High Density Lipoproteins (HDL) Interrupt the Sphingosine Kinase Signaling Pathway

The ability of high density lipoproteins (HDL) to inhibit cytokine-induced adhesion molecule expression has been demonstrated in their protective function against the development of atherosclerosis and associated coronary heart disease. A key event in atherogenesis is endothelial activation induced by a variety of stimuli such as tumor necrosis factor-α (TNF), resulting in the expression of various adhesion proteins. We have recently reported that sphingosine 1-phosphate, generated by sphingosine kinase activation, is a key molecule in mediating TNF-induced adhesion protein expression. We now show that HDL profoundly inhibit TNF-stimulated sphingosine kinase activity in endothelial cells resulting in a decrease in sphingosine 1-phosphate production and adhesion protein expression. HDL also reduced TNF-mediated activation of extracellular signal-regulated kinases and NF-κB signaling cascades. Furthermore, HDL enhanced the cellular levels of ceramide which in turn inhibits endothelial activation. Thus, the regulation of sphingolipid signaling in endothelial cells by HDL provides a novel insight into the mechanism of protection against atherosclerosis.

Numerous evidence from epidemiological, clinical, and genetic studies have clearly shown a potential protective role of high density lipoproteins (HDL) 1 against the development of atherosclerosis and associated coronary heart disease (1)(2)(3)(4). Several mechanisms have been proposed for the cardioprotective function of HDL. These include the promotion of the efflux of cholesterol from atherosclerotic plaques and reducing the atherogenicity of LDL by inhibition of their oxidative modification (4 -6). Recently, we and other groups have demonstrated an ability of HDL to inhibit endothelial adhesion protein expression, providing a new mechanistic explanation for its protective effect on atherosclerosis (7)(8)(9)(10)(11).
Atherosclerosis has been definitely characterized as an inflammatory disease (12). An important event in the initiation of atherosclerosis is adhesion of circulating monocytes to activated endothelial cells and their subsequent transendothelial migration to the subendothelium. This process is mediated by adhesion molecules such as vascular cell adhesion molecule-1 (VCAM-1), intercellular adhesion molecule-1 (ICAM-1), and E-selectin (13,14). The inappropriate expression of these adhesion proteins in response to the "injury" are induced by various inflammatory stimuli, including cytokines and noncytokines such as interleukin-1, tumor necrosis factor-␣ (TNF), and oxidized or native LDL (14 -16). Pathological studies have shown increased adhesion molecule expression in several components of the atherosclerotic plaque (17)(18)(19)(20), and there is also evidence for a role of adhesion molecules in the acute atherothrombotic process (21). Furthermore, a direct association between an increased plasma concentration of soluble adhesion molecules and the increase in risk of future cardiovascular diseases has recently been reported (22,23).
The ability of HDL to inhibit the cytokines-induced adhesion protein expression has been well documented. It has been reported that human HDL profoundly inhibit the expression of VCAM-1, ICAM-1, and E-selectin in human umbilical vein endothelial cells (HUVEC) activated by TNF or interleukin-1 (7). Total native HDL together with both HDL 2 and HDL 3 subfractions, or the reconstituted HDL particles showed the inhibitory effect in a concentration-dependent manner, although considerable variation existed among different experiments (7)(8)(9)(10)(11). The phenotype of inhibition on adhesion molecule expression by HDL differs from their well known function in promoting cholesterol efflux and protecting against lipid peroxidation, suggesting a distinct mechanism exists. We recently demonstrated a novel signaling pathway, sphingosine kinase (SphK) pathway, through the generation of sphingosine 1-phosphate (S1P), which is critically involved in mediating adhesion protein expression and endothelial cell activation after TNF stimulation (24). The SphK pathway has also emerged as a signaling pathway in mediating a variety of cellular functions such as cell growth, proliferation, and inflammatory reaction (24 -29). In the present report we show that HDL profoundly inhibit the TNF-induced SphK activity and S1P generation, and subsequently reduce the activation of ERK and NF-B signal cascades. We thus demonstrate that HDL interrupt a signaling cascade, the SphK pathway, which results in inhibition of endothelial activation. This could provide a new potential mechanism by which HDL protect against atherosclerosis, a cardiovascular inflammatory disease. Cell Culture and Flow Cytometry Analysis-HUVEC were isolated as described previously (30) and cultured on gelatin-coated culture flasks in Dulbecco's modified Eagle's medium containing 20% fetal calf serum, endothelial growth supplement (Collaborative Research) and heparin. In some experiments, cells were treated in Opti-MEM (Life Technologies, Inc.) containing 0.1% fatty acid-free bovine serum albumin as serum-free medium. The expression of cell-surface adhesion molecules was measured as described previously (24) by use of a Coulter Epics Profile XL flow cytometer.

Materials
Isolation and Preparation of Lipoproteins-As described previously (10), the lipoproteins were isolated from normal healthy adult donors by sequential ultracentrifugation in their appropriate density range: total HDL 1.07 Ͻ d Ͻ 1.21, HDL 3 1.13 Ͻ d Ͻ 1.21, and LDL 1.019 Ͻ d Ͻ 1.063 g/ml. The resulting preparations of lipoproteins were dialyzed against endotoxin-free PBS (pH 7.4) prior to use. Oxidized LDL was obtained by incubating LDL (500 g/ml) with confluent cultures of HUVEC in Dulbecco's modified Eagle's medium containing 10 M CuSO 4 for 24 h, and the oxidation was assessed by the increase of mobility on 1% agarose gel. Discoidal reconstituted HDL containing apoA-I and 1-palmityl-2-oleylphosphatidylcholine (POPC) were prepared by the cholate dialysis method described by Matz et al. (31).
125 I-TNF Binding Assay-The binding assay was performed in confluent HUVEC after preincubation with an increasing concentration of HDL 3 for 4 h. Cells were washed with M199 medium and incubated with 1 nM 125 I-TNF in the absence or presence of 500 nM unlabeled TNF in M199 medium containing 10% fetal calf serum. After 4 h of incubation at 4°C, cells were washed three times with ice-cold M199 and then solubilized in 1 N NaOH, 1% SDS and radioactivity was determined in a ␥-counter. Specific binding is defined as the difference between total binding and nonspecific binding with excess unlabeled TNF.
Metabolic Labeling and Sphingolipids Assay-HUVEC were labeled with [ 3 H]serine (10 Ci/ml) for 48 h as described previously (24). The cells were then washed three times and incubated for an additional 2 h in the presence or absence of HDL 3 . After treatment with TNF for the indicated times, cellular lipids were extracted and resolved by thin layer chromatography (TLC) in two different solvent systems: (a) chloroform:methanol:acetic acid:water (50:30:8:5, v/v) and (b) 1-butanol: acetic acid:water (3:1:1, v/v). The samples were concomitantly run with standard sphingolipids including sphingomyelin, ceramide, sphingosine, and S1P. Sphingolipid spots were visualized by fluorography, quantitated by scintillation spectrometry, and normalized by radioactivity recovered in total cellular lipids.
Ceramide Measurement-In addition to metabolic labeling assay as described above, cellular ceramide was quantified with the diacylglycerol kinase reaction as described previously (32). Briefly, the cellular lipids were extracted with CHCl 3 /CH 3 OH/HCl (1 N) (100:100:1) and resuspended into a sample buffer containing 7.5% n-octyl-␤-D-glucopyranoside, 5 mM cardiolipin, and 1 mM diethylenetriamine-pentaacetic acid. The samples were reacted with diacylglycerol kinase and [␥-32 P]ATP in enzyme buffer containing 20 mM Tris/HCl (pH 7.4), 10 mM dithiothreitol, and 15% glycerol for 30 min at 22°C. The product of the phosphorylation reaction, ceramide 1-phosphate, was extracted and resolved by TLC using CHCl 3 /CH 3 OH/HAc (65:15:5) as solvent, detected and quantified by the Phosphoimage system. To exclude a possible error caused by some factors in the extracts affecting diacylglycerol kinase (33), synthetic C 2 -ceramide was added in assays as an internal control. There were no changes in the phosphorylated C 2ceramide in this assay system. Measurement of SphK Activity-As described previously (24), cells were homogenized in 20 mM Tris buffer (pH 7.4) containing 20% glycerol, 1 mM dithiothreitol, 1 mM EDTA, 20 M ZnCl 2 , 1 mM Na 3 VO 4 , 15 mM NaF, 10 g/ml leupeptin and aprotinin, 1 mM phenylmethylsulfonyl fluoride, and 0.5 mM 4-deoxypyridoxine. After centrifugation at 13,000 ϫ g for 30 min, SphK activity was measured in the supernatant by incubation with 10 M sphingosine-bovine serum albumin complex and [␥-32 P]ATP for 20 min at 37°C. The labeled lipids were extracted and resolved two times by TLC in the solvent of CHCl 3 /CH 3 OH/NH 4 OH (65:35:8, v/v) and 1-butanol:acetic acid:water (3:1:1, v/v), respectively. The radioactive spots corresponding to authentic S1P were visualized and quantified by the Phosphoimage system. For kinetic studies, cell extracts were prepared from HUVEC treated with TNF for 5 min after preincubation with an increasing concentration of HDL 3 for 4 h. The kinase assay was performed with various concentrations of sphingosine (0, 2.5, 5, and 10 M).
To measure the SphK activity in vivo, the formation of S1P was measured in the permeabilized cells as described previously (25).
Electrophoretic Mobility Shift Assay-Nuclear extracts were prepared from HUVEC treated for 30 min with vehicle or the indicated agents after preincubation with or without HDL 3 . The double-stranded oligonucleotides used as a probe in these experiments included 5Ј-GGATGCCATTGGGGATTTCCTCTTTACTGGATGT-3Ј, which contains a consensus NF-B binding site in the E-selectin promoter that is underlined (34). Gel mobility shift of a consensus NF-B oligonucleotide was performed by incubating a 32 P-labeled NF-B probe with 4 g of nuclear proteins. The specific DNA-protein complexes were completely abolished by addition of a 50-fold molar excess of unlabeled E-selectin NF-B oligonucleotides.

HDL 3 Inhibits Adhesion Molecule Expression and Synthesis
in Response to TNF-In order to minimize possible confounding effects of the variations between the distinct subfractions of total HDL, HDL 3 (d ϭ 1.13ϳ1.21 g/ml) was used in the present study. As shown in Fig. 1A, HDL 3 inhibited by ϳ70% the TNF-induced expression of VCAM-1 and E-selectin in HUVEC, which was consistent with our previous reports (7,10). The inhibitory effect of HDL 3 was further identified by its reduction of E-selectin mRNA levels (Fig. 1B). To determine whether the inhibitory effect of HDL 3 on TNF-induced adhesion protein expression result from alterations of TNF access to its receptors, 125 I-TNF binding assay was performed. Fig. 1C shows that HDL 3 at concentrations of 0.25ϳ1 mg/ml (apo A-I) did not significantly affect binding of TNF to HUVEC, suggesting that the effect of HDL 3 is secondary to a perturbation of subsequent signaling pathways at postreceptor sites.
HDL 3 Inhibits SphK Activation and S1P Formation-Since we have recently identified the SphK pathway as a potent signaling pathway in mediating TNF-induced endothelial activation, the effect of HDL 3 on this pathway was determined. Consistent with our previous report (24), TNF stimulation of HUVEC caused a significant increase in cytosolic SphK activity. This activity was profoundly inhibited by HDL 3 preincubation at a physiological concentration (p Ͻ 0.01, Fig. 2A white bars). The inhibitory effect of HDL 3 was dose-dependent with half-maximal inhibition at about 0.41 mg/ml apoA-I (Fig. 2B).
To characterize the inhibition of SphK by HDL 3 , Lineweaver-Burk plots revealed that HDL 3 treatment altered the kinetics of SphK by decreasing its V max without significant changes in the K m (Fig. 2B, inset). Moreover, TNF-induced increases in S1P formation and its levels in intact cells were also markedly inhibited by HDL 3 to comparable levels as obtained by N, N-dimethylsphingosine (DMS), a competitive inhibitor of SphK ( Fig. 2A, gray and dark bars). These data further confirm the inhibitory effect of HDL 3 on activation of SphK in endothelial cells.
HDL 3 Does Not Inhibit Sphingomyelinase Activation by TNF-Given the evidence that HDL 3 inhibited SphK activity and S1P production, we tested the possibility that this inhibition was due to a reduction in sphingomyelin-ceramide turnover, an essential upstream event in S1P metabolic pathway. We previously reported that TNF stimulation of HUVEC rapidly reduced sphingomyelin content and consistently increased cellular ceramide levels by approximately 2 fold peaking at 30 min with return to near basal levels by 2 h (24). Pretreatment with HDL 3 did not interrupt the TNF-promoted sphingomyelinase activation, but significantly delayed the reversion of post TNF sphingomyelin levels to base line and sustained the increased ceramide levels (Fig. 3A). There was a significant difference in the comparison of both sphingomyelin and ceramide levels between pretreatment with and without HDL 3 after 2 h of TNF stimulation (p Ͻ 0.01, three separate experiments). The parallel delay in sphingomyelin reversion and accumulation of ceramide strongly argued against the proposition that HDL 3 decreased S1P production by inhibiting the response of sphingomyelin-ceramide cycle to TNF. Interestingly, the addition of exogenous cell-permeable ceramide (C 2 -ceramide) induced a dose-dependent inhibition of TNF-induced adhesion protein expression (Fig. 3B). Thus, the increased ceramide levels could be at least partly responsible for the inhibitory effect of HDL 3 .
Intact HDL Particles Are Required for Their Inhibitory Effect-To gain an insight into which components of HDL 3 are responsible for the inhibitory activity, the effect of apoA-I and lipids isolated from HDL 3 was investigated. In marked contrast to intact HDL 3 particles, lipid-free apoA-I, at the same concentration as HDL 3 , did not inhibit SphK activity. Similarly, HDL lipid constituents such as POPC in the form of small unilamellar vesicle also had no inhibitory effect (Fig. 4A). However, when apoA-I and POPC were reconstituted into discoidal HDL, the resulting complexes inhibited the TNF-induced SphK activation comparable to that seen with native HDL particles (Fig.  4A). In parallel, neither apoA-I nor POPC had any significant inhibitory activity on the TNF-induced expression of E-selectin (Fig. 4A, bottom panel), which was in agreement with our previous observations on VCAM-1 expression (10, 11).
As the cardioprotective ability of HDL in vivo appears to be dependent on the presence of LDL (3, 5), we tested whether the inhibitory effect of HDL on SphK is due to the interaction of LDL or other unknown factors in the serum. When cells were incubated in the serum-free conditions, the SphK activities in response to TNF in the presence or absence of HDL were to the same extent as that in the cells cultured with normal growth medium containing 20% fetal calf serum (compare Fig. 4B and 4A). The presence of LDL (250 g/ml of apoB) in the cultures did not influence SphK activities either at the basal levels or post TNF stimulation. Further LDL had no effect on the inhibitory activity of HDL on SphK activation (Fig. 4B). Additionally, in the presence of oxLDL (250 g/ml of apoB) HDL retained its ability to inhibit SphK activation (data not shown). Thus it is unlikely that the inhibitory effect of HDL resulted from the interaction of LDL, oxLDL or other unknown factors in the serum.

HDL 3 -induced Reduction of Adhesion Protein Expression Is Related to the Inhibition of SphK
Activity-Having shown that HDL inhibited SphK activity and the production of S1P, a novel identified inducer of adhesion protein expression, we further examined the linkage between the inhibition of SphK and reduction of endothelial activation. In the experiment illustrated in Fig. 5A, the HDL 3 -induced dose-dependent inhibition of SphK activity was plotted against the reduction of E-selectin expression. There was a significant linear correlation between the inhibitory effects of HDL 3 on SphK activity and E-selectin expression (r ϭ 0.953, Fig. 5A). Furthermore, when the formation of intracellular S1P was inhibited by DMS, a competitive inhibitor of SphK, the TNF-induced adhesion protein expression was also reduced (Fig. 5B). Conversely, both the HDL 3 and DMS inhibitory effects on TNF action were reversed by the addition of S1P (Fig. 5B). This demonstrated that the inhibition of SphK is an important event in the HDL 3 -mediated reduction of endothelial activation. As a control, neither DMS nor HDL 3 prevented S1P-induced adhesion protein expression (Fig. 5B), indicating a specific inhibitory effect on SphK. In addition, DMS did not change cellular ceramide levels (data not shown), suggesting the inhibition of SphK activity was not associated with altered ceramide levels.
HDL 3 Inhibits TNF-promoted ERK and NF-B Activation-The MAP kinase, ERK, has been proposed to be a downstream target in the SphK pathway mediating a variety of cellular functions including adhesion protein expression (24,26,29). Fig. 6A showed that both TNF and S1P were approximately equipotent in stimulating ERK activities. Treatment with DMS significantly inhibited TNF-activated ERK, indicating the involvement of SphK in the TNF-activated ERK signal cascade. Preincubation of HUVEC with HDL 3 also reduced TNF-stimulated ERK activation by 49 Ϯ 6.2% (p Ͻ 0.01), consistent with its effect on reducing cellular levels of S1P.
We have previously demonstrated a role of SphK pathway for TNF-promoted NF-B activation (24) that is essential for the transcription of adhesion molecule genes (35). Fig. 6B shows that HDL 3 significantly inhibited the TNF-induced activation of NF-B by 51 Ϯ 17% (p Ͻ 0.01), but did not inhibit that induced by S1P. This inhibition was comparable to that induced by DMS, the SphK inhibitor, suggesting that HDL 3 inhibited the SphK pathway resulting in an inhibition of NF-B which could account for the reduction of adhesion protein expression. Furthermore, HDL 3 treatment did not inhibit the phorbol ester-promoted NF-B activation (data not shown), indicating a specificity of HDL 3 effect in this pathway.

DISCUSSION
In this report, we show a novel mechanism of atheroprotection by HDL. In this model, HDL interrupt a signal transduction pathway, the SphK pathway, which is critically involved in endothelial cell activation and adhesion protein expression. The expression of adhesion proteins on activated endothelial cells plays an essential role for the inflammatory processes in the pathogenesis of atherosclerosis (12). The importance of adhesion molecules in atherogenesis is strongly supported by several lines of evidence: (i) adhesion molecules are present in atherosclerotic plaques (17)(18)(19)(20)(21); (ii) increased plasma levels of adhesion molecules are associated with the risks of atherosclerosis (22,23); and (iii) a deficiency of adhesion molecules significantly reduces the formation of atherosclerotic fatty streaks in knockout mice lacking the genes of ICAM-1, P-selectin, or both ICAM-1 and P-selectin (36,37).
The nature of the inflammatory signals and associated molecular mechanisms that activate adhesion molecule expression in endothelial cells in the atherogenic lesion are unknown. Factors such as TNF and interleukin-1 that are commonly found in inflammatory atherogenic lesions induce the expression of adhesion molecules in cultured endothelial cells. Thus, TNF-stimulated adhesion molecule expression on HUVEC provides a useful model to investigate the signaling pathways involved in the regulation of endothelial cell activation. In this model we have recently identified a novel signaling pathway, the SphK pathway, in mediating TNF-induced adhesion protein expression and endothelial cell activation (24). We found that TNF consistently stimulated SphK activity and the generation of S1P, and blockage of SphK by its inhibitor, DMS, inhibited NF-B activation and adhesion protein expression. An inhibitory effect of HDL 3 was clearly seen in this pathway:  6. Effect of HDL 3 on ERK and NF-B activation. A, HUVEC were preincubated with or without HDL 3 (1 mg/ml) for 4 h and treated with the indicated agents for 30 min. ERK activities were then assayed with myelin basic protein (MBP) as substrate after immunoprecipitation with antibodies against p42/p44 ERK . The kinase reaction products were separated on 10% SDS-PAGE. In parallel, an aliquot of the same cell lysates was blotted with anti-p42/p44 ERK antibodies to ensure equal ERK expression. B, NF-B binding activity was measured by electrophoretic mobility shift assay after treatment with a vehicle (lane HDL 3 inhibited (i) SphK activity, (ii) S1P generation, (iii) S1P levels, (iv) ERK activation and (v) nuclear translocation of NF-B. Moreover, HDL 3 -induced inhibition of SphK activity is linear correlating with the reduction of adhesion protein expression, and the inhibitory effects of HDL 3 were reversed by the addition of S1P (Fig. 5). Taken together, these results strongly indicated that the inhibition of SphK activation by HDL 3 could account for its inhibitory effect on adhesion protein expression and endothelial activation.
The finding that HDL 3 not only inhibited the activity and the V max of SphK in a dose dependent manner but also the generation of S1P and its levels in intact cells (Fig. 2) indicated a primary inhibitory effect of HDL 3 on the SphK pathway. On the other hand, it is possible that HDL 3 may affect endothelial phenotype by an effect on the sphingomyelin-ceramide turnover since HDL 3 increased the TNF-dependent ceramide generation and inhibited the reaccumulation of sphingomyelin (Fig. 3A). It is uncertain whether the ceramide accumulation is primarily due to prolonged hydrolysis by sphingomyelinase or to inefficient metabolism by downstream catalysis. The inhibition of adhesion molecule expression by exogenous ceramide (Fig. 3B) indicated a two-pronged inhibition on endothelial activation by HDL: reduction of S1P formation and increase in ceramide levels. Thus, it is assumed that HDL may reset the 'biostat' of ceramide and/or S1P to modulate cellular responses to TNF stimulation and to inhibit endothelial cell activation. This sphingolipid biostat has been proposed in regulating a variety of cellular functions such as cell growth, proliferation and cell death (38,39).
HDL may exert the protective effect against atherosclerosis by several mechanisms including (i) promoting cholesterol efflux from the peripheral tissues, (ii) reducing LDL oxidation, or (iii) protecting the vasculature against the cytotoxic effect of oxLDL (5,6,40). In addition to these effects of HDL, we now demonstrate a novel mechanism whereby HDL interrupt intracellular signaling involved in the pathogenesis of atherosclerosis. The inhibitory effect of HDL on the SphK pathway is very likely independent of the above-mentioned known antiatherogenic ability of HDL, since (i) an intact conformation of HDL particle is required for the inhibition; (ii) the delipidated apoA-I is unable to mimic HDL effect; (iii) the inhibition is serumindependent; and (iv) the interaction of LDL or oxLDL is unlikely to be involved in the inhibition of SphK activation (Fig. 4).
HDL may function on cells in either a receptor-dependent or -independent manner (40). Our preliminary data that HDL particles have no direct inhibitory effect on SphK activity in vitro (results not shown) suggested that the access of HDL to cell membrane and putative HDL-binding proteins could be necessary. However, the fact that lipid-free apoA-I is able to access putative HDL receptors such as SR-B1 (41,42) suggests that binding per se may not completely explain the inhibition of SphK pathway. It is possible that the apolipoprotein is critical to the effect but only when it is in an intact conformation that results from its association with phospholipids. Alternately, HDL induced inhibition of SphK pathway may act through other putative HDL-binding proteins in endothelial cells.
In conclusion, this is the first demonstration that HDL interrupt a signaling cascade-the SphK pathway that is involved in regulation of endothelial cell activation, a key event in atherogenesis. This provides a mechanistic explanation for the well-documented ability of HDL to protect against atherosclerosis and might ultimately lead to the development of novel strategies for the prevention and treatment of this disease.