fl TINSAL-T2D (Targeting In ammation Using Salsalate in Type 2 Diabetes) Study significance. Physiol Rev 2004;84:277–359 Team. The effects of salsalate on glycemic control in patients with type 2 di- 25. Cao Y, Dubois DC, Sun H, Almon RR, Jusko WJ. Modeling diabetes disease – abetes: a randomized trial. Ann Intern Med 2010;152:346 357 progression and salsalate intervention in Goto-Kakizaki rats. J Pharmacol Exp 5. Meex RC, Phielix E, Moonen-Kornips E, Schrauwen P, Hesselink MK. Stim- Ther 2011;339:896–904 ulation of human whole-body energy expenditure by salsalate is fueled by higher 26. Serizawa Y, Oshima R, Yoshida M, et al. Salicylate acutely stimulates lipid oxidation under fasting conditions and by higher oxidative glucose disposal 59-AMP-activated protein kinase and insulin-independent glucose transport in rat under insulin-stimulated conditions. J Clin Endocrinol Metab 2011;96:1415–1423 skeletal muscles. Biochem Biophys Res Commun 2014;453:81–85 6. Fu ZQ, Yan S, Saleh A, et al. NPR3 and NPR4 are receptors for the immune 27. Stanford KI, Middelbeek RJ, Townsend KL, et al. Brown adipose tissue signal salicylic acid in plants. Nature 2012;486:228–232 regulates glucose homeostasis and insulin sensitivity. J Clin Invest 2013;123: 7. de Haan W, van der Hoogt CC, Westerterp M, et al. Atorvastatin increases 215–223 HDL cholesterol by reducing CETP expression in cholesterol-fed APOE*3-Leiden. 28. Matsushita M, Yoneshiro T, Aita S, Kameya T, Sugie H, Saito M. Impact of CETP mice. Atherosclerosis 2008;197:57–63 brown adipose tissue on body fatness and glucose metabolism in healthy hu- 8. van der Hoogt CC, de Haan W, Westerterp M, et al. Fenofibrate increases mans. Int J Obes 2013;38:812–817 HDL-cholesterol by reducing cholesteryl ester transfer protein expression. J Lipid 29. Fedorenko A, Lishko PV, Kirichok Y. Mechanism of fatty-acid-dependent Res 2007;48:1763–1771 UCP1 uncoupling in brown fat mitochondria. Cell 2012;151:400–413 9. Westerterp M, van der Hoogt CC, de Haan W, et al. Cholesteryl ester transfer 30. Kopp E, Ghosh S. Inhibition of NF-kappa B by sodium salicylate and aspirin. protein decreases high-density lipoprotein and severely aggravates atheroscle- Science 1994;265:956–959 rosis in APOE*3-Leiden mice. Arterioscler Thromb Vasc Biol 2006;26:2552–2559 31. Goldfine AB, Fonseca V, Jablonski KA, et al.; Targeting Inflammation Using 10. Coomans CP, Geerling JJ, van den Berg SA, et al. The insulin sensitizing Salsalate in Type 2 Diabetes Study Team. Salicylate (salsalate) in patients with effect of topiramate involves KATP channel activation in the central nervous type 2 diabetes: a randomized trial. Ann Intern Med 2013;159:1–12 system. Br J Pharmacol 2013;170:908–918 32. Fleischman A, Shoelson SE, Bernier R, Goldfine AB. Salsalate improves 11. Boon MR, Kooijman S, van Dam AD, et al. Peripheral cannabinoid 1 receptor glycemia and inflammatory parameters in obese young adults. Diabetes Care blockade activates brown adipose tissue and diminishes dyslipidemia and obe- 2008;31:289–294 sity. FASEB J 2014;28:5361–5377 33. Yoneshiro T, Aita S, Matsushita M, et al. Recruited brown adipose tissue as 12. Zambon A, Hashimoto SI, Brunzell JD. Analysis of techniques to obtain plasma an antiobesity agent in humans. J Clin Invest 2013;123:3404–3408 for measurement of levels of free fatty acids. J Lipid Res 1993;34:1021–1028 34. Koska J, Ortega E, Bunt JC, et al. The effect of salsalate on insulin action 13. Rensen PC, van Dijk MC, Havenaar EC, Bijsterbosch MK, Kruijt JK, van and glucose tolerance in obese non-diabetic patients: results of a randomised Berkel TJ. Selective liver targeting of antivirals by recombinant chylomicrons— double-blind placebo-controlled study. Diabetologia 2009;52:385–393 a new therapeutic approach to hepatitis B. Nat Med 1995;1:221–225 35. van Marken Lichtenbelt WD, Schrauwen P. Implications of nonshivering 14. Bligh EG, Dyer WJ. A rapid method of total lipid extraction and purification. thermogenesis for energy balance regulation in humans. Am J Physiol Regul Can J Biochem Physiol 1959;37:911–917 Integr Comp Physiol 2011;301:R285–R296 15. Cariou B, Postic C, Boudou P, et al. Cellular and molecular mechanisms of 36. Vosselman MJ, van Marken Lichtenbelt WD, Schrauwen P. Energy dissi- adipose tissue plasticity in muscle insulin receptor knockout mice. Endocrinology pation in brown adipose tissue: from mice to men. Mol Cell Endocrinol 2013;379: 2004;145:1926–1932 43–50