NAG-1/GDF15 Exerts Anti-diabetic Effects through the Insulin Signaling Pathway

Nonsteroidal anti-inammatory drug-activated gene-1 (NAG-1) has a role in various diseases. Here, the anti-diabetic effects of NAG-1 were evaluated using a high-fat diet (HFD)/streptozotocin (STZ)-induced diabetic mouse model. NAG-1-overexpressing transgenic (NAG-1 Tg) mice exhibited lower bodyweight and fasting blood glucose and insulin levels than wildtype (WT) mice. The homeostatic model assessment of insulin resistance scores of NAG-1 Tg mice were three times lower than those of WT mice. Hematoxylin and eosin staining demonstrated that NAG-1 Tg mice exhibited a smaller lipid droplet size in the adipose tissues, lower lipid accumulation in the hepatocytes, and larger beta cell area in the pancreas, compared to WT mice. Immunohistochemical analysis revealed downregulated expression of cleaved caspase-3, an apoptosis marker, in the beta cells of NAG-1 Tg mice. The adiponectin and leptin mRNA levels were upregulated and downregulated in NAG-1 Tg mice, respectively. Additionally, the IRS1/PI3K/AKT signaling pathway, especially Foxo1, which regulates gluconeogenesis in the muscle and WAT, was downregulated in NAG-1 Tg mice. Furthermore, NAG-1 overexpression promoted As160 (AKT substrate of 160 kDa) expression in both muscles and adipocytes and the mRNA levels of the Nlrp3 pathway members were downregulated in NAG-1 Tg mice. Our ndings suggest that NAG-1 expression alleviates diabetes in mice. increased insulin sensitivity in mice.


Introduction
Nonsteroidal anti-in ammatory drug-activated gene 1 (NAG-1), which is also known as growth differentiation factor 15 (GDF15), macrophage inhibitory cytokine-1 (MIC-1), placental transforming growth factor β (TGF-β) (PTGFB), and prostate derived factor (PDF), is a member of the TGF-β superfamily. The signaling and function of NAG-1, which are mainly dependent on the health of the tissues and body, are associated with cell and tissue homeostasis. Previous studies have reported that patients with critical illnesses or those under stress exhibit upregulated serum NAG-1 levels 1 . In contrast, the expression of NAG-1 is downregulated under normal physiological conditions 2 . Recent studies have reported that NAG-1 is a major regulator of energy metabolism and appetite 3 . Additionally, NAG-1 is reported to contribute to the pathogenesis of various diseases, such as cardiovascular diseases, obesity, in ammation, cognitive impairment, and cancer 1 . NAG-1 exerts its effects, probably through the mature NAG-1 and pro-NAG-1 forms, which are secreted into the extracellular matrix [4][5][6] . In contrast to that of mature NAG-1, the biological activity of pro-NAG-1 in the extracellular matrix has not been investigated, except the nuclear translocation of intercellular pro-NAG-1 regulates the transcription of the members of the SMAD pathways 4 .
The increasing prevalence of diabetes is a global health burden. Type 2 diabetes mellitus (T2DM) is characterized by persistent hyperglycemia. The major phenotypes of T2DM include dysregulated insulin response in the corresponding cells and insu cient insulin production in the pancreas. Furthermore, prolonged high blood glucose levels can lead to other complications, such as diabetic retinopathy, diabetic neuropathy, diabetic foot, cognitive de cits and cardiovascular diseases 7 . Obesity and sedentary lifestyle are the major causes of diabetes that contribute to increased morbidity and mortality 8 .
There are contradictory reports on the role of NAG-1 in diabetes. For example, a study examining the correlation between NAG-1 and beta cell function demonstrated that the serum NAG-1 levels were upregulated in the prediabetes and diabetes groups. Additionally, the upregulated serum NAG-1 levels were correlated with HbA1c, glucose, insulin, baseline and dynamic indices of insulin sensitivity 9 . These results suggest that glucose upregulates serum NAG-1 levels in patients with diabetes and that serum NAG-1 could be a potential marker to identify individuals who are at risk for developing diabetes and obesity. On the other hand, NAG-1 Tg mice exhibit decreased energy metabolism and bodyweight and improved glucose and insulin response 10 . Moreover, the administration of recombinant NAG-1 decreased food intake and bodyweight, which resulted in decreased adiposity, and improved glucose tolerance and insulin sensitivity in obese and wildtype (WT) mice 11 . Interestingly, Metformin, a rst-line drug for T2DM, promotes bodyweight loss through the upregulation of NAG-1 12 . Although the role of NAG-1 in the pathogenesis of diabetes is not clear, NAG-1 expression, both mature NAG-1 and pro-NAG-1 forms, may play a role in the disease process.
The correlation between obesity-induced chronic in ammation and insulin resistance has been extensively investigated. The NLR family pyrin domain containing 3 (NLRP3) in ammasome, which plays an important role in regulating in ammatory cytokines, is associated with metabolic disorders 13 . The activation of the NLRP3 in ammasome promotes the secretion of interleukin-1β (IL-1β) and interleukin-18 (IL-18), which induce in ammation. The downregulated NLRP3 in ammasome activity in NAG-1 Tg mice may confer resistance against diet-induced obesity and improve insulin sensitivity 14 . Another study suggested that IL-1β and IL-18 are associated with obesity-associated in ammation and promote insulin resistance 15 . Thus, NLRP3 may be a novel therapeutic target to alleviate in ammasome hyperactivation in patients with diabetes.
This study aimed to elucidate the role of NAG-1 in insulin signaling, which is the major pathway that regulates cellular metabolic pathways, such as glucose and lipid metabolism. Mice were fed on high-fat diet (HFD) and administered with a low dosage of streptozotocin (STZ) to induce T2DM and the IRS1/AKT/PI3K signaling pathway was examined. The ndings of this study indicated that NAG-1 regulated the insulin pathway and thus NAG-1 could be a biomarker for determining the e cacy of antidiabetic agents.

NAG-1 Tg mice exhibit decreased bodyweight and blood glucose level
It has been reported that the bodyweight of NAG-1 Tg mice is reported to be lower than that of sibling WT mice in both genetically modi ed and HFD-induced obesity animal models 10,16,17 . In this study, the effect of NAG-1 expression on STZ-treated mice was measured. The bodyweight of NAG-1 Tg mice was signi cantly lower than that of WT mice ( Fig. 2A), which concurred with the ndings of a previous study on non-STZ-treated mice 10 . The incremental changes in bodyweight between weeks 6 and 11 in NAG-1 Tg mice (20.65±3.27%) were markedly lower than those in WT mice (46.24±15.35%) (Fig. 2B). The blood glucose levels in NAG-1 Tg mice signi cantly decreased, especially after week 9, which was the timepoint at which STZ treatment was initiated (Fig. 2C). The change in blood glucose level between baseline and nal week in WT mice (86.20±39.63 mg/dL) was markedly higher than that in NAG-1 Tg mice (9.60±12.85 mg/dL) (Fig. 2D). Thus, the bodyweight and blood glucose levels decreased in NAG-1 Tg mice, indicating that NAG-1 expression decrease the blood glucose level in the HFD/STZ-induced diabetic mouse model.

NAG-1 Tg mice exhibit increased insulin sensitivity
The levels of fasting blood glucose (FBS) and insulin were measured and intraperitoneal glucose tolerance test (IPGTT) and intraperitoneal insulin tolerance test (IPITT) were performed before the mice were euthanized. As shown in Fig. 3A (Fig. 3B). IPGTT and IPITT were performed to measure glucose tolerance and insulin sensitivity, respectively, in both NAG-1 Tg mice and WT mice. As shown in Fig. 3C and 3D, WT mice exhibited higher serum glucose levels than NAG-1 Tg mice during the 120-min period after glucose administration. The area under the curve (AUC) of glucose in NAG-1 Tg mice was signi cantly lower than that in WT mice than that in WT mice. These results demonstrated that the NAG-1 Tg mice were more sensitive to insulin than WT mice. Further, insulin resistance and insulin sensitivity were calculated using the mathematical indices homeostatic model assessment of insulin resistance (HOMA-IR) (Fig. 3H). The HOMA-IR score (2.16 ± 0.27) in NAG-1 Tg mice was markedly lower than that in WT mice (7.00 ± 2.50), which indicated that NAG-1 overexpression increased insulin sensitivity in mice.
Effects of diabetic dyslipidemia and hepatic steatosis on HFD/STZ-induced diabetic mouse model The results of histological staining of the liver, inguinal white adipose tissue (iWAT), epididymal white adipose tissue (eWAT), and brown adipose tissue (BAT) are shown in Fig. 4A. Compared with those in the hepatic sections of WT mice, lipid accumulation and the number of fat droplets were lower in NAG-1 Tg mice. This suggests that NAG-1 overexpression decreased hepatic steatosis. Furthermore, adipose tissue hypertrophy in NAG-1 Tg mice was lower than that in WT mice. The quanti cation of the histological data revealed that the size of lipid droplets in the iWAT, eWAT, and BAT of NAG-1 Tg mice was signi cantly smaller than that in the iWAT, eWAT, and BAT of WT mice (p < 0.001) (Figs. 4B-D). The number of crownlike structure (CLS), a histological hallmark of the pro-in ammatory process in the adipose tissue, of NAG-1 Tg mice was lower than that in the adipose tissue of WT mice (p < 0.001), which suggested that the in ammatory lesions decreased in NAG-1 Tg mice (Figs. 4E and F).
Effects of NAG-1 overexpression on pancreatic islet architecture T2DM development is accompanied by the loss of β cells in the pancreatic islets 18 . The effect of NAG-1 overexpression on pancreatic islets was examined using hematoxylin and eosin (H&E) staining and immunohistochemistry. H&E analysis of the pancreas (Figs. 5A-C) revealed that the islet area in NAG-1 Tg mice was signi cantly larger than that in WT mice ( 19 , in the pancreatic islets of NAG-1 Tg mice was lower than that in the pancreatic islets of WT mice (Fig. 5D). This indicated that NAG-1 overexpression may delays the development of T2DM and decreases apoptosis in the pancreatic islets.
Effect of NAG-1 overexpression on mRNA expression level of the insulin signaling pathway members To con rm the NAG-1 expression in the NAG-1 Tg mice, we performed RT-PCR using tissues from muscle, iWAT and BAT. Human NAG-1 gene was only expressed in the tissues from NAG-1 Tg mice, whereas no mRNA expression of human NAG-1 gene in WT mice (Fig. 6A). Further, to elucidate the molecular mechanisms underlying the NAG-1-mediated regulation of blood glucose level, quantitative real-time polymerase chain reaction (qRT-PCR) analysis were performed to analyze the expression levels of several genes related to the insulin signaling pathway. As shown in Fig. 6B, C, and D, NAG-1 Tg mice exhibited upregulated mRNA expression levels of Irs1, Glut4, Akt, Pi3k, and As160 in the muscle, iWAT, and BAT.
These results indicate that NAG-1 may improve glucose metabolism and attenuate insulin resistance through the activation of the IRS1/AKT/PI3K signaling pathway. These results were also con rmed by measuring the mRNA expression levels of the downstream effectors of the AKT pathway, including Gsk3b, Foxo1, and Mtor1. The expression levels of downstream effector genes were signi cantly downregulated in the muscles, iWAT, and BAT of NAG-1 Tg mice. Additionally, the expression level of Ptpn1, a negative regulator of AKT, was downregulated in the muscles, iWAT, and BAT of NAG-1 Tg mice.
Interestingly, the expression level of Mtor2 was signi cantly upregulated in the muscles but not in the iWAT and BAT. The evidence reported that Mtorc2 in muscle tissue contributes to glucose homeostasis by positively regulating insulin-stimulated phosphorylation of the Akt substrate As160 and negatively regulating basal glycogen synthase activity 20 . In contrast, Mtorc2 in adipose tissues controls expression of the lipogenic transcription factor ChREBPβ, which increase adipose tissue de novo lipogenesis and impairs hepatic insulin sensitivity 21 . The mRNA expression level of Pparg was signi cantly upregulated in the iWAT and BAT but not in the muscles. Since the adipose tissue is the major mediator of PPARγ activity on insulin sensitivity, PPARγ activation in the mature adipocytes induces the expression of several genes involved in the insulin signaling cascade and consequently contribute to enhancing insulin sensitivity 22 .
Effect of NAG-1 overexpression on the mRNA expression levels of adipocytokines Adiponectin and leptin, which are major adipokines, are important mediators of energy homeostasis. The downregulation of adiponectin has been associated with insulin resistance, dyslipidemia, and atherosclerosis. Leptin decreases appetite, stimulates thermogenesis, enhances fatty acid oxidation, decreases glucose level, and reduces bodyweight and fat mass 23 . The effect of NAG-1 overexpression on the mRNA expression levels of adiponectin and leptin was examined in NAG-1 Tg mice. The mRNA level of adiponectin was upregulated, while that of leptin was downregulated in the muscles and iWAT of NAG-1 Tg mice. However, the mRNA level of adiponectin in the BAT was not signi cantly different between WT and NAG-1 Tg mice (Fig. 7).
Effect of NAG-1 overexpression on the mRNA expression level of Nlrp3 in ammasome NLRP3 in ammasomes are involved in the secretion of in ammatory cytokines, which promote insulin resistance. Therefore, this study investigated the effects of NAG-1 overexpression on the expression of the NLRP3 in ammasome in HFD/STZ-induced diabetes mouse model. As shown in Figs. 8A-C, the mRNA expression levels of Nlrp3 and Casp1 in the muscles, iWAT, and BAT of NAG-1 Tg mice were signi cantly lower than those in the muscles, iWAT, and BAT of WT mice. Compared with those in the iWAT of WT mice, the mRNA levels of Asc were signi cantly downregulated in the iWAT but not in the muscle and BAT of NAG-1 Tg mice. Furthermore, the mRNA levels of Il18 were signi cantly downregulated in the muscles and iWAT of NAG-1 Tg mice when compared with those in the muscles and iWAT of WT mice. In contrast, the mRNA levels of Il1b in the BAT of NAG-1 Tg were signi cantly lower than those in the BAT of WT mice. The expression levels of Nlrp3 and Casp1 were downregulated in the muscles, iWAT, and BAT of NAG-1 Tg mice. This indicated that the decreased NLRP3 in ammasome activity in NAG-1 Tg mice may alleviate diet-induced obesity and enhance insulin sensitivity in HFD/STZ mouse model.

Discussion
Mature NAG-1/GDF15 signaling in the brain exerts anti-obesity effects by regulating appetite 24 and consequently decreases bodyweight and fat mass 17 . Consistently, this anti-obesity effect was inhibited in NAG-1 knockout mice 16 . Previously, we had reported that NAG-1 is a novel therapeutic target for obesity and insulin resistance as it modulates metabolic activity by promoting the expression of key thermogenic and lipolytic genes in the BAT and WAT 10 . As NAG-1 can decrease bodyweight, it has potential applications for the prevention and treatment of obesity and hyperglycemia, which is one of the main chronic symptoms of diabetes.
The molecular mechanisms of NAG-1 in diabetes have not been elucidated. However, recent studies have suggested that treatment with NAG-1/GDF15 inhibited diabetes in 53% of non-obese diabetic mice 25 . Non-obese diabetic mice are commonly used as models to screen the therapeutic agents for type 1 diabetes because they spontaneously develop autoimmune diabetes, which mimics the diabetic condition in humans. Further evidence of the anti-diabetic role of NAG-1 was obtained from the use of metformin, an anti-diabetic drug. Treatment with metformin consistently induces weight loss in individuals with or without T2DM. Additionally, metformin is reported to upregulate the expression of NAG-1/GDF15 12 , which indicated that the mechanism of action of metformin involves the induction of NAG-1 expression. However, it has been reported that the serum levels of NAG-1/GDF15 in patients with T2DM are higher than those in non-diabetic individuals 26 . Additionally, the serum levels of NAG-1/GDF15 are positively correlated with blood concentrations of glucose 27 . The mechanism underlying the upregulation of serum NAG-1/GDF15 levels in patients with diabetes has not been elucidated. One potential explanation is the presence of several forms of NAG-1 in the serum. NAG-1 is synthesized in the cytoplasm and secreted in at least two different forms of NAG-1 (pro-form and mature form). However, the molecular function of the pro-NAG-1 form in the serum has not been elucidated. The role of pro-NAG-1 in diabetes development must be examined and thus the ratio of mature form of NAG-1 to pro-form of NAG-1 could determine the progression of diabetes in patients. Furthermore, intercellular NAG-1 may also have a role in the adipose tissue, which exhibits NAG-1 expression. Previously, we had reported that pro-NAG-1 decreases SMAD activity in the nucleus through transcriptional regulation 4 , which possible contribute to regulate several other mechanisms of transcriptional regulation, including micro RNA and long non-coding RNA expression. The elucidation of pro-NAG-1 function can aid in determining the role of NAG-1 in the development of diabetes.
To elucidate the mechanism involved in NAG-1/GDF15-mediated enhanced insulin sensitivity, this study investigated the insulin signaling pathway using NAG-1 Tg mouse tissues. In addition to the adipose tissue, the skeletal muscle was considered as the target tissue in which NAG-1/GDF15 was expressed. Glucose uptake by this tissue accounts for the majority of glucose disposal during the postprandial period, as well as after the challenge with exogenous insulin. The expression of GLUT4 was upregulated in the skeletal muscle of NAG-1/GDF15 Tg mice (Fig. 6), supporting the hypothesis that NAG-1/GDF15 expression in the skeletal muscle may increase systemic insulin sensitivity. The upregulated expression of Glut4 is accompanied by the induction of the expression of As160, which is required for the insulin-stimulated translocation of glucose transporter to the plasma membrane. Indeed, the expression of Glut4 was also upregulated in the iWAT and BAT (Fig. 6).
In obesity and diabetes, PI3K/AKT is the major insulin pathway involved in physiological functions of various organs. The PI3K/AKT signaling pathway promotes lipid biosynthesis and inhibits lipolysis in the adipose tissue. Foxo1, which is the substrate of PI3K/AKT, regulates lipolysis by modulating the expression of adipose triglyceride lipase (ATGL). Additionally, the modulation of the PI3K/AKT signaling pathway and the downstream effectors is a potential therapeutic strategy for obesity and T2DM. However, the complex mechanism of the PI3K/AKT pathway must be elucidated in future studies. This study demonstrated that the metabolic functions of the PI3K/AKT signaling pathway in the muscle and adipose is critical for metabolism and that NAG-1 is a positive regulator of this pathway.
The mTOR protein modulates insulin signaling by regulating several downstream components 28 . The mTORC1 is a downstream effector of AKT, is involved in the conversion of BAT to WAT. Previous studies have reported that mTORC1 activation in the adipose tissue promotes the accumulation of lipids in BAT through the downregulation of brown adipocyte markers and the upregulation of WAT markers. In the skeletal muscle, Mtorc1 regulates muscle mass by regulating protein synthesis and degradation 29 . The overactivation of mTORC1 in the muscle of obese and high-fat-fed rodents is associated with increased inhibitory phosphorylation of insulin receptor substrate-1 (IRS-1) and impaired AKT activation, which lead to impaired insulin signaling, decreased glucose uptake by the muscle, and systemic insulin resistance 30 . The ndings of this study indicated that mTORC1 is downregulated in the BAT but not in the WAT, which suggested decreased conversion of BAT into WAT in NAG-1 Tg mice. This may be associated with low bodyweight and small size of droplets (Figs. 2 and 4). Moreover, Mtorc1 was also downregulated in the muscles, which indicated decreased insulin resistance in NAG-1 Tg mice. mTORC2, a component of the mTOR complex, promotes AKT signaling through phosphorylation 31 . The knockout of mTORC2 in the muscle leads to the impairment of insulin signaling through GLUT4 translocation 20 . Additionally, the knockout of Mtorc2 in the adipose tissues suppressed insulin-induced lipolysis, which increased the levels of circulating fatty acids and glycerol 32 . The ndings of this study indicated that the expression of Mtorc2 changes in the muscle but not in the adipose tissue of NAG-1 Tg mice. This suggested that the induction of Mtorc2 expression in the muscles of NAG-1 Tg mice is associated with the upregulation of As160, which leads to the translocation of GLUT4 to the plasma membrane. Interestingly, NAG-1 regulated the expression of leptin and adiponectin in this study. Studies on animal models have revealed that leptin and adiponectin play a critical role in the prevention and control of T2DM by promoting beta cell function and survival, improving insulin sensitivity, and regulating glucose metabolism 33 . Leptin and adiponectin regulate blood glucose through several mechanisms. In this study, NAG-1 overexpression downregulated the expression of leptin and upregulated the expression of adiponectin (Fig. 7). Previous studies have reported that increased leptin levels are associated with insulin resistance and T2DM development. Additionally, increased leptin concentrations are associated with increased cardiovascular risk in patients with T2DM 34 . In this study, NAG-1 Tg mice exhibited downregulated expression of leptin in the WAT, BAT, and muscle, and decreased insulin resistance when compared with WT mice. Adiponectin also contributes to improving insulin sensitivity by locally increasing GLUT4-mediated glucose uptake. However, adiponectin also inhibits BAT activation and thermogenesis in mice by suppressing Ucp1 expression, lipolysis, and brown adipocyte recruitment 35 . This demonstrated that increased adiponectin levels are sometimes associated with metabolic dysfunction in BAT 36 . These data are consistent with the improved insulin sensitivity and increase in thermogenesis in the BAT of NAG-1 Tg mice. These results are consistent with those of our previous study, which suggested that NAG-1 mice exhibit upregulated expression of thermogenesis-related genes in the BAT 10 .
NLRP3 in ammasome is the most fully characterized of the in ammasomes belongs to the nucleotidebinding oligomerization domain (NOD)-like receptor (NLR) family of pattern recognition receptors (PRRs) which found in cytosol to detecting pathogens invasion and initiating immune responses. The NLRP3 in ammasome contains 3 main components which are apoptosis-associated speck-like protein containing a CARD (ASC), proin ammatory caspase-1 and NLRP3. The activation of this in ammasome leads to secretion promotion of the proin ammatory cytokines interleukin 1β (IL-1β) and interleukin 18 (IL-18) which play roles in T2DM progression. Several studies reported that NLRP3 in ammasome and its components are crucial mechanisms that induces metabolic in ammation and insulin resistance 37 .
In this study, the results show the decreasing of NLRP3 and caspase-1 mRNA expression in muscle, iWAT and BAT followed by the reduction of IL1β and IL-18 mRNA expressions in different tissues of NAG-1 Tg diabetic mice, compared to WT mice. Similarly, in our previous research reported NAG-1 Tg mice with CHO, low-fat and high-fat diet exhibited no signi cant but lower expression of NLRP3 mRNA 14 and signi cant decline of caspase-1, ASC, IL1β and IL18 mRNA expression in WAT of NAG-1 Tg obese mice, compared with WT mice 10 . This suggests that NAG-1 Tg mice have negative correlation with NLRP3 in ammasomes which play important roles for protection against insulin resistance.
In summary, the ndings of this study demonstrate that NAG-1 may inhibit diabetic development in an HFD/STZ-induced diabetic mouse model. Based on the ndings of this study, we conclude that NAG-1 upregulates insulin signaling and consequently downregulates Gsk3β, mTorc1, and Foxo1, which may contribute to the prevention of HFD/STZ-induced diabetic mice (Fig. 9). Histological analysis revealed that NAG-1 mitigates the adverse effects of diabetes on the liver and adipose tissues. NAG-1 overexpression exerts anti-diabetic activity by increasing the pancreatic islet area and the beta cell population in the pancreas. These results also highlight the potential role of NAG-1 in the development of novel preventive approaches to manage insulin resistance.

Animals
The procedures for animal care and experiments were approved by the Institutional Animal Care and Use Committees of Seoul National University (SNU-170417-23-2).
Six-week-old male C57BL/6 mice (wildtype and NAG-1 Tg) were used in this study. The animals were housed in a temperature-controlled (21 ± 2°C) room with a 12-h dark/light cycle. The mice had free access to food and water.
All mice were maintained on HFD (60% from fat, D12492, Research Diets, Inc., New Brunswick, NJ) throughout the experimental period and the study was carried out in compliance with the ARRIVE guidelines.

Experimental design
Mice were grouped into the following two groups: WT mice and NAG-1 Tg mice (n = 5 for each group).
After three weeks of maintenance on HFD, the mice were and intraperitoneally injected with STZ (40 mg/kg bodyweight; #13104, Cayman Chemical, MI, USA) after 8 h of fasting. The mice were continued to be maintained on HFD (Fig. 1). At week 10, the mice were again administered with STZ. The levels of FBG were analyzed to con rm diabetes in mice at week 11.

Measurements of bodyweight and fasting FBG and insulin levels
Bodyweight and FBG were initially measured at week 6 and monitored every week thereafter until the end of the experimental period. The FBG levels were measured using the fresh blood samples collected from the mouse tail vein with Accu-Chek® Performa (Roche Diabetes Care, Indianapolis, IN, USA). The fasting serum insulin levels were determined using the ultra-sensitive mouse insulin ELISA kit (#90080, Crystal Chem, IL, USA), following the manufacturer's instructions.

IPGTT
The mice were allowed to fast for 18 h and subjected to IPGTT on week 11. Mice were intraperitoneally injected with fresh glucose solution (2 g/kg bodyweight, A2494001, Gibco, Gaithersburg, MD, USA) and the blood glucose level was measured at 0 (baseline),15, 30, 60, and 120 min post-glucose injection using Accu-Chek® Performa. The AUC value of insulin was calculated using GraphPad Prism software (GraphPad Prism 8 Software Inc., San Diego, CA, USA).

IPITT
At week 12, the mice were allowed to fast for 6 h before the experiment. The serum glucose and insulin levels were measured using Accu-Chek® Performa at baseline (0 min) and after intraperitoneal injection of insulin (0.75 U/kg bodyweight) (I0516, Sigma-Aldrich, St. Louis, MO, USA). The blood glucose levels and serum insulin levels were also determined at 15, 30, 60, and 120 min post-injection. Insulin resistance was calculated using the following formula: HOMA-IR index = glucose level × (serum insulin level/22.5)

Histological analysis
The pancreas, liver, muscle, iWAT, eWAT, and BAT were xed with paraformaldehyde and subjected to H&E staining. The tissues were xed in 10% neutral formalin, embedded in para n, sectioned (thickness: 5 µm), and stained with H&E. The images were captured using Panoramic SCAN (3DHISTECH, Budapest, Hungary).

Immunohistochemistry
The tissue blocks were xed in 10% formalin, embedded in para n, sectioned to 5-µm-thick sections. The sections were incubated in citrate buffer for 1 min in a microwave (Immunobioscience, Mukilteo, WA, USA) to retrieve the antigens. Endogenous peroxidase activity was blocked using 0.3% hydrogen peroxide (Sigma-Aldrich, St. Louis, MO, USA) in phosphate-buffered saline for 1 h at room temperature.
Immunoreactivity was detected using an ultra-sensitive ABC staining kit (Thermo Scienti c, Rockford, IL, USA), following the manufacturer's instructions. The sections were incubated with rabbit polyclonal anticleaved caspase-3 (#9661S, Cell Signaling Technology, Beverly, MA, USA) antibody at 4°C overnight. The negative control samples were incubated with secondary antibodies but not primary antibody. Quantitative reverse transcription polymerase chain reaction (qRT-PCR) Total RNA was extracted from tissues using the RNeasy mini kit (QIAGEN, Hilden, Germany), following the manufacturer's instructions. RNA (1 µg) was reverse transcribe into cDNA using the rst strand cDNA synthesis kit (K1612, Thermo Scienti c™, Waltham, MA, USA) in a MiniAmp Plus Thermal Cycler (A37835, Applied Biosystems, Marsiling Industrial Estate, Singapore). The primers are used for qRT-PCR analysis are shown in Table 1. The relative level of each RNA was measured using qRT-PCR with SYBR Green reagents (PowerUp™ SYBR™ Green Master Mix, A25741, Applied Biosystems, Thermo Scienti c™) in QuantStudio™ 1 real-time PCR system (Applied Biosystems, Marsiling Industrial Estate, Singapore). The expression level of the target gene was normalized to that of Gapdh (housekeeping gene). The relative gene expression was calculated using the comparative Ct (2 −ΔΔCt ) method.

Statistical analysis
Data processing and analysis were performed using GraphPad Prism 8.0.1. All experimental data are represented as the mean ± standard error of the mean for parametric data. The means between two groups were analyzed using the Student's t-test. The data of NAG-1 Tg mice were compared with those of controls (WT mice). The differences were considered signi cant at *p < 0.05, **p < 0.01, or ***p < 0.001.