Deletion of ATF4 in Agrp Neurons Promotes Fat Loss Mainly Via Increasing Energy Expenditure

640 Diabetes Volume 66, March 2017

Jiali Deng,1 Feixiang Yuan,1 Yajie Guo,1 Yuzhong Xiao,1 Yuguo Niu,1 Yalan Deng,1 Xiao Han,2 Youfei Guan,3 Shanghai Chen,1 and Feifan Guo1

Deletion of ATF4 in AgRP Neurons Promotes Fat Loss Mainly via Increasing Energy Expenditure

Diabetes 2017;66:640–650 | DOI: 10.2337/db16-0954

Although many functions of activating transcription liver steatosis (1,2). Changes in body weight normally re- factor 4 (ATF4) are identiﬁed, a role of ATF4 in the sult from an imbalance between energy intake and energy hypothalamus in regulating energy homeostasis is un- expenditure (2), controlled by the central nervous system, known. Here, we generated adult-onset agouti-related especially the hypothalamus (3). The center of this reg- peptide neuron–speciﬁc ATF4 knockout (AgRP-ATF4 KO) ulatory network is the arcuate nucleus (ARC) of the hy- mice and found that these mice were lean, with im- pothalamus, which contains sets of important neurons proved insulin and leptin sensitivity and decreased devoted to metabolic regulation including orexigenic neu- hepatic lipid accumulation. Furthermore, AgRP-ATF4 rons that coproduce agouti-related peptide (AgRP) and KO mice showed reduced food intake and increased neuropeptide Y, as well as anorexigenic neurons that con- energy expenditure, mainly because of enhanced ther- tain cocaine- and amphetamine-regulated transcript and mogenesis in brown adipose tissue. Moreover, AgRP- proopiomelanocortin (POMC)–derived peptides (3,4). AgRP ATF4 KO mice were resistant to high-fat diet–induced neurons increase feeding by opposing the anorexigenic obesity, insulin resistance, and liver steatosis and main- actions of POMC neurons, in part through the release of

METABOLISM tained at a higher body temperature under cold stress. Interestingly, the expression of FOXO1 was directly AgRP, a competitive inhibitor of melanocortin receptors regulated by ATF4 via binding to the cAMP-responsive (4). It also had an effect on energy expenditure via affecting element site on its promoter in hypothalamic GT1-7 sympathetic nervous system (SNS) activity or leptin sensi- cells. Finally, Foxo1 expression was reduced in the tivity (5,6). arcuate nucleus (ARC) of the hypothalamus of AgRP- Activating transcription factor 4 (ATF4), also known as ATF4 KO mice, and adenovirus-mediated overexpres- CREBP2, belongs to the CREBP families, characterized by sion of FOXO1 in ARC increased the fat mass in the presence of a leucine zipper dimerization domain and a AgRP-ATF4 KO mice. Collectively, our data demon- basic amino acid–rich DNA binding domain (7,8). ATF4 is strate a novel function of ATF4 in AgRP neurons of ubiquitously expressed in many tissues and some parts of the hypothalamus in energy balance and lipid metab- the brain, including the hypothalamus (8). It is involved in olism and suggest hypothalamic ATF4 as a potential the regulation of various processes, including memory for- drug target for treating obesity and its related meta- mation, osteoblast differentiation, amino acid deprivation, bolic disorders. and redox homoeostasis (9). Recent studies (9–12) have demonstrated a role of ATF4 in the control of glucose and lipid metabolism. A role of ATF4 in speciﬁcneuronsofthe Obesity is strongly associated with metabolic syndrome hypothalamus, however, has not been previously described. and predisposes to diseases including type 2 diabetes and The aim of our current study was to investigate the role of

1Key Laboratory of Nutrition and Metabolism, Institute for Nutritional Sciences, Received 5 August 2016 and accepted 12 December 2016. Shanghai Institute for Biological Sciences, Chinese Academy of Sciences, Grad- This article contains Supplementary Data online at http://diabetes ’ uate School of the Chinese Academy of Sciences, Shanghai, People s Republic of .diabetesjournals.org/lookup/suppl/doi:10.2337/db16-0954/-/DC1. China J.D. and F.Y. contributed equally to this study. 2Key Laboratory of Human Functional Genomics of Jiangsu Province, Jiangsu Diabetes Center, Nanjing Medical University, Nanjing, People’s Republic of China © 2017 by the American Diabetes Association. Readers may use this article as 3Advanced Institute for Medical Sciences, Dalian Medical University, Dalian, long as the work is properly cited, the use is educational and not for proﬁt, and the People’s Republic of China work is not altered. More information is available at http://www.diabetesjournals .org/content/license. Corresponding author: Feifan Guo, [email protected]. diabetes.diabetesjournals.org Deng and Associates 641

ATF4 expressed in AgRP neurons in energy homeostasis Cold Exposure Treatment regulation. The 2- to 3-month-old mice were housed in individual precooled 4°C cages for 3 h, and rectal temperatures of RESEARCH DESIGN AND METHODS mice were measured every 30 min during this period, as Generation of Mice With ATF4 Deletion in AgRP described previously (23). Body weight was measured im- Neurons and Animal Treatment mediately before the cold exposure. fl All animals were on C57BL/6J background. ATF4- oxed Blood Glucose, Serum Insulin, Glucose Tolerance mice (13) and AgRP Cre-ER mice (14) (provided by Joel K. Tests, Insulin Tolerance Tests, and HOMA-Insulin Elmquist and Tiemin Liu, UT Southwestern Medical Center, Resistance Index fl fl Dallas, TX) were bred to generate AgRP-Cre ATF4 ox/ ox The measurements of blood glucose and serum insulin, fl fl and ATF4 ox/ ox littermates, which were named AgRP- results of glucose tolerance tests (GTTs) and insulin ATF4 knockout (KO) and control mice, respectively. For in- tolerance tests (ITTs), and the calculation of the HOMA- ducing Cre expression and avoiding the possible toxic effect insulin resistance (IR) index were conducted as described of tamoxifen (15,16), both control and AgRP-ATF4 mice previously (19). were intraperitoneally injected with 150 mg/kg body weight Leptin Sensitivity Assay In Vivo tamoxifen (Sigma-Aldrich, St. Louis, MO) for 5 days, be- Mice were intraperitoneally injected with either PBS or tween the ages of 5 and 7 weeks (14). The basal metabolic 3 mg/kg leptin (R&D Systems, Minneapolis, MN) at 9:00 A.M. phenotypes in AgRP-ATF4 mice were analyzed by treating after fasting for 24 h, as described previously (24). Food them and control mice with corn oil (Standard Food, Shang- intakeandbodyweightweremeasuredat1and4hafter hai, People’sRepublicofChina)for5days.Forhigh-fatdiet the injection of leptin. (HFD) study, 4-week-old AgRP-ATF4 KO or control mice were maintained on a normal chow diet or HFD with 60% Metabolic Parameters Measurements kcal fat (Research Diets, New Brunswick, NJ) for 16 weeks. The body fat composition of mice was determined using the Pair-fed experiments (17) were conducted by feeding control Bruker Minispec mq10 NMR Analyzer (Bruker, Billerica, mice a normal chow diet in the same amounts of food eaten MA). Indirect calorimetry was measured in a Comprehensive by AgRP-ATF4 KO mice during the previous day. The effi- Lab Animal Monitoring System (Columbus Instruments, ciency for ATF4 deletion was evaluated by mating AI9 Columbus, OH), as described previously (11). Rectal tem- (tdTomato) reporter mice (18) with transgenic mice express- peratures were measured using a rectal probe attached to ing Cre under control of the AgRP promoter after tamoxifen a digital thermometer (Physitemp Instruments, Clifton, treatment. Body weight was monitored weekly throughout NJ). the experiments, and mice were kept as previously described (19). All the experiments were conducted in accordance with Serum and Liver Measurements the guidelines of the Institutional Animal Care and Use Serum and liver total glycerol, total cholesterol, and free Committee of the Institute for Nutritional Sciences. fatty acid levels were determined using Glycerol Assay Kit Reagent (SSUF-C, Shanghai, People’s Republic of China), Cell Culture and Treatments cholesterol reagent (SSUF-C), and NEFA C reagent (Wako, – Pshuttle vector constructed plasmids expressing ATF4 or Osaka, Japan), respectively. Serum norepinephrine (NE) a dominant-negative form of ATF4 (DN-ATF4) was made level was determined using ELISA kits (R&D Systems). All based on plasmids described previously (19). The recom- of these assays were performed according to manufac- binant adenoviruses (Ads) expressing mouse ATF4 (Ad- turer instructions. ATF4) or control green fluorescent protein (Ad-GFP) were generated as previously described (19). Hypotha- Protein and mRNA Analysis lamic GT1-7 cells were maintained as described previously Western blot analysis was performed with primary anti- (20). Plasmids and Ads indicated were transfected into bodies against actin (Sigma-Aldrich); ATF4, tribbles homo- GT1-7 cells with Lipofectamine 2000 (Invitrogen, Carls- log 3 (TRB3), and uncoupling protein 1 (UCP1) (Santa Cruz bad, CA). Biotechnology, Santa Cruz, CA); and t–hormone-sensitive lipase (HSL), phosphorylated (p)-HSL, p-cAMP-dependent ARC Administration Experiments protein kinase (PKA), and FOXO1 (Cell Signaling Technol- ARC administration experiments were conducted as pre- ogy, Danvers, MA); and visualized by ECL Plus (GE Health- viously described (21). Ad-FOXO1 (22) was provided by care, Chicago, IL), as described previously (11). RT-PCR, Professor Youfei Guan (Dalian Medical University, Dalian, with GAPDH as an internal control gene, was carried out People’s Republic of China). Mice were anesthetized and as described previously (11). The sequences of primers used received bilateral stereotaxic injections of Ad-GFP, Ad-Null, in the current study are available upon request. or Ad-FOXO1 (1 mL/5 3 109 plaque-forming units/side/ mice) into ARC (21.4 mm from bregma; 60.3 mm from Histological Analysis of White Adipose Tissue, Brown midline; 25.90 mm from dorsal surface), and metabolic Adipose Tissue, and Liver phenotypes were examined 1–2 weeks after Ad-FOXO1 White adipose tissue (WAT), brown adipose tissue (BAT), injection. and liver were fixed in 4% paraformaldehyde overnight and 642 ATF4 Deletion in AgRP Neurons Promotes Fat Loss Diabetes Volume 66, March 2017 stained with hematoxylin-eosin (H-E), as well as Oil Red O ThebodyweightofmaleAgRP-ATF4KOmicewaslower staining with optimal cutting temperature embedding (25). than that of control mice starting from 8 weeks of age (tamoxifen was given as treatment at 6 weeks of age) Immunofluorescence Staining and was associated with decreased fat and lean mass (Fig. Immunofluorescence (IF) staining was performed with 1A and B). Consistently, decreased weights of subcutane- antibodies against p-STAT3 (Cell Signaling Technology), ous WAT (sWAT), epididymal WAT (eWAT), BAT, and ATF4 (Santa Cruz Biotechnology), FOXO1 (Cell Signaling liver were observed in AgRP-ATF4 KO mice (Fig. 1C). Technology), and c-fos (Santa Cruz Biotechnology), as des- Histological analysis of eWAT showed that the loss cribed previously (24). of ATF4 expression in AgRP neurons resulted in a 40% Luciferase Assay reduction in adipocyte volume, but had no effect on pGL3-Foxo1 promoter (21,137 to 5) (26) were provided eWAT cell numbers as determined by DNA content anal- by Professor Xiao Han (Nanjing Medical University, Nan- ysis (Fig. 1D–F). The diminished adipocyte volume observed jing, People’sRepublicofChina).GT1-7cellswerecotrans- in AgRP-ATF4 KO mice suggested a possible enhanced li- fected with the internal control vector pRL Renilla (Promega, polysis. Consistently, levels of p-HSL, the rate-limiting en- Madison, WI) and plasmids expressing ATF4 using Lipo- zyme for triglyceride (TG) lipolysis, and substrate for PKA, fectamine 2000. The Firefly and Renilla luciferase activi- the kinase that phosphorylates HSL (27), were signifi- ties were assayed using Dual-Glo Luciferase Assay System cantly increased in eWAT of AgRP-ATF4 KO mice com- (Promega). pared with control mice (Fig. 1G). Genes related to other Chromatin Immunoprecipitation Assays lipid metabolism including lipogenesis, fatty acid oxidation fi Chromatin immunoprecipitation (ChIP) assays were per- and TG secretion (28,29), however, were not signi cantly formed according to the manufacturer protocol (EMD affected in the WAT of AgRP-ATF4 KO mice (Supplemen- Millipore, Billerica, MA) with anti-ATF4 antibodies (1:50; tary Fig. 1D). Santa Cruz Biotechnology) or normal rabbit IgG (1:50; AgRP-ATF4 KO Mice Have Decreased Food Intake and Santa Cruz Biotechnology) for negative control. Immuno- Increased Energy Expenditure fi precipitated FOXO1 promoter was quanti ed using PCR To assess the possible reasons for the reduced fat mass with primers designed to amplify the 220 base-pair region in AgRP-ATF4 KO mice, we measured food intake and encompassing the cAMP-responsive element (CRE) site energy expenditure, the two aspects determining body fat 9 9 (forward, 5 -TCAATTCTAAAGCATCCTAGCC-3 ; reverse, mass (27). Daily food intake and feeding efficiency (30) 9 9 5 -TGGGGCACAGCTCGTCTC-3 ) or a 220 base-pair up- were decreased in AgRP-ATF4 KO mice compared with stream region not involved in ATF4 response (forward, control mice (Fig. 2A). Total energy expenditure (24-h O 9 9 2 5 -AACCTTTGTATTGGGGGCAT TGATTG-3 ;reverse, consumption) adjusted to lean mass (31,32) or calculated 9 9 5 -CTGTTGCGATGAGAGCATTTGGTTA-3 ). by ANCOVA (33) and heat generation were signifi- Statistics cantly increased and RER (VCO2/VO2) was decreased in All results are expressed as the mean 6 SEM. Significant AgRP-ATF4 KO mice during both the dark and light – differences were assessed by two-tailed Student t test, phases (Fig. 2B D and Supplementary Fig. 1E). No differ- one-way ANOVA followed by the Student-Newman-Keuls ence in total physical activity was observed between test, or ANCOVA. For energy expenditure, respiratory AgRP-ATF4 KO mice and control mice (Fig. 2E). In con- exchange ratio (RER), and locomotor activity presented trast, rectal temperature was higher in AgRP-ATF4 KO with lines, GTTs, and ITTs, a t test was used to compare mice (Fig. 2F). The higher body temperature observed in the differences between different groups of mice at each AgRP-ATF4 KO mice was most likely caused by increased time point examined. P , 0.05 was considered to be thermogenesis, which is regulated by UCP1 in BAT (34). fi statistically significant. Consistently, BAT UCP1 expression was signi cantly elevated and the BAT cells were denser with fewer lipids in RESULTS AgRP-ATF4KOmicecomparedwithcontrolmice(Fig. 2G–I). BAT thermogenesis is activated by SNS with the Deletion of ATF4 in AgRP Neurons Promotes Fat Loss release of NE (34). Not surprisingly, AgRP-ATF4 KO mice In this study, we generated AgRP-ATF4 KO mice by had higher serum NE levels (Fig. 2J). A pair-fed experiment tamoxifen treatment that allows temporal control of Cre (17) showed that AgRP-ATF4 KO mice had lower body recombinase activity and can be combined with flox mice to weight, fat mass, lean mass, and tissue weight (liver and enable adult-onset deletion (14). IF staining of tdTomato adipose tissue), and higher body temperature (Supplemen- (reflecting AgRP neurons) and ATF4 showed that ATF4 tary Fig. 2). was colocalized with AgRP neurons in control mice but was absent in AgRP neurons of AgRP-ATF4 KO mice AgRP-ATF4 KO Mice Exhibit Improved Insulin (Supplementary Fig. 1A and B). In addition, Atf4 Sensitivity, Decreased Hepatic Lipid Accumulation, mRNA levels were significantly decreased in ARC, but and Increased Leptin Sensitivity not other brain areas or tissues, of AgRP-ATF4 KO mice We then investigated whether ablation of ATF4 in AgRP compared with control mice (Supplementary Fig. 1C). neurons had any impact on insulin sensitivity and liver diabetes.diabetesjournals.org Deng and Associates 643

Figure 1—AgRP-ATF4 KO mice are lean. A: Body weight at the age of 6, 8, and 16 weeks. Fat and lean mass (B) and tissue weights (liver, sWAT, eWAT, and BAT) (C). D: H-E staining of representative abdominal eWAT sections (scale bars, 250 mm). E: Analysis of abdominal eWAT cell volume. F: DNA content of total abdominal eWAT. G: p-PKA substrate proteins, p-HSL, and HSL Western blot and densitometric quantiﬁcation in eWAT. All studies were conducted in 12-week-old (or as indicated) male control or AgRP-ATF4 KO mice maintained on a normal chow diet. The data are expressed as the mean 6 SEM (n =5–7/group) and analyzed by two-tailed Student t test. *P < 0.05.

steatosis, which is associated with a change in fat mass effects of leptin were much more significant in AgRP- (1). Though levels of fed blood glucose and fed and fasting ATF4 KO mice (Supplementary Fig. 4A and B). Consistent serum insulin were comparable between two genotypes, with the stronger effect of leptin in AgRP-ATF4 KO mice, fasting blood glucose levels were decreased in AgRP-ATF4 leptin produced more signals of p-STAT3, the marker of KO mice (Supplementary Fig. 3A and B). Consistently, cellular leptin action (35), in ARC of hypothalamus of HOMA-IR was significantly reduced in AgRP-ATF4 KO these mice (Supplementary Fig. 4C). mice (Supplementary Fig. 3C). GTTs and ITTs further AgRP-ATF4 KO Mice Are Resistant to HFD-Induced revealed that AgRP-ATF4 KO mice had significantly in- Metabolic Disorders creased glucose clearance and improved insulin sensitivity To test whether AgRP-ATF4 mice may play a role (Supplementary Fig. 3D). Possibly because of the reduced under an HFD, we fed 4-week-old male AgRP-ATF4 KO fat mass in AgRP-ATF4 KO mice, hepatic lipid accumula- mice and control mice an HFD for 16 weeks. Similar tion was also slightly reduced in these mice as examined phenotypic changes were observed as those in mice by Oil Red O and H-E staining (Supplementary Fig. 3E). maintained on a normal chow diet. AgRP-ATF4 KO Furthermore, hepatic and serum TG levels were also re- mice showed thin appearance, decreased body weight, duced, although total cholesterol and free fatty acids were unchanged, in AgRP-ATF4 KO mice (Supplementary Fig. lean and fat mass, and weight of tissues, including – 3F and G). liver, sWAT, eWAT, and BAT (Fig. 3A C). Consis- Leptin signaling in hypothalamus is a key regulator tently, the size of adipocytes was much smaller in for energy homeostasis (35). To examine the effect of ATF4 AgRP-ATF4 KO mice, associated with increased levels deletion in AgRP neurons on leptin sensitivity, we in- of p-PKA substrates and p-HSL in eWAT (Fig. 3D and traperitoneally administered leptin (24) to AgRP-ATF4 E). Food intake was comparable between AgRP-ATF4 KO and control mice and monitored the effects of lep- KO mice and control mice under HFD, whereas HFD- tin injection on changes in food intake and body weight. fed AgRP-ATF4 KO mice exhibited markedly increased As shown previously (24), food intake and body weight oxygenconsumption,adjustedtoleanmassorcalcu- were reduced after 1 or 4 h after leptin injection in con- lated by ANCOVA (33), and rectal temperature, with trol mice (Supplementary Fig. 4A and B). Notably, the decreased RER and no change in locomotor activity 644 ATF4 Deletion in AgRP Neurons Promotes Fat Loss Diabetes Volume 66, March 2017

Figure 2—AgRP-ATF4 KO mice have decreased food intake and enhanced energy expenditure. Daily food intake and feeding efﬁciency (A), 24-h oxygen consumption normalized by lean mass (B) or analyzed by ANCOVA (C), RER (VCO2/VO2)(D), locomotor activity (E), and rectal temperature (F). Gene expression (G) and Western blot and densitometric quantiﬁcation (H) of UCP1 in BAT. I: H-E staining of representative BAT sections (scale bars, 250 mm). J: Serum NE levels. All studies were conducted in 10- to 12-week-old male control or AgRP-ATF4 KO mice maintained on a normal chow diet. The data are expressed as the mean 6 SEM (n =4–6/group) and analyzed by two-tailed Student t test in all panels except for C, or ANCOVA in C.*P < 0.05.

(Fig. 3F–K). Consistently, BAT UCP1 expression was sig- control mice (Fig. 4B and D). UCP1 expression and serum nificantly higher, cells were denser with fewer lipids, and NE levels were induced by cold exposure in BAT of control serum NE levels were elevated in AgRP-ATF4 KO mice mice; however, the extent was much higher in AgRP-ATF4 (Fig. 3L–O). In addition, the deletion of ATF4 in AgRP KO mice (Fig. 4E–G). neurons protected mice from HFD-induced IR and liver ATF4 Regulates Expression of FOXO1 via Direct steatosis (Supplementary Fig. 5). Binding to the CRE Site on Its Promoter AgRP-ATF4 KO Mice Maintains at a Higher Body It is shown that FOXO1 in AgRP neurons is critical for the Temperature Under Cold Exposure maintenance of energy homeostasis (36), leading us to Based on the above results, we investigated whether AgRP- investigate the possible involvement of FOXO1 in medi- ATF4 KO mice have enhanced thermogenic response ating the effects of ATF4 deletion in AgRP neurons. In- under cold exposure, a process that has been shown to hibition of ATF4 by a plasmid expressing DN-ATF4 (37), induce thermogenesis (34). After exposure to an ambient as evaluated by the increased ATF4 expression and de- temperature of 4°C, AgRP-ATF4 KO mice constantly creased expression of its downstream target gene TRB3 maintained at a relatively higher core body temperature (19), significantly decreased FOXO1 expression in AgRP- for the period examined, with significantly reduced expressing hypothalamic GT1-7 cells (20) (Fig. 5A and B). weight and cell volume of sWAT and eWAT, compared Opposite effects were observed when ATF4 was over- with control mice (Fig. 4A–C). Although BAT weight expressed (Fig. 5C and D). ATF4 regulates the expres- remained unchanged, BAT was much denser with few sion of downstream target genes via direct binding to lipid droplets in AgRP-ATF4 KO mice compared with CRE sites in their promoters (7). Similarly, ATF4 can diabetes.diabetesjournals.org Deng and Associates 645

Figure 3—AgRP-ATF4 KO mice are resistant to HFD-induced obesity. A: Body weight at the age of 12, 16, and 20 weeks. Fat and lean mass (B) and tissue weights (liver, sWAT, eWAT, and BAT) (C). D: H-E staining of representative eWAT sections (scale bars, 250 mm). E: p-PKA substrate proteins, p-HSL, and HSL Western blot and densitometric quantiﬁcation in eWAT. Daily food intake (F), 24-h oxygen consumption normalized by lean mass (G) or analyzed by ANCOVA (H), RER (VCO2/VO2)(I), locomotor activity (J), and rectal temperature (K). Gene expression (L) and Western blot and densitometric quantiﬁcation (M) of UCP1 in BAT. N: H-E staining of representative BAT sections (scale bars, 250 mm). O: Serum NE levels. All studies were conducted in 16- to 20-week-old (or as indicated) male control or AgRP-ATF4 KO mice under an HFD (HFD feeding starting from age of 4 weeks). The data are expressed as the mean 6 SEM (n =8–10/group) and analyzed by two-tailed Student t test for all panels except for H (analyzed by ANCOVA in H). *P < 0.05. 646 ATF4 Deletion in AgRP Neurons Promotes Fat Loss Diabetes Volume 66, March 2017

Figure 4—AgRP-ATF4 KO mice maintain at a higher body temperature under cold exposure. A: Rectal temperature of mice. B: Tissue weights (sWAT, eWAT, BAT). C: H-E staining of representative eWAT and sWAT sections (scale bars, 250 mm). D: H-E staining of representative BAT sections (scale bars, 250 mm). Gene expression (E) and Western blot and densitometric quantiﬁcation (F) of UCP1 in BAT. G: Serum NE levels. All studies were conducted in 10-week-old male control or AgRP-ATF4 KO mice maintained on a normal chow diet and exposed to a 4°C environment for 3 h. The data were expressed as the mean 6 SEM (n =4–6/group) and analyzed by two-tailed Student t test. *P < 0.05 for the effects of AgRP-ATF4 KO mice vs. control mice after cold exposure in A and B, for the effects of any group of mice vs. control mice prior to cold exposure in E–G;#P < 0.05 for the effects of AgRP-ATF4 KO mice vs. control mice after cold exposure in E–G.

bind to FOXO1 promoter as demonstrated by lucifer- of neuronal peptides, including Agrp, Npy,andPomc, ase assays and ChIP assays in hypothalamic GT1-7 except for Cart, in ARC of AgRP-ATF4 KO mice (Supple- cells transfected with plasmid or adenovirus expressing mentary Fig. 7). ATF4 (Fig. 5E and F). Overexpression of FOXO1 Increases Fat Mass in DISCUSSION AgRP-ATF4 KO Mice In this study, we used an AgRP-Cre-ER transgenic mouse As observed above in vitro, expression of Foxo1 was line that allowed for spatiotemporal gene manipulation decreased in ARC of AgRP-ATF4 KO mice (Fig. 6A). specifically in AgRP neurons after tamoxifen induction To confirm a role for FOXO1 in mediating the effects of Cre recombinase expression to avoid developmental of ATF4 deletion in AgRP neurons, we injected issues or compensating actions (14,40), which may hap- Ad-FOXO1 or control Ad-GFP to ARC of AgRP-ATF4 pen in AgRP-Cre mice (14). As expected, no difference KO and control mice. As predicted, IF staining showed in metabolic parameters examined was observed between that FOXO1 expression was increased in both control the two phenotypes before tamoxifen treatment when and AgRP-ATF4 KO mice, most of which were localized administered with corn oil as the control vehicle (Supple- at ARC (Supplementary Fig. 6A and B). Consistently, mentary Fig. 8). In contrast, after tamoxifen treatment, RT-PCR results showed that Foxo1 was overexpressed AgRP-ATF4 KO mice became lean and resistant to HFD- in ARC but not in other brain areas (Supplementary induced obesity, with improved insulin sensitivity and Fig. 6C). Although there was no significant difference in decreased lipid accumulation in liver. A pair-fed experi- body weight and lean mass, Ad-FOXO1 increased fat mass ment showed that the decreased fat mass was mainly and tissue weights, including sWAT and eWAT, and food caused by increased energy expenditure in AgRP-ATF4 intake; and decreased WAT p-HSL, rectal temperature, KO mice. Our current study for the first time demon- BAT UCP1 expression, and serum NE levels in AgRP- strated a novel function of ATF4 in hypothalamic AgRP ATF4 KO mice (Fig. 6B–M). As shown previously (36,38), neurons for systematic metabolic control. Our results also Ad-FOXO1 also reversed the suppressed neuronal ac- provided novel insights into understanding the signals in tivity,asmeasuredbyIFstainingofc-fos,amarker specific neurons that are critical for the regulation of used to reflect neuronal activity (39), and the expression energy homeostasis. diabetes.diabetesjournals.org Deng and Associates 647

Figure 5—ATF4 regulates the expression of FOXO1 via direct binding to the CRE site in its promoter in vitro. A: Western blot and densitometric quantiﬁcation of FOXO1, ATF4, and TRB3 in GT1-7 cells transfected with plasmids expressing DN-ATF4 or control vector. B: Gene expression of Foxo1 and Trb3 in GT1-7 cells transfected with plasmids expressing DN-ATF4 or control vector. C: Western blot and densitometric quantiﬁcation of FOXO1, ATF4, and TRB3 in GT1-7 cells transfected with plasmids expressing ATF4 or control vector. D: Gene expression of Foxo1 and Trb3 in GT1-7 cells transfected with plasmids expressing ATF4 or control vector. E: Luciferase activity was assessed in GT1-7 cells expressing the Foxo1 promoter vector with or without a plasmid expressing ATF4. F: ChIP assay was performed in GT1-7 cells infected with Ad-GFP or Ad-ATF4. NC, negative control. All studies were conducted in GT1-7 cells with at least three independent experiments. The data are expressed as the mean 6 SEM and analyzed by two-tailed Student t test. *P < 0.05, for any treatment compared with control group.

In this study, we found that the deletion of ATF4 in AgRP-ATF4 KO mice. Given the importance for leptin AgRP neurons also improves insulin sensitivity and reduces sensitivity in body weight control, we speculated that hepatic lipid accumulation in normal chow diet–fed mice the increased leptin sensitivity may contribute to the and protects mice from HFD-induced IR and liver stea- decreased food intake and enhanced energy expenditure tosis. We speculated that the above effects in AgRP-ATF4 in AgRP-ATF4 KO mice. This possibility, however, must KO mice may result from the decreased fat mass in these be studied in the future. mice, as many studies (1,35,41) have demonstrated that Because body fat mass is maintained by a balance a reduction in body weight helps to improve insulin sen- between food intake and energy expenditure, we explored sitivity and decrease lipid accumulation in liver. On the the possible reasons responsible for the decreased fat other hand, it might be directly regulated by signals from mass in AgRP-ATF4 KO mice from these two aspects. It is hypothalamus (42), possibly via vagus nerve (19,42,43). previously shown that global deletion of ATF4 had no Therefore, we could not exclude the possibility that ATF4 effect on food intake (9,11). It is also reported that the in AgRP neurons may have a direct effect on hepatic in- activation of a certain signal in ARC is sufficient to inhibit sulin sensitivity and lipid accumulation, particularly given food intake, associated with ATF4 overexpression (45). In the fact that hypothalamic ATF4 has a significant impact contrast, food intake was reduced in AgRP-ATF4 KO mice. on regulating hepatic insulin sensitivity via the vagus The difference in the effect of ATF4 deletion on food nerve (19). intake, however, might be due to the difference in the Leptin binds to its receptors expressed in the hypo- way of deleting ATF4 under each different circumstance. thalamus and regulates neural circuits that suppress food In addition, AgRP-ATF4 KO mice had increased energy intake and increase energy expenditure (35,44). Here, we expenditure and body temperature, which reflect an in- showed that the deletion of ATF4 in AgRP neurons im- crease in thermogenesis. BAT is of major importance in proves leptin sensitivity, as demonstrated by the much the regulation of thermogenesis and energy expendi- more significant inhibitory effects of leptin on food in- ture via affecting UCP1 expression (34). It is shown that take and body weight, and p-STAT3 signaling in ARC of UCP1 expression in BAT is induced by the activation of 648 ATF4 Deletion in AgRP Neurons Promotes Fat Loss Diabetes Volume 66, March 2017

Figure 6—Overexpression of FOXO1 increases fat mass in AgRP-ATF4 KO mice. A: Gene expression of Foxo1 in ARC. B: Body weight at the seventh day after virus injection. Fat and lean mass (C) and tissue weights (liver, sWAT, eWAT, and BAT) (D). E: H-E staining of representative eWAT sections (scale bars, 250 mm). F: Analysis of abdominal eWAT cell volume. G: P-PKA substrate proteins, p-HSL, and HSL Western blot and densitometric quantiﬁcation in eWAT. Daily food intake (H) and rectal temperature (I). Gene expression (J) and Western blot and densitometric quantiﬁcation (K) of UCP1 in BAT. L: H-E staining of representative BAT sections (scale bars, 250 mm). M: Serum NE levels. All studies were conducted 7 days after receiving Ad-GFP or Ad-FOXO1 bilaterally in ARC in 10- to 12-week-old male control or AgRP-ATF4 KO mice maintained on a normal chow diet. The data are expressed as the mean 6 SEM (n =6–7/group) and analyzed by one-way ANOVA followed by the Student-Newman-Keuls test. *P < 0.05 for the effects of any group of mice vs. control mice injected with Ad-GFP; #P < 0.05 for the effects of Ad-FOXO1 vs. Ad-GFP in AgRP-ATF4 KO mice. diabetes.diabetesjournals.org Deng and Associates 649

SNS, which stimulates the release of NE binding to expression. Consistently, we found neuronal activity as b3-adrenergic receptor on the surface of BAT (2,34). The examined by IF staining of c-fos, a marker reflecting neu- upregulation of UCP1 expression results in increased ther- ronal activity (39), and Agrp expression are changed in mogenesis and energy expenditure, which helps in protec- the ARC of AgRP-ATF4 KO mice after overexpression of tion from fat accumulation (34). In this study, we found FOXO1, indicating that similar mechanisms may mediate that AgRP-ATF4 KO mice exhibit increased body tempera- FOXO1 regulation of energy homeostasis in AgRP-ATF4 ture, BAT UCP1 expression, and serum NE levels, suggest- KO mice. Thus, our study demonstrates an important role ing increased thermogenesis. It is well known that fat for FOXO1 as a downstream target for ATF4. In addition, mobilization is promoted by SNS activation, which stimu- we identified a direct effect of ATF4 on the regulation of lates the release of NE binding to b3-adrenergic receptor on FOXO1 expression. In contrast to our observation, it is the surface of adipocytes and sequentially phosphorylates shown that FOXO1 expression is not affected by ATF4 in PKA and HSL (34). Our results show that WAT cell volume osteoblast cells (50). The difference might be caused by was decreased in AgRP-ATF4 KO mice, suggesting increased tissue-specific regulatory mechanisms, which require fur- lipolysis and altered nutrient partitioning that may contrib- ther investigation. The possible influence from FOXO1 ute to the maintenance of the energy balance. The altered expressed in other neurons by ARC injection and the nutrient partitioning in AgRP-ATF4 KO mice could be due involvement of other pathways in ATF4 regulation of to the insufficient energy intake (27,34), which is needed to energy homeostasis, however, needs to be studied in mobilize more fat to be used as an energy source. On the the future. other hand, however, it might also be induced by activated Taken together, these results demonstrate a critical role SNS activity or increased insulin sensitivity (34). The in- for ATF4 in hypothalamic AgRP neurons in the regulation creased thermogenesis in BAT and possibly the increased of energy homeostasis, and lipid and glucose metabolism lipolysis in WAT should certainly contribute to the de- mainly by the increased energy expenditure via affecting creased fat mass in AgRP-ATF4 KO mice. FOXO1 expression. These results also suggest hypotha- In this study, we also explored the possible role of lamic ATF4 as a potential novel drug target in treating ATF4 in AgRP neurons after treatment with an HFD or obesity and its related metabolic disorders. cold exposure. Thermogenesis is one of the most important adaptive changes in response to these environmental stimuli and functions to maintain metabolic homeostasis Acknowledgments. The authors thank Joel K. Elmquist and Tiemin Liu or to protect the organism from cold exposure (34). As (UT Southwestern Medical Center, Dallas, TX) for providing AgRP Cre-ER mice. observed in mice maintained on a control diet, we found Funding. This work was supported by grants from National Natural Science Foundation (81130076, 81325005, 31271269, 81300659, 81400792, 81471076, that AgRP-ATF4 KO mice also exhibit enhanced thermo- 81570777, 81500622, and 81390350), Basic Research Project of Shanghai genesis, resulting in the resistance to HFD-induced obe- Science and Technology Commission (16JC1404900), and International S&T Co- sity, IR, and liver steatosis, and a higher body temperature operation Program of China (Singapore 2014DFG32470) and by the international after cold exposure. Interestingly, food intake was com- Partnership Program For Creative Research Teams (CAS/SAFEA). F.G. was also parable between HFD-fed control and AgRP-ATF4 KO supported by the One Hundred Talents Program of the Chinese Academy of mice, which is different from the observed decreased Sciences. food intake in these mice maintained on a control diet. Duality of Interest. No potential conflicts of interest relevant to this article Although it is shown AgRP neurons are important for were reported. energy intake control, however, when palatable food is Author Contributions. J.D. and F.Y. researched the data and wrote, provided, AgRP neurons are dispensable for an appropri- reviewed, and edited the manuscript. Y. Guo and Y.N. researched the data. Y.X. and ate feeding response (46). These results suggest that AgRP Y.D. contributed to the writing of the manuscript and helpful discussion. X.H., Y.Gua., and S.C. provided research material. F.G. directed the project, contributed to neurons are indispensible in the feeding response under a discussion, and wrote, reviewed, and edited the manuscript. F.G. is the guarantor of control chow diet, but not under HFD, which might ex- this work and, as such, had full access to all the data in the study and takes plain the difference in food intake between AgRP-ATF4 responsibility for the integrity of the data and the accuracy of the data analysis. KO and control mice maintained on a normal chow diet or HFD. References Several signaling pathways in AgRP neurons have been 1. Grundy SM. Adipose tissue and metabolic syndrome: too much, too little or identified to be important regulators of energy homeo- neither. Eur J Clin Invest 2015;45:1209–1217 stasis (20,47–49). In this study, we found that the ef- 2. Spiegelman BM, Flier JS. Obesity and the regulation of energy balance. Cell – fects of ATF4 deletion in AgRP neurons are mediated by 2001;104:531 543 inhibited expression of FOXO1, based on the effects of 3. Morton GJ, Cummings DE, Baskin DG, Barsh GS, Schwartz MW. Central nervous system control of food intake and body weight. Nature 2006;443:289–295 FOXO1 overexpression in ARC on increasing fat mass in 4. Cowley MA, Smart JL, Rubinstein M, et al. Leptin activates anorexigenic AgRP-ATF4 KO mice. However, body weight was not af- POMC neurons through a neural network in the arcuate nucleus. Nature 2001; fected by FOXO1 overexpression, possibly because of the 411:480–484 simultaneously decreased tendency of lean mass. A pre- 5. Blouet C, Liu SM, Jo YH, Chua S, Schwartz GJ. TXNIP in Agrp neurons vious study (36) has shown that FOXO1 regulates energy regulates adiposity, energy expenditure, and central leptin sensitivity. J Neurosci homeostasis via affecting AgRP neuronal activity and Agrp 2012;32:9870–9877 650 ATF4 Deletion in AgRP Neurons Promotes Fat Loss Diabetes Volume 66, March 2017

6. Yang L, McKnight GS. Hypothalamic PKA regulates leptin sensitivity and 28. Tang Y, Wallace M, Sanchez-Gurmaches J, et al. Adipose tissue mTORC2 adiposity. Nat Commun 2015;6:8237 regulates ChREBP-driven de novo lipogenesis and hepatic glucose metabolism. 7. Karpinski BA, Morle GD, Huggenvik J, Uhler MD, Leiden JM. Molecular Nat Commun 2016;7:11365 cloning of human CREB-2: an ATF/CREB transcription factor that can negatively 29. Xiao F, Deng J, Guo Y, et al. BTG1 ameliorates liver steatosis by decreasing regulate transcription from the cAMP response element. Proc Natl Acad Sci USA stearoyl-CoA desaturase 1 (SCD1) abundance and altering hepatic lipid metab- 1992;89:4820–4824 olism. Sci Signal 2016;9:ra50 8. Vallejo M, Ron D, Miller CP, Habener JF. C/ATF, a member of the activating 30. Hinton AO Jr, He Y, Xia Y, et al. Estrogen receptor-a in the medial amygdala transcription factor family of DNA-binding proteins, dimerizes with CAAT/ prevents stress-induced elevations in blood pressure in females. Hypertension enhancer-binding proteins and directs their binding to cAMP response ele- 2016;67:1321–1330 ments. Proc Natl Acad Sci USA 1993;90:4679–4683 31. Himms-Hagen J. On raising energy expenditure in ob/ob mice. Science 9. Seo J, Fortuno ES 3rd, Suh JM, et al. Atf4 regulates obesity, glucose ho- 1997;276:1132–1133 meostasis, and energy expenditure. Diabetes 2009;58:2565–2573 32. Butler AA, Kozak LP. A recurring problem with the analysis of energy ex- 10. Li K, Zhang J, Yu J, et al. MicroRNA-214 suppresses gluconeogenesis by penditure in genetic models expressing lean and obese phenotypes. Diabetes targeting activating transcriptional factor 4. J Biol Chem 2015;290:8185–8195 2010;59:323–329 11. Wang C, Huang Z, Du Y, Cheng Y, Chen S, Guo F. ATF4 regulates lipid 33. Tschöp MH, Speakman JR, Arch JR, et al. A guide to analysis of mouse metabolism and thermogenesis. Cell Res 2010;20:174–184 energy metabolism. Nat Methods 2011;9:57–63 12. Yoshizawa T, Hinoi E, Jung DY, et al. The transcription factor ATF4 regulates 34. Lowell BB, Spiegelman BM. Towards a molecular understanding of adaptive glucose metabolism in mice through its expression in osteoblasts. J Clin Invest thermogenesis. Nature 2000;404:652–660 2009;119:2807–2817 35. Myers MG Jr, Olson DP. Central nervous system control of metabolism. 13. Schwartz MW, Woods SC, Porte D Jr, Seeley RJ, Baskin DG. Central nervous Nature 2012;491:357–363 system control of food intake. Nature 2000;404:661–671 36. Ren H, Orozco IJ, Su Y, et al. FoxO1 target Gpr17 activates AgRP neurons to 14. Wang Q, Liu C, Uchida A, et al. Arcuate AgRP neurons mediate orexigenic regulate food intake. Cell 2012;149:1314–1326 and glucoregulatory actions of ghrelin. Mol Metab 2013;3:64–72 37. He CH, Gong P, Hu B, et al. Identiﬁcation of activating transcription factor 15. Ye R, Wang QA, Tao C, et al. Impact of tamoxifen on adipocyte lineage 4 (ATF4) as an Nrf2-interacting protein. Implication for heme oxygenase-1 gene tracing: inducer of adipogenesis and prolonged nuclear translocation of Cre re- regulation. J Biol Chem 2001;276:20858–20865 combinase. Mol Metab 2015;4:771–778 38. Kim MS, Pak YK, Jang PG, et al. Role of hypothalamic Foxo1 in the regu- 16. Mottillo EP, Desjardins EM, Crane JD, et al. Lack of adipocyte AMPK ex- lation of food intake and energy homeostasis. Nat Neurosci 2006;9:901–906 acerbates insulin resistance and hepatic steatosis through brown and beige 39. Curran T, Morgan JI. Fos: an immediate-early transcription factor in neu- adipose tissue function. Cell Metab 2016;24:118–129 rons. J Neurobiol 1995;26:403–412 17. Herman AM, Ortiz-Guzman J, Kochukov M, et al. A cholinergic basal forebrain 40. Joly-Amado A, Denis RG, Castel J, et al. Hypothalamic AgRP-neurons feeding circuit modulates appetite suppression. Nature 2016;538:253–256 control peripheral substrate utilization and nutrient partitioning. EMBO J 2012;31: 18. Madisen L, Zwingman TA, Sunkin SM, et al. A robust and high-throughput 4276–4288 Cre reporting and characterization system for the whole mouse brain. Nat 41. Könner AC, Brüning JC. Selective insulin and leptin resistance in metabolic Neurosci 2010;13:133–140 disorders. Cell Metab 2012;16:144–152 19. Zhang Q, Yu J, Liu B, et al. Central activating transcription factor 4 (ATF4) 42. Pocai A, Obici S, Schwartz GJ, Rossetti L. A brain-liver circuit regulates regulates hepatic insulin resistance in mice via S6K1 signaling and the vagus glucose homeostasis. Cell Metab 2005;1:53–61 nerve. Diabetes 2013;62:2230–2239 43. van den Hoek AM, van Heijningen C, Schröder-van der Elst JP, et al. In- 20. Kaushik S, Rodriguez-Navarro JA, Arias E, et al. Autophagy in hypothalamic tracerebroventricular administration of neuropeptide Y induces hepatic insulin AgRP neurons regulates food intake and energy balance. Cell Metab 2011;14: resistance via sympathetic innervation. Diabetes 2008;57:2304–2310 173–183 44. Friedman JM, Halaas JL. Leptin and the regulation of body weight in 21. Nakajima K, Cui Z, Li C, et al. Gs-coupled GPCR signalling in AgRP neurons mammals. Nature 1998;395:763–770 triggers sustained increase in food intake. Nat Commun 2016;7:10268 45. Maurin AC, Benani A, Lorsignol A, et al. Hypothalamic eIF2a signaling 22. Li J, Chi Y, Wang C, et al. Pancreatic-derived factor promotes lipogenesis in regulates food intake. Cell Rep 2014;6:438–444 the mouse liver: role of the Forkhead box 1 signaling pathway. Hepatology 2011; 46. Denis RG, Joly-Amado A, Webber E, et al. Palatability can drive feeding 53:1906–1916 independent of AgRP neurons. Cell Metab 2015;22:646–657 23. Bal NC, Maurya SK, Sopariwala DH, et al. Sarcolipin is a newly identiﬁed reg- 47. Kitamura T, Feng Y, Kitamura YI, et al. Forkhead protein FoxO1 mediates ulator of muscle-based thermogenesis in mammals. Nat Med 2012;18:1575–1579 Agrp-dependent effects of leptin on food intake. Nat Med 2006;12:534–540 24. Zhang Q, Liu B, Cheng Y, et al. Leptin signaling is required for leucine 48. Liu T, Kong D, Shah BP, et al. Fasting activation of AgRP neurons requires deprivation-enhanced energy expenditure. J Biol Chem 2014;289:1779–1787 NMDA receptors and involves spinogenesis and increased excitatory tone. 25. Carson FL. Histotechnology: A Self-Instructional Text. Chicago, American Neuron 2012;73:511–522 Society for Clinical Pathology Press, 1997 49. Claret M, Smith MA, Batterham RL, et al. AMPK is essential for energy 26. Chen F, Zhu Y, Tang X, et al. Dynamic regulation of PDX-1 and FoxO1 homeostasis regulation and glucose sensing by POMC and AgRP neurons. J Clin expression by FoxA2 in dexamethasone-induced pancreatic b-cells dysfunction. Invest 2007;117:2325–2336 Endocrinology 2011;152:1779–1788 50. Kode A, Mosialou I, Silva BC, et al. FoxO1 protein cooperates with ATF4 27. Zechner R, Zimmermann R, Eichmann TO, et al. FAT SIGNALS–lipases and protein in osteoblasts to control glucose homeostasis. J Biol Chem 2012;287: lipolysis in lipid metabolism and signaling. Cell Metab 2012;15:279–291 8757–8768