A total of 40 overweight/obese patients (BMI ≥ 24, age 42.8 ± 4.5 y) and 40 lean control subjects (BMI < 24, age 44.6 ± 8.3 y) were recruited among clinical outpatients and the patients of a medical examination center at Peking Union Medical College Hospital. BMI was calculated as body mass divided by the square of height (kg m⁻²). All recruited subjects received physical and clinical examinations, and blood samples were collected after an overnight fast for the biochemical measurements. Fasting insulin was measured by a Siemens Centaur XP system (Siemens, Tarrytown, USA), and the homeostasis model assessment estimate of insulin resistance (HOMA-IR) was calculated as fasting insulin (mU L⁻¹) × fasting glucose (mmol L⁻¹)/22.5 (Wallace et al., 2004). Serum ZAG concentrations were determined by commercially available human ZAG enzyme-linked immunosorbent assay (ELISA) kits (Biovendor Laboratorni Medicina, Modrice, Czech Republic) according to the manufacturer's instructions. The intra-assay and inter-assay variation coefficients were 3.9 and 6.6%, respectively.

Human sWAT tissue was collected from 40 morbidly obese patients (BMI ≥ 40) or obese patients (35 ≦ BMI < 40) with obesity-related complications who underwent laparoscopic gastric banding surgery and 21 lean control subjects (BMI < 24, age 48.8 ± 11.3 y) who underwent laparoscopic surgery due to various benign diseases. Abdominal sWAT samples were dissected and then frozen immediately in liquid nitrogen for further studies. The study was approved by the ethics committee of Peking Union Medical College Hospital (No. S-364). All participants signed informed documents before participating in the study.

Total RNA extraction from human sWAT tissue and subsequent reverse transcription were performed as described in our previous studies (Zhu et al., 2015, 2016). In brief, total RNA was isolated from sWAT by use of an E.Z.N.A. Total RNA Kit I (Omega, San Diego, USA). Total RNA in a quantity of 0.5 μg was reverse-transcribed with an Omniscript RT Kit (Qiagen, Maryland, USA). PCR amplification was performed using an ABI 7500 PCR instrument (Applied Biosystems, California, USA), and procedure was conducted in duplicate for each gene. The primer sequences of target genes and internal control genes were listed in Supplementary Table 1. The results were calculated as fold changes in Ct-value relative to the control by using the 2^−ΔΔCt formula (Livak and Schmittgen, 2001).

Eight-week-old male ICR mice (weighing 27–30 g, n = 31, purchased from HFK Bioscience Co. LTD, Beijing, China) were housed at the Institute of Laboratory Animal Sciences, CAMS and PUMC, at an ambient temperature of 22 ± 1°C under a 12/12 h light/dark cycle and fed ad libitum. After 1 week of acclimatization, the mice were weighed and then randomly divided into standard food (SF) group (n = 10), which received SF (protein 22%, fat 11% and carbohydrates 67%), and HFD group (n = 21), which received HFD (protein 42%, fat 41%, and carbohydrates 17%). All animal experimental protocols were carried out in compliance with the standards of the Guide for the Care and Use of Laboratory Animals and approved by the ethics committee of Peking Union Medical College Hospital.

After 8 weeks, HFD-fed mice were weighed and randomly divided into simple HFD group (n = 13, HFD+pcDNA3.1(+) control plasmid) and ZAG overexpression group (n = 8, HFD+ZAG expression plasmid). Mice in the SF group were also intraperitoneally injected with an equal volume of pcDNA3.1(+) control plasmid (n = 10, SF+pcDNA3.1(+) control plasmid). Plasmid transfection of mice was performed as described in our previous study (Gong et al., 2009). In brief, ZAG expression plasmid (5 μg each) or pcDNA3.1(+) negative control plasmid (5 μg each) in 150 μL of OPTI-MEM medium (Invitrogen, Carlsbad, CA, USA) was mixed thoroughly with Lipofectamine 2000 (10 μL, Invitrogen, Carlsbad, CA, USA) in 150 μL of OPTI-MEM medium, and the total 300 μL of mixture was incubated at room temperature for 30 min, then intraperitoneally injected into mice. This injection was performed at a frequency of four times a week for 8 weeks. The body weights of all mice were routinely recorded once a week.

At the end of the intervention, the IPITT and the IPGTT were performed. For the IPITT, mice were fasted for 5 h, and then insulin (Humulin R, Lilly, USA, 0.75 U/kg) was administered intraperitoneally; blood glucose was measured in tail vein at 0, 30, 60, 90, and 120 min with a glucometer (Roche, Basel, Switzerland). For the IPGTT, mice were fasted overnight, then 50% glucose was injected intraperitoneally at a dose of 2 g/kg, and blood glucose was measured as in the IPITT. In addition, the area under the curve (AUC) of glucose concentration between 0 and 120 min was calculated for both the IPITT and the IPGTT.

After 8 weeks of ZAG intervention, WAT (including inguinal (iWAT), mesenteric (mVAT), perirenal (pVAT), and epididymal (eVAT) white adipose tissue) and BAT (interscapular brown adipose tissue) were excised and weighed. Tissue samples from the liver and right hind limb skeletal muscle were also collected. Serum total cholesterol (TC), triglycerides (TG), free fatty acids, low-density lipoprotein cholesterol (LDL-C), high-density lipoprotein cholesterol (HDL-C) and fasting blood glucose levels were determined by routine automated laboratory methods, and fasting serum insulin was determined by a commercially available mouse insulin ELISA kit (Linco Research Inc., MO, USA). The intra-assay coefficient of variation for insulin was 5.9%.

Hematoxylin and eosin (H&E) staining and UCP1 immunohistochemical staining were performed in mouse iWAT according to standard procedures. A primary rabbit polyclonal antibody against UCP1 (1:200 dilution, Abcam, Cambridge, UK) and a secondary goat anti-rabbit IgG antibody (1:2,000 dilution, ZSGB-BIO, Beijing, China) were used in this process. All sections for UCP1 staining were counter-stained with Harris hematoxylin (Histolab, Gothenburg, Sweden) and observed by light microscopy.

Total RNA extraction from mouse WAT, liver, and skeletal muscle tissue and subsequent reverse transcription were performed as described above for human adipose tissue. The primer sequences for the target genes and internal control genes were listed in Supplementary Table 2. The expression levels of target genes in WAT and skeletal muscle were normalized to that of peptidylprolyl isomerase A (PPIA), and the expression levels in the liver were normalized to that of β-actin. The results were expressed as fold changes in Ct-value relative to the control according to the 2^−ΔΔCt formula (Livak and Schmittgen, 2001).

Total proteins were extracted from mouse iWAT and pVAT using a Total Protein Extraction kit (Applygen, Beijing, China), and the protein concentrations were measured with a BCA protein assay reagent kit (Applygen, Beijing, China). Approximately 30 μg of protein was separated on 10% SDS-PAGE gels and transferred onto nitrocellulose membranes (Millipore, Billerica, MA, USA) through a wet transfer (BIO-RAD, California, USA). The membranes were incubated with the primary antibodies (anti-UCP1, Proteintech, Wuhan, China; anti-PGC1α, Proteintech, Wuhan, China; anti-ZAG, Santa Cruz, Dallas, USA; anti-β-actin, CST, Danvers, MA, USA; anti-GAPDH, CST, Danvers, MA, USA) at a 1:100 to 1:1,000 dilution at 4°C overnight, followed by incubation with a goat anti-mouse/rabbit secondary antibody at a 1:5,000 dilution for 1 h (ZSGB-BIO, Beijing, China). The specific protein bands were visualized by enhanced chemiluminescence (Applygen, Beijing, China). The protein bands were quantified by Quantity One software (Version 4.6.9, BIO-RAD, California, USA). The expression levels of target proteins in iWAT were normalized to that of β-actin, and the expression levels in pVAT were normalized to that of GAPDH.

All data are shown as the mean ± s.e.m. except where it is indicated otherwise. Normal distribution of the data was tested using Shapiro–Wilk test. Non-normally distributed parameters were logarithmically transformed to a normal distribution. One-way ANOVA and Dunnett's T3 post-hoc test were used to analyze the differences between groups. Pearson and partial correlation coefficients were used to determine the linear association between ZAG and other parameters. All statistical computations were performed in SPSS 20.0 for Windows (SPSS Inc, Chicago, IL, USA), P < 0.05 was considered statistically significant.

As displayed in Table 1, body weight, BMI, waist circumferences, fat mass, fasting glucose, fasting insulin, HOMA-IR, and diastolic blood pressure (DBP) were significantly higher in overweight/obese patients than in normal controls. However, serum ZAG levels in overweight/obese patients were significantly decreased by 23.6% compared with those in normal controls. Further association analysis found that serum ZAG levels were negatively correlated with BMI (r = −0.336), body weight (r = −0.224) and fat mass (r = −0.407) after adjusting for age and sex in all subjects.

As we expected, obese patients had higher body weight, BMI, waist circumference, DBP, and systolic blood pressure (SBP) than normal-weight controls; details are shown in Table 2 (all P < 0.05). In accordance with the above serological assay results, ZAG mRNA expression in the sWAT of obese patients was significantly decreased by 82.8% in comparison with normal-weight controls (Figure 1A). Partial correlation analysis found that ZAG expression had a significant negative correlation with body weight (r = −0.417), BMI (r = −0.688) and waist circumference (r = −0.699) after adjustment for age and sex (Table 3). Moreover, the expression levels of five WAT browning-related genes—UCP1, peroxisome proliferator-activated receptor gamma coactivator 1 alpha (PGC1α), PR/SET domain 16 (PRDM16), peroxisome proliferator-activated receptor gamma 2 (PPARγ2), and cell death-inducing DFFA-like effector a (CIDEA) were also significantly reduced in obese patients by 48.0, 64.8, 59.1, 44.8, and 59.9%, respectively, compared with the expression in normal-weight controls (all P < 0.01, Figure 1A). Further correlation analysis demonstrated that there was a significant positive relationship in mRNA levels between ZAG and WAT browning related genes, including UCP1 (r = 0.67), PGC1α (r = 0.60), PRDM16 (r = 0.32), CIDEA (r = 0.72), and PPARγ2 (r = 0.59), after adjustment for age, sex and BMI (all P < 0.05, Figures 1B–F).

After 8 weeks of HFD feeding, an obese mouse model was established successfully as evidenced by significant increased body weight, which reached 1.3 times the weight of mice fed with SF, as shown in Figure 2A. During the period of ZAG intervention, the body weight of obese mice in the HFD group always remained higher than that of SF mice, but gradually reduced after ZAG administration. A significant reduction in body weight was observed at 6 weeks after ZAG treatment (56.4 ± 8.5 g vs. 43.7 ± 2.3 g, P < 0.05). At the end of ZAG intervention, the body weight of mice in the ZAG group was reduced by an average of 8.2 g compared with that of the HFD group (Figure 2A). Interestingly, total WAT mass and the percentage of body weight consisting of WAT mass were significantly higher in the HFD group than in the SF group (3.9- and 2.5-fold, respectively), and they significantly decreased by 58.2 and 50.5%, respectively, after 8 weeks of ZAG administration (all P < 0.01, Table 4). The same changes were also observed in different deposits of adipose tissue including iWAT, eVAT, mVAT, and pVAT, which were all significantly increased in HFD mice and markedly decreased after ZAG treatment as presented in Figure 2B (all P < 0.05). However, no significant change was observed in brown fat mass between these three groups (Figure 2B). Further histological HE staining showed that the adipocyte size of iWAT in HFD mice was notably larger than that in the SF group, while ZAG administration markedly reduced adipocyte size as shown in Figure 2C.

The IPGTT showed that the blood glucose levels of ZAG-treated obese mice at 0, 60, 90, and 120 min were significantly reduced by 50.4, 52.8, 56.1, 38.5%, respectively, compared with those of HFD mice (Figure 3A), and the area under the curve (AUC) of glucose also showed a 35.8% decrease in comparison with that in the HFD group (Figure 3C). Furthermore, IPITT also showed a significant reduction in the blood glucose levels of ZAG-treated obese mice at 0, 30, 90, and 120 min by 35.5, 41.3, 71.6, and 71.0%, respectively, in comparison with HFD mice (Figure 3B), and the AUC of glucose also presented a remarkable decrease, falling to 41.8% of the AUC from HFD mice (Figure 3C).

As displayed in Table 4, the fasting glucose levels of HFD-induced obese mice were 53.3% higher than those of SF-fed mice. After these mice were treated with ZAG for 8 weeks, their glucose levels were notably decreased by 35.5% in comparison with those of simple HFD-fed mice. Similar changes were observed in HOMA-IR, which was higher in HFD mice and lower in ZAG-treated mice, although the differences were not statistically significant. In addition, the levels of low-density lipoprotein cholesterol (LDL-C), free fatty acids and uric acid in ZAG-treated mice were all significantly higher than those in the HFD group (all P < 0.01).

As shown in Figure 4A, staining for UCP1-positive adipocytes in the iWAT of HFD-induced obese mice was weaker than that of SF mice, while it was significantly stronger after these mice were treated with ZAG for 8 weeks. In line with these results, the protein levels of UCP1 in the iWAT of ZAG treated mice were significantly increased to 1.8-fold of that of the HFD mice (P < 0.05, Figure 4D). In addition, the mRNA and protein levels of PGC1α, widely known as a key regulator that promotes WAT browning, in the iWAT of ZAG treated mice were significantly increased to 2.2 and 5.3 times in HFD mice, respectively, as presented in Figures 4B,E (all P < 0.05). Further correlation analysis showed that Zag mRNA levels in iWAT were positively related with Pgc1α expression (r = 0.74) (P < 0.05, Figure 4C).

As with iWAT, the expressional changes of WAT browning-related genes after ZAG administration were also investigated in pVAT. The ZAG protein level in the pVAT of ZAG-treated mice was markedly increased by 4.13-fold when compared with that of HFD-fed mice (P < 0.05, Figure 5A). As shown in Figure 5B, the mRNA level of Pgc1a in the pVAT of HFD-induced obese mice was significantly reduced to 62.2% of the value found in SF mice (P < 0.05), while it was significantly increased to 1.9 times the value for HFD-induced obese mice after administration of ZAG for 8 weeks (P < 0.05). In agreement with the changes in transcription levels, the protein level of PGC1α in the pVAT of obese mice was also significantly increased to 3.6 times the value found in HFD-induced obese mice (Figure 5C, P < 0.05). In addition, the mRNA levels of Ucp1 in pVAT had a tendency to increase after ZAG overexpression (Figure 5B). However, there was no significant difference in Ucp1 or Pgc1a mRNA levels in BAT among these three groups of mice (Figure 5D).

As shown in Figure 6A, the mRNA levels of Pgc1a, glucose-6-phosphatase, catalytic (G6pc), phosphoenolpyruvate carboxykinase 1, cytosolic (Pck1) and glycogen synthase 2 (Gys2) in the liver tissue of HFD-induced obese mice were significantly decreased by 83.3, 49.2, 56.7, and 67.3%, respectively, in comparison with that of SF mice (all P < 0.05). After treatment of obese mice with ZAG for 8 weeks, the mRNA levels of Pgc1a were increased to 3.9 times, while the levels of G6pc were notably decreased by 64.1% compared with the HFD-induced obese mice (P < 0.05). The mRNA levels of Gys2 tended to increase after ZAG overexpression (Figure 6A). In addition, the mRNA levels of Pgc1a, glucose transporter type 4 (Glut4), insulin receptor substrate 1 (Irs1) and glycogen synthase 1 (Gys1) in skeletal muscle had no significant changes among these three groups as displayed in Figure 6B.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.