OUP user menu

★ editor's choice ★

Sphingosine-1-phosphate as a mediator of high-density lipoprotein effects in cardiovascular protection

Katherine Sattler, Bodo Levkau
DOI: http://dx.doi.org/10.1093/cvr/cvp070 201-211 First published online: 20 February 2009


Sphingosine-1-phosphate (S1P) has gained special attention in the high-density lipoprotein (HDL) field because HDL is the most prominent plasma carrier of S1P and because the S1P content of HDL may be responsible for many of the pleiotropic functions of HDL. This revelation has come from the evidence that HDL employ S1P receptors and signalling pathways to implement several HDL-ascribed biological effects as diverse as endothelial nitric oxide production, vasodilation, survival, and cardioprotection. This review focuses on HDL effects that are completely or partially mediated by the S1P content of the HDL particle and differentiates them from genuine HDL effects that are S1P-independent. In addition, the functional properties of ‘free’, HDL-unbound S1P are sometimes different from or even contrary to those of HDL-associated S1P. The nature of the physical interactions between HDL and local and systemic S1P production will be discussed as well as their consequences for organ function. Finally, we will elucidate the potential benefits and limitations of S1P analogues as a new class of functional HDL mimetics for cardiovascular therapy.

  • High-density lipoproteins (HDL)
  • Sphingosine-1-phosphate (S1P)
  • Cardiovascular protection

1. Pleiotropic effects of high-density lipoproteins in cardiovascular protection

Ever since the first descriptions of an association between low levels of high-density lipoprotein cholesterol (HDL-C) with coronary artery disease nearly 60 years ago,1,2 the role of high-density lipoproteins (HDL) as the best endogenous predictor of the development of coronary artery disease and cardiovascular mortality has been clearly established.35 During recent years, growing insight into the properties of HDL has changed our perception of HDL: from mere cholesterol carriers they have become global molecular players that impact on many different facets of cellular behaviour. The majority of the physiological functions of HDL influence the cardiovascular system in a favourable way either directly or indirectly. These pleiotropic beneficial effects of HDL on metabolism, vasculature, and heart make them a bona fide ‘maintenance and repair party’ of the cardiovascular system.

The most common molecular explanation for the cardiovascular protection conferred by HDL has been their fundamental role in the reverse cholesterol transport process by which excess cholesterol is shuttled from peripheral cells to the liver either for elimination via biliary excretion or reutilization in the entero-hepatic cycle.6 Especially the uptake of cholesterol from macrophages via ABC-transporters and phospholipid transfer proteins prevents the formation of foam cells in the atherosclerotic lesion,7 which is one of the first steps in the pathogenesis of atherosclerosis.8 The uptake of excess cholesterol by HDL not only prevents the formation of new lesions, but also affects the characteristics of established ones: high HDL-levels achieved by raising their endogenous levels or via administration of reconstituted, artificial HDL were shown to induce a more stable plaque morphology,9 reduce plaque lipid core,10,11 and even promote plaque regression.1214

However, there is clear evidence that HDL possess other biological functions than reverse cholesterol transport, which may independently contribute to the prevention of cardiovascular risk. HDL exert potent anti-inflammatory, anti-oxidative, anti-apoptotic, and vasodilatory properties, which are mediated by a multitude of signalling events HDL induce at the cellular level. Such properties may confer protection in a variety of cardiovascular disease settings as diverse as atherosclerosis, diabetes, metabolic syndrome, reperfusion injury, reperfusion-induced arrhythmias, and heart failure.1517 Although intrinsic to the HDL particle, the molecular basis for these pleiotropic functions is still little understood and even less well mapped to the different components of the HDL particle. The reasons lie in the complexity of the HDL subclasses, the substantial differences in HDL composition among individuals, and the existence of a variety of different biochemical entities inside HDL. This HDL diversity has hampered straight-forward mechanistic studies. Advanced protein analysis has shown that apart from the apolipoproteins apoAI, AII, AIV, E, CI to CIV, LI, M, F, D, and H that are mainly involved in lipid metabolism, the HDL particle contains a multitude of other proteins and enzymes18,19 that have diverse functions associated with immunity, the acute phase response, and complement regulation.18 Lipid profiling has revealed that in addition to free and esterified cholesterol a variety of different lipids are found in HDL including phospholipids [phosphatidylcholine, phosphatidylethanolamine (PE), PE-based plasmalogen, lysophosphatidylcholine, glycerophospholipid], free and esterified fatty acids (mono- and triacylglycerols), and different sphingolipids such as ceramide, sphingolipids/sphingomyelin species,20 sphingosine-1-phosphate (S1P), lysosulfatide, and sphingosylphosphorylcholine.2123 This review will concentrate on S1P as the main representative of the sphingolipids identified in HDL, which has gained special attention in the HDL field because of its ability to mimic many HDL functions and, most importantly, to actually mediate several of the biological effects of HDL. This revelation has come from the evidence that HDL employ S1P-specific signalling pathways for implementation of many of their physiological effects.

2. High-density lipoprotein is the major carrier and acceptor of sphingosine-1-phosphate in plasma

The major carrier of S1P in plasma is HDL, and plasma S1P levels positively correlate with HDL-C, apoAI, and apoAII levels.24 S1P occurs in plasma in a concentration of 200–1000 nM22,24 and is contained mainly in HDL (∼50–70%) and albumin (∼30%), followed by LDL and VLDL (<10%) when calculated per unit amount of protein.21 Accordingly, the concentration of S1P within different lipoproteins varies strongly: for HDL, the S1P concentration has been calculated as 140–300 pmol/mg protein (HDL-C concentrations in plasma of 40–70 mg/dL correspond to 0.8–1.2 mg HDL-protein/mL); for LDL and VLDL, S1P concentrations of 40 and 25 pmol/mg protein, respectively, have been estimated.22,23 When compared on a ‘per particle’ basis, HDL carry ∼seven-fold less S1P than LDL (9 vs. 65 mmol/mol) as calculated on the basis of 33 µM/L particles for HDL and 1500 nM/L for LDL, respectively, which correspond to ∼44 and ∼135 mg/dL HDL-C and LDL-C, respectively.21,25 However, it must be considered that there are ∼22-fold more HDL-particles than LDL-particles in plasma making HDL the primary source of S1P-exposure to cells. Any exogenous administration of HDL and LDL in physiological ratios would equal an application of five-fold higher S1P amounts administered with HDL than with LDL and distributed among 22-fold more HDL than LDL particles. Thus there are major differences in the biochemical packaging and biological activity of S1P dependent on the lipoprotein carrier. One clear indication for this is the protection against myocardial reperfusion injury conferred by HDL-associated but not LDL-associated S1P.26 Of all HDL fractions, HDL3—the small dense HDL particles—carry the highest amount of S1P with 2–3-fold higher S1P levels compared with HDL2 at a molar basis (40–50 mmol S1P/mol HDL3 compared with 15–20 mmol S1P/mol HDL2).27

The major source of plasma S1P are haematopoietic cells (mainly erythrocytes, platelets, and leukocytes) but vascular and lymphatic endothelial cells can also synthesize and release S1P.28,29 Inside the cell S1P moves freely among membranes but needs transport mechanisms for translocation to the outer leaflet of the cytoplasmic membrane because of the low propensity for spontaneous flip-flop.30,31 ABC-type transporters have been suggested to play a role in S1P export in some cell types such as platelets and mast cells,32,33 but whether they participate in the homeostasis of extracellular and specifically plasma S1P is unknown, especially as plasma S1P levels are not altered in any of the knockout mice for ABCA1, ABCA7, or ABCC1.34 The affinity of HDL for S1P is extremely high compared with other plasma carriers at a molar basis making them the primary acceptor of plasma-borne S1P (our unpublished observations).

On the following pages, we will review the cardiovascular effects of HDL specifically in respect to the possibility that they may be mediated either partially or entirely by the S1P content of HDL. To do this, we will ask not only if S1P may mimic effects of HDL but rather explore how much of the HDL effect can be abolished if S1P receptor signalling is interfered with. Vice versa, we will try to convey which of the genuine HDL effects may be S1P-independent. Furthermore, we will attempt to discriminate between the functional properties of ‘free’ S1P in contrast to HDL-associated S1P, which appear to de distinct. Finally, we will discuss the potential benefits and limitations of S1P analogues as a new class of functional HDL mimetics for the therapy of cardiovascular diseases.

3. Regulation of arterial tone

The most essential cellular messenger for the regulation of arterial tone induced by HDL is nitric oxide (NO). Native or reconstituted HDL have been shown to induce endothelium-dependent NO-mediated vasorelaxation in isolated mouse arteries ex vivo23,35 and to promote flow-induced vasodilation in hypercholesterolemic36 and HDL-deficient37 patients, respectively, via direct or indirect effects on endothelial nitric oxide synthase (eNOS) function. Furthermore, NO-dependent increases in myocardial perfusion in vivo have also been measured after administration of human HDL to mice.38

Induction of NO production by HDL in endothelial cells and vasodilation in isolated vessels is induced by a molecular mechanism completely dependent on the binding of HDL to the scavenger receptor type I (SR-BI).35 The resulting cholesterol efflux promotes phosphorylation of eNOS at multiple sites35,39,40 in a process regulated by phosphatidylinositol 3-kinase (PI3K)/protein kinase B (Akt) and mitogen-activated protein (MAP) kinases.41,42 As the two main binding partners for SR-BI on HDL apoAI and apoAII were unable to activate eNOS-mediated vasodilation,35 it must be concluded that either the sole binding of HDL to SR-BI alone via apoA is not sufficient for eNOS activation or that another constituent of HDL must exist that can activate eNOS. As it turns out, both possibilities are correct. ApoAI needs to be packaged together with 1-palmitoyl-2-oleoyl-phosphatidylcholine (POPC) to be able to induce cholesterol efflux-dependent SR-BI-mediated eNOS activation.39,40 However, is still unknown how exactly the cholesterol binding to SR-BI and its efflux leads to eNOS activation. Cholesterol efflux to HDL mediated by ABCG1 has also been shown to preserve eNOS functionality but, again, the exact mechanism has also remained elusive.43 This leaves us with the possibility that cholesterol efflux per se may be more important for eNOS function than the actual type of transporter mediating it. The requirements for the second postulate—that of an enigmatic HDL-constituent that is necessary for eNOS activation—were fulfilled by the discovery of S1P inside HDL,21 and the finding that HDL can mediate eNOS activation via S1P receptors.23 While the previously described loss of vasodilation after HDL-delipidation and the ability of S1P to activate eNOS have been hints and circumstantial evidence,21,23,44,45 the ultimate proof was provided by the observation that ∼50% of HDL-mediated vasodilation is lost in mice deficient for the S1P receptor S1P3.23 While the vasodilatory effect of free S1P used as a HDL mimic was completely abolished in S1P3-deficient mice,23 the nature of the remaining S1P3-independent but HDL-mediated vasodilation still remains enigmatic. In vitro, augmented NO production by HDL after statin treatment has been attributed to the upregulation of another S1P receptor, S1P1, while S1P3 or SR-BI appeared to play no role.46 However, siRNA to S1P1 inhibited only basal and not statin-induced NO production,46 making such conclusions difficult to generalize, especially as clear defects in eNOS signalling after HDL stimulation were detected in endothelial cells derived from S1P3-deficient mice.23

Numerous studies have been performed on the direct vasoactive effects of S1P alone. The consensus is that exogenous S1P is able to induce vasoconstriction in isolated resistance vessels (mesenteric, cerebral, and coronary arteries) but not in conduit vessels (aorta, carotid, and femoral arteries) in tension myograph studies through actions on vascular smooth muscle cells (VSMC).47 When vasoconstriction is induced by adrenergic stimulation in the same setting (e.g. by pre-contracting aortae with norepinephrine), S1P and S1P mimetics are able to induce vasodilation via eNOS activation in endothelial cells.23,48,49 In addition, intrinsic S1P sources in the vascular wall appear to play a role both in the homeostatic regulation of basal tone in resistance vessels and the myogenic response thus ensuring constant blood supply to the periphery.50,51 The sources of S1P presented to the vessel wall range from locally produced endogenous S1P29,51 to systemic, free of HDL-packaged S1P.47 This suggests that S1P exerts different and at times counteractive effects on arterial tone that are mediated by different cell types and depend on the underlying basal tone and vascular bed. These S1P effects are dynamically integrated into the overall regulation of regulation of vessel tone.47 Their consequence is a constant fine-tuning of blood flow in the periphery by local and systemic S1P.47

4. Endothelial barrier integrity and angiogenesis

Both native and reconstituted HDL have been shown to induce capillary tube-formation via the Ras/Raf/ERK and Akt/ERK/eNOS pathway and to promote endothelial cell proliferation in vitro.52,53 In addition, HDL was shown to augment endothelial cell motility and endothelial barrier integrity, with both effects being partially mediated by S1P1 as concluded from studies with pharmacological antagonists.54 It would not surprise if the S1P content of HDL acted as a functional mediator of HDL-induced angiogenesis and endothelial integrity because there is ample evidence that S1P promotes angiogenesis, migration, and proliferation in endothelial cells55,56 and enhances endothelial integrity and barrier function.5759 This appears to be S1P receptor subtype-specific as S1P1 and S1P3 strengthen the formation of endothelial cell junctions5961 while S1P2 weakens them.62,63 Several observations have also linked transient S1P-mediated Ca2+ increases to endothelial barrier stabilization and eNOS activation.45,61 Both processes are closely interrelated as the sealing of the endothelial barrier and prevention of microvascular leakage are inherent to NO. Since HDL have been shown to induce both Ca2+ mobilization and NO production in a partially S1P3-dependent manner,23 any of their effects on vascular integrity could be partially mediated by their S1P content. Other well-known HDL effects on the endothelium such as promotion of proliferation and protection against apoptosis may further help to preserve and promote endothelial integrity.64,65 The recently discovered stimulatory effect of HDL on endothelial progenitor cells (EPC) may be one more source of endothelial protection and repair.66 Reconstituted HDL have been shown to stimulate differentiation of human peripheral mononuclear cells into EPC, and enhance ischaemia-induced angiogenesis in the murine hind-limb ischaemia model,53 In the same model, S1P and its analogue FTY720 phosphate were shown to stimulate the capacity of therapeutically administered patient-derived EPC to improve blood flow recovery and neovascularization, but lost their effect when EPC from S1P3-deficent mice were used.67 Although the crucial experiment of testing the HDL effect in S1P3-deficient mice is still missing, we cannot avoid succumbing to the charm of possible causal interrelationships.

5. Anti-oxidative and cytoprotective effects

An important feature of HDL is their ability to reduce oxidative stress caused by the accumulation of deleterious oxygen radicals (ROS) and oxidatively damaged lipids and proteins.68 HDL particles carry anti-oxidative enzymes such as paraoxonase-1 and -3, and platelet-activating factor acetylhydrolase (PAF-AH), which counteract the oxidation of proteins, especially that of LDL,69,70 and prevent atherosclerotic lesion formation71 and myocardial injury.72 Recently, HDL were shown to inhibit genuine ROS production and protect endothelial cells against apoptosis.44,73,74 Both effects have been linked to S1P because the HDL fraction carrying S1P was the most efficacious one,44 and vice versa, the LDL oxidation and oxLDL-induced apoptosis were best attenuated by the HDL subclass with the highest S1P/sphingomyelin ratio: the small dense HDL3.27 While the signalling mechanisms by which HDL and its sphingolipids protect endothelial cells were identified to be Akt and NO-dependent,44,73,74 the first direct evidence of a causal contribution of the S1P content of HDL to cytoprotection was provided by experiments showing that inhibition of NAD(P)H oxidase by HDL as preponderant source of ROS in the vasculature was S1P3-dependent.75 Furthermore, HDL via its S1P content also inhibited the NAD(P)H oxidase-dependent synthesis of thrombin-induced monocyte chemotactic protein-1 (MCP-1), the key chemokine in monocyte recruitment to atherosclerotic lesions.75 While free S1P also inhibited NAD(P)H oxidase activation and MCP-1 production,75 it did not inhibit the increased basal production of MCP-1 in diabetic endothelial cells.76,77 In respect to apoptosis, free S1P is clearly a potent survival factor in endothelial cells.44,73,74

6. Inflammatory cell adhesion to activated endothelium

Leukocyte–endothelial interactions occur in a complex multi-step process mediated by several adhesion receptors. HDL restrain leukocyte adhesion by reducing these interactions in vitro and in vivo through inhibition of adhesion molecule expression and affinity.78 HDL have been reported to reduce endothelial adhesiveness in apolipoprotein E-deficient mice (ApoE−/−) in vivo, 79 to inhibit binding and transmigration of monocytes to cytokine-activated endothelium, and to reduce the expression of endothelial adhesion molecules such as VCAM-1, ICAM-1, and E-selectin in vitro.26,8082 HDL were also shown to inhibit expression of CD11b on the monocyte surface83 as well as that of MCP-1 in VSMC.75 The biological significance of the anti-adhesive effect of HDL in cardiovascular biology is eminent in two major disease complexes: atherosclerosis and acute vascular inflammation. Although it is difficult to evaluate the contribution of the anti-adhesive effects of HDL to atheroprotection separately from all other HDL effects, administration of native and reconstituted HDL has been shown to reduce plaque volume and promote lesion stabilization in patients9,10 and animal models84,85 with clear anti-inflammatory effects on plaque composition. The second disease complex to study effects of HDL on adhesion—acute vascular inflammation—can be more easily simulated in animal models in vivo, e.g. by application of TNFα or by inducing post-ischaemic inflammation. Both native human HDL and HDL mimetics consisting of apoAI-Milano and POPC have been shown to reduce inflammation in such models as shown for the protection against ischaemia–reperfusion injury of the myocardium26,86 and kidney,87 and for the prevention of organ injury in haemorrhagic shock.88

As the expression of many adhesion molecules are regulated by NF-κB and ROS-dependent mechanisms, HDL are perfectly equipped for interfering with them by their capacity to induce NO and prostacyclin (PGI2) production, inhibit ROS generation, and promote ROS elimination. Indeed, both NO and PGI2 have been shown to mediate the anti-adhesive effect of HDL in vitro and in vivo.26,82 Of the HDL constituents implicated as mediators, PAF-AH has been shown to participate in the process by increasing the anti-oxidative capacity of HDL towards oxidation products contained in pro-atherogenic lipoproteins.70,79 More recently, S1P has joined the group of potentially anti-adhesive HDL-compounds. However, the role of HDL-associated S1P in the regulation of inflammatory cell adherence is complex and not always congruent with that of free S1P. It has been clearly shown that HDL26,89,90 and exogenous free S1P76,90 inhibit TNFα-induced adhesion molecule expression and inflammatory cell adherence. The attenuation of the inhibitory effect of HDL on TNFα-induced adhesion molecule expression by siRNA against S1P1 and S1P3 in vitro supports a role for HDL-associated S1P in mediating this effect.90 In vivo, inhibition of post-ischaemic inflammation by HDL was abolished in S1P3-deficient mice.26 However, the issue becomes complicated, at least in vitro, when we take into account that TNFα has been shown to induce adhesion molecule expression by activating sphingosine kinase-1, and that free S1P by itself stimulates the expression of VCAM-1 and ICAM-1.9093 Both effects—the TNFα-mediated and the free S1P-mediated—can be inhibited by HDL,90 which depends both on SR-BI-mediated NO generation and NF-κB inhibition.90 Two explanations have been attempted to reconcile the inflammation-promoting effect of free S1P with the inflammation-inhibitory effect of HDL and HDL-associated S1P. The first one postulates the existence of yet unknown pro-inflammatory of intracellular S1P that is generated after TNFα-stimulation of sphingosine kinase-1. In this case, HDL would inhibit TNFα-activated sphingosine kinase-1 (again by yet unknown mechanisms), which would then prevent the increase of intracellular S1P.89,91 The second explanation suggests that S1P may exert two opposing effects on adhesion molecule expression via engagement of different receptors: accordingly, S1P1 would mediate the inhibitory effect of free S1P and HDL-bound S1P on TNFα-induction of adhesion molecules, while S1P3 would mediate the stimulatory effect of free S1P on adhesion molecules; in this scenario, G12/13 proteins that are activated only by S1P3 and not S1P1 would provide the molecular bias for pro-adhesive S1P effects.90 However, several questions remain unanswered. Why does free S1P preferentially activate pro-inflammatory signals when the Kds of S1P1 and S1P3 are similar and S1P1 is more abundant than S1P3 on endothelial cells? Why should only NO-generation mediated by SR-BI and not that mediated by S1P1 and S1P323,94 be involved in the inhibitory effect of HDL on TNFα? Could HDL be inhibiting any inflammatory S1P effects simply by ‘defusing’ it through incorporation before it could reach its receptors? Fortunately, the effects of free S1P on adhesion to inflamed endothelium in vivo are much more straightforward: S1P inhibits TNFα-mediated inflammatory cell adhesion in large vessels after TNFα stimulation76,77 and attenuates neutrophil recruitment to post-ischaemic inflammation during myocardial reperfusion injury.26 Interestingly, S1P3-deficiency abrogated completely not only S1P-mediated reperfusion injury but also that conferred by HDL, suggesting that under the conditions of rather mild myocardial injury caused predominately by endothelial cell damage, the S1P-content of HDL by engaging endothelial S1P3 accounts for the entire HDL-mediated cardioprotection.26

7. Prostaglandins

A different mechanism of promoting vasorelaxation besides endothelial NO production lies in the ability of HDL to induce the synthesis of the functional antagonist of thromboxane A2 prostacyclin (PGI2) in VSMC95 and endothelial cells96 by upregulation of cyclooxygenase-2 (COX-2) and the p38 MAP kinase pathway.95,97 PGI2 exerts its action through binding to the G-protein coupled prostacyclin receptors IP and EP with subsequent increases of intracellular cAMP and activation of potassium channels.98,99 Thus PGI2 not only promotes vasodilation but inhibits VSMC migration100,101 and platelet activation,102 and suppresses the production of pro-inflammatory cytokines.103 Clinically, the vasodilator capacity of PGI2 is applied in the therapy of pulmonary hypertension.104 PGI2 also affects cardiomyocyte biology as activation of cAMP by the IP-receptor inhibits cardiomyocyte hypertrophy,105 while the EP-receptor protects cardiomyocytes from damage or death from oxidative stress by opening of ATP-dependent potassium channels in the mitochondria.106 Interestingly, HDL have been shown to promote PGI2 production in the myocardium of isolated, Langendorff-perfused hearts,107,108 which has been suggested to partially account for the protection HDL confer against ischaemia/reperfusion injury in the same model.108 For the induction of PGI2 production by HDL, their protein composition seems to be important as HDL from hypoalphalipoproteinemic patients are apparently less capable in inducing PGI2 than HDL from healthy subjects.109 Again, a lipid factor inside HDL has been implied in the induction of PGI2 as delipidation of HLD has been shown to abolish the effect.110 The successive identification of this lipid factor as S1P has been due to the ability of S1P2 and S1P3 receptor antagonists to inhibit HDL-induced COX-2-mediated PGI2 release.111113

In contrast, the effects of free S1P on COX-2 and production of inflammatory prostaglandins are quite to the contrary. There is a large body of literature clearly implicating sphingosine kinase-1-generated S1P in mediating the effects of TNFα on the induction of COX-2 and the subsequent production of inflammatory prostaglandins such as PGE2.114,115 Knockdown of S1P phosphatase or S1P lyase augmented prostaglandin production along with the increase in S1P levels.114 This suggests that free S1P mediates COX-2-dependent pro-inflammatory effects of cytokines.

8. Direct effects on the heart

Any direct, primary cardioprotective effects of HDL that target specifically the myocardium must be distinguished from those secondary to the anti-atherogenic effect of HDL. While this is easily achieved under experimental conditions where myocardial ischaemia is induced, e.g. by mechanical occlusion of a coronary artery, it becomes extremely difficult when epidemiological human studies are considered. However, there is evidence in favour of direct atherosclerosis-independent cardioprotection mediated by HDL and even some evidence in favour of S1P being its mediator. Elevation of HDL for 16 weeks is considered short in respect to epidemiological studies but has proved beneficial in patients with acute coronary syndromes in the MIRACL trial, where a 1.4% risk reduction for recurrent adverse events was observed for each 1 mg/dL increment of HDL-C.116 Even shorter follow-up periods (30 days) have led to a lower incidence of mortality and major adverse cardiac events, respectively, in patients with high HDL-C levels compared with low HDL-C levels after implantation of a drug-eluting stent for acute coronary syndromes.117 The most recent study from our laboratory has shown that high HDL-C levels reduced the risk for myocardial injury during elective percutaneous coronary intervention and improved long-term prognosis when such injury did occur (K. Sattler et al., submitted for publication). Although these beneficial effects may be due to stabilization of vulnerable lesions by HDL, they also raise the question whether HDL may exert beneficial effects on the myocardium directly. In fact, the following observations argue in favour of such direct cardioprotective effect of HDL. Both in healthy individuals and patients with coronary artery disease, a positive association between HDL-C levels and left ventricular function has been observed.118,119 Although a mechanism has not been defined, the NO-dependent increase in cardiac perfusion by HDL38 may be involved as NO is important for maintenance of normal left ventricular function in healthy individuals in vivo.120 More direct evidence is provided by studies in which exogenous administration of native or reconstituted HDL improved functional recovery in isolated hearts after ischaemia/reperfusion.108,121 A concomitant enhancement of PGI2 release has been observed and ‘scavenging’ of myocardial TNFα by HDL proposed as possible explanation. The straightest argument in favour of direct HDL effects on the myocardium comes from in vivo studies of ischaemia/reperfusion, where administration of HDL potently reduced infarct size in an NO-dependent manner by inhibiting both post-ischaemic inflammation and cardiomyocyte apoptosis.26 These effects were mediated by the S1P3 receptor as HDL conferred no protection in S1P3-deficient mice.26 This is to our knowledge the first report to attribute the direct HDL effects on the heart to their S1P content. S1P itself has been clearly shown to be cardioprotective in the same and several other models,122,123 leading us to the suggestion that HDL may be viewed as a carrier of cardioprotective S1P that is made available to the endangered heart whenever and wherever needed.124

Clinically, ischaemia/reperfusion injury to the heart is extremely dangerous because of the increased arrhythmogenicity of the injured heart. Here, HDL has proved beneficial as well: administration of HDL was shown to dramatically decrease the incidence of ischaemia/reperfusion-induced ventricular arrhythmias in isolated perfused hearts by a mechanism possibly involving PGI2 and NO.125,126 In contrast to the advantageous HDL effect, the S1P1 agonist SEW2871 was demonstrated to induce irreversible tachyarrhythmias in the reperfusion period.127 However, SEW2871 was used at very high concentrations (1 mM), where receptor-promiscuous or even unspecific effects cannot be excluded. With this in mind, HDL and S1P1 agonists appear to have divergent arrhythmogenic effects. The effect of HDL on heart rate is difficult to obtain from complex epidemiological studies but there seems to be a negative correlation: in middle-aged, sedentary men the resting heart rate was inversely correlated with plasma HDL-C levels (especially HDL2 and HDL3),128 while after exercise, heart rate recovery was shown to be inversely related to the plasma triglyceride-to-HDL-C ratio.129 From the different effects S1P has on ion currents, its stimulatory effect on the inward rectifier potassium current (IK.ACh) results in a reduction of spontaneous pacing rate.130 Its inhibitory effect on the isoproterenol-induced increase in currents through L-type calcium channels (ICa,L) and the hyperpolarization-activated inward current (If) in an attenuation of the positive chronotropic effects of β-adrenergic stimulation in sino-atrial node cells and ventricular myocytes.130,131 S1P analogues such as FTY720 phosphate induce transient bradycardia in mice and men,132,133 and the S1P receptor involved has been identified as S1P3.132

9. High-density lipoprotein is not sphingosine-1-phosphate, and sphingosine-1-phosphate is not high-density lipoprotein: similarities and differences

Despite the unequivocal evidence that the S1P-content of HDL is biologically active and accounts for several of the HDL effects, there are clear functional differences between free S1P and HDL-associated S1P, as well as between native HDL and S1P associated with HDL. These statements are based on observations demonstrating that HDL have effects that are: (i) not at all or only partially attributable to their S1P content; (ii) opposite to the effect of free S1P; and (iii) even contrary to S1P effects in general. Some of them have been discussed in the previous chapters but we would like to explicitly underline these three scenarios by representative examples. An illustration for HDL effects that are independent of their S1P content is the induction of cholesterol efflux in macrophages, the most crucial factor in reverse cholesterol transport in the artery wall: it has been shown that particles containing only apoAI and POPC were as effective as HDL in promoting reverse cholesterol transport,39,40 and that reconstituted HDL without S1P was similarly effective in promoting cholesterol efflux as one containing S1P.134 An example of HDL effects only partially attributable to S1P is difficult to find when both the S1P-dependent and S1P-independent HDL effects are synergistic. However, this has been done concerning the vasodilatory effect of HDL: there, only 50% of the total vasodilation mediated by HDL was abolished in S1P3-deficient arteries while that of free S1P was completely abrogated.23 Nevertheless, the entire HDL-dependent vasodilation (both S1P-dependent and S1P-independent) was completely reliant on eNOS and a functional SR-BI receptor.35 Thus ∼50% of HDL-mediated vasodilation is not mediated by S1P but exerted by yet unknown mechanisms or HDL constituents. Finally, an example of HDL effects opposite to free S1P effects is the increase of cardiac perfusion by administration of HDL but its decrease by S1P.38 Lastly, an example for HDL being able to counteract S1P effects in general has been discussed earlier in view of the ability of HDL to inhibit S1P-induced adhesion molecule expression.135

10. How does the interaction between high-density lipoproteins and sphingosine-1-phosphate occur, where does it take place, and what consequences could it have?

There are many unsolved questions on the nature of the relationship between HDL and S1P. Why does HDL of all other molecules in plasma carry most of the S1P? How does the uptake of S1P by HDL take place? Where is S1P located topographically in the HDL particle? How much of it is biologically active? Plasma S1P levels are 20–100-fold higher than the Kd value of its receptors.21,136 Accordingly, studies have shown that the concentration of biologically active S1P in plasma is much lower (∼40-fold) than that of the total S1P concentration.21 This has led to the hypothesis that plasma proteins ‘buffer’ the large amounts of S1P to prevent erroneous activation of S1P receptors.137 On the other hand, the evidence presented in the reviewed literature strongly suggests that HDL-associated S1P (representing the majority of plasma S1P) is biologically active. How does this fit together?

Outside of the plasma compartment, 2–3-fold elevated local S1P levels have been shown at inflammation sites, and suggested to occur by activation of sphingosine kinase-1 through inflammatory mediators such as TNFα, IL-1β, LPS, and thrombin.138,139 On site, S1P presumably acts in a pro-inflammatory manner by inducing PGE2 and adhesion molecules,9193,114,115,135 retaining lymphocytes at the inflammation site,140 and promoting coagulation-induced activation of dendritic cells in the lymphatics.141 Quite to the contrary but occurring simultaneously, the same S1P acts in a negative feedback mechanism to limit the increase in endothelial permeability associated with inflammation139 by enhancing endothelial barrier function57,60,61 and inhibiting leukocyte adhesion.76,77 Therefore, locally produced S1P appears to be an important determinant of the build-up, magnitude, and duration of the inflammatory response.

The scenario we would like to propose for the functional interrelation between HDL and S1P is one in which HDL act as the master regulator of local S1P concentrations by ‘sucking up’ excess S1P or even ‘snatching’ it away from other carriers. HDL as well as other plasma proteins have been suggested to act as ‘sinks’ for S1P, thereby neutralizing any excess S1P and providing an explanation for the much higher plasma levels of S1P than those necessary for S1P receptor activation.21,142 From our own data, the capacity of HDL to take up S1P is enormous (up to 10-fold higher than the actual content in HDL per milligram of protein; unpublished observations). Accordingly, the presence and local concentration of HDL would determine how much S1P is biologically active and where. The ‘where’ may be very important because HDL and other lipoproteins are present in the interstitial space in amounts that correspond to ∼25% of their plasma concentration (in the case of HDL),143,144 and are known to circulate with the lymph fluid.145 Remarkably, the concentration of lipoproteins increases several-fold in inflammatory exudates.146 This increase would enable HDL to remove the excess S1P produced at sites of inflammation, buffer it and carry it away, thus helping in the resolution of inflammation. However, there may be something more to HDL than just that. All reported effects of HDL-associated S1P are potentially beneficent for cardiovascular homeostasis, while vice versa, not a single deleterious effect has been reported for HDL-bound S1P. In contrast, free, HDL-unbound S1P has the propensity of exerting pro-inflammatory, vasoconstrictive, and other potentially adverse effects as reviewed here. In contrast to scenarios suggested by others,142 we would argue that by incorporating free S1P in their macromolecular structure HDL may not only neutralize the deleterious excess of S1P but may also transform it from ‘bad’ free S1P to ‘good’ HDL-packaged S1P. Such benignity may require the docking of HDL to cell surface receptors such as SR-BI in order to allow presentation of HDL-associated S1P to adjacent S1P receptors. In this way a spatially confined activation of S1P receptors is achieved dependent on the presence of HDL receptors and HDL-S1P content. Obviously, this all has to be proven both by experimental S1P distribution studies as well as human patient studies, in which a ‘more’ of S1P-bound HDL will have to be associated with a better prognosis of disease. In vitro support for this hypothesis comes from the observation that a clearly defined biological effect of HDL has been shown to depend on the magnitude of S1P content within the HDL particle: loading of HDL with exogenous S1P was shown to increase their ability to inhibit oxLDL-induced apoptosis in endothelial cells.27

11. Is sphingosine-1-phosphate a marker of dysfunctional high-density lipoproteins?

Several studies have led to the idea that the absolute level of HDL-C is not the only criterion contributing to their athero-protective effect but that an enigmatic attribute termed ‘HDL-quality’ also exists.14,147,148 Some of these studies refer to the observation that low HDL-C increase risk in patients with low LDL-C levels but that vice versa, high HDL-C does not necessarily decrease risk.149 Others have observed a superior propensity of certain HDL mutations such as apoAI type Milano to mediate cholesterol efflux150 accompanied by an enhanced protection against atherosclerosis.151 The quality aspect of HDL is mirrored by the evidence that each known cardioprotective function of HDL can become defective and give rise to functionally impaired HDL.83,147,152 Such functionally impaired HDL particles have been described in patients with virtually all cardiovascular risk factors (metabolic syndrome, diabetes mellitus, obstructive sleep apnoea).153155 The biological characteristics of dysfunctional HDL extend to many of the known effects of HDL such as protection against LDL-oxidation and apoptosis,153,154 vasorelaxation, and macrophage adhesion.156,157 Therapeutic interventions which have shown to improve HDL dysfunctionality include treatments with high-fibre/low saturated fat diets,158 statins,157 and fibrates.159 The molecular origin of HDL dysfunction has been suggested to lie in alterations of HDL composition (e.g. apolipoprotein and lipid ratios) or in biochemical changes of individual HDL components such as apoAI, for which oxidative modifications and non-enzymatic glycation have been described.147,152 So far, although S1P has emerged as an important mediator of many regular HDL functions, there are no epidemiological or clinical studies that have analysed S1P levels in dysfunctional or even normal HDL. Such studies are clearly needed to find out if alterations of HDL-associated S1P participate in HDL dysfunctionality, and whether therapeutic treatments known to improve HDL function may be doing so via raising their S1P content. Finally, if S1P indeed proves important as a marker of HDL dysfunctionality, then the S1P content of HDL may itself constitute a novel predictor of cardiovascular risk.

12. Sphingosine-1-phosphate analogues as functional high-density lipoprotein mimetics

The understanding that HDL quality is clinically important has led to the development of HDL-based therapies. There is a growing family of HDL surrogates such as reconstituted HDL, apoAI, apoAI-Milano, and apoAI-mimetic peptides designed to imitate the structural requirements for the atheroprotective and anti-inflammatory properties of HDL.148 Reconstituted HDL has been shown to reduce volume and promote stabilization of atherosclerotic lesions in animals84 and patients,9 and to improve endothelial dysfunction.36 The apoAI-mimetic peptide D-4F was shown to promote HDL-mediated cholesterol efflux from macrophages,160 restore NO production,161 reduce atherosclerotic lesions162 in mice, and improve the anti-inflammatory properties of HDL in patients.163 ApoAI-Milano complexed with POPC reduced ischaemia/reperfusion injury in rabbits,86 diminished the lipid core and macrophage content in apoE−/− mice,164 and decreased the volume and thickness of atherosclerotic lesions in patients with overt coronary artery disease.165 Therefore, apoAI mimetic peptides are an increasingly important option for HDL-based therapy.

However, no matter which sort of apoAI mimetic has been used in vitro or in vivo, the ultimate biochemical mechanisms for its beneficial effect may not be related to the apoAI moiety alone. Lipid-free apoAI, apoAI mimetics, reconstituted HDL, and small unilamellar phosphatidylcholine vesicles may all have the same in common: they most certainly alter their composition once having entered the plasma. In fact, the differences in biological potency existing among them have been attributed to differences in their lipidation profile.166 If so, then an uptake of S1P from the local milieu could most certainly be part of this lipidation process. Once inside the particle S1P may then mediate part of the biological effects ascribed to the apoAI mimetics. Studies are needed to determine how an apoAI mimetic exactly changes its biophysical and biochemical configuration after entering plasma, and how much and how fast does S1P integrate into the particle.

Based on such considerations, S1P analogues may be considered functional HDL mimetics.48 The S1P analogue FTY720 (fingolimod) is the first member of a new class of immunosuppressive drugs currently in phase III clinical trials for prevention of allograft rejection and phase II for multiple sclerosis.167 In vitro and in vivo, FTY720 phosphate activates four of the five S1P receptors and mimics several of the functional properties ascribed to HDL-associated S1P such as vasodilation,168 inhibition of NAD(P)H oxidase and MCP-1 production,75 protection against ischaemia/reperfusion injury,169,170 and attenuation of atherosclerosis.124,171,172 The major drawback of using FTY720 for cardiovascular purposes is its immunosuppressive effect which, fortunately, is exclusively mediated by S1P1. Therefore, employing combinations of the upcoming receptor subtype-specific S1P analogues as functional HDL-mimetics will allow their consideration as tools for tailoring individual therapies for cardiovascular diseases.

Conflict of interest: none declared.


This work was supported by the Deutsche Forschungsgemeinschaft (LE940/4-1, LE940/3-1).


View Abstract