Global inactivation of carboxylesterase 1 (Ces1/Ces1g) protects against atherosclerosis in Ldlr−/− mice

Atherosclerotic cardiovascular disease is a leading cause of death in the western world. Increased plasma triglyceride and cholesterol levels are major risk factors for this disease. Carboxylesterase 1 (Ces1/Ces1g) has been shown to play a role in metabolic control. So far, the role of mouse Ces1/Ces1g deficiency in atherosclerosis is not elucidated. We generated Ces1/Ces1g −/− mice. Compared to wild-type mice, Ces1/Ces1g −/− mice had reduced plasma cholesterol levels. We then generated Ces1g −/− Ldlr −/− double knockout (DKO) mice, which were fed a Western diet for 16 weeks. Compared to Ldlr −/− mice, DKO mice displayed decreased plasma cholesterol and TG levels and reduced atherosclerotic lesions. Interestingly, knockdown of hepatic Ces1/Ces1g in Apoe −/− mice resulted in hyperlipidemia and exacerbated Western diet-induced atherogenesis. Mechanistically, global inactivation of Ces1/Ces1g inhibited intestinal cholesterol and fat absorption and Niemann-Pick C1 like 1 expression, and increased macrophage cholesterol efflux by inducing ATP-binding cassette subfamily A member 1 (ABCA1) and ABCG1. Ces1/Ces1g ablation also promoted M2 macrophage polarization and induced hepatic cholesterol 7α-hydroxylase and sterol 12α-hydroxylase expression. In conclusion, global loss of Ces1/Ces1g protects against the development of atherosclerosis by inhibiting intestinal cholesterol and triglyceride absorption and promoting macrophage cholesterol efflux.

cholesterol absorption and inducing macrophage cholesterol efflux and bile acid synthesis. Similar to loss of hepatic Ces3/Ces1d function, inactivation of hepatic Ces1/Ces1g in Apoe −/− mice aggravated the development of atherosclerosis. Our data indicate that Ces1/Ces1g plays a critical role in the pathogenesis of atherosclerosis.

Results
Generation and characterization of Ces1/Ces1g −/− mice. We have recently shown that hepatic Ces1/ Ces1g plays a critical role in maintaining hepatic and plasma lipid homeostasis and is an important TG hydrolase in the liver 11 . To better understand the role of Ces1/Ces1g in peripheral tissues, we generated global Ces1/Ces1g −/− mice by replacing Ces1g exon 1 and exon 2 with an nLacZ-neomyclin cassette (Fig. 1A). The successful creation of the Ces1/Ces1g −/− mice was confirmed by Southern blotting, PCR genotyping and Western blotting (Fig. 1B-H). Other isoforms of Ces1 were not affected in the liver of Ces1/Ces1g −/− mice ( Fig. 1D; for primer sequences see Supplementary Table 1). LacZ staining of tissues isolated from Ces1 +/− mice indicated that Ces1 was expressed in a number of tissues, including brain, lung, liver, stomach, gallbladder, small intestine, colon, pancreas, spleen, kidney, white adipose tissue (WAT), adrenal and brown adipose tissue (BAT), but not in heart or skeletal muscle (Fig. 1E). Since LacZ staining was not quantitative, we also measured mRNA and protein levels of Ces1g. Results from RNA profiling showed that Ces1/Ces1g was most abundantly expressed in the liver and to a less extent in other tissues (see Supplementary Fig. 1A). Western blotting results showed that Ces1/Ces1g protein was expressed in the liver, intestine and gallbladder (Fig. 1F), consistent with a previous finding by Quiroga et al. 10 . Interestingly, we did not observe Ces1g protein expression in the lung, kidney, WAT or BAT (Fig. 1F), which was likely due to very low expression of Ces1g in these tissues. There was a significant reduction in the expression of Ces1d protein (the faster migrating band, according to Quiroga et al. 10 ) in the intestine and gallbladder (Fig. 1F). The reason for the reduced Ces1d expression in the intestine or gallbladder is unclear. However, we believe the reduction of Ces1d expression should be secondary to Ces1g ablation as Ces1d expression did not decrease in the liver or lung of Ces1g −/− mice (Fig. 1F). In addition, Ces1/Ces1g was also expressed in macrophages, as demonstrated by There was no gross difference in body weight, size or food intake on a chow diet (data not shown). Ces1/Ces1g −/− mice have reduced plasma cholesterol levels. When 8-weeks-old Ces1/Ces1g −/− mice and their control littermates (Ces1/Ces1g +/+ ) were fed a chow diet for 6 months, plasma TG levels were not significantly different between the two genotypes ( Fig. 2A), but plasma cholesterol levels were reduced by ~30% in Ces1/Ces1g −/− mice (Fig. 2B). Similar data were obtained when 8-weeks-old Ces1/Ces1g −/− mice and their control littermates were fed a Western diet for 16 weeks (Fig. 2C and D). Analysis of plasma lipoprotein profile by fast protein liquid chromatography (FPLC) indicated that Ces1/Ces1g −/− mice had reduced LDL-C and HDL-C levels ( Fig. 2E). In the liver, TG or cholesterol levels were not different between the two genotypes when fed a chow or Western diet (see Supplementary Fig. 2A-D). Oil red O staining showed no difference in neutral lipid accumulation (Fig. 2F). Despite unchanged hepatic lipid levels, Ces1/Ces1g −/− mice had significantly reduced plasma AST and ALT levels (data not shown) and hepatic mRNA levels of inflammatory markers (Tnfα, Il6 and Il1β) (Fig. 2G). In addition, Western diet-fed Ces1/Ces1g −/− mice had similar body weight gain and body fat content compared to Ces1/Ces1g +/+ mice (see Supplementary Fig. 2E and F).
Liver-specific Ces1/Ces1g deficiency exacerbates atherosclerosis in Apoe −/− mice. Ces1/Ces1g is abundantly expressed in the liver (Fig. 1). To understand how global inactivation of CES1 ameliorates atherosclerosis, we knocked down hepatic Ces1/Ces1g expression in Apoe −/− mice using an adenovirus expressing Ces1 shRNA (Ad-shCes1). Previously, we showed that Ad-shCes1 can efficiently knock down hepatic Ces1 mRNA and protein levels by ~95% 11 . The selection of Apoe −/− mice is because these mice can develop spontaneous atherosclerosis even on a regular chow diet. Three weeks after Ad-shCes1 infection, hepatic Ces1/Ces1g mRNA and protein levels were reduced by ~90% (Fig. 4A) whereas intestinal CES1g protein levels were unchanged ( Supplementary Fig. 3). We did not observe fat malabsorption (data not shown). Similar to what we have reported previously 11 , knock down of

Ces1/Ces1g inactivation induces hepatic CYP7A1 and CYP8B1 expression and bile acid synthesis.
In the liver, cholesterol is converted to bile acids via cholesterol 7α-hydroxylase (CYP7A1) and sterol 12α-hydroxylase (CYP8B1) or secreted directly to the bile. Transgenic expression of CYP7A1 in the liver reduces both HDL-C and non-HDL-C levels 14 and protects against atherosclerosis 15 . In the liver of Ces1/Ces1g −/− mice, both Cyp7a1 and Cyp8b1 mRNA levels were induced by ~1.8 fold (Fig. 6A), which was accompanied by a 99% reduction in fibroblast growth factor 15 (Fgf15) in the intestine (Fig. 6A). We did not observe much change in mRNA levels of small heterodimer partner (Shp) (Fig. 6A), scavenger receptor class B type 1 (SR-BI), ATP-binding cassette subfamily G member 5 (Abcg5), Abcg8, HMG-CoA reductase (Hmgcr), Apoe or Ldlr (see Supplementary Fig. 4). Gain-and loss-of-function studies have demonstrated that intestinal FGF15 is an important repressor of hepatic CYP7A1 and CYP8B1 expression 16,17 . Consistent with these data, bile acid (BA) levels in gallbladder (Fig. 6B) and intestine (Fig. 6C) were increased whereas BA levels in the liver were unchanged (Fig. 6D). As a result, total BA pool size was increased (Fig. 6E). In addition, biliary cholesterol levels were also increased (Fig. 6F). Similarly, mRNAs levels of hepatic Cyp7a1 and Cyp8b1 were elevated and intestinal Fgf15 level was repressed in DKO mice compared with control littermates (Fig. 6G). In addition, BA levels in gallbladder or intestine were increased in DKO mice, resulting in increased total BA pool size (Fig. 6H). These data suggest that the induction of Cyp7a1 and Cyp8b1 may contribute to hypocholesterolemia in Ces1/Ces1g −/− mice.
Cholesterol efflux is increased in Ces1/Ces1g deficient macrophages. Macrophage cholesterol efflux is the first step in reverse cholesterol transport (RCT). Next, we determined whether loss of Ces1/Ces1g Representative en face aorta images are shown (H) and aortic lesion size was quantified (I). Representative aortic root images are also shown (J) and lesion size was quantified (K). *P < 0.05, **P < 0.01.
in macrophages affected cholesterol efflux. Peritoneal macrophages were isolated from Ces1/Ces1g +/+ mice and Ces1/Ces1g −/− mice and then treated with or without acetylated LDL (Ac-LDL). Ces1/Ces1g inactivation resulted in a 1.7-fold increase in cholesterol efflux to ApoA-I and this increase was further enhanced in the presence of Ac-LDL (Fig. 7A). In addition, Ces1 inactivation caused a 2.1-fold increase in cholesterol efflux to HDL (Fig. 7B).
In addition to human CES1, mouse hormone sensitive lipase (Hsl) 7,18 and neutral cholesterol ester hydrolase 1 (NCEH1) 8 are reported to promote hydrolysis of cholesterol esters in macrophages. ATP-binding cassette subfamily A member 1 (ABCA1) and ABCG1 are the major transporters that efflux free cholesterol to ApoA-I or HDL, respectively 19 . Ces1/Ces1g inactivation significantly increased Hsl but not Nceh1 mRNA levels (Fig. 7C). Ces1/Ces1g inactivation in macrophages also caused 2-and 3-fold increase in Abca1 and Abcg1 mRNA levels, respectively (Fig. 7C). In contrast, Ces1/Ces1g inactivation had no much effect on scavenger receptor class A (Sra), fatty acid translocase (Cd36) or acetyl-Coenzyme A acetyltransferase 2 (Acat2) (Fig. 7C and see Supplementary  Fig. 5A-C). T-0901317, a liver X receptor agonist, drastically induced Abca1 (Fig. 7D) and Abcg1 (Fig. 7E) mRNA levels in wild-type macrophages, and this induction was further increased in Ces1/Ces1g −/− macrophages ( Fig. 7D and E), suggesting that Ces1/Ces1g inactivation may lead to increased production of endogenous LXR agonists (oxysterols). In the absence of Ac-LDL treatment, the intracellular free cholesterol and total cholesterol contents were decreased in Ces1/Ces1g −/− macrophages compared with wild-type macrophages, but this decrease became less pronounced when macrophages were treated with Ac-LDL ( Fig. 7F and G). Taken together, these data suggest that Ces1/Ces1g inactivation increases macrophage cholesterol efflux by inducing both ABCA1 and ABCG1.

Discussion
The studies on the role of the carboxylesterase family in regulating lipid metabolism are challenged by the structural and functional similarities shared by their family members 9 . Mouse Ces3/Ces1d and Ces1/Ces1g share high amino acid identities with human CES1, and both Ces1d and Ces1g have been reported to play roles in lipid metabolism 11,21 . Ces3/Ces1d and Ces1/Ces1g share 76% amino acid identity and have the similar protein size, but display different tissue expression profiles. Global or liver-specific loss of Ces3/Ces1d/TGH is shown to improve blood lipids by attenuating VLDL secretion 22,23 . Ces3/Ces1d/TGH knockout mice are reported to have increased energy metabolism and glucose tolerance 23 . In contrast, loss of Ces1/Ces1g/Es-x is reported to induce obesity and hyperlipidemia 10 . In addition, loss of Ces3/Ces1d/TGH attenuates atherosclerosis in Ldlr −/− mice, but liver specific loss of Ces3/Ces1d/TGH promotes atherosclerosis in Ldlr −/− mice 12,13 . Similar results are found in this study when we inactive Ces1/Ces1g in Ldlr −/− mice or Apoe −/− mice. In this report, we provide the first evidence demonstrating that global Ces1/Ces1g inactivation markedly attenuates the development of atherosclerosis by reducing intestinal cholesterol absorption, promoting macrophage cholesterol efflux and M2 macrophage polarization, and inducing bile acid synthesis (Fig. 8F) whereas loss of hepatic Ces1/Ces1g aggravates the development of atherosclerosis.
Previously, Lehner group show that female, 5-6 months old Ces1/Ces1g/Es-x −/− mice are obese 10 . We do not see any difference in body weight or adiposity between male Ces1 +/+ mice and male Ces1g −/− mice that are fed either a chow diet or high fat diet. We realize that we may have used mice with different sex (male vs female) and background. Lehner group also reports that Ces1/Ces1g/Es-x −/− mice have increased secretion of chylomicron and decreased chylomicron clearance 24 . We find that loss of Ces1/Ces1g −/− leads to decreased TG and cholesterol absorption in the intestine. Interestingly, in our Ces1g −/− mice intestinal Ces1d expression is also reduced, which may contribute to the reduced fat absorption and overall lipid homeostasis. In our global Ces1/Ces1g −/− mice, we do not see any change in the levels of hepatic neutral lipids. Previously, we show that liver-specific knockdown of Ces1/Ces1g causes hepatic triglyceride accumulation and increased VLDL secretion and plasma lipids 11 . In global Ces1/Ces1g −/− mice, the reduction in intestinal fat and cholesterol absorption may be sufficient to counteract the increased lipogenic activity in the liver, thus resulting in reduced plasma cholesterol levels and unchanged hepatic lipids.
Reverse cholesterol transport begins with macrophage cholesterol efflux to the acceptors ApoA-I or HDL by ABCA1 or ABCG1. Prior to cholesterol efflux, cholesterol esters are hydrolyzed by cholesterol ester hydrolase. Loss-of-function studies have established both HSL and NCEH1 as important cholesterol ester hydrolases in macrophages 7,8 . Our data show that Ces1/Ces1g is present in peritoneal macrophages. So far, it is unclear whether mouse Ces1/Ces1g has cholesterol ester hydrolase activity. Interestingly, loss of Ces1/Ces1g in macrophages induces Hsl, Abca1 and Abcg1 expression. HSL over-expression in macrophages is reported to increase cholesterol ester hydrolysis 18 . The induction of Abca1 and Abcg1 suggests that intracellular oxysterols may be increased in Ces1/Ces1g −/− macrophages. Indeed, LXR activation is further enhanced in Ces1/Ces1g −/− macrophages. In the future, we will determine whether oxysterols and other lipid metabolites are changed in Ces1/ Ces1g −/− macrophages.
In addition to regulating cholesterol efflux, our data also show that Ces1/Ces1g inactivation promotes M2 macrophage polarization. M2 macrophages are anti-inflammatory and play an important role in regression of atherosclerosis 20 . Although the mechanism underlying the regulation of macrophage polarization remains to be determined, our data indicate that loss of Ces1/Ces1g in macrophages has two beneficial effects − promoting cholesterol efflux and inhibiting inflammation.
Liver is one of the most important organs regulating the pathogenesis of atherosclerosis. Our data show that Ces1/Ces1g knockdown in the liver promotes the development of atherosclerosis in Apoe −/− mice whereas global loss of Ces1/Ces1g protects against atherosclerosis in Ldlr −/− mice. Over-expression of hepatic Ces1g produces less atherogentic lipoproteins 11,25 , which may account for the atheroprotective effect of hepatic Ces1g. Since we used mice with different genetic background (Apoe −/− mice versus Ldlr −/− mice), we are not able to reach a firm conclusion at this point that hepatic Ces1/Ces1g deficiency is not part of the mechanism underlying the atheroprotective effect found in DKO mice. However, Apoe −/− mice and Ldlr −/− mice are both widely used for atherosclerosis study and the results achieved using these mouse lines are often very similar. The reduced intestinal TG and cholesterol absorption plays an important role in atheroprotection in DKO mice. In addition, hepatic Cyp7a1 and Cyp8b1 are induced in Ces1/Ces1g −/− mice likely as a result of inhibition of intestinal Fgf15. Since Cyp7a1 and Cyp8b1 are key enzymes in cholesterol catabolism in the liver, the induction of Cyp7a1 and Cyp8b1 may also contribute to the atheroprotective effect of Ces1/Ces1g inactivation.
In summary, we demonstrate that global inactivation of Ces1/Ces1g ameliorates the development of atherosclerosis by inhibiting intestinal cholesterol absorption, promoting macrophage cholesterol efflux and M2 macrophage polarization, and inducing bile acid production. In contrast, inactivation of hepatic Ces1/Ces1g aggravates the development of atherosclerosis likely by producing atherogenic lipoproteins 11 .
RNA isolation and quantitative real-time PCR. Total RNA was isolated using TRIzol Reagent (Life Technologies, NY). mRNA levels were determined by quantitative reverse-transcription polymerase chain reaction (qRT-PCR) on a 7500 real-time PCR machine from Applied Biosystems (Foster City, CA). Relative mRNA levels were calculated using the comparative cycle threshold (Ct) method and were normalized to the values of 36B4 mRNA levels.

Lipid analysis.
Approximately 100 mg liver was homogenized in methanol and lipids were extracted in chloroform/methanol (2:1 v/v). Hepatic and plasma triglyceride and cholesterol levels were quantified using Infinity reagents from Thermo Scientific (Waltham, MA). For analyzing plasma lipoprotein levels, FPLC was performed as described 11 . In brief, 100 μL plasma was injected into the BioLogic DuoFlow QuadTec 10 system (Bio-Rad, Hercules, CA). Lipoproteins were run at 0.5 mL/min in a buffer containing 0.15 M NaCl, 0.01 M Na 2 HPO4, 0.1 mM EDTA, pH7.5, and separated on a Superose 6 10/300 GL column (GE Healthcare). A 500 μL sample per fraction was collected for analyzing lipid levels using Infinity reagents. β-galactosidase (LacZ) staining. Peritoneal macrophages or tissues were washed with PBS, fixed with a fixative solution containing 0.5% glutaraldehyde, washed with PBS and stained with a solution containing 1 mg/ ml X-gal at 37 °C.
Peritoneal macrophage isolation. Wild-type and Ces1/Ces1g −/− mice were injected with 1 mL of 3% (w/v) brewer thioglycollate medium into their peritoneal cavity. Four days later, mice were euthanized. The outer skin of the peritoneum was cut with gentle and pulled back to expose the inner skin lining the peritoneal cavity. Dulbecco's Modified Eagle Medium (DMEM) medium with 10% fetal bovine serum (FBS) was injected into the peritoneal cavity using a 21.5 G needle, followed by gently massage of the peritoneum to dislodge any attached cells into the medium. The cells were collected into a 5-ml syringe. DMEM medium was injected 1-2 more time to retrieve more cells. Cell suspension was spun at 1500 rpm, resuspended and cultured in 6-well plates.
Intestinal fat absorption. After overnight fasting, mice were injected intravenously with 500 mg/kg of Triton WR1339. Immediately after the Triton injection, mice were given an intragastric 200 μL olive oil containing 7 μCi of [ 3 H]triolein (Perkin Elmer). Blood samples were drawn via retro-orbital bleeding at indicated time points after administration of [ 3 H]triolein. The amount of 3 H in plasma was determined using a liquid scintillation counter.

Intestinal cholesterol absorption.
Intestinal cholesterol absorption was determined by a dual-isotope plasma ratio method 27 . Briefly, mice were intravenously injected with 2.5 μCi 3 H-cholesterol in Intralipid (Sigma, St Louis, MO), immediately followed by oral gavage of 1 μCi 14 C-cholesterol in median-chain triglycerides (MCT oil, Mead Johnson, Evansville, IN). Mice were returned to cage with free access to food and water. After 72 h, blood samples were collected and the radioactivity of 14 C and 3 H were determined by scintillation counting. Intestinal cholesterol absorption was determined as the ratio of 14 C/ 3 H in 1 ml of plasma.
Cholesterol efflux. Cholesterol efflux was performed as described 28,29 . Peritoneal macrophages from wild-type and Ces1 −/− mice were isolated and incubated in DMEM containing 0.2% bovine serum albumin (BSA). Cells were then incubated with acetylated LDL (50 μg/mL) and [ 3 H] cholesterol (1 μCi/mL) for 24 hours. After 24 h, cells were washed twice with PBS, and further incubated in DMEM containing 0.2% BSA for 2-4 hours. Cholesterol efflux was performed in the presence of fresh DMEM containing 0.2% BSA, 50 μg/mL of HDL, ApoA-I or Ac-LDL. After 4-h incubation, the radioactivity in the medium and the cell lysate was measured. The percent efflux was calculated as (medium dpm)/(cell dpm + medium dpm) × 100%.
Bile acid analysis. Bile acids in gallbladder, liver and intestine were extracted with 95% ethanol overnight, followed by 80% ethanol for 2 hours and methanol/chloroform (2:1) for 2 hours at 50 °C. Total bile acids were determined using a bile acid assay kit (Diazyme, San Diego, CA).

Body composition analysis.
Body fat mass and lean mass were measured by Echo-MRI (EchoMRI, LLC, Houston, TX).
Atherosclerosis study. The aorta, including the ascending arch, thoracic and abdominal segments were dissected, gently cleaned of the adventitia and stained with Oil Red O as described previously 30,31 . In addition, the aortic roots were collected from the base of the heart including the atria and embedded in optimal cutting temperature (OCT) compound. Sections (5 μm) were obtained every 50 μm from the base of the aortic leaflets to 400 μm above. After staining with Oil Red O, images were captured with a microscope, and the lesion area for each aortic ring was analyzed using Image Pro (Media Cybernetics).
Statistical Analysis. The data were analyzed using unpaired Student t test and ANOVA (GraphPad Prisim, CA). All values were expressed as mean ± SEM. Differences were considered statistically significant at P < 0.05. Data availability. All the data are available upon request.