Membranes in Balance: Mechanisms of Sphingolipid Homeostasis

Abstract

Sphingolipids and their metabolites play key cellular roles both as structural components of membranes and as signalling molecules that mediate responses to physiologic cues and stresses. Despite progress during the last two decades in defining the enzymatic machinery responsible for synthesizing and degrading sphingolipids, comparatively little is known about how these enzymes are regulated to ensure sphingolipid homeostasis. Here we review new insights into how cells sense and control sphingolipid biosynthesis and transport. We also discuss emerging evidence that sphingolipid metabolism is closely coordinated with that of sterols and glycerolipids and with other processes that occur in the secretory pathway. An improved understanding of sphingolipid homeostasis promises to shed light on basic processes in cell biology and disease, including how cells establish and maintain the complex membrane composition and architecture that is a defining feature of eukaryotic cell biology.

Introduction

Eukaryotic cell membranes have a remarkably complex composition and organization that is critical to the wide range of cellular processes in which they participate. Sphingolipids and their metabolites are a major component of these membranes and also have key roles as signaling molecules. This functional diversity is enabled by their structural diversity: sphingolipids encompass a large and essential group of lipids including sphingosines, ceramides, sphingomyelin, and various phosphorylated and glycosylated derivatives (Figure 1). Sphingolipids form cellular membranes together with glycerolipids and sterols, but are distinguished from other lipids by their chemical structures, physical properties, and dedicated enzymatic machinery (van Meer et al., 2008). They utilize two key lipid building blocks – long-chain bases (LCBs) and fatty acids - and use serine rather than glycerol as the backbone to which acyl chains are attached (Figure 1A; see Table 1 for full list of abbreviations). These characteristic core features of sphingolipids, together with their head-group modifications, enable them to have unique and essential functions in eukaryotic cells.

Figure 1. Overview of sphingolipid metabolism in yeast and mammals.

(A) Structures of sphingolipid core components are shown, with primary sites of chemical modification and structural diversity highlighted. Long-chain bases (LCBs) are made from the condensation of serine with fatty acid-Coenzyme A (FA-CoA) conjugates. LCBs can be phosphorylated at position a, hydroxylated at position b (yielding phytosphingosine, the major yeast LCB), or desaturated at position c (yielding sphingosine, the major mammalian LCB). Ceramides are produced by N-acylation of LCBs with FA-CoA or very long-chain fatty acid-Coenzyme A (VLCFA-CoA) molecules. The N-linked fatty acids can be hydroxylated at position d and vary in chain length (as indicated by e). The head-group position of ceramides (a) can be phosphorylated as in LCBs. Additionally, the ceramide head-group can be derivatized with inositol and mannose moieties to yield the major yeast sphingolipids, with phosphorylcholine to yield sphingomyelin (SM), or with complex glycans to yield glycosphingolipids (GSLs). (B) Sphingolipid biosynthetic pathways in yeast (left) and mammals (right) are shown, with key steps highlighted. Biosynthesis begins in the ER (bottom) with LCB and ceramide (Cer) production. Ceramides are subsequently transported by vesicular and non-vesicular means to the Golgi, where they undergo modification at the head-group position. In mammalian cells, CERT (Ceramide transport protein) is a ceramide transporter, and FAPP2 (Four-phosphate adaptor protein 2) is a glucosylceramide (GluCer) transporter. In the Golgi, phosphoinositol and mannose moieties are added to yeast ceramides to yield inositolphosphorylceramide (IPC), mannosyl-inositolphosphorylceramide (MIPC), and mannosyl-diinositolphosphorylceramide (M(IP)₂C). Similarly, phosphorylcholine and glycan modifications are added to mammalian ceramides in the Golgi to yield SM and GSLs, respectively. Mature sphingolipids are ultimately trafficked to the plasma membrane (top), where they are most abundant. Many of these reactions are reversible, enabling breakdown of complex sphingolipids and generation of signaling molecules such as sphingosine-1-phosphate (S1P). See text for details.

Table 1.

List of abbreviations

Abbreviation	Standard name
25-HC	25-Hydroxycholesterol
Cer	Ceramide
CERT	Ceramide transport protein
CPE	Ceramide phosphoethanolamine
ER	Endoplasmic Reticulum
ERK	Extracellular signal-regulated kinase
FA-CoA	Fatty acid-Coenzyme A
FAPP2	Four-phosphate adaptor protein 2
FFAT	Di-phenylalanine in an acidic tract
GluCer	Glucosylceramide
GPCR	G protein-coupled receptor
GSL	Glycosphingolipid
IPC	Inositolphosphorylceramide
LCB	Long-chain base
LCB-P	Long-chain base phosphate
MAG	Myelin-associated glycoprotein
M(IP)2C	Mannosyl-diinositolphosphorylceramide
MIPC	Mannosyl-inositolphosphorylceramide
ORD	Oxysterol-binding-protein-related domain
ORMDL	ORM1-like (S. cerevisiae)
OSBP	Oxysterol-binding protein
PC	Phosphatidylcholine
PE	Phosphatidylethanolamine
PH	Pleckstrin homology
PI	Phosphatidylinositol
PI(4,5)P2	Phosphatidylinositol-4,5-bisphosphate
PI4P	Phosphatidylinositol-4-phosphate
S1P	Sphingosine-1-phosphate
SACM1L	Suppressor of actin mutations 1-like
SAM	Sterile alpha motif
SM	Sphingomyelin
SMase	Sphingomyelinase
SMS	Sphingomyelin synthase
SMSr	Sphingomyelin synthase-related
Sph	Sphingosine
SphK	Sphingosine kinase
SPOTS	Serine palmitoyltransferase, Orm1/2, Tsc3, Sac1
SPTLC	Serine palmitoyltransferase long-chain base subunit
SREBP	Sterol regulatory element binding protein
START	Steroidogenic acute regulatory protein-related lipid transfer
SUR	Suppressor of RVS161 and RVS167
TOR	Target of rapamycin
TORC2	Target of rapamycin complex 2
TSC	Temperature-sensitive CSG2 suppressor
VAP	Vesicle-associated membrane protein-associated protein
VLCFA	Very long-chain fatty acid
VLCFA-CoA	Very long-chain fatty acid-Coenzyme A

A full description of the many functions of sphingolipids is beyond the scope of this review, but below we highlight some of their primary cellular roles. First, sphingolipids serve as structural components of membranes whose chemical features modulate the physical properties of lipid bilayers and the activity of trans-membrane proteins. Sphingolipids therefore control many membrane-associated cellular processes including endocytosis, intracellular trafficking, and signal transduction by membrane receptors (Lingwood and Simons, 2010; Lippincott-Schwartz and Phair, 2010). For example, sphingolipid-rich membrane domains are thought to provide a platform for signaling by the T-cell and B-cell receptors (Lingwood and Simons, 2010). Sphingolipids are also highly abundant on the extracellular face of the plasma membrane, where they participate in cell-cell communication and host-pathogen interactions (Lopez and Schnaar, 2009; Tsai et al., 2003).

Second, sphingolipids have important functions as signaling molecules in a wide array of biological processes. In an intracellular context, ceramides are potent inducers of apoptosis in response to stresses such as ionizing radiation and chemotherapeutics (Deng et al., 2008; Hannun and Obeid, 2008; Taha et al., 2006), while LCBs and their derivatives participate in the heat-shock response in yeast (Dickson et al., 2006). Some sphingolipids also play critical roles as extracellular signaling molecules. For example, phosphorylated LCBs such as sphingosine-1-phosphate (S1P) regulate chemotaxis, cardiac development, the migration of osteoclast precursors between bone and peripheral blood, and the formation and barrier function of the vasculature (Fyrst and Saba, 2010; Ishii et al., 2009; Rosen et al., 2009; Spiegel and Milstien, 2002). S1P also has prominent roles in immune-mediated processes and pathologies including lymphocyte egress from lymphoid organs, T cell lineage specification, allergic inflammation, and sepsis (Liu et al., 2010a; Puneet et al., 2010; Rosen et al., 2009; Schwab and Cyster, 2007).

Three principal mechanisms have emerged by which sphingolipids mediate their diverse cellular roles. First, some sphingolipid functions are attributable to their effects on the physical properties of lipid bilayers. As reviewed recently by Lingwood and Simons (2010) and Lippincott-Schwartz and Phair (2010), the structural features of sphingolipids can influence the order of the lipid phase and the curvature and thickness of membranes. Such properties likely underlie the insulating and barrier functions of sphingolipid-rich membranes found in myelin and skin, respectively (Holleran et al., 2006; Stoffel and Bosio, 1997). Additionally, recent reports provide striking examples of sphingolipid-mediated membrane curvature during intra-luminal vesicle budding (Trajkovic et al., 2008) and during membrane remodeling events induced by viruses, protein toxins, and plasma membrane damage (Ewers et al., 2010; Römer et al., 2010; Tam et al., 2010). Sphingolipids also preferentially associate with each other and with sterols to form dynamic lateral membrane inhomogeneities or “rafts” (Lingwood and Simons, 2010). Such lipid domains can interact with membrane proteins and thereby modulate their biogenesis, trafficking, and activity (Lingwood and Simons, 2010; Lippincott-Schwartz and Phair, 2010).

Second, sphingolipid metabolites play critical roles as ligands that bind to and regulate the activity of enzymes and signaling proteins such as kinases and membrane receptors. This mode of action is primarily attributed to LCBs, LCB-Ps, and ceramides, with one of the best-characterized examples being the role of extracellular S1P as a ligand for G protein-coupled receptors (GPCRs) (Fyrst and Saba, 2010; Hannun and Obeid, 2008; Rosen et al., 2009). These S1P receptors, encoded by S1PR1-5, are responsible for many of the above-mentioned functions of S1P in development and disease and are also targets of the promising immuno-modulatory drug FTY720 (Rosen et al., 2009). Recently, exciting new functions have also been found for intracellular S1P in modulating histone deacetylase and ubiquitin ligase enzyme activities (Alvarez et al., 2010; Hait et al., 2009). It should be noted that these functions of S1P and other LCB metabolites may in part arise from their moderate aqueous solubility, which likely enables them to interact with target proteins outside of the lipid bilayer context (Hannun and Obeid, 2008). Lastly, a third basis for sphingolipid-mediated functions is seen for glycosphingolipids (GSLs), in which complex head-group glycans serve as surface-exposed membrane elements that can be recognized by a variety of proteins. For example, GSLs are bound by the SV40 virus capsid, lectins such as myelin-associated glycoprotein (MAG) and E-selectin, and, in autoimmune disease, by anti-GSL auto-antibodies (Ariga and Yu, 2005; Lopez and Schnaar, 2009; Tsai et al., 2003).

In the past several years our understanding of sphingolipid functions and metabolism has grown significantly. We now have a nearly complete inventory of the enzymes that are required for their synthesis and degradation and a growing appreciation of their many cellular roles. In light of these diverse and important functions, it is not surprising that the levels of sphingolipids and their metabolites are tightly regulated and that perturbations to sphingolipid metabolism cause cellular stress and organismal pathology. Indeed, it has been appreciated for over twenty years that cells adjust sphingolipid production in response to metabolic needs (van Echten et al., 1990). However, until quite recently little was known about the mechanisms by which sphingolipid homeostasis is achieved. The rudimentary state of understanding of sphingolipid homeostasis is particularly striking when we contrast it to studies of glycerolipids and sterols, for which multiple feedback pathways have been found that link the uptake, degradation, and synthesis of these lipids to their cellular levels (Brown and Goldstein, 2009; Nohturfft and Zhang, 2009; Zhang and Rock, 2008). Here, we discuss new insights into how cells sense and maintain sphingolipid levels. While these lipids undergo many interesting species-specific and tissue-specific modifications, such as glycosylation, we focus on the regulation of the conserved core ceramide structure of sphingolipids. We also call particular attention to the mounting evidence that sphingolipid metabolism is closely coordinated with that of other lipid classes and with other fundamental processes of the secretory pathway.

Biosynthesis of sphingolipids

Sphingolipids are synthesized by a series of membrane-embedded enzymes localized within the secretory pathway (Figure 1). Many studies over the last two decades have elucidated the enzymatic machinery responsible for sphingolipid synthesis, with especially important contributions coming from genetic screens in yeast that identified the roles of TSC (Temperature-Sensitive CSG2 suppressor) and SUR (Suppressor of RVS161 and RVS167) genes in sphingolipid metabolism (Beeler et al., 1998; Desfarges et al., 1993). Sphingolipid synthesis begins on the cytosolic face of the ER membrane with the condensation of serine and fatty acid-Coenzyme A (FA-CoA) conjugates (Figure 1) (Dickson et al., 2006; Hannun and Obeid, 2008). This reaction is the rate-limiting step in de novo sphingolipid synthesis and is catalyzed by the enzyme serine palmitoyltransferase (formed in yeast by Lcb1, Lcb2, and the accessory protein Tsc3). A reduction step then yields key intermediates known as sphingosines or long-chain bases. These LCBs can be phosphorylated to yield LCB phosphates (LCB-Ps), or they can be N-acylated with a second FA-CoA to produce ceramides. Notably, these amide-linked acyl chains are often saturated and tend to be longer than those seen in glycerolipids (Lingwood and Simons, 2010); as such, they are often generated by a distinct set of enzymes that elongate standard FA-CoAs to produce very long-chain fatty acid-Coenzyme A molecules (VLCFA-CoAs) (Jakobsson et al., 2006). Thus a major source of sphingolipid diversity arises from the synthesis of fatty acids of variable chain length by the elongase machinery and from the specificity of different ceramide synthase isoforms for particular FA-CoAs and VLCFA-CoAs (Figure 1A) (Denic and Weissman, 2007; Levy and Futerman, 2010; Menuz et al., 2009). Further complexity is introduced by enzymes that can hydroxylate or desaturate the LCB and N-linked fatty acid components of sphingolipids (Figure 1A) (Degroote et al., 2004). In yeast, the LCB moiety is typically hydroxylated to yield phytosphingosine-based ceramides, whereas LCB desaturation in mammalian cells generates sphingosine-based ceramides (Figure 1).

After ceramides are generated in the ER, they are then transported to the Golgi apparatus for further elaboration (Figure 1B). This transport can occur by either vesicular trafficking or by dedicated transporter proteins (in mammalian cells, CERT mediates ER-to-Golgi transport of ceramide while FAPP2 mediates intra-Golgi transport of glucosylceramide) (D’Angelo et al., 2007; Funato and Riezman, 2001; Hanada et al., 2003). In the Golgi, ceramides undergo species-specific modifications at the head-group position. Many of these modifications are glycosylation events, although in mammalian cells phosphorylcholine can also be coupled to ceramide to form sphingomyelin (SM) (Figure 1B). Additionally, while these glycosylations are relatively simple in yeast, mammalian cell glycosphingolipids possess a tremendous diversity of glycan structures that gives rise to sub-classes such as sulfatides, cerebrosides, and gangliosides (Degroote et al., 2004). Interestingly, these glycosylations occur in the lumen of the Golgi, indicating that ceramides “flip” by an as yet unknown mechanism from the head-group facing the cytosolic surface of the lipid bilayer to facing the luminal/extracellular surface (van Meer and Hoetzl, 2010). Ultimately, sphingolipids continue to move through the secretory pathway to the plasma membrane, where they are most abundant (Lingwood and Simons, 2010).

In considering the regulation of sphingolipid metabolism, two key features should be kept in mind. First, many sphingolipid enzymes and their substrates are membrane-embedded and not diffusible. As such, enzyme activities can be regulated independently at distinct sub-cellular locations, with corresponding local differences in sphingolipid amounts and functions. Second, many of the reactions in sphingolipid metabolism are reversible, allowing for the rapid inter-conversion of different metabolic intermediates in the pathway. This reversibility has been shown to be important both for the generation of signaling sphingolipid metabolites such as S1P and for catabolic recycling of complex sphingolipids. Nonetheless some key steps are thought to be effectively irreversible, including the initial step in LCB production catalyzed by serine palmitoyltransferase and the degradation of sphingolipids by an ER-localized lyase that cleaves LCB-Ps to acyl aldehydes and phosphoethanolamine (Figure 1B).

Regulation of sphingolipid metabolism

As has been seen for sterols and glycerolipids, the regulation of sphingolipid metabolism is likely to be of fundamental cellular importance and to be governed by multiple interconnected mechanisms. Complex regulation is particularly likely in light of the fact that many biosynthetic intermediates, in addition to the sphingolipid end products, have potent biological activities. For example, mutations or drugs that cause accumulation of LCB-Ps or ceramides cause rapid growth arrest in yeast (Nickels and Broach, 1996; Zhang et al., 2001). Similarly, in mammalian cells the relative amounts of ceramide and S1P are proposed to form a ‘rheostat’ in which the balance between S1P-mediated pro-survival signals and ceramide-mediated pro-apoptotic signals determines cellular responses to diverse stimuli (Spiegel and Milstien, 2002). Consistent with the roles of sphingolipids in many basic cellular processes, their misregulation causes cellular stress and has pathological effects. In yeast, inhibition of sphingolipid synthesis blocks growth, while over-production of sphingolipids has been shown to trigger the unfolded protein response and is toxic (Breslow et al., 2010; Dickson et al., 2006; Han et al., 2010). Similarly, mutations in any of several different enzymes needed for lysosomal sphingolipid catabolism cause sphingolipid accumulation and human diseases such as Tay-Sachs, Niemann-Pick, Farber’s Disease, Fabry Disease, and Gaucher Disease (Ginzburg et al., 2004).

Although many aspects of sphingolipid production are likely to be tightly controlled, we focus here on the mechanisms that regulate synthesis of the core ceramide backbone common to all sphingolipids. While sphingolipid export and uptake are also likely to be subject to regulation, these pathways remain incompletely understood. We also draw a conceptual distinction between mechanisms that modulate sphingolipid enzymes in order to generate acute signaling responses and mechanisms that set and adjust steady-state levels of sphingolipids and their metabolites, as the principles underlying these types of regulation may differ. In the case of signaling by metabolites such as ceramide and LCB-Ps, basal levels are typically quite low compared to the predominant cellular sphingolipids (Ejsing et al., 2009), and metabolic turnover is rapid (Alvarez-Vasquez et al., 2005). Thus, intracellular concentrations often increase several-fold in response to a stimulus, leading to downstream signal transduction, and then return to baseline quite rapidly (Alvarez-Vasquez et al., 2005). By contrast, for homeostatic mechanisms that sense and adjust steady-state levels of abundant sphingolipids, lipid levels likely must be continuously monitored and maintained in a narrow range, as even small concentration changes are physiologically important for properties such as membrane fluidity. This is in fact precisely the type of regulation that has been found for sterol homeostasis mediated by the SREBP pathway, in which even 50% changes in ER cholesterol levels induce a switch-like change in SREBP activity (Radhakrishnan et al., 2008). Below we describe known mechanisms that regulate sphingolipid metabolism.

Regulation of sphingolipid signaling

First we discuss briefly regulatory mechanisms in the context of sphingolipid signaling, especially with respect to S1P and ceramide. This topic remains an area of intensive research, and readers seeking a more detailed discussion are referred to recent reviews by Hannun and Obeid (2008) and Fyrst and Saba (2010). One common theme of sphingolipid-based signaling is that many features are shared with canonical paradigms of receptor-triggered signal transduction. For example, acute production of S1P is typically mediated by activation of sphingosine kinase 1 or sphingosine kinase 2 (SphK1/2). These soluble enzymes are activated by recruitment from the cytosol to the plasma membrane and by receptor-induced phosphorylation mediated by ERK1/2 or PKC (Johnson et al., 2002; Raben and Wattenberg, 2009). Similarly, ceramide generation during stress-induced apoptosis results from activation of sphingomyelinase (SMase) by membrane translocation and PKCδ phosphorylation (Zeidan and Hannun, 2007). Alternatively, ceramide can be produced by de novo synthesis via activation of serine palmitoyltransferase by an unknown, post-transcriptional mechanism (Perry et al., 2000). Interestingly, it appears that it is mitochondrial ceramide that is responsible for functions in apoptosis (Birbes et al., 2001), raising the question of how the intracellular site of ceramide accumulation is determined.

One variation on this theme is seen during heat-shock-induced LCB and ceramide production in S. cerevisiae. In this case, sphingolipid metabolite signaling results from stimulation of de novo synthesis by heat-induced changes in substrate availability rather than by direct enzyme activation. Specifically, endogenous fatty acid synthesis and uptake of extracellular serine were both shown to be required for heat-induced sphingolipid synthesis (Cowart and Hannun, 2007). Serine import is stimulated two-fold upon heat shock, and elevated uptake appears to be necessary and sufficient for transient elevation of LCB levels. Consistent with these observations, the affinity of serine palmitoyltransferase for serine is very similar to its intracellular concentration; thus, changes in internal serine levels will directly affect LCB synthesis (Dickson et al., 2000). Given that fatty acids are also shared with other lipid metabolic processes, the concentrations of substrates and their shunting between competing pathways may have important roles in sphingolipid signaling and metabolism.

Mechanisms of sphingolipid homeostasis

It was reported over twenty years ago that sphingolipid synthesis is regulated in response to metabolic demand. These initial studies showed that when neuronal cultures are supplied with exogenous LCBs, de novo sphingolipid synthesis is suppressed, as measured by the incorporation of radio-labeled serine (van Echten et al., 1990). This echoes similar results seen in cholesterol metabolism (Brown and Goldstein, 2009), and subsequent studies suggested that the decrease in de novo sphingolipid synthesis was due to a reduction in serine palmitoyltransferase activity (Mandon et al., 1991). This observation provided empirical evidence that cells have mechanisms that sense and coordinate sphingolipid production. Several studies have recently provided promising initial insights into the molecular details of such phenomena.

Two notable factors that regulate steady-state sphingolipid levels were first identified via efforts to characterize the core biosynthetic machinery. In the first case, the starting point was the TSC yeast genetic screen that had previously revealed several key sphingolipid enzymes. Mutations to genes encoding the TORC2 kinase complex, which includes the Target of Rapamycin (Tor) kinase and associated regulatory subunits, were also isolated in the TSC screen, suggesting a role for this kinase in sphingolipid production (Beeler et al., 1998). Aronova et al. have provided mechanistic insights into this genetic connection, finding that TORC2 activity is required for maximal ceramide synthase activity and thus for normal levels of LCBs, LCB-Ps and ceramides (Aronova et al., 2008). TORC2’s effects appear to be mediated by the Ypk2 kinase and to be opposed by the calcineurin phosphatase, a finding that is particularly interesting in light of reports that LCBs may directly activate Ypk2 (Liu et al., 2005). Aronova et al. thus propose a potential feed-forward mechanism in which LCBs are both substrates for and indirect activators of ceramide synthase. However, the exact means by which TORC2 increases ceramide synthase activity, the physiologic conditions that modulate TORC2 signaling, and the potential conservation of this regulation await further characterization.

A second factor regulating sphingolipid metabolism was initially identified by its homology to the two known SM synthase genes SMS1 and SMS2. SMSr (for SM synthase-related) is an ER-resident transmembrane protein but, unlike SMS1 and SMS2, SMSr does not synthesize SM (Vacaru et al., 2009). SMSr is instead a ceramide ethanolamine phosphotransferase that synthesizes ceramide phosphoethanolamine (CPE). Interestingly, Vacaru et al. found that RNA interference against SMSr causes a dramatic increase in ceramide and glucosylceramide (GluCer) levels that cannot be accounted for by loss of SMSr’s rather modest CPE-producing activity. The authors propose instead that SMSr is a ceramide sensor that may act via its sterile alpha motif (SAM) domain to modulate sphingolipid production in response to ceramide levels. When this sensor’s function is blocked, unregulated ceramide accumulation occurs, leading to structural defects in the secretory pathway including Golgi fragmentation. Two immediate questions that arise from this interesting model is how SMSr and/or CPE production regulate sphingolipid synthesis and how this regulation is coupled to ceramide levels.

Although TORC2 and SMSr activity are required for normal sphingolipid metabolism, it is not yet clear if they fit the canonical model of feedback regulation in which components both control and sense sphingolipid levels. Compelling evidence for this type of regulation is seen however for the mammalian ceramide transporter CERT. First discovered by Hanada et al., CERT is a soluble protein possessing in vitro ceramide transfer activity between distinct lipid vesicles (Figure 2A). This activity is attributable to its ceramide-binding START domain and to two membrane recruitment elements: an FFAT motif that mediates binding to the ER-localized protein VAP and a PH domain that recognizes Golgi-enriched phosphatidylinositol-4-phosphate (PI4P) (Hanada et al., 2003). An appealing model therefore is that these ER- and Golgi-binding domains allow CERT to hand-off or shuttle ceramide from the ER to the Golgi, where SM and GSL synthesis continues.

Figure 2. CERT and Orm proteins regulate sphingolipid homeostasis.

(A) CERT catalyzes inter-membrane transport of ceramide (Cer; indicated in green) from the ER (top) to the Golgi (bottom). CERT binding to ER membranes is mediated by an FFAT motif that interacts with the ER-localized protein VAP; binding to Golgi membranes is mediated by a PH domain that recognizes phosphatidylinositol-4-phosphate (PI4P; indicated in orange), a phosphoinositide primarily found in the Golgi. CERT’s START domain binds to ceramides and mediates lipid transfer. The activity of CERT is inhibited by phosphorylation, and this phosphorylation is regulated in response to sphingolipid levels. When sphingolipid levels are low, CERT is dephosphorylated and ceramide transport is stimulated (left). When sphingolipid levels are high, CERT is phosphorylated and ceramide transport is reduced (right). (B) Orm proteins are homeostatic regulators of serine palmitoyltransferase, the ER-localized enzyme that catalyzes the condensation of serine with FA-CoAs in the first and rate-limiting step in de novo sphingolipid synthesis. In yeast, Orm1/2 form a complex (termed the SPOTS complex) that includes serine palmitoyltransferase (Lcb1, Lcb2, and the accessory protein Tsc3) and the phosphoinositide phosphatase Sac1. When cellular sphingolipid levels are sufficient for metabolic needs (left), Orm proteins act as inhibitors of serine palmitoyltransferase activity and limit sphingolipid production. However, when sphingolipid levels are inadequate (right), the Orm proteins are inactivated by phosphorylation, enabling compensatory up-regulation of sphingolipid production to meet metabolic demand. This phosphorylation-induced change in Orm protein activity also reduces the ability of Orm proteins and their binding partners to assemble into higher-order multimeric states.

Hanada and others have subsequently found that CERT-mediated transport of ceramides is regulated. Specifically, CERT is inhibited by phosphorylation, and this phosphorylation is reduced when sphingolipid metabolism is perturbed (Kumagai et al., 2007) (Figure 2A). Incubation of HeLa cells with myriocin, a potent serine palmitoyltransferase inhibitor, or treatment with recombinant bacterial SMase to degrade plasma membrane SM both resulted in CERT activation (i.e. a shift from partially phosphorylated to fully dephosphorylated CERT) (Kumagai et al., 2007). This regulation may therefore serve as a feedback mechanism to increase ceramide transport when sphingolipid biosynthesis is disrupted. One potential caveat is that up-regulation of ceramide trafficking may need to be coupled to similar increases in sphingolipid synthesis to adequately compensate for changes in sphingolipid levels.

Orm family proteins and sphingolipid homeostasis

How then might cells adjust sphingolipid synthesis according to metabolic demand? One mechanism for this type of feedback control comes from our investigation of the ORM gene family (Breslow et al., 2010). ORM genes are evolutionarily conserved and include two isoforms in S. cerevisiae (ORM1 and ORM2) and three isoforms in humans (ORMDL1-3) (Hjelmqvist et al., 2002). They encode transmembrane proteins localized to the ER, but lack any known protein domains. Beginning from a functional genomics approach based on large-scale measurement of genetic interactions in yeast, we found that over-expression of ORM genes produces a phenotypic signature very similar to that seen when serine palmitoyltransferase activity is reduced. Deletion of ORM2 produces the opposite phenotype, suggesting that Orm proteins are negative regulators of sphingolipid production. Consistent with this prediction, deletion of yeast ORM genes causes elevated levels of sphingolipids, whereas increased ORM gene activity reduces sphingolipid concentrations. Orm proteins negatively regulate sphingolipid metabolism by forming a conserved, roughly stoichiometric complex (termed the SPOTS complex) with serine palmitoyltransferase and the phosphoinositide phosphatase Sac1 (Figure 2B). Han et al. (2010) have reported a similar function for Orm proteins and have also shown that loss of Orm proteins activates stress responses pathways and alters inositol metabolism. Specifically, Han et al. found that the over-production of sphingolipids in yeast lacking Orm proteins leads to induction of the unfolded protein response, further highlighting the cellular stress that results from misregulation of sphingolipid metabolism.

A key feature of Orm1/2’s homeostatic function is their regulation by phosphorylation in response to changing sphingolipid levels (Figure 2B). Specifically, phosphorylation of N-terminal residues relieves the inhibitory activity of Orm proteins when sphingolipid production is disrupted (for example, by treatment with the serine palmitoyltransferase inhibitor myriocin). This inactivation of Orm1/2 provides a compensatory feedback mechanism that stabilizes cellular sphingolipid levels in the face of external perturbations. Mutation of the phosphorylation sites on Orm1/2 renders cells unable to survive disruptions to sphingolipid synthesis, as the inhibitory effect of Orm proteins cannot be released. Conversely, constitutive loss of ORM gene function causes toxic over-production of sphingolipids. Although the precise mechanism by which Orm1/2 inhibit serine palmitoyltransferase requires further investigation, Orm proteins (and potentially their binding partners) appear to be assembled into a higher-order multimeric complex whose stoichiometry is controlled by the same phosphorylation events.

Thus, Orm proteins are key mediators of sphingolipid homeostasis in yeast. In human cells, ORMDL genes also regulate sphingolipid levels, and human serine palmitoyltransferase binds to Ormdl proteins and forms a high-molecular-weight complex (Breslow et al., 2010; Hornemann et al., 2007). However, it is not yet known to what extent feedback regulation of Orm protein activity is conserved. The importance of further investigation in this area is further highlighted by the recent identification of a major genetic risk factor for childhood asthma near the human ORMDL3 gene (Moffatt et al., 2007). Although more work is required to understand the molecular nature of this risk factor, polymorphisms at this locus have also been linked to a number of other inflammatory diseases including type I diabetes (Barrett et al., 2009), Crohn’s disease (Barrett et al., 2008), ulcerative colitis (McGovern et al., 2010), and primary biliary cirrhosis (Hirschfield et al., 2010; Liu et al., 2010b). Furthermore, sphingolipids have a number of well-established roles in the lung and in modulating inflammatory processes and immune system function (Rivera et al., 2008; Ryan and Spiegel, 2008; Uhlig and Gulbins, 2008). Together, these observations raise the testable hypothesis that sphingolipid misregulation contributes to asthma development.

Collectively, significant progress has been made recently toward understanding the mechanisms by which cells ensure sphingolipid homeostasis. The roles of TORC2, SMSr, CERT, and Orm proteins are providing our first molecular insights into how sphingolipid supply is matched to cellular demand and suggest that exciting new developments will be coming soon.

Links between sphingolipids and other membrane-associated processes

Membrane proteins, sphingolipids, glycerolipids, and sterols are all synthesized within the secretory pathway, and their interactions collectively determine the properties of cellular membranes. Given this physical and functional relationship, it is likely that their biogenesis must be closely coordinated. Co-regulation may be especially important because the relative amounts of different lipid classes have profound effects on the physical characteristics of lipid bilayers. Such regulatory connections would help, for example, to ensure that as rates of cell growth increase or decrease, the synthesis of all lipid types is coordinately adjusted and membrane imbalances are avoided. In recent years, there has been a growing appreciation for the links between sphingolipid metabolism and other secretory pathway processes, and below we describe some of these emerging insights (Figure 3).

Figure 3. An integrated view of regulatory pathways in sphingolipid metabolism.

Several interconnected pathways link sphingolipid metabolism to glycerolipid and sterol production. Shown here are select regulatory mechanisms for mammalian lipid biosynthesis; see text for details. Dotted lines indicate functional relationships that are not yet fully characterized or that may be indirect. Metabolites and pathways for their inter-conversion are shown in black. Proteins and their regulatory activities are shown in blue. Abbreviations are as follows: Ormdl1/2/3 (ORM1-like [S. cerevisiae] 1/2/3), Sptlc1/2/3 (Serine palmitoyltransferase long-chain base subunit 1/2/3), Sacm1l (Suppressor of actin mutations 1-like), CERT (Ceramide transport protein), OSBP (Oxysterol-binding protein), SREBP (Sterol regulatory element binding protein).

Coordination between sphingolipids and sterols

Perhaps the most extensive and best-understood examples of co-regulation are seen between sphingolipids and sterols (Gulati et al., 2010). Members of these two lipid classes associate physically and have a similar and cooperative activity in altering membrane order and nucleating lipid microdomains (Lingwood and Simons, 2010). In mammalian cells, their extracellular trafficking occurs in the same lipoprotein particles (Nilsson and Duan, 2006), and mutations to enzymes needed for the turnover of either sterols or sphingolipids cause accumulation of both lipid types (Bektas et al., 2010; Ginzburg et al., 2004). Similarly, in yeast, mutations to enzymes of sterol metabolism often cause changes specifically in sphingolipid levels and vice versa (Guan et al., 2009). Functional interdependence is also suggested by the many genetic interactions seen between yeast sphingolipid and sterol genes (Guan et al., 2009).

How are sterol and sphingolipid metabolism coordinated at a mechanistic level? One likely explanation is that, due to their physical association, changes in sphingolipid levels affect the amount and intracellular location of free sterols. These sterols are in turn sensed by a number of well-characterized adaptive pathways that exert feedback control on sterol metabolism (see reviews by Brown and Goldstein (2009) and Chang et al. (2006)). For example, elegant studies have shown that low ER cholesterol levels cause a compensatory increase in activity of SREBP2, a major transcriptional regulator of cholesterol uptake and biosynthesis (Figure 3), and a decrease in storage of cholesterol as ester derivatives (Brown and Goldstein, 2009; Nohturfft and Zhang, 2009). The opposite responses occur when cholesterol levels are high. Artificial decreases in sphingolipid levels have been shown to cause redistribution of plasma membrane cholesterol to intracellular membranes including the ER (Lange and Steck, 1997; Slotte and Bierman, 1988). Consequently, SREBP activity decreases, cholesterol synthesis is inhibited, and cholesterol ester formation is induced (Figure 3) (Lange and Steck, 1997; Scheek et al., 1997; Slotte and Bierman, 1988). Conversely, treatment of cells with exogenous sphingolipids appears to siphon cholesterol to the plasma membrane; thus, ER cholesterol levels fall, and feedback pathways are activated to restore ER sterol concentrations (Gatt and Bierman, 1980; Puri et al., 2003). In this way, perturbations to sphingolipid levels are coupled indirectly to changes in sterol metabolism.

Similar to the examples described above, changes in sterol levels have also been shown to impact sphingolipid metabolism. These effects are unlikely to be mediated by the canonical SREBP pathway, as enzymes of sphingolipid metabolism are not known to be targets of this transcriptional regulator (Nohturfft and Zhang, 2009). Although our mechanistic understanding is currently incomplete, one component of the metabolic link between these lipid groups appears to be due to the ability of sterols to modulate sphingolipid transport. For example, depletion of plasma membrane cholesterol by cyclodextrin treatment causes de-phosphorylation and activation of the ceramide transporter CERT in the same way that is seen in response to myriocin or upon exposure of cells to SMase (Kumagai et al., 2007). Recent work has identified the oxysterol-binding protein OSBP as a potential mediator of this sterol-dependent regulation of CERT (Figure 3). OSBP has a function analogous to that of CERT, but instead mediates inter-membrane transfer of sterols. This functional similarity is also reflected in OSBP’s domain architecture: like CERT, it contains a PI4P-binding PH domain and a VAP-binding FFAT motif (Raychaudhuri and Prinz, 2010). OSBP’s binding to sterols is mediated by an ORD domain, and this domain is required for in vitro transfer activity. It should be noted that OSBP may also be a sterol sensor, as ligands such as 25-hydroxycholesterol (25-HC) regulate its oligomerization and localization, and these properties of OSBP in turn control ERK signaling (Raychaudhuri and Prinz, 2010).

Ridgway and colleagues (2006) have characterized how sterols regulate sphingolipid metabolism by showing that 25-HC causes a significant increase in SM synthesis that is dependent on OSBP, CERT, and their shared binding partner VAP (Figure 3). The precise mechanism is not yet known, but OSBP appears to activate CERT by promoting its recruitment to membranes and its binding to VAP. Interestingly, not only does OSBP link sterols to CERT activity, but OSBP is itself activated by SM depletion, providing further evidence for coordinated regulation of these lipid types (Ridgway et al., 1998). Finally, it is noteworthy that 25-HC and cyclodextrin treatment not only stimulate sphingolipid transport via CERT but also induce de novo LCB synthesis (Leppimaki et al., 1998; Ridgway, 1995). Although the mechanistic basis for these latter effects remains unclear, these observations indicate that sterols may regulate multiple steps in sphingolipid metabolism.

From a broader perspective, despite the mounting evidence for cross-regulation of sphingolipids and sterols, key questions concerning the physiological significance of these phenomena remain largely unanswered. Do the observed forms of regulation represent adaptive coordination or do they only reflect indirect effects of sphingolipids on intracellular sterols and vice versa? The net effect of SREBP-mediated responses to sphingolipid perturbations appears to be to stabilize the ratio of sterols to sphingolipids in a narrow range, but at the expense of fluctuations in the absolute levels of these lipids that potently affect membrane order. Other examples described above appear to make the opposite trade-off. How do these forms of cross-regulation affect membrane function and what do they indicate about the lipid concentrations and membrane properties that cells sense and regulate? Future investigation of these questions will be critical for a complete understanding of membrane homeostasis.

Coordination between sphingolipids and glycerolipids

As has been seen for sphingolipids and sterols, there is also growing recognition of functional links between sphingolipids and glycerolipids. One notable difference however is that sphingolipids and glycerolipids share several direct metabolic connections. For example, in mammalian cells, the glycerolipid phosphatidylcholine (PC) must be converted to diacylglycerol (DAG) in order to generate SM from ceramide. Similarly, phosphatidylinositol (PI) provides the phosphoinositol source for inositolphosphorylceramide (IPC) generation in yeast. As a result, perturbations that alter PC and PI levels may also impact sphingolipid metabolism (Brice et al., 2009). Conversely, sphingolipid metabolites can also be used to synthesize glycerolipids such as PE via the action of the LCB-P lyase. This enzyme cleaves LCB-Ps to acyl aldehydes and ethanolamine phosphate, which can be recycled for use in PE synthesis. The importance of this pathway and the direct manner in which sphingolipids can regulate glycerolipid metabolism are highlighted by work on the SREBP pathway in insects. Specifically, the Drosophila SREBP pathway, which senses and regulates PE rather than cholesterol, is highly sensitive to regulation by LCBs and ceramides because the LCB-P lyase enables these sphingolipid metabolites to be used for PE production (Dobrosotskaya et al., 2002). The ability of sphingolipid metabolites to support glycerolipid synthesis and the shared dependence of both pathways on fatty acid precursors raises the important question of how cells regulate the distribution of lipid precursors between sphingolipid and glycerolipid synthesis.

In addition to these direct metabolic connections, there is also evidence for regulatory mechanisms that coordinate production of glycerolipids and sphingolipids. For example, Roelants et al. (2010) recently found that the yeast kinases Ypk1, Ypk2, Fpk1, and Fpk2 form a regulatory network that controls bilayer asymmetry of phospholipids in a sphingolipid-dependent manner. Specifically, Ypk and Fpk kinases regulate each other, and Fpk1 in turn activates lipid flippases that return aminophospholipids (i.e. PE and phosphatidylserine) to the inner leaflet of the plasma membrane. Inhibition of sphingolipid synthesis causes changes in Fpk1 and Ypk1 activity that result in down-regulation of phospholipid flippase activity (Roelants et al., 2010). These findings therefore indicate that sphingolipid levels influence glycerolipid bilayer asymmetry and that these two determinants of membrane function may be coordinately regulated. A second example of cross-regulation between sphingolipids and glycerolipids is seen for the Drosophila ceramide synthase enzyme schlank. Bauer et al. (2009) recently found that schlank is not only responsible for ceramide synthesis, but also regulates organismal body fat and glycerolipid levels. These latter effects reflect schlank’s apparent ability to activate the SREBP transcriptional program, possibly via its N-terminal homeobox domain.

Lastly, several recent studies have linked sphingolipids to phosphoinositide glycerolipids, and by extension, to a range of phosphoinositide-regulated processes that occur within the secretory pathway. Some of the first links between sphingolipids and phosphoinositides were again found in yeast, with the observation that MSS4, which encodes a PI4P kinase, is a TSC gene (Beeler et al., 1998). It was subsequently found that STT4, which encodes the PI kinase that generates PI4P, also shows a TSC phenotype and exhibits other complex genetic interactions with genes involved in sphingolipid metabolism (Tabuchi et al., 2006). A direct role for phosphoinositides in yeast metabolism is suggested by our recent finding that the SPOTS complex contains the PI4P phosphatase Sac1 in addition to Orm proteins and serine palmitoyltransferase (Breslow et al., 2010). Loss of SAC1 leads to elevated LCB levels and exhibits synthetic lethality with deletion of ORM genes, suggesting a direct role for Sac1 as a negative regulator of sphingolipid production (Figure 3). This model is consistent with the finding that cells lacking SAC1 are strongly resistant to the serine palmitoyltransferase inhibitor myriocin (Breslow et al., 2010). The mechanism by which Sac1 activity affects sphingolipid synthesis remains unknown, although a direct role for PI4P or PI in regulating serine palmitoyltransferase is an exciting possibility. It is also unknown whether Sacm1l, the mammalian homolog of Sac1, shares this function in sphingolipid regulation.

One case where the effect of PI4P is better understood is in mammalian sphingolipid transport. As mentioned above, the transport of both ceramide by CERT and glucosylceramide by FAPP2 is dependent on PI4P-binding PH domains found within these proteins (Figure 2A and Figure 3). These PH domains target CERT and FAPP2 to the PI4P-enriched Golgi membrane (D’Angelo et al., 2007; Hanada et al., 2003) and may have a role in determining the donor and acceptor membranes for lipid transfer. For example CERT’s ER-binding FFAT motif may specify loading of ceramide from the ER into the START domain while the PH domain instructs delivery to the Golgi (Figure 2A). It will be interesting to determine whether Sac1’s association with the sphingolipid biosynthetic machinery in the SPOTS complex also has relevance to sphingolipid transport.

Phosphoinositides not only have a role in sphingolipid biosynthesis but also participate in protein synthesis and transport in the secretory pathway. The enrichment of PI(4,5)P₂ at the plasma membrane, PI4P at the Golgi, and PI in the ER serve as identifying markers for these organelles that are critical for vesicular transport of proteins (Behnia and Munro, 2005). Thus, dynamic regulation of phosphoinositides may simultaneously regulate both protein delivery to the plasma membrane and sphingolipid production. In this regard, Sac1 was recently found in both yeast and mammalian cells to be regulated in response to nutrient availability (Figure 3). Specifically, glucose starvation in yeast or growth factor deprivation in mammalian cells causes re-localization of Sac1 to the Golgi, which reduces Golgi PI4P levels and slows protein trafficking (Blagoveshchenskaya and Mayinger, 2009). It is tempting to speculate that this mechanism to adjust protein secretion according to nutrient levels may also impact sphingolipid transport by CERT and FAPP2 (by altering PI4P levels) and sphingolipid synthesis by the SPOTS complex (by changing complex localization and/or composition) (Figure 3). The SPOTS complex may therefore act as a dynamic coordinating center that senses sphingolipid and nutrient levels to coordinate lipid and protein biosynthesis.

Perspective and future directions

As described above, we are now beginning to understand the molecular mechanisms by which sphingolipid production is regulated according to metabolic need and in conjunction with other basic cellular processes. Despite these advances, there are many areas in which our understanding of sphingolipid homeostasis remains rudimentary. We discuss below these directions for future research and emerging technologies that will enable these questions to be addressed.

One immediate goal is to define the full spectrum of mechanisms by which sphingolipid metabolism is regulated. Indeed it is quite likely that our current inventory of regulators is highly incomplete, particularly given the large number of sphingolipid metabolites that have potent biological activities that must be tightly controlled. The need for multiple feedback controls is further underscored by the existence of distinct sub-cellular pools of sphingolipids that may each need to be precisely regulated.

A specific area in which our understanding remains poor is transcriptional control of sphingolipid metabolism. Given that master transcriptional regulators have been identified for glycerolipids and sterols in both yeast and mammals (Nohturfft and Zhang, 2009), it is surprising that this aspect of sphingolipid biosynthesis is not yet understood. Although it is possible that sphingolipid regulation is primarily post-transcriptional, there is preliminary evidence for coordinated transcription of sterol and sphingolipid genes in yeast. For example, the transcription factors Ecm22 and Upc2 that control ergosterol production also have consensus binding sequences in the promoters of the serine palmitoyltransferase genes LCB1 and LCB2 (Vik and Rine, 2001). Similarly, in the course of re-examining published yeast microarray studies (Ihmels et al., 2002), we have found evidence for co-regulation of ergosterol and sphingolipid biosynthesis across a range of experimental conditions (our unpublished observations). It is also interesting to note that RPD3 and SIN3, which encode the yeast homologs of the S1P-regulated histone deacetylases characterized by Hait et al. (2006), were previously identified as TSC genes (Beeler et al., 1998) and are known to participate in transcriptional control of glycerolipid metabolism (Wagner et al., 2001). Whether LCB-Ps also regulate Rpd3/Sin3 in yeast and how this deacetylase controls sphingolipid metabolism will be exciting questions for future investigation.

Beyond obtaining a comprehensive list of sphingolipid regulators, a second major area for future research is understanding how cells modulate the activity of these regulators in response to sphingolipid levels. Particularly, what is the physical basis for sensing sphingolipids and what cellular machinery performs this function? Given that sphingolipid metabolites are membrane-embedded and vary widely in their physical properties and cellular concentrations, sensing sphingolipids is hardly a trivial task. Furthermore, sensors likely need to distinguish subtle chemical features because small differences in acyl chain length or unsaturation can be functionally important (as in Alvarez et al., 2010 and Menuz et al., 2009).

Broadly speaking, two potential mechanisms for detecting sphingolipids can be imagined. In the first case, sphingolipid sensors may bind specifically to lipid metabolites in a fashion similar to canonical ligand-receptor interactions. Alternatively, sphingolipid sensors could instead detect sphingolipid-dependent changes in membrane properties. To date, evidence for both mechanisms exists. For example, specific binding of sphingolipid metabolites is carried out by S1P-specific GPCRs (Rosen et al., 2009), the ubiquitin ligase TRAF2 (Alvarez et al., 2010), and the histone deacetylases HDAC1 and HDAC2 (Hait et al., 2009), and crystallographic studies are now providing structural insights into such mechanisms of protein-sphingolipid recognition (Kudo et al., 2008; Malinina et al., 2006). There is also precedent for sensors that detect the physical properties of membranes. For example, sphingolipids impact membrane thickness (Lingwood and Simons, 2010), and a mechanism for responding to changes in membrane thickness has recently been described in bacteria (Cybulski et al., 2010). Similarly, dynamic association of a sensor protein with sphingolipid-dependent membrane microdomains could be used to monitor sphingolipid levels. Intriguingly, the yeast trans-membrane protein Nce102 was recently shown to utilize this type of membrane domain-dependent regulation to control Pkh kinase activity (Frohlich et al., 2009). Given that sphingolipids interact with functional effectors by a variety of mechanisms, it seems likely that multiple modes of molecular recognition are utilized to ensure sphingolipid homeostasis.

Efforts to identify and characterize sensors and regulators of sphingolipid metabolism will likely depend on a number of newly emerging technologies. Indeed, the substantial progress made in the past several years has been enabled by new methodologies such as lipidome analysis by mass spectrometry (Shevchenko and Simons, 2010), identification of sphingolipid-binding proteins with affinity resins (Hait et al., 2009), and the use of functional genomic approaches to characterize new regulators of sphingolipid production (Aguilar et al., 2010; Breslow et al., 2010; Denic and Weissman, 2007). Moving forward, technologies that allow researchers to modulate and monitor sphingolipid metabolism with high temporal and spatial resolution would be of particularly value. In this regard, in vivo reporters based on fusions of fluorescent proteins to lipid-binding domains have been key tools to study phosphoinositides and are now being developed for sphingolipids (Bakrac et al., 2010). Similarly, previous work on phosphoinositides suggests that inducible targeting of sphingolipid-generating and sphingolipid-degrading enzymes to specific sub-cellular locations will be a powerful method to reveal local functions of sphingolipid metabolites (Birbes et al., 2001).

Enabled by new technologies, future studies will be poised to meet the key challenge of understanding how different regulators of sphingolipid metabolism act together to ensure sphingolipid homeostasis. For example, it is unlikely that Orm proteins, SMSr, and CERT regulate serine palmitoyltransferase, ceramide levels, and ceramide transport by three isolated, independent mechanisms. Rather, multiple sphingolipid regulatory pathways likely act in concert, and characterizing the nature of these connections will be critical to understanding the systems-level properties of sphingolipid metabolism. More broadly, a comprehensive and integrated view of sphingolipid homeostasis promises to provide fundamental insights into lipid signaling and membrane function in cell biology and disease.