Saturated fatty acids activate ERK signaling to downregulate hepatic sortilin 1 in obese and diabetic mice.

Hepatic VLDL overproduction is a characteristic feature of diabetes and an important contributor to diabetic dyslipidemia. Hepatic sortilin 1 (Sort1), a cellular trafficking receptor, is a novel regulator of plasma lipid metabolism and reduces plasma cholesterol and triglycerides by inhibiting hepatic apolipoprotein B production. Elevated circulating free fatty acids play key roles in hepatic VLDL overproduction and the development of dyslipidemia. This study investigated the regulation of hepatic Sort1 in obesity and diabetes and the potential implications in diabetic dyslipidemia. Results showed that hepatic Sort1 protein was markedly decreased in mouse models of type I and type II diabetes and in human individuals with obesity and liver steatosis, whereas increasing hepatic Sort1 expression reduced plasma cholesterol and triglycerides in mice. Mechanistic studies showed that the saturated fatty acid palmitate activated extracellular signal-regulated kinase (ERK) and inhibited Sort1 protein by mechanisms involving Sort1 protein ubiquitination and degradation. Consistently, hepatic ERK signaling was activated in diabetic mice, whereas blocking ERK signaling by an ERK inhibitor increased hepatic Sort1 protein in mice. These results suggest that increased saturated fatty acids downregulate liver Sort1 protein, which may contribute to the development of dyslipidemia in obesity and diabetes.

were maintained on a standard chow diet and water ad libitum and were housed in a room with a 12 h light (6 AM to 6 PM) and 12 h dark (6 PM to 6 AM) cycle. WT mice were fed either a standard chow diet or a Western diet (Harlan Teklad TD.88137, 42% fat calories, 0.2% cholesterol) for four months. For U0126 injection, mice were fasted for 16 h and given a single dose of U0126 (5 mg/kg, in sterile saline containing 1% DMSO) via tail vein. Control mice received vehicle alone. Insulin-defi cient C57BL/6 mice were generated via intraperitoneal injection with 7.5 mg/kg streptozocin (STZ) once daily for fi ve consecutive days. Control mice were injected with vehicle only (sodium citrate buffer, pH 4.5). One week after the last injection, mice with hyperglycemia (nonfasting blood glucose > 400 mg/dl) were used. All studies were conducted in conformity with the Public Health Service Policy on Humane Care and Use of Laboratory Animals and were approved by the Institutional Animal Care and Use Committee.

Human liver samples
Human liver samples were obtained from the University of Kansas Liver Center Tissue Bank. Liver samples used as controls were from individuals who were without the presence of significant liver steatosis by histological analysis and who had a body mass index (BMI) р 25. Steatotic liver samples were selected based on the presence of steatosis у 20% without evidence of fibrosing steatohepatitis or cirrhosis by histology and BMI у 30. Histology was read by multiple board-certifi ed pathologists.

Cell culture and fatty acid treatment
The human hepatoma cell line HepG2 was obtained from the American Type Culture Collection (Manassas, VA). Primary human hepatocytes were obtained from the Liver Tissue and Cell Distribution System of the National Institutes of Health (S. Strom, University of Pittsburgh, PA) or from the Biospecimen Core facility, University of Kansas Medical Center (KUMC). Primary mouse hepatocytes were obtained from the Biospecimen Core facility at KUMC. Palmitate (50 mM) or oleate (50 mM) was fi rst dissolved in 100% ethanol by heating at 80°C to yield a clear solution, which was then added to serum-free medium containing 10% fatty acid-free BSA (BSA) and incubated at 37°C for 2 h to obtain a fi nal concentration of palmitate at 5 mM. For cell treatment, the 5 mM palmitate solution was further diluted in serum-free culture medium containing 1% BSA and incubated at 37°C for 2 h before addition to cells. Control cells were treated with vehicle prepared using the same procedure.

Measurement of Sort1 ubiquitination
HepG2 cells were transfected with an HA-tagged ubiquitin expression plasmid (PRK5-HA-Ubiquitin-WT, Addgene plasmid 17608) that was deposited by Dr. Ted M. Dawson (Johns Hopkins University School of Medicine) ( 16 ). On the same day, HepG2 cells were then infected with Adenovirus expressing a C-terminal fl ag-tagged Sort1 at multiplicity of infection of 1. After 24 h, Flagtagged Sort1 was immunoprecipitated with anti-fl ag antibody (M2) conjugated to magnetic beads (Sigma). Ubiquitinated Sort1 was detected with anti-HA antibody in Western blot. HepG2 cells that did not express Flag-Sort1 or HA-Ubiquitin were used as a negative control.

Recombinant adenovirus
Adeno-GFP was a gift from Dr. Li Wang (University of Utah, Salt Lake City, UT). Adenovirus-expressing human full length Sortilin 1 was a generous gift from Dr. Anders Nykjaer (Lundbeck Foundation Research Center MIND, Department of Medical Biochemistry, Aarhus University, Denmark). Adenovirus expressing a C-terminal fl ag-tagged human Sortilin1 was generated using AdEasy Adenoviral Vector System (Agilent Technologies, Sortilin 1 (Sort1), a transmembrane-sorting receptor ( 11 ), has recently been identifi ed as a novel regulator of hepatic apoB metabolism and plasma LDL-cholesterol (LDL-C) concentrations ( 12,13 ). A number of genome-wide association studies (GWAS) fi rst identifi ed a strong association between LDL-C and SNPs at the 1p13.3 locus containing SORT1 and several other genes ( 12 ). It was further shown that the homozygous causal minor allele was associated with increased hepatic Sort1 mRNA and lower plasma LDL-C, which translated into signifi cant reduction of myocardial infarction risk in humans ( 12 ). Sort1 mainly localizes in the trans-Golgi network, and it is involved in traffi cking of various target proteins to the secretory pathway or to endosome/lysosome compartment ( 11 ). A small amount of Sort1 ( ‫ف‬ 10%) also localizes on the cell membrane and is involved in ligand internalization or intracellular signaling ( 11 ). Mechanistic studies in mice suggest that liver Sort1 may reduce plasma LDL-C via modulating apoB metabolism: fi rst, intracellular Sort1 recognizes apoB as a ligand and directs newly synthesized apoB for lysosomal degradation, leading to reduced hepatic apoB secretion; second, cell surface Sort1 may facilitate extracellular apoB-containing LDL internalization for subsequent lysosomal degradation ( 12,14,15 ).
Because abnormal hepatic apoB metabolism critically contributes to dyslipidemia and because the molecular mechanisms for altered apoB metabolism in obesity and diabetes are still not fully understood, we asked whether hepatic Sort1 expression is changed in obesity and diabetes, and if so, what the impact of altered hepatic Sort1 levels on the development of diabetic dyslipidemia is. We fi rst found that hepatic Sort1 protein was signifi cantly reduced in obese and diabetic mice and in steatotic livers of obese humans. We then provided in vitro and in vivo evidence demonstrating that elevated FFA-mediated MAPK activation results in posttranslational Sort1 degradation, which provides a novel molecular basis for hepatic apoB/VLDL overproduction in diabetes. In addition, we showed that overexpression of Sort1 in the liver was suffi cient to reduce plasma cholesterol and triglycerides in obese and diabetic mice, suggesting that therapeutic approaches that can either restore or further raise liver Sort1 levels may provide benefi ts in improving plasma lipid homeostasis in obesity and diabetes.

Animals
Male wild-type (WT) C57BL/6J mice and ob / ob mice were purchased from the Jackson Laboratory (Bar Harbor, ME). All mice Western diet, a model of diet-induced obesity, hepatic Sort1 protein was also reduced by more than 80%, but hepatic Sort1 mRNA was only reduced by ‫ف‬ 35% ( Fig. 1C, D ). To fi nd whether obesity and fatty liver is associated with decreased Sort1 expression in humans, we compared Sort1 levels in human liver samples from individuals with steatosis and obesity to that of controls. Results showed that Sort1 protein levels, but not mRNA, were markedly lower in the steatosis/obese group ( Fig. 1E, F ).
It was shown that increasing hepatic Sort1 expression in apobec1 Ϫ / Ϫ ; apoB transgenic mice that had elevated LDL-C signifi cantly reduced plasma total cholesterol, LDL-C and triglycerides, while siRNA-mediated knockdown of hepatic Sort1 in these mice raised plasma cholesterol and triglycerides ( 12 ). To test whether decreased hepatic Sort1 under obese and diabetic conditions might play a causative role in the development of hyperlipidemia, we increased hepatic Sort1 expression in both WT and ob / ob mice by adenovirus-mediated gene delivery. Interestingly, liver-specifi c Sort1 overexpression signifi cantly lowered plasma triglyceride and cholesterol levels in WT mice ( Table 1 ). In addition, Santa Clara, CA). Adenovirus was purifi ed from HEK293A cells by CsCl centrifugation. Adenovirus titer was determined with an Adeno-X rapid titer kit from Clontech (Mountain View, CA). Adenovirus was administered at 1 × 10 9 pfu/mouse via tail vein. Experiments were carried out seven days after injection.

Measurement of apoB secretion
Measurement of apoB secretion into culture medium was performed essentially the same as previously published ( 17 ), with minor modifi cations. Briefl y, primary mouse hepatocytes or HepG2 cells were infected with Ad-GFP or Ad-Sort1 at MOI of ‫ف‬ 5 for 24 h. Cells were then cultured in methionine/cysteine-free medium (Sigma, St. Louis, MO) for 15 min, followed by incubation with medium containing 100 Ci S-35 methionine/cysteine (EasyTag protein labeling mix, Perkin Elmer, Waltham, MA) for 20 min. Cells were washed three times with culture medium and chased for an additional 90 min, after which apoB in the culture medium was immunoprecipitated with apoB antibody and resolved on a 4-15% SDS-PAGE.
Dried gels were exposed to a phosphoimager, and labeled apoB was detected with a Typhoon imaging system. The apoB band intensity was normalized to total cellular protein. Each assay was performed in duplicate.

RNA isolation and quantitative real-time PCR
Total RNA was purifi ed with Tri-reagent (Sigma). Real-time PCR assays were performed with SYBR primers or Taqman primers/probe (Applied Biosystems, Foster City, CA) on an Applied Biosystems 7300 real-time PCR system. Amplifi cation of mt18s or Ubiquitin C (UBC) was used as an internal control. Relative mRNA expression was quantifi ed using the comparative CT (Ct) method and expressed as 2 Ϫ ⌬ ⌬ Ct .

Immunoblot
Cell or tissue protein samples were prepared in RIPA buffer followed by brief sonication. Protein concentrations were determined by a BCA assay kit (Rockford, IL). Equal amount of protein was used for SDS-PAGE and Western blotting.

Lipid analysis
Cholesterol, triglyceride, and free fatty acids were measured with colorimetric assay kits (Biovision, Milpitas, CA) as described previously ( 18 ).

Statistical analysis
Results were expressed as mean ± SE unless noted. Statistical analysis was performed by Student t -test or ANOVA. P < 0.05 was considered statistically signifi cant.

Hepatic Sort1 levels are markedly decreased in obese mice and steatotic livers from obese individuals
To investigate whether obesity, diabetes, and hepatic steatosis are associated with altered hepatic Sort1, we fi rst measured liver Sort1 expression in genetic obese ob/ob mice. Hepatic Sort1 protein levels were decreased by more than 90% in ob / ob mice as estimated by densitometry ( Fig. 1A ). Unexpectedly, Sort1 mRNA was modestly elevated in ob / ob mice ( Fig. 1B ). As controls, the mRNA and protein of microsome triglyceride transfer protein (MTTP) were higher in ob / ob mice ( Fig. 1A, B ), consistent with a previous report ( 19 ). In mice fed a high saturated fat-containing JNK or PKC signaling showed no effect on basal Sort1 protein or palmitate-induced Sort1 repression in HepG2 cells ( Fig. 3B ). Furthermore, treating HepG2 cells with C6ceramide at 10 M induced ERK phosphorylation and repressed Sort1, which was abolished by the ERK inhibitor U0126 ( Fig. 3C , upper panel). Similar results were obtained in primary mouse hepatocytes ( Fig. 3C , lower panel). To further test whether ERK signaling regulates Sort1 expression in vivo, we injected mice with U0126 and measured hepatic Sort1 protein after 6 h. Mice treated with U0126 showed a signifi cant increase in hepatic Sort1 increasing hepatic Sort1 levels in ob / ob mice was able to reduce plasma cholesterol and triglycerides to levels seen in WT control mice ( Table 1 ). We also found that Sort1 overexpression reduced apoB secretion from primary mouse hepatocytes and HepG2 cells (supplementary Fig. I), consistent with previous studies ( 12 ). Taken together, these results suggest that obesity and diabetes are associated with impaired hepatic Sort1 function, which may contribute to the development of diabetic hyperlipidemia.

Saturated fatty acid represses Sort1 protein, but not mRNA, in HepG2 cells and primary human hepatocytes
The differential expression of liver Sort1 mRNA and protein observed in obese mice and humans implies potential posttranslational downregulation of hepatic Sort1 protein. Increased circulating and tissue-saturated FFAs are thought to play critical roles in hepatic VLDL overproduction in obesity and diabetes ( 20 ). We tested the hypothesis that saturated FFAs downregulate hepatic Sort1 in obesity and diabetes. As shown in Fig. 2A , both palmitate and stearic acid, the two most common saturated fatty acids in mammals, repressed Sort1 protein in primary human hepatocytes. Similarly, palmitate treatment dosedependently inhibited Sort1 protein in HepG2 cells ( Fig. 2B ). In contrast, treating HepG2 cells with monosaturated fatty acid oleate did not repress Sort1 protein ( Fig. 2C ). Interestingly, palmitate treatment had no effect on Sort1 mRNA expression in primary human hepatocytes or in HepG2 cells ( Fig. 2D, E ), further implying a role of palmitate in mediating the posttranslational downregulation of Sort1 in obese mice. Palmitate is the main substrate in de novo biosynthesis of ceramides, which act as important downstream lipid mediators of elevated FFA effects on infl ammation and insulin resistance in diabetes ( 21 ). Treating HepG2 cells with a ceramide analog, C6-ceramide, dose-dependently repressed Sort1 protein without affecting Sort1 mRNA expression ( Fig. 2F, G ), suggesting that palmitate inhibition of Sort1 protein could be at least partially mediated by the production of ceramides.

Activation of ERK signaling downregulates hepatic Sort1 in diabetic mice
To further delineate the mechanisms mediating saturated FFA-induced Sort1 downregulation, we fi rst aimed to identify the downstream signaling pathway involved. Saturated FFAs have been shown to activate MAPK and PKC signaling ( 7,8 ). Consistently, palmitate treatment induced the phosphorylation of ERK, JNK, and PKC in HepG2 cells ( Fig. 3A ). Interestingly, blocking ERK signaling with U0126 signifi cantly blocked the repressive effect of palmitate on Sort1 protein ( Fig. 3B ). In contrast, blocking   ( Table 2 ).
To further test whether increased FFAs are associated with reduced hepatic Sort1 in vivo, we took advantage of the STZ-treated type I diabetic mice, which have increased circulating FFAs ( Table 2 ). STZ-treated mice also showed elevated plasma cholesterol and triglyceride levels ( Table  2 ). Interestingly, the hepatic Sort1 protein, but not mRNA, was markedly lower in STZ-treated mice over controls ( Fig. 4A , B ), and such decrease in Sort1 protein was associated with higher liver ERK phosphorylation ( Fig. 4C ). Hepatic phosphorylation of AKT was not significantly different between the two groups after overnight fasting ( Fig. 4C ). Analysis of hepatic lipid contents showed that  is well known that hepatic apoB production is mainly regulated by posttranslational presecretory degradation via both proteasomal-dependent and lysosomal-dependent pathways ( 22 ). Because insulin signaling was shown to promote proteasomal-dependent apoB degradation, hepatic apoB overproduction under diabetic conditions has mainly been attributed to impaired hepatic insulin resistance ( 22 ). As recent studies demonstrated a signifi cant impact of Sort1-mediated lysosomal-dependent apoB degradation on plasma cholesterol and triglyceride levels ( 12,14 ), impaired hepatic Sort1 function may also contribute significantly to higher apoB production in obesity and diabetes. It should be noted, however, that two independent studies found that whole-body knockout of sort1 in mice resulted in decreased hepatic apoB secretion and lower plasma cholesterol levels ( 13,14 ). These results are opposite to that from mice with liver-specifi c Sort1 overexpression or siRNA-mediated hepatic Sort1 knockdown, which showed an inverse relationship between hepatic Sort1 and plasma lipid levels ( 12 ). Although further studies are necessary to address such discrepancies from mechanistic perspectives, it is possible that complete and whole-body absence of Sort1 may result in different metabolic outcomes than those caused by variations in hepatic Sort1 levels. Besides results obtained from mouse models, only one study has so far demonstrated that the genetic variation that results in higher Sort1 mRNA expression in human livers is associated with lower plasma LDL-C ( 12 ). Here we show that Sort1 protein, but not mRNA, is signifi cantly lower in individuals with obesity and hepatic steatosis, which are often associated with higher hepatic apoB/VLDL production. We further showed in both normal mice and obese and diabetic mice that increasing Sort1 expression in the liver signifi cantly reduced plasma cholesterol and triglyceride levels. Our study, therefore, provides new evidence supporting an inverse relationship between hepatic Sort1 and plasma lipid levels, and it further suggests a potential role of Sort1 in the pathogenesis and/or progression of diabetic dyslipidemia. Increased circulating FFAs play a critical role in the pathogenesis of hepatic steatosis, insulin resistance, and infl ammation ( 23 ), all of which contribute to hepatic VLDL overproduction and hyperlipidemia. Our study suggests a novel molecular link between FFAs, ERK activation, and hepatic VLDL overproduction via hepatic Sort1 downregulation. On one hand, abnormal activation of MAPKs and PKC isoforms is associated with insulin receptor and insulin receptor substrate phosphorylation and inactivation, leading to insulin resistance ( 7,8 ); on the other hand, ERK activation downregulates hepatic Sort1. These changes presumably result in decreased apoB degradation in both proteasomes and lysosomes and, therefore, apoB overproduction and dyslipidemia. Consistent with these fi ndings, a large body of evidence has suggested that abnormal ERK activation, not only in the liver but also adipose and muscle, is critically involved in the development of obesity, insulin resistance, and altered lipid and glucose metabolism (24)(25)(26), whereas decreased ERK activation or genetic ERK1 inactivation is linked to improved metabolic

Palmitate-induced ERK activation promotes Sort1 protein degradation
To obtain further mechanistic insights, we tested whether palmitate represses Sort1 by regulating Sort1 protein stability. By treating cells with cyclohexamide (CHX) to block protein synthesis, we estimated that Sort1 protein has a half-life of approximately ‫ف‬ 8 h in HepG2 cells ( Fig. 5A ). Treating HepG2 cells with palmitate signifi cantly accelerated the degradation rate of endogenous Sort1 ( Fig. 5A ) as well as the degradation rate of ectopically expressed fl ag-tagged Sort1 ( Fig. 5B ). In addition, palmitate-induced Sort1 degradation was blocked by the ERK inhibitor U0126, supporting our hypothesis that activation of ERK results in Sort1 protein degradation ( Fig. 5C ). Palmitateinduced Sort1 degradation was also blocked by the proteasome inhibitor MG132 ( Fig. 5D ). The detection of poly-ubiquitinated Sort1 in an ubiquitination assay further suggests the possible involvement of proteasomaldependent degradation of Sort1 ( Fig. 5E ). On the other hand, treating HepG2 cells with the lysosome inhibitor chloroquine alone resulted in increased Sort1 degradation, which was not further affected by palmitate ( Fig. 5D ).

DISCUSSION
Results from this study suggests that FFA-mediated posttranslational downregulation of hepatic Sort1 may promote the development and progression of diabetic dyslipidemia, the underlying cause of increased CVD risk among obese and diabetic patients. Hepatic apoB production is an important determining factor for hepatic VLDL secretion and plasma LDL-C and triglyceride levels ( 22 ). It with higher circulating FFAs but lower hepatic tri glycerides and FFAs ( Table 2 ). These results suggest that downregulation of hepatic Sort1 in these mice was not a direct result of hepatic fat accumulation per se. It is reported that STZ-treated rats showed a marked increase in C16:0 esterifi ed ceramides in the circulation ( 30 ), and uptake of circulating LDL-ceramides is involved in infl ammation and insulin resistance ( 31 ). It is possible that increased hepatic uptake of circulating ceramides plays a role in hepatic Sort1 downregulation in STZtreated mice. Elevated circulating ceramides in obesity and type II diabetes may also play a role in hepatic Sort1 downregulation.
In this study, we provide evidence demonstrating that hepatic Sort1 levels cannot only be changed by genetic homeostasis ( 27,28 ). Palmitate is an essential substrate for de novo synthesis of ceramides, and it is well documented that oversupply of saturated FFAs, mainly palmitate, results in elevated ceramides in the circulation and insulinresistant tissues in diabetic humans and rodent models ( 21 ). Elevated ceramides are not only implicated in infl ammation and insulin resistance but also in the development of coronary artery disease ( 29 ). In an attempt to identify the downstream mediators of palmitate, we found that C6-ceramide treatment repressed Sort1 protein, but not mRNA, in HepG2 cells and mouse hepatocytes. These results indicate that it is possible that the palmitate inhibition of Sort1 in diabetic mice may be partially mediated by elevated circulating and tissue ceramides. Interestingly, in STZ-treated mice, lower hepatic Sort1 protein was associated variations that affect sort1 gene transcription but also by posttranslational mechanisms that affect Sort1 protein stability. We show that ERK-mediated Sort1 protein degradation may be a mechanism underlying the dissociation of hepatic Sort1 protein and mRNA in obese mice. Although it is still not fully clear how ERK activation results in Sort1 degradation, our results suggest the potential involvement of Sort1 ubiquitination and proteasome-dependent degradation. Interestingly, as we prepared this article, one study reported that cellular Sort1 turnover is regulated by Sort1 mono-ubiquitination and lysosomaldependent degradation in HeLa cells ( 32 ). In contrast, our results from HepG2 cells ( Fig. 5E ) showed a typical pattern of poly-ubiquitination of Sort1, which is more frequently associated with proteasomal-dependent protein degradation. In addition, our results in Fig. 5D showed that blocking lysosome activity by chloroquine itself resulted in Sort1 degradation. After traffi cking to the lysosome, Sort1 is retrograde transported back to Golgi apparatus via a retromer complex ( 11 ). It is possible that disruption of this process via lysosomal inhibition somehow leads to Sort1 protein degradation. However, our results do not rule out the possible lysosomal-dependent Sort1 turnover under normal conditions. Further study is underway to investigate the posttranslational regulation of Sort1 protein.
Our results showing decreased Sort1 protein in steatotic liver of obese individuals are important evidence of impaired hepatic Sort1 function in obese humans. However, we did not observe increased ERK activation in steatotic human livers as measured by ERK phosphorylation due to large variations (supplementary Fig. II). It should be noted that reduction of hepatic Sort1 likely resulted from chronic metabolic alterations; however, numerous known and unknown factors could acutely alter ERK signaling in human livers. This is very different from experimental conditions in which baseline levels are well controlled.
Increasing evidence suggests that saturated fatty acid accumulation plays a critical role in causing ER stress, which underlies many lipid disorders associated with diabetes and fatty liver diseases ( 33 ). Palmitate is a strong inducer of ER stress in hepatocytes ( 34 ). One recent study reports that sort1 gene transcription is repressed in obese mice by mechanisms involving activation of ER stress, mTORC1, and an ER stress-induced transcriptional factor ATF3 ( 35 ). Our results showed decreased hepatic Sort1 mRNA in Western diet-fed mice but not in ob / ob mice or in human steatotic livers. It is yet to be determined whether activation of ER stress may also account for posttranslational Sort1 downregulation in obesity and diabetes.
Despite the newly emerged role of liver Sort1 in regulating plasma lipid homeostasis in humans, experimental data on how Sort1 can be regulated is still very limited. Further insights on Sort1 regulation are important in improving our understanding of not only the physiological and pathological roles of Sort1 in lipid metabolism but also the potentials of targeting Sort1 to improve lipid homeostasis in obesity, diabetes, and fatty liver diseases.