Endostatin Prevents Dietary-Induced Obesity by Inhibiting Adipogenesis and Angiogenesis

Obesity and its consequent dysregulation, as important risk factors for type 2 diabetes, nonalcoholic fatty liver disease (NAFLD), cardiovascular diseases, and various types of cancer, represent significant threats to global health (1,2). When food intake exceeds energy expenditure, excess nutrients are stored as fat. This process causes white adipocyte tissue (WAT) expansion and, ultimately, obesity (3). The expansion of WAT is a complex process that involves the enlargement of existing adipocytes and an increased number of adipocytes through adipogenesis (4). A better understanding of adipogenesis and its physiological function in adipose tissue can help us to understand and prevent obesity and its associated consequences in human health.

Adipocyte fate determination and differentiation are mediated by several master transcription factors, including peroxisome proliferator–activated receptor-γ (PPAR-γ) and CCAAT/enhancer–binding proteins (C/EBPs) (4,5). Actually, transient expression of C/EBP-β is one of the earliest events in adipogenesis after the stimulation by adipogenic signals. It then triggers the transcription of C/EBP-α and PPAR-γ (6–8). PPAR-γ is considered to be the master and proximal regulator of adipogenesis, which coordinately activates and maintains the expression of adipocyte-specific genes (8,9).

Recently, Sam68 (Src-associated substrate in mitosis; 68 kDa) was reported to inhibit adipogenesis and obesity development by regulating alternative RNA splicing of mammalian target of rapamycin (mTOR) (10). Sam68^−/− mice exhibit a lean phenotype and are protected against dietary-induced obesity (10). Moreover, adipocytes from Sam68^−/− mice have decreased expression of PPAR-γ and its downstream adipocyte-specific genes, which can inhibit adipogenesis (10). Sam68, as a prototypical member of the signal transduction activator of RNA (STAR) family, functions in RNA metabolism, mRNA recruitment, and alternative splicing (10,11). Sam68^−/− mice retain intron 5 within the mTOR transcript. This induces an error in the RNA splicing of mTOR, which leads to a reduced expression of wild-type mTOR and, consequently, results in defects in adipogenesis (10).

Similar to tumor tissue, WAT development in physiological and pathological contexts is accompanied by the formation of vascular networks (12,13), and several angiogenesis inhibitors have been reported to exert an antiobesity effect by inhibiting angiogenesis in adipose tissue (13,14). As a result of these studies, adipose tissue angiogenesis has a great potential as a therapeutic target for obesity and metabolic diseases (12). Endostatin is a well-documented angiogenesis inhibitor. Despite these intriguing studies, the mechanistic role of endostatin in adipogenesis and dietary-induced obesity, which is the most common form of human obesity (15), remains to be elucidated.

Here we report that endostatin directly inhibits adipogenesis and angiogenesis, thereby protecting mice from dietary-induced obesity and its associated metabolic disorders, including insulin resistance, glucose intolerance, and hepatic steatosis. The antiobesity and antiadipogenic functions of endostatin provide new insights into the biological relevance of this protein, which leads to potential clinical therapeutics.

Escherichia coli–expressed recombinant human endostatin was obtained from Protgen Ltd. (Beijing, China). The antibodies used in this study are described in the Supplementary Experimental Procedures.

C57BL/6J male mice were purchased from Beijing Vital River (Beijing, China). All animal studies were approved by the Tsinghua University Institutional Animal Care and Use Committee. Seven-week-old mice were fed with a high-fat diet (HFD) providing 60% kcal from fat (Research Diets, Inc., New Brunswick, NJ) or a normal diet (ND), providing 10% kcal from fat. The HFD-fed mice were randomly assigned to each therapeutic group and were treated with endostatin (12 mg/kg/day) or saline (0.9% sodium chloride) by daily intraperitoneal injection for 60 days (the first injection was designated as day 0). All drugs were diluted with saline. Mice body weight was measured every 3 days. Weight of WAT, liver, heart, lung, and kidney was measured after treatments were completed. Liver and fat pads were collected and fixed in 4% paraformaldehyde for further analysis.

Insulin tolerance tests (ITTs) and glucose tolerance tests (GTTs) were performed as previously described (10,16) after mice were treated with endostatin (12 mg/kg/day) or saline for 60 days. For the ITT, mice were given a dose of insulin (0.5 units/kg; Novolin R; Novo Nordisk) by intraperitoneal injection. For the GTT, mice were starved overnight and then orally fed with 1 mg glucose (20 mg/mL, solution water) per gram of body weight. Blood glucose was measured with a OneTouch Basic glucose meter (Roche).

WAT and liver tissue were fixed in 4% paraformaldehyde, embedded in paraffin, and then sliced and stained with hematoxylin and eosin according to standard procedures. The detailed description of histology and the matrigel plug assay is provided in the Supplementary Experimental Procedures.

For differentiation, 3T3-L1 preadipocytes (3T3-L1s) were treated as described previously with slight modification (17). As shown in Fig. 2A, 2-day postconfluent 3T3-L1s were incubated with 0.5 mmol/L 3-isobuty-1-methylxanthine, 1 μmol/L dexamethasone, and 10 μg/mL insulin (MDI) in DMEM with 10% FBS (MDI medium) (designated as day 0). After 2 days, the medium was changed to insulin medium for additional 2 days. Insulin medium is 10 μg/mL insulin in DMEM with 10% FBS, and FBS medium is DMEM with 10% FBS. Adipogenesis was detected by Oil Red O staining.

Endostatin or BSA was coupled with CNBr-activated Sepharose 4B (GE Healthcare), according to the manufacturer’s instructions, and incubated with 3T3-L1s lysates overnight at 4°C. The samples were subjected to SDS-PAGE or immunoblotting. Gel slices containing protein bands were analyzed by liquid chromatography–mass spectrometry, as previously described (18).

3T3-L1s at day 2 of differentiation were incubated with endostatin (0, 25, 50 μg/mL) for 3 h at 37°C. Afterward, RNA-binding assay was performed, as previously described (10). The RNA sequences of two 5′ biotin-tagged putative Sam68-binding sites (SBS1 and SBS2) (10) were synthesized by GenePharma (Shanghai, China).

Tube formation assay, transwell migration assay, and scratch wound–healing assay were performed on SVEC4-10 cells as previously described (18). In each assay, SVEC4-10 cells were treated with DMEM, adipocyte-conditioned media (adipocyte-CM), or a combination of endostatin and adipocyte-CM.

Immunoprecipitation was performed as previously described (18). The detailed description of immunoprecipitation is provided in the Supplementary Experimental Procedures.

pcDNA3.0-Sam68 and pcDNA3.0-Sam68-KH-delete as templates were obtained from Addgene (Addgene plasmid 17690 and 17688). Hemagglutinin (HA)-tagged Sam68 truncations of functional domains (Fig. 3F) were constructed by the QuikChange Site-Directed and transfected with TurboFect in vitro transfection reagent (Fermentas), according to the manufacturer’s instructions.

RNA was extracted and cDNA generated as described (19). Primers designed for quantitative PCR are listed in Table 1.

3T3-L1s (day 2 of differentiation) cultured on coverslips were incubated with 25 μg/mL Rh-conjugated endostatin (Rh-endostatin) (20) for 3 h at 37°C. Afterward, cells were immunofluorescence stained as previously described (20).

The quantitative data are shown as mean ± SD, unless noted otherwise. The two-sided Student t test was used for comparisons between two groups. One-way ANOVA, followed by post hoc analysis using the Dunnett test, was performed to test the significant difference between the control group and treatment groups. All P values <0.05 were considered statistically significant.

To determine whether endostatin affects the development of dietary-induced obesity, we fed 7-week-old C57BL/6J male mice with the HFD or ND. Animals in the experimental groups were systemically treated with endostatin by daily intraperitoneal injection for 60 days. We found that the HFD-fed mice treated with endostatin were resistant to dietary-induced body weight gain, which was 38% lower than that of the untreated HFD-fed mice (Fig. 1A and Table 2), whereas no significant body weight change was observed in ND-fed group (Supplementary Fig. 1A). Necropsy showed that endostatin-treated HFD-fed mice contained deposits of epididymal fat that were reduced in mass volume compared with untreated HFD-fed animals (Fig. 1B). Similarly, the weight of total WAT was significantly decreased in endostatin-treated HFD-fed mice compared with the untreated HFD-fed mice (Fig. 1C and Table 2). Interestingly, the weight of other organs, including heart, lungs, and kidneys was unaffected in endostatin-treated HFD-fed mice (Table 2). We also performed a detailed histological analysis of WAT, which revealed that the sizes of adipocytes in endostatin-treated HFD-fed mice were smaller than those in untreated HFD-fed animals (Fig. 1D). Thus, endostatin is effective in the inhibition of obesity development in HFD-fed mice.

Adipocyte dysfunction is tightly linked to metabolic disorders such as insulin resistance, glucose intolerance, and hepatic steatosis (4,8). The glucose levels of the HFD-fed mice with or without endostatin treatment were determined. Decreased glucose levels were observed in the endostatin-treated group compared with HFD-fed mice (Supplementary Fig. 1B). We further tested whether endostatin could affect HFD-induced insulin resistance and glucose intolerance in HFD-fed mice. The results showed that the glucose levels were decreased in endostatin-treated HFD-fed mice, whereas untreated HFD-fed mice developed insulin resistance and glucose intolerance (Fig. 1E and F).

The phosphorylation of Akt serves as a powerful indicator of insulin sensitivity. Blocking the activity of Akt can cause insulin resistance (21). Consistently, we observed a lower phosphorylation level of Akt at Thr308 in HFD-fed mice, whereas compensating effects were observed by endostatin treatments (Fig. 1J and Supplementary Fig. 10A). Furthermore, we observed that the weight of liver was significantly decreased in endostatin-treated HFD-fed mice compared with untreated HFD-fed mice (Fig. 1G and Table 2). The histological analyses of liver showed that endostatin alleviated hepatic steatosis, which was typical in HFD-fed mice (Fig. 1H and I), suggesting that it might suppress the progression of NAFLD. Taken together, all of these results support that endostatin mitigates dietary obesity–induced metabolic disorders.

To determine whether endostatin affects adipogenesis in vitro, we investigated its effect on the differentiation of 3T3-L1s. As shown in Fig. 2B, lipid droplet (LD) formation occurred at day 3 of adipocyte differentiation, the point when a difference between the groups with and without endostatin treatment could first be detected. At days 5 and 6, a clear difference between the endostatin-treated and -untreated group was visible: most endostatin-treated 3T3-L1s contained fewer and smaller LDs, whereas cells in the control group contained more and larger LDs (Fig. 2B). By day 8, fewer 3T3-L1s in the endostatin-treated groups were differentiated into adipocytes compared with the control group (Fig. 2B). Furthermore, Oil Red O staining of differentiated 3T3-L1s on day 8 confirmed that endostatin inhibited adipogenesis in a dose-dependent manner (Fig. 2C).

C/EBPs and PPAR-γ are master transcription factors of adipogenesis (8). Their expressions are always increased during adipogenesis. We hypothesized that endostatin might have an inhibitory effect on these factors. Consistently, endostatin significantly reduced the RNA and protein levels of PPAR-γ and C/EBP-α and -β (Fig. 2D and F and Supplementary Fig. 10B). We then evaluated the effect of endostatin on adipocyte fatty acid–binding protein (aP2), cluster of differentiation 36 (CD36), and glucose transporter 4 (Glut4), which are the downstream genes of PPAR-γ and C/EBP-α and considered as adipogenesis markers. The results showed that endostatin significantly reduced the expression of aP2, CD36, and Glut4 during adipogenesis (Fig. 2F). These effects were consistent with changes in cell morphology induced by endostatin during adipogenesis.

To further determine whether endostatin inhibits adipogenesis in vivo, we detected the expression levels of PPAR-γ, C/EBP-α, C/EBP-β, aP2, CD36, and Glut4 in HFD-fed mice. Compared with the ND, the HFD increased the expression levels of PPAR-γ, C/EBP-α, C/EBP-β, aP2, and CD36, whereas endostatin attenuated these effects (Fig. 2E and Supplementary Figs. 2A and 10C). However, the HFD decreased the expression level of Glut4 in vivo, and endostatin treatment significantly reversed this effect (Supplementary Fig. 2A).

To investigate the mechanism that endostatin inhibits adipogenesis, we set out to identify potential endostatin-interacting proteins in 3T3-L1s. 3T3-L1 cell lysates were applied to endostatin-linked CNBr-activated Sepharose 4B column (CNBr-endostatin). After elution and separating bound proteins by SDS-PAGE, mass spectrometry analysis revealed the presence of Sam68: seven distinct peptides comprising the polypeptide sequences of Sam68 were identified (Fig. 3A and B). Furthermore, Western blot confirmed the presence of Sam68 in the eluted solution of CNBr-endostatin (Fig. 3C). Next, the result of reciprocal immunoprecipitation confirmed that endostatin physically interacted with Sam68 (Fig. 3D). To determine the subcellular compartment in which endostatin binds to Sam68, we performed internalization and immunofluorescence assays and found that endostatin was internalized by 3T3-L1s (Supplementary Fig. 3) and then colocalized with Sam68 in the nuclei (Fig. 3E). Thus, endostatin physically interacts with Sam68 in the nuclei of 3T3-L1s.

To determine which structural domain of Sam68 (11) interacts with endostatin, we constructed a series of HA-tagged Sam68 truncations in which different functional domains were deleted (Fig. 3F) and then detected their abilities to interact with endostatin. This analysis revealed that the binding of Sam68Δ96-157 or Sam68Δ96-279 to endostatin was dramatically disrupted compared with Sam68WT, whereas the interactions between other truncated protein constructs and endostatin were unaffected (Fig. 3G). These findings suggest that endostatin interacts with the NK (N-terminal of KH) domain of Sam68, which resides in the amino acid region 96-157. This domain regulates the specificity of interactions between Sam68 and RNA elements (11), which implies that the binding of endostatin to the NK domain of Sam68 affects Sam68-mediated RNA metabolism.

Sam68 deficiency leads to an error in alternative splicing of mTOR, producing a mutated unstable form of mTOR mRNA, termed “mTORi5,” which is a mutated transcript terminated within intron 5 (10). Thus, we detected whether endostatin directly interfered with the interactions between Sam68 and the two putative Sam68-binding sites (SBS1 and SBS2) within intron 5 of mTOR (10) using RNA-binding assay. Our result showed that endostatin attenuated the binding of Sam68 to SBS1 and SBS2 in a dose-dependent manner (Fig. 4A). These findings suggested that endostatin impaired the binding of Sam68 to intron 5 of mTOR, which predicted that endostatin led to intron 5 retained within the mTOR transcript and then reduced the expression of wild-type mTOR.

To determine whether endostatin indeed affects the mTOR transcript during adipogenesis, we measured the levels of mTOR, mTORi5, and Sam68 at days 0 and 4 of differentiation in the endostatin-treated or -untreated group. Consistent with previous observations (10), the expression of mTOR, mTORi5, and Sam68 was increased at day 4 compared with day 0 of undifferentiated 3T3-L1s (Fig. 4B and D and Supplementary Fig. 10E). Moreover, in the endostatin-treated group, wild-type mTOR expression was decreased whereas mTORi5 expression was increased compared with those in the control groups at day 0 and day 4 (Fig. 4B–D and Supplementary Fig. 10D and E). However, the expression of Sam68 was unaffected by the endostatin treatment (Fig. 4D and Supplementary Fig. 10E). We further detected the expression level of mTOR in vivo and found that HFD-induced upregulation of mTOR was inhibited by endostatin treatment in WAT (Fig. 5C and Supplementary Fig. 10H).

We next investigated whether the overexpression of Sam68 could rescue the inhibitory effect of endostatin on mTOR transcription. If so, which domains of Sam68 are responsible for it? Interestingly, endostatin-induced upregulation of mTORi5 and downregulation of wild-type mTOR were abrogated by the overexpression of Sam68WT and Sam68Δ1-96 but not by the overexpression of Sam68Δ96-157 and Sam68Δ96-279, in which the endostatin-interacting domains were deleted (Figs. 4E and 5D and Supplementary Fig. 10I). Thus, it is through the competitive interaction with Sam68 that endostatin inhibits Sam68-mediated alternative splicing of mTOR, which results in retaining intron 5 within the mTOR transcript, leading to the downregulation of mTOR.

The inhibitory effect of endostatin on mTOR expression would be predicted to influence the activity of mTOR complex 1 (mTORC1). To determine whether endostatin influences mTORC1, we tested its effects on phosphorylation of eIF4E-binding protein 1 (4E-BP1), p70 S6 ribosomal kinase (S6K), and the ribosomal protein S6 (RPS6), which indicate the activities of downstream kinases of mTORC1 (22,23). The results showed that the phosphorylation levels of 4E-BP1 and S6K at Thr389 were decreased in a dose-dependent manner in vitro (Fig. 5A and Supplementary Figs. 4A and 10F). The decreased phosphorylation level was also observed at Ser240/244 of RPS6 (Fig. 5A and Supplementary Fig. 10F), which acted as the downstream effector of S6K activation in adipocytes or preadipocytes (22,23). In addition, the phosphorylation levels of 4E-BP1, S6K (Thr389), and RPS6 (Ser240/244) in WAT were lower in endostatin-treated mice than those in HFD-fed mice (Fig. 5C and Supplementary Figs. 4B and 10H). Raptor is a representative component of mTORC1. Wang et al. (24) reported that the phosphorylation of raptor at Ser863 is important for the activation of mTORC1 toward substrates of 4E-BP1 and S6K. Consistently, the protein level and phosphorylation level of raptor at Ser863 were also downregulated in the endostatin-treated groups in vitro and in vivo (Supplementary Fig. 5A and B).

We then tested whether Sam68 overexpression could rescue the inhibitory effects of endostatin on the activities of S6K and RPS6. As shown in Fig. 5D and Supplementary Fig. 10I, overexpression of Sam68WT and Sam68Δ1-96, but not Sam68Δ96-157 and Sam68Δ96-279, which lack the endostatin-interacting domain, could rescue the inhibitory effects of endostatin on the activities of S6K and RPS6. Furthermore, the inhibitory effect of endostatin on adipocyte differentiation also could be abrogated by overexpression of Sam68WT and Sam68Δ1-96 but not Sam68Δ96-157 and Sam68Δ96-279 (Fig. 5E and F). Thus, through directly interacting with Sam68, endostatin impaired the activity of mTORC1, consequently leading to the inhibition of adipogenesis.

Rictor is a representative component of mTORC2. S6K has been reported to directly phosphorylate rictor at Thr1135. We found that endostatin inhibited Thr1135 phosphorylation of rictor in vitro and in vivo (Supplementary Fig. 5A and B), which is consistent with our observation that endostatin suppresses the activity of S6K. However, the relationship between the phosphorylation of rictor at Thr1135 and the activity of mTORC2 was not well defined (25,26). Therefore, we further examined the activity of mTORC2 by assessing Ser473 phosphorylation of Akt (27) at days 0 and 2 of differentiation. The results showed that only a high dose of endostatin (50 μg/mL) treatment, but not a low dose (25 μg/mL), significantly suppressed the phosphorylation of Akt at Ser473 (Fig. 5B and Supplementary Fig. 10G). Meanwhile, there was no significant change in the phosphorylation level of Akt at Ser473 in the HFD-fed mice treated with endostatin (12 mg/kg/day) (Fig. 5C and Supplementary Fig. 10H). Lamming et al. (28) reported that inhibition of mTORC2 might not be good over the long-term because it leads to insulin resistance. Thus, endostatin did not affect the mTORC2 pathway in vivo, which is consistent with our observation that it improved insulin resistance in HFD-fed mice. Actually, the situation in vivo is more complicated than that of in vitro. A possible explanation of the contradictory observations of mTORC2 activity in vitro and in vivo is caused by the effect of endostatin on insulin resistance, which could be a secondary effect to its antiobesity function.

We also tested mTOR-regulated cell behaviors, including cell apoptosis and proliferation in vitro. Endostatin significantly promoted the apoptosis of 3T3-L1s during adipogenesis (Supplementary Fig. 6A and B). However, the proliferation of 3T3-L1s remained unchanged after endostatin treatment (Supplementary Fig. 6C and D). These results are consistent with our previous observations (29,30). Thus, the apoptosis-promoting effect of endostatin may contribute to its antiadipogenic function.

To study whether endostatin exerts an antiobesity effect via its antiangiogenic activities, we examined the degree of vascularization in adipose tissues of endostatin-treated or -untreated mice. Our result showed that adipose tissue was highly vascularized in HFD-fed mice compared with ND-fed mice, as shown by immunostaining for the vascular endothelial marker CD31 (Fig. 6A and B). Compared with untreated HFD-fed mice, a striking reduction in vascular density in the WAT of endostatin-treated HFD-fed mice was observed (Fig. 6A and B). We further observed that endostatin inhibited HFD-induced upregulation of angiogenic factors, including vascular endothelial growth factor (VEGF), fibroblast growth factor (FGF)-2, and placental growth factor (PlGF) in WAT (Supplementary Fig. 7A and B). Consistently, VEGF, FGF-2, and PlGF were also highly expressed in mature adipocytes compared with undifferentiated preadipocytes, which could be compensated by endostatin treatment during adipogenesis in vitro (Supplementary Fig. 7C).

Many angiogenic factors could be secreted by adipocytes in vitro and in vivo (31). We consistently found that proangiogenic factors, including VEGF, FGF-2, and PlGF, were also secreted by adipocytes in vitro (Supplementary Fig. 7D), which prompted us to explore whether endostatin could counteract the angiogenic activity induced by such secretions. We treated SVEC4-10 murine endothelial cells with adipocyte-CM, and tested whether endostatin inhibited this treatment-induced endothelial cell vascular tube formation and migration in vitro. Our results showed that adipocyte-CM led to increased tube formation and transwell migration of SVEC4-10 murine endothelial cells and that these effects were reversed by endostatin treatment in a dose-dependent manner (Fig. 6C, D, and F). Similarly, in a scratch wound–healing assay, we found that wound healing was significantly delayed in the endostatin-treated groups (Fig. 6E and F). In addition, we used the in vivo matrigel plug assay to confirm these in vitro results. Consistent with our in vitro results, endostatin markedly reduced blood vessel formation in the matrigel plugs supplemented with adipocyte-CM in a dose-dependent manner (Fig. 6G and H). Thus, the well-known antiangiogenic activity of endostatin is an additional mechanism for its antiobesity function.

Endothelial cell angiogenesis involves activation of multiple signaling pathways, such as FAK, p38 mitogen-activated protein kinase (MAPK), and extracellular signal–related kinase (ERK) (20), prompting us to determine whether endostatin inhibits any of these pathways activated by secretory products of adipocytes. As shown in Fig. 6I and Supplementary Fig. 10J, adipocyte-CM stimulated the phosphorylation of FAK, p38 MAPK, and ERK1/2 in endothelial cells, which was attenuated by endostatin in a dose-dependent manner.

Here, the dietary-induced obesity model, in which the HFD can dramatically stimulate adipogenesis, angiogenesis, and then the expansion of WAT compared with the ND, was used to study the antiobesity function of endostatin. This study reports that endostatin protects mice against dietary-induced obesity and its related metabolic disorders through both antiadipogenic and antiangiogenic mechanisms. We find that endostatin inhibits adipogenesis by the mechanism of impairing Sam68-mediated alternative splicing of mTOR and reducing the activity of the mTORC1 pathway. In addition, we demonstrate that the well-known antiangiogenic activity of endostatin also contributes to its antiobesity effect, which is consistent with previous observations that several angiogenesis inhibitors, including angiostatin and TNP-470, prevent obesity in mice (13,14).

We reveal that the antiadipogenic function of endostatin is mediated by competitive interaction with Sam68 in the nuclei of 3T3-L1s and then inhibits Sam68-mediated RNA splicing of mTOR. However, the mechanism by which endostatin is internalized and then translocated to the nuclei of 3T3-L1s is still unclear. Our previous study discovered that cell surface nucleolin is the functional receptor of endostatin (29). Nucleolin, which is mainly located in the nuclei, can translocate to the cell surface when endothelial cells are stimulated by extracellular matrix and VEGF (32). When endostatin binds to cell surface nucleolin, both can be internalized and then translocated to the nuclei of endothelial cells (20,29,33). Here we found that adipogenesis stimulated nucleolin expression on the cell surface of differentiating 3T3-L1s at days 2, 4, and 6 of differentiation (Supplementary Fig. 8). Moreover, the expression of cell surface nucleolin dramatically decreased in nondifferentiating preadipocytes and differentiated adipocytes (Supplementary Fig. 8). Together, we infer that cell surface nucleolin mediates the internalization and translocation of endostatin in 3T3-L1s during adipogenesis so that endostatin can specifically target adipogenesis in WAT.

Our study shows that the antiadipogenic function of endostatin is mechanistically linked to the mTORC1 pathway. Upon physically binding to Sam68 in the nuclei of preadipocytes, endostatin competitively inhibits the interaction between Sam68 and intron 5 of mTOR, thereby it can decrease the expression of wild-type mTOR and finally inhibit the activity of the mTORC1/S6K pathway. Consistent with our findings, Huot et al. (10) reported that Sam68 is a critical mediator of mTOR alternative splicing in WAT. Sam68^−/− mice show inhibition of mTORC1 pathway and are protected from obesity (10). In addition, mTOR inhibition decreases cell size of adipocytes and prevents obesity in humans (23,34). The inhibition of mTORC1/S6K consistently downregulates the expression of C/EBP-α and PPAR-γ and then blocks adipogenesis (35). Adipose tissue–specific mTORC1 knockout mice were lean and resistant to dietary-induced obesity (36). Thus, our results showed that the activity of mTORC1/S6K in the HFD-fed mice was suppressed by endostatin, which could explain why endostatin treatment led to the reduction of body weight.

Like tumor tissues, WAT contains a diversity of cell types, including adipocytes and other adipocyte stromal cells, such as preadipocytes, endothelial cells, and inflammatory cells (12), comprising a complex adipose microenvironment. Interestingly, there is evidence for a functional link between adipogenesis and angiogenesis during deposition of fat mass, whereby the angiogenic capacity of WAT may determine the extent of adipogenesis and propensity of a subject to gain weight (37,38). In addition, several studies have shown that endothelial cells control adipogenesis in WAT through VEGF and matrix metalloproteinases pathways (39–42). Accordingly, we revealed that factors derived from endothelial cells stimulated adipogenesis of 3T3-L1s. Furthermore, the conditioned media from endothelial cells precultured in the presence of VEGF accelerated adipocyte differentiation (Supplementary Fig. 9A and B). In addition, these adipogenic-promoting effects on adipogenesis were reversed by the pretreatment of endostatin on endothelial cells (Supplementary Fig. 9A and B). These data indicate that vascular endothelial cells promote adipogenesis in a paracrine manner and that endostatin can attenuate this effect. In this study, we treated the mice with endostatin systemically, and therefore cannot exclude the possibility that its action on endothelial cells also contributed to the antiadipogenic effect of endostatin.

In the current study, endostatin was observed to inhibit angiogenesis by suppressing FAK, p38 MAPK, and ERK1/2, which is consistent with a previous study (20). In addition, the angiogenesis factors including VEGF, FGF-2, and PlGF were also downregulated in the endostatin-treated group. FGF-2 was reported to activate S6K (43,44). Moreover, Pende et al. (43) reported that the S6K pathway mediates CREB phosphorylation in response to protein kinase C activation and FGF-2 stimulation. In addition, rapamycin inhibits the expression of adipogenic control transcription factors, including C/EBP-α and PPAR-γ, by blocking the mTOR/S6K pathway (35). These together could be a possible explanation why the antiangiogenic effect of endostatin led to the inhibitory effect on adipogenesis through mTOR/S6K.

Insulin resistance and glucose intolerance are important characteristics of metabolic disorders. Here we found that systemically administered endostatin improved insulin resistance and glucose intolerance in mice, suggesting that it may provide a potential therapeutic target for the prevention of type 2 diabetes. Um et al. (45) reported that mTORC1 promotes insulin resistance in adipose tissue through the S6K-mediated inhibition of insulin signaling, which is consistent with our finding that endostatin improves insulin resistance and impairs the activity of S6K.

Our results showed that endostatin inhibited the expression of Glut4 during adipogenesis in vitro. However, Glut4 was downregulated in the WAT of HFD-fed mice, whereas this effect was significantly reversed by endostatin treatment. Studies have consistently reported that Glut4 is downregulated in WAT and skeletal muscle of obese mice that were fed with an HFD over long-term and that the mice developed insulin resistance and glucose intolerance (46,47). A possible explanation of the contradictory observations of Glut4 expression changes in vitro and in vivo is due to the effects of endostatin on insulin resistance and glucose intolerance, which could be a secondary effect to its antiobesity function.

Although endostatin has been tested as an antitumor drug in human clinical trials, this is not a good model to determine the effect of endostatin on body weight because of the combined use of endostatin and chemotherapy drugs. Clinical trials of the Pichia pastoris–expressed endostatin were terminated at phase II studies. The China Food and Drug Administration (CFDA) subsequently approved E. coli–expressed N-terminal–modified endostatin. The likely reason for this difference was that the P. pastoris–expressed endostatin suffered from N-terminal truncations that influenced its correct folding and further decreased its antiangiogenic capacity. Our previous work reported that correct refolding and N-terminal integrity were critical for the activities of this molecule (48); therefore, we used E. coli–expressed recombinant endostatin in the current study.

Many therapeutic approaches, including restricting food intake, increasing physical exercise, medication, and surgical intervention, have been developed to reduce obesity. However, these approaches are limited by their poor efficacy for treating some types of obesity, poor long-term adherence rates, and serious adverse effects (12). Thus, new approaches to obesity prevention and treatment are urgently needed. We demonstrated in this report that endostatin effectively protects mice against dietary-induced obesity and related metabolic disorders. Even so, the correlation between obesity and the endostatin concentration in blood circulation was not well defined (49,50). The basal levels of endostatin in blood circulation of obese individuals need to be investigated in the future. However, endostatin still has a great potential to be used in antiobesity therapy and in the prevention of obesity-related metabolic disorders.

Acknowledgments. The authors thank Lin Li (the Luo Laboratory member) for her helpful suggestions.

Funding. This work was partly supported by the General Programs of the National Natural Science Foundation of China (nos. 81171998 and 81272529) and the National Science and Technology Major Project for “Major New Drugs Innovation and Development” (2013ZX09509103).

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

Author Contributions. H.W. designed and performed experiments, analyzed results, and wrote the manuscript. Y.C., X.-a.L., G.L., and Y.F. analyzed results and reviewed the manuscript. Y.L. designed experiments, analyzed results, and wrote the manuscript. Y.L. 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.