Rosiglitazone Inhibits Acyl-CoA Synthetase Activity and Fatty Acid Partitioning to Diacylglycerol and Triacylglycerol via a Peroxisome Proliferator–Activated Receptor-γ–Independent Mechanism in Human Arterial Smooth Muscle Cells and Macrophages Free

Oleic acid (OA) (sodium salt) was purchased from Nucheck Prep (Elysian, MN). RPMI-1640, fetal bovine serum, and penicillin/streptomycin were obtained from Gibco/Invitrogen (Grand Island, NY). CoA, ATP, essentially fatty acid–free BSA, sodium thioglycollate, dithiothreitol, EDTA, EGTA, leupeptin, aprotinin, pepstatin, phenylmethylsulfonyl fluoride, and benzamidine were obtained from Sigma-Aldrich (St. Louis, MO). Triascin C, T0070907, 5-aminoimidazole-4-carboxaide 1-β-d-ribo-furanoside (AICAR), 15-deoxy-Δ^12,14 prostaglandin J₂ (15-dPGJ₂), and Wy14643 were from Biomol Research Laboratories (Plymouth Meeting, PA). Rosiglitazone was obtained from Alexis Biochemical (San Diego, CA). Anti-rabbit IgG horseradish peroxidase–linked antibodies, [9,10(n)-³H]-OA (259 GBq/mmol), [1-¹⁴C]-OA (2.07 GBq/mmol), and [U-¹⁴C]-glycerol (5.9 GBq/mmol) were obtained from Amersham Biosciences (Piscataway, NJ). Arachidonic acid (AA) [5,6,8,9,11,12,14,15,³H(N)] (6.66–8.88 TBq/mmol, 9.25 MBq) was obtained from Perkin Elmer Life and Analytical Sciences (Wellesley, MA). Recombinant human platelet-derived growth factor-BB was from Upstate Biotechnology (Lake Placid, NY).

Normal human aortic smooth muscle cells (hSMCs) were isolated by an explant method and grown as previously described (11). Incubation with fatty acids was performed in serum-free medium in the presence of 0.5% fatty acid–free BSA. All investigation on human cells was conducted according to the principles of the Declaration of Helsinki.

Monocytes were purified from buffy coats obtained from human donors by the Red Cross (Portland, OR), as described (12). Monocytes were purified by using density-gradient centrifugation with Ficoll-Paque Plus (Amersham Pharmacia Biotech), followed by positive selection with magnetic CD-14 microbeads (Miltenyi Biotec, Auburn, CA). Monocytes were differentiated in RPMI medium 1640 supplemented with penicillin/streptomycin and 10% fetal bovine serum and buffered with 10 mmol/l HEPES. The monocytes were differentiated to macrophages at a concentration of 5 × 10⁵ cells per milliliter with 100 ng/ml granulocyte macrophage colony stimulating factor (Leukine; Drug Services, University of Washington).

Normal murine SMCs were isolated from the thoracic aorta of male C57BL/6 mice via enzymatic digestion (13).

Peritoneal macrophages were collected from male and female C57BL/6 (25–30 g) mice 5 days after intraperitoneal injection of 2 ml thioglycollate (40 mg/ml), as described previously (14). The harvested cells were incubated in an erythrocyte lysis buffer (155 mmol/l NH₄Cl, 10 mmol/l KHCO₃, 0.1 mmol/l EDTA) for removal of residual erythrocytes before plating. All animal work was approved and performed by following the guidelines for the use and care of laboratory animals at the University of Washington.

Total RNA was isolated from quiesced hSMCs (passage 4–11), murine SMCs (passage 2–5), human monocyte–derived macrophages, or murine peritoneal macrophages, using an RNeasy Mini Kit (Qiagen, Valencia, CA). A total amount of 250 ng RNA was reverse-transcribed using random hexamers and Omniscript reverse transcriptase (0.2 units/μl; Invitrogen, Grand Island, NY). The mixture was incubated for 60 min at 37°C and heat-inactivated for 5 min at 94°C. The resultant cDNA template was amplified by PCR with specific oligonucleotide primers targeted for human and murine ACSL1, ACSL3, ACSL4, ACSL5, and ACSL6 (Table 1). Primers were designed to recognize all known splice variants, including the newly described splice variants for human ACSL1 and ACSL6 (15). PCR was performed using a GeneAmp PCR system (Applied Biosystems, Foster City, CA) with a primer concentration of 400 nmol/l. PCR products were then separated via gel electrophoresis in 2% agarose gels and visualized with ethidium bromide. Murine brain and liver were used as positive controls for the Acsl3, Acsl5, and Acsl6 primers. A human neuroblastoma cell line (SH-SY5Y) was used as a positive control for human ACSL6.

Lysates were prepared as previously described (12,14,16), separated using 7.5% Tris-glycine SDS-PAGE, and transferred to polyvinylidene difluoride membranes. The membranes were then incubated overnight at 4°C with rabbit polyclonal anti-rat Acsl1 or Acsl4 antibodies (1:10,000) generated against the amino acid sequences MEVHELFRYFRMPELIDIR and EIHSMQSVEELGSKPENSSI, respectively (17). Horseradish peroxidase–linked anti-rabbit secondary antibodies were used at a final dilution of 1:5,000 for 1 h at room temperature. Signal detection was performed by enhanced chemiluminescence. Recombinant rat Acsl4 (3 μg) was used as a positive control. Rat liver high-speed supernatant was used as a positive control for ACSL1 expression. Briefly, rat liver was homogenized (1:3; weight:volume) using a Potter-Elvehjem homogenizer in a 320 mmol/l sucrose buffer containing Tris-HCl, pH 7.4 (20 mmol/l), EDTA (1 mmol/l), EGTA (1 mmol/l), leupeptin (20 μg/ml), benzamidine (1 mmol/l), aprotinin (10 μg/ml), pepstatin (5 μg/ml), dithiothreitol (1 mmol/l), and phenylmethylsulfonyl fluoride (0.1 mmol/l). The homogenate was centrifuged at 7,700g for 2 min. The supernatant was then centrifuged at 12,000g for 12 min and then at 177,000g for 1 h. The resulting high-speed supernatant was used for analyses.

Total acyl-CoA synthetase activity was measured via a modified protocol by Kim et al. (10), involving the formation of OA-CoA from OA. Briefly, cells were scraped into ice-cold cell sonication buffer containing 50 mmol/l potassium phosphate (pH 7.4), 10% glycerol, 1 mmol/l EDTA, 20 μg/ml leupeptin, 5 μg/ml pepstatin, 10 μg/ml aprotinin, and 5 mmol/l benzamidine and sonicated with two bursts of 10-s duration at 50 W. Protein concentration was determined via BCA protein assays (Pierce, Rockford, IL). The lysate (200 μg/sample) was then incubated in the presence of increasing concentrations of rosiglitazone, triacsin C, 15-dPGJ₂, AICAR, or Wy16463 for 20 min at 37°C in a reaction mixture containing 175 mmol/l Tris (pH 7.4), 8 mmol/l MgCl₂, 5 mmol/l dithiothreitol, 10 mmol/l ATP, and 250 μmol/l CoA with 50 μmol/l OA trace-labeled with [³H]-OA (1 μCi) in a final volume of 200 μl. The reaction was initiated by the addition of total cell lysate and terminated by the addition of 1 ml Doles reagent (2-propanol:hexane:H₂SO₄; 80:20:2), followed by 2 ml heptane, 0.5 ml water, and vigorous vortexing. The upper layer was removed and the lower (aqueous) phase was washed with 2 ml heptane. The radioactivity of the lower phase was evaluated by a liquid scintillation counter. The radioactivity measured was used to calculate the total enzymatic activity and was expressed as picomoles of OA-CoA ester formed per unit time. The results were corrected for blanks (samples without cell lysates added and samples analyzed in the absence of CoA or ATP) and for protein content. All reactions were confirmed to occur in the linear range.

Analysis of fatty acid partitioning into cellular lipids was measured as [¹⁴C]-OA, [¹⁴C]-glycerol, and [³H]-AA incorporation into phospholipids, diacylglycerol (DAG), and triacylglycerol (TAG) after separation by thin-layer chromatography, as previously described (16). For analysis of AA incorporation, neutral and phospholipid spots were visualized by exposure to iodine vapor, scraped, and quantified via liquid scintillation spectroscopy.

The dominant-negative form of PPAR-γ harbors mutations of Leu⁴⁶⁸ and Glu⁴⁷¹ of the full-length PPAR-γ1 sequence into Ala and has been shown to block the effects of endogenous PPAR-γ in different cells (18,19). This PPAR-γ mutant retains ligand and DNA binding, but exhibits markedly reduced transactivation due to impaired co-activator recruitment (18). The dominant-negative PPAR-γ was subcloned into a recombinant type 5 adenovirus (Adx-D/N-PPAR-γ). An adenovirus expressing the LacZ gene was used as a control vector. Human SMCs were infected with 40 plaque-forming units/cell via a modified protocol by Bruemmer et al. (20).

Values are shown as means ± SE. Curve fitting and data analysis were performed using GraphPad Prism (GraphPad, San Diego, CA). Statistical significance of difference was assessed by either one-way or two-way ANOVA and, in some experiments, the Student's t test. Simultaneous multiple comparisons were based on a post hoc comparison test using a Student-Newman-Keuls test.

To evaluate the role of ACSLs in fatty acid partitioning in SMCs and macrophages, we characterized gene expression of ACSL isoforms present in human and murine SMCs, human monocyte–derived macrophages, and murine peritoneal macrophages. Using primers specific for each isoform (Table 1), we amplified mRNA for the different ACSL isoforms. Human SMCs and monocyte-derived macrophages expressed ACSL1, ACSL3, ACSL4, and ACSL5 mRNA (Fig. 1A). In hSMCs, the ACSL1 splice variant 1 and 2 primers detected a band of the correct size, whereas ACSL1 variant 3 was not detected (data not shown). Detectable levels of ACSL5 mRNA were found in hSMCs at cycle numbers above 30 (data not shown). Murine SMCs expressed Acsl1 and Acsl4 mRNA, and murine peritoneal macrophages expressed Acsl1, Acsl3, and Acsl4 mRNA. None of the Acsl6 splice variants were detected in any of the cell types studied (Fig. 1A).

Antibodies that recognize human and murine ACSLs are not yet available for all isoforms. However, an anti-rat Acsl1 antibody recognized a band of the predicted 68 kDa in high-speed supernatant from rat liver (Fig. 1B) and in human and murine SMCs, monocyte-derived macrophages, and murine peritoneal macrophages (Fig. 1B). The Acsl4 antibody recognized a band of 74 kDa in hSMCs that co-migrated with recombinant rat Acsl4 (Fig. 1C; upper panel). ACSL4 was expressed in hSMCs and monocyte-derived macrophages and in murine SMCs, while no band was seen in murine peritoneal macrophages (Fig. 1C; lower panel).

Previous studies have demonstrated that TZDs directly inhibit recombinant Acsl4 activity (10,21). This effect is selective to the Acsl4 isoform, because recombinant Acsl1, Acsl3, Acsl5, and Acsl6 are not inhibited by rosiglitazone until concentrations of >10 μmol/l are reached (10,21). Furthermore, rosiglitazone has not been reported to affect other enzymes with acyl-CoA synthetase activity, such as the fatty acid transport proteins. Triacsin C, an alkenyl-N-hydrozytrazene fungal metabolite, inhibits Acsl1, Acsl3, and Acsl4, but not Acsl5 and Acsl6 activity, at concentrations <10 μmol/l (10,21).

Initially, the ability of rosiglitazone to inhibit ACSL activity in cell lysates was evaluated. Rosiglitazone dose-dependently inhibited ACSL activity in both human and murine SMCs (Fig. 2A and E). The non-TZD PPAR-γ ligand 15dPGJ₂, the PPAR-α agonist Wy14643, and the AMPK activator AICAR had no effect on ACSL activity at the concentrations tested (Table 2). Rosiglitazone did not inhibit ACSL activity in murine peritoneal macrophages at concentrations up to 100 μmol/l, consistent with the nondetectable expression of Acsl4 protein in these cells (Fig. 2G). However, contrary to the effects seen in murine macrophages, rosiglitazone inhibited ACSL activity in human monocyte–derived macrophages (Fig. 2C).

Triacsin C dose-dependently inhibited total ACSL activity in both human and murine SMCs and macrophages, with similar half-maximal inhibition (IC₅₀) values (Fig. 2B, D, F, and H).

Because ACSL4 is likely to play an important role in fatty acid partitioning, we next investigated the effect of rosiglitazone on fatty acid partitioning in SMCs and macrophages. We had previously demonstrated, using thin-layer and high-performance liquid chromatography, that in SMCs, OA is incorporated into phospholipids (phosphatidylcholine, phosphatidylserine, and phosphatidylethanolamine), monoglycerides, DAG, and TAG (16,22). In this study, we analyzed the effects of rosiglitazone and triacsin C on the kinetics of OA incorporation into these lipid pools. As seen in our previous studies, [¹⁴C]-OA was incorporated into cellular phospholipids, DAG, TAG, and, to a small degree, cholesteryl esters in hSMCs (Fig. 3A and B). However, the kinetics of OA incorporation into these pools were distinctly different. OA incorporation into phospholipids occurred rapidly (<2 min), while incorporation into DAG was first observed at the 5-min time point and reached a maximum after 4 h (Fig. 3A and B). Incorporation into TAG was first observed at the 1-h time point (Fig. 3A and B) and reached a maximum at 6 h (data not shown).

In hSMCs, preincubation with rosiglitazone caused an initial decrease of ∼30% in OA incorporation into phospholipids (2–5 min after addition of OA; Figs. 3A and 4). This effect is most likely due to inhibition of phospholipid re-acylation because de novo phospholipid synthesis was undetectable at these time points (data not shown). Furthermore, in hSMCs (Figs. 3A and 5A) and macrophages (Fig. 5C), rosiglitazone inhibited [¹⁴C]-OA incorporation into DAG and TAG. Similarly, [¹⁴C]-glycerol incorporation into TAG was also inhibited by rosiglitazone in SMCs (65.2 ± 3.3% of untreated control, P < 0.01), indicating the attenuation observed is due to a decrease in de novo TAG formation. Similar results were seen with [³H]-AA–treated hSMCs (Fig. 5B). The non-TZD PPAR-γ ligand 15-dPGJ₂ (Fig. 5A, □), the PPAR-α ligand Wy14643, and the AMPK activator AICAR (data not shown) were without effect. Thus, rosiglitazone results in a transient inhibition of phospholipid reacylation and a more sustained inhibition of DAG and TAG synthesis in hSMCs. As expected, rosiglitazone had no effect on [¹⁴C]-OA incorporation into these lipid pools in murine peritoneal macrophages (Fig. 5D).

Triacsin C was used as a positive control to investigate the role of triacsin C–sensitive ACSL isoforms in SMCs and macrophages. Preincubation with triacsin C dramatically attenuated OA and AA incorporation into the DAG, TAG, and phospholipids (Figs. 3B, 4, and 5A and B) in hSMCs. As with rosiglitazone, the inhibitory effects of triacsin C on [¹⁴C]-OA incorporation into phospholipids were seen early (2–5 min) after addition of the fatty acid, at a time when [¹⁴C]-glycerol incorporation was not detectable, indicating that this effect was due to inhibition of phospholipid reacylation (Figs. 3B and 4). At later times, preincubation with triacsin C also attenuated [¹⁴C]-glycerol incorporation into phospholipids (8.6 ± 2.8% of control; P < 0.05), indicating that both reacylation and de novo pathways of phospholipid synthesis were inhibited.

As in hSMCs, triacsin C significantly attenuated [¹⁴C]-OA incorporation into the phospholipid, DAG, and TAG pools in human monocyte–derived macrophages (Fig. 5C). Similarly, in murine peritoneal macrophages, triacsin C inhibited [¹⁴C]-OA incorporation into phospholipids and TAG (Fig. 5D). These results suggest that triacsin C mimics some of the effects of rosiglitazone, but in addition has more dramatic effects on fatty acid incorporation into lipid pools, consistent with its ability to inhibit several ACSL isoforms.

To confirm that the effects of rosiglitazone on OA incorporation into DAG and TAG in hSMCs were not mediated by PPAR-γ activation, we used two approaches. First, hSMCs were treated with T0070907, a noncompetitive antagonist of PPAR-γ (23). Rosiglitazone caused a 50% decrease in OA incorporation into DAG, an effect that was not influenced by pretreatment with T0070907 (Fig. 6A). Interestingly, while treatment with the noncompetitive inhibitor alone attenuated OA incorporation into TAG, it had no effect on the rosiglitazone-induced effects. Second, we transfected hSMCs with an adenovirus containing a dominant-negative PPAR-γ mutant. Similar to T0070907, neither treatment with an adenovirus-containing LacZ, nor the dominant-negative form of PPAR-γ, reversed the inhibitory effects of rosiglitazone on OA incorporation into DAG and TAG (Fig. 6B). Conversely, the dominant-negative PPAR-γ mutant resulted in a significant stimulation of platelet-derived growth factor-BB–mediated [³H]-thymidine incorporation into DNA in these cells (27,580 ± 765 cpm/mg protein in hSMCs expressing LacZ versus 64,343 ± 5,087 cpm/mg protein in SMCs expressing dominant-negative PPAR-γ; n = 3; P < 0.05), consistent with the growth inhibitory actions of PPAR-γ (20).

The main finding of this report is that rosiglitazone inhibits ACSL activity and fatty acid partitioning into DAG and TAG in human SMCs and macrophages through a PPAR-γ–independent mechanism. Several findings point to a direct inhibition of ACSL4 enzymatic activity as a mechanism of action of rosiglitazone on fatty acid partitioning. First, recombinant rat Acsl4 activity is inhibited by TZDs through a mixed-type inhibition with respect to all of its substrates: fatty acid, CoA, and ATP (10). Of all the ASCL isoforms, only ACSL4 is inhibited by rosiglitazone (10,21). Second, human SMCs and macrophages and murine SMCs express ACSL4, and rosiglitazone inhibited ACSL activity in cell lysates, whereas murine macrophages did not express Acsl4 and were insensitive to the effects of rosiglitazone on acute fatty acid partitioning. Third, the effect of rosiglitazone on ACSL activity and fatty acid partitioning into DAG and TAG was not mimicked by the PPAR-γ agonist 15-dPGJ₂, the PPAR-α agonist Wy14643, or the AMPK activator AICAR. Finally, the inhibitory effects of rosiglitazone on OA partitioning into DAG and TAG were not blocked by pharmacological inhibition of PPAR-γ nor by inhibition of PPAR-γ action by expression of a dominant-negative PPAR-γ mutant. Thus, in addition to the well-described ability of rosiglitazone and other TZDs to activate AMPK (9,24) and p38 mitogen-activated protein kinase (25), ACSL4 appears to be an important contributor to the PPAR-γ–independent effects of rosiglitazone, at least in human and murine SMCs and in human macrophages. In fact, it is possible that the ability of rosiglitazone to activate AMPK might be due to its inhibitory action on ACSL4. ACSL4 is highly expressed in peroxisomes and mitochondrial-associated membranes and may be important in activating fatty acids for peroxisomal oxidation (26). Activity of AMPK, which is regulated by changes in the AMP:ATP ratio, might be upregulated by a reduced ACSL4 activity, leading to decreased fatty acid oxidation.

Rosiglitazone-induced inhibition of ACSL4 may have important biological effects in human macrophages and SMCs. Interestingly, AA is one of the preferred substrates for recombinant ACSL4 (27), and uterine tissues isolated from Acsl4^+/− mice show increased levels of prostaglandin E₂ (PGE₂) and 6-keto-PGF_1α (prostaglandin F_1α) (28). It is therefore possible that inhibition of AA re-acylation by rosiglitazone-mediated ACSL4 inhibition may lead to increased levels of bioactive eicosanoids in the vascular wall. Indeed, rosiglitazone increases lipopolysaccaride-induced prostanoid production in the human monocytic cell line U937 (29). On the other hand, the PPAR-γ–mediated effects of rosiglitazone are likely to increase cellular fatty acid uptake. For example, in adipocytes, rosiglitazone induces the expression of oxidized LDL receptor-1, which not only increases LDL uptake but also increases fatty acid uptake (30). CD36 is another PPAR-γ target gene induced by rosiglitazone in macrophages (31). CD36 can act as a fatty acid transporter (32), and increased expression therefore might increase fatty acid uptake. Overall, these studies show that several different pathways are likely to mediate the effects of rosiglitazone on fatty acid metabolism in macrophages and SMCs.

The concentrations of rosiglitazone used in the present study are similar to maximal plasma concentrations (1.1 μmol/l) achieved in humans by oral dosing of 8 mg/day (33) and are also similar to concentrations found to activate AMPK. Thus, rosiglitazone inhibits ACSL activity with an average IC₅₀ of 15.5 μmol/l in hSMCs (this study) and activates AMPK in the skeletal muscle cell line H-2K^b above 5 μmol/l (9). Therefore, the clinical effects of rosiglitazone, as well as other TZDs, may well be mediated not only by activation of PPAR-γ, but also by activation of AMPK and inhibition of ACSL4.

Based on differences in subcellular localization, regulation, tissue distribution, and fatty acid substrate preferences, it has been proposed that the different ACSL gene products channel fatty acids to different fates in the cell (34). Although this hypothesis will have to be carefully addressed, our data appear to support this notion, but also indicate that there is no absolute specificity in ACSL isoform function. Inhibition of ACSL4 by rosiglitazone resulted in an inhibition of fatty acid partitioning into DAG and TAG in both human SMCs and macrophages. Inhibition of ACSL4 has only a minor and transient effect on fatty acid partitioning into phospholipids. This appears to be primarily an effect on the re-acylation pathway, as our experiments demonstrated no [¹⁴C]-glycerol incorporation at these early time points (data not shown). In contrast, triacsin C, which inhibits ACSL1 and ACSL3 in addition to ACSL4, had dramatic effects on fatty acid partitioning into all lipid pools, including phospholipids. Although triacsin C might affect other fatty acid–binding proteins in the cell, these results suggest that ACSL4 is involved in de novo synthesis of DAG and fatty acid storage in TAG and that ACSL1 and/or ACSL3 have additional effects on fatty acid partitioning into phospholipids.

However, our results on murine macrophages show that in some cell types, one ACSL isoform appears sufficient to mediate fatty acid partitioning into all lipid pools, in addition to other fates, such as peroxisomal and mitochondrial β-oxidation. In these cells, although Acsl1 is the principal isoform expressed at detectable levels, murine macrophages efficiently incorporate fatty acids into all lipid pools.

Our results also demonstrate that human monocyte–derived macrophages express ACSL4 and exhibit sensitivity to rosiglitazone-inhibited fatty acid partitioning, whereas murine thioglycollate–elicited macrophages do not. While this study compared human monocyte–derived macrophages and murine thioglycollate–elicited peritoneal macrophages, preliminary studies in our laboratory show that bone marrow–derived macrophages from mice exhibit an Acsl expression profile very similar to that of murine peritoneal macrophages, indicating that the source of macrophages does not significantly alter the Acsl expression profile (J.E. Kanter and K.E. Bornfeldt, unpublished data). Important species differences have previously been demonstrated for other enzymes, such as the cyclic nucleotide phosphodiesterases (35).

TZDs have been shown to reduce atherosclerosis in diabetic and insulin-resistant mice (36,37). However, while rosiglitazone reduced the incidence of type 2 diabetes, it did not reduce the rate of myocardial infarction or stroke in the DREAM trial (38). On the contrary, patients receiving rosiglitazone developed more frequent congestive heart failure (38). One could speculate that the fact that rosiglitazone has several targets in addition to PPAR-γ, including ACSL4, might potentially explain the lack of beneficial cardiovascular effects in humans despite promising effects on atherosclerosis in mice (36,37). It will therefore be important to verify results obtained in murine models on isolated human macrophages and in clinical trials.

This study was supported by National Institutes of Health Grants HL62887 and HL076719 to K.E.B., by DK21723 and HL 44948 to J.A.B., and by DK59935 to R.A.C.

We would like to thank Dr. Harald Frankowski for providing us with RNA from SH-SY5Y cells and Dr. Xia Shen for designing the ACSL1_v1-2 and ACSL1_v3 primers.