Osteoprotegerin Promotes Liver Steatosis by Targeting the ERK–PPAR-γ–CD36 Pathway

Nonalcoholic fatty liver disease (NAFLD) has been considered an important public health problem worldwide and represents a spectrum of diseases, including steatosis, nonalcoholic steatohepatitis, and cirrhosis (1–4). NAFLD is also closely associated with insulin resistance (IR), obesity, type 2 diabetes mellitus (T2DM), and dyslipidemia (5,6) and thereby tends to be related to excess mortality and health expenditures. In recent decades, obesity has been thought to be an important cause of NAFLD. The prevalence of hepatosteatosis and steatohepatitis is higher in obese subjects than in lean subjects (7,8). Although obesity is closely related to NAFLD in human and animal studies, the molecular mechanism of hepatosteatosis in obese individuals remains poorly understood.

The liver, as the main metabolic organ of the human body, exerts a key role in the regulation of metabolism and energy equilibrium (9). The first step in the development of NAFLD is the accumulation of hepatic lipid, mainly triglycerides (TGs), which results in an imbalance between fat synthesis and lipolysis (10). Therefore, a crucial mechanism contributing to the development of NAFLD involves the rates of fat synthesis and lipolysis. Recent studies have revealed that some transcription factors and cytokines play crucial roles in the regulation of lipogenesis and lipolysis in the liver (11–13). It is thus desirable to elucidate the roles of these genes in hepatic lipogenesis and lipolysis.

Accumulating evidence suggests that NAFLD may be related to other common chronic wasting diseases, especially bone loss and osteoporosis (14–16). It has also been reported that the cross talk between cytokines and hormones, such as leptin and osteocalcin, in bone, liver, and adipose tissue contributes to the regulation of energy metabolism and glucose homeostasis. Alterations of these cytokines and hormones may lead to obesity and related diseases such as NAFLD and T2DM (17–20).

Osteoprotegerin (OPG) is a soluble glycoprotein that is a member of the tumor necrosis factor (TNF) receptor superfamily and is produced in many tissues, including liver, kidney, bone, and lung (21). This protein was initially found to be a key regulator in bone metabolism. Later, several studies demonstrated that circulating OPG levels in individuals with obesity, T2DM, polycystic ovary syndrome, and obesity-related diseases were significantly decreased or increased (22–25). In addition, circulating OPG levels were significantly increased or decreased in NAFLD patients relative to healthy individuals (26,27), whereas OPG levels decreased in a stepwise fashion from healthy subjects to NAFLD patients without and with nonalcoholic steatohepatitis (28). The levels of circulating OPG were also recently reported to be related to HDL cholesterol and HOMA-IR (29). All these reports, however, are from cross-sectional population-based studies. The role of OPG in NAFLD and other metabolic diseases is still unclear. Therefore, it is important to investigate the role of OPG in obesity-related NAFLD to further identify the targets for therapeutic intervention of NAFLD.

In the current study, we established an OPG knockout (OPG^−/−) mouse model and fed it a high-fat diet (HFD) to evaluate the effects of OPG in hepatic TG accumulation and associated metabolic disorders. We subsequently performed functional analyses and provided evidence that OPG plays a key role in the regulation of hepatic lipid metabolism.

C57BL/6J (wild-type [WT]), CD36 knockout (CD36^−/−), adiponectin knockout (Adipoq^−/−), OPG^−/−, ob/ob, and db/db mice aged 8 weeks were purchased from the Animal Centers of Chongqing Medical University or Shanghai Biomodel and Ganismsci & Tech Develop Co., Ltd. (Shanghai, China). The 8-week-old mice were adapted to the environment for 1 week before the experiments. OPG^−/− mice were generated and identified by PCR screening of their tail DNA (Supplementary Table 1). Male WT, OPG^−/−, Adipoq^−/−, ob/ob, and db/db mice (9 weeks old) were fed either a standard chow diet (SD) (protein, 20%; fat, 10%; carbohydrates, 70%) (#MD 12031; Medicine Inc, Jiangsu, China) or an HFD (protein, 20%; fat, 45%; carbohydrates, 35%) (#MD 12032; Medicine Inc) for 12 weeks. For metabolic experiments, 8-week-old WT and OPG^−/− mice were divided into four groups (n = 5–7/each group) and fed the SD or HFD for 12 weeks. Animals were weighed weekly, and food intake was measured daily for a week. Mice were sacrificed after 12 weeks of dietary feeding. Blood and tissue samples were collected for further analysis. All animal protocols were approved by the animal care and use review committee of Chongqing Medical University (Chongqing, China).

Hepatic tissues from humans were obtained via liver biopsy from eight patients with NAFLD. Eight liver transplantation donors were obtained from the Department of Surgery, the First Affiliated Hospital, Chongqing Medical University, or Xinqiao Hospital, Third Military Medical University. The protocols were reviewed and supported by the Ethics Committee of Xinqiao Hospital, Third Military Medical University, Chongqing China.

Adenoviral vectors (Adv) expressing OPG (Adv-OPG) and shRNA against OPG (Adv- shOPG) (sequence 5′-TCCAAAGTACCTTCATTAT-3′) were constructed using the AdEasy Adenoviral Vector System (Qbiogene) or the pAdxsi system (SinoGenoMax Co. Ltd., Beijing, China), as previously described (30). Adenovirus encoding green fluorescence protein (Adv-GFP) was used as a control for Adv-OPG, and an Adv expressing the control sequence (Adv-control, 5′-TTCTCCGAA CGTGTCACGT-3′) was used as another control for Adv-shOPG. Large-scale amplification and purification of adenoviruses were performed with the ViraBind Adenovirus Purification Kit (Cell Biolabs, San Diego, CA), and adenoviruses were stored at −80°C. For the study of OPG expression, 8-week-old male OPG^−/− mice were fed the HFD for 10 weeks and injected with Adv-OPG or Ad-GFP via the tail vein 12 days prior to the in vivo study.

Mouse primary hepatocytes (MPHs) were isolated from WT, OPG^−/−, and CD36^−/− mice aged 8 weeks, as previously reported (31). MPHs were cultured in DME/F12. L02, Hepa1-6, and HepaG2 cells were cultured in DMEM with 10% FBS. When cells reached 70% confluence, they were treated with free fatty acid (FFA) mixture (1 mmol/L) at a 2:1 ratio of oleate/palmitate or BSA for 24 h after infection with Adv-OPG, Adv-shOPG, Adv-GFP, or Adv-control for 48 h. For some studies, cells were treated with recombinant OPG protein (10 ng/mL) (OPG-Fc) (Sigma-Aldrich, St. Louis, MO) or PBS after treatment with the FFA mixture for 24 h. For the extracellular signal–regulated kinase (ERK)–peroxisome proliferator–activated receptor-γ (PPAR-γ) pathway studies, L02 cells were treated with or without FFAs for 24 h, and/or GW9662 (a PPAR-γ inhibitor), and/or SCH772984 (an ERK inhibitor) for 8 h, and/or OPG-Fc for 6 h.

CD36 reporter genes (−2,584/+165 base pair [bp], −1,138/+165 bp, −483/+165 bp, −292/+165 bp, and −99/+165 bp) were prepared by PCR to generate various segments of the sequences between −2,584 and +165 bp in the CD36 gene promoter (Supplementary Table 2). Supplementary Table 2 also lists primer sequences for site-directed mutations. Then, all fragments were cloned into the pGL3-luciferase plasmid (TIANGEN Biotech, Beijing, China) and sequenced to confirm orientation and sequence. Transient transfections were performed on L02 cells. Luciferase activities were assayed after 24 h of transfection (32).

Paraffin-embedded livers from humans or mice were processed, and paraffin sections (5 µm) were stained with hematoxylin and eosin (H&E) or immunohistochemistry (IHC) by a rabbit polyclonal anti-human OPG antibody (Santa Cruz Biotechnology, Dallas, TX). Frozen liver sections, L02, and Hepa1-6 cells, and MPHs on a six-well plate were stained with 0.15% Oil Red O by standard procedures.

Lipids were extracted with chloroform-methanol (2:1). Total TGs, total cholesterol, HDL cholesterol, LDL cholesterol, FFAs, alanine transaminase, and AST in serum were measured using a commercial kit according to the manufacturer’s protocol. TG contents were expressed as micrograms of lipids per milligram of cellular protein or per gram of tissue weight. Serum OPG levels were measured using an ELISA kit according to the manufacturer’s protocol (R&D Systems, Minneapolis, MN).

Real-time quantitative PCR (RT-qPCR) was performed as described previously (33). Supplementary Table 3 lists primer pairs. Protein analysis was performed with Western blots, as described previously (33). Primary antibodies included anti-OPG and anti–PPAR-γ (Santa Cruz Biotechnology)/phosphorylated (p)-PPARγ (Bioss, Beijing, China), anti-CD36 (Abcam, Cambridge, MA), anti-Akt/p-Akt, anti-protein kinase A (PKA)/p-PKA, anti-ERK/p-ERK, anti-P38/p-P38, anti–c-Jun N-terminal kinase (JNK)/p-JNK, anti-signal transducers and activators of transcription (STAT)5/p-STAT5 (all from Cell Signaling Technology, Danvers, MA), and anti–β-actin (Santa Cruz Biotechnology).

Gene expression profiles were analyzed using liver tissues of HFD-fed WT and OPG^−/− mice. RNA sequencing (RNA-seq) was performed as reported previously (34). Raw reads were mapped to hg19 using the TopHat 2.0.13 software. Heat maps were generated using Cluster 3.0 and TreeView.

The data are presented as the mean ± SD or SE. Statistical analysis was performed via Microsoft Excel and Prism software (GraphPad, La Jolla, CA). Statistical significance was evaluated using the two-tailed paired or unpaired Student t test and among more than two groups by one-way ANOVA. A P value <0.05 was considered statistically significant.

We first investigated the expression pattern of OPG in WT mice using RT-qPCR. We found that OPG was expressed in all tissues examined, with the highest levels detected in bone and liver tissues and the lowest levels detected in the spleen (Supplementary Fig. 1A). In addition, we evaluated the effects of nutritional status on OPG mRNA expression in vitro and in vivo. We found that when the glucose concentration was between 5.6 and 22.2 mmol/L, the expression of OPG mRNA did not increase in L02 cells but increased significantly under high glucose conditions (33.3 mmol/L) (Supplementary Fig. 1B). Importantly, OPG mRNA expression was decreased dose dependently by FFA stimulation in L02 cells (Supplementary Fig. 1C), suggesting that FFA has a more significant effect on OPG mRNA expression than glucose. In WT mice, 24-h fasting led to an increase in OPG mRNA expression in the liver; however, 6 h of refeeding decreased hepatic OPG mRNA expression (Supplementary Fig. 1D).

To determine the involvement of OPG in obesity-related fatty liver, we examined hepatic OPG expression in SD- or HFD-fed WT, OPG^−/−, Adipoq^−/−, ob/ob, and db/db mice and patients with NAFLD. As shown in Fig. 1A–C, lower levels of OPG mRNA and protein were found in the livers of HFD-fed WT mice, ob/ob, and db/db mice compared with the SD-fed WT mice. However, decreased OPG expression was not observed in the livers of the Adipoq^−/− mice (data not shown). Furthermore, the OPG mRNA expression in the adipose tissues of db/db and HFD-fed WT mice was significantly lower than that in the adipose tissues of SD-fed WT mice (Supplementary Fig. 2A and B).

In addition, consistent with the results in animals, the expression levels of OPG mRNA/protein were significantly downregulated in the livers of NAFLD patients compared with normal subjects (Fig. 1D). Furthermore, prominent hepatosteatosis in the livers of NAFLD patients was shown by H&E staining, and IHC also indicated the decreased expression of OPG (Fig. 1E).

To investigate the effects of OPG on hepatic steatosis, Advs expressing OPG (Adv-OPG) and small hairpin RNA against the coding region of OPG (Adv-shOPG) were constructed to regulate OPG expression in cells. As expected, the mRNA and protein levels of OPG were markedly decreased in the L02 cells infected with Adv-shOPG (Supplementary Fig. 3) but were increased in L02 cells infected with Adv-OPG (Supplementary Fig. 4). In Adv-OPG–infected L02 cells and Hepa1-6 cells treated with FFAs, cellular TG contents were markedly increased compared with those of the Adv-GFP controls (Fig. 2A and C). Consistent with these findings, Oil Red O staining showed more lipid droplets in FFA-treated L02 cells (Fig. 2B) and Hepa1-6 cells (Fig. 2D) infected with Adv-OPG than in Adv-GFP controls. To determine whether an increase in OPG in the medium can regulate hepatosteatosis, MPHs isolated from WT mice were treated with OPG-Fc for 6 h after treatment with FFAs for 24 h. As shown in Fig. 2E, OPG-Fc treatment resulted in increasing TG content in MPHs. Consistent with increasing TG content, Oil Red O staining showed increased lipid stores when MPHs cells were treated with both OPG-Fc and FFAs compared with FFA treatment alone (Fig. 2F). In contrast, Adv-shOPG treatment resulted in decreased TG contents (Fig. 2G) and lipid droplets (Fig. 2H) in L02 cells treated with FFAs compared with the control cells.

The metabolic and biochemical parameters are shown in Supplementary Table 4. In the HFD-fed OPG^−/− mice, body weight (BW), epididymal white adipose tissue (eWAT), TG, FFA, LDL cholesterol, and serum OPG levels were significantly lower than those of the HFD-fed WT mice (Supplementary Table 4). As shown in Fig. 3A, the BW of the OPG^−/− mice fed the SD or HFD decreased markedly between weeks 7 and 12 compared with the BW of the WT mice, but food intake was not obviously different (data not shown). A representative photograph of SD- and HFD-fed WT and OPG^−/− mice is shown in Supplementary Fig. 5A and B. The OPG^−/− mice had a lower accumulation of body fat than the WT mice. When fed the HFD, liver weights were significantly reduced (Fig. 3B), and the livers showed a red appearance (Fig. 3C) in OPG^−/− mice. These effects were accompanied by reduced levels of fasting blood glucose, eWAT, alanine transaminase, and AST as well as a decreased eWAT-to-BW ratio in OPG^−/− mice (Supplementary Table 4 and Supplementary Fig. 6). Under both SD and HFD feeding conditions, the hepatic TG contents of the OPG^−/− mice were significantly decreased compared with the hepatic TG contents in WT mice (Fig. 3D). In accordance with the decreased liver TG content, Oil Red O and H&E staining of the liver revealed less steatosis in the HFD-fed OPG^−/− mice compared with the steatosis of the control littermates consuming the same diet (Fig. 3E and F). In the in vitro study, the cellular TG content was significantly lower in MPHs from HFD-fed OPG^−/− mice than in MPHs from the WT littermates fed the same diet (Fig. 3G). In parallel, Oil Red O staining showed decreased lipid stores in MPHs from OPG ^−/− mice relative to MPHs from WT mice (Fig. 3H).

To identify candidate targets of OPG, gene expression profiles were analyzed in the livers of HFD-fed WT and OPG^−/− mice by RNA-seq. As expected, a differential gene expression pattern was observed in the livers of these mice (Fig. 4A). Among the differentially expressed genes, CD36 and PPAR-γ, which are closely related to lipid metabolism, were markedly downregulated in the livers of OPG^−/− mice (Fig. 4A). To further confirm the transcription effects of OPG on these genes, we transfected L02 cells with Adv-OPG, Adv-shOPG, Adv-GFP, or Adv-control and further screened for the expression of genes related to lipid metabolism by RT-qPCR. We found that the mRNA expression of genes related to fatty acid uptake (CD36) was dramatically increased in L02 cells infected with Adv-OPG (Fig. 4B), whereas the mRNA expression levels of genes related to fatty acid synthesis and oxidation were not changed (Supplementary Fig. 7A and B). In contrast to the increased expression of CD36 mRNA in L02 cells infected with Adv-OPG, RT-qPCR analysis demonstrated that CD36 mRNA expression was significantly reduced in L02 cells infected with Adv-shOPG (Supplementary Fig. 8).

In an in vivo study, the mRNA expression levels of lipid metabolism genes were screened in SD- and HFD-fed WT and OPG^−/− mice. We found that the mRNA expression levels of some genes responsible for fatty acid uptake, including CD36, fatty acid transport protein 2 (FATP2), FATP4, and FATP5, were dramatically decreased in OPG^−/− mice, and among these genes, the changes in CD36 were the most obvious (Fig. 4C). Moreover, consistent with the results of the in vitro study, fatty acid synthesis and oxidation genes did not change in OPG^−/− mice (Supplementary Fig. 7C and D). Collectively, these findings show that OPG is associated with fatty acid uptake in the liver.

Based on the change in CD36 mRNA induced by OPG, we further explored the effect of OPG on the expression of CD36 protein after transcription. We infected L02 cells with Adv-OPG, Adv-GFP, Adv-shOPG, or Adv-control in the presence or absence of FFAs. The protein levels of CD36 in OPG-overexpressing cells were markedly increased compared with the protein levels of CD36 in the control cells (Fig. 5A), whereas knockdown of OPG in L02 cells led to downregulation of CD36 protein expression (Fig. 5B). To further confirm the effects of OPG on CD36, we observed changes in CD36 protein levels in the livers of SD- or HFD-fed WT and OPG^−/− mice. Western blotting results revealed that the protein expression of CD36 in the livers of OPG^−/− mice was significantly reduced relative to that of WT mice (Fig. 5C).

Based on the above observations, we speculated that if OPG’s promotion of hepatosteatosis in the liver relied on the upregulation of CD36, it would be anticipated that OPG should have no effect when CD36 is depleted in the livers. To investigate this possibility, we treated MPHs from CD36^−/− mice with OPG-Fc. Oil Red O staining showed that there was no obvious difference in the lipid droplets of MPHs treated with OPG-Fc and those that were not treated (Fig. 5D) and that the TG contents were also similar in these MPHs (Fig. 5E). Therefore, these results further indicate that the role of OPG in the promotion of TG accumulation is dependent on CD36 upregulation.

To further seek the molecular basis for OPG’s regulation of CD36 transcription, we identified a potential OPG binding site in the CD36 gene promoter in L02 cells. In accordance with the modulation of CD36 expression, OPG increased CD36 transcriptional activity (Fig. 5F). To confirm the binding site of OPG in the CD36 promoter, we analyzed the mouse CD36 promoter ranging from −2,584 to +165 bp in L02 cells. With a series of truncated promoters, luciferase activity assay revealed that deletion of three sites (−2,584/−1,138 bp, −2,584/−483 bp, and −2,584/−292 bp) had little effect on OPG-induced CD36 transcription, but deletion of the −2,584/99 bp region containing the PPAR response element (PPRE) site almost blocked the transcriptional activity of the CD36 promoter (Fig. 5F). Site-directed mutation in the PPRE binding site (located between −291 bp and −99 bp) completely blocked the effect of OPG on CD36 transcriptional activity in a transient transfection measurement (Fig. 5G). The results revealed that the PPRE binding site was indispensable for OPG activation of the CD36 promoter.

CD36 is a shared target of orphan nuclear receptors, including the liver X receptors (LXRs), PPAR-γ, and the pregnane X receptor (PXR) (35). We, therefore, examined the effects of OPG on orphan nuclear receptors both in vitro and in vivo. The results showed that Adv-OPG treatment led to a significant increase in PPAR-γ mRNA expression in L02 cells in both the presence and absence of FFAs, whereas the expression levels of LXRs, RXR, and PXR were unaffected (Fig. 6A). Consistent with this finding, the Western blotting results showed that the protein levels of PPAR-γ were also markedly increased in Adv-OPG–treated cells (Fig. 6B), while PPAR-γ phosphorylation remained unchanged (Supplementary Fig. 9). Accordingly, in the in vivo study, PPAR-γ expression (mRNA and protein levels) was significantly reduced in OPG^−/− mice compared with WT mice under both SD and HFD feeding conditions (Fig. 6C and D).

To further verify the association of OPG with CD36 and PPAR-γ, we explored whether the presence of PPAR-γ is required for the functional regulation of CD36 by OPG. We treated L02 cells with or without GW9662, a PPAR-γ inhibitor, in the presence and absence of FFAs. The Western blotting results showed that GW9662 treatment completely eliminated OPG-induced CD36 protein expression (Fig. 6E). Furthermore, a dual-luciferase reporter assay showed that the transcriptional activity of CD36 induced by Adv-OPG was completely inhibited by GW9662 (Supplementary Fig. 10). Collectively, these data demonstrate that the regulation of CD36 expression by OPG relies on the PPAR-γ pathway.

To investigate the molecular mechanism by which OPG might upregulate PPAR-γ expression, we explored potential regulatory pathways. We found that OPG overexpression resulted in a marked inhibition of ERK phosphorylation in cultured L02 cells, while the P38, Akt, PKA, and JNK pathways were unaffected (Fig. 7A). In contrast to the above findings, the downregulation of OPG expression by Adv-shOPG increased ERK phosphorylation in L02 cells (Fig. 7B). In the in vivo study, we found that the phosphorylation levels of ERK were significantly elevated in the livers of OPG^−/− mice compared with the phosphorylation in WT mice under both SD and HFD feeding conditions (Fig. 7C), but STAT5 phosphorylation was unaffected (Supplementary Fig. 11).

To further identify the molecular basis of the regulation of PPAR-γ and CD36 by OPG, we examined the effects of ERK signaling on OPG-mediated PPAR-γ and CD36 expression in L02 cells. As expected, Adv-shOPG treatment led to a significant downregulation in the protein levels of both CD36 and PPAR-γ in L02 cells treated with FFAs (Fig. 7D). However, the combined treatment of Adv-shOPG and SCH772984 (an ERK inhibitor) considerably attenuated the effects of Adv-shOPG on CD36 and PPAR-γ expression in these cells (Fig. 7D). The results collectively demonstrated that ERK was required for OPG-regulated expression of CD36 and PPAR-γ.

To validate these results in multiple cell lines, we repeated some experiments using HepaG2 cells. HepaG2 cells were infected with Adv-OPG or Adv-GFP, and results similar to those of the experiments with the L02 and Hepa1-6 cells were obtained (Supplementary Fig. 12).

Since OPG deficiency inhibits hepatic steatosis and CD36 expression induced by HFD feeding, we next wanted to examine whether restoring OPG expression in OPG^−/− mice could reverse these changes. As expected, OPG mRNA and protein expression levels were significantly increased in the livers of OPG^−/− mice infected with Adv-OPG (Supplementary Fig. 13A and B). Indeed, when we reexpressed OPG in HFD-fed OPG^−/− mice, hepatic TG contents and lipid droplets were markedly increased compared with those in the Adv-GFP–treated mice (Supplementary Fig. 13C–E). Furthermore, the protein expression levels of CD36 and PPAR-γ were also significantly increased, while ERK phosphorylation was inhibited in the liver (Supplementary Fig. 13B).

In the current study, we identified OPG as a secretory protein involved in the regulation of hepatic lipid metabolism in the ERK–PPAR-γ–CD36–dependent pathway and further confirmed the roles of OPG in hepatic steatosis in both in vitro and in vivo studies. First, we investigated OPG expression in obese and IR mice as well as in patients with NAFLD, which is the hepatic manifestation of obesity and IR. The downregulation of OPG expression in the liver of these mice and humans suggests that OPG may be related to hepatosteatosis.

Second, overexpression or knockdown of hepatic OPG resulted in increased or decreased TG accumulation, respectively, in primary and L02 hepatocytes. RNA-seq and RT-qPCR revealed that OPG regulated CD36 and PPAR-γ expression in tissues and cells.

Third, genetic deletion of OPG decreased hepatic steatosis and CD36 expression, whereas OPG-Fc treatment of MPHs from CD36^−/− mice did not decrease lipid accumulation in the presence of FFAs, indicating that OPG regulates hepatic lipogenesis through CD36 in vivo.

Finally, the studies of the underlying mechanisms revealed that the regulation of CD36 expression by OPG is mediated by the ERK–PPAR-γ pathway. OPG positively regulates CD36 transcription by binding to a PPRE site, thereby increasing lipid accumulation and resulting in hepatosteatosis. However, in the current study, it is surprising and interesting that the expression of OPG was downregulated in the livers of obese and IR animals and NAFLD patients, while in OPG^−/− mice, the accumulation of body fat and TGs and hepatic lipogenesis were reduced. The reason for the “paradoxical” downregulation of OPG in subjects with obesity and/or NAFLD is unknown. Because OPG is a secretory protein, this phenomenon might be a compensatory downregulation to counteract the metabolic stress imposed by hepatic steatosis. Indeed, when we reexpressed OPG in HFD-fed OPG^−/− mice, hepatic lipid accumulation was markedly increased. As reported previously, the decline in OPG in NAFLD may reflect the degree of the lesion (28).

The limitation of this study is the lack of tissue-specific OPG^−/− mice. We did not use hepatic-specific knockout mice mainly for the following reasons:

1. The liver-specific knockout of OPG, as a secretory protein, may lead to incomplete OPG knockout and increased secretion in other tissues (36).

2. As a secretory protein, OPG receptor knockout specifically in the liver may increase the binding of OPG to other tissue receptors.

3. It is not clear whether OPG specifically binds to its receptor.

Therefore, it is difficult to perform the specific knockout of OPG or its receptor in the livers of mice. All of these factors may affect the OPG phenotype as a whole.

Currently, the potential mechanism of the regulation of hepatic lipid metabolism by OPG is still unclear. Several studies have shown that circulating OPG levels are significantly decreased in obese individuals (22) and increased in NAFLD and in patients with T2DM (26,37), demonstrating seemingly paradoxical results. However, in humans and mice, we further confirmed that OPG expression was downregulated in obesity-related fatty liver. Downregulated OPG expression thus may represent a common characteristic of NAFLD. Two previous studies have revealed that OPG is expressed and produced in the adipose tissue of rats and humans (27,38). In the current study, OPG^−/− mice showed a lower accumulation of body fat and marked protection against hepatic steatosis versus WT mice, suggesting that OPG may be involved in lipid metabolism related to liver and adipose tissue in vivo. To address this possibility, we screened the effect of OPG on the expression of some genes related to lipid metabolism by RT-qPCR and Western blotting in vivo and in vitro. We found that up- or downregulation of OPG led to altered CD36 expression, whereas gene expression related to fatty acid synthesis and oxidation remained unchanged in vitro. In OPG^−/− mice, the expression of genes related to fatty acid uptake and transport, including CD36, FATP2, FATP4, and FATP5, in the liver was significantly reduced. Differences in gene expression related to fatty acid uptake between hepatic tissues and hepatocytes may be influenced by other confounding factors, such as stromal vascular fraction and Kupffer cells. Nevertheless, these results suggest that OPG may regulate fatty acid uptake and transport in hepatocytes.

CD36 is a mediator of FFA uptake in many tissues and may contribute to hepatic steatosis because hepatic fatty acid uptake plays an important role in steatosis (39,40). Koonen et al. (41) reported that CD36 overexpression in the liver increases TG accumulation and the development of hepatic steatosis and that hepatic tissues obtained from NAFLD patients have significantly increased CD36 expression levels (42). We thus speculate that CD36 may, at least in part, underlie the mechanisms linking the OPG regulation of hepatic steatosis. Indeed, OPG-Fc treatment did not lead to an increase in TG content in the hepatocytes of CD36^−/− mice, further confirming that CD36 is required for the OPG-mediated steatotic effects. Therefore, the lean phenotype and lower visceral fat content without significant changes in food intake in OPG^−/− mice under both SD and HFD conditions might be partly due to decreased lipogenesis mediated by CD36 and an increase in energy expenditure in vivo.

It has been reported that orphan nuclear receptors, including LXRs, PPAR-γ, and PXR, are related to lipogenesis (35). Moreover, CD36 has also been reported to be a common target of orphan nuclear receptors in their regulation of lipid metabolism (35). We thus screened the expression levels of five nuclear receptors in L02 cells infected with Adv-GFP or Adv-OPG and in the livers of WT or OPG^−/− mice. We found that the protein expression of PPAR-γ was significantly increased by OPG overexpression, whereas PPAR-γ phosphorylation and other nuclear receptors remained unchanged in L02 cells. These results suggest that OPG may induce an increase in PPAR-γ synthesis. In parallel with the results of the in vitro investigations, PPAR-γ expression was significantly decreased in OPG^−/− mice. All of those results suggest the participation of PPAR-γ in OPG-mediated hepatic steatosis. Based on the relationship between PPAR-γ and CD36, we speculated that PPAR-γ might be involved in the OPG regulation of CD36 expression. As expected, a PPAR-γ inhibitor treatment nearly completely blocked the OPG-mediated CD36 upregulation in L02 cells. Therefore, the present findings reveal that PPAR-γ mediates the CD36 expression induced by OPG.

A variety of protein kinases participate in adipogenesis and lipolysis in vivo. The mitogen-activated protein kinase (MAPK) signaling pathways, including JNK, p38 MAPK, and ERK (43), have also been found to play key roles in lipid metabolism in mammals (44). In addition, other protein kinases, including Akt and PKA, have been reported to be related to hepatic lipogenesis. For example, Akt has been found to stimulate hepatic lipogenesis via both mammalian target of rapamycin complex 1 (mTORC1)-dependent and mTORC1-independent pathways (45). Lu et al. (46) also found that PKA phosphorylates SREBP1c and consequently suppresses lipogenic activity.

In addition, it has been revealed that lipid metabolism in the liver and thermogenesis in fat are regulated by cAMP-dependent signaling pathways (PKA/Akt/AMPK) (47). In recent studies, STAT5 has also been found to play a role in CD36 regulation and hepatic steatosis (48–50). Therefore, we screened the expression or phosphorylation of these protein kinases in L02 cells with OPG overexpression or in OPG^−/− mice to further explore the regulatory mechanism of OPG in lipogenesis. Somewhat surprisingly, OPG overexpression significantly inhibited the expression of ERK protein kinases, while other protein kinases and STAT5 were unaffected. In parallel, the deletion or downregulation of OPG increased the phosphorylation of ERK in vivo and in vitro, and treatment with an ERK inhibitor greatly blocked the effects of Adv-shOPG on PPAR-γ and CD36 expression. These results suggest that the regulation of PPAR-γ and CD36 expression by OPG is closely related to the inhibition of ERK. Based on previous reports and our observations, it can be deduced that ERK–PPAR-γ–CD36 cascades may be responsible for the regulation of hepatic steatosis by OPG. However, how OPG acts on the ERK–PPAR-γ–CD36 pathway is unknown. In the current study, treatment with Adv-OPG or OPG-Fc increased CD36 and PPAR-γ expression and inhibited ERK phosphorylation in vitro. Therefore, we speculate that OPG, like other secreted proteins related to bone metabolism, such as Nel-like protein (Nell-1), inhibits the ERK pathway by binding to the cell membrane receptor integrin b1 (51,52). Nevertheless, this possibility needs to be further explored.

In summary, the current study demonstrates that OPG is an important metabolic regulator. OPG promotes lipid accumulation in the liver, mainly through enhancement of CD36 expression, which is dependent on the activation of downstream ERK–PPAR-γ cascades (Fig. 8). Therefore, decreasing OPG expression or disrupting the connection between OPG and the ERK–PPAR-γ–CD36 pathway may be a promising therapeutic target for NAFLD and other metabolic diseases.

Funding. This study was supported by the grants from the National Natural Science Foundation of China (81670755, 81873658, IRT1216, 81870755, and 81570752) and the Graduate Student Scientific Research Foundation of Chongqing Medical University (CYB18159).

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

Author Contributions. C.Z., X.L., J.C., B.Z., and M.Y. researched the data. R.L. and D.L. analyzed the data and wrote the manuscript. H.F.G., Z.Z., and H.Z. contributed to writing the manuscript and helpful discussion. L.L. and G.Y. directed the project, contributed to discussion, and wrote, reviewed, and edited the manuscript. L.L. and G.Y. are the guarantors of this work and, as such, had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.

Prior Presentation. Parts of this study were presented in abstract form at the 54th Annual Meeting of the European Association for the Study of Diabetes, Berlin, Germany, 1–5 October 2018.