Compartmentalization of Sphingolipid Metabolism: Implications for signaling and therapy

Abstract

Sphingolipids (SLs) are a family of bioactive lipids implicated in a variety of cellular processes, and whose levels are controlled by an interlinked network of enzymes. While the spatial distribution of SL metabolism throughout the cell has been understood for some time, the implications of this for SL signaling and biological outcomes have only recently begun to be fully explored. In this review, we outline the compartmentalization of SL metabolism and describe advances in tools for investigating and probing compartment-specific SL functions. We also briefly discuss the implications of SL compartmentalization for cell signaling and therapeutic approaches to targeting the SL network.

INTRODUCTION

Since the discovery of their signaling functions in the 1980s, bioactive sphingolipids (SL) such as ceramide (Cer), sphingosine (Sph), and Sph 1-phosphate (S1P) have been implicated in many biologies, including proliferation, migration, and cell death. Dysregulation of SL levels has been reported in cancer, diabetes and neurodegenerative diseases, among others [1, 2]. Mutations in SL enzymes, most commonly linked with lysosomal storage disorders, have also begun to emerge in association with other pathologies [3, 4]. Consequently, there is a growing interest in the possibility of targeting the SL network as a therapeutic approach.

Although often referred to as singular lipids, the advent of mass spectrometry analysis has revealed that many bioactive SLs comprise a family of closely related molecular species. For example, Cer can have variations in length and desaturation of its acyl chain, the length of the sphingoid backbone, and the presence of hydroxyl groups on the sphingoid base or the 1-position, which, when combined, produces over 200 molecular species in the Cer family alone. Such broad heterogeneity ultimately challenged the idea that Cer and other SL had singular functions and led to the ‘Many Ceramides’ paradigm, described by Hannun and Obeid in 2011 [5]. This proposed that “individual ceramide molecular species are likely regulated by specific biochemical pathways….and execute distinct functions.” The increased attention on molecular species spurred by this paradigm has begun to show promise with growing evidence of different biophysical properties of distinct Cer species [6, 7] and identification of proteins that interact with one specific SL moiety [8, 9]. However, the “Many Ceramides” also contains a key component that is only recently coming to the fore: the compartmentalization of SLs.

As hydrophobic molecules, many SLs often remain within the membranes and compartments they are produced. This has important implications both for metabolism and signaling as they are influenced by the enzymes and proteins that are present in their local environment. From here, it is easy to see how the same lipid – indeed, even the same molecular species of lipid – could have distinct signaling effects if produced in the plasma membrane vs. the endoplasmic reticulum (ER) or mitochondria, as an example. This clearly has implications for cell biology but also for therapeutics and may demand a more surgical targeting of SL metabolism for the desired outcome to be achieved. In this review, we describe the compartmentalization of SL metabolism and discuss the growing number of tools available to probe SL function in cellular compartments. We will also outline some compartment-specific functions of SL signaling and will briefly discuss therapeutic implications of compartmentalized SL metabolism.

COMPARTMENTALIZATION OF SL METABOLISM

Cellular SL levels are tightly controlled by an interlinked metabolic network of enzymes that is distributed throughout the cell in multiple cellular compartments. Here, we will briefly describe the SL network, with particular emphasis on the cellular locations of major arms of the pathways. For a more detailed description of the pathways and enzymes within the network, the review by Hannun & Obeid is recommended [1]. The compartmentalization of SL metabolism is shown in Fig. 1

Figure 1. Compartmentalization of SL metabolism and signaling.

SL synthesis starts in the ER to form Cer, and is transported via vesicular or by active protein transport to the Golgi apparatus. Cer in the Golgi can be metabolized to the diverse branches of complex SLs (for simplicity, reduced to SM, HxCer, and acylceramide, AC). These are distributed to different subcellular membranes where they undergo further metabolism. Several biological pathways have been related to SL species in specific membranes. For example, Cer in the PM can trigger the extrinsic apoptotic signal by clustering cell death receptors in Cer-rich Rafts. PM Cer can also function as a second messenger by directly activating protein phosphatases (here, PP1 regulating cytoskeletal dynamics). S1PR, also localized in PM, co-regulates the cytoskeleton during cell polarization. SL in different membranes can also independently regulate different pathways, such as mitochondrial-Cer regulating the intrinsic apoptotic pathway.

ER:

De novo SL synthesis typically begins with the condensation of palmitate and serine by serine palmitoyltransferase (SPT) to produce 3-ketodihydro-Sph. This is subsequently metabolized to dihydroSph (dhSph), dihydroceramide (dhCer) and finally Cer which forms the central hub of the network. The de novo SL pathway and the enzymes that mediate them have been primarily localized to the ER [10–12], although they may also occur in ER-associated membrane e.g. mitochondria-associated membranes (MAMs). From here, de novo generated Cer can be trafficked to the Golgi by Cer transfer protein (CERT) or vesicular transport for further metabolism (see below) [13, 14]. A recently described metabolic pathway found that de novo Cer can be converted to acylceramide by DGATs and shunted into lipid droplets [15]. Evidence also suggests ER-generated Cer can transfer to the mitochondria or MAMs where it could be subject to further metabolism [16, 17]. Finally, the protein SAMD8, also known as sphingomyelin synthase-related protein (SMSr), can convert Cer to Cer-phosphoethanolamine and was proposed to act as a sensor of Cer levels in the ER [18].

Golgi:

In recent years, the Golgi apparatus has gained appreciation as a key switch track in the SL network that controls the flow of Cer into various metabolic routes. It is a major site of synthesis of complex SLs, including sphingomyelin (SM), glucosylceramide (GluCer), and lactosylceramide (LacCer), as well as their metabolism to gangliosides and cerebrosides [19–22]. Conversion of Cer to Cer-1-phosphate by Cer kinase (CERK) [23] and to Sph by Alkaline Ceramidase 2 (ACER2) [24] or neutral ceramidase (CDase) can also occur in the Golgi [25]. From the catabolic side, neutral sphingomyelinase-2 (nSMase2) has been found in the Golgi [26] and may revert Golgi-resident SM back to Cer. Of note, the two known pathways that bring Cer from the ER were found to connect to different metabolic fates. Thus, CERT delivers Cer that is specifically used for SM and C1P synthesis [23, 27, 28] while vesicular trafficking of Cer is coupled to GluCer synthesis [29]. For the latter pathway, the generated GluCer can then be delivered for ganglioside synthesis by the transport protein FAPP2 [30]. Subsequent to the Golgi, many of the metabolic products of Cer are transported to different parts of the cell through vesicular trafficking. C1P can also be transported to other membranes through the recently discovered and characterized transfer protein CPTP [31, 32].

Plasma Membrane (PM):

The regulation of local SL metabolism at the PM following activation of receptors generates various bioactive SL messengers. The stimulation of SM hydrolysis to Cer through activation of sphingomyelinases (SMases) has been reported for stimuli such as cytokines, chemotherapies, radiation, and mitogenic signals [33–39]. This can be mediated through translocation of acid SMase [35], through the secreted form of acid SMase [40], or through neutral SMase [33, 34, 38, 39]. Generated Cer can be further metabolized to Sph by neutral CDase which can, in turn, be converted to S1P [38, 39]. This is typically mediated by Sph kinase 1 (SK1), which translocates to the PM in response to pro-growth and pro-inflammatory stimuli [41, 42]. The generated S1P at the plasma membrane may act as an intracellular messenger [43], be exported by lipid transporters such as SPSN2 [44] or may interact with S1P receptors to activate various signaling pathways [45–47]. Coordination in the regulation of Cer, and S1P signaling at the plasma membrane is suggested to regulate the dynamic cycle of phosphorylation and dephosphorylation of several cytoskeletal proteins, including the ERM family, allowing adhesion and detachment during cell migration [48]. Finally, the sphingomyelin synthase SMS2 is also found at the PM where it could potentially terminate Cer-dependent signaling by reconverting it back to SM [49].

Lysosomes:

As with other macromolecules, complex SLs undergo catabolism in the late endosomes/lysosomes with acid SMase breaking down SM and glucosidases breaking down GluCer [50, 51]. The product of both reactions is Cer, which can be further metabolized to Sph by acid CDase [50]. Prior studies have suggested that Sph may escape the lysosome and be used for resynthesis of Cer a process termed the ‘salvage pathway’ [51, 52] although other studies have suggested that Sph may also accumulate in the lysosomes [53]. Of note, mutations in lysosomal SL enzymes and disruption of lysosomal SL metabolism has been linked to many lysosomal storage diseases including Niemann-Pick, Farber, Fabry, Tay-Sachs, and Sandhoff disease among others [3, 4]. This underscores the importance of lysosomal SL metabolism within the cell.

Nucleus:

In recent years, there has been a growing appreciation for a distinct SL metabolic network within the nucleus. SM is the most abundant SL in the nucleus where it has been reported to play both functional and structural roles [54, 55]. However, a number of other SLs have also been identified there including Cer, Sph, and S1P. The presence of multiple lipids in the nucleus implies that there is metabolic turnover and consistent with this, activity of many SL enzymes has been detected in nuclear extracts [54, 55]. Furthermore, specific enzymes have been found within the nucleus including Sph kinase 2 (SK2) and neutral sphingomyelinase-2 [56–59]. In some cases, studies have reported stimulated trafficking of enzymes both into and out of the nucleus where they have locally affected SL levels. This infers that there are functional signaling roles for nuclear SL. It should be noted that many studies into SLs in the nucleus have not determined the relevant nuclear compartments where the enzymes and/or lipids are localized. This is particularly important to keep in mind given that the nuclear envelope is an extension of the ER.

Mitochondria:

The connections of Cer and other SLs to cell death led to considerable interest in SL metabolism in mitochondria. Although ER-generated Cer was reported to transfer to the mitochondria or MAMs [16, 17], localization of SL metabolizing enzymes within the mitochondria has also been reported. From the anabolic perspective, Cer synthase (CerS) activity was localized to the outer and inner mitochondrial membranes from rat liver mitochondria [60]. This was further supported by studies localizing many CerS isoforms (1, 2, and 6) in rat brain mitochondria with CerS6 being found in the outer mitochondrial membrane and CerS2 in the inner mitochondria membrane [61]. The association of the yeast SMase ISC1 with mitochondria [62] led to speculation of a mammalian mitochondrial SMase, which was first identified in zebrafish and subsequently isolated in mouse tissues [63]. However, it should be noted that a human homolog has yet to be found. Moreover, the enzyme appeared to have dual localization of mitochondria and ER-associated membranes dependent on the cell type [64]. From the catabolic perspective, neutral CDase activity was described in rat liver mitochondria and consistent with this, a cloned neutral CDase localized to mitochondria [65, 66]. Of note, subsequent studies suggested that a reverse neutral CDase activity could generate Cer in the mitochondria of liver, providing an alternate route of Cer synthesis [67]. Finally, downstream of ceramidase, the SK2 isoform (but not SK1) was localized to mitochondria of cardiomyocytes, most probably in the inner mitochondrial membrane [68]. This regulated production of mitochondrial S1P, although it was noted by the authors that S1P in their system also came from extramitochondrial sources.

TOOLS AND TECHNIQUES FOR PROBING SL COMPARTMENTALIZATION

In the early days of SL research, cellular lipid levels were often probed with radioactive labelling, classical Bligh-Dyer lipid extractions, and thin layer chromatography (TLC) [69]. In such studies, quantification was done by scraping of plates and scintillation counting. Since then, almost three decades later, techniques for measuring cellular SLs have evolved enormously and now allow us to analyze both the membrane composition and function of SLs at the subcellular compartment level. Here, we will discuss several techniques that have been used to identify specific pools of bioactive lipids and assess their functions, or those techniques that are building towards that goal. A summary of the tools and techniques described can be found in Table 1.

Table 1:

Summary of Tools and techniques for probing SL metabolism

Technique	Compartment	Comments	Refs
Cell lysates	Whole cells	Cells are pelleted or directly collected from the culture support with organic solvents. Combined with TLC or HPLC/MS to measure whole cell SL profile. Topological information is lost.	[124, 125]
Cell fractionation	Enrichment in large organelles	Differential centrifugation or selective detergent organelle purification. Expect cross-contamination between organelles.	[67, 68, 70]
Gradient centrifugation	Purer organelles, and lipidrafts	This method gets pure membranes. However, it requires cell disruption and long protocols that could affect lipid composition.	[51, 72, 73]
Lipid binding proteins (including antibodies)	Outer leaflet plasma membrane, diffuse intracellular staining	Exceptional for imaging. Some pairs are very well established (lysenin/SM; Cholera toxin/GM1; Cer / Ab). Specificity and sensitivity are an issue for some probes.	[77, 79, 81, 87, 88, 91, 126]
Fluorescent lipid	Golgi, diffused membranes	Useful to study metabolism when combined with HPLC. Imaging studies are not that clear. Tend to accumulate as SM in Golgi.	[93, 127]
Enzymatic assays in situ	Plasma membrane	Allow acute measurement of several lipids. Limited to plasma membrane.	[117, 120]
Targeted-enzymes	Organelle membrane	Combine with HPLC allows measuring lipid composition and lipid changes in different organelles in situ. However, this is not an acute system and the cells will adapt to lipid changes.	[121, 122]
Clickable lipids	Whole cell	Have shown potential by being metabolized as endogenous lipids and being clicked with fluorophores. Different species cannot be differentiated yet. Combined with PLA and Cer antibody have been used to give Cer specificity to the assay.	[99, 102, 104]
Crosslinkable lipids	Whole cell	Powerful tool to study lipid-protein interaction in vivo. Compounds are metabolized and different species cannot be differentiated.	[101, 102]
Caged lipids	Whole cell, Organelle specific	Allows delivery in different organelles, and keeps SLs inactive until released. Can be combined with clickable and crosslinkable moieties.	[103, 108]
FRET probes	Endosomes, Golgi	They allow to measure and potentially visualize enzymatic activity in vivo.	[109]

Determining the subcellular SL localization

Cell fractionation

The most common technique for detecting and measuring compartmentalized SLs is biochemical fractionation followed by analysis of the lipid content by different techniques, including radioactivity counts, chromatography, or mass spectrometry (MS). Of the various protocols, differential centrifugation followed by liquid chromatography-MS/MS has been the most used and studies utilizing this have revealed asymmetric SL distribution. For example, comparative studies in SL composition between mitochondria and ER from the brain [70], muscle [71], liver [72], and heart tissue showed that not only do ER and mitochondria have different SL composition, but also that the effects of stimulation was organelle specific. Thus, ischemia-reperfusion increased mitochondrial Cer in brain tissue, while insulin deprivation increased ER Cer in muscle cells. Cell fractionation followed by gradient centrifugation has also been used to identify bioactive Cer in the PM [73] and in mitochondria [51]. These and other studies started to point out that SLs responding to physiological inputs, mediating signaling, and biological events were compartment-specific, where SLs in other organelles were unaffected.

Lipid rafts and Cer rich platforms:

Although lipid rafts remain controversial, there is no doubt that the idea of raft-regulated signaling is compelling and imply a sub-compartment of the membranes that they exist in. While there has been debate on the precise definition of rafts, they have typically been defined as rich in SLs and cholesterol and the tools to study them have not changed since the 90s [74]. Lipid rafts are primarily isolated by their resistance to cold detergent, usually Triton X100, followed by flotation in sucrose gradients. Since Cer is considered a bioactive molecule, hydrolysis of raft-associated SM might not disintegrate rafts, but may instead generate a new signaling platform. Indeed, consistent with this hypothesis, Cer-enriched platforms have been reported and purified based on its insolubility in cold Triton X100. Studies into the description and signaling mechanism of such platforms have also been based on observation of Cer using anti-Cer-specific antibodies (see below).

Microscopy: visualizing SLs in situ

Although cell fractionation remains a powerful tool for subcellular analysis of lipid composition, it is a long and laborious protocol, particularly if combined with gradient centrifugation. In contrast, fluorescent and electron microscopy allow the visualization of probes in different membranes without disrupting the cell. In microscopy, the resolution is defined by the Abbe diffraction limit [75], which depends on the wavelength of light and the numerical aperture of the objective, with a representative resolution of 250 nm. While electron microscopy offers a significant improvement in resolution (0.2 nm), it lacks the flexibility of fluorescent microscopy on using different labels in the same sample, the capability to use life imaging and having simple protocols from treatment to visualization. However, the development of different fluorescent microscopy techniques is helping to improve the detection limit and bring it closer to the electron microscopy. These are structured illumination microscopy (SIM, typical resolution 100nm), stimulated emission depletion microscopy (STED 30–80nm), photoactivated localization microscopy (PALM, 10–50nm), and stochastic optical reconstruction microscopy (STORM, 20–80nm) [76]. Nonetheless, the challenge with fluorescent microscopy is how to visualize endogenous SLs and their dynamics. Here we discuss some of the available probes for SLs:

SL-binding proteins:

Lysenin is a 41KDa protein from the earthworm Eisenia foetida that was reported to cause lysis of erythrocytes [77], CHO cells, and other cell lines. The observation that pre-treatment with a bacterial sphingomyelinase (bSMase) made cells resistant to lysenin-induced lysis [78, 79] led to the discovery that lysenin binds SM, preferentially in membranes enriched with cholesterol [80], and forms pores of 3nm diameter [79]. From this, lysenin gained utility as a probe to stain cellular SM and was used to show polarization of SM in basolateral and apical membranes across many cell lines, as well as staining membrane microdomains suggestive of rafts [80]. A mutant form of lysenin that cannot form pores was subsequently developed and became a powerful to study SM topology, including its role in lipid rafts [81, 82]. Analysis of PM vs digitonin -permeabilized cells was used to visually demonstrate that PM sphingomyelin is regulated by SMS2, whereas Golgi SM is regulated by SMS1 [83]. Equinatoxin-II is a protein from the sea anemone Actinia equina and is another SM binding protein. Intriguingly, in liposome models, equinatoxin-II has been shown to bind dispersed SM whereas lysenin preferentially binds clustered SM. How this translates in cell staining is unclear, but differential and overlapping staining in both PM and intracellular membranes has been found with both toxins [83]. They have also helped to distinguish between SM in apical (dispersed) and basolateral (clustered) membranes in MDCK II cells [84]. It is also worth mentioning that other toxins (such as aegerolysins, and ostreolysin A) have been reported to specifically bind membranes enriched in both SM and cholesterol, and may have additional utility as probes. Many different microscopy techniques have been used to visualize SM in cells, such as NanoSIMS [85], dSTORM, AFM. Combining SM-binding toxins and electron microscopy has revealed images of clustered SM forming domains in cellular membranes [84]. Kobayashi’s lab took this further using freeze-fracture replica-labelling immunoelectron microscopy to separate outer and inner monolayers. Using lysenin, they described clustered SM both in the outer and inner-leaflet of the PM. Of note, SM in the inner leaflet formed domains of 60–180nm radius [85, 86]. Notice that the terminology can be confusing. It is defined as clustered because is detected with lysenin, and the term domain refers that detection is confined in small, defined areas in the membrane.

Cer antibodies:

An alternative to using toxins and other probes came with the development of anti-Cer antibodies with one of the first being developed by the Okazaki group (IgM isotype, named NHCER-2) as a way to study Cer changes following Dox treatment [87]. However, while there appeared to be some utility of these antibodies, there were also issues raised with these initial attempts. Indeed, a comparison of commercially available antibodies from GlycoTech (polyclonal, IgM isotype, MAS 0010) and Alexis (monoclonal IgM isotype against C14-Cer, MID 15B4 still commercialized by Millipore-Sigma) found that both antibodies also recognized Cer metabolites such as dhCer, keto-Cers and N-methylCer. On lipids spotted in nitrocellulose, MID 15B4 also recognized phosphatidylcholine as well as Cer. To attempt to address this, more recently, the Bieberich group developed an IgG-isotype Cer antibody raised against C18-Cer in rabbit that was found to be specific for Cer when compared to other typical lipids. Moreover, lipid overlay assays in nitrocellulose membrane showed the antibody recognized all tested Cer species with a preference for C18. Further analysis by immunofluorescence showed that the antibody clearly stained plasma membrane protrusions and the Golgi apparatus, and that staining was reduced when SL synthesis was blocked with myriocin [88]. Of note, this antibody was also used to localize Cer in the leading edge of expanding lamellipodia and to modulate Cer signaling in the PM [89]. Anti-Cer antibodies have also been used to develop the evolving idea of Cer-enriched platforms, as visualized using Clone 15B4 in capping lymphocytes in response to CD95 signaling [90] and further revised by the Bieberich group [87]. Finally, an anti-Cer antibody was reported by Kolesnick and colleagues (IgM, 2A2) selected after screening against C16-Cer [91], and validated using bSMase and bCDase to generate and deplete PM Cer. This antibody was also used to visualize Cer generated at the PM in HeLa cells treated with doxorubicin and following nSMase2 overexpression [73].

Fluorescent SL analogs:

A number of fluorescent SL molecules such as N-[7(−4 -nitrobenzo-2-oxa −1,3-diazole)] −6-aminocaproyl Sph (C6-NBD-Cer) [92] and N-[5-(5,7-dimethyl BODIPY)-1-pentanoyl]-D-erythro-Sph (C5-DMB-Cer) [93] were used in early attempts to visualize cellular localization and metabolism of SLs. These probes combined with fluorescent microscopy, and chromatographic systems (TLC, HPLC) showed metabolism of the fluorescent lipids towards complex lipids, their accumulation in Golgi apparatus and vesicle transport to the PM [93]. Although they have proved to be very useful to measure SL-metabolizing enzyme activities in in vitro and in cellulo assays [94–96], they cannot be used to localize specific SL in the cells due to their rapid metabolism. Additionally, the large fluorophore groups of fluorescent SLs have raised concerns that they do not behave as the endogenous structures. Indeed, NBD- and BODIPY- lipids do not behave as natural SLs in membranes models [97].

Clickable SLs:

Progress on bio-orthogonal click chemistry [98] led to the development of SLs containing small modifications in their structure (for example, azide or alkyne groups), and solving the issue of having large fluorescent structures. These compounds with small modification can then react with fluorophores or other label groups to reveal their localization. The Delgado group first reported the synthesis of w-azidosphingoid bases that, when added to cells, were found to be metabolized to 1-phosphate sphingoid bases, their Cer, SM and HexCer analogs, probing their metabolism in the SL pathway and following a similar pattern as endogenous SLs [99]. Of note, clickable Cer tend to accumulate in the Golgi apparatus [100], similar to their fluorescent counterparts, probably to its metabolism to complex SLs. Valentin Wittmann generated a series of clickable glucosylceramides and lactosylceramides with different acyl chain length and the azido group either in the glucose moiety or on the N-acyl chain. These compounds were incorporated in the plasma membrane in the presence of defatted BSA and kept at 4C to avoid internalization and clicked with fluorophor for visualization. One of the advantages of having the azido-group in the sugar moiety is that lipid catabolism will lose the tag, and the labeling in membranes. Although SL click chemistry represents a still novel tool to probe SL metabolism and compartmentalization, it still has the limitation that the fluorescence from click functional group in the lipid moiety can not distinguish between SL species.

As clickable functionalized SLs have been shown to behave as endogenous SLs, a recent evolution of clickable SLs has led to bi- and tri-functional structures. Here, in addition to the clickable group for illuminating the SL, a biotin or other bait can be used for pull-down assays, or other groups, such as diazirine, are suitable for photoactivated crosslinking between the SL and interacting proteins [101, 102]. Finally, in a recent case, a third functional group was added to generate caged (see below), crosslinkable, and clickable Sph [103]. It should also be noted that a combination of these techniques has also been explored. For example, the proximity ligation assay (PLA) using an anti-Cer antibody and crosslinkable SLs was used by the Bieberich group to specifically identify Cer binding partners crosslinked to pacCer [104], ignoring other metabolites with the pacSphingoid structure. This approach led to the identification of acetylated tubulin [105], VCAD1 [106], and Hsd17b4 [107] as Cer interacting proteins in ciliary microtubules, mitochondria, and MAMs respectively.

Caged SLs:

Caging of a lipid refers to the attachment of a photocleavable group into its structure such that its bioactive function is blocked. When a lipid is uncaged, a specific wavelength of light is used to cleave the chemical group and liberate the functional lipid. For a summary of caged lipids, including SLs, we recommend the review by Schultz [108]. Caging techniques gained a topological meaning with Andre Nadler’s work on uncaging lipids in specific subcellular compartments. In those, bioactive lipids, including Sph, were caged with coumarin and tagged with organelle targeting moieties, including mitochondria, PM, ER, and lysosomes. Caged-Sph localization was confirmed using a confocal microscope and calcium release upon uncaging was monitored, showing that similar Sph production in different organelles has different effects on calcium release. The potential of organelle targeted SLs, using different species, or combined with the other mentioned functional modifications can easily be imagined.

FRET probes:

Fluorescence resonance energy transfer (FRET) is a process where fluorescent energy is transferred from one fluorophore to another, when two dyes with overlapping emission/excitation spectra (FRET pair) are close by. This property has been exploited to determine if two molecules are in close proximity or are interacting, and if such interaction is lost upon treatment. For SLs, FRET assays have been developed to quantify activity of SL-metabolizing enzymes using intact cells as a substitute for tedious enzymatic assays requiring derivatizations of enzymatic products, as well as radioactive or fluorescent compounds that need to be resolved by TLC or HPLC. More importantly, such assays also maintain the topological information that is lost in assays with cell lysates. At present, a few groups have developed FRET probes based on the SL structure. The Arenz lab has developed Cer and SM structures that contain FRET pairs within the same structure. This enables measurement of the FRET radiation in intact compounds but once the probes are hydrolyzed by CDases and SMases, the fluorophores go with each hydrolyzed product and the FRET signal is lost [109]. Thus, the enzyme activity can be assessed as a a function of the loss of FRET signal. Using this Cer probe, the double-labelled compound was found to localize to the Golgi and a FRET signal was recorded proving compatibility with an in vivo system but suggesting that CDase activity was not detected. The lack of activity was not further explored, but may reflect an absence of CDase activity in the Golgi or possible metabolism of the compounds to SM, thereby escaping the action of CDase. Future work targeting the compound to lysosomes and mitochondria has also been discussed [109]. In vitro use of the SM FRET probe, found that compound was substrate for the acid but not for the neutral SMase so may have more limited utility in cells. However, possible modifications in the probe to make it a substrate for neutral SMase would broaden its utility and be a useful complementary tool for studying topology of SM hydrolysis in the cell. The authors also discussed the possibility of changing delivery system to ensure the compounds reach several organelles [109]. While other SL FRET probes have also been developed, their main functions – as yet – were not for characterizing SL topology. For example, FRET analog probes of SM and enantiomeric SM were used in biophysical studies to show that homo-SM molecules interact closer that hetero natural/enantiomeric SM [110]. Finally, the Arenz group also used Cer FRET probes to develop an in vitro assay for CERT activity as a means to identify inhibitors of this activity [111]. Overall, although not quite there yet, is clear that changes in FRET imaging have a lot of potential as a tool for localizing SL metabolizing enzymes activity in the cell.

Imaging Mass Spectrometry:

Some of the previously listed strategies combine microscopy with different techniques to visualize the localization of SLs within the cell. These localizations are based on indirect readings, probes such as anti-SL antibodies, SL binding proteins, or synthesized analogs of sphingoid structures mimicking natural compounds. Although these techniques offer the possibility to localize the SL function in specific subcellular sites, they lose information on the immense diversity in SL structures within each SL species. Imaging mass spectrometry applied to lipids, specifically to SLs, is closing the bridge between the power of MS/MS and cellular visualization. A growing number of works have reported the heterogeneity in sphingolipid distribution within animal tissue [112, 113]. However, these studies seek enough resolution to report the topology of the sphingolipidomics within a single cell. Although this point is not reached, d’Angelo group has reported MALDI-MSI of 205 annotated lipids with a resolution of 25–50 μm2 [114], allowing not only to reveal cell to cell heterogeneity but to combine to other single-cell techniques, allowing to relate specific SLs to pathways and biological processes.

Modulating Compartment SLs

In addition to visualizing SL localization, a number of tools have been developed to modulate SL levels in specific compartments as a means to probe function. Of all compartments, the outer leaflet of the PM is the most accessible to study without the need for cell disruption or long and complex protocols and several approaches have been used to study this compartment. However, studies have also begun to explore tool targeted to intracellular organelles.

Methyl-α-cyclodextrin (MαCD) for extracting lipids:

Methyl-beta-cyclodextrins (MbCD) have been commonly used to manipulate cholesterol and phospholipids in cells, as a means to infer lipid raft involvement in cellular functions. In comparison, the cavity in MαCD is too small for cholesterol but still interacts with glycerolipids and SLs. Taking advantage of this, the lab of Erwin London’s lab developed a technique to replace SLs and glycerolipids from the outer leaflet of the plasma membrane [115, 116]. This allowed the study of lipid composition of the plasma membrane but also it has the potential to research SL function in the plasma membrane.

Recombinant bacterial sphingomyelinase (bSMase):

When applied to the culture medium, bSMase will act on the outer leaflet of the PM but will not get internalized. It has been used to measure SM content and composition from the outer leaflet of the PM [117]. It has also been used as a means to manipulate endogenous SM and Cer levels and interrogate the novel functions of Cer specific from the plasma membrane [73].

Recombinant bacterial ceramidase (bCDase).

Similar to bSMase, bCDase has also been used to manipulate Cer in the plasma membrane without affecting Cer in other compartments. The use of this technique helped confirm that there is little to no Cer in the PM in basal conditions. It was also used as a means to detect acute changes in Cer content in the plasma membrane in response to specific inducers, such as chemotherapy [118].

Purified kinases:

Diacylglycerol and Sph kinases combined with radioactive [³²P]ATP, have been used to phosphorylate Cer and Sph respectively in intact cells, and radioactive counts were used to quantify Cer and Sph content in the plasma membrane [119, 120].

Organelle-targeted bSMases CDases, and CERT:

Expanding on the strategy of using bacterial enzymes at the PM, studies have also taken advantage of intracellular targeting of these enzymes. This approach was first used by the Obeid lab was able to target bSMase to different organelles including the inner-leaflet of the plasma membrane, ER, Golgi nucleus membrane and mitochondria. Interestingly, only bSMase targeted to the mitochondria, but not to other organelles, resulted in apoptotic cell death [121]. Later, using these constructs, we show that SM, Cer, and Sph content could be manipulated in different organelles by a combination of targeted bSMase and bCDase [122]. This also allowed us to find that the SL profile seems to follow a pattern being enriched in very long acyl-fatty acid chains in the ER membranes, towards long chain in the way through Golgi apparatus, with a lipid gradient orthogonal to the PM. A similar targeting approach was used by the Holthuis group as a means to redirect CERT-mediated trafficking of Cer from the ER to other organelles. In their initial studies, CERT redirected Cer to the mitochondria and was used in combination with a mitochondria-targeted bacterial CDase to interrogate Cer specific functions [123]. Clearly, these are powerful tools for interrogating organelle-specific functions of SM and Cer.

IMPLICATIONS OF SL COMPARTMENTALIZATION FOR SIGNALING AND THERAPEUTICS

Signaling

Much of the effort to understand the molecular mechanisms by which SLs work has been thought of in two ways: as a second messenger, where the SL directly interacts with and regulates an effector protein, or the SL acts as a modulator of the biophysical properties of membranes. However, as discussed in the Introduction, SL species are often functionally treated as unique entities e.g., Cer triggers apoptosis, cell differentiation, and not as a family of related molecules. Moreover, much of the early work probing the signaling functions of Cer combined analysis of cellular SLs with the addition of exogenous lipid. The presence of SL metabolizing enzymes in many organelles, as discussed above, means that the same SL can be generated in many compartments by a different enzyme but may exert different functions owing to the specific environment of that compartment. Here, we will briefly discuss this in the context of Cer, perhaps the most well-studied SL, focusing on Cer in specific compartments. Some of the signaling effects of Cer in specific compartments is shown in Fig. 1

Cer in Mitochondria:

While this was followed up by studies confirming that endogenously generated Cer functioned as a mediator of apoptosis, it was still a number of years until the idea of compartment specific functions of Cer was addressed. In that study, Obeid and colleagues targeted bSMase proteins to various intracellular organelles. This established there were pools of SM at various organelles throughout the cell but, crucially, also found that only selective generation of mitochondrial Cer was able to induce cell death [121]. Follow up studies showed that mitochondrial Cer generation led to activation of Bax [128], and that mitochondrial targeted Cer analogues also induced cell death [129]. These findings were subsequently validated by Holthuis and colleagues who found that redirecting CERT-mediated transport of Cer to the mitochondria promoted Bax-dependent apoptosis and this was prevented by a mito-targeted bCDase [123]. Taken together, these provide compelling evidence that mitochondrial generation of Cer can promote cell death. However, while prior research has focused on apoptosis, it should also be noted that more recent studies have found that C18-Cer accumulation in mitochondria promotes lethal mitophagy, independently of apoptosis [130]. This is a useful reminder that species specific effects must also be kept in mind when considering compartmentalization.

Cer in the PM:

In contrast to mitochondrial Cer, early studies suggested that generation of Cer at the outer PM was not required for programmed cell death [131, 132], despite being able to induce hemolysis of red blood cells [133]. In contrast, we were able to find that exogenous bSMase treatment led to a profound reorganization of the cytoskeleton that was mediated by dephosphorylation of the ezrin family of proteins (ERM). Building upon these findings, the acute generation Cer at the PM was linked to the activation of an anti-adhesion, pro-migration signaling protein. Crucially, this was linked to chemotherapy activation of nSMase2 – which had previously been localized to the inner PM [73, 82].

Cer in the ER:

As discussed above, the levels of Cer in the ER are regulated by a number of metabolic pathways and an alteration in the balance of these pathways has been reported to have a number of effects. A substantial Cer accumulation in the ER has been linked to induction of cell death – although this has been attributed to a transfer of Cer into mitochondria [17, 18]. Related to this, accumulation of Cer in cases of lipid overload have been implicated in activation of ER stress pathways. Studies in pancreatic B-cells with palmitate overload suggested this was due to alterations of the biophysical properties of ER membranes which lead to a disruption of protein export from the ER [134]. In contrast, treatment of intestinal epithelial cells with myristate, but not palmitate, let to activation of the XBP1 and IRE1 arms of ER stress and cytokine production through generation of C14-Cer [135]. This, again, emphasizes that effects of specific lipids species must be kept in mind, even in the same compartment.

Cer in extracellular vesicles:

As well as SL signaling in intracellular compartments, there is emerging evidence that Cer can also function as a signaling molecule in extracellular vesicles. The best studied example of this is exosomes, which are 30–100nm secreted vesicles that carry protein, lipids, nucleic acids, and RNA as cargo. Recent research in this area was stimulated by the finding that the nSMase2-Cer axis regulated exosome formation and secretion [136]. Since then, studies have suggested that nSMase2 can also influence the mRNA and microRNA content of exosomes, although this might be cell type specific [137, 138]. Biologically, studies have linked the nSMase2-exosome pathway to reprogramming of the tumor microenvironment with a recent study in melanoma linking this to the response to immunotherapy [139]. Outside of cancer, the enrichment of Cer in exosomes was also linked to the development of amyloid plaques in Alzheimer’s disease. In those studies, Cer in astrocyte-derived exosomes (termed astrosomes) was found to bind Ab peptide and promoted exosome aggregation with subsequent uptake of Ab-bound astrosomes by neurons leading to activation of cell death pathways [140]. Of note, more recent studies suggested a similar role for nSMase2-mediated exosome release in a-synuclein aggregation in a mouse model of Parkinson’s disease [141]. Thus, although in its infancy, further understanding of SLs in exosomes could provide additional opportunities for therapeutic intervention.

Although not an exhaustive discussion, the above examples serve as a good illustration of how the generation of Cer in three different compartments can influence cellular signaling and outcomes. Mechanistically, the reasons for these outcomes may well be due to the local environment, both in terms of the metabolic pathways accessible to the lipids or the different binding partners accessible there. This is most clearly suggested in the mitochondrial context by the connecting of Cer and other metabolites with VDAC2 and Bax/Bak respectively [129, 142, 143].

Therapeutics

The dysregulation of SL metabolism that has been linked with many diseases has led to substantial interest in therapeutically modulating cellular SL levels. In the context of cancer treatment, this approach has primarily focused on increasing the intracellular levels of Cer, given its well established role in mediating cell death. Strategies to achieve this are primarily through addition of exogenous lipids, inhibition of Cer metabolizing pathways, or a combination of the two. These approaches are not without pitfalls. For example, non-transformed cells can be susceptible to increased Cer levels as cancer cells [144, 145]. Exogenous Cer treatment can upregulate Cer metabolic pathways which can ultimately lead to a resistant phenotype [146]. Finally, there are many known pathways of Cer metabolism and targeting many of these pathways at once could be a difficult exercise. Nonetheless, despite these challenges, there has been some progress with the development of Cer nanoliposomes that have shown success in pre-clinical models and are currently in clinical trials [147, 148].

As discussed here, the compartmentalization of SL metabolism and signaling can have profound consequences for biological responses, which now presents an additional consideration for development of SL-based therapies. When considering a strategy to increase cellular levels of Cer, for example, there is a need to understand that any exogenously added Cer (or other lipid) is reaching the correct compartment to elicit the desired response. In the case of cancer, this would be an effective cell death. In this context, lipid analogues that are designed to accumulate in specific cellular organelles – as described above – could be valuable tools. Similar consideration must also be made when combining exogenous lipids with inhibitors of metabolic pathways. In this sense, understanding the local metabolic environment at the desired destination of the exogenous lipid is important when selecting the appropriate inhibitors to use. In this sense, inhibitors of specific pathways might have no benefit the exogenous Cer (or other lipid) is not being funneled into that pathway. By understanding the ultimate cellular location and tracing the metabolism of exogenously added lipids – both in the cell and in whole animals – would provide guidance for the best combinatorial treatments. As exogenous lipids may also upregulate specific metabolic pathways as a detoxification mechanism, this may also provide insight mechanisms of treatment resistance.

CONCLUSIONS

In the last two decades, our understanding of the complexities of SL signaling has grown considerably. While early studies established localization of many of the SL enzymes, the impacts of compartmentalized signaling remained relatively unexplored. With the development of a host of targeted constructs, probes, and sensors, we now have a toolkit that can be used to probe the signaling and biologies associated with compartment specific SL metabolism can now proceed. We firmly believe that such studies are essential for unravelling the mysteries of SL signaling, for understanding how disruption of SL metabolism leads to pathological conditions, and to best allow us to exploit SL metabolism for therapeutics.