Salicylate improves macrophage cholesterol homeostasis via activation of Ampk[S]

Atherosclerosis stems from imbalances in lipid metabolism and leads to maladaptive inflammatory responses. The AMP-activated protein kinase (Ampk) is a highly conserved serine/threonine kinase that regulates many aspects of lipid and energy metabolism, although its specific role in controlling macrophage cholesterol homeostasis remains unclear. We sought to address this question by testing the effects of direct Ampk activators in primary bone marrow-derived macrophages from Ampk β1-deficient (β1−/−) mice. Macrophages from Ampk β1−/− mice had enhanced lipogenic capacity and diminished cholesterol efflux, although cholesterol uptake was unaffected. Direct activation of Ampk β1 via salicylate (the unacetylated form of aspirin) or A-769662 (a small molecule activator), decreased the synthesis of FAs and sterols in WT but not Ampk β1−/− macrophages. In lipid-laden macrophages, Ampk activation decreased cholesterol content (foam cell formation) and increased cholesterol efflux to HDL and apoA-I, effects that occurred in an Ampk β1-dependent manner. Increased cholesterol efflux was also associated with increased gene expression of the ATP binding cassette transporters, Abcg1 and Abca1. Moreover, in vivo reverse cholesterol transport was suppressed in mice that received Ampk β1−/− macrophages compared with the WT control. Our data highlight the therapeutic potential of targeting macrophage Ampk with new or existing drugs for the possible reduction in foam cell formation during the early stages of atherosclerosis.

thromboxane A 2 synthesis and other anti-coagulant therapies do not display the same cardioprotective effect ( 36,37 ). Upon ingestion, aspirin is rapidly deacetylated in the circulation to salicylate ( 38 ). Given the gastrointestinal and thrombotic side-effects of higher doses of aspirin and salicylate, a pro-drug of salicylate (salsalate), which is better tolerated, is now being used in clinical trials for CVD (TINSAL-CVD: NCT00624923). At clinical salsalate concentrations, salicylate directly activates Ampk and is required for increasing rates of FA oxidation in hepatocytes ( 25 ). The role of salicylates in regulating macrophage Ampk activity has not been studied.
Here we show that deletion of Ampk results in higher FA and cholesterol synthesis. Under lipid-laden conditions, Ampk ␤ 1-defi cient ( ␤ 1 Ϫ / Ϫ ) macrophages have more lipid accumulation and lower cholesterol effl ux. In addition, direct activation of Ampk was able to restore cholesterol homeostasis in lipid-loaded macrophages through the suppression of lipid synthesis and foam cell formation, as well as stimulation of cholesterol effl ux in vitro and in vivo.

Mice
The generation and characterization of the Ampk ␤ 1 Ϫ / Ϫ mice has been previously described ( 39,40 ). The Ampk ␤ 1 Ϫ / Ϫ and littermate WT control mice used in these studies were housed in specifi c pathogen-free micro-isolators, maintained on a 12 h light/12 h dark cycle with lights on at 0700, and had unlimited access to standard rodent chow and water. All animal experimental protocols used were approved by the McMaster University Animal Ethics Research Board and the University of Ottawa Animal Care Committee.

Cell culture
Bone marrow-derived macrophages (BMDMs) were generated as previously described ( 32 ). Briefl y, mice were euthanized, tibias and femurs isolated, and the ends of each bone cut off. The tibia and femur from each leg were placed into a sterile 0.5 ml microfuge tube that had a hole punctured in the end with an 18 gauge needle, which was then placed inside a 1.5 ml microfuge tube before the addition of 100 l of DMEM (Invitrogen) to the 0.5 ml tube. Bone marrow cells were collected by centrifuging at 2,000 rpm for 4 min, resuspended and plated in 100 ml of DMEM supplemented with 10% FBS (Invitrogen) and 1% penicillin/ streptomycin (Invitrogen) in a T175 fl ask, and incubated at 37°C in a humidifi ed atmosphere at 5% CO 2 . After 4 h, cells were plated into 10 cm tissue culture dishes in the presence of 20% L929 medium (as a source of macrophage colony stimulating factor) and left to differentiate for 7-8 days. One day prior to the experiment, cells were lifted into suspension in the existing L929-supplemented DMEM by gentle scraping and seeded into the appropriate plate for subsequent experiments.

Foam cell formation and lipid determinations
BMDMs were subjected to two distinct foam cell protocols. In foam cell protocol 1, BMDMs were incubated in the presence or absence of acetylated LDL (acLDL) (50 g/ml) [Biomedical Technologies Inc. (BTI)] for 30 h in the presence or absence of salicylate (3 mM) or A-769662 (100 M). In foam cell protocol 2, critical for the uptake of modifi ed lipoproteins within the subendothelial space ( 7 ). The accumulation of lipid-laden macrophages (foam cells) results in the progression of atherosclerotic plaques. As atherosclerosis progresses, increased lipid content (mainly cholesteryl esters) of macrophages is intimately linked to the increased infl ammatory tone in the plaque environment, and the infi ltration and activation of numerous innate and adaptive immune cells ( 2,3,5,8 ). Because the lipid burden of macrophage foam cells contributes directly to the infl ammatory tone and risk of plaque rupture ( 3 ), strategies aimed at lowering the lipid content of foam cells may be useful therapeutically.
The physiological process known as reverse cholesterol transport (RCT), whereby cholesterol is removed from peripheral tissues and transported by HDLs to the liver for excretion through bile and ultimately the feces ( 9 ), has garnered therapeutic interest. In the macrophage, excess cholesterol is esterifi ed to cholesteryl esters and stored in lipid droplets, leading to foam cell formation. However, cholesterol can be mobilized from lipid droplet stores ( 10,11 ) and effl uxed via active transport to extracellular acceptors, HDL and lipid-poor apoA-I, a process mediated by the ATP binding cassette transporters, Abcg1 and Abca1, respectively (12)(13)(14). There is now strong evidence to suggest that increased cholesterol effl ux from macrophage foam cells is a strong predictor of improved CVD risk profi le ( 9,(15)(16)(17). Thus a mechanistic understanding of signaling pathways that promote cholesterol effl ux and RCT is crucial for developing future therapeutic interventions.
Aspirin (acetylsalicylate) is one of the most widely prescribed medications world-wide for the primary and secondary treatment of CVD. Aspirin irreversibly inhibits cyclo-oxygenases to disrupt prostaglandin synthesis ( 34 ) and reduces coagulation by inhibiting thromboxane A 2 production in platelets ( 35 ). It was initially thought that the benefi ts to aspirin therapy were entirely mediated by its anti-platelet effects; however, genetic disruption of

Western blotting
Cellular lysates were prepared, and Western blotting and quantifi cation were performed as previously described ( 32 ). All antibodies were from Cell Signaling Technologies.

Flow cytometry
Cells were washed in PBS and removed from the wells via gentle pipetting into a 96-well 0.2 ml plate. Cells were then washed in PBS containing 1% BSA and pelleted at 500 × g . Cells were then incubated with 0.5 g Fc block (CD16/32; BioLegend) for 30 min followed by incubation with the conjugated primary antibodies APC-CD36 (1:100; Miltenyl Biotec), FITC-SR-A (1:20; Miltenyl Biotec), and PE-SR-BI (1:20; Novus) for a further 30 min in a volume of 50 l covered from light. Cells were then washed twice more with PBS containing 1% BSA and resuspended in a fi nal volume of 200 l. Cells were analyzed using a CyAn TM ADP analyzer and fl uorescence intensity was calculated as fold increase over unstained genotype and treatment controls by FlowJo software (Tree Star).

In vivo RCT
In vivo RCT was performed as previously described ( 11,48 ). In brief, BMDMs from Ampk ␤ 1 Ϫ / Ϫ and WT control mice were plated into 10 cm tissue culture plates. acLDL (50 g/ml; BTI) pre-equilibrated with 5 Ci/ml [ 3 H]cholesterol (Perkin Elmer) in DMEM supplemented with 5% LPDS (BTI) for 12 h, was then added to the BMDMs for a further 30 h. Cells were washed and harvested by gentle scraping. Approximately 8 × 10 6 cells were injected into the intraperitoneal cavity of WT C57BL/6 mice. The radioactivity of each genotype preparation was determined prior to injection to account for changes in specifi c activity between isolations . Cellular cholesterol mass was also determined from each genotype to ensure equal lipid loading. Blood was collected at 48 h via cardiac puncture and livers were removed and weighed. Feces were collected over the 48 h period and total feces radioactivity was measured. Radioactivity was determined in serum, liver, and total feces (of equivalent wet weight) by LSC. cells were incubated with acLDL (50 g/ml) for 30 h and then equilibrated for 12-16 h in 0.2% BSA DMEM stimulated in the presence or absence of salicylate (3 mM) and A-769662 (100 M). At the completion of the incubation, cells were washed twice with PBS and cholesterol determined using the Amplex cholesterol kit (Invitrogen) as described ( 41 ).

Lipogenesis and lipid extraction
BMDMs were labeled with [ 3 H]sodium acetate in the presence of 5 mM unlabeled sodium acetate, in the presence or absence of salicylate (3 mM), A-769662 (100 M), or DMSO vehicle control for 16 h. Cells were then washed twice with ice-cold PBS and scraped in 200 l PBS for lipid extraction. Lipids were extracted as described previously ( 44,45 ). Radioactivity was determined by counting the chloroform phase containing total lipids or after saponifi cation and extraction with petroleum ether, as previously described ( 45 ).

RNA isolation, cDNA synthesis, and quantitative PCR
RNA was isolated using TRIzol reagent (Invitrogen) according to the manufacturer's instructions. Total RNA was DNase-I treated (Invitrogen) and fi rst strand synthesis was performed using SuperScript III reverse transcriptase (Invitrogen). cDNA was diluted 1:40 into ultrapure water, and mRNA expression determined using TaqMan assays (Invitrogen). For Sr-a determinations, primers were used as previously described ( 46 ). Relative expression was calculated using the ⌬ ⌬ Ct method, as previously described ( 32,47 ). Duplicate gels were run for total Acc determination. WT and Ampk ␤ 1 Ϫ / Ϫ samples are loaded on the same gel, dotted line represents a cropped lane. Gels are representative of three separate bone marrow isolations per genotype, performed in duplicate. Data represent mean ± SEM, where *** P < 0.001 compared within genotype and # P < 0.05 and ### P < 0.001 are differences between treatment groups compared with vehicle control. Similar to primary hepatocytes ( 22,39 ), deletion of the Ampk ␤ 1 subunit in BMDMs resulted in a higher lipogenic rate (incorporation of [ 3 H]acetate into lipids) compared with WT controls ( Fig. 2A, B ). The dysregulation of lipid synthesis resulted in a higher incorporation of labeled acetate into both FAs and sterols in Ampk ␤ 1 Ϫ / Ϫ BMDMs compared with WT cells ( Fig. 2C, D ). Upon Ampk activation with the direct ␤ 1 activators salicylate and A-769662, both FA and sterol synthesis were inhibited in WT BMDMs, but were unaffected in the Ampk ␤ 1 Ϫ / Ϫ cells ( Fig. 2A-D ).
The progression of atherosclerosis stems from the unregulated uptake of modifi ed lipoproteins by macrophages in the sub-endothelial space. We next treated WT and Ampk ␤ 1 Ϫ / Ϫ macrophages with acLDL to induce foam cell formation both in the presence and absence of Ampk activators. Exposure to acLDL caused a marked increase in cellular cholesterol content in cells from both genotypes ( Fig. 3A ), with no differences in triacylglyceride content (data not shown). When cells were lipid-loaded in the presence of salicylate or A-769662, there was a significant reduction in total cholesterol accumulation and foam cell formation in WT cells, but not Ampk ␤ 1 Ϫ / Ϫ cells. To further investigate how Ampk may regulate macrophage All [ 3 H]tracer measurements are expressed relative to the initial injected dose.

Statistics
All results are shown as mean ± SEM. Results were analyzed using a two-tailed Student's t -test or two-way ANOVA where appropriate using GraphPad Prism software. A Bonferonni post hoc test was used to test for signifi cant differences revealed by the ANOVA. Signifi cance was accepted at P р 0.05.

RESULTS
We fi rst sought to determine the effi cacy of the direct Ampk ␤ 1 activators in primary macrophages. Similar to hepatocytes ( 25 ), salicylate increased Ampk activity as indicated by increased phosphorylation of Acc1 (pSer 79) in a dose-dependent manner, as did A-769662, a result that we have shown previously ( 32 ) ( Fig. 1 ). This effect was abolished in Ampk ␤ 1 Ϫ / Ϫ cells ( Fig. 1 ). Neither treatment affected Ampk Thr172 phosphorylation (data not shown), which is in keeping with their allosteric mode of activation ( 24,25,27,49 ). Phosphorylation of Acc1 by Ampk on Ser 79 inhibits its activity, resulting in reduced FA synthesis ( 22 ). . Total lipids were then saponifi ed and radioactivity determined from saponifi able (FAs) (C) and nonsaponifi able (sterols) (D) fractions, after 16 h. Data represent mean ± SEM and are from three separate bone marrow isolations per genotype, performed in triplicate, where * P < 0.05, ** P < 0.01, and *** P < 0.001 are differences compared within genotype and # P < 0.05, ## P < 0.01, and ### P < 0.001 are differences between treatment groups compared with vehicle control. scavenger receptors) is in keeping with the notion that Ampk activation has no effect on acLDL uptake.
The ability of macrophages to effl ux cholesterol to extracellular acceptors is a critical component of RCT ( 9,52 ). We treated BMDM cells with acLDL to induce foam cell formation, and then assessed cholesterol effl ux to both HDL (mediated mainly by Abcg1) and lipid-poor apoA-I (mainly to Abca1) in the presence and absence of Ampk activators. Cholesterol effl ux was lower in Ampk ␤ 1 Ϫ / Ϫ cells compared with WT control cells ( Fig. 4A, B ). Ampk activation via salicylate and A-769662 resulted in increased cholesterol effl ux to both HDL and apoA-I ( Fig. 4A, B ), but only in WT macrophages, thus suggesting that Ampk-dependent processes contribute to macrophage RCT. Importantly, treatment with salicylate or A-769662 after the period of lipid loading cholesterol metabolism, we monitored the uptake of Dil-acLDL in BMDM cells that had been lipid-loaded in the presence or absence of the Ampk ␤ 1 activators; however, there were no signifi cant genotype or treatment differences ( Fig. 3B ). We next investigated the transcript and surface expression of the main scavenger receptor involved in acLDL uptake, SR-A, as well as SR-BI and CD36. There were no differences in transcript expression between genotype or treatment when cells were lipid-loaded in the presence of salicylate or A-769662 (supplementary Fig. 1A-C). To investigate further, we determined that the surface expression of SR-A, SR-BI, and CD36 were also unaltered ( Fig. 3C-E ). Because SR-A has been demonstrated to be the main effector by which acLDL is taken into the macrophage ( 50,51 ), unaltered transcript and surface expression of SR-A (as well as other extent in WT BMDMs ( Fig. 5C ). Exogenous lipid loading suppressed Srebp2 expression, although this effect was prevented in the Ampk ␤ 1 Ϫ / Ϫ macrophages ( Fig. 5D ). Ampk activation was associated with further increases in Lxr-␣ ( Fig. 5C ) and further suppression of Srebp2 ( Fig. 5D ). We next investigated the role of macrophage Ampk in RCT in vivo. Radiolabeled, lipid-loaded macrophages from WT and Ampk ␤ 1 Ϫ / Ϫ mice (acLDL-[ 3 H]cholesterol) were injected into WT mice. We assessed radioactivity in the serum, liver, and feces after 48 h ( Fig. 6A-C ). Consistent with the observed genotype difference in cholesterol effl ux, radioactivity in the serum ( P < 0.05), liver ( P < 0.01), and feces ( P < 0.001) were all signifi cantly lower in mice that received macrophages from Ampk ␤ 1 Ϫ / Ϫ mice, compared with WT ( Fig. 6 ). Taken together, these data are entirely consistent with a regulatory role for Ampk in mediating the reverse transport of cholesterol.

DISCUSSION
Hyperlipidemia and other associated risk factors that predispose individuals to atherosclerosis lead to the overabundance of modifi ed LDL-cholesterol and unregulated uptake via scavenger receptors, ultimately causing foam cell formation and atherogenesis ( 7 ). We show that Ampk plays an integral role in regulating macrophage cholesterol accumulation. The deletion of the Ampk ␤ 1 subunit in macrophages results in a higher lipogenic rate and lower cholesterol effl ux. Moreover, activation of Ampk using the direct Ampk ␤ 1 activators salicylate and A-769662 resulted in a protective decrease in lipogenesis, cholesterol accumulation, and foam cell formation, and an increase in cholesterol effl ux.
Ampk inhibits multiple facets of lipid metabolism. The synthesis of FAs is inhibited via phosphorylation of Acc1 at Ser 79, which reduces the production of the FA precursor malonyl-CoA ( 22,29 ). Ampk also phosphorylates Hmgcr at Ser 871 (in mice), the rate-limiting enzyme in cholesterol synthesis ( 31 ); however, the physiological role of this regulation has yet to be examined in detail. In addition, Ampk directly phosphorylates Srebp1c (Ser 372) and Srebp2 (site unknown) ( 54 ), which are master transcriptional regulators that govern lipid levels via modulation of FA and cholesterol synthetic pathways, respectively. Upon activation of Ampk in macrophages, the synthesis of FAs and sterols was inhibited in WT cells, but not Ampk ␤ 1 Ϫ / Ϫ cells ( Fig. 2 ). Ampk ␤ 1 defi ciency is associated with higher Srebp2 gene expression compared with WT control cells, and in response to Ampk activators, Srebp2 expression was signifi cantly inhibited in WT cells, but not Ampk ␤ 1 Ϫ / Ϫ cells ( Fig. 5D ). The acute Ampk-specifi c inhibition of FA and sterol synthesis by salicylate and A-769662 after 4 h was likely mediated by the acute phosphorylation and inactivation of Acc1 and Hmgcr, respectively. However, during chronic experiments (24-30 h), the continued inhibition of lipogenesis and the modulation of lipid levels may refl ect the net contribution of all affected pathways, including did not alter cholesterol mass (supplementary Fig. 2), therefore specifi c activity was unaffected.
Given the signifi cant effect of Ampk deletion on effl ux to HDL and apoA-I, as well as the protective increase in efflux upon Ampk activation, we next assessed the transcript expression of the main cholesterol transporters in macrophages ( 48,53 ). The expression of the Abcg1 and Abca1 transporters were reduced in Ampk ␤ 1 Ϫ / Ϫ BMDMs under vehicle-treated conditions ( Fig. 5A, B ), and both salicylate and A-769662 increased Abcg1 and Abca1 expression in WT but not Ampk ␤ 1 Ϫ / Ϫ cells ( Fig. 5A, B ). In addition to lipid transporter expression, we also monitored the expression of two important transcriptional regulators responsible for governing cholesterol homeostasis, liver X receptor ␣ (Lxr-␣ ) and sterol regulatory element binding protein (Srebp)2. Similar to the expression of Abca1 and Abcg1, Lxr-␣ expression was induced upon lipid loading, which occurred to a greater Data represent mean ± SEM and represent at least four separate bone marrow isolations per genotype, performed in triplicate, where * P < 0.05 and *** P < 0.001 are differences compared within genotype and ### P < 0.001 are differences between treatment groups compared with vehicle control. uptake reported ( 55 ). In our experiments (supplementary Fig. 1) and those of Li et al. ( 55 ), this effect on foam cell formation was independent of alterations in scavenger receptor gene expression. Our investigation determined that the surface expression of key scavenger receptors (SR-BI and CD36), as well as the main receptor responsible for acLDL uptake, SR-A, were unaffected by Ampk signaling ( Fig. 3C-E ). This data suggests that Ampk activation alters cholesterol accumulations via uptake-independent mechanisms.
The ability of macrophages to effl ux and the ability of HDL and other acceptors to transport cholesterol for the purpose of RCT have garnered therapeutic interest ( 15,52,56,57 ). Many studies have used various cholesterol mobilization and transport protein knockout models to assess their role in macrophage RCT. Li et al. ( 55 ) demonstrated that 5-aminoimidazole-4-carboxyamide ribonucleoside, an indirect Ampk activator, increased Abcg1-mediated effl ux to HDL in oxidized LDL-loaded macrophages and endothelial cells ( 58 ), although the authors did not assess effl ux to apoA-I, nor did they note a basal difference with transient transfection of a dominant negative Ampk. Human monocyte-derived macrophages incubated with the Ampk activators metformin and heme also demonstrate an increase in cholesterol effl ux and stemming of foam cell formation ( 59 ). In the current study we found that exposure to direct Ampk activators Srebp2. The relative contributions of the possible downstream regulators remain unclear.
In spite of differences in foam cell formation in the presence of Ampk activators ( Fig. 3A ), neither endogenous Ampk signaling, nor Ampk activation had any effect on the uptake of acLDL into lipid-loaded macrophages ( Fig. 3B ). This is consistent with previous studies demonstrating reductions in cholesterol accumulation with oxidized LDL and 5-aminoimidazole-4-carboxyamide ribonucleoside treatment, although there were no measures of Srebp2 (D) were determined. Transcripts are shown relative to WT control in the absence of acLDL and expressed relative to ␤ actin. Data represent mean ± SEM and represent three separate bone marrow isolations per genotype, performed in quadruplicate, where * P < 0.05, ** P < 0.01, and *** P < 0.001 are differences compared within genotype and # P < 0.05, ## P < 0.01, and ### P < 0.001 are differences between treatment groups compared with vehicle (acLDL) control . increased Abcg1 and Abca1 expression and cholesterol effl ux to both HDL and apoA-I ( Figs. 4, 5 ). Basal levels of effl ux and associated transporter expression were lower in Ampk ␤ 1 Ϫ / Ϫ macrophages, indicating that Ampk is important for controlling endogenous cholesterol effl ux. In addition, we show for the fi rst time that macrophage Ampk plays a critical role in mediating effi cient in vivo RCT ( Fig.  6 ). Interestingly, the process of in vivo RCT was diminished in Ampk ␤ 1 Ϫ / Ϫ macrophages ( ‫ف‬ 30%) to a similar extent as cells that lack either Abca1 or Abcg1 ( 48,53 ). This strongly suggests that Ampk is linked to important regulatory pathways that govern this process. Our current study highlights a protective role for macrophage Ampk in regulating cholesterol metabolism. These effects are mediated through reductions in cholesterol synthesis under normal conditions and reductions in uptake and enhanced cholesterol effl ux/RCT under lipid-loaded conditions. Importantly, we have tested A-769662 and salicylate and show that they reduce macrophage lipid synthesis and enhance cholesterol effl ux via an Ampk ␤ 1-dependent pathway. While A-769662 has poor bioavailability ( 24 ), following salsalate and high-dose aspirin ingestion, salicylate is present in circulation at concentrations similar to those used in our studies (1-3 mM) ( 60,61 ). Because Ampk ␤ 1 is the predominant subunit in human macrophages ( 33 ), it is interesting to speculate that the activation of macrophage Ampk might mediate a portion of the benefi cial effects of these drugs on CVD. Future studies testing this hypothesis in mouse models of atherosclerosis are warranted.