Adipocyte Metrnl Antagonizes Insulin Resistance Through PPARγ Signaling

Insulin resistance is a common feature of morbid obesity and a critical risk factor for metabolic syndrome, type 2 diabetes, and cardiovascular diseases (1–4). New insulin-sensitizing therapies are needed (1,3). Adipose tissue releases various adipokines for metabolic homeostasis and disease (3,5). Both brown adipose tissue (BAT) and subcutaneous white adipose tissue (WAT) appear metabolically protective (1,4–8). To discover potential beneficial adipokines, we have screened secretory proteins from either brown fat or subcutaneous white fat (9). Most of our previous studies are focused on the function and molecular mechanism of Visfatin, which is highly expressed in brown fat, and we demonstrate this protein acting via Nampt enzyme activity in fat and other tissues (10–13). Recently, we have identified a novel adipokine Metrnl, also known as Subfatin (9) and Cometin (14), whose expression is abundant in subcutaneous white fat and relatively low in brown fat (9). The function of Metrnl is largely unknown, due to very limited studies on this protein (9,14–16), only two of which investigated the function of this protein (14,15).

Metrnl is undetected (14) or at low (9) expression in the adult brain, but it is expressed in the restricted sites of the brain during development and may act as a neurotrophic factor to promote neurite outgrowth (14). Our previous data show that Metrnl is a secreted protein, highly expressed in rat, mouse, and human subcutaneous WAT, and easily detected in the incubation medium of WAT (9), suggesting a role of Metrnl in white adipose biology and metabolic homeostasis. We thus generated both adipocyte-specific knockout and overexpression mice to explore the function of Metrnl.

Very recently, progress has been made for Metrnl function on metabolism by the Spiegelman laboratory (15). The study showed that Metrnl expression was upregulated in WAT upon acute cold exposure at 4°C and in muscle after an acute bout of concurrent exercise or forced overexpression of PGC1α4. Increasing circulating levels of Metrnl induced browning of WAT via an eosinophil-dependent increase of IL4 expression and M2 macrophage activation. However, this action chain only lasted for <1 week after either intravenous injection or adipose local injection of adenoviral vectors encoding Metrnl (15). The chronic effect of adipose Metrnl on metabolism and its mechanism are still unclear.

In this study, we demonstrate that Metrnl promotes differentiation of white adipocytes in vitro. Through creating conditional knockout and transgenic overexpression mice, we show that adipocyte Metrnl is an insulin-sensitizing factor in vivo. Furthermore, we show that the insulin sensitization of Metrnl is attributed to the improvement of adipose dysfunction and is mediated by the PPARγ pathway.

A targeting vector was constructed by inserting LoxP sites flanking exon 3 and the coding region of exon 4 of the Metrnl gene with a neomycin cassette flanked by FRT sites (Supplementary Fig. 1A). After homologous recombination in C57BL/6-derived ES cells and selection with neomycin, positive surviving clones were identified with Southern blot and used for generating chimeras, which were back-crossed with C57BL/6 mice to generate Metrnl^loxP-neo/wt in Ozgene Pty. Ltd. (Perth, Australia). Metrnl^loxP-neo/wt mice were verified with Southern blot and PCR using primers in Supplementary Table 1 (Supplementary Fig. 1B and C). Fabp4-Cre mice [B6.Cg-Tg(Fabp4-cre)1Rev/J; JAX stock 005069] were used to generate adipocyte-specific Metrnl knockout mice (Metrnl^−/−). The breeding strategy for Metrnl^−/− was provided in Supplementary Fig. 1D–G. Littermate Cre-negative Metrnl^loxP/loxP mice were used as control, similar to previous studies (8). All animal experiments were performed in accordance with the National Institutes of Health Guide for the Care and Use of Laboratory Animals and were approved by the animal ethical committee of the Second Military Medical University.

The cDNA of mouse Metrnl tagged with his₆ was subcloned into the NotI site of a plasmid that contains a 5.4-kb promoter/enhancer of Fabp4 (Addgene). Transgenic mice expressing Metrnl-his₆ gene were obtained by microinjection of fertilized eggs from C57BL/6J mice at the Shanghai Research Center for Model Organisms (Shanghai, China). These transgenic mice were back-crossed to C57BL/6J mice for four to six generations before being used for experiments as transgenic mice with adipocyte-specific overexpression of Metrnl (Metrnl Tg). Littermate wild-type mice were used as control.

Heterozygous leptin knockout mice (leptin^+/−) were purchased from the Shanghai Research Center for Model Organisms and crossed with Metrnl Tg to obtain leptin^+/−Metrnl Tg, which were further crossed with leptin^+/− to generate adipocyte-specific overexpression of Metrnl in leptin-knockout mice (leptin^−/−Metrnl Tg) (Supplementary Fig. 3B). Littermate leptin^−/− mice were used as control.

Human serum samples were collected from a group of males undergoing a physical examination. The written informed consents were obtained from all subjects, and the research was approved by the medical ethical committee of the Second Military Medical University. The basic clinical data of these subjects were described in Supplementary Table 2.

Male mice were used for the study. Animals were housed under controlled temperature (23–25°C) and lighting (12-h light-dark cycle) with free access to tap water and standard chow (Sino-British SIPPR/BK Laboratory Animal Ltd.) or high-fat diet (HFD) containing about 60% of calories from lipids (D12492; Research Diets). Energy expenditure and respiratory exchange ratio of mice were measured by indirect calorimetry using an Oxymax system (TSE Systems). In brief, mice were housed in individual chambers. CO₂ and O₂ levels were detected every 21 min for a period of 3 days after 2 days of acclimation.

Metrnl Tg and wild-type mice were fed HFD or HFD with PPARγ inhibitor (GW9662, 8 mg/kg/day; Sigma-Aldrich) for 16 weeks. Otherwise, Metrnl Tg and wild-type mice were fed HFD for 16 weeks and subsequently administered another PPARγ inhibitor (30 mg/kg/day; bisphenol A diglycidyl ether [BADGE], Sigma-Aldrich) (17) or vehicle via intraperitoneal injection for 3 weeks. Then, glucose tolerance test (GTT) and insulin tolerance test (ITT) were performed to evaluate insulin sensitivity of these mice. To achieve PPARγ knockdown in vivo, lentivirus containing the scramble or PPARγ short hairpin RNA (shRNA) (18) was administrated to Metrnl Tg and wild-type mice fed HFD via intravenous injection (10⁸ PFU/mouse). GTT was performed 10 days after injection.

For treatment of Metrnl^−/− mice with recombinant Metrnl protein, a catheter was placed into an external jugular vein of the mouse and tunneled subcutaneously and exteriorized between the scapulae under anesthesia. After 2 days of recovery, the conscious mouse was intravenously given recombinant Metrnl (1.75 μg/mouse/day) or vehicle for 7 days.

Both GTT and ITT were performed as we previously described (12).

Hyperinsulinemic-euglycemic clamp studies were performed as previously described (19), with some modifications. In brief, mice were anesthetized with sodium pentobarbital (75 mg/kg), and an infusion catheter was inserted into the right jugular vein. Insulin (Novo Nordisk) was given primarily (150 mU/kg) and was administered continuously (150 mU/kg/h) via intravenous infusion. Blood glucose concentration was monitored every 5 min via tail vein bleeding. And the intravenous infusion rate of a 20% glucose solution was adjusted to maintain blood glucose at 5 mmol/L for >30 min.

Blood glucose was detected with a OneTouch UltraSmart Glucose Meter (Lifescan). Serum triglycerides were measured with a serum triglyceride determination kit (Sigma-Aldrich). Liver triglycerides were extracted by the method of Bligh and Dyer (20) and measured with the triglyceride determination kit. Total cholesterol and LDL and HDL cholesterol in serum were measured by enzymatic assays using a Hitachi 7600 modular chemistry analyzer with the corresponding reagent (ZhengKang Biotech). Mouse and human serum Metrnl levels were examined with the corresponding ELISA kits (R&D Systems or Antibodies Online). Insulin, leptin, adiponectin, TNFα, and IL6 levels were measured by ultrasensitive mouse insulin ELISA kit (Mercodia), mouse leptin and adiponectin ELISA kits (Millipore), and mouse TNFα and IL6 ELISA kits (R&D Systems), respectively. Lipase activity was measured with a lipase activity fluorometric assay kit (Biovision). Malondialdehyde (MDA) was measured with Lipid Peroxidation MDA Assay Kit (Beyotime Biotechnology), and reactive oxygen species were detected with the CM-H2DCFDA probe (Life Technology).

Lipid overload test was performed as described elsewhere (21).

These experiments were performed as we previously described (9,11). For real-time PCR, RNA was abstracted with RNeasy Plus Universal Tissue Mini Kit (Qiagen). Primers were provided in Supplementary Table 4. For Western blot, antibodies for PPARγ (Santa Cruz Biotechnology), Metrnl (produced by ourselves) (9), AKT, and p-AKT(Ser473) (Cell Signaling Technology) were used at dilution 1:1,000. For flow cytometry, adipose tissue was minced and digested with 0.1% collagenase (Sigma-Aldrich) and filtered through a 100-μm cell strainer. After centrifugation, the cell pellet was resuspended in PBS and stained with fluorescent-labeled antibodies (Miltenyi Biotec). For immunohistochemistry, antibody for Metrnl (Sigma-Aldrich) was used at dilution 1:100.

Culture and differentiation induction of 3T3-L1 (Shanghai Institutes for Biological Sciences), mouse primary embryonic fibroblasts, and C3H10T1/2 (American Type Culture Collection) cells were performed as we previously described (9,22). Culture medium of CHO cells with or without Metrnl overexpression and Metrnl recombinant protein (R&D Systems) were added into the culture medium when adipocytes were induced to differentiation.

To construct Metrnl shRNA lentivirus, three targeting sequences were designed, including 5′-GGGCGCTCATTGTTAACCT-3′, 5′-GCTTCCAGTATGAGCTGATGA-3′, and 5′-CACGCTTTAGTGACTTTCAAA-3′. Only the last target displayed a knockdown efficiency >90% (Fig. 1B). To construct Metrnl-expressing lentivirus, the coding sequence of Metrnl was obtained from the previously constructed plasmid pCI-Metrnl-his₆ (9); the overexpression of Metrnl was verified with Western blot (Fig. 1C). Cells were infected at a multiplicity of infection of 20.

For collecting the incubation medium of adipose tissue, WAT was incubated with 4 mL Opti-DMEM medium (Life Technologies) for 24 h. Metrnl concentration in the medium was normalized to the weight of adipose tissue. For collecting culture medium of primary adipocytes, WAT was digested with 0.1% collagenase (Sigma-Aldrich) for 60 min. Undigested tissue was removed by filtration with 400-μm nylon mesh. Primary adipocytes were washed and cultured with 4 mL Opti-DMEM medium for 24 h. Metrnl concentration in the medium was normalized to the total volume of adipocytes. For collecting culture medium of mature 3T3-L1 adipocytes, 3T3-L1 preadipocytes with or without lentivirus-mediated Metrnl overexpression were induced to differentiation. Culture medium was changed to Opti-DMEM medium and collected 24 h later. The concentration of Metrnl in the medium was measured with an ELISA kit (R&D Systems).

Oil Red O staining was performed as we previously described (22).

All data are expressed as mean ± SEM. Statistical analysis was performed with Prism 5.0 software (GraphPad Software). All data were tested for normality applying the Kolmogorov-Smirnov test. If normality was given and there were no significant differences in variance between groups (F test), a two-tailed Student t test was used to evaluate the differences. Otherwise, nonparametric Mann-Whitney U test was applied. P < 0.05 denoted the presence of a statistically significant difference.

We first explored the effect of Metrnl on adipogenesis using the cell culture model. Differentiation induction in 3T3-L1 preadipocytes overexpressing or silencing Metrnl demonstrated that Metrnl upregulated adipocyte mature-specific markers and promoted lipid accumulation (Fig. 1A–C). Besides, overexpression of Metrnl increased the expression of PPARγ in primary mouse embryonic fibroblasts and C3H10T1/2 cells after induced differentiation to adipocytes (Fig. 1D).

Consistent with our previous data (9), Metrnl was detectable in the incubation medium of WAT (Fig. 1E). Metrnl was also detectable in the culture medium of either primary adipocytes or 3T3-L1 adipocytes. As expected, the medium Metrnl levels were increased in 3T3-L1 adipocytes overexpressing Metrnl. Moreover, PPARγ expression was induced by culture medium from Metrnl-overexpressed CHO cells (Fig. 1F) or specifically by Metrnl recombinant protein in a concentration-dependent manner (Fig. 1G), indicating that the secreted Metrnl can regulate adipocyte differentiation.

To understand the in vivo long-term effects of Metrnl in adipose biology and metabolic homeostasis, we created conditional knockout mice with adipocyte-specific ablation of Metrnl (Metrnl^−/−, i.e., Metrnl^loxP/loxPFabp4-Cre) (Fig. 2A). Metrnl expression was reduced by 57% in adipose tissue of Metrnl^−/− mice, whereas Metrnl serum concentration was comparable between Metrnl^−/− and control mice (Fig. 2B).

Under normal chow diet (NCD), Metrnl^−/− mice displayed no changes in body weight, food intake, and energy expenditure, compared with control mice (Supplementary Fig. 2A and B). Blood glucose and serum insulin levels also demonstrated no significant differences between Metrnl^−/− and control mice during GTT (Supplementary Fig. 2C). Further, the 16-week-old mice were exposed to HFD for 2 months. Body weight, food intake, energy expenditure, lean/fat mass, liver weight, and adipose weight had no significant differences (Fig. 2C and Supplementary Fig. 2D–F), whereas insulin resistance detected by both GTT (Fig. 2D) and ITT (Fig. 2E) became more severe in Metrnl^−/− mice, indicating local adipose Metrnl deficiency deteriorates HFD-induced insulin resistance. Meanwhile, white adipocyte size was obviously decreased in Metrnl^−/− mice (Fig. 2F). Metrnl expression was also decreased in BAT of Metrnl^−/− mice (Fig. 2B), which did not significantly affect BAT weight (Supplementary Fig. 2F), BAT gene expressions relating to lipid metabolism and thermogenesis (Supplementary Fig. 2G), and energy expenditure (Supplementary Fig. 2B and D).

To further investigate the function of Metrnl in white adipose biology and insulin sensitivity, transgenic mice with adipocyte-specific overexpression of Metrnl were generated (Fig. 3A). Metrnl Tg mice were verified with PCR and immunohistochemistry (Fig. 3A and Supplementary Fig. 3A). Metrnl expression was markedly increased in adipose tissue of Metrnl Tg mice, whereas Metrnl serum concentration only demonstrated an upregulation trend (P = 0.057) in Metrnl Tg mice (Fig. 3A).

Under NCD, no significant differences of body weight, food intake, and insulin sensitivity were observed between Metrnl Tg and wild-type mice (Supplementary Fig. 4A–C), whereas mean adipocyte size was slightly increased in Metrnl Tg mice (Supplementary Fig. 4D). Under HFD, body weight, food intake, lean/fat mass, energy expenditure, and distribution of adipose tissue were not altered in Metrnl Tg mice (Fig. 3B and Supplementary Fig. 4E–G). However, insulin resistance induced by HFD was obviously attenuated in all transgenic clones in GTT (Fig. 3C and Supplementary Fig. 4H and I). Metrnl Tg mice also showed more sensitivity to insulin than wild-type mice did in ITT (Fig. 3D). The insulin sensitization of Metrnl was further verified by a significant increase in the glucose infusion rate during the hyperinsulinemic-euglycemic clamp study in Metrnl Tg mice (Fig. 3E). Meanwhile, white adipocyte size was significantly increased in Metrnl Tg mice fed HFD (Fig. 3F). In addition, acute intravenous injection with recombinant Metrnl did not lower blood glucose in HFD-fed mice (Fig. 3G), in contrast to chronic hypoglycemic effect in Metrnl Tg mice.

To verify that Metrnl can prevent not only HFD-induced but also hyperphagia-induced insulin resistance, we further prepared a genetic model with adipocyte-specific overexpression of Metrnl in leptin^−/− obese mice (leptin^−/−Metrnl Tg). Body weight was still unchanged (Fig. 3H), whereas insulin resistance was markedly improved in leptin^−/−Metrnl Tg compared with leptin^−/− mice (Fig. 3I). Similarly, intravenous injection of recombinant Metrnl had no acute hypoglycemic effect in leptin^−/− mice (Fig. 3J).

To further know the relationship between serum Metrnl and BMI in humans, we performed an investigation in 87 human subjects undergoing a physical examination. Metrnl serum concentration was measurable, and no correlation was found between Metrnl serum concentration and BMI (Fig. 3K). Serum Metrnl levels were also not correlated with serum glucose, total cholesterol, or triglyceride levels (Supplementary Table 3).

The insulin signaling pathway was assessed by analysis of insulin-induced phosphorylation of AKT at Ser473 in white fat, brown fat, muscle, and liver. Insulin-stimulated phosphorylation of AKT in WAT was significantly increased in Metrnl Tg mice and decreased in Metrnl^−/− mice. However, no alterations were observed in other examined tissues in either Metrnl^−/− or Metrnl Tg mice (Fig. 4).

Adipokines and inflammation play crucial roles in obesity-associated local and systemic insulin resistance (5,23). Serum adiponectin and leptin were unchanged in Metrnl knockout and transgenic mice (Supplementary Fig. 5). TNFα expression in adipose tissue significantly increased in Metrnl^−/− mice and decreased in Metrnl Tg mice under HFD, whereas IL6 had no alterations (Fig. 5A and B). Expressions of anti-inflammatory factors IL4, IL10, and IL13 in adipose tissue were not significantly changed in Metrnl Tg mice under HFD (Fig. 5C–E). Further, different concentrations of lipopolysaccharide (LPS) were injected intraperitoneally. The elevated TNFα serum concentrations were similar between Metrnl Tg and wild-type mice under LPS stimulations (Fig. 5F). Hence, Metrnl inhibits chronic HFD-induced, but not acute LPS-induced, inflammation.

Considering that lipid overload induces insulin resistance (1,24) and adipose tissue can buffer lipid flux (25), we wondered about the effect of Metrnl on lipid metabolism. For blood lipid profiles, Metrnl attenuated HFD-induced hypertriglyceridemia (Fig. 5G), but not hypercholesterolemia (Supplementary Fig. 6A). Accumulation of liver triglyceride was not significantly changed in Metrnl^−/− mice under HFD (Fig. 5H). During acute lipid overload test, overexpression of Metrnl enhanced serum triglyceride clearance (Fig. 5I). Blood triglyceride is degraded by lipase (26). The activity of lipase was elevated in adipose tissue of Metrnl Tg mice, but not in muscle (Fig. 5J). Lipoprotein lipase was significantly increased in adipose tissue of Metrnl Tg mice (Supplementary Fig. 6B).

Oxidative stress plays a role in the regulation of insulin sensitivity (27). MDA, a primary biomarker of free radical–mediated lipid damage and oxidative stress, in adipose tissue, liver, and muscle was not significantly changed in either Metrnl^−/− or Metrnl Tg mice fed HFD (Supplementary Fig. 7). Consistently, reactive oxygen species were also not altered in 3T3-L1 adipocytes either overexpressing or silencing Metrnl.

Adipocyte differentiation is a key factor in forming functional fat for insulin sensitivity and lipid metabolism (7,28,29). Gene markers relating to adipocyte differentiation and lipid metabolism were detected in both Metrnl knockout and transgenic mice (Fig. 5K and L), showing that Metrnl upregulated key transcription factors for adipocyte differentiation (PPARγ and C/EBPα) and lipid metabolism genes for lipid transport (FABP4 and CD36), lipogenesis (ACC and FASN), lipolysis (Lipe and PNPLA), and lipid storage (Perilipin).

PPARγ is the master regulator of adipocyte differentiation, lipid metabolism, and adipose inflammation in forming functional fat (7,30,31), and PPARγ activation in adipocytes is sufficient for systemic insulin sensitization (32). In addition to mRNA level, PPARγ protein was increased markedly in adipose tissue of Metrnl Tg mice (Fig. 5L). Hence, we speculate that PPARγ plays a critical role in Metrnl-mediated beneficial effects. Consistent with this speculation, PPARγ inhibition by long-term treatment with either of the small-molecule inhibitors, GW9662 and BADGE (17), completely abolished the insulin-sensitizing effect of Metrnl in HFD-fed Metrnl Tg mice (Fig. 6). In addition, knockdown of PPARγ via lentivirus-mediated shRNA interference abolished the Metrnl-induced upregulation of adipose marker genes in 3T3-L1 adipocytes (Fig. 7A–F) and canceled the Metrnl-improved glucose tolerance in GTT (Fig. 7G and H).

It has been reported that Metrnl can induce white adipose browning through an eosinophil-mediated increase of IL4 expression and activation of M2 macrophages in WAT in a short period (15). We thus used inguinal subcutaneous WAT to verify whether this mechanism is involved in the chronic beneficial effects of Metrnl in our models. No significant differences were observed for gene expression of anti-inflammatory factors (IL4, IL10, and IL13), thermogenesis-associated genes (UCP1, PGC1α, Dio2, and ERRα), eosinophil, and M2 macrophage markers (Siglec F, Ccr3, Mrc-1, Clec10a, and Retnla) in both Metrnl^−/− and Metrnl Tg models compared with their corresponding controls, under either NCD or HFD. No significant changes were found for eosinophils and M2 macrophages in Metrnl Tg mice fed either NCD or HFD (Fig. 8A–F and Supplementary Fig. 8). Moreover, thermogenesis-associated genes, anti-inflammatory factors, and activation of M2 macrophages displayed no differences between Metrnl Tg and wild-type mice after a 72-h cold challenge (Supplementary Fig. 9).

We used C57 mice to study pharmacokinetics of Metrnl after a single injection of recombinant Metrnl (1.75 μg/mouse) via a venous catheter in conscious mice (Fig. 8G). Serum Metrnl dramatically increased at 15 min (226 ng/mL) and then rapidly decreased within 4 h (8 ng/mL at 4 h) but was still markedly above basal level at 24 h after injection. These serum Metrnl levels within 24 h were obviously higher than those in Metrnl^−/− mice (Fig. 2B) and Metrnl Tg mice (Fig. 3A). We thus used this dose of Metrnl (1.75 μg/mouse) while performing the rescue experiment in HFD-fed Metrnl^−/− mice. After 1-week treatment, GTT indicated no effect of this Metrnl therapy on insulin resistance of Metrnl^−/− mice (Fig. 8H).

In this study, adipocyte-specific Metrnl deficiency was produced by the Fabp4-cre driver. This driver is also expressed in the central nervous system and activated macrophages (33,34). We performed further experiments to know whether Metrnl expression was altered in neurons and macrophages, thus affecting the function of these cells in Metrnl^−/− mice. Metrnl expression appeared unchanged in cultured primary neurons of Metrnl^−/− mice, and the neurite outgrowth was also unaltered (Supplementary Fig. 10A and B). In cultured primary macrophages of Metrnl^−/− mice, Metrnl expression showed no significant changes. And neither TNFα nor IL6 expression was altered between Metrnl^−/− and control mice after treatment with vehicle or LPS (Supplementary Fig. 10C).

In this study, we demonstrate an insulin sensitization of Metrnl using both adipocyte-specific knockout and overexpression models. The results support that adipocyte Metrnl ameliorates overall insulin resistance through acting on local adipose tissue in an autocrine/paracrine fashion. First, serum Metrnl concentrations remain unchanged in the Metrnl^−/− mice, while the phenotype of deteriorated insulin resistance in Metrnl^−/− mice is obvious and unambiguous. Second, insulin-stimulated phosphorylation of AKT is enhanced by adipocyte Metrnl in WAT, but not in other major metabolic tissues. Third, using intravenous administration of recombinant Metrnl (1.75 μg/mouse) to increase circulating Metrnl level for 1 week is unable to rescue insulin resistance exacerbated by adipocyte deficiency of Metrnl. Although our present data stated above suggest that adipocyte Metrnl controls insulin sensitivity via its autocrine/paracrine action, we do not exclude the endocrine action of Metrnl under certain conditions. To know whether Metrnl may regulate insulin sensitivity in an endocrine fashion, future study needs to test longer periods of treatment with higher doses of Metrnl.

In both knockout and transgenic overexpression mice, WAT is remodeled obviously and expression of genes relating to adipocyte differentiation, lipid metabolism, and inflammation is regulated remarkably by adipocyte Metrnl. These obvious changes in WAT are in contrast with no significant changes in BAT, indicating that the role of Metrnl in white fat is more important than that in brown fat. In addition, energy expenditure is also unaltered in adipocyte-specific Metrnl knockout and transgenic overexpression mice. Thus, the regulation of adipocyte Metrnl on insulin sensitivity is mediated by WAT, rather than BAT.

The normal and proper function of adipose tissue plays a critical role in maintaining insulin sensitivity (35). Improved adipose tissue function and expandability in parallel with adipose remodeling can remedy metabolic complications (36). In this study, Metrnl expands adipocytes significantly, upregulates lipid metabolism–related genes, improves insulin sensitivity, and attenuates hypertriglyceridemia under HFD. Under NCD, although Metrnl Tg mice do not show insulin sensitization and hypolipidemic effects, the increased adipocyte size and enhanced lipid disposal ability are observed. These suggest the improved adipocyte function and expansion occur prior to its measurable beneficial effect on insulin resistance and hypertriglyceridemia.

Metrnl may regulate both lipogenesis and lipolysis in white adipocytes. It promotes lipid accumulation and lipogenesis-related gene expression during adipogenesis in vitro and enlarges adipocytes and upregulates lipogenesis-related genes in vivo, suggesting a prolipogenesis effect of Metrnl. Meanwhile, Metrnl upregulates lipolysis-related genes in vitro and in vivo and enhances lipase activity in WAT, suggesting a prolipolysis of Metrnl. Along with the above-mentioned phenotypes, the adipose content is constant, implying triglyceride turnover is increased but balanced. It is proposed that improving adipose triglyceride turnover may be a potential target for treatment of insulin resistance (37), which could be a possible mechanism involved in the insulin sensitization of Metrnl.

PPARγ is the key regulator of adipocyte differentiation and function. And PPARγ mutants can exacerbate insulin resistance and reduce adipocyte size (38). In our study, adipose PPARγ expression is dramatically upregulated by Metrnl in vitro and in vivo, suggesting it is involved in Metrnl-enhanced adipose function and expansion. Consistently, both inhibition and knockdown of PPARγ can abolish Metrnl-improved insulin resistance, proving that Metrnl antagonizes insulin resistance via the PPARγ pathway. Regarding the links between Metrnl and PPARγ, further studies are needed. In particular, the initial mechanism remains to be clarified; does Metrnl act as a ligand exerting function through a receptor signaling pathway, or does it act as an enzyme exerting function through a biochemical reaction pathway?

We have not observed white adipose browning in the transgenic mice, which appears inconsistent with the previous study (15). The discrepancy could be explained by different experimental models: relatively acute models in the previous study and chronic models in the current study. In the previous study, both administration of Metrnl-expressing adenoviral vectors and high dose of Metrnl protein cause an abrupt increase of Metrnl in adult mice. The plasma Metrnl levels increase by more than fivefold in the previous study, much higher than those in the transgenic mice of the current study. Of note, the previous study shows that Metrnl-driven adipose browning occurs only in a very short period between days 5 and 7 postinjection (15) and disappears 1 week after injection (15). This transient phenotype is compatible with our results of no adipose browning in the permanent transgenic mice.

We use the Fabp4 promoter to create adipocyte-specific knockout and overexpression mice. Because Fabp4 is also expressed in nonadipose tissues, especially in the central nervous system and activated macrophages (33,34), one limitation of the current study is potentially a nonspecific effect from nonadipocytes. Our further experiments exclude the possible effects from neurons and macrophages. Neuronal Metrnl expression and neurite outgrowth are unaltered in Metrnl^−/− mice. Also, Metrnl expression and LPS-induced upregulation of TNFα and IL6 in macrophages are unchanged in Metrnl^−/− mice. Nevertheless, the generation of more adipocyte-specific knockout models (e.g., using Adipoq-Cre) will be helpful to prove the unique role of adipocyte Metrnl in insulin sensitivity.

In conclusion, the current study demonstrates an important role of Metrnl in white adipose biology (Fig. 8I). Adipocyte Metrnl antagonizes obesity-induced insulin resistance by improving adipose function, including adipocyte differentiation, metabolism activation, and inflammation inhibition. The insulin sensitization of adipocyte Metrnl is through the PPARγ pathway. Considering adipocyte Metrnl improves insulin resistance without increasing body weight, it may be a new promising therapeutic target for metabolic syndrome and type 2 diabetes.

Funding. This work was supported by grants from the National Natural Science Foundation of China (81130061, 81202572, 81422049, and 81373414), the National Basic Research Program of China (2009CB521902), and the National Science and Technology Major Project (2009ZX09303-002).

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

Author Contributions. Z.-Y.L. performed the experiments and wrote the manuscript. J.S., S.-L.Z., M.-B.F., Y.-F.G., Y.Q., J.X., and P.W. performed the experiments. All authors contributed to the analysis of experimental data. C.-Y.M. designed the study and wrote the manuscript. C.-Y.M. 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.