Caveolin-1 regulates lipid droplet metabolism in endothelial cells via autocrine prostacyclin–stimulated, cAMP-mediated lipolysis

Lipid droplets (LD) are dynamic organelles involved in intracellular lipid metabolism in almost all eukaryotic cells, and LD-associated proteins tightly regulate their dynamics. One LD coat protein is caveolin-1 (Cav-1), an essential component for caveola assembly in highly differentiated cells, including adipocytes, smooth muscle cells, and endothelial cells (EC). However, the role of Cav-1 in LD dynamics is unclear. Here we report that EC lacking Cav-1 exhibit impaired LD formation. The decreased LD formation is due to enhanced lipolysis and not caused by reduced triglyceride synthesis or fatty acid uptake. Mechanistically, the absence of Cav-1 increased cAMP/PKA signaling in EC, as indicated by elevated phosphorylation of hormone-sensitive lipase and increased lipolysis. Unexpectedly, we also observed enhanced autocrine production of prostaglandin I2 (PGI2, also called prostacyclin) in Cav-1 KO EC, and this PGI2 increase appeared to stimulate cAMP/PKA pathways, contributing to the enhanced lipolysis in Cav-1 KO cells. Our results reveal an unanticipated role of Cav-1 in regulating lipolysis in non-adipose tissue, indicating that Cav-1 is required for LD metabolism in EC and that it regulates cAMP-dependent lipolysis in part via the autocrine production of PGI2.

Caveolins, including caveolin-1 (Cav-1), Cav-2, and Cav-3, are the major coat proteins driving caveola biogenesis (2). Among these isoforms, Cav-1 is necessary for caveola assembly in EC and adipocytes because blood vessels and adipose tissue isolated from Cav-1-deficient mice exhibit complete loss of caveolae (3,4). Caveolae have been implicated in many cellular processes, including cholesterol homeostasis, endocytosis, signal transduction, mechanosensing, and lipid metabolism (2). Physiologically, mice lacking Cav-1 are lean and exhibit lipodystrophy with elevated plasma triglyceride (TG) and cholesterol levels (5). Breeding of Cav-1-deficient mice to atherosclerotic mice markedly reduces the extent of atherosclerosis and reduces the infiltration of LDLs into the artery wall in the face of elevated plasma lipids (6,7). Importantly, reintroducing Cav-1 only into EC reverses these phenotypes, showing the significance of Cav-1 in EC during atherogenesis (7). These results suggest a potential role of Cav-1 in regulating endothelial lipid metabolism, although it has not been investigated in significant detail.
Lipid droplets (LD) are phospholipid monolayer spheres that function as intracellular sites for neutral lipid storage. LD are prominently expressed in energy-storing cell types such as adipocytes and hepatocytes; however, almost all eukaryotic cells are capable of forming and metabolizing LD (8). LD dynamics are regulated by the recruitment of key molecules promoting fatty acid storage and metabolism. For instance, diglyceride acyltransferase 2 (DGAT2), the rate-limiting enzyme catalyzing TG synthesis, translocates to the surface of LD and promotes TG storage into cytosolic LD (9). CTP:phosphocholine cytidyltransferase, a key enzyme of phospholipid synthesis, is also found on the surface of LD and can regulate LD size (10). Notably, Cav-1 has been localized to LD in several cell types (11)(12)(13), and in adipocytes, Cav-1 can regulate LD homeostasis in a perilipin (PLN)-dependent (14) and a lipid-induced mechanosensitive manner (15). However, it is still not fully understood how Cav-1 regulates LD dynamics, particularly in non-adipose tissue such as EC.
Our recent work demonstrated that EC exhibit the machinery to form LD and that these LD are coated with Cav-1 (16). In this study, we report that the loss of Cav-1 in EC reduces the number and TG content of LD and that this is due to enhanced ATGL-dependent lipolysis but not reduced TG synthesis or FA cro ARTICLE uptake. Mechanistically, the loss of Cav-1 increases cAMP/ PKA signaling in EC, potentially because of the autocrine production of PGI 2 . Consequently, our results document an unanticipated role of Cav-1 in regulating lipolysis in EC.

Cav-1 does not affect fatty acid uptake or TG synthesis in EC
A previous study in mouse embryonic fibroblasts demonstrated that Cav-1 regulates FA uptake via the levels of the fatty acid transporter CD36 (19). Therefore, we tested whether Cav-1 regulates LD formation by mediating FA uptake and CD36 levels in EC. The loss of Cav-1 had no effect on the timedependent uptake of [ 14 C]oleate into EC ( Fig. 2A). Immunoblotting of CD36 and FATP4, a major long-chain FA transporter in EC, showed comparable abundance among the genotypes (Fig. 2B), implying that endothelial Cav-1 regulates LD formation without affecting the FA uptake or expression of these FA transporters.
Next we tested whether Cav-1 influenced TG synthesis, thereby regulating LD formation. DGAT activity was measured

Caveolin-1 regulates lipid droplets in the endothelium
using [ 14 C]oleoyl-CoA and diacylglycerol (DAG) as co-substrates for TG synthesis, and the formation of radioactive TG was assessed by TLC. As seen in Fig. 3, A and B, the formation of TG was similar among the three cell types, as was the DAG dependence of the TG formation (Fig. 3, C and D). Additionally, analysis of the content of fatty acyl species in TG by mass spectrometry showed that Cav-1 KO EC had decreased content of all fatty acyl species measured in TG before and after OA treatment (Fig. 4A). Although Cav-1 KO EC exhibited slight differences in the percentage of C16.0 and C20.4 FA of TG when cultured with basal medium, the fatty acyl composition of TG was similar to WT and Cav-1 RC EC after OA treatment (Fig.  4B), indicating that the nature of the fatty acids in TG is not affected in the absence of Cav-1.

Lipolysis is augmented in the absence of Cav-1
Next we assessed whether the loss of Cav-1 influenced TG lipolysis. EC were loaded with OA, followed by removal of the fatty acid, and the time-dependent release of fatty acid and glycerol into the medium was examined. Indeed, Cav-1 KO EC released more FA and glycerol over time than WT EC, and the effect in Cav-1 KO EC was normalized in Cav-1 RC cells (Fig. 5, A and B). The increased metabolism of TG was supported by direct measurement of lipase activity, where Cav-1 KO EC had higher activity that was blunted by the adipocyte triglyceride lipase (ATGL) inhibitor Atglistatin (Fig. 5C). Finally, inhibiting ATGL during OA treatment resulted in normalized TG accumulation in all three types of EC (Fig. 5D) as well as the presence of LD via imaging (Fig. 5E), suggesting that loss of Cav-1 in EC enhanced lipolysis and consequently led to reduced LD formation.

Cav-1 negatively regulates lipolysis in a cAMP-PKAdependent mechanism
In adipocytes, lipolysis can be induced by agonists that couple via the G protein ␣ subunit G s to adenylate cyclase (AC) activating the cAMP-mediated PKA pathway. To examine whether Cav-1 influenced the cAMP-PKA pathway in EC, intracellular cAMP levels were measured during OA loading. As seen in Fig To determine whether the elevated cAMP/PKA signaling pathway regulated lipolysis in EC, PKA-mediated phosphorylation of hormone-sensitive lipase (HSL on Ser-563) (20, 21), a known regulator of lipolytic flux, was examined. EC were loaded with OA overnight, followed by stimulation with FSK for 0 -30 min. OA loading increased basal HSL and CREB-1 phosphorylation, and treatment with FSK increased phosphorylation further in Cav-1 KO EC compared with WT and Cav-1 RC EC ( Fig. 7A; quantified in Fig. 7, B and C). The increase in PKA activity was not due to changes in levels of PKA subunits, including catalytic PKAC␣ and regulatory RI ␣/␤ subunits (Fig.  7A). Next, FSK-mediated lipolysis (measured as glycerol release) in OA-loaded cells was measured. The loss of Cav-1 KO EC increased FSK-stimulated glycerol release (Fig. 7D), an effect attenuated by inhibition of PKA with H-89 (Fig. 7E). Similarly, H-89 treatment during OA treatment also increased TG accumulation in Cav-1 KO EC to a greater extent than in WT and Cav-1 RC (Fig. 7, F and G). Taken together, these results suggest that loss of Cav-1 in EC augments PKA activity, resulting in enhanced lipolysis and impaired LD formation.

Basal and OA-stimulated increases in PGI 2 contribute to lipolysis in EC
The enhanced basal levels of cAMP, PKA-mediated phosphorylation, and lipolysis in Cav-1 KO EC suggested that perhaps Cav-1 KO cells generated a mediator that promoted autocrine activation of the cAMP-PKA pathway. In EC, the major arachidonic acid metabolite generated via COX) is prostaglandin I 2 (PGI 2 , also called prostacyclin). PGI 2 binds to its cognate G protein-coupled receptor (IP), which couples through G s to AC and PKA activation (22). Indeed, Cav-1 KO EC produced more PGI 2 (as assessed by the stable metabolite 6-keto PGF 1␣) under basal and OA-stimulated conditions, which was abolished by blockade of COX using indomethacin (Indo) (Fig. 8A). Treatment with Indo or the IP receptor antagonist CAY10441 increased absolute and relative TG levels during OA loading in Cav-1 KO cells (Fig. 8, B and C). Interestingly, treatment with

Caveolin-1 regulates lipid droplets in the endothelium
CAY10441 reduced elevated FA and glycerol release in Cav-1 KO cells (Fig. 8, D and E). These data suggest that the autocrine production of PGI 2 acts as a stimulus for the activation of the cAMP/PKA pathways, contributing to the enhanced lipolysis in Cav-1 KO cells.

Discussion
The central finding of this study shows that Cav-1, an important structural component essential for the assembly of caveola organelles, is necessary for LD metabolism in EC and that it regulates cAMP-dependent lipolysis in part via the autocrine production of PGI 2 . Previous studies in adipocytes have highlighted the role of Cav-1 in lipid metabolism in regulating aspects of fatty acid uptake and LD size (23); however, the role of Cav-1 in regulating LD in EC has not been studied until now. Recent work has shown that Cav-1 is enriched in purified LD from EC, and LD in EC serve as reservoirs for neutralizing toxic lipids and can readily metabolize their TG to release FA for energy or utilization by other cells (16).
Global Cav-1 KO mice are lean with hypertriglyceridemia (5), and their adipocytes are smaller with less LD accumulation (13,14). Mechanistically, the reduction in LD size was linked to defective FA uptake through the fatty acid-binding protein CD36 in adipocytes and in fibroblasts (13,19,24). Moreover, the loss of Cav-1 reduces CD36 levels in whole-protein lysates prepared from Cav-1 KO aortic tissue (6). In contrast to previous work, the loss of Cav-1 in intact vessels and in isolated EC reduces LD formation in EC by virtue of enhancing LD degradation but not the uptake of FA or synthesis of TG, effects rescued by genetic reconstitution of Cav-1 back into EC. In isolated EC, the time-dependent uptake of labeled [ 14 C]OA as well as CD36 and FATP4 protein levels are not affected by the loss of Cav-1. In addition to TG synthesis data, MS data confirm that that the composition of fatty acyl groups within TG is not affected by the loss of Cav-1. However, we did not test whether phospholipid composition is altered in LD isolated from Cav-1 KO EC, as demonstrated previously in Cav-1-deficient adipocytes (25).
Surprisingly, the metabolism of TG quantified by the release of glycerol and FA is enhanced in Cav-1 KO EC, explaining why Cav-1 KO EC have fewer TG-rich LD. Enhanced lipolysis during OA loading in Cav-1KO EC is normalized by chemical inhibition of ATGL, implying that Cav-1 modulates LD formation by regulating lipolysis. The loss of Cav-1 in EC enhances OAand FSK-stimulated cAMP levels and PKA substrate phosphorylation (including HSL, CREB, and eNOS), and enhanced lipol-

Caveolin-1 regulates lipid droplets in the endothelium
ysis in Cav-1 KO EC is normalized by inhibition of PKA. These findings in EC are in contrast to prior work showing blunted lipolytic responses in Cav-1 KO adipocytes in response to ␤3 receptor agonists or FSK. Impaired lipolysis in Cav-1-deficient adipocytes was mediated via defective cAMP/PKA-dependent phosphorylation of PLN1 and HSL (14,26). Because PLN1 is expressed exclusively in adipocytes and is not found in EC (16), a different mechanism must exist in non-adipose cell type such as EC. Although PLN2 (formally named ADRP) is thought to be functionally equivalent to PLN1 in non-adipocytes, the molecular regulation of ADRP in lipolysis remains unclear, including its regulation by PKA signaling and HSL. Nevertheless, our results clearly demonstrate that HSL protein levels are elevated in Cav-1-deficient EC and that its phosphorylation is enhanced by FSK treatment, demonstrating a unique mechanism of Cav-1 regulation of lipolysis in endothelium versus adipocytes.
The increase in basal and OA-stimulated cAMP in Cav-1 KO EC prompted the search for factors that are produced by EC and regulate cAMP signaling. We show that OA treatment enhances 6-keto PGF1␣ release in EC, an unexpected link between FA metabolism and prostacyclin production. This idea is supported by a recent study showing that storage-operated calcium entry regulates lipolysis via cAMP up-regulation (28); thus it is possible that enhanced PGI 2 production occurs as a result of enhanced extracellular calcium entry and adenylyl cyclase activation during OA loading. Furthermore, our results show marked release of 6-keto PGF1␣ in Cav-1 KO EC, suggesting that Cav-1 can regulate aspects of arachidonic acid (AA) release or COX metabolism. A relationship between Cav-1 and AA metabolism in EC has been suggested because COX-2 expression and PG metabolites are elevated in lungs from Cav-1-deficient mice because of loss of COX-2 degradation (29). Moreover, a recent MS study indicated that silencing Cav-1 in EC increases intracellular FFA, including AA metabolites (30). Importantly, in our experiments, both indomethacin and the IP receptor antagonist rescue LD formation in Cav-1 KO EC, implying that autocrine production of PGI 2 can contribute to regulation of PKA and that lipolysis in EC and Cav-1 functions as an endogenous break on the system.
In conclusion, we describe a novel pathway for Cav-1mediated cAMP activation in EC and a connection between prostacyclin production and lipolysis. Perhaps elevated prostacyclin and the attendant changes in EC lipolysis contribute to the well-established findings that Cav-1 KO mice are protected from atherosclerosis in the face of elevated lipids (6, 7). Additional experiments examining LD metabolism in EC-specific conditional Cav-1-deficient mice will improve our understanding of how Cav-1 regulates LD homeostasis during atherogenesis.

Reagents and antibodies
Oleic acid, Atglistatin, forskolin, CAY10499, CAY10441, H-89, isobutylmethylxanthine, and indomethacin were purchased from Cayman Chemical. BODIPY 493/503 was obtained from Life Technologies. The antibodies used in this study were from the following resources, with the indicated dilution for Western blotting: anti-Cav-1 (BD Biosciences,   (n ϭ 4ϳ5). Statistical analysis in A and B was performed by two-way ANOVA. Statistical analysis in C and D was determined by unpaired t test. N.S., non-significant. *, p Ͻ 0.05 relative to WT and Cav-1 RC.

Animals
Cav-1 Ϫ/Ϫ and Cav-1 Ϫ/Ϫ mice carrying the canine Cav-1 transgene under the preproendothelin-1 promoter (Cav-1 Rec ) were generated as reported previously (17,18). All mouse strains were congenic on a C57Bl6 background and maintained on pelleted rodent chow with free access to water. Mice were maintained in a specific pathogen-free animal facility with a steady-temperature room (25°C) under a fixed 12-h light/dark cycle. All procedures were approved by the Institutional Animal Care and Use Committee of Yale University.

Caveolin-1 regulates lipid droplets in the endothelium
bated at 4°C for 20 min and centrifuged at 14,000 rpm for 10 min. Supernatants were collected, and protein concentration was measured with the DC TM protein assay kit (Bio-Rad.) Equal amounts of protein in each experiment were resolved by SDS-PAGE and transferred to nitrocellulose membranes for 2 h at 90 V using Mini Trans-Blot Cell (Bio-Rad.) Membranes were blocked using 0.1% (w/v) casein (Bio-Rad) in TBS for 50 min and incubated with primary antibodies overnight at 4°C. Membranes were washed with TBS-T (TBS solution containing 0.1% Tween 20) and incubated with conjugated secondary antibodies for 30 min at room temperature. Membranes were washed with TBS-T and developed using the Odyssey system (Li-Cor.) The densitometry of membranes was analyzed with ImageJ software (National Institutes of Health).

Lipid droplet formation and detection in vessels
Whole-mount thoracic aortas were isolated from 8-to 10-week-old WT, Cav-1 Ϫ/Ϫ , and Cav-1 Rec mice and incubated with OA conjugated to FA-free albumin (1 mM) in complete EBM-2 medium overnight in a 37°C incubator with 5% CO 2 , followed by en face immunostaining as described previously (16). Aortae were cut longitudinally, pinned down with EC facing upward, washed three times with PBS, and fixed with 4% paraformaldehyde in PBS. Fixed samples were further blocked with TNB blocking buffer (0.1 M Tris-HCl (pH 7.5), 0.15 M NaCl, and 0.5% (w/v) blocking reagent (PerkinElmer Life Sciences, FP1020)) overnight at 4°C. Aortae were further incubated with anti-PECAM-1 antibody (EMD Millipore, MAB1398Z, 1:250) diluted in TNB blocking buffer overnight at 4°C. Samples were washed three times with PBS and incubated with fluorescence-conjugated secondary antibody (Cy TM 3 AffiniPure Goat Anti-Armenian Hamster, Jackson ImmunoResearch Laboratories, 127-165-099, 1:200) at room temperature for 3 h. BODIPY 493/503 diluted in PBS at a final concentration of 4 g/ml was applied for 30 min to delineate LD, and DAPI (Sigma, 0.1 ng/ml) was used to high-

Caveolin-1 regulates lipid droplets in the endothelium
light the nucleus. Samples were mounted and imaged by confocal microscopy (Leica SP5).

Lipid droplet formation and detection in cultured EC
EC were grown to confluence on coverslips precoated with 0.1% gelatin in PBS solution. After the designated treatment, cells were washed with 3ϫ PBS and fixed with 4% paraformaldehyde in PBS solution for 15 min. Fixed cells were washed with 3ϫ PBS and stained with BODIPY 493/503 diluted in PBS at a final concentration of 0.1 g/ml for 15 min to delineate LD and DAPI (Sigma, 0.1 ng/ml) to highlight the nucleus. Coverslips were mounted with Fluoromount TM aqueous mounting medium (Sigma) and imaged by laser-scanning confocal microscopy (Leica SP5) in the sequential scan mode with an HCX PL APO blue 63ϫ/1.40 oil objective lens at room temperature and analyzed with ImageJ software.

TG measurements
TG content was determined by mass spectrometry (Fig. 1C). TG content (Figs. 5D, 7F, and 8B) was determined by a colorimetric assay kit (BioVision Inc.) following the instructions of the manufacturer and normalized by total protein amount in each sample (determined by DC TM protein assay kit (Bio-Rad).

Fatty acid uptake assay
The measurement of fatty acid uptake was performed as described previously with modifications (31). In brief, trace amounts of [ 14 C]oleic acid (PerkinElmer Life Sciences) mixed with non-radioactive oleic acid were dissolved in 10% free fatty acid BSA solution to a final concentration of 12 mM. This stock solution was further diluted with complete EBM-2 medium to a concentration of 1 mM and incubated with the confluent WT/Cav-1 KO/Cav-1 RC EC monolayer for 1, 2, 3, 4, and 24 h. The medium after incubation was collected, and aliquots were used to determine the remaining radioactivity. The uptake reaction was stopped by aspiration of the solution followed by addition of ice-cold solution containing 0.5% BSA and 200 M phloretin for 2 min. Cells were washed three times with ice-cold PBS and lysed with 1 M NaOH. Protein concentration was determined using the DC TM protein assay. Radioactivity in cell lysates and medium were determined with the addition of scintillation mixture solution (American Bioanalytical) using Tri-Carb 2900TR liquid scintillation analyzers. The percentage of fatty acid uptake was determined by radioactivity in cell lysates over total radioactivity added, normalizing to protein amount between each cell type.

DGAT activity assay
Enzymatic activity of DGAT was performed as described previously (32). 50 g of protein lysate (in 250 mM sucrose and 50 mM Tris-HCl (pH 7.4) supplemented with complete protease inhibitors (Roche)) was incubated at 37°C with 200 M diacylglycerol (dissolved in acetone) and 25 M [ 14 C]oleoyl-CoA. The reaction was stopped by methanol:chloroform (3:1), followed by addition of H 2 O. The organic phase was extracted and dried by N 2 . Dried lipids were suspended in chloroform and further subjected to TLC using hexane:diethylether:acetic acid (80:20:1) as solvent. The TLC plate was developed with a storage phosphor screen (Molecular Dynamics). Radioactivity was detected and imaged by Typhoon Trio Imager (GE Healthcare). The densitometry of TLC images was determined by Image Quant software (GE Healthcare).

Mass spectrometry
The TG content and molecular species distribution was determined using electrospray ionization-MS analysis as described previously with minor modifications (33)(34)(35). Briefly, lipids were extracted from EC using a modified Bligh and Dyer

Caveolin-1 regulates lipid droplets in the endothelium
technique in the presence of the internal standard tri-17:1 TG. Because of the low abundance of TG species in non-OA-loaded cells, TG species were enriched into the hexane fractions through liquid/liquid partitioning using hexane-methanolwater (36). Electrospray ionization-MS analyses of hexane fractions (non-OA-loaded cells) or chloroform lipid extracts (OA-loaded cells) were performed utilizing a TSQ Quantum Ultra triple-quadrupole mass spectrometer (Thermo Fisher Scientific, San Jose, CA) equipped with an automated nanospray apparatus (Nanomate HD, Advion Bioscience Ltd., Ithaca, NY). TG molecular species were analyzed in the positive ion mode as Na ϩ adducts or alternatively as Li ϩ adducts (in the presence of a small amount of LiOH). TG molecular species identification and fatty acyl compositional analysis were performed in positive ion mode as Li ϩ adducts or NH 4 ϩ adducts (in the presence of 10 mM ammonium acetate) using tandem mass spectrometry at collision energies of 35 or 25 electron volt, respectively.

Measurement of FFA and glycerol release
WT/Cav-1 KO/Cav-1 RC EC were grown to confluence and treated overnight with OA (1 mM). Medium containing OA was aspirated, washed with 2ϫ PBS, and placed into serum-free EBM-2 medium containing 50 M FFA-BSA. 1% of the medium was collected at the designated time points. FFA and glycerol concentration in the collected medium were determined using colorimetric assay kits (BioVision Inc.). FFA and glycerol release was normalized by total protein content in each sample, determined by DC TM protein assay kit (Bio-Rad).

Lipase activity assay
General lipase activity was measured based on the lipase activity colorimetric assay kit (Bio Vision Inc.) In brief, MLEC of the three genotypes were loaded with 1 mM OA overnight to induce lipolysis. Lysates were collected with the designated buffer, and protein concentration was determined by DC TM protein assay kit (Bio-Rad). 50 g of total protein lysates was used in the reactions containing exogenous TG as substrates. Lipase activity was determined based on glycerol formation rate following the instructions of the manufacturer.

Cyclic AMP assay
EC were grown to confluence. After designated treatments, cells were lysed with 0.1 M HCl and chilled on ice. Lysates were collected, and cAMP was acetylated by addition of KOH and acetic anhydride. Protein concentration was determined by DC TM protein assay kit (Bio-Rad). Cyclic AMP was determined by cyclic AMP EIA kit (Cayman Chemical) and normalized by total protein content.

6-Keto PGF1-␣ assay
EC were grown to confluence, and the medium was replaced with medium or without 1 mM OA as well as with or without 10 M indomethacin or 10 M CAY10441. Conditioned medium was collected, and total protein amount was determined as described under "Western blot analysis". 6-keto PGF1-␣ concentration in the conditioned medium was determined by 6-keto PGF1-␣ EIA kit (Cayman Chemical) and normalized by total protein content.

Statistical analysis
Statistical analyses were performed using Prism 6 software (GraphPad) using tests as described in each figure. The data are presented as mean Ϯ S.E.