ECs: Using anti-sense oligonucleotides and an RNAi strategy, we have demonstrated that S1P mediates HDL-induced cell survival through S1P₁/G_i/ERK pathways and migration through the S1P₁ and S1P₃/G_i/PI3K/p38MAPK pathway in HUVECs[11,30]. HDL-associated S1P, through the G_i/Ras/ERK pathway, has been shown to induce angiogenesis in human coronary artery ECs[58]. HDL-like lipoproteins are also present in the follicular fluid of the ovary and induce angiogenesis through S1P in association with the activation of ERK, protein kinase C, and Akt in ECs[59]. Calcium/calmodulin-dependent protein kinase kinase (CaMKK) and serine-threonine kinase LKB1-mediated AMP-activated protein kinase (AMPK) activation seems to be involved in the HDL- and S1P-induced activation of PI3K, Akt, and eNOS in ECs[60,61]. Nofer et al[35] have reported that HDL, possibly through its associated lysosphingolipids, activated PI3K and Akt, stimulates eNOS and NO synthesis, and inhibits apoptosis[13] and TNF-α-induced E-selectin expression[62] in ECs. The role of S1P₃ receptors in the HDL-induced eNOS activation has been proposed on the basis of the inhibition of S1P actions by S1P₃ deficiency[35]. The receptor subtypes involved, however, are controversial. Argraves et al[63] have reported that the S1P₁ receptor mediates HDL-induced Akt activation and subsequent stimulation of EC barrier integrity. The role of the S1P₁ receptor in the activation of eNOS through the PI3K/Akt pathways has been reported in HUVECs[63-66], bovine aortic ECs[67], and lung microvascular ECs[68]. Stimulation of bovine aortic ECs[69] and HUVECs[70] with statins, such as pitavastatin and simvastatin, has resulted in the enhancement of HDL-induced eNOS activation. The enhancement appears to be in part due to the increase in S1P₁ receptor expression[69]. Moreover, Hedrick’s group also have provided evidence that S1P₁ receptors mediate eNOS-dependent inhibition of adhesion of monocytes to the endothelium in in vivo and ex vivo mouse models[71,72]. In contrast, Norata et al[73] have reported that HDL induces transforming growth factor (TGF)-β2 expression and Smad activation through PI3K/Akt, and increases expression of long pentraxin 3 (PTX3), an acute phase protein, through G_i/PI3K[74] in ECs. The siRNA experiments have suggested that HDL-induced PTX3 expression is mediated by S1P₁ and S1P₃ receptors.

SMCs: Sachinidis et al[33] have reported that lipid molecules closely related to SPC and S1P mediate LDL- and HDL-induced Ca²⁺ mobilization and ERK activation in VSMCs. Chrisman et al[75,76] have shown that HDL-associated S1P induces the desensitization of guanylyl cyclase B, a receptor for C-type natriuretic peptide (CNP), through G_i-proteins and thereby inhibits CNP-induced cGMP accumulation and proliferation in VSMCs. HDL markedly inhibits platelet-derived growth factor (PDGF)-induced migration of VSMCs through S1P₂ receptors[77,78]. Monocyte chemoattractant protein-1 (MCP-1) has been shown to be involved in monocyte recruitment to the site of vascular inflammation, and to be elevated in atherosclerotic lesions. Treatment of VSMCs or isolated aortas with HDL inhibits thrombin-induced MCP-1 mRNA and protein expression, which is associated with suppression of NADPH oxidase, reactive oxygen species (ROS) production, and Rac1 activation. The HDL-induced actions are abolished by the S1P receptor antagonist, VPC23019, and mimicked by S1P and SPC. Moreover, HDL, S1P and SPC fail to inhibit MCP-1 production and ROS generation in aortas from S1P₃-deficient mice[79]. These results suggest that HDL-associated S1P and SPC mediate the lipoprotein-induced inhibition of ROS and MCP-1 production through S1P₃ receptors in VSMCs. González-Díez et al[80] have shown that HDL-S1P induces cyclooxygenase (COX)-2 expression and prostaglandin I₂ production through S1P₂ and S1P₃ receptors in human VSMCs, and thereby protects the cardiovascular system. They also have shown that simvastatin potentiates the HDL- and S1P-induced COX-2 expression by enhancing expression of S1P₃ receptors[80].

Cardiomyocytes: In cardiomyocytes, S1P_1-4 receptors seem to be expressed[81]. HDL and S1P have been reported to protect cardiomyocytes against apoptosis in vitro[82,83]. Frias et al[84] have recently shown that HDL-S1P activates ERK and STAT3 through S1P₂ receptors in rat ventricular cardiomyocytes, which might be involved in protection from apoptosis[85]. Tao et al[86] have shown that HDL protects mouse cardiomyocytes from apoptosis via S1P₁ and S1P₃, through mechanisms that involve the activation of ERK, PI3K, and Akt in a hypoxia/re-oxygenation model in vitro.

Monocytes and macrophages: Human monocytes and macrophages express S1P₁, S1P₂ and S1P₄, and during differentiation of monocytes into macrophages, S1P₃ is induced[87]. Although S1P actions have been extensively investigated in monocytes and macrophages[88], the roles of HDL-S1P have not been well characterized in cells. In human peripheral blood monocytes and monocyte-derived macrophages, HDL inhibits expression of chemokines, including macrophage inflammatory protein (MIP)-1α, MIP-1β, and IL-8, induced by lipopolysaccharide (LPS), a Toll-like receptor (TLR) 4 agonist; however, S1P attenuates TLR2- rather than TLR4-mediated nuclear factor (NF)-κB activation and chemokine production through S1P₁ and S1P₂ receptors[89]. This suggests that components other than S1P seem to mediate the HDL-induced actions.

Nofer et al[35] have shown that deficiency of S1P₃ receptors abolishes HDL-induced vasodilation. They also have shown that intra-arterial administration of HDL and S1P lowers mean arterial blood pressure in rats. S1P has been shown to improve ischemia/reperfusion-induced injury in vivo[90]. Theilmeier et al[83] have shown that HDL and S1P dramatically attenuate infarction size in an in vivo mouse model of myocardial ischemia/reperfusion, which is associated with inhibition of inflammatory neutrophil recruitment and cardiomyocyte apoptosis in the infarcted area. They also have observed that the HDL- and S1P-induced actions are abolished by pharmacological NOS inhibition, and are completely absent in S1P₃-deficient mice. Thus, HDL and its constituent, S1P, protect the heart against ischemia/perfusion injury in vivo via the S1P₃-mediated and NO-dependent pathway in ECs and cardiomyocytes. The same group has previously reported that HDL administration increases myocardial perfusion in a manner dependent on eNOS activation, which reflects a vasodilatory effect on the coronary circulation in vivo. However, the vasodilatory effect of HDL is unchanged in S1P₃-deficient mice, which implicates an unknown system other than S1P₃ receptors in the HDL-induced eNOS/vasorelaxation effect in vivo[91].

Recent studies have shown that SR-BI is also expressed in ECs and mediates HDL-associated apoA-I-induced stimulation of AMPK, Akt, and eNOS activation, which results in inhibition of adhesion molecule expression and monocyte adhesion to ECs[18,21,22]. The roles of SR-BI in HDL-induced vasorelaxation in isolated aorta[92] and the repair of endothelium after injury in vivo[93] have been confirmed using SR-BI-deficient mice. In support of the role of SR-BI as a receptor that links to intracellular signaling pathways, knockdown or knockout of the adapter protein of SR-BI, termed PDZK1[94], which contains four PSD-95/Dlg/ZO-1 (PDZ) domains, has been shown to attenuate the HDL-induced activation of eNOS and subsequent inhibition of NF-κB activation and adhesion of monocytes to HUVECs[66], and carotid artery re-endothelialization following perivascular electric injury in vivo[95]. Thus, although the roles of SR-BI have been suggested by experiments using in vitro receptor knockdown and in vivo receptor knockout strategy, there is still controversy on the role of SR-BI as a receptor coupling to intracellular signaling pathways. Nofer et al[35] have proposed that SR-BI plays a role, through apoA-I, as an anchor for HDL-associated S1P and other lysolipids to interact with S1P receptors to stimulate intracellular signaling pathways. Their proposal is based on the finding that reconstituted apoA-I with cholesterol and phospholipids is ineffective to stimulate eNOS and MAPKs, whereas antibodies against apoA-I and SR-BI effectively attenuate HDL-induced enzyme activation[92]. Tölle et al[79] recently have reported that aortas from either S1P₃ receptor-deficient mice or SR-BI-deficient mice have shown attenuation of the inhibitory actions of not only HDL, but also S1P on MCP-1 production, even though S1P₃ receptor expression remains unchanged by SR-BI deficiency. This result raises the possibility that the S1P₃ receptor signal transduction system is also damaged by SR-BI deficiency. Thus, we must interpret the results of SR-BI deficiency with caution; the reduction of the activities in the cells or tissues from SR-BI-deficient mice does not always eliminate the possible involvement of the S1P receptors in the HDL-induced actions.

However, Assanasen et al[96] have found that the reconstituted apoA-I with phospholipids, but without cholesterol or with a reduced level of cholesterol, can stimulate SR-BI, which results in activation of eNOS and MAPKs. They have speculated that cholesterol efflux from the intracellular space is necessary for SR-BI activation by apoA-I. The ability of rHDL or reconstituted apoA-I with phospholipids but without cholesterol to stimulate eNOS activation has been confirmed by our research group[66]. We have also shown that HDL-induced eNOS activation and subsequent inhibition of NF-κB-mediated adhesion molecule expression are attenuated by SR-BI or S1P receptor siRNA, and completely blocked by the combination of siRNAs against SR-BI and S1P receptors in HUVECs. Moreover, the rHDL preparation, in which no S1P exists, stimulates eNOS activation in a manner that is sensitive to SR-BI siRNA but not to S1P receptor siRNA, and S1P stimulates the enzyme activity in an opposite manner[66]. The enhancement of SR-BI expression by simvastatin results in enhancement of HDL- and rHDL-induced, but not S1P-induced, eNOS activation and subsequent inhibition of adhesion molecule expression[70], which supports the role of SR-BI in HDL-induced anti-inflammatory actions. rHDL preparation has also been shown to repair damaged endothelium[97] and to enhance ischemia-induced angiogenesis through stimulation of endothelial progenitor cells (EPCs)[98] in vivo. Moreover, rHDL increases circulating EPCs in patients with type 2 diabetes[99]. Feng et al[100] recently have reported that the transfer of apoA-I and subsequent increase in the level of circulating HDL attenuates artery-transplantation-induced neointima formation, in association with stimulation of EPC incorporation into the endothelium and acceleration of its regeneration. They also have demonstrated that attenuation of allograft vasculopathy requires expression of SR-BI in bone-marrow-derived EPCs, using SR-BI-deficient mice. These results suggest that SR-BI and S1P receptors are independently involved in the mature HDL-induced stimulation of intracellular signaling pathways in ECs.

S1P has been shown to exert diverse actions, which are beneficial and in some cases detrimental, to the cardiovascular system, depending on the expression profile of S1P receptor subtypes[18-20]. For example, in vascular ECs, S1P stimulates migration and proliferation or anti-apoptotic action in ECs, which might be important for maintaining intact ECs and repairing injured ECs. S1P also activates AMPK, Akt, and eNOS[60,61,68] and inhibits cytokine-induced adhesion molecule expression[71,101]. In rat neonatal cardiomyocytes, S1P rescues cardiomyocytes from hypoxic cell death[82]. S1P administration improves ischemia/reperfusion-induced injury[83,90]. S1P has also been shown to inhibit, through S1P₂ receptors, SMC migration[78], which is suggested to be involved in the formation of neointima formation. In contrast to these beneficial effects, S1P alone has also been reported to induce the expression of adhesion molecules[101-104], which might cause the stimulation of leukocyte interaction with ECs, and leukocyte penetration into the subendothelial space or the arterial intima. Thus, adhesion of leukocytes to ECs is thought to be a crucial early step of atherogenesis. S1P has also been reported to stimulate the expression of tissue factor, an essential factor for blood coagulation through the activation of transcriptional factors NF-κB and early growth response-1 (Egr-1) in HUVECs[105], and endothelial exocytosis of Weibel-Palade bodies that contain several factors, including P-selectin, von Willebrand factor, and tissue plasminogen activator in human aortic ECs[106].

Thus, S1P has the potential to exert pro- and anti-atherogenic actions. What factors determine S1P as either a pro- or anti-atherogenic signal? As described above, although S1P stimulates adhesion of monocytes to ECs, the lysolipid simultaneously inhibits TNF-α-induced adhesion of monocytes to ECs. The stimulatory and inhibitory actions might be mediated by different S1P receptor subtypes; the stimulatory action involves NF-κB activation predominantly through the S1P₃ receptor, and the inhibitory action involves PI3K/eNOS activation predominantly through the S1P₁ receptor[71,101]. Apart from the question of whether S1P-induced actions are pro- or anti-atherogenic, S1P receptor subtypes mediate, in some cases, opposite actions in the cardiovascular system[18]. For example, S1P₁ and S1P₃ receptors stimulate proliferation and migration of SMCs, whereas S1P₂ receptor inhibits these cellular activities. As for angiogenesis, the S1P₁ receptor stimulates, whereas the S1P₂ receptor inhibits, the EC migration and Rac activity[53]. As shown previously[101], pro-atherogenic or detrimental actions usually require a higher concentration of S1P than that required for anti-atherogenic or beneficial actions. Thus, pro-atherogenic adhesion molecule expression and vasoconstriction might occur at the platelet clot site, where a very high concentration of S1P might be present, under pathological conditions[40]. It remains unknown, however, whether lipoprotein-deficient plasma or albumin-associated S1P, which might be released from erythrocytes and bound to plasma albumin at a level as high as 50 mg/mL, is anti- or pro-atherogenic. In addition to the difference in the potency of S1P between beneficial and detrimental effects, HDL-associated apoA-I might play an important role in abolishing the detrimental actions of S1P. For example, pro-atherogenic adhesion molecule expression elicited by S1P disappears in the presence of physiological concentrations of HDL in a manner sensitive to SR-BI[66]. This result implies that, as far as S1P is associated with HDL, S1P never exerts pro-atherogenic actions. Thus, whether S1P is a pro- or anti-atherogenic signal seems to be determined by the pattern of S1P receptor subtypes expressed, the S1P concentration, the S1P sources (plasma or local cells), or the S1P binding proteins (HDL or albumin).

LDL contains S1P to the extent of 20%-40% of the S1P content in HDL when expressed as pmol/mg protein[10,11]. In agreement with this, LDL, similar to HDL, inhibits serum-starvation-induced EC death in association with ERK activation, and stimulates EC migration in association with p38MAPK activation[11,30,107]. However, LDL never inhibits monocyte adhesion and expression of adhesion molecules, such as vascular cell adhesion molecule-1 and intercellular adhesion molecule-1, in ECs, whereas HDL and S1P clearly inhibit these activities[66]. As reported above, HDL and S1P inhibit PDGF-induced migration of VSMCs through S1P₂ receptors. However, LDL and mildly oxidized-LDL (ox-LDL) stimulate rather than inhibit migration at even high concentrations, in which a sufficient amount of S1P exists to inhibit the migration response to PDGF[78]. The lack of LDL effects might be partly related to its susceptibility to oxidation.

The oxidative modification of LDL is a crucial step for the initiation and development of atherosclerosis[8,108]. In contrast, HDL is protective against oxidation and inhibits LDL oxidation, which is most likely due to the presence of anti-oxidants, anti-oxidative apoA-I, and anti-oxidative enzymes, such as paraoxonase, which hydrolyze ROS[5,9,109]. HDL also inhibits ox-LDL-induced inflammatory responses[9]. Spontaneous oxidation or mild oxidation with Fe²⁺ of LDL facilitates the production of lysophosphatidic acid (LPA)[78,110], which has been shown to be present in ox-LDL in atherosclerotic lesions[110]. Severe oxidation with Cu²⁺ causes the breakdown of S1P in association with LPA production in LDL[11,78]. LPA has been shown to stimulate expression of adhesion molecules in ECs[111] and migration of VSMCs[78]. Phospholipid oxidation products, such as 1-palmitoyl-2-arachidonyl-sn-glycero-3-phosphorylcholine and 1-palmitoyl-2-(5,6 epoxyisoprostanoyl)-sn-glycero-3-phosphocholine, are also accumulated in ox-LDL at the site of atherosclerotic lesions and activate ECs, thereby enhancing monocyte/EC interactions[112] and stimulating migration of VSMCs[78]. HDL, through the S1P component, inhibits oxidized phospholipid-induced inflammatory cytokine production in ECs[113]. Thus, the lack of the inhibitory LDL and ox-LDL effects on adhesion molecule expression, despite the presence of S1P, could be accounted for by the accumulation of LPA and/or oxidized phospholipids, which act oppositely to S1P. Supporting this notion, when LDL-associated LPA is degraded by monoglyceride lipase, or LPA₁ receptors on coronary artery SMCs are antagonized by Ki16425, an LPA₁ receptor antagonist, even LDL is able to inhibit the PDGF-induced migration through S1P₂ receptors[78]. These results suggest that at least LPA in the LDL particle masks the inhibitory S1P action on VSMCs, and that a balance of LPA and S1P contents in the lipoprotein is important to determine whether the lipoprotein is a positive or negative regulator of VSMC migration. It remains unknown, however, how LPA is synthesized during oxidation of LDL.

Although the plasma concentration of S1P is 200-900 nmol/L, the potency of S1P to trigger S1P receptors is around 10-100 nmol/L in in vitro assays. If plasma S1P is fully active, all the S1P receptors are saturated with S1P and fully activated. The addition of plasma that contains a low level of S1P (low-S1P plasma), which is prepared by extensive treatment of plasma with charcoal, causes a marked rightward shift of the concentration-dependent phosphatidylinositol phosphate response to S1P. For example, the addition of 10% low-S1P plasma to the assay medium increases the EC₅₀ by about 10 times. Based on the inhibitory activity of the low-S1P plasma, we anticipated that, under physiological conditions, i.e. in the presence of 100% plasma, the active S1P content would be only 7.3 nmol/L, whereas the total S1P would be 357 nmol/L[10]. Thus, only 2% of plasma S1P is speculated to be active. As reported above, we found that S1P is bound to high-molecular-weight molecules, such as lipoproteins and albumin, in plasma with a rank order of HDL > LDL > VLDL > albumin when expressed as pmol/mg proteins[10]. However, because the total albumin content is very high, plasma S1P is roughly distributed equally (50%) between both lipoproteins and lipoprotein-deficient albumin fractions when expressed as pmol/mL plasma.

At first, we speculated that HDL is the major plasma component to inhibit the activity of S1P because the HDL traps S1P most effectively among plasma components. However, HDL-S1P is as potent as albumin-S1P (in the presence of 0.1% bovine serum albumin unless otherwise specified) to stimulate ERK[11], the migration response[30], and NOS[66,101] when plotted as total S1P content in the assay system, which suggests that HDL-S1P is biologically fully active. Therefore, we tentatively speculate that a high concentration of plasma albumin (approximately 50 mg/mL or 5%) is a major cause of the interference with S1P to interact with its receptor under physiological conditions. In contrast, when long-term functions such as cell survival are concerned, HDL-S1P is 10-30 times more potent than albumin-S1P[11]. The difference in the potency of HDL-S1P and albumin-S1P between the short-term response, such as ERK (5 min) and NOS (10 min) activation, and the long-term response, such as survival (24 h), can be explained, at least in part, by the difference in the fourfold longer half-life of HDL-associated S1P than that of albumin-S1P[11]. This suggests that binding to HDL protects S1P from degradation by ectoenzymes, such as lipid phosphate phosphohydrolases. The anti-oxidative nature of HDL might also protect S1P from oxidative degradation. Thus, HDL might allow a stable reservoir of S1P in the blood and modulate S1P actions, especially long-term anti-atherogenic actions of the lipid mediator.