Sonic hedgehog signaling instigates high-fat diet–induced insulin resistance by targeting PPARγ stability

Obesity is a major risk for patients with chronic metabolic disorders including type 2 diabetes. Sonic hedgehog (Shh) is a morphogen that regulates the pancreas and adipose tissue formation during embryonic development. Peroxisome proliferator-activated receptor γ (PPARγ) is a member of the nuclear receptor superfamily and one of the most important regulators of insulin action. Here, we evaluated the role and mechanism of Shh signaling in obesity-associated insulin resistance and characterized its effect on PPARγ. We showed that Shh expression was up-regulated in subcutaneous fat from obese mice. In differentiated 3T3-L1 and primary cultured adipocytes from rats, recombinant Shh protein and SAG (an agonist of Shh signaling) activated an extracellular signal–regulated kinase (ERK)-dependent noncanonical pathway and induced PPARγ phosphorylation at serine 112, which decreased PPARγ activity. Meanwhile, Shh signaling degraded PPARγ protein via binding of PPARγ to neural precursor cell-expressed developmentally down-regulated protein 4-1 (NEDD4-1). Furthermore, vismodegib, an inhibitor of Shh signaling, attenuated ERK phosphorylation induced by a high fat diet (HFD) and restored PPARγ protein level, thus ameliorating glucose intolerance and insulin resistance in obese mice. Our finding suggests that Shh in subcutaneous fat decreases PPARγ activity and stability via activation of an ERK-dependent noncanonical pathway, resulting in impaired insulin action. Inhibition of Shh may serve as a potential therapeutic approach to treat obesity-related diabetes.

Obesity is a major risk for patients with chronic metabolic disorders including type 2 diabetes. Sonic hedgehog (Shh) is a morphogen that regulates the pancreas and adipose tissue formation during embryonic development. Peroxisome proliferator-activated receptor ␥ (PPAR␥) is a member of the nuclear receptor superfamily and one of the most important regulators of insulin action. Here, we evaluated the role and mechanism of Shh signaling in obesity-associated insulin resistance and characterized its effect on PPAR␥. We showed that Shh expression was up-regulated in subcutaneous fat from obese mice. In differentiated 3T3-L1 and primary cultured adipocytes from rats, recombinant Shh protein and SAG (an agonist of Shh signaling) activated an extracellular signal-regulated kinase (ERK)-dependent noncanonical pathway and induced PPAR␥ phosphorylation at serine 112, which decreased PPAR␥ activity. Meanwhile, Shh signaling degraded PPAR␥ protein via binding of PPAR␥ to neural precursor cell-expressed developmentally down-regulated protein 4-1 (NEDD4-1). Furthermore, vismodegib, an inhibitor of Shh signaling, attenuated ERK phosphorylation induced by a high fat diet (HFD) and restored PPAR␥ protein level, thus ameliorating glucose intolerance and insulin resistance in obese mice. Our finding suggests that Shh in subcutaneous fat decreases PPAR␥ activity and stability via activation of an ERK-dependent noncanonical pathway, resulting in impaired insulin action. Inhibition of Shh may serve as a potential therapeutic approach to treat obesity-related diabetes.
Insulin is a crucial anabolic hormone that stimulates the metabolism of carbohydrates, lipids, and proteins. Adipose tissue receives insulin signals and is responsible for plasma glucose clearance by stimulating its uptake and utilization (1). An inability of tissues to adequately respond to insulin leads to impaired glucose homeostasis and the pathogenesis of type 2 diabetes (2). Obesity resulting from inappropriate food intake is commonly accompanied by metabolic disorders, such as electrolyteimbalance,hyperuricemia,dyslipidemia,andinsulinresis-tance (3). Type 2 diabetes is frequently associated with an accumulation of visceral adipose tissue (4). Nonetheless, the predominant storage of fat in the subcutaneous adipose tissue is indicative of a better metabolic profile (5). The molecular mechanism behind these clinical observations remains unknown. In adipocytes, insulin signaling starts from the binding of insulin to its receptor, leading to the phosphorylation of insulin receptor substrate-1 (IRS-1). 2 Activated phosphoinositide 3-kinase (PI3K) triggers the production of specific phosphoinositides, which recruit protein kinase B (Akt) to the plasma membrane (6). Subsequent Akt phosphorylation is responsible for most of the metabolic actions of insulin, such as glucose uptake by inducing the translocation of glucose transporter 4 from intracellular storage to the plasma membrane (7). Thus, Akt phosphorylation normally serves as a downstream marker of insulin. ERK and c-Jun N-terminal kinase (JNK) activations negatively regulate insulin signaling through serine phosphorylation of IRS-1 (8).
Hedgehog (Hh) proteins, including Shh, Indian hedgehog (Ihh), and Desert hedgehog (Dhh), were initially defined as morphogens that regulate the embryo development (9). In mammals, Hh binding to the inhibitory receptor Patched leads to the activation of the cell-surface receptor Smoothened (Smo) and subsequently drives the transcription of Gli target genes (10). Although most Hh-induced biological effects result from the transcriptional program, several noncanonical pathways have been reported, such as mitogen-activated protein kinase ERK (11). Hh signaling is normally quiescent in adults, except during various pathologies, such as cancers (12), immune-demyelinating neuropathy (13), and polycystic kidney disease (14). Shh inhibits adipocyte differentiation by diverting preadipocytes away from adipogenesis (15), which is implicated in the development of white adipose tissue (16). The Hh-Gli2 axis drives lipogenesis in adipocytes, leading to increased obesity in adult mice (17). Moreover, Shh-related ciliopathy is involved in obesity, cognitive impairment, and limb deformities, which has been defined as Bardet-Biedl syndrome in humans (18). Shh regulates epithelial and ␤-cell expansion during early pancreas morphogenesis (19). Excessively activated Hh signaling impairs ␤-cell function and insulin secretion, resulting in glucose intolerance in adult mice (20,21). Furthermore, hepatic Hh signaling contributes to the progression of nonalcoholic fatty liver diseases by promoting liver inflammation (22). Similarly, its signaling in myeloid cells of adipose tissue is suggested to be a trigger of type 2 diabetes (23).
PPAR␥ is a member of the nuclear receptor superfamily and a master regulator of insulin action. Although specific deletion of PPAR␥ in muscle, liver, macrophages, or brain alters glucose homeostasis and induces insulin resistance (24 -28), adipose tissue is the major site for insulin sensitizing by PPAR␥ (29). Rosiglitazone, a selective agonist of PPAR␥, has been used to treat type 2 diabetes with a glucose lowering effect. Due to side effects including heart failure, rosiglitazone has been withdrawn from the first-line therapy in clinical practice (30). Hence, identifying PPAR␥ regulators as therapeutic targets would be a novel approach in the treatment of type 2 diabetes.
In this study, we demonstrate that Shh in subcutaneous adipose tissue decreases the activity and stability of PPAR␥ by activating the ERK-dependent noncanonical pathway. Conversely, vismodegib, an inhibitor of Shh signaling, improves obesity-related insulin resistance by stabilizing PPAR␥.

Shh is induced in subcutaneous adipose tissue from obese mice
To determine whether Hh ligands are implicated in obesity, C57Bl/6 mice were first fed with a HFD for 12 weeks. The mice developed pronounced obesity ( Fig. S1A) with no difference in food intake (Fig. S1B). Glucose tolerance test (Fig. S1C) and insulin tolerance test (Fig. S1D) revealed impaired insulin sensitivity in these mice. Next, we measured the expressions of Shh, Ihh, and Dhh in subcutaneous, epididymal and brown adipose tissues. ELISA was used to detect the protein level of Shh because of its better sensitivity. Compared with lean mice, both mRNA ( Fig. 1A) and protein ( Fig. 1B) levels of Shh were significantly increased in subcutaneous fat from obese mice. Although a slight augmentation of Shh mRNA was observed in epididymal and brown adipose tissues, no difference was found at the protein level. No differential expression of Ihh (Fig. 1C) and Dhh (Fig. 1D) was observed in subcutaneous, epididymal, and brown adipose tissues.

Shh decreases PPAR␥ protein level via an ERK-dependent noncanonical pathway
To determine the involvement of Shh signaling in adipocytes, we treated the differentiated 3T3-L1 adipocytes with recombinant Shh protein or SAG, a Smo agonist, for the indicated times or with different doses. Recombinant Shh and SAG activated the mitogen-activated protein kinase pathway with a maximum ERK phosphorylation at 1 h after stimulation ( Fig. 2A). ERK phosphorylation induced by recombinant Shh and SAG occurred in a dose-dependent manner (Fig. 2B). In primary rat adipocytes, ERK phosphorylation was also increased by recom-binant Shh and SAG (Fig. 2C). These results indicate that Shh signaling activates the ERK pathway in adipocytes.
To study the cross-talk between the ERK-dependent noncanonical and Gli-dependent canonical pathways we investigated the effect of PD98059 on Gli1 activation in 3T3-L1 and primary rat adipocytes. SAG-induced Gli1 activation was not attenuated by PD98059 (Fig. 2J). These data confirm that ERK phosphorylation and Gli activation are 2 independent pathways. Taken together, these findings suggest that Shh-increased ERK phosphorylation promotes PPAR␥ phosphorylation at Ser-112 and decreases the protein level of PPAR␥.

Shh decreases PPAR␥ stability via a ubiquitin-dependent mechanism
To explore how Shh signaling decreased the PPAR␥ protein level, PPAR␥-transfected HEK 293 cells were pretreated with cycloheximide and then exposed to SAG. As shown in Fig. 3A, SAG significantly increased PPAR␥ degradation. Pretreatment

Vismodegib treatment in type 2 diabetes
with the proteasome inhibitor MG132 significantly prevented the SAG-triggered decrease in PPAR␥ protein level (Fig. 3B), indicating a role of the proteasome in PPAR␥ degradation. Depletion of the E3 ubiquitin ligase NEDD4-1 led to PPAR␥ accumulation (Fig. 3C). Co-immunoprecipitation indicated that the binding of NEDD4-1 to PPAR␥ was largely augmented by SAG treatment (Fig. 3D). Next, we determined if Shh signaling promoted PPAR␥ ubiquitination. The level of ubiquitinated PPAR␥ was markedly higher in SAG-treated cells (Fig. 3E). Col-lectively, these data indicate that Shh signaling enhances PPAR␥ degradation via NEDD4-1-dependent ubiquitination.

Vismodegib improves insulin sensitivity in obese mice induced by a HFD
We assessed the potential therapeutic effects of vismodegib on insulin resistance in mice fed a HFD. Vismodegib treatment slightly decreased the body weight of both mice fed with a normal (ND) or HFD. No statistical differences were observed Figure 2. PPAR␥ protein level is decreased via the ERK pathway in adipocytes. A-C, differentiated 3T3-L1 adipocytes were serum-starved and then incubated with Shh (0.5 g/ml) or SAG (0.5 mol/liter) for the indicated times (A). Serum-starved 3T3-L1 adipocytes were treated with different concentrations of Shh (0, 0.1, 0.2, 0.5, 1, and 2 g/ml) or SAG (0, 0.1, 0.2, 0.5, 1, and 2 mol/liter) for 60 min (B). C, primary rat adipocytes (2 ϫ 10 5 cells/well) after serum starvation were treated with Shh (0.5 g/ml) or SAG (0.5 mol/liter) for 60 min. The level of p-ERK and ERK was analyzed using immunoblotting. D-F, before incubation with Shh (0.5 g/ml) for 60 min, 3T3-L1 or primary rat adipocytes were transfected with siRNA against Smo (D) or pretreated with vismodegib (100 nmol/liter, E) for 30 min. F, PD98059 (PD, 10 mol/liter)-pretreated cells were exposed to SAG (0.5 mol/liter) for 60 min. Levels of p-PPAR␥ and PPAR␥ was analyzed by using immunoblotting. G-I, Smo siRNA-transfected (G) or vismodegib-pretreated (H) cells were incubated with Shh (0.5 g/ml) for 24 h. I, PD98059 (PD, 10 mol/liter)-pretreated cells were exposed to SAG (0.5 mol/liter) for 24 h. Levels of PPAR␥ and ␤-actin were analyzed using immunoblotting. Immunoblots shown are representative of 3 independent experiments. J, differentiated 3T3-L1 (left) and primary rat adipocytes (right) were pretreated with or without PD98059 (PD, 10 mol/liter) for 30 min, and then exposed to SAG (0.5 mol/liter) for 24 h. Cell lysates were analyzed to determine the mRNA level of Gli1. The results were normalized to the level of cyclophilin mRNA. Data were from 3 independent experiments performed in triplicate. *, p Ͻ 0.05; NS, no significance. A, cells were pretreated with cyclohexamide (CHX) (5 g/ml) for 30 min before exposure to SAG (0.5 mol/liter), followed by immunoblotting to detect PPAR␥ and ␤-actin (left). Quantifications of band intensity normalized to ␤-actin (right). B, cells were incubated with SAG (0.5 mol/liter) for 24 h in the presence or absence of MG132 (10 mol/liter) pretreatment. C, cells were infected with scramble shRNA lentivirus or different lentiviral shRNA constructs against NEDD4-1. PPAR␥, NEDD4-1, and ␤-actin levels were analyzed by immunoblotting. D, immunoblotting (IB) of whole cell lysates (Input) and immunoprecipitates (IP) from PPAR␥-overexpressing HEK 293 cells with or without SAG (0.5 mol/liter) treatment for 24 h. MG132 (10 mol/liter) was added to the medium 12 h before collecting. E, PPAR␥-transfected HEK 293 cells were pretreated with MG132 (10 mol/liter) and then incubated with or without SAG (0.5 mol/liter) for 24 h. PPAR␥ was immunoprecipitated from cell lysates and immunoblotted with an anti-ubiquitin antibody (left). Band intensities were normalized to that of IgG (right). Immunoblots shown are representative of 3 independent experiments. *, p Ͻ 0.05.

Vismodegib treatment in type 2 diabetes
between the two groups (Fig. 4A). Vismodigib did not affect body weight in either ND-or HFD-fed mice after 16 days of treatment (Fig. 4B). Food intake did not differ with vismodegib treatment (Fig. 4C). However, vismodegib significantly improved HFD-induced impairment of glucose tolerance (Fig.  4D) and insulin sensitivity (Fig. 4E). A HFD impaired insulin secretion after glucose stimulation as assessed by the insulin level in the blood. Vismodegib did not affect the glucose-induced insulin level in mice fed with either a ND or HFD (Fig.  4F). Accordingly, the level of p-Akt (Ser-473), a downstream marker of insulin signaling, was increased in ND-fed mice with vismodegib treatment compared with the vehicle group. Importantly, vismodegib restored the HFD-attenuated p-Akt level (Fig. 4G). These results indicate that inhibition of Shh signaling improves HFD-induced insulin resistance.

Vismodegib inhibits ERK activation in subcutaneous adipose tissue of obese mice
Because Gli1 induction is a main reporter of Hh signaling activation, we investigated the effect of vismodegib on Gli1 activation. Expression of Gli1 was up-regulated in adipose tissue of HFD-fed mice. Surprisingly, the induction was not attenuated

Vismodegib restores PPAR␥ protein level in adipose tissue of obese mice
To confirm the effect of vismodegib on the decreased PPAR␥ protein level as demonstrated in vitro, we investigated the PPAR␥ level in subcutaneous fat of mice fed with a ND and HFD. Compared with the vehicle group, the PPAR␥ protein level did not significantly change in ND mice upon vismodegib treatment. However, vismodegib significantly restored the PPAR␥ protein level in HFD mice (Fig. 6A). Vismodegib did not affect the PPAR␥ mRNA level in either ND-or HFD-fed mice (Fig. 6B). Furthermore, it also restored HFD-decreased expression of PPAR␥ target genes, including adiponectin and CD36 (Fig. 6C).

Discussion
Obesity is a medical condition associated with the accumulation of excess body fat, triggering metabolic disorders, such as type 2 diabetes (32). Perturbation of fatty acid metabolism is a major factor contributing to whole-body insulin resistance. Fatty acid-induced inflammation and cellular lipid overload inhibit insulin signaling by triggering endoplasmic reticulum stress and oxidative stress (33).
Clinical studies have demonstrated that excess accumulation of visceral adipose tissue is associated with cardio-metabolic risk factors (4). Subcutaneous adipose tissue shows a better metabolic outcome (5). The underlying molecular mechanism remains poorly understood. Our study provides a detailed analyses of the expression of three Hh ligands in epididymal, subcutaneous, and brown fat tissues. Here, we report that Shh expression was increased in subcutaneous fat, whereas it remained unchanged in epididymal and brown fat. This may explain the observation that the PPAR␥ protein level was decreased mainly in subcutaneous adipose tissue. We did not detect differential expression of Ihh or Dhh in any of the three adipose tissues. In mammals, the 3 Hh genes are located on different chromosomes but have highly conserved sequences at the protein level. Shh shares 91 and 76% identity with Ihh and Dhh, respectively. Ihh has 80% identity with Dhh (34). All three ligands share the same downstream signaling.
Although investigations typically rely on the Gli-dependent canonical pathway, several noncanonical signaling pathways have been reported. For instance, rapid induction of Src family kinases by Shh is implicated in spatial axonal outgrowth (35). Hh signaling triggers Warburg-like metabolic reprogramming via a Smo-Ca 2ϩ -AMP-activated protein kinase axis (36). Shh, Ihh, and Dhh drive cytoskeletal rearrangement and endothelial migration via a Smo-Rho A axis (37). Furthermore, Shh activates PI3K to promote mural cell migration and recruitment into neovessels (11). We found that ERK phosphorylation was enhanced by Shh signaling in adipocytes. This noncanonical pathway of ERK activation was mediated by Smo, a member of the G protein-coupled receptor superfamily.
The nuclear receptor PPAR␥ is a transcription factor and a master regulator of adipogenesis and glucose metabolism (38). In addition to its transcriptional regulation, activity and stability of the PPAR␥ protein are under precise posttranscriptional modifications, including phosphorylation, ubiquitination, SUMOylation, O-GlcNAcylation, and S-nitrosylation (27, 39 -41). Site-specific phosphorylation of PPAR␥ at Ser-112 by MAPK inhibits PPAR␥ activity during adipogenesis and insulin action (42), whereas PPAR␥ agonists cause its phosphorylation at serine 273 by cyclin-dependent kinase 5, leading to its nuclear translocation (31). Our data demonstrated that Shh signaling-induced PPAR␥ phosphorylation at Ser-112 was responsible for the decreased PPAR␥ activity
One question of interest is whether other nuclear receptors could be regulated by the Hh-triggered and phosphorylationdependent ubiquitination. In this regard, we looked at the PPAR␦ protein level, another member of the PPAR subfamily expressed in adipose tissues. Vismodegib did not have a significant effect on the PPAR␦ protein level in adipose tissues of either ND-fed or HFD mice (Fig. S3). The third member of this subfamily, PPAR␣, is mainly expressed in the liver, but not in adipose tissue (52). RXR␣, the dimerization partner of PPARs, is not affected by a HFD in mice (53). PPAR␥ degradation by Shh signaling may explain the impaired formation of the PPAR␥/RXR heterodimer in insulin resistance. In addition, LXR agonists increase GLUT4 expression, resulting in insulin-mediated glucose uptake into adipose tissue (54). SREBP is also a key transcription factor controlling the synthesis of cholesterol and triglycerol (55). Potential cross-talk between Hh signaling and these metabolic regulators certainly warrants further investigation.
Recently, the Hh signaling pathway was implicated in metabolism and the pathogeneses of several metabolic diseases. Shh signaling switches preadipocytes away from adipogenesis and instead drives them toward osteogenesis. Shh-related ciliopathy is involved in Bardet-Biedl syndrome in humans with severe obesity (18). The Hh-activated Smo-Ca 2ϩ -AMP-activated protein kinase axis is involved in Warburg-like metabolic reprogramming (36). Hh signaling promotes the hepatic inflammatory state and controls the progression of nonalcoholic fatty liver diseases (22). Insulin resistance is triggered by Hh signaling in inflammatory cells of obese mice (23). In this study, we provide evidence that Shh-mediated PPAR␥ ubiquitination via NEDD4-1 is a novel mechanism to illustrate the impaired insulin action in type 2 diabetes.
Vismodegib is the first Hh pathway inhibitor approved in the United States and Europe for the treatment of basal cell carcinoma (56). Recently, Kwon et al. (22) found that intraperitoneal administration of vismodegib decreases hepatic fat accumulation in mice via anti-inflammatory effects. In our study, vismodegib slightly decreased body weight in both ND-fed and HFD mice with statistical significance observed only at a single time point. However, the body weights before and after treatment were not statistically significantly different. Because the sample size was relatively small, we cannot conclude that vismodegib affects body weight. In addition to demonstrating a novel mechanism by which Hh signaling attenuates PPAR␥ action,

Vismodegib treatment in type 2 diabetes
our study also provides the first proof of concept that Hh inhibition may represent a new therapeutic strategy to treat metabolic disease featuring with insulin resistance.
Taken together, our data demonstrate that Shh signaling in subcutaneous adipose tissue leads to PPAR␥ degradation via NEDD4-1-mediated ubiquitination. These results also provide evidence that Shh-induced ERK phosphorylation is a trigger of decreased PPAR␥ activity and stability (Fig. 7). Improved insulin sensitivity by vismodegib treatment indicates that Shh signaling could be a potential therapeutic target in type 2 diabetes.

Animal procedures
Male Sprague-Dawley rats and C57BL/6 mice were bred with controlled temperature (25°C) and a 12-h light and dark cycle. Animal experiments were approved by the institutional ethics review board of Xi'an Jiaotong University (XJTULAC2015-404), and performed in accordance with the NIH guidelines for the care and use of animals. Mice at 8 weeks were fed a HFD (60% fat) or kept on ND (Research Diets, New Brunswick, NJ). Twelve weeks later, mice were treated with intraperitoneal injections of vehicle or vismodegib (5 mg/kg) every other day. Food intake and body weight were measured before every injection and followed every other day.

Glucose tolerance test and insulin tolerance test
For the glucose tolerance test, mice were food-deprived overnight for 14 h with free water supply. Glucose (2 mg/kg) was intraperitoneally injected and blood glucose was determined at 15, 30, 60, and 120 min after the injection. For the insulin tolerance test, mice intraperitoneally received recombinant insulin (0.8 units/kg, Humulin, Eli Lilly and Co., IN). Blood glucose was determined at 0, 15, 30, 60, and 120 min after the injection.

RNA extraction, reverse transcriptase-quantitative PCR (RT-qPCR)
Total RNA was extracted by using TRIzol (Invitrogen). RT-qPCR was performed using SYBR Green (Promega, Madison, WI). Primer sequences were described in Table S1. Cyclophilin was used as an internal control.

Immunoblotting, enzyme-linked immunosorbent assay (ELISA), and immunoprecipitation
Proteins were extracted in RIPA buffer supplemented with protease and phosphatase inhibitors. Protein concentrations Increased Shh expression in subcutaneous fat of obese mice activates p-ERK, which phosphorylates PPAR␥ at serine 112, leading to an impaired PPAR␥ activity. On the other hand, Shh-activated ERK phosphorylation promotes NEDD4-1-dependent PPAR␥ ubiquitination and degradation. Transcription of PPAR␥-targeted genes is therefore reduced, leading to an impaired insulin action. Thus, by targeting the Shh receptor Smo, vismodegib instigates HFDinduced insulin resistance in obese mice.

Vismodegib treatment in type 2 diabetes
were measured using the BCA protein assay. Immunoblotting was performed with appropriate primary antibodies and horseradish peroxidase-conjugated secondary antibodies followed by ECL detection. The concentration of Shh protein and serum insulin level was measured using the ELISA kit (Elabsciences, Wuhan, China) according to the manufacturer's instructions.
For immunoprecipitation, cell lysates were incubated with the appropriate antibodies or control IgG at 4°C overnight followed by incubation with protein A/G-Sepharose beads. Immunoprecipitates were washed with NETN buffer (100 mM NaCl, 1 mM EDTA, 20 mM Tris, pH 8.0, and 0.5% Nonidet P-40). Band intensity was quantified with ImageJ and normalized to loading control.

Statistical analyses
Results are reported as mean Ϯ S.D. We examined the differences between groups using Student's t test (comparison of two groups) or one-way analysis of variance (comparisons more than two groups) test with a Bonferrioni test to correct for multiple comparison testing. For body weight, glucose tolerance test, and insulin tolerance test, we conducted a repeated measures analysis of variance followed by a LSD post hoc test and a Bonferrioni test to correct for multiple comparison testing. p values Ͻ0.05 were considered significant.