ATF4/ATG5 Signaling in Hypothalamic Proopiomelanocortin Neurons Regulates Fat Mass via Affecting Energy Expenditure

Decreased energy expenditure and/or increased food intake contributes to the development of obesity (1,2). The central nervous system (CNS), especially the hypothalamus, plays an important role in the regulation of the balance between food intake and energy expenditure (2–4). It has been shown that certain populations of neurons in the arcuate nucleus (ARC) of the hypothalamus play key roles in the regulation of energy homeostasis (5,6). These neurons include those expressing orexigenic neuropeptides neuropeptide Y and agouti-related protein, along with neurons expressing anorexigenic neuropeptides cocaine and amphetamine-related transcript and proopiomelanocortin (POMC) (5,6).

The activating transcription factor 4 (ATF4) belongs to the family of basic zipper-containing proteins (bZIP) (7,8), with broad expression in various tissues, including brain (9). Previous studies have shown that ATF4 global knockout mice are lean (8,10) and resistant to high-fat diet (HFD)– or high-carbohydrate diet–induced obesity and hyperglycemia (10,11). Recently, studies using tissue-specific knockout mice have identified that ATF4 regulates glucose metabolism through its expression in liver and osteoblasts (12,13). In contrast, the role of ATF4 in POMC neurons in metabolic control is largely unknown.

Autophagy is a cellular process through which cells engulf and degrade damaged cytoplasmic components (14). Recent studies have implicated the involvement of autophagy in the regulation of obesity and energy expenditure (15–21). For example, autophagy is reduced in the hypothalamus of aged mice or mice fed an HFD (17,22). Consistently, mice with POMC neuron knockout of autophagy-related gene 7 (ATG7), one of the key regulators of autophagy (23), are obese and have lower energy expenditure (16,17,21), and autophagy protein 5 (ATG5)-transgenic mice have the opposite phenotype (19). Previous studies have shown that ATF4 can regulate autophagy and the expression of some of the autophagy regulators in vitro under different stimulations (24–27); however, the relationship has not been tested in vivo yet.

Given the knowledge mentioned above, the aim of our current study was to explore the role of ATF4 in POMC neurons in the regulation of obesity and energy expenditure and the possible involvement of autophagy in this regulation.

C57BL/6J wild-type (WT) mice were purchased from Shanghai Laboratory Animals Co. Ltd (Shanghai, China). To generate hypothalamic POMC neuron–specific ATF4 knockout mice (PAKO) mice, ATF4-floxed mice (13) were intercrossed with POMC-Cre mice (28), provided by Joel K. Elmquist and Tiemin Liu from Southwestern Medical Center (Dallas, TX). The mice lacking ATF4 and ATG5 in the POMC neurons (double knockout [DKO]) were generated by intercrossing ATF4-floxed and ATG5-floxed mice (29) with POMC-Cre mice. To visualize POMC protein–expressing neurons under fluorescence microscope, POMC-Cre, PAKO, and DKO mice were intercrossed with tdTomato reporter/Ai9 mice (30). All animals were under the C57BL/6J background and housed in animal cages with a 12-h dark/light cycle at 25°C, with free access to water and normal chow diet (NCD). For HFD studies, 8-week-old male mice were fed an HFD with 60% kcal from fat (Research Diets, New Brunswick, NJ) for 3 months. At the age of 22 weeks, mice were sacrificed by CO₂ inhalation. All of the experiments were conducted in accordance with the guidelines of the Institutional Animal Care and Use Committee of the Institute for Nutritional Sciences, Shanghai Institutes for Biological Sciences, Chinese Academy of Sciences.

The mice body composition was measured by a nuclear magnetic resonance system (Bruker, Rheinstetten, Germany). Indirect calorimetry was measured in a comprehensive laboratory animal-monitoring system (Columbus Instruments, Columbus, OH), as previously described (31). Rectal temperature of mice was measured at 14:00 by a rectal probe attached to a digital thermometer (Physitemp, Clifton, NJ). The measurement of food intake was conducted as reported previously (6).

Leptin sensitivity assay was performed as reported previously (32). Mice were individually housed and intraperitoneally (i.p.) injected with PBS for 5 days prior to i.p. injection of leptin (R&D Systems, Minneapolis, MN) twice a day (at 8:00 and 19:00) for 3 days.

Levels of blood glucose and serum insulin were measured by a glucometer and a Mercodia Ultrasensitive Rat Insulin ELISA kit (ALPCO Diagnostic, Salem, NH), respectively. Glucose tolerance tests (GTTs) and insulin tolerance tests (ITTs) were performed by i.p. injection of 2 g/kg glucose to mice after overnight fasting or 0.75 units/kg insulin to mice after 4-h fasting, respectively, at 21 weeks old. HOMA of insulin resistance index and areas under the curve were calculated as described previously (33).

Serum norepinephrine (NE), leptin, growth hormone, and corticosterone levels were measured by an NE ELISA kit (Novus Biologicals, Littleton, CO), leptin ELISA kit (Merck Millipore, Frankfurter, Germany), growth hormone ELISA kit (Merck Millipore), and corticosterone ELISA kit (Novus Biologicals), respectively, according to the manufacturer’s instructions.

Paraformaldehyde-fixed, paraffin-embedded sections of white adipose tissue (WAT) and brown adipose tissue (BAT) were stained with hematoxylin and eosin for histology.

Immunofluorescence (IF) staining with anti-ATF4 antibody (Santa Cruz Biotechnology, Santa Cruz, CA), anti–α-melanocyte–stimulating hormone (α-MSH) antibody (Merck Millipore), and anti-p62 antibody (Progen, Heidelberg, Germany) were performed as described previously (31). Phosphorylated (p-)STAT3 staining was performed as described previously (32,34).

The coding sequence region of ATF4 was amplified from mouse hepatic cDNA and inserted into the eukaryotic expression plasmid pCMV-myc. The ATG5 promoter (−2085 to +1) was amplified from mouse genomic DNA and inserted into the pGL3-basic report plasmid. The ATG5 promoter with the bZIP sites (A: −1962/−1958 and B: −1321/−1317) deleted was generated by site-directed mutagenesis (35).

The primary culture of hypothalamic neurons was performed as described previously (31). Recombinant adenoviruses expressing green fluorescent protein (GFP), ATF4, or dominant-negative (DN) ATF4 were purified and administrated at the dose of 1 × 10⁷ plaque-forming units/well in 12-well plates for 48 h (36). The autophagic flux assays were performed as reported previously (37). 293T cells were cultured in DMEM with 10% FBS. Plasmids were transfected by Lipofectamine 2000 (Invitrogen, Carlsbad, CA) according to the manufacturer’s instructions.

The RNA isolation and RT-PCR were performed as described previously (31). The sequence of primers used is available upon request.

Western blot was performed as described previously (31) with the following primary antibodies: anti-LC3, anti–hormone-sensitive lipase (HSL), and anti–phosphorylated (p-)HSL^Ser660 (Cell Signaling Technology, Beverly, MA); anti–uncoupling protein 1 (UCP1), anti-ATF4, and anti–Tribbles homolog 3 (TRB3; Santa Cruz Biotechnology); anti-p62 (Abcam, Cambridge, U.K.); anti-ATG5 (Proteintech, Chicago, IL); and anti–β-actin (Sigma-Aldrich, St. Louis, MO).

293T cells were cotransfected with the internal control plasmid pRL Renilla (Promega, Madison, WI) or plasmids as indicated; the firefly and Renilla luciferase activities were measured by Dual-Glo Luciferase Assay System (Promega).

Chromatin immunoprecipitation (ChIP) assays were conducted according to the manufacturer’s instructions (Millipore, Danvers, MA) with anti-ATF4 antibody or normal rabbit IgG (both from Santa Cruz Biotechnology). Immunoprecipitated DNA was used as the template to amplify the bZIP sites in the promoter of ATG5 with the primers designed as following. Primers for site A are: forward, 5′-AAGGAGAGGGAAATCTCACCCAAGGGAAAGG-3′ and reverse, 5′-CTAAATAAACTCCGTTCTATGCTATGCCTGT-3′. Primers for site B are: forward, 5′-CTCCACCTACTCAAAGCAGAAATCTCTACTAT-3′ and reverse, 5′-TCTTGTCAGACTTCTGTTGAGGAGAAGCTGGG-3′. The sequences of negative control primers to amplify a DNA fragment 600 bp upstream and not containing the bZIP site are: forward, 5′-GTCAAACTAGAAATTCAGGTCGTCAAG-3′ and reverse, 5′-CAAATAGTGCCTGGCCAGCCTCTTCTG-3′.

Transmission electron microscope analysis was conducted as reported previously (38).

All values are presented as means ± SEM. Differences between groups were analyzed by either Student t test or one-way ANOVA followed by the Student-Newman-Keuls (SNK) test. For GTTs and ITTs, Student t test or one-way ANOVA followed by the SNK test was used to compare the difference between or among different groups of mice at each time point examined. Differences in which P was <0.05 were considered statistically significant.

To investigate whether ATF4 in POMC neurons regulates obesity and energy expenditure, PAKO mice were generated. Furthermore, to label the POMC-expressing neurons in vivo, some of the PAKO mice were intercrossed with Ai9 (tdTomato reporter) mice (30). IF staining showed that ATF4 was colocalized with POMC-expressing neurons in ARC of control mice, but this colocalization was largely reduced in PAKO mice, as evaluated by counting the number of ATF4-positive POMC neurons (Fig. 1A and B). Because the POMC promoter also drives Cre recombinase expression in the pituitary (6), we examined whether the functions of the pituitary–adrenal axis were affected by ATF4 knockout. Serum contents of hormones secreted from the pituitary, including corticosterone and growth hormone (6,17), were comparable between PAKO and control mice (Supplementary Fig. 1A and B). Expression of pituitary genes, including T box transcription factor (Tpit), Pomc, corticotrophin-releasing hormone receptor 1 (Crhr1), growth hormone (Gh), pituitary-specific positive transcription factor 1 (Pit1), and thyroid-stimulating hormone β-chain (Tshb) (6) were also not changed (Supplementary Fig. 1C). Furthermore, anatomical evaluation of POMC neurons throughout the ARC of PAKO and control mice revealed no changes in neuronal population size and distribution (Supplementary Fig. 2).

The body weight, fat mass component, and abdominal fat mass were significantly decreased, whereas the lean mass component was not changed in PAKO compared with control mice maintained on an NCD (Fig. 1C–F). WAT cell volume was decreased, and the protein levels of p-HSL, a key enzyme regulating lipolysis (31,34), were increased in PAKO mice (Supplementary Fig. 3A–C).

The energy homeostasis is maintained by a balance between food intake and energy expenditure (1,4). Food intake was not changed, but the oxygen consumption and energy expenditure were higher, and the respiratory exchange ratio (RER; VCO₂/VO₂) was lower in PAKO mice (Fig. 1G–J). Although the physical activity was not changed, body temperature was much higher in PAKO mice (Fig. 1K and L). The higher body temperature observed in PAKO mice was most likely caused by increased thermogenesis, which is regulated by UCP1 in BAT and serum NE that activates the sympathetic nervous system (SNS) (31,34). Consistently, the cell volume and lipid drops in BAT were decreased (Supplementary Fig. 3D and E), whereas BAT UCP1 expression and serum NE levels were increased in PAKO mice under an NCD (Fig. 1M–O). The expression of browning-related genes peroxisome proliferator-activated receptor γ coactivator 1-α (Pgc1a), Ucp1, and cell death–inducing DFF-like effector A (Cidea) (6) were also increased in the subcutaneous WAT of PAKO mice (Supplementary Fig. 4).

The expression of ATF4 is increased in ARC of mice fed an HFD (34), suggesting that ATF4 in POMC neurons may play a role in HFD-induced obesity. Then, we analyzed the metabolic parameters of PAKO and control mice maintained on an HFD for 3 months. As observed for PAKO mice under an NCD, the body weight, fat mass component, and abdominal fat mass were also lower, and the lean mass component was unchanged in PAKO mice compared with control mice under an HFD (Fig. 2A–D). WAT cell volume was lower, and the levels of p-HSL were higher in HFD-fed PAKO mice (Supplementary Fig. 5A–C). Moreover, HFD-fed PAKO mice had higher oxygen consumption, energy expenditure, and body temperature, with lower RER and no change in food intake and physical activity (Fig. 2E–J). BAT cell volume was lower (Supplementary Fig. 5D and E), whereas BAT UCP1 expression and serum NE levels were higher in HFD-fed PAKO mice (Fig. 2K–M).

Because HFD-induced obesity is normally accompanied by glucose intolerance and leptin resistance, and hypothalamic signaling has an effect on body glucose homeostasis and leptin sensitivity (4,34,39,40), we therefore explored the possible involvement of POMC ATF4 in this regulation in PAKO and control mice maintained on an HFD. As expected, the levels of fed or fasted blood glucose and serum insulin were lower in PAKO compared with control mice maintained on an HFD (Supplementary Fig. 6A–D). Consistently, HOMA of insulin resistance index was significantly lower in HFD-fed PAKO mice (Supplementary Fig. 6E). The glucose tolerance and clearance were also improved in HFD-fed PAKO mice, as demonstrated by GTTs and ITTs, respectively (Supplementary Fig. 6F–I).

In addition, serum leptin levels were lower in HFD-fed PAKO mice (Fig. 3A). To examine the effect of POMC ATF4 on leptin sensitivity, we i.p. administered leptin (32) to PAKO and control mice and compared the effects of leptin injection on changes in food intake and body weight. The effect of leptin on reducing food intake and body weight was much more significant in PAKO mice (Fig. 3B and C). Consistent with the stronger effect of leptin in PAKO mice, leptin injection produced more signals of p-STAT3, the marker of leptin signaling activation (32), in POMC neurons of PAKO mice (Fig. 3D and E). Similar effects of POMC ATF4 on glucose metabolism and leptin sensitivity were also observed in mice maintained on an NCD (Supplementary Figs. 7 and 8).

It is shown that ATF4 can regulate the activity of autophagy and the expression of autophagy regulators in different cell lines under different treatments (24–27), and several autophagy regulators have been indicated in metabolic control (16–20,41), suggesting that POMC ATF4 may regulate energy homeostasis by affecting the expression of some of the autophagy regulators. To test this possibility, we examined the protein levels of autophagy markers LC3-II (a positive autophagy marker) and p62 (a negative autophagy marker) (37,41) in primary cultured hypothalamic neurons with ATF4 overexpression or inhibition. The effects of ATF4 overexpression or inhibition were reflected by the corresponding changes in the expression of TRB3, a well-known downstream effector of ATF4 (42) (Fig. 4A and B). Autophagy was increased (as demonstrated by the increased LC3-II and decreased p62 protein levels) in the hypothalamic neurons infected with adenoviruses expressing DN ATF4 (Ad-DN ATF4) (Fig. 4A). The opposite effect was observed when ATF4 was overexpressed by adenoviruses expressing ATF4 (Ad-ATF4) (Fig. 4B). Similar results were obtained in the ARC of PAKO mice under either diet (Fig. 4C and D). Moreover, p62 staining in POMC neurons of PAKO mice and transmission electron microscope analysis of the number of autophagosomes in the ARC of PAKO mice under an HFD also indicated that the autophagy was increased in the hypothalamus of PAKO mice (Supplementary Fig. 9).

Furthermore, the autophagic flux was increased by ATF4 inhibition and decreased by ATF4 overexpression, respectively, in primary hypothalamic neurons transfected with GFP-LC3, simultaneously infected with Ad-DN ATF4 or Ad-ATF4. Because autophagy is a dynamic process, we further treated the primary hypothalamic neurons with chloroquine (CQ), a lysosomotropic weak base, which blocks the fusion of autophagosome with lysosome (37). In the presence of CQ, the amount of autophagosome was increased in control cells, but the increase was much more significant in Ad-DN ATF4 cells or smaller in Ad-ATF4 cells (Fig. 4E and F).

In order to identify through which autophagy-related regulators ATF4 regulates autophagy, we measured the mRNA expression of autophagy-related proteins Atg4a, Atg5, Atg6 (Becn1), and Atg7, all important for the production of autophagosome (23), in primary hypothalamic neurons with ATF4 overexpression or inhibition. The mRNA levels of Atg4a and Atg5 were increased by ATF4 inhibition, whereas only Atg5 mRNA levels were inhibited by ATF4 overexpression (Fig. 5A and B). Consistently, Atg5 mRNA levels were increased, with inconsistent changes in other Atgs expression, were observed in the ARC of PAKO mice under either diet (Fig. 5C and D). The ATG5 protein was also similarly regulated by ATF4 in vitro and in vivo (Fig. 5E–H). Because two potential ATF4 binding bZIP sites (A: −1962/−1958 and B: −1321/−1317) were identified on the promoter of ATG5 by the online tool MatInspector (43), we conducted a luciferase activity assay to investigate whether ATF4 can regulate ATG5 transcription by direct binding to these sites. As predicted, overexpression of ATF4 inhibits the promoter activity of ATG5, and the inhibitory effect was abolished by the deletion of the two bZIP sites on the promoter of ATG5 (Fig. 5I). Moreover, ChIP assay showed that the two bZIP sites in the promoter of ATG5 were immunoprecipitated by ATF4 antibodies in the ARC of WT mice (Fig. 5J).

To confirm a role of ATG5 in mediating the effects of ATF4 knockout in POMC neurons, we generated POMC neuron ATF4 and ATG5 DKO mice and maintained them and corresponding control mice on an HFD. Knockout of ATG5 (as confirmed by RT-PCR) largely reversed the effects of ATF4 knockout in POMC neurons on body weight, fat mass component, and abdominal fat mass under an HFD (Fig. 6A–C and E). Moreover, the higher oxygen consumption, energy expenditure, body temperature, serum NE levels, BAT UCP1 expression, and WAT p-HSL protein levels and the lower RER, WAT, and BAT cell volume were also largely reversed by ATG5 knockout in POMC neurons of PAKO mice under an HFD, with no changes in lean mass component, food intake, and physical activity (Fig. 6D and F–N and Supplementary Fig. 10). Furthermore, ATG5 knockout in POMC neurons also attenuated the improved glucose tolerance and leptin sensitivity in PAKO mice under an HFD (Fig. 7 and Supplementary Fig. 11).

ATG5 is an important regulator for autophagy (23), and autophagy has been indicated in the regulation of energy homeostasis (15–17,21,41), suggesting that the effect of ATF4 knockout in POMC neurons is possibly mediated via ATG5-dependent autophagy. Autophagy was activated (as demonstrated by the decreased p62 staining) in POMC neurons and non-POMC neurons of PAKO mice under an HFD; however, this increased autophagy was largely reversed by ATG5 knockout (Fig. 8A–C). Furthermore, autophagy affects POMC protein splicing to produce α-MSH (17), a key regulator released to the paraventricular nucleus (PVN) of the hypothalamus controlling energy homeostasis (6,16,17). α-MSH staining, using validated α-MSH antibody (Supplementary Fig. 12), was increased in the PVN of HFD-fed PAKO mice, and the increased α-MSH staining was also reversed by ATG5 knockout (Fig. 8D and E). The processing enzymes prohormone convertase 1 (Pc1/3), prohormone convertase 2 (Pc2), carboxypeptidase E (Cpe), a-amidating monooxygenase (Pam), and prolylcarboxypeptidase (Prcp) are involved in the processing of POMC protein to produce α-MSH, consistently. The expression of Pc2 and Cpe was increased in the hypothalamus of PAKO mice, and the increase was reversed by ATG5 knockout, whereas the expression of the other processing enzymes showed no significant changes (Fig. 8F).

Many populations of neurons in the hypothalamus play key roles in metabolic control (1,4). In this study, we demonstrated a novel function for ATF4 expressed in POMC neurons in the regulation of fat mass: deletion of ATF4 in POMC neurons results in a lean phenotype in mice maintained on an NCD and helps to prevent HFD-induced obesity. It suggests that ATF4 in POMC neurons may be a new potential therapeutic target for treating obesity. POMC promoter also drives Cre recombinase expression in the pituitary (6,17), a region also involved in the regulation of metabolism (44), suggesting that signals from pituitary might contribute to the phenotype in PAKO mice. We speculated that, however, the phenotypes observed in PAKO mice were largely caused by ATF4 deletion in POMC neurons, as there were no significant defects in the pituitary–adrenal axis observed in PAKO mice.

Because body fat mass is maintained by a balance between energy intake and energy expenditure (4,34), the lean phenotype in PAKO mice is likely to be caused by increased energy expenditure, as food intake is not changed in PAKO mice. In contrast to the unaltered physical activities in PAKO mice, thermogenesis, another aspect responsible for energy expenditure occurring in BAT via increased UCP expression by activation of SNS (45), was increased in PAKO mice. Consistently, the expression of browning-related genes (6) were also increased in subcutaneous WAT of PAKO mice. These results suggest that the lean phenotype in PAKO mice is caused by the increased thermogenesis. In addition, lipolysis is another factor that regulates the size of fat mass that is also controlled by the activation of the SNS (31,34). Our results showed that the activity of HSL, the key enzyme regulating lipolysis (17,31,34), is increased in WAT of PAKO mice, suggesting increased lipolysis might also contribute to the lean phenotype in PAKO mice. The increased energy expenditure observed in PAKO mice is consistent to that obtained in ATF4 global knockout mice (8,10), which further demonstrates an important role for ATF4 expressed in POMC neurons in the regulation of energy homeostasis.

A change in fat mass is always associated with a change in glucose metabolism and leptin sensitivity, which controls blood glucose and energy homeostasis, respectively (34,46). In this study, the glucose tolerance and clearance and leptin sensitivity were also improved in PAKO mice under either diet at 21 weeks old, which might be a consequence of the reduced fat mass in these mice, as no differences in glucose tolerance and clearance were observed between PAKO and control mice at 8 weeks old when there was no difference in body weight (Supplementary Fig. 13). Furthermore, leptin-induced STAT3 phosphorylation was increased not only in ARC but also in the ventromedial nucleus, dorsomedial nucleus, and lateral hypothalamic area of PAKO mice (Supplementary Fig. 14). Another possibility is that POMC ATF4 may have a direct effect on glucose metabolism and leptin sensitivity, as adenovirus-mediated ATF4 knockdown in hypothalamus improves insulin sensitivity (36), and many signaling pathways in POMC neurons regulate leptin sensitivity (2,4,6,47). These possibilities, however, require further investigation.

Several autophagy regulators have been demonstrated to play important roles in energy homeostasis (16–19,21,41). For example, knockout of ATG7 or ATG12 in POMC neurons causes obesity and decreases energy expenditure (16,17,21,41). It has been previously shown that knockdown of ATF4 expression activates autophagy in human HT1080 fibrosarcoma cells (24), and ATF4 is also required for stress-induced autophagy regulators ATG5, ATG7, and BECN1 expression in human cancer cells or mouse embryonic fibroblasts (26,27). The regulatory mechanisms underlying ATF4 regulation of ATG5 expression and whether this relationship exists in vivo, however, have not been reported. We showed that the transcription of ATG5 was regulated by ATF4 via direct binding to the bZIP sites in its promoter and also provided the first evidence showing that ATG5 expression is regulated by ATF4 in vivo.

The important role of ATG5 in ATF4-regulated energy homeostasis on an HFD was demonstrated by the larger amount of fat mass and lower energy expenditure observed in DKO mice compared with PAKO mice. In contrast, Malhotra et al. (41) reported that mice with ATG5 knockout in POMC neurons have no obvious phenotypes in body weight and fat mass compared with control mice under an HFD. The induced ATG5 expression in POMC neurons of HFD-fed PAKO mice in our study may be responsible for the different observations. However, this possibility requires further study in the future.

It is well established that ATG5 is required for the production of autophagosome (23), and autophagy in POMC neurons has a significant effect on energy homeostasis (15–17,21). In addition, POMC is a prohormone that needs posttranslational cleavage by several convertases to generate α-MSH (6,48), the process of which is shown to be attenuated by the inhibition of autophagy in mice (17). Inhibition of autophagy also decreases the POMC-derived α-MSH projections to PVN of the hypothalamus by attenuating POMC neuron axon projections in development (15,16). Our results showed that PAKO mice exhibited stimulated autophagy and production of α-MSH in PVN, both of which were largely reversed by ATG5 knockout in POMC neurons of these mice under an HFD, suggesting that POMC ATF4 may regulate obesity and energy expenditure via stimulated ATG5-dependent autophagy and α-MSH levels in the hypothalamus. Though it was unclear, we speculated that the increased α-MSH levels in PVN of PAKO mice was possibly caused by the increased cleavage of POMC precursors and the change in POMC neuron axon projections in development. These possibilities, however, require nvestigation in the future.

Although the hypothalamus of PAKO mice exhibits significant alterations in leptin sensitivity, autophagy activity, and α-MSH levels, factors that are important for the regulation of food intake (6,17), we did not observe any changes in daily or fasting-induced food intake in PAKO mice (Supplementary Fig. 15). Our results are consistent with some other studies showing that pharmacological or genetic blockade of melanocortin receptors in the CNS or global deletion of ATF4 has no effect on food intake (49,50), and mice with XBP1 overexpressed in POMC neurons remain unaltered with regard to food intake in the presence of enhanced leptin sensitivity (47). We hypothesized that the additional signals in POMC neurons or signals in other neurons or brain areas may play dominant roles in the regulation of food intake in PAKO mice.

Taken together, our results identify a new function for central ATF4/ATG5 axis in the regulation of energy expenditure, obesity, and obesity-related glucose metabolism and leptin sensitivity. It also suggests that ATF4 in the CNS may be a new potential therapeutic target for treating obesity and obesity-related metabolic diseases.

Acknowledgments. The authors thank Joel K. Elmquist and Tiemin Liu from Southwestern Medical Center for providing POMC-Cre mice. The authors also thank Noboru Mizushima from the University of Tokyo for providing ATG5 floxed mice.

Funding. This work was supported by grants from National Natural Science Foundation of China (81325005, 81471076, 81570777, 81390350, 81130076, 31271269, 81400792, 81500622, and 81600623), Basic Research Project of Shanghai Municipal Science and Technology Commission (16JC1404900 and 17XD1404200), and International S&T Cooperation Program of China (Singapore 2014DFG32470), and research was supported by the Chinese Academy of Sciences/State Administration of Foreign Experts Affairs International Partnership Program for creative research teams. F.G. was also supported by the 100 Talents Program of the Chinese Academy of Sciences.

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. Y.X. researched data and wrote, reviewed, and edited the manuscript. Y.D., F.Y., T.X., and H.L. researched data. Z. Li, Z. Liu, H.Y., and S.C. provided research materials. Y.L. and Q.Z. contributed to discussion. F.G. directed the project, contributed to discussion, and wrote, reviewed, and edited the manuscript. F.G. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.