Deletion of CGI-58 or adipose triglyceride lipase differently affects macrophage function and atherosclerosis[S]

Cellular TG stores are efficiently hydrolyzed by adipose TG lipase (ATGL). Its coactivator comparative gene identification-58 (CGI-58) strongly increases ATGL-mediated TG catabolism in cell culture experiments. To investigate the consequences of CGI-58 deficiency in murine macrophages, we generated mice with a targeted deletion of CGI-58 in myeloid cells (macCGI-58−/− mice). CGI-58−/− macrophages accumulate intracellular TG-rich lipid droplets and have decreased phagocytic capacity, comparable to ATGL−/− macrophages. In contrast to ATGL−/− macrophages, however, CGI-58−/− macrophages have intact mitochondria and show no indications of mitochondrial apoptosis and endoplasmic reticulum stress, suggesting that TG accumulation per se lacks a significant role in processes leading to mitochondrial dysfunction. Another notable difference is the fact that CGI-58−/− macrophages adopt an M1-like phenotype in vitro. Finally, we investigated atherosclerosis susceptibility in macCGI-58/ApoE-double KO (DKO) animals. In response to high-fat/high-cholesterol diet feeding, DKO animals showed comparable plaque formation as observed in ApoE−/− mice. In agreement, antisense oligonucleotide-mediated knockdown of CGI-58 in LDL receptor−/− mice did not alter atherosclerosis burden in the aortic root. These results suggest that macrophage function and atherosclerosis susceptibility differ fundamentally in these two animal models with disturbed TG catabolism, showing a more severe phenotype by ATGL deficiency.

Female mice were either fed a standard chow diet [containing 4% fat and 21% protein (R/M H; Ssniff, Soest, Germany)], challenged with a Western type diet (WTD) [TD88137mod; 21% fat, 0.2% cholesterol (Ssniff)] or a high-fat/high-cholesterol diet (HF/HCD) (E15126-34 EF R/M; 30% fat, 1% cholesterol) for 10-30 weeks starting at the age of 4-6 weeks. Mice were kept with water ad libitum on a regular light-dark cycle (12 h light, 12 h dark) in a clean environment. Body weights were measured weekly and plasma lipid parameters once a month.
For atherosclerosis studies using antisense oligonucleotide (ASO)-mediated knockdown of CGI-58, 6-week-old male LDLR Ϫ / Ϫ mice were fed a diet enriched in 0.2% (w/w) cholesterol and 20% of energy as lard for 16 weeks in conjunction with weekly injections (50 mg/kg) of either a nontargeting control ASO or CGI-58 ASO, as previously described ( 9 ). Plasma samples were collected by submandibular vein puncture at baseline (chow-fed animals, 6 weeks of age), and after 4, 8, and 16 weeks of diet and ASO treatment for subsequent lipid and lipoprotein analyses.
To assess in vitro lipopolysaccharide (LPS)-induced acutephase response, macrophages were treated with saline (control) or LPS (100 ng/ml) for 16 h. IL-6 concentrations in supernatants were determined by ELISA (Enzo Life Sciences, Lausen, Switzerland). hydrolase for the initial and rate-limiting step in lipolysis ( 2 ). Lipolysis has been extensively studied in adipocytes, where under basal conditions CGI-58 binds to the surface of lipid droplets through interaction with perilipin1. Hormonal stimulation of lipolysis leads to phosphorylation of perilipin1 and hormone-sensitive lipase, resulting in the release of CGI-58 from perilipin1 to interact with and activate ATGL, which then converts TG to diacylglycerol and FA ( 3 ). Both mice and humans affected with ATGL or CGI-58 defi ciency suffer from systemic TG accumulation, a condition called neutral lipid storage disease (NLSD) in humans ( 4 ). Of note, specifi c phenotypical alterations are observed depending on whether ATGL or CGI-58 is defective. The most apparent difference is the severe epidermal skin defect observed in mice and humans with CGI-58 defi ciency ( 2,5 ), which is absent in both species lacking ATGL ( 6,7 ). This fi nding resulted in different classifi cations of the respective human diseases, namely NLSD with myopathy in ATGL defi ciency ( 6 ), whereas CGI-58 deficiency leads to NLSD with ichthyosis ( 8 ). Several ongoing studies using tissue-specifi c CGI-58 and ATGL-defi cient ( Ϫ / Ϫ ) mice aim at studying the proteins' shared and individual contribution to lipid metabolism. In addition, functional differences observed in hepatocytes between mice with CGI-58 and ATGL defi ciency argue for an ATGLindependent function of CGI-58 in this cell population ( 9,10 ) and maybe other tissues as well ( 11 ).
Because CGI-58 Ϫ / Ϫ mice die shortly after birth due to a severe skin barrier defect ( 5 ), we generated myeloidspecifi c CGI-58 (macCGI-58) Ϫ / Ϫ mice to investigate the consequences of CGI-58 defi ciency in macrophages. In the present study, we examined: i ) whether CGI-58 Ϫ / Ϫ macrophages mimic the TG accumulation phenotype observed in ATGL Ϫ / Ϫ macrophages; ii ) whether CGI-58 defi ciency affects macrophage function; and iii ) whether the altered phenotype culminates in increased atherosclerosis susceptibility in macCGI-58/ApoE-double KO (DKO) animals. We have previously shown that loss of ATGL in macrophages affects macrophage phenotype and function, such as TG-rich lipid droplet accumulation, increased apoptosis ( 12 ) and endoplasmic reticulum (ER) stress ( 13 ), reduced migration ( 14 ), and decreased phagocytosis ability ( 15 ). In addition, transplantation of ATGL Ϫ / Ϫ bone marrow into LDL receptor (LDLR) Ϫ / Ϫ mice revealed that the lack of ATGL in immune cells attenuates atherosclerosis susceptibility ( 16 ). Being the coactivator of ATGL, we predicted that the absence of CGI-58 in macrophages leads to TG-rich lipid droplet accumulation. We hypothesized that loss of the ATGL coactivator CGI-58 in myeloid cells affects macrophage function in vitro and in vivo and impacts atherosclerosis susceptibility.

LPL activity in macrophages
Macrophages were incubated in 6-well plates with 300 l medium, 2% FA-free BSA (Sigma-Aldrich, St. Louis, MO), and 2 units/ml of heparin for 1 h at 37°C under continuous shaking. For the substrate preparation per sample, 0.6 Ci [ 3 H]triolein, 920 ng glycerol trioleate, and 0.1% Triton X-100 in chloroform were evaporated under a stream of nitrogen. Forty microliters of 1 M Tris-HCl (pH 8.6) and 80 l ddH 2 O were added, and the mixture was sonicated six times (1 min on and 1 min off) on ice. Then 40 l of heat-inactivated human serum containing ApoC-II as activator (obtained from a pool of donors, heated at 50°C for 1 h, and stored at 20°C) and 40 l of 10% FA-free BSA were added to the substrate. Analysis was performed as previously described ( 15 ).

Real time PCR
Total RNA from macrophages was isolated using a PerfectPure RNA cultured cell kit (5Prime, Hamburg, Germany). RNA concentrations were measured at 260 nm on a NanoDrop instrument (Thermo Scientifi c, Wilmington, DE). Two micrograms of total RNA were reverse transcribed by using the high capacity cDNA reverse transcription kit (Applied Biosystems, Foster City, CA). Quantitative real time PCR was performed on a LightCycler 480 (Roche Diagnostics, Rotkreuz, Switzerland) using the Quan-tifastTM SYBR ® Green PCR kit (Qiagen, Hilden, Germany). Amplifi cation of murine hypoxanthine-guanine phosphoribosyltransferase (HPRT) as housekeeping gene was performed on all samples as internal controls for variations in mRNA amounts. Expression profi les and associated statistical parameters were determined using the public domain program Relative Expression Software Tool-REST 2008 ( 22 ). Primer sequences are listed in the supplementary material.
For the studies using ASO-mediated knockdown, elicited peritoneal macrophages were collected 4 days after injection of 1 ml of 10% thioglycolate into the peritoneal cavities of C57BL/6 mice that had been treated with ASOs and fed a chow diet for 6 weeks, as previously described ( 19 ). Following 2 h of culture, nonadherent cells were removed by washing three times with PBS, and remaining adherent macrophages were harvested for Western blotting using methods previously described ( 9 ).

Plasma lipid parameters
Blood was collected from 12 h-fasted mice or from 12 hfasted/2 h-refed mice, and plasma was prepared by centrifugation at 5,200 g for 7 min at 4°C. Plasma TG, total cholesterol (TC), free cholesterol (FC), and nonesterifi ed FA concentrations were measured enzymatically by commercially available kits (DiaSys, Holzheim, Germany; Wako Chemicals GmbH, Neuss, Germany). For atherosclerosis studies using ASOmediated knockdown of CGI-58, total plasma concentrations of TC and TG were measured enzymatically by commercially available kits (Wako Chemicals, Richmond, VA). In addition, plasma lipoproteins were separated by fast protein liquid chromatography, and cholesterol concentrations in lipoprotein fractions were measured using an enzymatic assay as previously described ( 19 ).

Lipid parameters in macrophages
Macrophages were plated for 2 h in serum-free DMEM. After washing the cells three times with PBS, lipids were extracted with 2 ml hexane:isopropanol (3:2, v:v) for 1 h at 4°C. One hundred microliters of 1% Triton X-100 in chloroform were added and the lipid extract was dried under a stream of nitrogen. The samples were dissolved in 100 l ddH 2 O for 15 min at 37°C in a water bath. TG, TC, and FC concentrations were measured enzymatically by using 30 l of the sample with the above mentioned kits. The readings were normalized to protein concentrations. Protein was quantitated using a Lowry assay (Bio-Rad Laboratories, Hercules, CA) after dissolving the proteins of cells in 2 ml NaOH (0.3 M) for 2 h at room temperature.

Nile Red staining and fl uorescence microscopy
Macrophages were plated on chamber slides in DMEM containing 10% LPDS and 1% penicillin/streptomycin for 24 h. Cells were washed three times with PBS and fi xed with 10% formalin (30 min). Lipid droplets were visualized after Nile Red staining (2.5 g/ml) by confocal laser scanning microscopy using an LSM 510 META microscope system (Carl Zeiss GmbH, Vienna, Austria). Pictures (×63 magnifi cation) were taken at excitation 543 nm and signals were recorded using a 560 nm long pass fi lter.

TG and CE hydrolase activity assays
Macrophages were lysed with 100 l of lysis buffer [100 mM potassium phosphate, 250 mM sucrose, 1 mM EDTA, 0.1 mM DTT (pH 7)], sonicated on ice twice for 10 s with 10 s interval, and protein concentrations were measured using a Lowry assay (BioRad Laboratories). The TG substrate contained 17 nmol via a cannula in the left ventricular apex. Mice were perfused with 10% formalin (Carl Roth GmbH, Vienna, Austria) for 15 min. After fi xing the hearts in 10% formalin, serial sections (8 m) were cut (HM 560 Cryo-Star; Microm International GmbH, Walldorf, Germany). Images of the atherosclerotic lesion areas in Oil Red Ostained (Sigma-Aldrich) sections were taken with ScanScope T3 whole slide scanner (Aperio Technologies, Bristol, UK). Plaque areas were quantitated by ImageJ software. Mean lesion area was calculated from 10 consecutive Oil Red O-stained sections, starting at the appearance of the tricuspid valves. Sections were stained immunohistochemically for the presence of macrophages using a monoclonal rat anti-mouse Moma-2 antibody (1:600) (Acris, Hiddenhausen, Germany), as well as for collagen content using Masson's trichrome staining kit (Sigma-Aldrich). For en face analysis in macCGI-58/ ApoE-DKO mice, aortas were dissected and plaques were stained with Oil Red O as described recently ( 23 ). Images were analyzed using ImageJ software.
For studies in ASO-treated mice, hearts and aortae were carefully separated and hearts were immediately slow frozen in OCT for cross-sectional lesion analysis of the aortic sinus. Whole aortae were fi xed in 10% neutral buffered formalin for subsequent en face analysis. For atherosclerosis quantifi cation in the aortic sinus, histological cross-sections were stained with Oil Red O. Images were captured using a Leica DMR microscope (W. Nuhsbaum Inc., McHenry, IL) equipped with a Q imaging Retiga EX camera. Images were analyzed using Image-Pro Plus 7.0 (MediaCybernetics, Rockville, MD). Additionally, en face lesion area was determined for the whole aorta as described previously ( 19 ).

Cellular cholesterol effl ux
Macrophages were incubated with 50 g acLDL [preloaded with 0.5 Ci/ml [ 3 H]cholesterol (ARC Inc., St. Louis, MO)] and 30 g/ml nonlabeled cholesterol in DMEM/0.2% FA-free BSA for 32 h at 37°C. After washing the cells twice with PBS, the cells were cultivated for 16 h in equilibration medium (DMEM/0.2% FA-free BSA). We determined cholesterol effl ux after incubating the cells in DMEM/0.2% FA-free BSA in the absence or presence of 15 g/ml ApoA-I (Calbiochem, La Jolla, CA) or 100 g/ml HDL 3 . Radioactivity in 80 l medium and in the cells was measured by scintillation counting after 1, 3, 6, and 9 h of incubation. Cholesterol effl ux is expressed as the percentage of total cell [ 3 H]cholesterol present in the medium after 1, 3, 6, and 9 h. Basal effl ux in the absence of ApoA-I and HDL 3 was subtracted from the data shown.

Electron microscopy
Macrophages were cultured on an Aclar fi lm and fi xed in 2.5% (w/v) glutaraldehyde and 2% (w/v) formaldehyde in 0.1 M phosphate buffer (pH 7.4, 2 h), postfi xed in 2% (w/v) osmium tetroxide (2 h) at room temperature, dehydrated in graded series of ethanol, and embedded in a TAAB epoxy resin.
Ultrathin sections (75 nm) were cut with a Leica UC 7 Ultramicrotome and stained with lead citrate (5 min) and with uranyl acetate (15 min). Images were taken using a FEI Tecnai G2 20 transmission electron microscope (FEI, Eindhoven, The Netherlands) with a Gatan ultrascan 1000 CCD camera. Acceleration voltage was 120 kV.

Mitochondrial respiration measurement
Macrophages were plated in XF96 polystyrene cell culture microplates (Seahorse Bioscience ® , North Billerica, MA) at a density of 60,000 cells per well. After 24 h, cells were washed and preincubated for 30 min in XF assay medium supplemented with sodium pyruvate (1 mM) with or without glutamine (2 mM) and glucose (25 mM) at 37°C in a nonCO 2 environment. The oxygen consumption rate (OCR) was subsequently measured every 7 min using an XF96 extracellular fl ux analyzer (Seahorse Bioscience ® ). A standard protocol with 15 min basal measurement followed by 10 M oligomycin, addition of 0.3 M carbonyl cyanide p -trifl uoromethoxyphenylhydrazone (FCCP), and 2.5 M antimycin A was performed. Oxygen consumption was either normalized to protein content (pmol O 2 /min × g protein) or expressed as a percentage of the maximal mitochondrial respiration in the presence of 0.3 M FCCP.

Phagocytosis of fl uorescein-labeled Escherichia coli particles
Macrophages were plated in black 96-well Clear plates (Greiner Bio-One GmbH, Solingen, Germany). After 24 h of preincubation in DMEM, 10% LPDS, and 25 mM glucose, cells were incubated in DMEM, 10% LPDS, and 0, 6, or 25 mM glucose for 1 h, respectively. Cells were washed and incubated with 100 l of fl uorescein-labeled E. coli BioParticles (Vybrant TM phagocytosis assay, Molecular Probes, Invitrogen; suspended in Hanks' balanced salt solution; 2 h). The suspension was removed and subsequently 100 l of trypan blue was added (1 min) to quench the extracellular probe. After aspiration of trypan blue, the fl uorescence was measured at 484 nm (excitation) and 535 nm (emission) on a Victor 1420 multilabel counter (PerkinElmer Life Sciences, Turku, Finland). Fluorescence was normalized to the protein content of each well.
To analyze phagocytosis in vivo, mice were injected intraperitoneally with 200 l of fl uorescein-labeled E. coli BioParticles suspended in Hanks' balanced salt solution. After 2 h, macrophages were collected by fl ushing the peritoneal cavity with 10 ml PBS containing 1 mM EDTA and incubated in DMEM containing 25 mM glucose and 10% LPDS for 90 min. The cells were washed three times with PBS, and fl uorescence was measured before and after adding trypan blue to obtain total and intracellular fl uorescence, respectively. Experimental readings were normalized to protein content.

Apoptosis assay
Apoptosis was assayed by annexin V and propidium iodide (PI) costaining (Annexin-V-Fluor staining kit; Roche, Vienna, Austria). Two hundred thousand cells were washed twice with 200 l PBS; 50 l staining buffer was added and cells were incubated for 10 min. Macrophages were immediately analyzed on a FACScalibur fl ow cytometer (BD Biosciences, San José, CA).

Glucose tolerance test
Animals were fasted for 6 h (6 AM to 12 PM) with free access to drinking water . Blood was taken from the tail vein before and 15, 30, 60, 120, and 180 min after an ip injection of glucose (2.0 g/kg body weight). Glucose concentrations from blood were determined using a portable glucometer (AccuCheck).

Preparation of histological sections and lesion analysis
We analyzed atherosclerotic lesions in the aortic root and aorta of ApoE Ϫ / Ϫ and macCGI-58/ApoE-DKO animals after 10 weeks of HF/HCD feeding. Mice were euthanized and the arterial tree was perfused in situ with PBS (100 mm Hg) for 10 min preferences are comparable between macrophages of both genotypes (inset Fig. 2F ).
To investigate whether CGI-58 defi ciency in macrophages results in a compensatory upregulation of genes involved in intra-and extracellular TG hydrolysis, we determined mRNA expression of ATGL, hormone-sensitive lipase, lysosomal acid lipase, and LPL. We found no signifi cant differences in the mRNA expression of these genes in CGI-58 Ϫ / Ϫ compared with Wt macrophages (supplementary Fig. IIA). ATGL protein expression was unchanged as well (supplementary Fig. IIB). To further address whether additional metabolic changes interfere with TG homeostasis in CGI-58 Ϫ / Ϫ macrophages, we examined LPL activity. Like in ATGL Ϫ / Ϫ macrophages, LPL activity was unchanged in CGI-58 Ϫ / Ϫ macrophages (supplementary Fig. IIC).

Apoptosis and ER stress are not induced in CGI-58 Ϫ / Ϫ macrophages
We have previously shown that the mitochondrial apoptosis pathway is induced in ATGL Ϫ / Ϫ macrophages ( 12 ). mRNA expression levels of the two anti-apoptotic markers Bcl-XL and Mcl-1, however, were unchanged in CGI-58 Ϫ / Ϫ macrophages ( Fig. 3A ). Furthermore, Western blotting analysis revealed no differences in the protein expression of Bax and cytosolic cytochrome C ( Fig. 3B ), which are key players in mitochondria-dependent apoptosis ( 24 ). Unlike ATGL Ϫ / Ϫ macrophages, mitochondria in CGI-58 Ϫ / Ϫ macrophages are electron dense and have intact cristae, as in Wt macrophages ( Fig. 3C ). Finally, Annexin V/PI staining revealed the same amount of alive, early apoptotic, and necrotic cells. The number of late apoptotic cells was RESULTS Plasma lipid parameters, body weight, and glucose levels are unaffected in macCGI-58 Ϫ / Ϫ mice macCGI-58 Ϫ / Ϫ mice are viable with no apparent changes in skin phenotype (not shown). We observed no differences in lipid parameters and body weight between female Wt and macCGI-58 Ϫ / Ϫ mice, neither on chow ( Fig. 1A, B ) nor on WTD (data not shown). Glucose concentrations were decreased at 39 weeks of age but comparable to control levels at all other time points ( Fig. 1C ). Glucose tolerance in mice fed WTD was unchanged as well ( Fig. 1D ).
We also observed no signifi cant differences between both genotypes when we repeated the experiments in male mice (data not shown).
Decreased TG hydrolase activity and TG-rich lipid droplet accumulation in CGI-58 Ϫ / Ϫ macrophages We confi rmed the absence of CGI-58 expression in macrophages by real time PCR ( Fig. 2A ) and Western blotting ( Fig. 2B ). TG hydrolase activity of lysates was signifi cantly ( Ϫ 29%) decreased in CGI-58 Ϫ / Ϫ compared with Wt macrophages ( Fig. 2C ). CGI-58 Ϫ / Ϫ macrophages showed an increased number of lipid droplets as evidenced by immunofl uorescence and electron microscopy ( Fig. 2D ). Biochemical measurements revealed a specifi c accumulation of TG ( Fig. 2E ). CE hydrolase activities were comparable in macrophages from CGI-58 Ϫ / Ϫ and Wt mice (supplementary Fig. I), which is in accordance with unaltered TC and unesterifi ed FC concentrations ( Fig. 2E ). Analysis of FA composition within TG revealed increased concentrations of all FA species analyzed ( Fig. 2F ). Unchanged relative distribution of FAs within TG indicates that hydrolysis capacity in CGI-58 Ϫ / Ϫ compared with Wt macrophages, independent of glucose availability ( Fig. 4A ). To elucidate whether decreased ␤ -oxidation might contribute to the reduced phagocytic capacity of CGI-58 Ϫ / Ϫ macrophages, we analyzed the expression of PPAR ␣ target genes. mRNA expression levels of carnitine palmitoyl-transferase 1 ␣ (Cpt1 ␣ ), fatty acyl-CoA oxidase (Aox), very long chain acyl-CoA dehydrogenase (Vlcad), and medium chain acyl-CoA dehydrogenase (Mcad), however, were comparable between CGI-58 Ϫ / Ϫ and Wt macrophages ( Fig. 4B ). Next, we analyzed whether basal and maximal respiration rates are changed in mitochondria of CGI-58 Ϫ / Ϫ macrophages in the presence or absence of glucose and glutamine. Measurement of the absolute OCR revealed that mitochondria of CGI-58 Ϫ / Ϫ macrophages respire less compared with mitochondria from Wt mice in both conditions ( Fig. 4C ).
Relative OCR (presented as percent of maximal respiration) demonstrates that mitochondria from CGI-58 Ϫ / Ϫ macrophages are still responsive to oligomycin treatment and chemical uncoupling by FCCP ( Fig. 4D, E ). In addition, we observed unchanged mitochondrial surface area (supplementary Fig. IIIA) and protein expression of increased by 3.6-fold in CGI-58 Ϫ / Ϫ macrophages but lacked statistical signifi cance ( Fig. 3D ). Next, we investigated whether ER stress might be activated in CGI-58 Ϫ / Ϫ macrophages due to the TG-rich lipid droplet accumulation, as observed in ATGL Ϫ / Ϫ macrophages ( 13 ). These analyses revealed unchanged mRNA levels of the ER-resident chaperones Pdi and Erdj4 ( Fig.  3E ), unaltered protein expression of IRE1 ␣ ( Fig. 3F ), which is responsible for the X-box-binding protein 1 splicing during ER stress ( 13 ), and no protein expression of the cell death executor CHOP ( Fig. 3G ) in CGI-58 Ϫ / Ϫ macrophages ( 25 ). These fi ndings demonstrate that CGI-58 Ϫ / Ϫ macrophages lack any signs of mitochondrial apoptosis, ER stress, and mitochondrial dysfunction as observed in ATGL Ϫ / Ϫ macrophages.

CGI-58 Ϫ / Ϫ macrophages polarize toward M1
Macrophages are a heterogeneous and phenotypically polarized cell population consisting of classically activated
Comparable to macCGI-58/ApoE-DKO mice ( Fig. 6A ), ASO-mediated knockdown of CGI-58 did not alter atherosclerosis burden in the aortic sinus ( Fig. 7G ) or thoracic aorta of LDLR Ϫ / Ϫ mice as measured by en face morphometry ( Fig. 7H ). Collectively, these results indicate that diminished CGI-58 in macrophages, liver, and adipose tissue driven by CGI-58 ASO treatment or defi ciency of CGI-58 in myeloid cells has minimal effects on atherosclerosis progression.

DISCUSSION
Due to the functional changes in macrophages lacking ATGL (12)(13)(14)(15), we proposed that the absence of its coactivator CGI-58 results in similar alterations of macrophage function. To investigate the consequences of CGI-58 defi ciency in macrophages, we generated a mouse model lacking CGI-58 exclusively in myeloid cells including monocytes, mature macrophages, and granulocytes (macCGI-58 Ϫ / Ϫ ). These mice are viable with no apparent changes in skin phenotype. Compared with Wt mice, the lack of CGI-58 in myeloid cells does not affect body weight, lipid parameters, glucose levels, and glucose tolerance even when mice are challenged with WTD. Unchanged glucose tolerance in WTD-fed mice is contradictory to the results shown recently by Miao et al. ( 28 ), who reported impaired glucose tolerance in high-fat diet-fed male macCGI-58 Ϫ / Ϫ mice . This discrepancy is difficult to explain as IL-6 concentrations (supplementary Fig. IVA)

between DKO and ApoE
Ϫ / Ϫ mice. Macrophages from both genotypes show the same polarization (supplementary Fig.  IVB), which is in contrast to what we observed in CGI-58 Ϫ / Ϫ compared with Wt macrophages. Visual inspection of Oil Red O-stained aortic root sections revealed no differences in lesion formation ( Fig. 6A ). Moma-2 and Masson's trichrome staining, which were used to identify lesion macrophages and collagen, were comparable in sections from ApoE Ϫ / Ϫ and DKO mice ( Fig. 6A ). Quantitative analysis of plaque development in aortic arches (en face analysis) revealed a 1.3-fold increase in lesion size in thoracic aortic arches of macCGI-58/ApoE-DKO mice ( Fig. 6B ).
To investigate the effect of CGI-58 defi ciency on cholesterol transport from macrophages to exogenous lipid acceptors, we measured cholesterol effl ux to ApoA-I and HDL. As shown in Fig. 6C , cholesterol effl ux from DKO macrophages to ApoA-I was increased after 6 h compared with ApoE Ϫ / Ϫ macrophages, but unchanged to both acceptors at all other time points. This result is in line with unaltered mRNA levels of genes involved in cholesterol uptake (Cd36, SrB1) and effl ux (Abca1, Abcg1) in CGI-58 Ϫ / Ϫ compared with Wt macrophages (not shown).
To further interrogate the role of CGI-58 in atherosclerosis progression, we utilized ASO-mediated knockdown of CGI-58 in hyperlipidemic LDLR Ϫ / Ϫ mice. ASO-mediated knockdown has previously been shown to reduce expression levels of CGI-58 in the liver, white adipose tissue, and kidney ( 9 ). Here we show that thioglycolate-elicited macrophages from ASO-treated mice have a marked reduction in macrophage CGI-58 expression as well ( Fig. 7A ).
Using this system, we found that CGI-58 knockdown in  et al. ( 28 ) found increased TC and FC concentrations and reduced CD36 mRNA expression in macrophages of male macCGI-58 Ϫ / Ϫ mice. Whether the differences in the two studies are due to sex-specifi c differences or cultivation conditions of the macrophages (LPDS versus FCS) remain to be elucidated. Measurement of FA composition within TG of CGI-58 Ϫ / Ϫ macrophages revealed increased concentrations of all saturated, unsaturated, and polyunsaturated FAs analyzed with the highest change in arachidonic acid, oleic acid, and linoleic acid, respectively. FA composition has not been determined in ATGL Ϫ / Ϫ macrophages, but in white adipose tissue of ATGL Ϫ / Ϫ mice. Our fi ndings are slightly different to these results ( 29 ), where the authors showed that ATGL hydrolyzes long-chain FA esters in vivo with a modest substrate preference for C16:1.
Fragmented mitochondria in ATGL Ϫ / Ϫ macrophages are indicative of the mitochondrial apoptosis pathway being triggered as a consequence of defective lipolysis ( 12 ). We expected the same phenotype in CGI-58 Ϫ / Ϫ macrophages. Typical markers of programmed cell death, such as externalization of phosphatidylserine on the plasma both diets (WTD and high-fat diet) can be used to affect glucose tolerance. Because we have repeated the experiments in male mice that also lacked signifi cant differences between the phenotypes (data not shown), a sex difference leading to the contradictory results can be excluded.
CGI-58 defi ciency leads to decreased TG hydrolysis activity and a TG-rich lipid droplet accumulation in macrophages, identical to ATGL Ϫ / Ϫ macrophages ( 12,15 ). These results suggest that ATGL activity can be increased by CGI-58 as activator protein ( 2 ) also in macrophages. Unchanged CD36 and LPL mRNA as well as LPL activity argue against differences in FA uptake between CGI-58 Ϫ / Ϫ and Wt macrophages. Because LPL is responsible for the extracellular hydrolysis of lipoprotein-associated TG and the subsequent uptake of FAs in underlying cells and tissues, these results indicate that FAs generated by the action of LPL are taken up similarly by CGI-58 Ϫ / Ϫ and Wt macrophages. Unchanged TC and FC concentrations were associated with comparable CE hydrolase activities between Wt and CGI-58 Ϫ / Ϫ macrophages, as observed in ATGL Ϫ / Ϫ macrophages ( 15 ). In contrast to our results, Miao hydrolase activity, which is suffi cient to rescue the cell from mitochondrial apoptosis. However, macrophage TG concentrations are comparable between ATGL Ϫ / Ϫ and CGI-58 Ϫ / Ϫ macrophages, suggesting other factors to be responsible for programmed cell death, mitochondrial dysfunction, and ER stress in ATGL Ϫ / Ϫ macrophages. The in vitro phagocytic capacity was affected by the lack of ATGL ( 15 ) and CGI-58 in a similar manner with reduced phagocytosis in glucose-containing and glucosefree medium compared with Wt macrophages. It has been demonstrated that hydrolysis of cellular TG by ATGL is necessary to produce FAs as ligands for PPAR activation and that ATGL defi ciency leads to severely disrupted mitochondrial substrate oxidation and respiration ( 30 ). Although mRNA levels of PPAR ␣ target genes in macrophages were unchanged, decreased mitochondrial respiration might contribute to the reduced phagocytic capacity of CGI-58 Ϫ / Ϫ macrophages. Because the surface area of mitochondria was unaffected by CGI-58 defi ciency, the reduced respiration is likely due to the decrease in FAs as energy substrate. A decreased OCR was also described in CGI-58-silenced RAW264.7 macrophages ( 28 ). Our in vitro fi nding of reduced phagocytosis could not be confi rmed in vivo in macCGI-58 Ϫ / Ϫ mice, where we observed a trend to reduced phagocytosis, which, however, lacked statistical signifi cance. We hypothesize that the reason for this discrepancy is the difference in the mouse models: in macCGI-58 Ϫ / Ϫ mice, CGI-58 is absent in myeloid cells including macrophages, in which phagocytosis is affected in vitro. In contrast, the decreased in vivo phagocytosis ability was demonstrated in whole-body ATGL Ϫ / Ϫ mice ( 15 ). Phagocytosis is a highly energy demanding process; reduced FA concentrations in ATGL Ϫ / Ϫ mice ( 7 ) versus unaltered FA levels in macCGI-58 Ϫ / Ϫ mice might explain the observed in vivo changes in phagocytic capacity.
Due to the pronounced differences observed between ATGL Ϫ / Ϫ and CGI-58 Ϫ / Ϫ macrophages, we were particularly interested in the impact of macCGI-58 defi ciency on atherosclerosis susceptibility. Unexpectedly, plaque formation, number of macrophages, and collagen content was identical in aortic root sections of macCGI-58/ApoE-DKO and ApoE Ϫ / Ϫ mice. In line with these results, en face analyses of the aortic arch area revealed comparable lesion sizes in both genotypes. When we analyzed lesion per total area of the thoracic aorta, we found slightly increased plaque formation in DKO mice. In agreement, ASO-mediated knockdown of CGI-58 in LDLR Ϫ / Ϫ mice did not alter atherosclerosis burden in the aortic root, but very modestly increased atherosclerosis in the aorta. Collectively, these fi ndings are again contrary to what we observed in the absence of ATGL, where transfer of ATGL Ϫ / Ϫ bone marrow into LDLR Ϫ / Ϫ mice resulted in markedly reduced plaque formation ( 16 ). Although different mouse models were used in these two studies, our results indicate that the absence of ATGL or CGI-58 in myeloid cells results in different phenotypes: reduced lesion development by ATGL defi ciency and unaltered or even slightly increased plaque formation by lack of CGI-58. This difference in plaque formation is unlikely due to membrane measured by fl ow cytometry after annexin V/PI costaining, mRNA and protein expression of proapoptotic markers, and lack of fragmented mitochondria prior to cell death, however, revealed that mitochondrial apoptosis is not induced in CGI-58 Ϫ / Ϫ macrophages. Absent CHOP protein expression and unaltered mRNA expression of the ER-resident chaperones Erdj4 and Pdi indicate that CGI-58 defi ciency in macrophages does not cause ER stress. These results are different from fi ndings in ATGL Ϫ / Ϫ macrophages, where we observed induction of apoptosis ( 12 ) and ER stress ( 13 ). Interestingly, incubation of Wt macrophages with VLDL resulted in the same apoptotic phenotype and fragmentation of mitochondria as observed in ATGL Ϫ / Ϫ macrophages ( 12 ). From these results, we had initially concluded that intracellular TG accumulation is linked to mitochondrial dysfunction and programmed cell death in macrophages. The results from the present work are therefore counterintuitive. Why TG accumulation leads to mitochondrial dysfunction in ATGL Ϫ / Ϫ and VLDL-loaded Wt macrophages ( 12 ), but fails to affect CGI-58 Ϫ / Ϫ macrophages, is elusive. Because ATGL is present in CGI-58 Ϫ / Ϫ macrophages, it might be claimed that there is still basal ATGL-mediated TG that the lack of CGI-58 in macrophages affects the infl ammatory response to LPS. This fi nding is in line with results from ASO-mediated CGI-58 knockdown experiments, in which lipid second messengers generated by CGI-58 were shown to be critically involved in maintaining the balance between infl ammation and insulin action, thereby predicting a role of CGI-58 in cytokine signaling ( 31 ). On an ApoE Ϫ / Ϫ background, however, plasma IL-6 concentrations and macrophage polarization markers were no longer different between mice that express or lack CGI-58 in macrophages. Because ApoE differences in cholesterol effl ux capacities of the cells, because macrophages lacking either ATGL or CGI-58 show comparable cholesterol effl ux as control cells. The M2-like polarization of ATGL Ϫ / Ϫ macrophages and the M1-like phenotype of CGI-58 Ϫ / Ϫ might contribute (at least in part) to the differences in plaque formation. In accordance with our results, Miao et al. ( 28 ) have recently shown that macrophage CGI-58 defi ciency activates the ROS-NLRP3 infl ammatory pathway to promote insulin resistance in mice. We found increased IL-6 secretion from LPS-treated CGI-58 Ϫ / Ϫ macrophages, suggesting promotes macrophage conversion from M1 to the M2 phenotype ( 32 ), it might be speculated that the absence of ApoE directs macrophages into an M1-like phenotype, which cannot be further boosted by CGI-58 defi ciency. In contrast, reduced plaque formation in aortic roots of LDLR Ϫ / Ϫ mice transplanted with ATGL Ϫ / Ϫ bone marrow was associated with decreased macrophage IL-6 concentrations ( 16 ).
The data presented in this study provide evidence that the TG-rich lipid droplet accumulation within ATGL Ϫ / Ϫ macrophages is likely not the reason for attenuated atherosclerosis development as initially hypothesized. Although loss of CGI-58 in macrophages results in comparable TG accumulation, several results were different between CGI-58 Ϫ / Ϫ and ATGL Ϫ / Ϫ macrophages: Absence of ER stress, mitochondrial apoptosis, and mitochondrial dysfunction, as well as M1-like polarization of CGI-58 Ϫ / Ϫ macrophages, and differences in atherosclerosis susceptibility argue for different and/or additional function(s) of CGI-58 in macrophages beside activation of ATGL. We conclude that lack of the enzyme (ATGL) induces a more severe phenotype than loss of the activator (CGI-58).