Diabetes Volume 64, January 2015 117

Min Park,1 Ja-Woon Yi,1 Eun-Mi Kim,1 Il-Joo Yoon,1 Eun-Hee Lee,1 Hwa-Youn Lee,1 Kon-Young Ji,1 Kwang-Ho Lee,1 Ji-Hun Jang,1 Seung-Su Oh,1 Chul-Ho Yun,1 Seung-Hyung Kim,2 Ki-Mo Lee,3 Mun-Gyu Song,4 Dong-Hoon Kim,4 and Hyung-Sik Kang1

Triggering Receptor Expressed on Myeloid Cells 2 (TREM2) Promotes Adipogenesis and Diet-Induced Obesity

Diabetes 2015;64:117–127 | DOI: 10.2337/db13-1869

Triggering receptor expressed on myeloid cells 2 (TREM2) Obesity is one of the major factors contributing to the is known to be involved in the anti-inflammatory re- development of metabolic syndromes, such as insulin sponse and osteoclast development. However, the role resistance, type 2 diabetes (1), and hepatic steatosis (2). of TREM2 in adipogenesis or obesity has not yet been Obesity is characterized by an enlargement of the adipose BST STUDIES OBESITY defined. The effect of TREM2 on adipogenesis and tissue resulting from an increase in the number and/or obesity was investigated in TREM2 transgenic (TG) mice size of the adipocytes, which is accompanied by chronic on a high-fat diet (HFD). To block TREM2 signaling, a inflammation (2,3). These adipocytes secrete chemokines neutralizing fusion specific for TREM2 (TREM2- that recruit macrophages into the adipose tissue, leading Ig) was used. TG mice were much more obese than to insulin resistance and inflammation due to the pro- wild-type mice after feeding with an HFD, independent duction of inflammatory cytokines such as tumor necrosis of the quantity of food intake. These HFD-fed TG mice factor-a (TNF-a), interleukin (IL)-6, and IL-1b (2,3). Ad- manifested adipocyte hypertrophy, glucose and insu- ditionally, the enlarged adipose tissue releases excess free lin resistance, and hepatic steatosis. The expression fatty acids (FFAs), which are taken up by hepatocytes and of adipogenic regulator , such as peroxisome stored as triglycerides, resulting in hepatic steatosis and proliferator–activated receptor g and CCAAT/enhancer- insulin resistance (2,4). binding protein a, was markedly increased in HFD-fed TG Adipogenesis is the process of differentiating mesenchy- mice. Additionally, HFD-fed TG mice exhibited decreased mal stem cells into mature adipocytes (5). CCAAT/enhancer- Wnt10b expression and increased GSK-3b (glycogen binding (C/EBPs) b and d are key transcription synthase kinase-3b)–mediated b-catenin phosphoryla- factors in the early stage of adipogenesis as well as inducers tion. In contrast, the blockade of TREM2 signaling using of the expression of two principal adipogenic transcription TREM2-Ig resulted in the inhibition of adipocyte differen- factors, C/EBPa and peroxisome proliferator–activated tiation in vitro and a reduction in body weight in vivo by receptor g (PPARg)(5).C/EBPa and PPARg act synergisti- downregulating the expression of adipogenic regulators. cally to enhance the expression of adipogenesis-associated Our data demonstrate that TREM2 promotes adipogen- genes and subsequent adipogenic differentiation. The con- esis and diet-induced obesity by upregulating adipo- stitutive activation of Wnt/b-catenin pathway maintains genic regulators in conjunction with inhibiting the the preadipocytes in an undifferentiated state by inhibiting Wnt10b/b-catenin signaling pathway. the expression of C/EBPa and PPARg (6,7). Consequently,

1School of Biological Sciences and Technology, Chonnam National University, Received 11 December 2013 and accepted 22 July 2014. Buk-gu, Gwangju, Republic of Korea This article contains Supplementary Data online at http://diabetes 2 Institute of Traditional Medicine and Bioscience, Daejeon University, Daejeon, .diabetesjournals.org/lookup/suppl/doi:10.2337/db13-1869/-/DC1. Republic of Korea E.-M.K. is currently affiliated with the Department of Internal Medicine, University 3Traditional Korean Medicine Converging Research Division, Korea Institute of of Iowa and Veterans Affairs Medical Center, Iowa City, IA. Oriental Medicine, Daejeon, Republic of Korea 4Department of Pharmacology, Korea University College of Medicine, Seoul, © 2015 by the American Diabetes Association. Readers may use this article as Republic of Korea long as the work is properly cited, the use is educational and not for profit, and the work is not altered. Corresponding author: Hyung-Sik Kang, [email protected]. 118 TREM2 Promotes Adipogenesis Diabetes Volume 64, January 2015 canonical signaling through Wnt inhibits adipogenesis 35 kcal% of energy from carbohydrate) (Research Diets, via the glycogen synthase kinase-3b (GSK3b)/b-catenin– New Brunswick, NJ) for 16 weeks under free-feeding dependent mechanism (6–8). conditions. Mice were individually housed and kept on Triggering receptor expressed on myeloid cells 2 a 12 h light/dark cycle at 22 6 1°C. All of the mice experi- (TREM2) is expressed on the cell membrane of dendritic ments were conducted in accordance with the guidelines cells (9), macrophages (10), microglia (11,12), and osteo- of the Institutional Animal Care Committee of Chonnam clasts (13) as an orphan cell surface receptor of the Ig National University. The TREM2-Ig or control human Ig superfamily. TREM2 associates with the adapter protein (hIg) was administered intraperitoneally twice per week DAP12 containing an immunoreceptor tyrosine–based ac- for 16 weeks at 700 mg/mouse. tivation motif, which provides a docking site for the Syk (14). Cell Culture and Isolation of Primary Mouse Embryonic The recruitment and phosphorylation of Syk activate down- Fibroblasts stream intracellular signaling through phosphatidylinositol 3T3-L1 mouse preadipocytes were obtained from the – 3-kinase and extracellular signal related kinase pathways American Type Culture Collection (Manassas, VA). The fl (15). TREM2 signaling negatively regulates the in am- primary mouse embryonic fibroblasts (MEFs) were iso- matory response via Toll-like receptor signaling in macro- lated from embryos of WT and TG mice at 13.5 days phages (14). In microglia, TREM2 signaling also inhibits the postcoitus, as described previously (19). inflammatory response by suppressing TNF-a and nitric Induction of Adipocyte Differentiation oxide synthase-2 genes and the phagocytic clearance of ap- MEFs and 3T3-L1 cells were cultured in DMEM supple- optotic neurons (12). mented with 10% bovine calf serum. Two days after reaching Recently, two independent groups have reported that confluence (day 0), the medium was replaced with DMEM TREM2 expression is gradually upregulated in the supplemented with 10% FBS containing DMI (0.25 ng adipose tissue of dogs on a high-fat diet (HFD) (16) and in dexamethasone, 250 mmol/L methylisobutylxanthine, 5 the mesenteric adipose tissue of insulin-resistant/diabetic mg/mL insulin) (Sigma, St. Louis, MO). After 3 days, the db/db mice (17). However, the role of TREM2 in adipo- medium was replaced with DMEM containing 10% FBS, genesis and obesity has been poorly defined. Moreover, the involvement of TREM2 in the Wnt signaling pathway dexamethasone, and insulin; and was replenished every other day for 10 days in 3T3-L1 cells or for 15 days in has not yet been elucidated. MEFs. Here, we demonstrate for the first time that TREM2 induces obesity by promoting adipogenesis and upregulat- Indirect Calorimetry ing the expression of adipogenic regulators. The current Energy expenditure was measured by indirect calorimetry study supports the notion that the Wnt10b/b-catenin sig- (Oxylet; Panlab, Cornellà, Spain) for 24 h after 16 weeks naling pathway may play a crucial role in TREM2-induced of SD or HFD, as described previously (20). Briefly, mice adipogenesis and obesity. were acclimated individually in calorimetry chambers for 1 week prior to indirect calorimetry measurement. The RESEARCH DESIGN AND METHODS values of VO2 and VCO2 were continuously monitored Generation of TREM2 Transgenic Mice for 3 min at 30-min intervals. Food and water were freely Transgenic (TG) mice overexpressing TREM2 were available. Energy expenditure was expressed as VO2 (in generated by pronuclear microinjection, as described milliliters per minute) per mouse with no normalization previously (18). The forward primer (59-GAATTCGC to body weight, and respiratory quotient (RQ) was calcu- 9 9 CCTTGGCTGGCTGCTGGCA-3 ) and reverse primer (5 - lated by dividing VCO2 by VO2. 9 GTACGTGAGAGAATTC-3 ) for mouse TREM2 were used Semiquantitative RT-PCR and quantitative real-time to amplify TREM2 cDNA by RT-PCR. The PCR product PCR was cloned into pGEM-T Vector (Promega, Madison, WI) RT-PCR and quantitative real-time PCR were performed Eco and digested using RI restriction enzyme, which was as described (4). The primers used for RT-PCR and quan- then cloned into pcDNA3.1 expression vector (Invitrogen, titative real-time PCR are listed in Supplementary Table 1. Carlsbad, CA). The pure isolated recombinant TREM2 gene was microinjected into a fertilized egg of a C57BL/6 mouse Western Blotting and Flow Cytometry Analysis according to a standard protocol (Macrogen, Seoul, Repub- Western blotting was performed on cell lysates, as lic of Korea). The integration of the transgene was con- described previously (6). Cells were stained with the in- fl firmed by Southern blotting. Wild-type (WT) littermates dividual antibodies and analyzed by ow cytometry (BD and TG mice were maintained in a C57BL/6 genetic back- Biosciences, San Diego, CA) (21). For intracellular stain- fi ground, and any offspring was genotyped by RT-PCR prior ing, cells were xed with 1.5% paraformaldehyde and to use in the experiments. stained with the individual antibodies after permeabiliza- Mice, HFD, and Administration of TREM2-Ig tion with cold methanol. Eight-week-old male WT littermates and TG mice were fed Glucose and Insulin Tolerance Tests with either standard diet (SD) or HFD D12451 (45 kcal% After 16 weeks of feeding, mice fasted overnight were of energy from fat, 20 kcal% of energy from protein, and intraperitoneally injected with glucose (1 g/kg) or insulin diabetes.diabetesjournals.org Park and Associates 119

Table 1—TREM2 regulates plasma metabolic parameters and liver enzymes SD HFD WT mice TG mice WT mice TG mice CHOL (mg/L) 73 6 6.4 79 6 7.8 125 6 11.3 243 6 7.2* Triglycerides (mg/dL) 81 6 3.0 88 6 4.8 105 6 2.0 119 6 2.1† LDL (mg/dL) 8 6 1.4 10 6 2.1 20 6 5.7 67 6 2.8* FFA (mEq/L) 1,154 6 72.1 1,202 6 32.5 1,342 6 24.0 1,773 6 46.7* GOT (IU/L) 95 6 4.1 100 6 3.4 128 6 10.9 215 6 8.8† GPT (IU/L) 42 6 2.1 50 6 3.0 68 6 4.0 99 6 6.1† All values correspond to the mean 6 SEM of five experiments in each group (n = 5/group). GOT, glutamic oxaloacetic transaminase; GPT, glutamate-pyruvate transaminase. *P , 0.01; †P , 0.05.

(0.5 units/kg) for the glucose tolerance tests and insulin MEFs, cells were fixed in 10% formalin, washed twice with tolerance tests. Blood was collected from the tail vein at 0, 60% isopropanol, and stained with Oil Red O. 15, 30, 45, 60, and 75 min after injection. The blood Statistical Analysis glucose level was measured using a SureStep Plus The significance was calculated using two-way ANOVA Glucometer (LifeScan, Milpitas, CA). and Student t test. The results were considered statisti- Measurement of Plasma Metabolic Parameters cally significant when the P values were ,0.05. The metabolic parameters were determined using an RESULTS Auto Chemistry Analyzer (Chiron, Emeryville, CA). The plasma levels of adiponectin, insulin, glucose, and leptin Generation of TREM2 TG Mice – were determined using ELISA kits according to the A 732 fragment of mouse TREM2 cDNA was Eco Eco manufacturer’s instructions (R&D Systems, Minneapo- inserted into RI/ RI site of pcDNA3.1(+) expression A lis, MN). vector containing pCMV promoter (Fig. 1 ). To verify the expression of TREM2 in TG mice, TREM2 gene expression Histological Analysis and Oil Red O Staining was analyzed by RT-PCR in various tissues. TG mice The liver and adipose tissues were embedded in frozen OCT exhibited an overexpression of the TREM2 gene in all compound and paraffin, respectively. The tissue sections tissues examined in comparison with WT mice (Fig. 1B). were stained with hematoxylin-eosin (H&E) or Oil Red TG mice also overexpressed TREM2 protein in epididymal O (Sigma). For the Oil Red O staining of 3T3-L1 cells and white adipose tissue (EWAT), retroperitoneal white

Figure 1—Generation and characterization of TREM2 TG mice. A: Schematic of the expression construct used to generate TREM2 TG mice. B: RT-PCR analysis of TREM2 gene expression in various tissues isolated from WT and TREM2 TG mice. C: Western blot analysis of TREM2 protein expression in the EWAT, RWAT, and liver tissue. D: Flow cytometry analysis of TREM2 expression on the liver cell surface. The data are representative of five independent experiments. bp, base pair. 120 TREM2 Promotes Adipogenesis Diabetes Volume 64, January 2015 adipose tissue (RWAT), and liver tissue (Fig. 1C). The cell The weights of the EWAT, RWAT, and liver tissue were surface expression levels of TREM2 in the liver tissue of comparable between WT and TG mice being fed an SD but TG mice were also much higher than those in WT mice were greatly increased in HFD-fed TG mice (Fig. 2E). One (Fig. 1D). of the classic rodent models of obesity is leptin-deficient ob/ob TREM2 TG Mice Are Much More Obese Than WT Mice mice (22). Thus, to determine whether TREM2 was After HFD Feeding related to obesity, TREM2 gene expression was measured ob/ob To elucidate the role of TREM2 in obesity, both WT and in mice. Interestingly, TREM2 was markedly in- ob/ob TG mice were fed with HFD. Whereas we observed no creased in the EWAT, RWAT, and liver tissue of significant difference in body size between WT and TG mice (Supplementary Fig. 1), implying that TREM2 may mice, the bodies of TG mice were much fatter than those be involved in the development of obesity. of WT mice on an HFD (Fig. 2A). The body weight of TG Histologically, although the adipocyte size in the EWAT mice was considerably increased from 4 to 16 weeks after was not different between WT and TG mice on an SD (Fig. F a b beginning the HFD feeding compared with WT controls 2 , and ), the adipocytes from HFD-fed TG mice were F c d (57.87 6 1.20 g in TG vs. 46.48 6 1.27 g in WT at 16 considerably enlarged (Fig. 2 , and ). The liver tissue size weeks) (Fig. 2B). of HFD-fed TG mice paralleled that of the EWAT of the F e–h Next, we determined whether this difference in body same mice (Fig. 2 , ). Additionally, a histochemical anal- weight was due to a disparity in food intake. No ysis using Oil Red O staining clearly indicated that the lipid considerable difference was observed in food intake droplets in the liver tissue of HFD-fed TG mice were larger F i–l between WT and TG mice on an SD or an HFD, suggesting than in those of WT mice (Fig. 2 , ). that the increased body weight of HFD-fed TG mice was TREM2 TG Mice Show Lower Energy Expenditure Than not due to a difference in food intake (Fig. 2C). Consis- WT Mice on an HFD tently, the quantities of the EWAT and RWAT were mark- Obesity is caused by excessive energy intake relative to edly increased in HFD-fed TG mice, with comparable expenditure (23). To determine whether changes in en- quantities in WT and TG mice on an SD (Fig. 2D). The ergy expenditure contribute to body weight gain in TG liver tissue from TG mice was pale and enlarged compared mice, energy expenditure was measured by indirect calo- with that in WT mice being fed an HFD (Fig. 2D), suggest- rimetry. When energy expenditure was calculated without ing that hepatic steatosis may have occurred in TG mice. adjustment for body weight (24,25), nocturnal and

Figure 2—Increased body weight and adipocyte hypertrophy of HFD-fed TREM2 TG mice. Changes in the body size (A), body weight (B), and food intake (C) of WT and TG mice after 16 weeks of SD or HFD feeding (total n = 25/group). Representative images (D) and weights (E) of the EWAT, RWAT, and liver tissue isolated from WT and TG mice after 16 weeks of SD or HFD feeding (total n = 25/group). F: H&E staining of EWAT (a–d) and liver tissue (e–h) of WT and TG mice after 16 weeks of SD or HFD feeding and Oil Red O staining of the liver tissue (i–l). These images were captured under a light microscope at 3100 or 3200 magnification. The results are shown as the mean 6 SEM and are representative of five independent experiments. *P < 0.05; **P < 0.01; ***P < 0.001. diabetes.diabetesjournals.org Park and Associates 121 diurnal energy expenditure was not significantly different insulin tolerance. We then performed glucose and insulin between WT and TG mice on an HFD (Fig. 3A). Whereas tolerance tests in mice on an SD or HFD. The plasma nocturnal energy expenditure was significantly lower in glucose levels did not significantly differ between WT SD-fed TG mice than in WT controls, diurnal energy ex- and TG mice on an SD, but increased in TG mice on an penditure was slightly lower. Nocturnal/diurnal RQ ratios HFD, indicating that HFD-fed TG mice were more glucose were not significantly different between WT and TG mice and insulin resistant than WT controls (Fig. 4E and F). on SDs and HFDs (Fig. 3B). Brown adipose tissue (BAT) The levels of insulin resistance–related genes, such as the has been known to play a crucial role in energy expendi- insulin receptor, insulin receptor substrate 1/2, and glu- ture by promoting the expression of thermogenic genes, cose transporter 4, were reduced in the liver tissue and such as uncoupling protein 1 and b-3-adrenergic receptor EWAT of HFD-fed TG mice (Supplementary Fig. 2A). Glu- (b-3AR) (26,27). While the expression of uncoupling pro- cokinase (GK) gene expression was reduced in the liver tein 1 and b-3AR genes was decreased (Fig. 3C), the size tissue of HFD-fed WT and TG mice compared with those and weight of BAT were remarkably increased in HFD-fed of SD-fed controls, which were more prominent in HFD- TG mice compared with WT controls (Fig. 3D and E). fed TG mice than in HFD-fed WT controls (Supplemen- tary Fig. 2B). The phosphorylation levels of IRS1 at TREM2 TG Mice Exhibit Severe Systemic Glucose and Ser302, AKT at Ser473, and Gsk3b at Ser9 were much Insulin Resistance on HFD lower in the EWAT, RWAT, and liver tissue of HFD-fed The plasma adiponectin levels were decreased in HFD-fed TG mice than in SD-fed TG mice (Supplementary Fig. 2C). TG mice compared with WT controls (Fig. 4A), but the Adipocyte-derived inflammatory chemokines, such as leptin levels were elevated (Fig. 4B). However, food intake MCP-1 and chemokine ligand 5, recruit macrophages was not significantly different between the two groups of and T cells into the WAT, which leads to increased levels mice (Fig. 2C). These results are consistent with the pre- of inflammatory cytokines and thereby contributes to the vious findings that leptin is a critical regulator of obesity induction of insulin resistance (29). The expression of and appetite, but that plasma levels of leptin are not IL-6, TNF-a, MCP-1, and chemokine ligand 5 was higher always correlated with food intake (28). in the stromal vascular fraction (SVF) isolated from the To determine whether the difference in adipose tissue EWAT of HFD-fed TG mice than in that of WT controls mass leads to alteration of glucose metabolism mediated (Supplementary Fig. 3A). The percentages and absolute by TREM2 in mice on an HFD, we measured the plasma numbers of CD4+ and CD8+ TcellsandF4/80+CD11c+ fasting glucose and insulin levels. In HFD-fed TG mice, M1 macrophages were higher in the SVFs of HFD-fed TG the plasma levels of glucose and insulin were significantly mice than in those of WT controls (Supplementary Fig. 3B higher than those in WT controls (Fig. 4C and D), suggest- and C). However, the numbers of B220+ B cells and ing that HFD-fed TG mice exhibited impaired glucose and F4/80+CD206+ M2 macrophages were comparable between

Figure 3—Energy expenditure and thermogenesis-related gene expression in HFD-fed WT and TG mice. Energy expenditure (VO2; A)andRQ ratios (B) were measured by indirect calorimetry in WT and TG mice after 16 weeks of SD or HFD feeding. RQ was calculated by dividing VCO2 by VO2. Quantitative real-time PCR analysis for thermogenesis-related gene expression (C), representative images (D, a–d)andH&Estaining (D, e–h), and weight (E) of BAT after 16 weeks of SD or HFD feeding. The data are represented as the mean 6 SEM of three independent experiments. *P < 0.05; **P < 0.01. 122 TREM2 Promotes Adipogenesis Diabetes Volume 64, January 2015

Figure 4—Systemic glucose tolerance and insulin resistance in HFD-fed TREM2 TG mice. The plasma levels of adiponectin (A), leptin (B), glucose (C), and insulin (D) were quantified by ELISA. Glucose (E) and insulin (F) tolerance tests were performed on the WT and TG mice after 16 weeks of SD or HFD feeding (n = 10). *P < 0.05; **P < 0.01.

WT and TG mice on SDs and HFDs. These data indicate levels increase during adipogenic differentiation; in con- that TREM2 overexpression in mice on an HFD may play trast, PPARg1 is constitutively expressed (34). In agree- an important role in the development of insulin resistance. ment with these findings, we found that the expression of g TREM2 TG Mice Have Elevated Serum Metabolic the PPAR 1 gene did not vary among the groups of mice Parameters Than WT Mice on an HFD studied here. Elevated serum levels of cholesterol (CHOL), triglycerides, The activation of Wnt/b-catenin signaling pathway is – FFAs, and LDL are commonly observed in obesity (30,31). required to inhibit adipogenesis (6 8). Wnt10b gene TG mice fed with an HFD manifested elevated serum levels expression was dramatically reduced in the EWAT of ofCHOL,triglycerides,LDL,andFFAs,whereasnosignif- TG mice on an SD and an HFD (Fig. 5), suggesting icant differences between WT and TG mice fed with an SD that TREM2 may regulate Wnt10b signaling. To test were observed (Table 1). There were no significant changes this hypothesis, we assessed the activation state of the in the plasma levels of glutamic oxaloacetic transaminase TREM2 downstream signaling pathway. The phosphor- and glutamate-pyruvate transaminase in SD-fed WT and TG ylation of GSK3b at Ser9 was decreased in HFD-fed TG E mice, but levels increased considerably in HFD-fed TG mice. mice compared with WT mice (Fig. 5 ), indicating that GSK3b wasactivated.Ingoodaccordance,GSK3b– Upregulation of Adipogenic Regulators and Inhibition mediated b-catenin phosphorylation at Ser33/37/Thr41 of Wnt10b/b-Catenin Signaling in HFD-Fed TREM2 TG was increased in HFD-fed TG mice. These results imply Mice that TREM2 on an HFD upregulated the expression of PPARg and C/EBPa are regulators involved in promoting adipogenic regulators and inhibited Wnt10b/b-catenin adipogenesis (5). Elevated levels of the lipogenic enzyme signaling. fatty acid synthase (FAS) are associated with obesity (32), and CD36 plays a role in the transport of fatty acids in TREM2 Promotes Adipocyte Differentiation by adipocytes (33). We thus evaluated the expression of Regulating Adipogenic Regulators these adipogenic regulators in the EWAT of WT and TG Differentiation of 3T3-L1 preadipocytes and MEFs into mice on an SD or an HFD. Gene expression levels of adipocytes was induced by DMI treatment. The degree of PPARg2 and C/EBPa were overexpressed in SD-fed TG cell differentiation was determined based on the in- mice compared with WT controls (Fig. 5A and B). Whereas tracellular accumulation of lipid droplets, indicated by the expression of PPARg2, C/EBPa, FAS, and CD36 genes Oil Red O staining. The accumulation of cytoplasmic lipid was markedly increased in HFD-fed TG mice compared droplets in 3T3-L1 cells was gradually enhanced during with WT controls, the adiponectin levels were reduced, adipocyte differentiation (Fig. 6A). as determined by RT-PCR (A) and quantitative real-time Next, to investigate whether TREM2 is involved in PCR (B) (Fig. 5). Regarding PPARg1 gene expression, pre- adipocyte differentiation, we examined TREM2 gene ex- vious reports have demonstrated that PPARg2 transcript pression during the differentiation process. TREM2 gene diabetes.diabetesjournals.org Park and Associates 123

Figure 5—Elevated expression of adipogenic regulator genes and inhibition of Wnt10b/b-catenin signaling pathway in HFD-fed TREM2 TG mice. RT-PCR (A and C) and quantitative real-time PCR (B and D) analysis of the EWAT of WT and TG mice after 16 weeks of SD or HFD feeding. E: Western blot analysis of total and phosphorylated (p) GSK3b and b-catenin. The data in B and D are presented as relative fold changes, and the error bars represent the mean 6 SEM of five independent experiments. *P < 0.05; **P < 0.01; ***P < 0.001.

expression was increased in a differentiation stage– (TREM2-Ig). The cytoplasmic lipid accumulation and dependent manner (Fig. 6B). Similarly, the expression droplet formation were substantially lower after treat- levels of adipogenic regulator genes, such as those encod- ment of 3T3-L1 cells with TREM2-Ig than with an hIg ing PPARg1 and 2, C/EBPa, FAS, and CD36, also gradu- control (Fig. 6F). Additionally, the expression of the ally elevated over time during differentiation (Fig. 6C). PPARg2, C/EBPa, CD36, and FAS genes was suppressed These data suggest that TREM2 may promote the differ- in 3T3-L1 cells treated with TREM2-Ig (Fig. 6G). In con- entiation of adipocytes through the regulation of adipo- trast, Wnt10b gene expression was increased in TREM2- genic regulator expression. To evaluate this possibility, Ig–treated 3T3-L1 cells, suggesting that signaling via the degree of adipocyte differentiation and the expression TREM2 may promote adipogenesis by downregulating of adipogenic regulator genes were evaluated in primary Wnt10b expression. Consistent with the data in 3T3-L1 MEFs isolated from WT and TG mice. On days 10–15 of cells, intracellular lipid accumulation in primary MEFs was the differentiation period, the degree of intracellular lipid greatly decreased following TREM2-Ig treatment compared droplet accumulation in TG MEFs was much higher than with hIg-treated controls (Fig. 6H). Furthermore, the treat- in WT MEFs (Fig. 6D). Consistent with our data on the ment of primary MEFs with TREM2-Ig resulted in the accumulation of lipid droplets, the gene expression levels suppression of PPARg,C/EBPa, and FAS gene expression of adipogenic regulators were markedly increased in TG but elevated the expression of Wnt10b (Fig. 6I). MEFs on day 5 after adipogenic induction (Fig. 6E). These results indicate that TREM2 expression is increased dur- Blockade of TREM2 Signaling Inhibits HFD-Induced ing adipogenesis, which may accelerate the accumulation Obesity of lipid droplets by upregulating adipogenic regulators. TREM2-Ig decreased the magnitude of HFD-induced To confirm the involvement of TREM2 in adipocyte increase in the (final) body weight of both WT mice and differentiation, the accumulation of lipid droplets and the TG mice, although the same phenomenon was not expression of adipogenic regulator genes were evaluated observed in hIg-administered mice (Fig. 7A–C). The following the blockade of TREM2 signaling by treatment TREM2-Ig–administered WT and TG mice also manifested with a neutralizing fusion protein specific for TREM2 a reduction in the sizes and weights of their EWAT and 124 TREM2 Promotes Adipogenesis Diabetes Volume 64, January 2015

Figure 6—Promotion of adipocyte differentiation through upregulation of adipogenic regulator gene expression by TREM2. Oil Red O staining (A), quantitative real-time PCR analysis (B), and RT-PCR analysis (C) of DMI-induced differentiation of 3T3-L1 preadipocytes into adipocytes. Oil Red O staining (D) and RT-PCR analysis (E) of DMI-induced differentiation of MEFs isolated from WT and TREM2 TG mice into adipocytes. The data in B are presented as relative fold changes, and the error bars represent the mean 6 SEM of five independent experiments. F and H: Oil Red O staining of 3T3-L1 preadipocytes or MEFs pretreated with hIg or TREM2-Ig and cultured in the presence of DMI for 10 or 15 days, respectively. The hIg and TREM2-Ig were replenished every other day. G and I: RT-PCR analysis (left), and the relative expression ratio compared with hIg (right). The error bars represent the mean 6 SEM of five independent experiments. *P < 0.05; **P < 0.01; ***P < 0.001. diabetes.diabetesjournals.org Park and Associates 125

Figure 7—Inhibition of HFD-induced obesity through the downregulation of adipogenic regulators by blockade of TREM2 signaling. Changes in the body size (A), body weight (B), and final body weight (C) of the hIg- or TREM2-Ig–administered WT and TG mice after 16 weeks of HFD feeding (total n = 25/group). The weights (D) and representative images (E) of the EWAT, RWAT, and liver tissue of these mice. RT-PCR (F), quantitative real-time PCR (G), and Western blot (H) analyses of the EWAT. The error bars represent the mean 6 SEM of five independent experiments. *P < 0.05; **P < 0.01; ***P < 0.001.

RWAT (Fig. 7D and E). However, the weight and size of DISCUSSION – the liver tissue was only reduced in TREM2-Ig administered Here, we provide the first demonstration that TREM2 TG mice and not in WT controls. We also found that the promoted adipocyte differentiation through the regula- HFD-induced increase in the expression of the PPARg2, tion of adipogenic regulators in vitro. Furthermore, C/EBPa, FAS and CD36 genes was markedly inhibited in TREM2 TG mice showed the increased body weight, the TREM2-Ig–administered WT and TG mice compared adipocyte hypertrophy, insulin resistance, and hepatic with the hIg-administered control mice (Fig. 7F and G and steatosis while being fed an HFD. In addition, HFD-fed Supplementary Fig. 4A). In contrast, adiponectin and TREM2 TG mice manifested reduced Wnt10b expression, Wnt10b gene expression was increased in both groups of suggesting that TREM2 may inhibit Wnt/b-catenin sig- TREM2-Ig–administered mice. In addition, the expres- naling pathway and promote diet-induced obesity. There- sion levels of insulin resistance–related genes were fore, TREM2 may act as a previously unrecognized inducer increased in the EWAT and liver tissue of TREM2-Ig– of obesity as well as a novel proadipogenesis factor. administered HFD-fed WT mice, suggesting that The onset of obesity is strongly associated with age TREM2-Ig administration may improve adipose and he- (35). The initial body weight of TG mice was lower than patic insulin resistance (Supplementary Fig. 4B). Control that of WT mice for up to 16 weeks after birth while being hIg-administered HFD-fed TG mice showed a decreased fed an SD (Fig. 2B). Despite the fact that TG mice weighed phosphorylation of GSK3b compared with WT controls, less at the onset of the experiment, after 12 weeks WT but TREM2-Ig administration caused an increase in its and TG mice weighted approximately the same. Thereaf- phosphorylation (Fig. 7H). In contrast, the administration ter, the body weight of TG mice gradually increased with of HFD-fed TG mice with TREM2-Ig inhibited the in- age to a greater extent than that of WT mice. Although creased phosphorylation of b-catenin in hIg-administered the increase in the body weight of TG mice appeared control mice. minimal, there was, in fact, a very large difference 126 TREM2 Promotes Adipogenesis Diabetes Volume 64, January 2015 between WT and TG mice when the body weight changes regulators in conjunction with inhibiting the Wnt10b/ were expressed as a percentage of the initial body weight b-catenin signaling pathway. Consequently, TREM2 is re- (30% increase in WT mice and 61% increase in TG mice quired for adipocyte differentiation, although the over- after 16 weeks of the experiment) (Supplementary Fig. 5). expression of TREM2 causes the development of obesity As shown in Fig. 2, no significant differences were found coupled with insulin resistance and hepatic steatosis. Our in the degree of obesity between young (up to 12 weeks findings suggest that TREM2 acts as a novel marker and old) WT and TG mice on an SD. However, aged regulator of adipogenesis and that the blockade of TREM2 (15 months old) SD-fed TG mice showed an increased signaling is a potential therapeutic target in obesity and body size and weight compared with WT controls (Sup- insulin resistance. plementary Fig. 6A and B). A pronounced increase in the size and weight of the EWAT and liver tissue was also observed in SD-fed TG mice (Supplementary Fig. 6C and Funding. This work was supported by the Basic Science Research Program D). Histological analysis revealed that the adipocyte size of through the National Research Foundation of Korea (NRF), which was funded by the EWAT from SD-fed TG mice was much larger than that the Ministry of Education, Science and Technology (grants NRF-2009-0077425 from WT controls (Supplementary Fig. 6E). Gene expres- and NRF-2012R1A1A2004948). fl sion of adipogenic regulators such as PPARg2, C/EBPa, Duality of Interest. No potential con icts of interest relevant to this article were reported. FAS, and CD36 was elevated, but the levels of adiponectin Author Contributions. M.P. designed the experiments, researched the and Wnt10b were reduced in the EWAT of SD-fed TG mice data, contributed to the discussion, and wrote the manuscript. J.-W.Y. designed F (Supplementary Fig. 6 ). Therefore, obesity also could be the experiments and researched the data. E.-M.K. contributed to the generation induced in TG mice by an HFD as well as by an SD, imply- of the transgenic mouse model. I.-J.Y., S.-H.K., M.-G.S., and D.-H.K. researched ing that TREM2 overexpression likely accelerates the de- the data. E.-H.L., H.-Y.L., K.-Y.J., K.-H.L., J.-H.J., and S.-S.O. contributed to the velopment of age-related obesity, regardless of diet. assist techniques and reagents. C.-H.Y. contributed to critical reading of the The expression of adipogenic regulators, such as manuscript and useful discussions. K.-M.L. edited the manuscript. H.-S.K. PPARg, C/EBPa, FAS, and CD36 was increased in HFD- designed the experiments; analyzed the data; and cowrote, reviewed, and edited fed TG mice, but obesity-induced insulin resistance was the manuscript. H.-S.K. is the guarantor of this work and, as such, had full not improved. This may be partly due to impaired glucose access to all the data in the study and takes responsibility for the integrity of and insulin metabolism by inhibiting GK and GLUT4 ex- the data and the accuracy of the data analysis. fi g a pression, insuf cient binding sites for PPAR and C/EBP References transcription factors by downregulating their target genes 1. Kopelman PG. Obesity as a medical problem. Nature 2000;404:635–643 (GK, GLUT4), and FAS- or CD36-mediated inhibition of 2. Festi D, Colecchia A, Sacco T, Bondi M, Roda E, Marchesini G. Hepatic insulin signaling. Further investigations of the roles of steatosis in obese patients: clinical aspects and prognostic significance. Obes these genes in obesity-induced insulin resistance from Rev 2004;5:27–42 HFD-fed TG mice are required. 3. Kahn BB, Flier JS. Obesity and insulin resistance. J Clin Invest 2000;106: While TREM2 has been known primarily as a negative 473–481 regulator of inflammatory response in macrophages, 4. Boden G, She P, Mozzoli M, et al. Free fatty acids produce insulin resistance microglial cells, and dendritic cells (11–14), a recent study and activate the proinflammatory nuclear factor-kappaB pathway in rat liver. (36) has reported that TREM2 amplifies the inflammatory Diabetes 2005;54:3458–3465 response induced by Toll-like receptor in inflammatory 5. Farmer SR. Transcriptional control of adipocyte formation. Cell Metab 2006; – bowel disease. However, whether TREM2 is involved in 4:263 273 6. Cawthorn WP, Bree AJ, Yao Y, et al. Wnt6, Wnt10a and Wnt10b inhibit the inflammatory or anti-inflammatory response in adi- adipogenesis and stimulate osteoblastogenesis through a b-catenin-dependent pose tissues has not yet been investigated. Here, HFD-fed mechanism. Bone 2012;50:477–489 TG mice revealed overexpression of chemokines, leading 7. Ross SE, Hemati N, Longo KA, et al. Inhibition of adipogenesis by Wnt to recruitment of macrophages and T cells into adipose signaling. Science 2000;289:950–953 tissue. The elevated level of proinflammatory cytokines 8. Prestwich TC, Macdougald OA. Wnt/beta-catenin signaling in adipogenesis was also found in the SVFs of HFD-fed TG mice, which and metabolism. Curr Opin Cell Biol 2007;19:612–617 may be expressed by the infiltrated macrophages and 9. Bouchon A, Hernández-Munain C, Cella M, Colonna M. A DAP12-mediated T cells. The elevation of proinflammatory cytokines also pathway regulates expression of CC chemokine receptor 7 and maturation of contributes to the induction of insulin resistance (29). human dendritic cells. J Exp Med 2001;194:1111–1122 Our finding that insulin resistance was increased in 10. Chung DH, Seaman WE, Daws MRT. Characterization of TREM-3, an fi HFD-fed TG mice is consistent with the finding that the activating receptor on mouse macrophages: de nition of a family of single Ig domain receptors on mouse 17. Eur J Immunol 2002;32: infiltration of macrophages and T cells in adipose tissue 59–66 plays a crucial role in the induction of insulin resistance fl 11. Schmid CD, Sautkulis LN, Danielson PE, et al. Heterogeneous expression of by upregulating proin ammatory cytokines (29). Thus, the triggering receptor expressed on myeloid cells-2 on adult murine microglia. our study elucidated the idea that TREM2 may be in- J Neurochem 2002;83:1309–1320 volved in the inflammatory response in adipose tissues. 12. Takahashi K, Rochford CDP, Neumann H. Clearance of apoptotic neurons In summary, our study is the first to demonstrate that without inflammation by microglial triggering receptor expressed on myeloid TREM2 promotes adipogenesis by upregulating adipogenic cells-2. J Exp Med 2005;201:647–657 diabetes.diabetesjournals.org Park and Associates 127

13. Paloneva J, Mandelin J, Kiialainen A, et al. DAP12/TREM2 deficiency results 27. Sell H, Berger JP, Samson P, et al. Peroxisome proliferator-activated re- in impaired osteoclast differentiation and osteoporotic features. J Exp Med 2003; ceptor gamma agonism increases the capacity for sympathetically mediated 198:669–675 thermogenesis in lean and ob/ob mice. Endocrinology 2004;145:3925–3934 14. Turnbull IR, Gilfillan S, Cella M, et al. Cutting edge: TREM-2 attenuates 28. Mars M, de Graaf C, de Groot CP, van Rossum CT, Kok FJ. Fasting leptin macrophage activation. J Immunol 2006;177:3520–3524 and appetite responses induced by a 4-day 65%-energy-restricted diet. Int 15. Colonna M. TREMs in the immune system and beyond. Nat Rev Immunol J Obes (Lond) 2006;30:122–128 2003;3:445–453 29. Matter CM, Handschin C. RANTES (regulated on activation, normal T cell 16. Grant RW, Vester Boler BM, Ridge TK, Graves TK, Swanson KS. Adipose expressed and secreted), inflammation, obesity, and the metabolic syndrome. tissue transcriptome changes during obesity development in female dogs. Physiol Circulation 2007;115:946–948 Genomics 2011;43:295–307 30. Schonfeld G, Felski C, Howald MA. Characterization of the plasma lipo- 17. Fujimoto S, Goda T, Mochizuki K. In vivo evidence of enhanced di-methylation proteins of the genetically obese hyperlipoproteinemic Zucker fatty rat. J Lipid of histone H3 K4 on upregulated genes in adipose tissue of diabetic db/db mice. Res 1974;15:457–464 Biochem Biophys Res Commun 2011;404:223–227 31. Ji G, Zhao X, Leng L, Liu P, Jiang Z. Comparison of dietary control and 18. Ge JH, Zhang LZ, Li JX, et al. Replication and gene expression of mutant atorvastatin on high fat diet induced hepatic steatosis and hyperlipidemia in rats. hepatitis B virus in a transgenic mouse containing the complete viral genome Lipids Health Dis 2011;10:23–32 with mutant s gene. World J Gastroenterol 2004;10:3141–3145 32. Boizard M, Le Liepvre X, Lemarchand P, Foufelle F, Ferré P, Dugail I. 19. Kim KA, Kim JH, Wang Y, Sul HS. Pref-1 (preadipocyte factor 1) activates Obesity-related overexpression of fatty-acid synthase gene in adipose tissue the MEK/extracellular signal-regulated kinase pathway to inhibit adipocyte dif- involves sterol regulatory element-binding protein transcription factors. J Biol ferentiation. Mol Cell Biol 2007;27:2294–2308 Chem 1998;273:29164–29171 20. Kim HJ, Zhang XH, Park EY, et al. Metformin decreases meal size and 33. Abumrad NA, el-Maghrabi MR, Amri EZ, Lopez E, Grimaldi PA. Cloning of number and increases c-Fos expression in the nucleus tractus solitarius of obese a rat adipocyte membrane protein implicated in binding or transport of long-chain mice. Physiol Behav 2013;110-111:213–220 fatty acids that is induced during preadipocyte differentiation. Homology with 21. Kim EM, Lee HY, Lee EH, et al. Formaldehyde exposure impairs the function human CD36. J Biol Chem 1993;268:17665–17668 and differentiation of NK cells. Toxicol Lett 2013;223:154–161 34. Saladin R, Fajas L, Dana S, Halvorsen YD, Auwerx J, Briggs M. Differential 22. Kennedy AJ, Ellacott KL, King VL, Hasty AH. Mouse models of the metabolic regulation of peroxisome proliferator activated receptor gamma1 (PPARgamma1) syndrome. Dis Model Mech 2010;3:156–166 and PPARgamma2 messenger RNA expression in the early stages of adipo- 23. Seidell JC. Dietary fat and obesity: an epidemiologic perspective. Am J Clin genesis. Cell Growth Differ 1999;10:43–48 Nutr 1998;67(Suppl.):546S–550S 35. Villareal DT, Apovian CM, Kushner RF, Klein S; American Society for Nu- 24. Butler AA, Kozak LP. A recurring problem with the analysis of energy expenditure trition; NAASO, The Obesity Society. Obesity in older adults: technical review and in genetic models expressing lean and obese phenotypes. Diabetes 2010;59:323–329 position statement of the American Society for Nutrition and NAASO, The Obesity 25. Tschöp MH, Speakman JR, Arch JR, et al. A guide to analysis of mouse Society. Am J Clin Nutr 2005;82:923–934 energy metabolism. Nat Methods 2012;9:57–63 36. Correale C, Genua M, Vetrano S, et al. Bacterial sensor triggering receptor 26. Sell H, Deshaies Y, Richard D. The brown adipocyte: update on its metabolic expressed on myeloid cells-2 regulates the mucosal inflammatory response. role. Int J Biochem Cell Biol 2004;36:2098–2104 Gastroenterology 2013;144:346–356.e3