On the future of mass-spectrometry-based lipidomics

Authors


Correspondence

U. Loizides-Mangold, Department of Biochemistry, University of Geneva, 30 Quai Ernest-Ansermet, CH-1211 Geneva, Switzerland

Fax: +41 22 379 6470

Tel: +41 22 379 6198

E-mail: ursula.loizides-mangold@unige.ch

Abstract

Lipids have highly diverse functions that go beyond cellular membrane structure and energy storage. One of the great challenges in lipid research will be to understand how the enormous complexity of lipid homeostasis is maintained. Genetic approaches combined with mass spectrometry-based lipidomics will help to elucidate how cells create and maintain their nonrandom lipid distribution within tissues, cells, organelles and lipid bilayers. Lipid homeostasis is crucial for many cellular processes and we are currently only beginning to understand the specific functions of lipids and the local environment that they create.

Abbreviations
EI

electron ionization

ESI

electrospray ionization

GC

gas chromatography

GlcCer

glucosylceramide

GPI

glycosylphosphatidylinositol

MS

mass spectrometry

Introduction

Lipids have been recognized as a highly diverse class of metabolites [1, 2]. Their role was originally seen in the context of membrane structure and energy storage, but it has long been established that lipids have much more diverse functions, which include important roles as regulatory and signaling molecules [3, 4]. Because lipids are not uniformly distributed among organelles and also display asymmetric transbilayer organization, cells must conduct an enormous effort to maintain their lipid homeostasis [5-7]. The functional consequences of this lipid diversity are still not fully understood and novel tools are needed to investigate why cells undertake such an enormous effort to maintain their lipid homeostasis.

The molecular biology of cellular lipids

Lipids have often been described as a group of compounds that are hydrophobic and therefore highly soluble in organic solvents [8]. However, certain lipids, such as complex glycosphingolipids, are equally soluble in water and organic solvents and a definition based on solubility is incomprehensive. In addition, precursors of lipid biosynthesis and products of lipid degradation are water-soluble metabolites that would not be included in a definition based on solubility [9].

This predicament has led to alternative definitions such as that proposed by Christie who categorizes lipids as ‘fatty acids, their derivatives and substances related biosynthetically or functionally to these compounds' [10].

More recently, the LIPID MAPS consortium, a multi-institutional group created in 2003 by six US-based lipid laboratories, has defined lipids based on their structural and biosynthetic characteristics as hydrophobic or amphiphilic small molecules that may originate either entirely or in part from condensation of ketoacyl or isoprene subunits [11, 12].

The most common lipid classes that play important roles in membrane structure and energy storage are glycerophospholipids, sphingolipids, sterols and triglycerides.

Glycerophospholipids and triglycerides consist of fatty acids linked by ester bonds to glycerol, whereas sphingolipids are sphingoid bases linked by amide bonds to fatty acids.

Although lipids are composed of a limited number of building blocks, they have the potential to generate a highly diverse array of individual molecular species [13]. They may contain phosphoric acids, organic bases and carbohydrates. Also, the type of linkage at the sn-1 position of glycerol can vary because fatty acids may be linked by ether instead of ester bonds.

In mammalian tissue, the number of different fatty acid species found in lipids is typically in the order of 30–60 [2, 14]. The fatty acyl chain length normally lies between 12 and 26 carbons, with up to six double bonds, and can contain hydroxyl groups. Thus, the enormous structural diversity found in lipids arises mainly through various combinations of the different fatty acid chain lengths, the number and location of the double bonds and the different functional head groups linked to the glycerol or sphingoid base backbone (Fig. 1) [2, 13, 14].

Figure 1.

Diversity of glycerophospho- and sphingolipid species. (A) Glycerophospholipids share the same sn-1, -2 diacylglycerol structure with a phosphate residue in position sn-3. Individual species differ by the number of carbon atoms and double bonds, the type of linkage of the fatty acyl chains and their head group moieties. R refers to possible head groups that give rise to phosphatidic acid (PA), phosphatidylcholine (PC), phosphatidylethanolamine (PE), phosphatidylserine (PS), phosphatidylinositol (PI), phosphatidylglycerol (PG). (B) Sphingolipids share the sphingoid base backbone that is linked via an amide bond to a fatty acid. Individual species differ by the type of sphingoid base backbone (sphinganine, sphingosine, phytosphingosine) and the diversity of the fatty acyl chain and head group modifications. Most prominent sphingolipids are ceramide, sphingomyelin, GlcCer, galactosylceramide and complex glycosphingolipids such as gangliosides.

The three major classes of membrane lipids in eukaryotes are glycerophospholipids, sphingolipids and sterols. The glycerophospholipids can be classified according to their headgroup into phosphatidylcholine, phosphatidylethanolamine, phosphatidylinositol, phosphatidylserine, phosphatidylglycerol and phosphatidic acid (Fig. 1A). Using the techniques currently available, around 100 molecular species per glycerophospholipid class can be routinely detected in crude lipid extracts [2], which gives a total of ~ 600 glycerophospholipids. However, the theoretical number is much higher and has been estimated to be in the order of ~ 10 000 glycerophospholipids [2, 14].

Mitochondrial membranes of eukaryotic cells possess another class of glycerophospholipids, which are called cardiolipins. Cardiolipins are found almost exclusively in the inner mitochondrial membrane and in certain membranes of bacteria. They are composed of two phosphatidic acids that are connected via a glycerol backbone. Because cardiolipins have four distinct acyl chains the potential for complexity is enormous. However, eukaryotic cells only use a limited array of different fatty acids for their cardiolipin biosynthesis, keeping the lipid diversity relatively small [15, 16].

The other major class of membrane lipids, sphingolipids, belongs to a family of lipid metabolites that share a common structural feature, the sphingoid base backbone (Fig. 1B). The sphingoid base is synthesized de novo from serine and a long-chain fatty acyl-CoA, normally palmitoyl-CoA (Fig. 1) by serine palmitoyltransferase [17, 18]. However, serine palmitoyltransferase can also metabolize different acyl-CoAs and other amino acids such as l-alanine and glycine, which generates a spectrum of atypical sphingolipids such as deoxysphingolipids [19]. Free sphingoid bases are usually present in very low amounts and are commonly linked by amide bonds to long- and very-long-chain fatty acids to form ceramide, which can then be further modified via a polar headgroup giving rise to sphingomyelin or the complex glycosphingolipids (Fig. 1B) [11]. Also, the total number of estimated sphingolipid metabolites varies widely and has been proposed to yield up to 100 000–200 000 distinct sphingolipid moieties, which is mostly due to the combinatorial chemistry of the complex glycosphingolipids [14, 20].

Both, glycero- and sphingolipid biosynthesis are closely linked to the important bioactive lipids that are derived from them [3]. Phosphoinositides, lysobisphosphatitic acid, mono- and diacylglycerol, as well as platelet-activating factor are glycerolipid-derived regulatory metabolites, whereas sphingosine, its related sphingoid bases and sphingosine-1-phosphate, as well as ceramide and ceramide-1-phosphate originate from sphingolipid metabolism [3].

Although the bulk amount of cellular lipids are organized in membranes, storage lipids such as triacylglycerols and cholesteryl esters form the core of cytosolic lipid droplets and of lipoprotein particles, which are secreted or endocytosed [21]. Whereas cholesteryl esters consist of a limited number of lipids, triacylglycerols display an enormous array of complexity and have been proposed to comprise up to 64 000 different molecular species [14].

In total, the cellular lipidome has been estimated to encompass ~ 180 000–200 000 different lipid species [14, 22]. Because this complexity has survived evolution it must in some ways be of functional use [9]. To assess the significance of this enormous complexity new technology and approaches are needed.

A comprehensive classification system for lipids, which is compatible with current bioinformatic requirements, was introduced by the LIPID MAPS consortium [11]. This classification system has been under the leadership of the International Lipid Classification and Nomenclature Committee (ILCNC). In 2009, the classification system was updated to encompass lipid structures from non-mammalian sources such as plants, bacteria and fungi [23]. Today, complex glycosphingolipids from Saccharomyces cerevisiae such as inositolphosphoryl ceramide, mannosyl inositolphosphoryl ceramide and mannosyl di-inositolphosphoryl ceramide can be found within the LIPID MAPS database [24]. Ongoing interest into the lipidomes of different model systems will further expand this database and fuel our understanding of the enormous variety of the existing lipid structures.

Lipid analysis, lipidomics

Lipid analysis in the context of cell biology emerged with the pioneering work of Eugene P. Kennedy on phospholipid biosynthesis [25, 26]. Traditionally, lipid analysis has relied on analytical techniques with low resolution and sensitivity, such as thin-layer chromatography where lipid analysis is confined to the study of entire lipid classes. Although this method is limited it is still commonly used today because of its relative ease of use and low experimental costs. With the development of gas chromatography the separation of different fatty acid components from simple mixtures became available [27]. However, the identification and quantification of individual lipids from complex mixtures such as crude lipid extracts remained challenging. The field was advanced when gas chromatography (GC) was combined with mass spectrometry (MS) [28]. However, lipids have an enormous complexity, which could not be assessed at the time. In past decades, the field of lipid research has made great progress by impressive advances in MS, in particular, soft ionization techniques such as electrospray ionization (ESI) [29-32] and MALDI [33-35]. With the current advances in MS it is now possible to approach the entire lipidome of cells or individual organelles [36, 37].

The name ‘lipidome’ appeared first in the literature in 2001 and described the complete lipid composition within a cell, tissue or organism [38]. Around the same time the term ‘functional lipidomics’ was introduced by Rilfors and Lindblom who coined it to describe ‘the study of the role played by membrane lipids’ [39]. At that time, lipidomics was seen mainly in the context of other functional ‘omics’ technologies such as genomics, and did not refer to large-scale MS-based analyses. In 2003, Han and Gross defined lipidomics as a global analysis of cellular lipidomes by a comprehensive MS approach [40].

Today, lipidomics is seen in a broader context as a field of systems level analysis. It involves the comprehensive identification and quantification of all individual cellular lipid species and the characterization of their interactions with neighboring lipids and proteins [2, 13]. This also includes the expression of proteins involved in lipid metabolism and function, including gene regulation [41].

Commonly, lipids are extracted using the chloroform- and methanol-based protocol of Folch [42], or Bligh and Dyer [43]. Recently, another type of lipid extraction, the so-called methyl-tert-butyl ether extraction [44] has emerged which extracts lipid equally well but has the advantage that lipids partition into the upper layer of a two-phase extraction system and are free of contaminants that could block the electrospray of nanometer-size nozzles [45]. Alternatively, lipids can also be extracted by pyridine extraction, which leads to the effective extraction of inositol- containing sphingolipids and phosphatidylcholine from intact S. cerevisiae [46]. For the analysis of sphingolipids by MS an additional step of mild alkaline hydrolysis is recommended to reduce ion suppression during MS analysis by other lipids such as glycerophospholipids.

In a typical lipidomics experiment, the lipid sample is delivered to the mass spectrometer, ionized and vaporized, and the resulting ions are sorted according to their m/z ratio in the mass analyzer [47]. If a crude lipid extract is introduced directly into the mass spectrometer without previous chromatographic separation it is called shotgun lipidomics [32, 48, 49]. Shotgun lipidomics is inherently biased towards the more abundant and easily ionized lipids, but the dynamic range of lipid detection can be greatly improved by certain additives, the use of different electrospray polarity and tandem MS [13, 47]. Alternatively, the lipid extract is introduced to the mass spectrometer after liquid chromatography (LC-MS or HPLC-MS), which increases the number of detectable lipids due to reduced ion suppression [50] but can also introduce problems concerning ionization conditions and column memory.

ESI is a soft ionization technique that is usually employed in lipidomics although other ionization techniques such as MALDI [33, 34] and atmospheric pressure ionization have also been used [51-53]. Recent advances allow an even more detailed structural characterization such as position of the fatty acyl group on the glycerol backbone or the elucidation of the double bond position in unsaturated lipids [54, 55]. With these new advances in MS that allow the unambiguous identification of isobaric lipid species, lipidomics will reach a new level of complexity.

Analysis of lipids by ESI-MS is based on the ability of each class of lipids to acquire positive or negative charges when in solution during ionization [56]. Therefore, electron ionization in combination with GC-MS is often used to analyze neutral lipids such as cholesterol, cholesteryl esters and triglycerides. In electron ionization, the sample is vaporized into the mass spectrometer ion source. There, it is impacted by a beam of electrons, which causes the molecular ion to undergo structure-specific fragmentation.

In addition, lipids can be directly identified from crude lipid extracts solely by their mass. This so-called top-down lipidomics requires high-resolution mass spectrometers such as the Fourier transform ion cyclotron resonance and FT-Orbitrap instruments [49, 57]. Both are usually commercialized as linear ion trap hybrids, e.g. LTQ-FT and LTQ-Orbitrap. These instruments offer the advantage of fast scanning and high mass accuracy (1–3 p.p.m.) [47, 49]. In top-down lipidomics, the goal is mainly to distinguish differences in the lipid pattern rather than quantifying individual lipid species although it can also be used for this purpose.

However, most lipidomics experiments are performed using tandem mass spectrometry (MS/MS) [49]. An MS/MS experiment consists of two mass analyzers separated by a collision cell (containing collision gas) and is capable of a number of different MS/MS scan options (precursor ion scan, product ion scan, neutral loss scan) that produce structure-specific fragment ions. This approach has been termed bottom-up lipidomics [49, 58]. To obtain maximum sensitivity when using MS/MS, selected reaction monitoring or multiple reaction monitoring (MRM) is often performed and signals are only recorded when a specified precursor/fragment mass pair is detected [47]. The disadvantage of this approach is its targeted nature, which does not allow the detection of unanticipated lipid metabolites.

In lipidomics, the study of one lipid class or subclass is referred to as targeted lipidomics [13, 59]. In a targeted lipidomics approach the different lipid classes are analyzed separately using extraction and analytical protocols designed specifically for each lipid class. This approach was adopted by the LIPID MAPS consortium for the analysis of mammalian lipids [60]. However, targeted proteomics refers to the technique of selected reaction monitoring, which allows the detection and quantification of predetermined precursor/fragment ion pairs of proteotypic peptides [61]. Because technically selected reaction monitoring within the context of proteomics or lipidomics is comparable, it would be desirable if both definitions were congruent.

Currently available lipidomics tools for annotation (LIPID MAPS, LipidBank), shotgun MS/MS software tools such as the publicly available lipidxplorer [62], LC-MS/MS tools and databases containing reference spectra such as the Human Metabolome Database or the MassBank database (www.massbank.jp) demonstrate the growing need for detailed bioinformatics resources to facilitate lipid research. A detailed summary of the available resources can be found in a recent review by Hartler et al. [63]. New bioinformatics strategies have emerged that allow the study of molecular lipid profiles in the context of known metabolic pathways [64]. However, currently available pathway-level representation of lipids in databases such as the KEGG database [65] or SphinGOMAP are limited and lack the level of detail that is available by modern MS-based approaches.

Lipid distribution and trafficking

Lipids are not homogeneously distributed between subcellular organelles, or between the two leaflets of organelle membranes [5]. Most lipids are synthesized in the endoplasmic reticulum and Golgi and then transported to different cellular compartments. For example, sterols are synthesized in the endoplasmic reticulum but are gradually enriched along the secretory pathway with the highest level at the plasma membrane [6]. In addition, endocytic pathways and lipid recycling contribute to the intracellular transport and heterogeneous distribution of lipids. This cellular homeostasis is achieved by the concerted actions of numerous enzymes, proteins and receptors that synthesize, bind and transport lipids. In return, this lipid complexity is needed for biological processes such as vesicle fusion and fission, membrane sorting and signal transduction [66].

Our current knowledge about the localization of phospholipid biosynthesis and distribution originates from pioneering studies that used cell fractionation and metabolic labeling [67]. Phospholipid synthesizing enzymes were found predominantly on the endoplasmic reticulum and in the mitochondria although some phosphatidylcholine biosynthesis takes place in the nucleus [68]. However, recent data have emerged which indicate that phospholipid biosynthesis might be more compartmentalized [69]. The exchange of lipids and proteins between organelles can occur by vesicular transport and components of this molecular machinery, which include Rab GTPases [70], SNARE proteins [71], coat complexes [72] and tethering factors [73] have been described in great detail. However, organelles such as mitochondria and peroxisomes are not connected to the vesicular transport systems. Lipids must exit and enter in a nonvesicular manner and nonvesicular transport through membrane contact sites or lipid transfer proteins must play an important role in interorganellar lipid transport.

Lipids diffuse rapidly within one bilayer through lateral movement, but polar lipids such as glycerophospholipids do not easily cross the hydrophobic membrane interior because of the polarity of their head groups [5]. To maintain lipid asymmetry, eukaryotic cells contain phospholipid flippases like the family of type 4 P-type ATPases, which translocate phospholipids such as phosphatidylserine and phosphatidylethanolamine from the exoplasmic to the cytoplasmic leaflet of the lipid bilayers [74]. Several studies have indicated that type 4 P-type ATPases also play a role in vesicle formation, which suggests that in addition to lipid translocation, flippases may also cause an imbalance in total lipid mass between the two bilayers, which leads to curvature of the membrane [74].

Lipids without polar head groups like cholesterol or ceramide can flip spontaneously between lipid bilayers. However, nonvesicular lipid transport between organelles involves the role of specific lipid-binding proteins. In the case of ceramide, this is performed by the ceramide transport protein CERT, which facilitates the transport of newly synthesized ceramide from the endoplasmic reticulum to the Golgi, where it is metabolized to sphingomyelin [75]. The transport activity of CERT and consequently the biosynthesis of sphingomyelin is mediated through the phosphorylation status of CERT and through its phosphoinositide-binding (PtdIns4P) domain [76]. However, CERT deficiency does not affect glucosylceramide (GlcCer) biosynthesis, which indicates that ceramide destined for GlcCer biosynthesis is transported differently [75, 77]. The transport of ceramide which is metabolized to GlcCer and complex glycosphingolipid biosynthesis might be connected to the glycosylphosphatidylinositol (GPtdIns) anchor biosynthesis because it was recently shown that the transport of very-long-chain ceramide destined for GlcCer and GM3 biosynthesis is coordinated with the endoplasmic reticulum to Golgi transport of GPtdIns-anchored proteins [78].

Biosynthesis of complex glycosphingolipids requires the Golgi-associated four phosphate adaptor protein 2 (FAPP2). FAPP2 specifically binds and transports GlcCer but not ceramide or sphingomyelin [77], however, it is still under debate whether FAPP2 mediates the nonvesicular transport of GlcCer from the cis-Golgi to the trans-Golgi network or if it is rather involved in the retrograde transfer of GlcCer back to the endoplasmic reticulum [77, 79].

Cholesterol trafficking and distribution are highly dynamic and can be achieved by vesicular and nonvesicular processes [80]. The mechanisms by which intracellular cholesterol is transported are still largely unknown, but several players have emerged that facilitate the nonvesicular transport of cholesterol between organelles. One of them is the family of oxysterol-binding protein (OSBP) homologs, the OSBP-related protein family, which constitutes a conserved family of cytoplasmic lipid-binding/transfer proteins. Some family members are involved in the transport of cholesterol, whereas others function as sterol sensors [81]. Family members have also been found at membrane contact sites where they might facilitate the exchange of cholesterol. Currently, it is still under debate whether in mammalian cells, OSBP related proteins function as true lipid transport proteins. Instead of carrying sterols through an aqueous cytosolic environment, OSBP related proteins might act locally at membrane contact sites, facilitating the ‘flipping’ of sterols from one membrane to another. In fact, membrane contact sites, which are zones where signals and small molecules such as lipids are exchanged are still poorly understood and have been often overlooked as impurities during organelle preparation [82].

Receptor-mediated endocytosis of lipoproteins, such as low-density lipoprotein, and hydrolysis of their cholesteryl ester content represents another major source of cellular cholesterol. The exit of cholesterol from the late endosomal/lysosomal compartments requires the help of at least two proteins: the Niemann-Pick type C protein 1 and 2, whose exact functions are still unclear. Mutations in those genes are associated with Niemann-Pick disease, type C, which is a devastating neurodegenerative lysosomal storage disorder in which free, unesterified cholesterol and glycolipids accumulate in late endosomes and lysosomes [83, 84]. Niemann-Pick type C protein 1 has a putative sterol-sensing domain and Niemann-Pick type C protein 2 has a cholesterol-binding domain, but their exact function in sterol sensing and cholesterol transport is still under investigation.

Once at the plasma membrane, lipids laterally segregate in a way that allows them to fulfill their complex bioactive tasks. It has been proposed that highly dynamic nanodomains exist on the plasma membrane, which are enriched in sphingolipids, cholesterol and certain proteins (raft hypothesis) [85]. Those lipid ordered domains are supposed to serve as signaling platforms, however, the size and lifespan of these domains is still under debate. In addition, it has been proposed that ceramide-rich platforms can be generated on the exoplasmic leaflet of the plasma membrane through the activity of acid sphingomyelinase [86]. Ceramide-rich platforms display a high degree of lipid order, which promotes the reorganization of receptor and signaling molecules in response to diverse stimuli.

In summary, lipid localization and trafficking has been a very exciting area of research but many challenges still remain. Currently, there is a lack of tools to visualize lipid protein interactions, specialized lipid domains or lipid trafficking. For example, studies on cholesterol localization have relied greatly on its interaction with filipin, which is a mixture of different polyene antibiotics [87]. Cholesterol binds filipin but the structure of the cholesterol/filipin complex has not been determined. Similar concerns exist for the currently used anti-ceramide [88] or anti-sphingosine specific antibodies. In addition, many lipid visualization tools use chemically modified lipid analogs that do not show the same molecular behavior as their unmodified counterparts. New visualization tools are therefore needed and progress in this area will depend crucially on the development of new technical approaches.

Lipidomics of cellular pathologies

Lipid homeostasis is critical to maintain health because of the central role of membranes in cellular biology. Not surprisingly, lipids are involved in the pathogenesis of various diseases and defects in lipid metabolism are central to many devastating diseases such as insulin-resistant diabetes, Alzheimer's disease, cancer, atherosclerosis, steatohepatitis and obesity. A comprehensive analysis of the lipid metabolic changes is therefore crucial for the understanding of cellular pathologies and lipidomics will be a useful tool in the elucidation and characterization of defects in lipid homeostasis.

Lipids have only been recently implicated in neurodegenerative disorders such as Alzheimer's disease [89] although it has been known for a long time that the brain of Alzheimer patients contains lipid deposits suggesting a defect in lipid metabolism [90]. Recent lipidomics approaches on Alzheimer's disease point to ganglioside abnormalities [91]. Other lipidomics studies have shown alterations in the plasma sphingolipidome [92] and a loss of sulfatides combined with an increase in ceramides at the earliest stage of Alzheimer's disease [93].

Besides the well-studied role of bioactive lipids such as ceramides or sphingosine-1 phosphate in inflammation, apoptosis and cell proliferation, there has been renewed interest in the general role of lipids in cancer [3, 94]. Increased lipogenesis is an early event in carcinogenesis and a central hallmark of many cancers [95]. Healthy cells rely primarily on dietary fatty acids, whereas cancer cells including prostate and breast cancer cells use de novo lipid biosynthesis. To address the role of lipid metabolism in mammalian malignancies, shotgun lipidomics has been applied to characterize the membrane lipid composition during breast cancer progression [96]. The results imply that the viability of breast cancer cells depends on de novo lipid biosynthesis and that phospholipids such as palmitate containing phosphatidylcholines may have diagnostic potential. The introduction of lipidomics into clinical diagnostics has been further explored by a shotgun lipidomics study, which suggests the potential use of urinary phospholipids as candidate biomarkers for prostate cancer [97]. More data will emerge in the future from the characterization of urinary and plasma lipidomes with the potential identification and validation of lipid biomarkers that translate into clinical practice.

Altered regulation of glycolysis and mitochondrial function are considered a hallmark of cancer (Warburg hypothesis) [98]. Shotgun lipidomics of brain mitochondria from mice susceptible to glioma, showed altered phospholipid levels and a change in the molecular cardiolipin composition [99]. Cardiolipin alterations have also been described in various other pathological conditions such as Barth's syndrome, diabetes and heart failure [16]. Recent shotgun lipidomics studies revealed changes in the molecular cardiolipin content and composition in diabetes [48] and showed that alterations in cardiolipin hydrolysis and remodeling are present at the earliest stages of diabetes [100].

Also for atherosclerosis, a chronic vascular disease in which lipid-containing plaques build up in the walls of arteries, several lipidomics studies have led to useful diagnostic tools and further characterization of the disease [101]. Nutritional aspects influence the development of atherosclerosis, but the atherogenicity of a diet can even depend on its isobaric lipid composition. Lard and tallow both contain approximately the same amount of palmitic acid (C16:0), but results suggest that the amount of palmitic acid at the sn-2 position of triglycerides can influence their atherogenicity [102].

More recently, data on ceramide-synthase-deficient mice have underscored the importance of a specific lipid chain length on membrane homeostasis because sphingolipid acyl chain length plays a role in insulin receptor signaling, brain function, protection against hepatocarcinoma and skin barrier function [103-106].

Also the pathophysiology of obesity is tightly connected to alterations in lipid metabolism. Both genetic and environmental factors are involved in the etiology of obesity, which has reached epidemic proportions. Several lipidomics studies have focused on the different aspects of metabolic syndrome and obesity and have provided a deeper understanding of the metabolic changes observed in those pathologies [107, 108]. In this aspect, lipidomics may also serve as a powerful tool for the study of lipoprotein metabolism [109]. Recent lipidomic studies on the lipid composition of lipoprotein particles have offered new insight into their metabolism and function in health and disease [110, 111]. Lipoprotein particles are large aggregates of lipids and proteins that allow the transport of lipids in an aqueous environment such as biological fluids. A detailed characterization of their molecular composition will help to identify novel biomarkers in lipid metabolism.

Lipid storage disorders such as Niemann-Pick type C disease comprise a group of inherited metabolic disorders that are part of the family of lysosomal storage diseases. Another group of inherited metabolic disorders that led to abnormal levels of certain lipid metabolites are peroxisomal disorders [112]. Peroxisomal disorders are characterized by defects in peroxisome function or biogenesis, which lead to the accumulation of certain lipid metabolites such as very-long-chain fatty acids and to defects in ether lipid biosynthesis. Mammalian cells contain in addition to diacyl glycerophospholipids significant amounts of plasmalogens, which possess a vinyl ether moiety at the sn-1 position of the glycerol backbone. The molecular functions of plasmalogens are still not fully understood, but the high susceptibility of the vinyl ether bond to oxidative damage indicates a possible antioxidative role of this lipid class [113]. Plasmalogens are found in numerous tissues with the highest amount in the nervous system where they make up to 70% of the myelin sheath phosphoethanolamines [114]. Lysosomal storage and peroxisomal disorders show the devastating consequences of a grossly unbalanced lipid metabolism. But also less pronounced defects in lipid homeostasis e.g. caused by type 4 P-type ATPase deficiencies are linked to a various array of diseases such as liver disease, obesity, diabetes, hearing loss, neurological deficits, immune deficiency and reduced fertility [115]. Taken together, this shows that most diseases have a lipid component and underscores the importance of lipid research in the context of human pathologies.

Pathway analysis of lipidomics data

Although it is impossible to access the actual number of lipids per cell it was proposed that eukaryotic cells contain up to 200 000 different molecular species [14]. New approaches are therefore needed to assess the significance and function of this enormous lipid complexity. In order to understand how lipid homeostasis regulates biological processes it will be necessary to combine lipidomics with already existing genetic tools and molecular techniques. To incorporate function van Meer proposed a term ‘cellular lipidomics’ [9, 66]. Cellular lipidomics aims to determine not only which lipids are present, but also the concentration of each lipid at each specific intracellular location in time and its lipid interaction partners. However, the question remains how cells globally sense and regulate their lipid homeostasis. Therefore, we would like to propose a term called ‘systematic lipidomics’ which combines pathway analysis with MS-based lipidomics. Systematic lipidomics aims to investigate how the perturbance of one pathway influences the lipidome of cells (Fig. 2). Pathway targeting in the context of lipid analysis will lead to an increase or decrease in certain lipid metabolites, either directly or indirectly through a cascade of signaling events. There are two approaches imaginable, in the hypothesis-driven approach, prior knowledge is used to select a pathway that is hypothesized to have a lipid phenotype. In the data-driven approach, large-scale sets of biological pathways are selected to uncover novel pathway–lipid phenotype relationships (Fig. 2). This type of lipidomics approach could be supported by techniques such as pathway analysis of genomic data or metabolite pathway enrichment analysis [116, 117]. One example of such a systematic lipidomics approach is a yeast study on mutants of the ergosterol biosynthesis pathway, which showed remarkable effects on sphingolipid homeostasis [118]. Another more targeted MS-based example is the study on mutants of the GPtdIns anchor biosynthesis pathway [78].

Figure 2.

Systematic lipidomics – pathway analysis of lipidomics data. In the hypothesis-driven approach (red) prior knowledge is used to select a pathway that is expected to have a lipid phenotype. In the data-driven approach (blue), large-scale sets of biological pathways are selected to uncover novel pathway–lipid phenotype relationships. In both cases, lipid homoeostasis is assessed for each mutant and data interpretation is supported by appropriate analytical tests. Finally, available proteomics and genomic data complement the biological interpretation.

Lipid anchoring of proteins to the outer leaflet of the plasma membrane is essential for cellular function and development [119]. One prominent lipid anchor is a complex glycolipid called glycosylphosphatidylinositol (GPI). After biosynthesis, the GPI anchor is attached post translationally to the newly synthesized C-terminus of certain eukaryotic proteins. In mammalian cells, at least three organelles, endoplasmic reticulum, Golgi and peroxisomes are involved in the biosynthesis and remodeling of the GPI anchor. After protein attachment, the GPI anchor undergoes complex lipid and glycan remodeling that begins in the endoplasmic reticulum and is continued in the Golgi. Glycan remodeling is crucial for sorting of GPI-anchored proteins into endoplasmic reticulum exit sites and their subsequent endoplasmic reticulum to Golgi transport [120]. From the Golgi, GPI-anchored proteins are then transported to the plasma membrane where they associate preferentially with glycosphingolipids and cholesterol. Lipid remodeling is likely to be important for this association because unremodeled GPI-anchored proteins, which carry unsaturated fatty acids, are no longer enriched in detergent-resistant membrane fractions [121]. We systematically analyzed the lipidome of cells defective at different steps along the GPI anchor biosynthesis pathway and found various effects on cellular sphingolipid levels. GPI-anchor-deficient cells that accumulate short truncated GPI anchor intermediates in the endoplasmic reticulum display a decrease in very-long-chain sphingolipid levels, predominantly GlcCer and GM3, which suggests that very-long-chain ceramides and GPI anchor molecules might be cotransported. Cells that have a defect in GPI anchor remodeling showed a general increase in GlcCer levels, which was due to endoplasmic reticulum stress. By contrast, cells that generate no GPI-anchored proteins but have complete free GPI anchor molecules had unchanged sphingolipid levels, indicating that the effect was not due to the absence of GPI-anchored proteins. This example shows how pathway analysis combined with MS-based lipidomics helps to elucidate the effect of one metabolic pathway, in this case GPI anchor biosynthesis, on the lipidome of mammalian cells. Those results are promising and encourage further studies such as large-scale screenings related to lipid metabolism.

Acknowledgements

This work was supported by the Swiss National Science Foundation and the Swiss SystemsX.ch initiative, evaluated by the Swiss National Science Foundation.

Ancillary