Principles of bioactive lipid signalling: lessons from sphingolipids

Key Points

• The sphingolipids constitute an important class of bioactive lipids, including ceramide and sphingosine-1-phosphate (S1P). Ceramide can be considered to function as a hub in sphingolipid metabolism, and it mediates or regulates antiproliferative responses such as growth inhibition, apoptosis, differentiation and senescence, whereas S1P is a key regulator of cell motility and proliferation.

• The study of bioactive lipids in general and sphingolipids in particular presents several hurdles to molecular cell biologists. These include the hydrophobicity and biophysical properties of these molecules, the metabolic interconnections of the active metabolites, and the predominantly hydrophobic nature of the enzymes that regulate their metabolism.

• Enzymes of sphingolipid metabolism function as an interconnected network that regulates the levels and interconversions of the bioactive sphingolipids. Many of these enzymes, such as sphingomyelinases, sphingosine kinases and ceramide synthases, serve to couple the action of extra- and intracellular agonists to downstream effectors.

• Ceramide can be formed from the de novo pathway or following activation of the sphingomyelinase pathway, in which it functions in metabolic regulation and stress responses. Ceramide action is governed by the specific pathways that regulate its formation, their subcellular localization and their specific mechanisms of regulation.

• S1P is a product of sphingosine kinases, and acts predominantly on a family of G protein-coupled receptors that, in turn, mediate its action on cell growth, migration, transcription and signal transduction.

• The cellular actions of ceramide, S1P and other bioactive sphingolipids are increasingly thought to be crucial for the study of angiogenesis, inflammation, immune responses, diabetes, ageing, cancer biology and degenerative diseases.

Abstract

It has become increasingly difficult to find an area of cell biology in which lipids do not have important, if not key, roles as signalling and regulatory molecules. The rapidly expanding field of bioactive lipids is exemplified by many sphingolipids, such as ceramide, sphingosine, sphingosine-1-phosphate (S1P), ceramide-1-phosphate and lyso-sphingomyelin, which have roles in the regulation of cell growth, death, senescence, adhesion, migration, inflammation, angiogenesis and intracellular trafficking. Deciphering the mechanisms of these varied cell functions necessitates an understanding of the complex pathways of sphingolipid metabolism and the mechanisms that regulate lipid generation and lipid action.

Main

Although the concept of 'bioactive lipids' — broadly defined as changes in lipid levels that result in functional consequences — has been decades in the making, it has only started to gain traction in the past 20 years, and promises to occupy centre-stage in cell biology research in the twenty-first century. This belated recognition has its roots in earlier preconceived ideas about exclusive roles for lipids in energy metabolism and in membrane structure, which have prevented wider thinking about functional lipids by non-lipidologists. These impediments primarily arise from the many inherent difficulties of working with lipids, their enzymes and their targets.

In their pioneering studies in the 1950s, Hokin and Hokin observed the rapid turnover of inositol phospholipids in pancreatic slices that were stimulated with acetylcholine¹. However, the significance of those observations for cell function lay relatively dormant for several years until diacylglycerol (DAG) and inositol-1,4,5-trisphosphate (Ins(1,4,5)P₃) were subsequently implicated in specifically regulating protein kinase C (PKC) and calcium release², respectively, leading to various cellular responses. Conceptually, the demonstration of direct activation of PKC by the lipid DAG cemented the idea that a lipid (which already had an established role in intermediary metabolism) could regulate cell signalling, and ushered in the era of bioactive lipids.

Studies over the past four decades have also defined the eicosanoids (and other products of arachidonic acid metabolism) as key inter- and intracellular lipid signalling molecules, which are primarily involved in mediating or resolving inflammatory responses³. Further studies have shown that many minor products of inositol phospholipid metabolism, such as phosphatidylinositol-3,4,5-trisphosphate (PtdIns(3,4,5)P₃), also serve as key intermediates in cell signalling. Other glycerolipid-derived regulatory molecules include phosphatidic acid (PA), monoacylglycerols and anandamide (candidate ligands for cannabinoid receptors), lyso-phosphatidic acid and platelet activating factor (PAF).

Another group of bioactive lipids that have come to the foreground in the past two decades are the sphingolipids. The first of these to be identified was sphingosine, which exerts pleiotropic effects on protein kinases and other targets. Sphingosine and its related sphingoid bases have roles in regulating the actin cytoskeleton, endocytosis, the cell cycle and apoptosis⁴ (Fig. 1). Even more emphasis has been placed on the sphingolipids ceramide and sphingosine-1-phosphate (S1P). Ceramide mediates many cell-stress responses, including the regulation of apoptosis⁵ and cell senescence⁶, whereas S1P has crucial roles in cell survival, cell migration and inflammation⁷ (Fig. 1). Recent additions to the family of bioactive sphingolipids include ceramide-1-phosphate (C1P, which has roles in inflammation and vesicular trafficking^8,9,10), glucosylceramide (which is involved in post-Golgi trafficking and in drug resistance^11,12), lyso-sphingomyelin and dihydroceramide.

Figure 1: An overview of the roles of sphingolipids in biology.

The scheme shows potential participation of the bioactive lipids ceramide, sphingosine and sphingosine-1-phosphate (S1P) in cell biological responses. Ceramide can be generated by the breakdown of sphingomyelin by sphingomyelinases (SMases) or synthesized de novo by serine palmitoyl transferase (SPT) and ceramide synthase. Both of these processes can be induced by diverse stimuli (beige boxes). Sphingosine and S1P can be generated by ceramidases (CDases) and sphingosine kinases (SKs). These interconnected metabolites interact with specific protein targets such as phosphatases, kinases and G protein-coupled receptors (S1P receptors), which in turn mediate the effects of these lipids and, at least in part, also mediate the effects of the inducers on specific cell responses. The numbers in brackets indicate the relative levels of these sphingolipids. CAPP, ceramide-activated Ser–Thr phosphatase; IGF; insulin-like growth factor; IL-1, interleukin-1; oxLDL, oxidized low-density lipoprotein; PDGF; platelet-derived growth factor; PKC, protein kinase C; PKH, PKB homologue; TNFα, tumour necrosis factor-α; VEGF, vascular endothelial growth factor; YPK, yeast protein kinase.

In addition to the vast number of cellular processes that sphingolipids are associated with, further levels of complexity in sphingolipid signalling are only now being deciphered. This complexity arises from the metabolic interconnection of bioactive lipids (Fig. 2), with changes in one lipid exerting metabolic 'ripple' effects with functional consequences. Moreover, the enzymes of lipid metabolism (the input into these pathways) are often hydrophobic membrane proteins that defy simple biochemical and cellular investigation. The varied biophysical properties of the regulatory lipids also introduce complications concerning their subcellular localization and mechanisms of action.

Figure 2: Sphingolipid metabolism and interconnection of bioactive sphingolipids.

Ceramide is considered to be the central hub of sphingolipid metabolism, and is synthesized de novo from the condensation of palmitate and serine to form 3-keto-dihydrosphingosine (not shown). In turn, 3-keto-dihydrosphingosine is reduced to dihydrosphingosine followed by acylation by (dihydro)-ceramide synthase (also known as Lass or CerS). Ceramide is generated by the action of desaturases. From here, ceramide can be converted to other interconnected bioactive lipid species. The only exit pathway from the sphingolipid pathways is mediated by sphingosine-1-phosphate (S1P) lyase, which metabolizes S1P. Both ceramide and dihydroceramide have been shown to be bioactive and, in fact, >50 distinct molecular species may be classified as ceramide (Fig. 3). Each may have its own separate metabolic network, resulting in unprecedented levels of complexity in sphingolipid signalling. In this figure, only a pathway for a generic ceramide and a generic dihydroceramide (shown in the dotted outline box) are depicted. It does not show the multiplicity or subcellular localization of these pathways (Fig. 5). CDase, ceramidase; CK, ceramide kinase; DAG, diacylglycerol; GCase, glucosyl ceramidase; GCS, glucosylceramide synthase; PC, phosphatidylcholine; SK, sphingosine kinase; SMase, sphingomyelinase; SMS, sphingomyelin synthase; SPPase, sphingosine phosphate phosphatase; SPT, serine palmitoyl transferase.

Therefore, a comprehensive understanding of bioactive sphingolipids (and other lipids in general) requires at least two levels of investigation. At a basic level, there is a need to delineate the various specific mechanisms by which sphingolipid metabolism is regulated and the molecular mechanisms by which the bioactive lipid products transmit their respective signals. At an advanced and lipid-specific level, such an understanding requires knowledge of the metabolic organization of enzymes of sphingolipid metabolism and their inter-relationships, the subcellular and submembranous localization of lipid-mediated pathways, the metabolic transformation of bioactive lipids, and the integration or coordination of overall responses. This review aims to inform the reader about the basic organization and principles of sphingolipid-mediated cell regulation, and to provide a current understanding about specific sphingolipid signalling pathways. In the process, we attempt to elucidate principles that govern the operation of bioactive lipids in general.

Introducing the sphingolipid family

Sphingolipid metabolism mainly occurs in eukaryotes, but is also found in the Sphingomonas bacterial genus. Pathways of sphingolipid metabolism have a unique metabolic entry point (serine palmitoyl transferase; SPT), which forms the first sphingolipid in the de novo pathway, and a unique exit point, S1P lyase, which breaks down S1P into non-sphingolipid molecules. The multiple metabolic steps in between constitute a highly complex network that connects the metabolism of many sphingolipids (Fig. 2). In this network, ceramide (and dihydroceramide to a lesser degree) can be considered to be a metabolic hub because it occupies a central position in sphingolipid biosynthesis and catabolism.

Briefly, sphingolipids are synthesized de novo from serine and palmitate, which condense to form 3-keto-dihydrosphingosine through the action of SPT^13,14. In turn, 3-keto-dihydrosphingosine is reduced to dihydrosphingosine, followed by acylation by a (dihydro)-ceramide synthase (also known as Lass or CerS)¹⁵. Ceramide is formed by the desaturation of dihydroceramide¹⁶ (Fig. 2).

In sphingolipid biosynthetic pathways (Fig. 2), ceramide can be phosphorylated by ceramide kinase¹⁷, glycosylated by glucosyl or galactosyl ceramide synthases¹⁸, or can receive a phosphocholine headgroup from phosphatidylcholine (PC) in the biosynthesis of sphingomyelin (SM) through the action of SM synthases¹⁹, which thereby also serve to generate DAG from PC.

Breakdown of complex sphingolipids (Fig. 2) proceeds through the action of specific hydrolases, leading to the formation of glucosylceramide and galactosylceramide. In turn, specific β-glucosidases and galactosidases hydrolyse these lipids to regenerate ceramide^20,21. The breakdown of SM is catalysed by one of several sphingomyelinases (SMases)²². These include acid SMase, neutral SMases and alkaline SMase.

Ceramide may be broken down by one of many ceramidases^23,24, leading to the formation of sphingosine, which seems to have one of two fates. Sphingosine may be 'salvaged' or recycled into sphingolipid pathways (Box 1), or it can be phosphorylated by one of two sphingosine kinases²⁵, SK1 and SK2 (Fig. 2). The product S1P can be dephosphorylated to regenerate sphingosine through the action of specific intracellular S1P phosphatases²⁶ and, possibly, non-specific extracellular lipid phosphate phosphatases^27,28. Alternatively, S1P lyase can irreversibly cleave S1P to generate ethanolamine phosphate and hexadecenal (which, in turn, can be reduced to palmitate and subsequently reincorporated into lipid metabolic pathways)²⁹.

Complexities of sphingolipid signalling

The conceptualization of cell signalling is largely dominated by the canonical cyclic AMP paradigm, whereby a linear signalling pathway exists that involves receptors, cyclase, cAMP and cAMP-dependent protein kinase, and in which cAMP has a dedicated signalling function. However, sphingolipid signalling manifests additional layers of complexity compared with the cAMP pathway. Unlike the relative simplicity of the biochemically isolated cAMP pathway, enzymes of lipid metabolism are intimately related to each other, generating an interconnected network (Fig. 2) that serves to regulate not only the levels of individual bioactive lipids, but also their metabolic interconversion. Moreover, there are multiple pathways that can operate in parallel (Fig. 3). Also, unlike the soluble cAMP molecule, bioactive sphingolipids exhibit hydrophobic properties; therefore, the physiological environment of bioactive lipids is, for the most part, restricted to biological membranes. So, a crucial and unique set of properties operates in the cell to define the subcellular localization of sphingolipid-mediated pathways and the transport of bioactive lipids across and between membranes. These complexities are discussed below.

Figure 3: Parallel networks of sphingolipid signalling.

a | The figure shows the function of specific (dihydro)-ceramide synthase (also known as Lass or CerS) enzymes in regulating the formation of distinct ceramides that are distinguished by their N-linked fatty acid groups. For example, CerS1 selectively regulates the synthesis of C18-ceramide, whereas CerS5 and CerS6 result in the preferential formation of C16-ceramide. Studies that mostly rely on overexpression of CerS suggest that they localize to the endoplasmic reticulum (ER); however, caution should be exercised in interpreting results that are based on the overexpression of hydrophobic proteins. These enzymes may localize to different regions of the ER or to distinct ER-associated membranes such as the nuclear membrane, mitochondrial-associated membranes and others. b | The generic 'ceramide' is a family of >50 distinct molecular species, as ceramide may exist without the double bond (dihydroceramide), with the double bond (ceramide), with a 4′-hydroxy sphingoid base (phytoceramide), with a 2′-hydroxy (α-hydroxyceramide), or both hydroxyl groups (α-hydroxy-phytoceramide), or even with an w-hydroxy (not shown). Each of these can have various N-linked acyl chains. Evidence thus far shows distinct functions for ceramides and dihydroceramides, and is beginning to show distinct functions for different acyl-chain ceramides. dh, dihydro.

The interconnectivity of sphingolipid metabolism. The unprecedented complexity in sphingolipid biochemical interconnections enables cells to orchestrate cellular responses by regulating sphingolipid interconversions. For example, activation of SMase generates ceramide as an immediate lipid product. However, the subsequent action of ceramidase, ceramide kinase, SM synthase or glucosylceramide synthase may potentially convert the ceramide signal to one that is mediated by sphingosine (and subsequently S1P), C1P, DAG or glucosylceramide, respectively (Fig. 2).

Interestingly, the cellular levels of these bioactive sphingolipids render these scenarios highly likely³⁰. For example, in most cell types, SM is present in concentrations that are an order of magnitude higher than those of ceramide; therefore, small changes in SM can result in profound changes in ceramide. In addition, ceramide is often detected in concentrations that are more than an order of magnitude higher than those of sphingosine. Therefore, immediate hydrolysis of only 3–10% of newly generated ceramide may double the levels of sphingosine. Similarly, phosphorylation of 1–3% sphingosine may double the levels of S1P (Fig. 1).

Given these metabolic considerations, it is possible that when a sphingolipid enzyme is implicated in a process, the direct product of this enzyme may not be the actual signalling effector. Therefore, it is important to determine which lipid product or substrate actually mediates the signal. For example, by implicating SMase or SKs in a specific cell response, one may not immediately conclude that it is ceramide or S1P that directly mediates such function. Fortunately, tools and approaches are now available to tease these processes out. These include highly sensitive mass spectrometry methods for sphingolipid analysis^30,31, molecular tools that have been generated by the nearly complete molecular identification of all known enzymes of sphingolipid metabolism, and RNA interference, which has been applied successfully to cell biology studies of lipid metabolism and function.

Parallel networks of sphingolipid signalling. The multiplicity of several of the pathways of sphingolipid metabolism generates another layer of complexity, which is superimposed on the basic blueprint of sphingolipid metabolism discussed above. This is best illustrated with the ceramide synthases¹⁵ (Fig. 3). Six individual ceramide synthases are now known to catalyse the formation of dihydroceramides or ceramides (depending on whether the substrate is dihydrosphingosine or sphingosine, respectively). Importantly, these ceramide synthases show distinct preferences for the different fatty acyl-CoA substrates and, therefore, they generate distinct ceramides with unique N-linked fatty acids. Distinct ceramides may localize to distinct subcompartments and may mediate distinct functions. As another example, both acid SMase and neutral SMase can act on SM to generate ceramide. However, these two enzymes are regulated differentially, localize to distinct subcellular compartments (Box 1) and may even generate ceramides with distinct molecular species.

Subcellular localization of bioactive sphingolipids. Most enzymes of sphingolipid metabolism show specific subcellular localization(s) (Box 1), and this exerts profound effects on the signalling and regulatory functions of the generated sphingolipid. For example, acid SMase localizes primarily to the endolysosomal pathway but, under certain conditions, can also relocate to the plasma membrane, presumably at the outer leaflet³². Neutral SMase2 appears to localize to the inner leaflet of the plasma membrane. SK1 and SK2 seem to act preferentially on a sphingosine substrate that resides in membranes, which requires the translocation of SK enzymes from their cytosolic compartment to the plasma membrane (and other membranes) and binding to anionic lipids³³. It is therefore more likely that SK enzymes have multiple subcellular localizations that could dictate distinct functional responses.

These considerations, coupled with the poor solubility of lipids (Box 2) in cells, impose clear restrictions on the subcellular localization of bioactive lipids. In the absence of specific mechanisms of transport for these lipids, the site of generation of a bioactive lipid is likely to dictate its site of action, unless proven otherwise.

Bioactive sphingolipids (and lipids in general) may be divided into separate classes that can be distinguished by their biophysical properties (Box 2). Some bioactive lipids, such as C1P and phosphatidylinositol-3-phosphate, carry ionic charges at neutral pH and contain two hydrophobic chains. These lipids are most likely to reside in their compartment of generation and are unlikely to flip spontaneously across bilayers. A second group, which includes ceramide and DAG, is composed of neutral hydrophobic molecules that are also restricted to their compartments of formation, but may readily flip-flop across membranes³⁴.

A third group is composed of single-chain lipids, and these include sphingosine, S1P, lyso-sphingomyelin and lyso-phosphatidic acid. These molecules have sufficient aqueous solubility to move between membranes. For example, we have estimated that ∼70% of sphingosine resides in membranes at physiological pH and the remaining 30% is soluble³⁵. These molecules can therefore equilibrate quickly among membranes. More importantly, given their amphipathic nature (that is, soluble in both water and organic solvents), these molecules may exert surfactant activity. Therefore, it is not surprising that their cellular levels tend to be among the lowest of all bioactive lipids.

Transport and flipping of bioactive sphingolipids. For bioactive sphingolipids to effect functional responses, they must be able to interact with relevant direct mediators. The subcellular restriction of bioactive sphingolipids and some of their targets poses several questions as to how interacting partners are able to meet (Fig. 4). For example, S1P is probably generated at the inner leaflet of the plasma membrane in response to tumour necrosis factor-a (TNFa) and other agonists^36,37. Yet, it appears that it has to reach the outer leaflet of the plasma membrane to interact with S1P receptors (S1PRs)³⁸. Along those lines, two members of the ABC transporter superfamily have been suggested to regulate S1P transport (Fig. 4). The cystic fibrosis transmembrane regulator (CFTR) has been implicated in the internalization of S1P from the plasma membrane³⁹, whereas the ABC transporter ABCC1 has recently been reported to facilitate efflux of S1P⁴⁰.

Figure 4: Transport and transbilayer movement of bioactive sphingolipids.

Shown are model membranes that represent the plasma membrane, vesicular membranes and intracellular membranes, and the mechanism by which sphingolipids may be transported between and across them. Sphingomyelin (SM) has two aliphatic chains and a zwitterionic head group. Therefore, it only occasionally (if at all) flip-flops across bilayers and has negligible aqueous solubility, but may possibly have significant lateral movement. This movement may be hindered by self-aggregation and/or interaction with membrane sterols. Ceramide (Cer) contains two aliphatic chains but has a neutral headgroup, so exhibits less aqueous solubility than SM, but is readily able to flip-flop in model membranes. It is not known if interactions with other lipids or partitioning in microdomains would hinder this flipping. Both SM and Cer require vesicles or specific transfer proteins to move between distinct membranes. CERT functions as a specific ceramide transfer protein that acts in the endoplasmic reticulum (ER)-to-Golgi transport of Cer, either at the ER–Golgi interface or as a freely movable transfer protein. Sphingosine (Sph) has one aliphatic chain (usually 18 carbons in length) and appears to flip-flop freely and transfer among membranes. Sphingosine-1-phosphate (S1P) has a zwitterionic headgroup (including a charged phosphate), and is likely to move freely among membranes, but is unlikely to flip-flop spontaneously. Accordingly, two ABC transporters have been suggested to have roles in S1P transport: CFTR (cystic fibrosis transmembrane regulator) is implicated in S1P entry into the cell, and ABCC1 is implicated in S1P efflux. CDase, ceramidase; SMase, sphingomyelinase; SK, sphingosine kinase; SMS, sphingomyelin synthase; SPPase, sphingosine phosphate phosphatase.

The ceramide transfer protein CERT has been shown to effect ceramide transport from its site of synthesis in the endoplasmic reticulum (ER) to the Golgi and to be required for synthesis of SM (but not for glucosylceramide synthesis)⁴¹ (Box 1). Given the interaction of CERT with phosphatidylinositol phosphates and its potential regulation by phosphorylation, this step of ceramide metabolism may be regulated by pathways of inositol lipid metabolism, and by protein kinases⁴² and phosphatases.

As a neutral lipid, ceramide is expected to flip-flop across membranes with ease, and indeed, studies in model membranes and membranes from erythrocytes confirm these predictions³⁴ (Fig. 4). However, it is not clear if the ceramide that forms in more complex biological membranes flips with the same efficiency, or if the organization of ceramide into microdomains (these are specialized lipid domains in the plasma membrane; see the review by van Meer and colleagues in this issue) may restrict flipping of ceramide from the outer leaflet into the inner leaflet. Restricted flipping (or flopping from the inner to outer leaflet) could have immense effects on the signalling functions of ceramide. For example, ceramide that is generated by acid SMase in the outer leaflet may exert distinct functions compared with ceramide that is generated in the inner leaflet by neutral SMase2 (Ref. 22).

In summary, the biophysical and metabolic considerations outlined above introduce a level of complexity that is not seen in other signalling pathways. There are at least 26 distinct enzymes that act on ceramide as either a substrate or product, and which therefore have the potential to regulate ceramide levels. These enzymes show specific and varied subcellular localizations. Moreover, ceramide itself is a family of closely related molecules with at least 50 distinct molecular species that can be distinguished by their acyl chains, hydroxylations and desaturation (Fig. 3). Therefore, ceramide levels may be regulated in distinct compartments by distinct mechanisms at different times. So, asking the question of what ceramide does is not useful, because ceramides that are formed by different mechanisms in distinct compartments, by necessity, exert different functions, and more probing questions should query the specific pathways of ceramide generation.

Mechanisms of action: ceramide and S1P

Understanding what bioactive lipids do and how they transmit signals requires elucidation of the mechanisms by which these lipids act. One could envision two general mechanisms for the action of lipids: lipid–lipid interactions, whereby the candidate bioactive lipid affects membrane structure and/or the interaction of membrane proteins with the membrane bilayer (see the review by van Meer and colleagues in this issue); or lipid–protein interactions, whereby changes in lipids modulate the function of target proteins that interact specifically with the candidate bioactive lipid.

The physiological levels of bioactive lipids clearly influence (or even dictate) their mechanisms of action. Trace lipids such as S1P interact with high-affinity receptors that are capable of sensing their low levels. Lipids that are found at intermediate membrane concentrations, such as ceramide or DAG (which often constitute 0.1–1.0% of total membrane lipids), act on targets with intermediate affinity. By contrast, it is difficult to envisage how abundant lipids such as SM can have specific targets because the affinity of interactions must be low. This does not rule out proteins that may bind to these lipids with high affinity, but in this case, such proteins would not be able to sense changes in the levels of these lipids. The high levels of these lipids, however, impart on them the ability to change overall membrane properties and substructure if and when their levels change. For ceramide and S1P, interest has focused on defining the direct protein targets of action. Several candidate proteins have been shown to interact with ceramide in vitro and in cells (Fig. 5). These include ceramide-activated Ser–Thr phosphatases (CAPPs), such as PP1 and PP2A, which bind ceramide in vitro⁴³. They are modestly activated by ceramide and show a preference for the natural stereoisomers. Studies in cells have linked the action of ceramide to CAPPs and the induction of protein dephosphorylation. For example, ceramide-inducing agents (such as TNFa or loading the cell with palmitate) induce dephosphorylation of the retinoblastoma gene product RB⁴⁴, PKCα⁴⁵, protein kinase B (PKB or AKT)⁴⁶ and other proteins in a ceramide-dependent manner. However, to demonstrate the direct cellular activation of phosphatases by ceramide, experimental evidence is needed (for example, observation of translocation). Defining the specific binding sites of this activation process could result in improved understanding of how these interactions occur and could also provide specific tools for investigating these pathways.

Figure 5: Examples of ceramide signalling pathways and their role in stress responses.

a | The de novo pathway in palmitate signalling and insulin resistance. Extracellular palmitate is taken up by cells where it is acylated to form palmitoyl CoA, which serves along with serine as the initial substrates for sphingolipid synthesis. Excess palmitate increases de novo synthesis, which could result in higher ceramide (Cer) levels. Accumulation of Cer has been shown to activate protein phosphatase-2A (PP2A), resulting in inhibition of AKT, a key mediator of the metabolic effects of insulin. Excessive Cer also activates PP1 phosphatase, and activation of both PP1 and PP2A can enhance cytotoxic and apoptotic responses. b | Regulation of acid sphingomyelinase (SMase) by PKCδ and oxidative stress. UV and ionizing radiation and several chemotherapeutic agents have been shown to activate acid SMase. UV activates PKCδ and induces reactive oxygen species (ROS), which are also shown to activate PKCδ. Both ROS and PKCδ have been implicated in activation of acid SMase, which may act in the lysosome to generate Cer, which in turn activates cathepsin D, leading to cleavage of the pro-apoptotic protein BID and induction of apoptotic responses. Acid SMase may also translocate to the plasma membrane where it acts on outer leaflet sphingomyelin (SM), resulting in Cer formation. Cer can enhance the formation of membrane microdomains (capping) and/or launch signalling pathways that are mediated by protein phosphatases (for example, PP2A) and kinases (for example, c-Jun N-terminal kinase (JNK) and PKCζ). CerS, ceramide synthase; DES, dihydroceramide desaturase; dhCer, dihydroceramide; dhSph, dihydrosphingosine; IRS, insulin receptor signalling; PI3K, phosphoinositide 3-kinase; PtdIns(3,4,5)P₃, phosphatidylinositol-3,4,5-trisphosphate; SPT, serine palmitoyl transferase.

In addition, ceramide has been shown to activate PKCζ^47,48, the kinase KSR⁴⁹ and cathepsin D⁵⁰. The latter has been proposed as a specific target for lysosomally generated ceramide, and may couple the action of lysosomal acid SMase to the mitochondrial pathway of apoptosis⁵⁰. Activation of PKCζ by ceramide has been implicated in the regulation of membrane potential, inhibition of AKT and pro-apoptotic functions^48,51.

S1P is one of the more soluble sphingolipids and is present in low nanomolar concentrations in the cell (but in high nanomolar concentrations in serum where it is associated with lipoproteins and albumin⁵²). It interacts with S1PRs⁵³, which are high-affinity G protein-coupled receptors (GPCRs) and the only known receptors for S1P that have been identified so far. The five S1PRs (S1PR1–5) display selective tissue expression that is crucial for their biological functions and use well-known GPCR intracellular signalling pathways to mediate their specific effects (Fig. 6). S1P also seems to exert S1PR-independent actions intracellularly, such as induction of calcium release⁸, although these mechanisms remain unknown. Because an intracellular molecular target for S1P has not been identified, its intracellular role continues to be a subject of speculation and significant interest.

Figure 6: SK–S1P receptors and signalling.

Multiple growth factors and cytokines activate sphingosine kinase (SK), probably by inducing its phosphorylation in a phospholipase D (PLD)-, protein kinase C (PKC)- and ERK-dependent manner. This leads to its translocation to the plasma membrane through binding to phosphatidic acid (PA) and phosphatidylserine (PS), where it may access its substrate sphingosine (Sph). Sphingosine-1-phosphate (S1P) is then generated intracellularly and may act on as-yet-unidentified intracellular targets. S1P is also released and binds to one of five S1P receptors (S1PR1–5), leading to differential activation of distinct G-protein-responsive pathways to activate Rac, Ras–ERK, PI3K–AKT–Rac, phospholipase C (PLC) and Rho. Different receptors are differentially expressed in different cells and tissues and couple to different G proteins, leading to distinct signalling pathways and cellular responses (reviewed extensively in Refs 38,94). ERK, extracellular signal-regulated kinase; PI3K, phosphoinositide 3-kinase.

Sphingolipid-mediated cell regulation

Research into bioactive sphingolipids continues to grow at an accelerated pace, with >18,000 PubMed entries to date on ceramide, sphingosine and S1P, mostly related to their signalling functions. As such, a comprehensive discussion of the many facets of these pathways is no longer tenable within the scope of a single review. This section will focus on three specific pathways of ceramide signalling and on the SK1–S1P pathway in inflammation and angiogenesis in order to introduce the key players and provide a glimpse of our current knowledge of sphingolipid-mediated cell regulation.

De novo synthesis of sphingolipids and stress responses. The de novo pathway of sphingolipid biosynthesis, which resides in the ER, has emerged as a key mechanism in eukaryotic cells for regulating the levels of ceramide and other sphingolipids. In mammalian cells, de novo synthesis of ceramide is enhanced in response to some chemotherapeutic agents, such as etoposide and daunorubicin^54,55, and other inducers of apoptosis such as stimulation of the B-cell receptor⁵⁶. The pro-apoptotic FAS pathway stimulates de novo synthesis that has been shown to activate PP1. This results in dephosphorylation of SR proteins, which are regulators of RNA splicing, the function of which is regulated by phosphorylation⁵⁷. This tilts the balance of pro- and anti-apoptotic mRNAs in favour of the pro-apoptotic caspase-9 and BCL-X⁵⁷.

Interestingly, the de novo pathway is metabolically geared to respond to changes in serine and palmitate concentrations because SPT displays K_m values for the two substrates that are in the range of their usual intracellular concentrations⁵⁸ (Fig. 1). Indeed, a recent study in yeast demonstrated that heat stress induces an acute influx of serine into the ER that drives de novo synthesis⁵⁹. Activation of this pathway in yeast results in transient accumulation of sphingoid bases, followed by sphingoid base phosphates and then ceramide. Multiple studies have indicated that sphingoid bases that are generated in this pathway mediate specific responses to heat stress⁶⁰, which include regulation of nutrient permeases⁶¹, cytoskeletal changes⁶², cell-cycle arrest⁵⁹ and RNA translation⁶³.

Regulation of de novo synthesis by palmitate may have a key role in diabetes and the metabolic syndrome (Fig. 5a). Emerging evidence suggests that palmitate loading results in the accumulation of ceramide, which activates PP2A and results in dephosphorylation and inactivation of AKT⁴⁶, a key mediator in insulin signalling and metabolic control. Ceramide may exert additional effects on AKT and on other targets that, together, result in a deterioration in insulin responsiveness and in the death of islet cells^64,65 (a process termed lipoptosis) and may contribute to other diabetic complications. A recent study in mice demonstrated that inhibition of ceramide synthesis, either by pharmacological treatments or genetic manipulations, prevented insulin resistance induced by fatty acids, obesity or glucocorticoids⁶⁵.

Although the significance of this pathway is now appreciated, several key questions remain to be addressed. Increased flux through the de novo pathway does not necessarily lead to ceramide accumulation unless further metabolic transformation (for example, to SM, C1P or glucosylceramide) is limiting or inhibited (as has been shown in some cases of apoptosis in which SM synthase and glucosylceramide synthase (GCS) are suppressed). Also, it is not clear if the de novo pathway acts exclusively in the ER or also in other ER-associated membranes (such as mitochondrial-associated membranes or the nuclear membrane). The location of the de novo pathway could possibly tie it to the regulation of ER stress, but this has yet to be evaluated.

Acid SMase and the salvage pathway in stress signalling. Multiple stress stimuli, such as ultraviolet (UV) and ionizing radiation, ligation of death receptors and chemotherapeutic agents (including platinum, paclitaxel and histone deacetylase inhibitors) have been shown to activate acid SMase, usually within a few minutes of cell stimulation (Fig. 5b). This has been demonstrated primarily by showing increased in vitro activity of the enzyme, and thus is an indication of some post-translational modification of acid SMase. In a few studies, activation of SMase was shown to be accompanied by translocation to the plasma membrane and the simultaneous generation of ceramide⁶⁶.

Beyond these correlations, several studies have provided more direct evidence for the involvement of acid SMase in ceramide generation and in downstream responses, either through using genetic mutants of the enzyme (obtained from patients with Niemann–Pick disease)⁶⁷, knockout mice that lack acid SMase⁶⁸, small interfering (si)RNA-mediated knockdown⁶⁹ or pharmacological inhibitors⁷⁰ such as one of the tricyclic amines, desipramine or imipramine. However, such pharmacological studies need to be interpreted with caution because these inhibitors have other targets, including acid ceramidase⁷¹. Gene knockout or knockdown studies have now clearly implicated acid SMase in mediating apoptotic and stress responses to ionizing and UV radiation^68,72, and to lower concentrations of FAS ligands⁷³. However, other studies have demonstrated increased sensitivity to FAS but reduced mitogenic stimulation of T cells in the knockout mouse⁷⁴.

The mechanisms that are involved in activation of acid SMase in response to stress agents are poorly defined. Recent results show that the enzyme is activated by reactive oxygen species and, perhaps more specifically, by nitrosative stress⁷⁵. Other recent studies provided evidence for regulation of acid SMase by phosphorylation in response to activation of PKCδ, and this was implicated in mediating the effects of UV radiation on activation of acid SMase⁷². These studies are beginning to provide important mechanistic links between stress response pathways that were previously thought to be distinct (acid SMase, nitrosative stress and PKCδ).

Ceramide that is generated by acid SMase has been proposed to reside either in the lysosome or at the plasma membrane (Box 1). Both locations are plausible because acid SMase has been conclusively shown to exist at both subcellular sites. In the lysosome, ceramide has been shown to interact directly with the protease cathepsin D and to activate it, which possibly causes cleavage and activation of the pro-apoptotic protein BID⁵⁰ (Fig. 5b). At the plasma membrane, the actions of ceramide are less well defined, but have been suggested to mediate receptor capping (microscopic clustering) and/or microdomain formation⁷⁰.

In an illustration of the complex nature of ceramide metabolism, it was recently shown that activation of acid SMase in response to phorbol esters also leads to a concomitant increase in ceramide formation through the salvage (or recycling) pathway⁷⁶. In this pathway, sphingosine is recycled to ceramide through the action of ceramide synthases (Box 1). This could explain some of the apparent discrepancies in the literature in which both acid SMase and the de novo pathway (based on inhibition of ceramide synthase by fumonisin B1) have been implicated in several cellular responses (for example, to doxorubicin). Importantly, fumonisin B1 inhibits ceramide synthases and does not distinguish de novo synthesis from recycling (Box 1). Therefore, it could be suggested in these cases that acid SMase is coupled to recycling.

Neutral SMase2 in cytokine action. Multiple studies have coupled activation of neutral SMase to the action of several extracellular cytokines and stress responses⁷⁷, and the list of inducers shows much overlap with activators of acid SMase. However, progress in the molecular dissection of neutral SMase-mediated pathways only developed following the recent identification of neutral SMase-2 (nSMase2) as a bona fide sphingomyelinase^78,79. Importantly, knockout of the enzyme was found to result in short stature, and a natural mutation in nSMase2 (fro/fro mouse) results in bone fragility^80,81.

nSMase2 has been implicated in several cell responses. It is acutely activated by the cytokines TNFa and interleukin (IL)-1 and has been shown to mediate the effects of IL-1 on signal transduction (in the phosphorylation of c-Jun N-terminal kinase (JNK))⁸², and the effects of TNFa on gene induction (involving endothelial nitric oxide synthase (eNOS))⁸³. The enzyme has also been shown to be induced during ageing in mouse hepatocytes and to mediate a decreased responsiveness of hepatocytes to IL-1 signalling⁸⁴. nSMase2 has been shown to mediate, at least in part, the effects of TNFa on cell adhesion and migration⁸⁵.

Several studies have implicated nSMase2 in the cytotoxic action of amyloid peptide-β, demonstrating attenuation of these responses in cells in which nSMase2 was knocked down with siRNA^86,87. Long-term induction of the enzyme has been observed upon cell confluence (the enzyme was initially cloned as the cell confluence-associated (CCA) gene⁸⁸) and has been implicated in mediating cell-cycle arrest that is induced by cell contact⁸⁹.

The mechanisms involved in activation of nSMase2 are not well defined. TNFa has been shown to induce translocation of the enzyme to the plasma membrane, in a mechanism that depends on activation of p38 mitogen-activated protein kinase (MAPK)⁸⁵. In confluent cells and tissue cells (hepatocytes), nSMase2 resides basally at the plasma membrane. Importantly, it appears to localize to the inner leaflet of the plasma membrane where, presumably, a minor pool of SM may reside⁹⁰.

SK1–S1P in inflammation and angiogenesis. An overwhelming amount of primary literature has established the participation of the SK1–S1P–S1PR axis in multiple signalling pathways^38,91. Many growth factors, such as epidermal growth factor and platelet-derived growth factor, as well as the cytokines TNFa and IL-1, activate SK1 acutely, resulting in transient elevations in the levels of S1P (usually 2–3 fold)²⁵. This stimulation requires PKC⁹², phospholipase D (PLD)⁹³ and/or the extracellular signal-regulated kinase (ERK) MAPKs, which act upstream of SK1 (Ref. 94). PKC may act directly or indirectly through PLD and ERK, whereas it has been proposed that ERK directly phosphorylates SK1 on residue Ser225 (Ref. 94), which is required for its translocation to the plasma membrane. PLD generates phosphatidic acid, which may directly bind SK1 and lead to its membrane association³³.

S1P that is generated by this process has been detected predominantly in the extracellular space and, given its zwitterionic character, these results suggest the need for active transport mechanisms. Indeed, it has been proposed recently that the ABC transporter ABCC1 is required for transporting S1P from the inner leaflet of the plasma membrane to the cell exterior⁴⁰. Alternatively, SK1 itself has also been detected extracellularly. Once outside the cell, S1P binds S1PRs with high affinity and launches typical GPCR signalling pathways⁹⁵ (Fig. 6).

At the cellular level, the SK1–S1P pathway has been best studied in relation to cytokine action, with multiple functions related to the pro-inflammatory actions of TNFa (and IL-1). Within 10 minutes of stimulation with TNFa, SK1 is activated in a mechanism that is dependent on TRAF (TNFa receptor-associated factor) and ERK⁹⁶. Knockdown studies have implicated SK1 in mediating the effects of TNFa on induction of cyclooxygenase-2 (COX2), the production of prostaglandins^37,97, the induction of adhesion molecules⁹⁸ and activation of eNOS⁸³. Interestingly, knockdown of S1P phosphatase or S1P lyase (the key enzymes that metabolize S1P) augmented prostaglandin production, concomitant with augmentation of S1P levels³⁷. This specifically indicates that S1P is the mediator of SK1 action and not subsequent metabolites.

At the organismal level, the SK1–S1P–S1P1R pathway has been clearly implicated in the regulation of angiogenesis because S1P1R is present at high levels on endothelial cells and is known to have important roles in angiogenesis; in addition, knockout of S1P1R results in embryonic lethality due to poor vascular development⁹⁹. The combined knockout of SK1 and SK2 also recapitulated this phenotype¹⁰⁰. Studies in endothelial and smooth muscle cells support key roles for S1P in regulating endothelial cell proliferation, migration and tube formation, and the activation of smooth muscle migration and proliferation¹⁰¹.

Furthermore, S1P1R along with S1P3R and S1P4R are enriched in lymphoid organs and have key roles in regulating lymphoid sequestration and/or egress from lymph nodes¹⁰². Recently, FTY270, a sphingosine analogue, was shown to be phosphorylated by SK2 and to act as a potent agonist of S1P receptors. FTY720 is currently in clinical studies for its roles in immune modulation (for example, in multiple sclerosis¹⁰³), thus underscoring the importance of S1P in the regulation of lymphocyte function and immunity.

It is also important to remember that SK1 and SK2 serve a key metabolic function in the penultimate breakdown of all sphingolipids, whereby the irreversible breakdown of the product S1P leads to metabolic exit from the sphingolipid pathway. Thus, changes in SK activity may have long-term effects, not only on the levels of S1P but also on ceramide, complex sphingolipids and sphingoid bases. This is well demonstrated in siRNA studies that show the significant accumulation of ceramide in response to knockdown of SK1 (Ref. 104).

Conclusions and future directions

The above examples focus on specific, regulated sphingolipid pathways and enzymes that have been more extensively studied than other enzymes and pathways, such as SM synthases, ceramidases, ceramide synthases, GCS, CERT, S1P phosphatase and S1P lyase, which are currently subject to intensive investigation. Sphingolipids other than ceramide and S1P are emerging as candidate bioeffector molecules. These include C1P, which has roles in activation of phospholipase A2 (Refs 105,106), regulation of vesicular trafficking, phagocytosis¹⁰, macrophage degranulation⁹ and in mitogenesis¹⁰⁷. Glucosylceramide has been implicated in resistance to chemotherapeutic agents¹², and may serve as the endogenous cargo for the P-glycoprotein transporter MDR1 (Ref. 108). Dihydroceramide, which had been shown to be inactive in apoptosis, has been implicated in mediating the growth inhibitory actions of fenretinide (a retinoid analogue used in the treatment of neuroblastoma), which has been shown to inhibit the activity of the desaturase^109,110,111. Lyso-sphingomyelin has been shown to induce multiple cellular effects, possibly mediated by binding to specific GPCRs¹¹². These additional pathways and bioactive metabolites clearly represent areas of future investigation.

The examples described here highlight important roles for sphingoid bases, ceramide and S1P in regulation of the cell cycle, stress responses, pro-inflammatory pathways and cell migration, with key roles in apoptosis, inflammation and angiogenesis. These cellular roles extend to organismal function and pathobiology and have implications for the understanding of cancer biology, arthritis and inflammation, diabetes, immune function and neurodegenerative disorders^113,114,115. Notwithstanding the many difficulties and complexities in studying bioactive lipids, these examples demonstrate the power of modern cell biology and biochemistry in elucidating these pathways through the use of knockout mice, siRNA technology, liquid chromatography–mass spectrometry-based 'lipidomic' analysis, confocal microscopy, chemical biology, mathematical modelling and systems biology approaches.

There are several take-home messages. First, this ever-enlarging spectrum of bioactive lipids has, in effect, reversed our assumptions about the putative functions of lipid metabolism and the roles of minor lipids. Past generations of investigators may have been sceptical of the significance of lipids in cell biology, but we are now at a point where one can almost assume that lipids, which have regulated levels, are 'bioactive until proven otherwise'. Second, it is clear that these pathways are now accessible to the cell biologist, provided that due attention is given to the peculiarities of bioactive lipids and their pathways, as discussed in this review. Third, sphingolipid-mediated pathways operate at the level of individual organelles, and thus should be studied as such. Functions of ceramide in the ER are clearly different from those at the plasma membrane; even in the membrane, ceramide seems to regulate different functions depending on whether it is formed at the inner or outer leaflet. Similar considerations may apply to SK1 and SK2 (Ref. 116). Fourth, the complexity of sphingolipid metabolism and the metabolic interconversion of bioactive sphingolipids provide the cell with an extremely rich repertoire for not only generating multiple signals, but also for fine-tuning specific responses. Therefore, the field of bioactive sphingolipids constitutes its own area of biological '-omics', the 'sphingolipidome'. This generates the anticipation that many modern analytical techniques and the mathematical approaches of systems biology will be brought to bear on this field, as has been trialled with modelling of the sphingolipid metabolic pathways of yeast¹¹⁷.

The study of bioactive lipids is only now coming to the forefront of cell biology, representing possibly one of the last frontiers in molecular cell biology. Clearly, more work is required to dissect each of the many regulated pathways of bioactive sphingolipids and define the mechanisms of regulation of enzymes, roles of the pathways in specific cell responses, and mechanisms by which individual bioactive sphingolipids, which act in specific subcellular compartments, mediate those actions.

Box 1 | Compartmentalization of pathways of sphingolipid metabolism

De novo synthesis of sphingolipids occurs in the endoplasmic reticulum (ER) and possibly in ER-associated membranes, such as the perinuclear membrane and mitochondria-associated membranes (MAMs; see figure). Ceramide (Cer) formed in this compartment is transported to the Golgi, which is the site of synthesis of sphingomyelin (SM) and glucosylceramide (GluCer), with the latter serving as the precursor for complex glycosphingolipids (GSLs). The transport of Cer to the Golgi occurs either through the action of the transfer protein CERT, which specifically delivers Cer for SM synthesis, or through vesicular transport, which delivers Cer for the synthesis of GluCer. In turn, transfer of GluCer for GSL synthesis requires the action of the recently identified transport protein FAPP2 (Ref. 118). GluCer appears to be synthesized on the cytosolic side of the Golgi, and requires flipping to the inside of the Golgi for the synthesis of complex GSLs, possibly with the aid of the ABC transporter, P-glycoprotein (also known as MDR1) (not shown). Delivery of SM and complex GSLs to the plasma membrane appears to occur by vesicular transport. Acid sphingomyelinase (SMase) that is present in the outer membrane leaflet or neutral SMases in the inner leaflet of the bilayer can metabolize SM to Cer and subsequently other bioactive lipids. The mechanisms by which sphingolipids may cross lipid bilayers are discussed in Fig. 4. This metabolic network also extends to the circulatory system, where many of the enzymes (acid SMase, neutral ceramidase (CDase) and sphingosine kinase (SK)) have been detected, often in association with lipoproteins that are rich in SM and Cer. Internalization of membrane sphingolipids proceeds through the endosomal pathway whereby SM and GluCer may reach the lysosomal compartment, where they are degraded by the actions of SMases and glucosidases to form Cer. Cer is hydrolysed by acid CDase to form sphingosine (Sph). Sph may exit the lysosome, although its ionizable positive charge favours partitioning in lysosomes. Therefore, it is postulated that the action of SK1 at or near the lysosome may be required to 'trap' Sph through phosphorylation. Subsequent dephosphorylation by S1P phosphatases in the ER would recycle the salvaged Sph to Cer (a similar role for SK in delivering exogenous sphingoid bases to internal metabolic pathways has been demonstrated in yeast). Sph formed from this salvage pathway is also sufficiently soluble in cytosol to move among membranes. The salvage pathway is shown by the dashed arrows. 3KdhSph, 3-keto-dihydrosphingosine; dhCer, dihydroceramide; dhSph, dihydrosphingosine; S1P, sphingosine-1-phosphate.

Box 2 | Unique and particular properties of bioactive lipids

What makes a lipid bioactive? The central tenet of 'bioactivity' of lipids stipulates that changes in lipid levels result in functional consequences. This simplest (but not exclusive) formulation requires a module of lipid signalling that is composed of a regulated enzyme, the bioactive lipid and specific downstream targets. For example, sphingosine kinase is activated in response to cytokines and growth factors, and results in the formation of sphingosine-1-phosphate, which proceeds to act on G protein-coupled receptors to initiate intracellular signalling (Figs 1, 6). As such, specific modules of lipid signalling do not differ organizationally from the canonical paradigm of cell signalling that has been established with cyclic AMP. Beyond this apparent simplicity, however, lies a vast complexity that arises from the metabolic interconnection of bioactive lipids (Fig. 2) and the hydrophobic nature of enzymes of lipid metabolism. The lipid mediators themselves are difficult to study as a result of their hydrophobic nature (Figs 2, 3). Bioactive sphingolipids are often present at low levels, meaning that there is a paucity of suitable quantitative analytical methods for their characterization. Poor solubility generates problems in the cellular pharmacology of bioactive lipids, and this has necessitated the development of either 'unusual' delivery devices (for example, albumin conjugates, cyclodextrin loading) or the use of amphiphilic solvents such as ethanol or dodecane–ethanol. Each of these generates its own limitations because they may inadvertently modulate lipid composition and/or membrane structure. As a complementary and alternative approach, investigators are increasingly able to probe the cellular functions of endogenous sphingolipids through the use of pharmacological inhibitors or RNA interference. The significant hydrophobicity of bioactive lipids also restricts their action to their site of production in specific subcellular compartments, a problem that is rarely encountered with non-lipid mediators. This restriction may even apply to membrane 'sidedness' and/or to microdomains within the same membrane. A familiarity with these properties is necessary for understanding and predicting the behaviour of bioactive lipids.