HDL-bound sphingosine 1-phosphate acts as a biased agonist for the endothelial cell receptor S1P₁ to limit vascular inflammation

Abstract

The sphingosine 1-phosphate receptor 1 (S1P₁) is abundant in endothelial cells, where it regulates vascular development and microvascular barrier function. In investigating the role of endothelial cell S1P₁ in adult mice, we found that the endothelial S1P₁ signal was enhanced in regions of the arterial vasculature experiencing inflammation. The abundance of proinflammatory adhesion proteins, such as ICAM-1, was enhanced in mice with endothelial cell–specific deletion of S1pr1 and suppressed in mice with endothelial cell–specific overexpression of S1pr1, suggesting a protective function of S1P₁ in vascular disease. The chaperones ApoM⁺HDL (HDL) or albumin bind to sphingosine 1-phosphate (S1P) in the circulation; therefore, we tested the effects of S1P bound to each chaperone on S1P₁ signaling in cultured human umbilical vein endothelial cells (HUVECs). Exposure of HUVECs to ApoM⁺HDL-S1P, but not to albumin-S1P, promoted the formation of a cell surface S1P₁–β-arrestin 2 complex and attenuated the ability of the proinflammatory cytokine TNFα to activate NF-κB and increase ICAM-1 abundance. Although S1P bound to either chaperone induced MAPK activation, albumin-S1P triggered greater G_i activation and receptor endocytosis. Endothelial cell–specific deletion of S1pr1 in the hypercholesterolemic Apoe^−/− mouse model of atherosclerosis enhanced atherosclerotic lesion formation in the descending aorta. We propose that the ability of ApoM⁺HDL to act as a biased agonist on S1P₁ inhibits vascular inflammation, which may partially explain the cardiovascular protective functions of HDL.

INTRODUCTION

Vascular inflammation and endothelial dysfunction represent early events in cardiovascular disease (1). During vascular inflammation, endothelial function is altered, characterized by impaired nitric oxide release and an increase in the abundance of leukocyte adhesion molecules such as intercellular adhesion molecule–1 (ICAM-1) and vascular cell adhesion molecule–1 (VCAM-1), leading to the recruitment of proinflammatory immune cells into the vascular wall (2). Endothelial inflammation is affected by hemodynamic shear forces. In the descending aorta, laminar shear forces promote endothelial cell alignment, which suppresses inflammation. In contrast, in the aortic arch, the lesser curvature, and the branch points, turbulent flow induces random alignment of cells and vascular inflammation. Such regions are prone to atherosclerotic plaque formation (3). Plasma lipoproteins play a crucial role for either protection from or promotion of atherosclerosis (4). Indeed, altered concentrations of high-density lipoprotein (HDL) and low-density lipoprotein (LDL) represent a well-established risk factor for cardiovascular disease. High plasma concentrations of LDL injures the endothelium and promotes vascular inflammation (5, 6), whereas HDL is protective, partially by promoting endothelial function (7). Although the atheroprotective effect of HDL was initially attributed primarily to cholesterol removal from tissues, anti-inflammatory and endothelial protective functions of HDL may also be important (8). HDL particles are composed of apolipoproteins (such as ApoA-I, ApoA-II, and ApoM), enzymes involved in antioxidative mechanisms such as paraoxonase-1, lecithin-cholesterol acyltransferase, and diverse lipid species, including cholesterol esters, triglycerides, phospholipids (9), and bioactive sphingolipids such as sphingosine 1-phosphate (S1P) (10).

S1P is a lipid mediator that regulates many physiologic functions, especially those in the vascular and immune systems (11–13). It is concentrated in plasma (~1 µM) and carried primarily by the ApoM⁺ subfraction of HDL (~65%; ApoM⁺HDL) (14), with the remainder bound to plasma albumin (~35%). S1P signals through specific G protein (heterotrimeric guanine nucleotide–binding protein)–coupled receptors (GPCRs) [sphingosine 1-phosphate receptors 1 to 5 (S1P_1–5)] and regulates various cellular responses. In particular, S1P₁ inhibits vascular permeability and is required for vascular development (15) as well as immune cell trafficking. Moreover, a functional antagonist of S1P₁ (fingolimod/Gilenya) has been approved to treat relapsing-remitting multiple sclerosis (16). S1P₁ is abundant in endothelial cells and is critical for laminar shear stress signaling and stabilization of the nascent vascular network (17). The ability of various chaperones to modify the functions of S1P is not well understood.

The role of the S1P₁ receptor in vascular pathology is unclear. In nonobese diabetic (NOD) mice, S1P₁ is an anti-inflammatory receptor, its stimulation leads to reduced adhesion of monocytes to endothelial cells (18). In addition, S1P₁ stimulation in macrophages promotes an anti-inflammatory M2-like phenotype in vitro (19). In vivo, pharmacological agents (KRP203, an S1P₁-specific agonist, or FTY720, which targets all S1P receptors except S1P₂) attenuate atherosclerotic plaque development in Ldlr^−/− and Apoe^−/− mouse models (20–22). In contrast, stimulation of endothelial cells with S1P increases the surface abundance of adhesion molecules (VCAM-1 and ICAM-1), although it inhibits tumor necrosis factor–α (TNFα)–induced increases in the abundance of these molecules (23, 24). This inhibitory effect has been ascribed to S1P₁ and/or S1P₃ stimulation (24, 25). Additionally, S1P cooperates with lipopolysaccharide and induced inflammatory molecules and leukocyte adhesion on endothelial cells in vitro (26). These somewhat disparate findings underscore the need to clarify the role of this abundant lipid mediator in vascular inflammation and atherosclerosis.

Here, we report that mouse endothelial S1P₁ acted as an anti-inflammatory and anti-atherosclerotic GPCR in vivo. In addition, ApoM⁺HDL engaged human endothelial S1P₁ to form an anti-inflammatory signaling complex with β-arrestin 2 that suppressed nuclear factor κB (NF-κB)–dependent inflammatory pathways. Collectively, these data suggest that S1P bound to the chaperone ApoM⁺HDL activates the endothelial S1P₁ GPCR as a biased agonist to suppress inflammatory responses and atherosclerosis.

RESULTS

S1P₁ is differentially activated along the aortic endothelium

We have shown (17) that S1P₁ is localized on the plasma membrane in the descending aorta. In contrast, in the lesser curvature, S1P₁ is intracellular and colocalized with endosomal markers (17) (fig. S1A). The abundance of the proinflammatory cell adhesion molecules, ICAM-1 and VCAM-1 (fig. S1, B and C), was increased in the lesser curvature compared to the descending aorta. To study the role of S1P₁ signaling and its role in inflammation, we used the S1P₁ reporter mice [S1P₁–green fluorescent protein (GFP) reporter mice] (27). In these mice, endogenous activation of S1P₁ leads to the recruitment of a β-arrestin/TEV (tobacco etch virus) protease fusion protein, resulting in the release of a transcription factor that stimulates the expression of nuclear-localized GFP reporter. In the lesser curvature, subclavian-carotid bifurcation, thoracic branching point, and aortic valves (Fig. 1), substantial numbers of endothelial cells had nuclear GFP, suggesting an increase of S1P₁ signaling. In contrast, nuclear GFP was not detected in the regions of the aortic tree exposed to laminar shear stress such as medial aspect of the carotid branch point or descending aorta (Fig. 1). These results suggest a correlation between disturbed shear forces, inflammation, and exaggerated S1P₁ signaling/internalization.

Fig. 1. S1P₁ activation in the aortic endothelium exposed to disturbed shear forces.

Aortae from S1P₁-GFP signaling mice (27) were dissected, and nuclei were counterstained in en face preparations. Nuclear GFP abundance was analyzed by immunofluorescence in aortic valve (yellow lining), lesser curvature (red lining), subclavian-carotid bifurcation (purple lining), carotid artery (pink lining), descending aorta (green lining), and thoracic aorta branch point (blue lining). Images are representative of eight mice obtained in five independent experiments.

Endothelial S1pr1 expression suppresses vascular inflammation

To study the role of endothelial S1P₁ in vascular inflammation, we used the endothelial cell–specific inducible knockout mouse line (S1pr1^f/f Cdh5-Cre^ERT2, referred to as S1pr1ECKO) (17) and deleted endothelial S1pr1 at 4 weeks of age. Efficiency of deletion has been previously assessed by quantitative reverse transcription polymerase chain reaction analysis of endothelial cells isolated from lung tissue (17) as well as by immunofluorescence of the aortic tissue (Fig. 2A), which displayed an almost complete loss of S1P₁ immunofluorescence in the aortic endothelium. We did not detect differences in vascular endothelial (VE)–cadherin immunofluorescence between wild-type and S1pr1ECKO aortic endothelium. However, S1pr1ECKO aorta showed an increase in VCAM-1 and ICAM-1 immunofluorescence (Fig. 2, B and C), especially in the descending aorta.

Fig. 2. Genetic modulation of endothelial S1P₁ expression and inflammatory marker expression.

Aortae from wild-type (WT) (S1pr1^f/f) and S1pr1ECKO (S1pr1^f/f Cdh5-Cre^ERT2+/−) mice were dissected, and intima was stained in en face preparations. (A) Immunostaining of S1P₁ and VE-cadherin. S1P₁ immunofluorescence was undetectable in S1pr1ECKO aortae. (B and C) Immunostaining for VCAM-1 (B) and ICAM-1 (C) in the presence or absence of endothelial S1P₁. Images are representative of 10 mice per genotype obtained in 8 to 10 independent experiments. Aortae from WT (S1pr1^f/stop/f) and S1pr1ECGOF (S1pr1^f/stop/f Cdh5-Cre^ERT2+/−) mice were dissected, and intima was stained in en face preparations. (D) Immunostaining of S1P₁ and VE-cadherin. S1P₁ immunofluorescence was increased in S1pr1ECGOF aortae. (E) Immunostaining for ICAM-1 with and without endothelial S1P₁ overexpression. Images are representative of 10 mice per genotype obtained in five to seven independent experiments.

To further validate the anti-inflammatory effect of S1P₁, we immunoblotted dissected aortic extracts for ICAM-1 and VCAM-1. The abundance of ICAM-1 and to a lesser extent that of VCAM-1 was significantly increased in S1pr1ECKO aortae compared to control, especially in the descending aorta (fig. S2, A and B). Cell surface biotinylation assays of excised aortae showed increased ICAM-I abundance at the surface of S1pr1ECKO aorta (fig. S2C). Together, these results suggested that deletion of S1pr1 in endothelial cells promoted a proinflammatory state in the aortic endothelium.

To further confirm the involvement of S1P₁ in the control of vascular inflammation, we next used the inducible endothelial-specific S1P₁ overexpression mouse line (S1pr1^{f /stop/ f} Cdh5-Cre^ERT2, referred to as S1P₁ ECGOF) (17). Overexpression of endothelial S1pr1 was induced at 4 weeks of age. Efficiency of the overexpression was assessed by immunofluorescence of the aortic tissue (Fig. 2D), which showed a strong, albeit mosaic increase in S1P₁ staining. Overexpression of S1P₁ in the vascular endothelium led to increased membrane localization of S1P₁ in the lesser curvature of S1pr1ECGOF aorta. Similar to S1pr1ECKO aorta, we did not detect substantial differences in VE-cadherin staining. In contrast, S1pr1ECGOF animals showed a decrease in ICAM-1 staining (Fig. 2E) in both lesser curvature and descending aorta, suggesting that S1P₁ suppressed endothelial inflammatory gene expression.

HDL-bound S1P attenuates cytokine-induced increases in adhesion molecule abundance in endothelial cells

On the basis of the anti-inflammatory properties of endothelial S1P₁ in vivo, we hypothesized that S1P₁ induces an anti-inflammatory signal in endothelial cells. In human umbilical vein endothelial cells (HUVECs) treated with the proinflammatory cytokine TNFα, ICAM-1 abundance increased in a time-dependent fashion, a response that was attenuated (~70%) by co-incubation with HDL from wild-type mice (ApoM⁺ HDL) (Fig. 3, A and B). HDL particles contained ~4.79 ± 0.975 pmol S1P per microgram of protein, as determined by liquid chromatography–tandem mass spectrometry. We estimated the ApoM concentration in HDL preparations to be ~160 ng/µg protein. Thus, under our experimental conditions, HUVECs were exposed to ~1.5 to 2 µM S1P and ~2.5 µM ApoM. In contrast, HDL isolated from Apom^−/− mice (ApoM⁻HDL) did not significantly suppress the TNFα-induced increase in ICAM-1 expression. These HDL particles do not contain detectable ApoM and had very low S1P concentrations (0.038 ± 0.008 pmol S1P/µg protein). Albumin-S1P did not suppress the TNFα-induced increase in ICAM-1 abundance to the same extent as HDLS1P (Fig. 3, A and B). These results suggest that the anti-inflammatory effect of S1P depended on its association with ApoM⁺HDL. Dose-response analysis of HDL isolated from wild-type and Apom^−/− mice (Fig. 3, C and D) and albumin-S1P (fig. S3, A and B) further illustrated that the TNFα-induced increase in ICAM-1 abundance was inhibited in a dose-dependent manner by ApoM⁺HDL but not by ApoM⁻HDL particles or albumin-S1P. Flow cytometry revealed that VCAM-1 abundance was increased by TNFα treatment (fig. S3C), a response that was partially suppressed (~40%) by S1P bound to recombinant ApoM but not by recombinant ApoM alone, and to a lesser extent by albumin-bound S1P (~20%) (fig. S3C). Together, these data suggest that S1P bound to ApoMin HDL particles potently suppressed cytokine-induced increases in adhesion molecule abundance in endothelial cells.

Fig. 3. S1P chaperone–dependent inhibition of TNFα-induced increase in ICAM-1 abundance.

(A) Serum-starved HUVECs were stimulated with human recombinant TNFα for the indicated times, in the absence or presence of HDL isolated from WT (ApoM⁺HDL) or Apom^−/− (ApoM⁻HDL) mice or albumin (Alb)–S1P. ICAM-1 abundance was assessed by immunoblot of total cell lysates. (B) ICAM-1 abundance was quantified from three independent experiments. Data are shown as means ± SD. *P < 0.05 by two-way analysis of variance (ANOVA) followed by Dunnett’s multiple comparison (no treatment compared to each treatment for every time point). A.U., arbitrary unit. (C) Serum-starved HUVECs were stimulated with human recombinant TNFα and with increasing doses of ApoM⁻HDL or ApoM⁺HDL as indicated. ICAM-1 abundance was determined by immunoblot analysis. (D) Data are shown as means ICAM abundance ± SD from three independent experiments. *P < 0.05, compared to TNFα alone by Kruskal Wallis test followed by Dunnett’s multiple comparison.

Cytokine-induced NF-κB activation is attenuated by HDL-bound S1P

The NF-κB pathway is a key inflammatory signaling pathway induced by TNFα in endothelial cells, and its activation requires the phosphorylation of the p65 NF-κB subunit (28). TNFα stimulation of HUVECs increased the phosphorylation of p65 within 10 min, a response that was suppressed by co-incubation with human HDL (fig. S4) in a dose-dependent manner (Fig. 4, A and B). Similarly, HDL isolated from wild-type mice (ApoM⁺HDL) inhibited the phosphorylation of p65 in a dose-dependent manner, whereas ApoM⁻HDL was less effective (Fig. 4, C and D). These observations suggest that S1P bound to HDL antagonized the cytokine-induced NF-κBpathway in endothelial cells.

Fig. 4. HDL-S1P suppresses cytokine-induced NF-κB activation.

(A and C) Serum-starved HUVECs were stimulated with TNFα with or without increasing doses of human HDL (A) or WT mouse HDL (ApoM⁺HDL) and mouse HDL from Apom^−/− animals (ApoM⁻HDL) (C). Phosphorylation (p) of the NF-κB subunit p65 was analyzed by immunoblot of total (tot) cell lysates. (B and D) Data in (B) and (D) represent combined analysis from three independent experiments and are shown as means ± SD. *P < 0.05, **P < 0.01, compared to TNFα alone by one-way ANOVA followed by Dunnett’s multiple comparison.

HDL-S1P induces the persistent presence of S1P₁ at the cell membrane

HDL-S1P enhances endothelial barrier function by activating S1P₁ at the cell membrane (29).We investigated the localization of S1P₁ in HUVECs expressing GFP-tagged S1P₁ (14) and stimulated with various chaperonebound forms of S1P. GFP-tagged S1P₁ was localized at the cell membrane under basal conditions, which was not affected by TNFα stimulation or HDL treatment. In contrast, stimulation with albumin-S1P or exposure to P-FTY720 induced receptor endocytosis (Fig. 5A). Cell surface biotinylation assays (fig. S5, A and B) confirmed that albumin- S1P and P-FTY720, but not HDL treatment, induced the disappearance of membrane S1P₁. These results suggested that HDLS1P activation of endothelial cells promoted cell surface S1P₁, whereas albumin-S1P promotes rapid endocytosis of S1P₁.

Fig. 5. HDL-S1P acts as a biased agonist on S1P₁.

(A) HUVECs were transduced with lentivirus encoding GFP-tagged S1P₁ and after serum starvation, were stimulated with TNFα or with human HDL (huHDL), albumin- S1P, or P-FTY720 for the indicated times. Confocal microscopy was performed to assess the localization of GFP-tagged S1P₁. Images are representative of three independent experiments. (B) Serum-starved HUVECs were stimulated with albumin-S1P or HDL-S1P for the indicated times. MAPK phosphorylation was assessed by immunoblot of total lysates. (C) Serum-starved HUVECs were preincubated for 30 min with 3-isobutyl-1-methylxanthine (IBMX), and then stimulated with increasing doses of albumin-S1P, HDL-S1P, or ApoM⁻HDL in the presence of forskolin for 30 min. Results are expressed as picomole of cAMP per microgram of protein. Data represent combined analysis of five separate experiments and are expressed as means ± SD. **P < 0.01 by Friedman test followed by Dunnett’s multiple comparisons.

Cellular responses to S1P are chaperone-dependent

Activation of S1P₁ by its ligand S1P leads to G_i-dependent signaling, which activates the mitogen-activated protein kinase (MAPK) extracellular signal–regulated kinase (ERK) and inhibits adenylate cyclase (30). Both albumin-S1P and HDL-S1P induced rapid phosphorylation of ERK [Fig. 5B and (14)], in an S1P₁-dependent manner (fig. S6). In contrast, forskolin-activated increases in cyclic adenosine 3′,5′-monophosphate (cAMP) concentrations were inhibited by albumin-S1P, but not by HDL-S1P (Fig. 5C), suggesting that HDL-S1P acts as a biased agonist for S1P₁.

β-Arrestin 2 plays a crucial role in HDL-induced S1P₁ anti-inflammatory signaling

After ligand stimulation, β-arrestins associate with GPCRs to promote desensitization and endocytosis (31, 32). β-Arrestins also provide an additional GPCR signaling platform in addition to the heterotrimeric G protein–regulated mechanisms (32, 33), leading to the activation of intracellular signaling responses including the inhibition of the NF-κB pathway (34, 35). We therefore asked whether β-arrestin was involved in mediating endothelial cell anti-inflammatory pathways.

HDL, but not albumin-S1P, treatment increased the association of S1P₁ with β-arrestin 2 (Fig. 6A). In HUVECs, cell membrane localization of DsRed-tagged β-arrestin 2 was induced by HDL-S1P but not by albumin- S1P (fig. S7). The ability of HDL to suppress the phosphorylation of p65 after TNFα treatment was blocked by short hairpin RNA (shRNA)–mediated knockdown of β-arrestin 2 but was not affected by overexpression of β-arrestin 2 (Fig. 6, B and C).

Fig. 6. The role of β-arrestin 2 in the biased signaling of HDL-S1P.

(A) Serum-starved HUVECs were treated with albumin-S1P or human HDL-S1P for the indicated times. Cell lysates were subjected to immunoprecipitation for S1P₁ and immunoblotted (IB) with β-arrestin 2 antibody. Immunoblots are representative of three independent experiments. (B) HUVECs stably overexpressing β-arrestin 2 complementary DNA or an shRNA directed against β-arrestin 2 were serum-starved and stimulated with TNFα, with or without human HDL at the indicated doses. p65 phosphorylation was analyzed by immunoblot assay. (C) Data represent combined analysis of p65 phosphorylation from four independent experiments. *P < 0.05, **P < 0.01, ^#P < 0.005 by two-way ANOVA followed by Dunnett’s multiple comparison. (D) Serum-starved HUVECs were stimulated with TNFα in the absence or presence of albumin-S1P or ApoM⁻HDL or human HDL as indicated. In the P-FTY720–treated experiments, serum-starved HUVECs were pretreated with P-FTY720 in the absence or presence of albumin-S1P, ApoM⁻HDL, or human HDL before TNFα stimulation in the presence of P-FTY720. ICAM-1 abundance was assessed by immunoblotting. (E) Data represent the combined analysis of ICAM-1 abundance from three independent experiments. **P < 0.01 by unpaired Student’s t test.

To confirm the involvement of S1P₁ in this response, HUVECs were treated with the S1P₁ inhibitor P-FTY720, which attenuated the ability of HDL-S1P to suppress TNFα-induced increases in ICAM-1 abundance (Fig. 6, D and E). These data suggest that a specific ApoM⁺HDL-induced signaling complex consisting of S1P₁ and β-arrestin 2 was critical to inhibit cytokine-induced inflammatory responses.

Endothelial S1pr1 abundance influences atherosclerosis development

Exaggerated inflammation in the vascular wall promotes atherosclerosis in the presence of hypercholesterolemia and hyperlipidemia (36). To determine the role of endothelial S1P₁ in atherosclerosis, we compared Apoe^−/− S1pr1ECKO mice with Apoe^−/− S1pr1 wild-type littermates. As for S1pr1ECKO mice (Fig. 2C), ICAM-1 immunofluorescence was increased in Apoe^−/− S1pr1ECKO mice, especially in the descending aorta (fig. S8).

When placed on a high-fat diet for 16 weeks, weight gain (fig. S9A) and plasma cholesterol concentrations (fig. S9B) were similar between Apoe^−/− S1pr1ECKO and Apoe^−/− S1pr1 wild-type mice. In addition, monocyte (CD11b⁺, CD115⁺, Ly6G⁻), neutrophil (CD11b⁺, Ly6C^int, Ly6G^hi), T lymphocyte (CD4⁺ and CD8⁺), and B lymphocyte (CD19⁺) numbers were similar in Apoe^−/− S1pr1ECKO and Apoe^−/− S1pr1 wild-type mice (fig. S9, C and D). Also, plasma concentrations of ceramide and sphingosine species did not differ (fig. S10, A and B).

We next examined the development of atherosclerotic plaques using Oil Red O staining of en face preparations of the aorta (Fig. 7A). Quantification of plaque areas in the aortic arch (Fig. 7B) did not show significant differences. However, plaque development was significantly greater in the branch points of the descending aorta in Apoe^−/− S1pr1ECKO mice than in Apoe^−/− S1pr1 wild-type mice, likely as a consequence of increased ICAM-1 and VCAM-1 abundance in this specific area (Fig. 7B and fig. S8). Consequently, total plaque area was significantly higher in the Apoe^−/− S1pr1ECKO mice than in the Apoe^−/− mice (Fig. 7B). Histological analysis of the lesions with hematoxylin-eosin, Masson’s trichrome, or elastin staining indicated that plaque composition was similar in the two genotypes (fig. S11A). However, ICAM-1 staining was increased in plaques in Apoe^−/− S1pr1ECKO animals (fig. S11B). Plaque areas in aortic roots (fig. S12, A and B) were similar in Apoe^−/− S1pr1ECKO and Apoe^−/− S1pr1 wild-type mice. Although macrophage immunofluorescent staining in the aortic root did not appear to differ between the two genotypes (Fig. 7, C and D),more macrophage immunofluorescent signals were detected in plaques of the descending aorta in Apoe^−/− S1pr1ECKO mice (Fig. 7, E and F). Together, these results suggest that endothelial S1P₁ exerts a region-specific antiatherosclerotic effect in the descending aorta.

Fig. 7. Endothelial S1P₁ suppresses atherosclerotic lesion development.

Male Apoe^−/− S1pr1 WT and Apoe^−/− S1pr1ECKO littermates were placed on high-fat diet for 16 weeks. (A) Aortae were dissected; en face preparation followed by Oil Red O staining was done. Shown here is a representative image of aorta from each group. (B) Quantification of plaque areas in each section of aorta (aortic arch, descending aorta, and total aorta) is shown (six Apoe^−/− WT mice and eight Apoe^−/− S1pr1ECKO mice). ns, not significant. (C) Aortic root sections from Apoe^−/− S1pr1 WT and Apoe^−/− S1pr1ECKO mice were immunostained with MOMA-2 to visualize the infiltration of macrophages and foam cells. (D) Quantification of MOMA-2 infiltration in the aortic valve was assessed (n = 5 mice for each genotype). (E) Plaques from descending aorta of Apoe^−/− S1pr1 WT and Apoe^−/− S1pr1ECKO mice were isolated, and macrophage was visualized by MOMA-2 staining followed by confocal microscopy. (F) Quantification of MOMA-2 staining in the descending aorta lesion was assessed (n = 2 animals, two separate plaques analyzed for each aorta). Data are shown as means ± SD. *P < 0.05, **P < 0.01 by Mann-Whitney test.

DISCUSSION

Although sphingolipids, which function together with sterols in cellular membranes, are thought to be critical in cardiovascular disease, their specific functions in vascular pathophysiology remain unclear. S1P and its receptors are of particular interest, because the ligand is enriched in plasma, and the receptors are present on vascular and immune cells. Indeed, S1P₂ regulates myeloid cell retention in tissues during sterile inflammatory responses and in mouse models of atherosclerosis, and S1pr2 knockout mice exhibit reduced atherosclerotic plaques (37, 38). In contrast, S1P₁ stimulation of monocytes-macrophages induces an anti-inflammatory phenotype (19), suggesting a potential function of S1P receptors in myeloid cells during vascular disease. However, despite its high abundance in the endothelium, the role of S1P₁ during vascular inflammation and atherosclerosis is still unknown.

A major finding of this work is that S1P₁ localization on the plasma membrane of aortic endothelial cells was inversely correlated with inflammatory adhesion molecule abundance. The surface amount of S1P₁ was substantially lower in endothelial cells of the lesser curvature, which are exposed to disturbed flow, than in those in the descending aorta. Analysis of the S1P₁-GFP reporter mice revealed that areas exposed to laminar shear did not show detectable S1P₁ signaling, whereas areas exposed to disturbed shear forces showed high signaling, which correlated with enhanced internalization of S1P₁ in these areas. Supraphysiological activation of S1P₁ results in enhanced endocytosis followed by degradation of the receptor (39, 40). The reason for this gradient of signaling, despite a uniform distribution of the ligand in the circulation, is not clear. A possible explanation for the enhanced signaling could be the increased permeability of disturbed shear areas of the endothelium (41), which would allow plasma S1P to access S1P receptors on the basolateral surface of the endothelium. Alternatively, disturbed shear stress per se or local inflammatory response could induce exaggerated S1P₁ signaling.

Endothelial S1P₁ signaling was analyzed with tissue-specific genetic loss of function and overexpression in mice. S1pr1ECKO mice showed increased VCAM-1 and ICAM-1 abundance, whereas mice that overexpressed S1pr1 on endothelial cells had decreased abundance of these inflammatory markers in the aortic endothelium, suggesting that it suppressed inflammatory responses. In HUVECs, TNFα-induced NF-κB activation and adhesion molecule expression were suppressed by HDL-S1P, specifically ApoM⁺HDL.HDL-S1P induced the association of β-arrestin 2 with S1P₁ and promoted receptor localization on the plasma membrane, whereas S1P bound to albumin, its other chaperone in the plasma, suppressed forskolin-induced increases in cAMP concentrations. The ability of HDL-S1P to suppress inflammation depended on S1P₁, because P-FTY720 pretreatment, which triggers the degradation of S1P₁, antagonized it. However, both HDL-S1P and albumin- S1P activated the MAPK pathway in an S1P₁-dependent manner. Together, these data suggest that HDL-S1P acts as a biased agonist for S1P₁. The ability of specific chaperones to impart unique biological actions of S1P may be a general phenomenon, because we have described a unique function of HDL-S1P to suppress lymphopoiesis in the bone marrow (42). The mechanism underlying the differences in receptor activation by HDL-S1P and albumin-S1P is not known but could depend on co-receptors for HDL such as SR-B1 (25, 43).

Using defined HDL preparations (ApoM⁺HDL and ApoM⁻HDL), we showed that the association between S1P₁ and β-arrestin 2 correlated with the anti-inflammatory activity of HDL-S1P. In addition to the role of β-arrestins in the desensitization and internalization of GPCRs (33), they can act as G protein–independent signal transducers (44).For instance,β-arrestins can inhibit the proinflammatory cascade at different levels, by interacting either with inhibitor of nuclear factor κB α (IκBα) to enhance its stability (34, 35) or with the IκB kinase (IKK) complex directly (34). Moreover, studies on the receptor GPR120, which is activated by an ω-3 polyunsaturated fatty acid, showed that recruitment of β-arrestins (45) prevented downstream activation of the IKK complex and NF-κB activation. By using shRNA down-regulation in HUVECs, our data showed that β-arrestin 2 was essential for HDL-S1P to inhibit cytokine-induced inflammatory responses.

In the Apoe^−/− background, S1pr1ECKO mice show enhanced plaque size, especially at branch points that are subject to disturbed shear stress and show exaggerated S1P₁ signaling. These mice did not show alterations in plasma cholesterol or sphingolipids, suggesting that attenuated signaling of the S1P₁ receptor in the endothelial compartment is responsible for the accelerated atherosclerosis. This suggests that endothelial S1P₁ protects the vessel wall from inflammation and atherosclerosis. These observations are consistent with findings that show that P-selectin–dependent leukocyte rolling is enhanced in postcapillary venules of S1P₁ endothelial knockout mice (46). In conclusion, our results highlight ApoM⁺HDL chaperone– dependent anti-inflammatory and antiatherosclerotic function of S1P in the endothelium. This occurs through the activation of S1P₁, which transduces an anti-inflammatory signal in response to ApoM⁺HDL-bound S1P. The ability of the S1P chaperone ApoM⁺HDL to impart specific signal transduction mechanisms and an anti-inflammatory function in the endothelium may partially explain the cardiovascular protective functions of HDL.

MATERIALS AND METHODS

Mice

Animal experiments were approved by the Weill Cornell Medical College Institutional Animal Care and Use Committee. All experiments were done in male animals. S1pr1^f/f mice were crossed to Cdh5-Cre^ERT2 mice (a gift from R. Adams, Max Planck Institute) (47) to generate S1pr1^f/f Cdh5-Cre^ERT2 mice. Gene deletion was achieved by intraperitoneal injection of tamoxifen (150 µg/day) to mice at 4 weeks of age for five consecutive days. For the atherosclerosis lesion development studies, S1pr1^f/f Cdh5-Cre^ERT2 mice were crossed with Apoe^−/− mice (Jackson Laboratory) for >4 generations to obtain mice that were >95% C57Bl/6 background, as determined by single-nucleotide polymorphism analysis. Apoe^−/− S1pr1^f/f littermates (Cdh5-Cre^ERT2–negative) were used as controls. After S1P₁ deletion, male mice received high-fat chow (Teklad, TD88137; adjusted calories diet, 42% from fat) for 16 weeks.

S1pr1^f/stop/f was generated as described (17) and crossed to Cdh5-Cre^ERT2 mice to generate S1pr1^f/stop/f Cdh5-Cre^ERT2 mice. Gene overexpression was achieved by intraperitoneal injection of tamoxifen (150 µg/day) at 4 weeks of age for five consecutive days. Apom^−/− mice were a gift of L. B. Nielsen and C. Christoffersen of Rigshospitalet. S1P₁-GFP reporter mice have been previously described (27).

Cell culture

HUVECs (passages 3 to 8) (Clonetics) were grown in MEM1999 supplemented with 10% heat-inactivated fetal bovine serum (FBS), 10 mM Hepes, heparin-stabilized, penicillin (50 µg/ml), streptomycin (50 µg/ml), and endothelial cell growth factor as previously described (48). GFP-tagged S1P₁ overexpression was done as previously described (49).

En face protocol and staining

After cervical dislocation, the superior vena cava was opened, and mice were flushed with cold phosphate-buffered saline (PBS) containing 5% heparin and a cocktail of phosphatase inhibitors (1 mM Na₃VO₄, 1 mM NaF, 10 mM β-glycerophosphate) followed by intracardiac injection of cold 4% paraformaldehyde (PFA) in PBS. Aortae were dissected and then fixed for 20 min in 4% PFA. After a brief wash in glycine solution (100 mM), followed by a10-min incubation in 0.1% Triton X-100, arteries were carefully opened, and descending aorta, lesser curvature, and greater curvature were isolated. Tissues were then incubated for 2 hours in blocking solution [75 mM NaCl, 18 mM Na₃ citrate, 2% FBS, 1% bovine serum albumin (BSA), 0.05% Triton X-100] as described previously (50), followed by incubation with primary antibodies diluted in blocking solution at 4°C under constant agitation.

After a 24-hour incubation with primary antibodies, arteries were washed for 2 hours in washing solution (75 mM NaCl, 18 mM Na₃ citrate, and 0.05% Triton X-100), incubated with fluorescent secondary antibodies (Alexa Fluor dyes, Molecular Probes) for 1 hour, counterstained with TO-PRO-3, mounted, and then imaged using an Olympus FluoView FV10i confocal microscope.

In vivo aortic protein biotinylation

Male age-matched mice were anesthetized with ketamine/xylene. The mice underwent thoracotomy, and an incision was made in the right ventricle. The left ventricle was perfused with cold PBS for 5 min and then by cell-impermeable Sulfo-NHS-Biotin (1 mg/ml) for 10min using an infusion pump (VWR) at a constant rate of 1 ml/min. Excess biotin was rinsed out with consecutive 10-ml infusions of cold 50mMNH₄Cl in PBS followed by cold PBS alone. The thoracic aorta was harvested and immediately frozen in liquid nitrogen.

Immunoprecipitation

After the indicated treatments, HUVECs were washed in ice-cold PBS containing phosphatase inhibitors and lysed using modified radio-immunoprecipitation assay (RIPA) buffer [50 mM tris (pH 7.5), 150 mM NaCl, 1 mM EDTA, 1% NP-40, 0.5% sodium deoxycholate, 0.1% SDS, 1% Triton X-100, 0.5% FOS-choline, containing also 1 mM Na₃VO₄, 1 mMNaF, 10 mMβ-glycerophosphate, and protease inhibitor cocktail]. Lysates were then incubated for 20 min at 4°C using a tube rotator, and then centrifuged for 20 min at 13,500 rpm. Lysates were then cleared by protein A/G beads (Pierce Protein A Agarose, Thermo Scientific; Protein G Sepharose, GE Healthcare) and subjected to overnight immunoprecipitation using a monoclonal anti-S1P₁ immunoglobulin G antibody (H60, Santa Cruz Biotechnology). Bead-bound proteins were then separated on NuPAGE Novex 4 to 12% bis-tris gel (Invitrogen).

Cell surface biotinylation

After stimulation, cells were treated with a nonpermeable biotinylation agent (EZ-Link Sulfo-NHS-SS-Biotin) (Thermo Scientific) and lysed with a mild detergent following the supplier’s protocol. Labeled proteins were then isolated using NeutrAvidin Agarose beads (Thermo Scientific) and subjected to Western blotting analysis.

Oil Red O staining of atherosclerotic plaques

Aortic lipid content was analyzed and quantified using Oil Red O staining, as described previously (51). Briefly, after CO₂ asphyxiation, aortae were harvested from heart to iliac bifurcation, opened in situ to expose the intima, and fixed in 4% PFA overnight. Lipids in plaques were stained using 0.5% Oil Red O solution. Once stained, aortae were digitally photographed, and plaque quantitation was performed using the Image-Pro Analyzer software. Aortic root lesion was determined as described before (38).

Monocyte/macrophage staining of atherosclerotic plaques

Cryosections of aortic root were incubated for 1 hour in blocking solution (75 mM NaCl, 18 mM Na₃ citrate, 2% FBS, 1% BSA, 0.05% Triton X-100) before immunostaining using rat anti-mouse MOMA-2 (Serotec). After incubation with fluorescent secondary antibody and nuclear counterstaining using DAPI (4′,6-diamidino-2-phenylindole) solution, sections were imaged using confocal microscope system (Olympus, FV10i).

Plasma cholesterol quantification

Cholesterol was quantified using Cholesterol E CHOD-DAOS method kit (Wako) according to the manufacturer’s instructions. Results are expressed as microgram of cholesterol per deciliter of plasma.

Blood cell collection and flow cytometry

Mice were euthanized with CO₂ asphyxiation. Blood was recovered by hepatic portal vein puncture and collected in tubes containing 35 µl of 0.1 M EDTA. Whole blood (400 µl) was removed to a new tube, centrifuged at 2000g for 15 min, and plasma was removed and stored at −20°C. Cells were subjected to ammonium chloride erythrocyte lysis, washed with PBS, and stained with eFluor 718 fixable viability dye (eBioscience) according to the manufacturer’s instructions. Cells were then briefly fixed for 30 min in 0.1% PFA, washed, and Fc receptors were blocked with TruStain FcX (BioLegend) in 5% FBS overnight at 4°C. Cells were then stained with the following antibody mixtures for 30 min on ice: CD4–phycoerythrin (PE) (RM4-4; BioLegend), CD8- PerCP/eFluor 710 (H35-17.2; eBioscience), CD19–eFluor 450 (ID3; eBioscience), CD11b–APC–eFluor 780 (M1/70; eBioscience), Ly6C–FITC (fluorescein isothiocyanate) (HK1.4; Southern Biotechnology), Ly6G-PECy7 (1A8; BioLegend), and CD115-APC (AFS98; BioLegend). Liquid counting beads (BD) were added to each sample, and data were obtained on a BD LSR II (Becton Dickson) and analyzed using FlowJo software (Tree Star Inc.).

HDL isolation

HDL₃ was isolated by fractionation of outdated pooled plasma (New York Blood Center) or freshly isolated mouse plasma from wild-type and Apom^−/−mice. Sequential ultracentrifugation, using increasing density solutions, was used to separate the different lipoprotein fractions according to their flotation (52). Endotoxin contamination was assessed using the Limulus assay (Enzo Life Science). Albumin contamination was analyzed by Western blotting, and only endotoxin-free, albumin-free HDL₃ isolations were used.

Immunoblot experiments

HUVECs were washed with cold PBS containing phosphate inhibitors (1 mM Na₃VO₄, 5mMNaF, and 10 mM β-glycerophosphate) and lysed with modified RIPA buffer [1 mM ETDA, 50 mM tris (pH 7.5), 150 mM NaCl, 1% NP-40, 0.5% sodium deoxycholate, 0.1% SDS, 1% Triton X-100, 0.5% FOS-choline] containing phosphatase inhibitors (1 mM Na₃VO₄, 1mMNaF, and 10mMβ-glycerophosphate) and protease inhibitor cocktail (Calbiochem). After a freeze/thaw cycle, lysates were sonicated, protein concentrations were determined by bicinchoninic acid assay (Pierce), and samples were briefly denatured at 95°C for 5 min after addition of 10% β-mercaptoethanol. Equal amounts of protein were separated into 10% polyacrylamide gel and transblotted on polyvinylidene difluoride membrane. Transferred proteins were then probed with the antibodies as indicated. Antibodies were obtained from Cell Signaling (pp65 NF-κB, β-arrestin 2, phospho P44-42 MAPK, and total ERK) or Santa Cruz Biotechnology (S1P₁, total p65 NF-κB, ICAM-1, and VCAM-1).

cAMP measurement

Cells were starved for 3 hours. Cells were then preincubated for 30 min with 1 mM IBMX (Tocris Bioscience) followed by stimulation for 30 min with albumin-S1P, ApoM⁻ HDL, or ApoM⁺HDL, in the presence of IBMX (1 mM) and forskolin (20 µM) (Tocris Bioscience). cAMP measurement was then assessed using the Direct cAMP ELISA (enzyme-linked immunosorbent assay) kit (Enzo Life Sciences) following the supplier’s protocol.

Statistical analysis

Western blot quantification was done with ImageJ software (National Institutes of Health). Data are expressed as means ± SD. Statistical analysis was performed as mentioned using Prism software (Prism, GraphPad). P values <0.05 were considered significant.