- 1.1 Introduction
- 1.2 Most Common Experimental Methods in Lipidomics
- 1.2.1 DI and Shotgun Methods
- 1.2.2 LC-MS Methods
- 1.2.3 GC-MS Methods
- 1.3 Emerging Experimental Methods in Lipidomics
- 1.3.1 Low-flow-rate-LC
- 1.3.2 IM-MS
- 1.3.3 MSI
- 1.4 Sample Preparation and Derivatisation
- 1.4.1 Liquid and Solid Phase Extractions
- 1.4.2 Liquid Extraction for Surface Analysis (LESA)
- 1.4.3 LESA for MSI
- 1.4.4 Derivatisation for Lipidomics
- 1.5 Databases Searches, Nomenclature and Reporting
- 1.6 Conclusions
- Conflict of Interest Statement
- References
CHAPTER 1: Lipidomics Basics
-
Published:13 Jan 2020
-
Special Collection: 2020 ebook collection
W. J. Griffiths, E. Yutuc, D. Davies, A. Dickson, R. Angelini, D. El Assad, ... Y. Wang, in Lipidomics: Current and Emerging Techniques, ed. W. Griffiths and Y. Wang, The Royal Society of Chemistry, 2020, pp. 1-24.
Download citation file:
Lipidomics can be regarded as the quantitative profiling of the entire lipid composition of a defined system, be that a cell, tissue, biofluid or intact organism. Lipidomics is a descendent of what was previously known as “metabolite profiling” and is a relative of the related “omic” disciplines of genomics, transcriptomics and proteomics. Lipidomics can be regarded as a sub-division of metabolomics. In this chapter we will discuss the current methodologies popular in lipidomics, highlighting gas-chromatography mass spectrometry (GC-MS), liquid-chromatography MS (LC-MS) and direct-infusion, also known as “shotgun” MS (DI-MS). We will also introduce newer methods including low-flow-rate-LC MS, ion mobility MS (IM-MS) and MS imaging (MSI), described in detail in later chapters. Chemical derivatisation is not a new idea to lipidomics; however, its use goes in and out of vogue, hence we also introduce some of its benefits and disadvantages here and in a later chapter.
1.1 Introduction
The term lipidomics was introduced at around the turn of the century.1,2 It refers to the quantitative profiling of the lipidome, which is the lipid composition of a system. The lipidome is a subdivision of the metabolome which includes all the small molecules, usually thought of as having a mass < 1500, in a system (Figure 1.1). Both lipidomics and metabolomics are modern versions of “metabolite profiling” introduced as far back as the early 1970s.3 Currently, the methods of lipidomics are dominated by the use of mass spectrometry (MS) although nuclear magnetic resonance spectrometry (NMR) can also be used.4 The most popular ionisation method for MS-based lipidomics is electrospray-ionisation (ESI) which is readily linked with liquid-chromatography (LC) separations and with direct-infusion (DI) or “shotgun” sample-introduction formats. Other ionisation methods include desorption-ESI (DESI)5 and matrix-assisted laser/desorption ionisation (MALDI),6 both widely used for MS imaging (MSI) and electron ionisation (EI) incorporated in most gas-chromatography (GC)-MS methods.7
The lipidome is extremely complex with eight different categories, six of which dominate the mammalian lipidome (Figure 1.2)8 and lipidomics' methods are highly dependent on the nature of the problem to be solved. If hundreds, or even thousands of samples are to be analysed in a study, then it is most practical to perform only a minimum of sample pre-treatment and to perform rapid MS analysis using DI-MS (e.g. ref. 9 and 10). On the other hand, if the object of the study is to determine the most comprehensive measurement of a lipidome, with only a few samples to be analysed, then extensive sample preparation protocols followed by multiple analysis may be the best route.11,12 The reader is directed to the Lipid Maps website, specifically http://www.lipidmaps.org/resources/protocols/, for protocols for the analysis of different lipid categories.
1.2 Most Common Experimental Methods in Lipidomics
1.2.1 DI and Shotgun Methods
The terms “shotgun” and abbreviation DI are essentially interchangeable according to current lipidomic terminology. Simply, a sample is infused into the ESI source of a MS in the absence of chromatography. This may be achieved by injecting a sample from an LC-autosampler, in the absence of a column, via a syringe pump, via a nano-ESI needle or via a chip-based nano-ESI nozzle as commercialised by Advion Inc. The method then relies on the MS to perform the identification. This, as initially exploited by Han and co-workers, may be realised by utilising multiple consecutive scan modes on a tandem quadrupole MS (Figure 1.3)1,2,13 or using high-resolution scans often on Orbitrap instruments9,14 using both positive- and negative-ion modes. Lipids fragment in tandem mass spectrometry (MS/MS) and multistage fragmentation (MSn) experiments according to well established ion-chemistry rules, allowing the prediction of fragment ion spectra, a significant advantage for compound identification in DI-MS(MS/MS) methods.15–17 The interested reader is directed to the text “Tandem Mass Spectrometry of Lipids: Molecular Analysis of Complex Lipids” by Dr Robert Murphy, one of the pioneers of modern lipid analysis.18 The major advantages of DI-MS methods are simplicity in sample handling and speed of sample analysis. The disadvantage is that DI-MS spectra are dominated by the most abundant or easily ionised lipids. This limitation can be overcome by performing dedicated sample preparation protocols targeted at one class of analytes at a time19 or by performing specific derivatisations targeting a defined functional group.20
(A) Important MS/MS scan modes utilised in DI-MS/MS studies performed on tandem quadrupole instruments. See Chapter 10. (B) Lipids fragment according to well established rules of ion chemistry. Abbreviations: PC, phosphocholine (glycerolphosphocholine); PE, phosphoethanolamine (glycerophosphoethanolamine); PI, phosphoinositol (glycerophosphoinostol) and PS, phosphoserine (glycerophosphoserine).
(A) Important MS/MS scan modes utilised in DI-MS/MS studies performed on tandem quadrupole instruments. See Chapter 10. (B) Lipids fragment according to well established rules of ion chemistry. Abbreviations: PC, phosphocholine (glycerolphosphocholine); PE, phosphoethanolamine (glycerophosphoethanolamine); PI, phosphoinositol (glycerophosphoinostol) and PS, phosphoserine (glycerophosphoserine).
1.2.2 LC-MS Methods
The advantage of utilising LC is that it can provide molecular separation, the penalty of its use is that it consumes time. LC-MS for lipidomics is often performed using reversed phase (RP) or hydrophilic interaction liquid chromatography (HILIC) columns that separate according to chain length and head group polarity, respectively.21,22 The two stationary phases can be utilised in a two-dimensional-LC (2D-LC) format, with the first separation according to the hydrophobic portion of the molecules (RP column) followed by separation of lipid classes based on polarity (HILIC column).23 LC separations with RP and HILIC columns may take up to 1 hour, although application of ultra-performance-LC (UPLC), with chromatography performed at higher pressures, can reduce the run time (Figure 1.4). Both RP and HILIC utilise solvent systems compatible with ESI. This is not generally so with normal phase (NP) chromatography where alternative ionisations methods, e.g. atmospheric pressure chemical ionisation (APCI) may be more appropriate.
(A) UPLC-MS separation of lipids in the NIST SRM1950 of human plasma in (A) positive-ion mode and (B) negative-ion mode. (C) Bile acid analysis in the negative-ion mode after solid phase extraction. Data generated on an Acquity UPLC, with a 2.1 mm × 100 mm C18 column exploiting a Xevo-G2-XS quadrupole time-of-flight (QTOF) MS (Waters Corporation). In (A) and (B) the reproducibility of chromatography is demonstrated by three overlaid chromatograms. Abbreviations: PC, phosphocholine (glycerolphosphocholine); SM, sphingomyelin (phosphosphingolipid); PG, phosphoglycerol (glycerophosphoglycerol); PE, phosphoethanolamine (glycerophosphoethanolamine); DG, diacylglycerol (diradylglycerol); CE, cholesterol ester (sterol ester); TG, triacylglycerol (triradylglycerol); PI, phosphoinositol (glycerophosphoinostol) and PS, phosphoserine (glycerophosphoserine). The NIST SRM1950 is used as a standard reference material (SRM) in lipidomics.12,24,25
(A) UPLC-MS separation of lipids in the NIST SRM1950 of human plasma in (A) positive-ion mode and (B) negative-ion mode. (C) Bile acid analysis in the negative-ion mode after solid phase extraction. Data generated on an Acquity UPLC, with a 2.1 mm × 100 mm C18 column exploiting a Xevo-G2-XS quadrupole time-of-flight (QTOF) MS (Waters Corporation). In (A) and (B) the reproducibility of chromatography is demonstrated by three overlaid chromatograms. Abbreviations: PC, phosphocholine (glycerolphosphocholine); SM, sphingomyelin (phosphosphingolipid); PG, phosphoglycerol (glycerophosphoglycerol); PE, phosphoethanolamine (glycerophosphoethanolamine); DG, diacylglycerol (diradylglycerol); CE, cholesterol ester (sterol ester); TG, triacylglycerol (triradylglycerol); PI, phosphoinositol (glycerophosphoinostol) and PS, phosphoserine (glycerophosphoserine). The NIST SRM1950 is used as a standard reference material (SRM) in lipidomics.12,24,25
1.2.3 GC-MS Methods
GC-MS has been used for the profiling of lipids for decades.26–28 It is still the ideal method for measuring underivatised short chain fatty acids and is also widely used for analysing fatty acids as their methyl esters (Figure 1.5).29,30 With the exception of small volatile compounds, a disadvantage of using GC-MS in lipidomics is the requirement for derivatisation to enhance volatility and stability. There is a huge volume of literature describing derivatisation reactions appropriate for lipids. The interested reader is directed to the classical books by Blau and King,31 Blau and Halket32 and more recent reviews by Halket and Zaikin.33 GC has the advantage over LC of providing a greater number of theoretical plates, hence superior resolution. GC-MS is extensively exploited for steroid and sterol analysis where robust and well documented analytical protocols are in place (Figure 1.5B).34–36
(A) Analysis of short-chain fatty acids in the absence of derivatisation by GC-MS. The column was a DB-WAXetr, the run time was 15 min and the temperature gradient was as follows: initial temperature 50 °C, raised to 150 °C at 15 °C min−1, then to 170 °C at 5 °C min−1 and to 250 °C at 50 °C min−1. Abbreviations: AA, acetic acid (ethanoic acid); PA, propionic acid (propanoic acid); iBA, isobutyric acid (2-methylpropanoic acid); BA, butyric acid (butanoic acid); iV, isovalaric acid (3-methylbutanoic acid); VA, valaric acid (pentanoic acid); 4-MVA, 4-methylvalaric acid (4-methylpentanoic acid). (B) Analysis of cholesterol precursors derivatised to trimethylsilane (TMS) ethers. Lathosterol, 24,25-dihydrolanosterol and lanosterol are challenging to analyse by LC-MS, but readily separated by GC-MS. Total ion chromatograms (TIC) recorded on a GC-qTOF (Agilent).
(A) Analysis of short-chain fatty acids in the absence of derivatisation by GC-MS. The column was a DB-WAXetr, the run time was 15 min and the temperature gradient was as follows: initial temperature 50 °C, raised to 150 °C at 15 °C min−1, then to 170 °C at 5 °C min−1 and to 250 °C at 50 °C min−1. Abbreviations: AA, acetic acid (ethanoic acid); PA, propionic acid (propanoic acid); iBA, isobutyric acid (2-methylpropanoic acid); BA, butyric acid (butanoic acid); iV, isovalaric acid (3-methylbutanoic acid); VA, valaric acid (pentanoic acid); 4-MVA, 4-methylvalaric acid (4-methylpentanoic acid). (B) Analysis of cholesterol precursors derivatised to trimethylsilane (TMS) ethers. Lathosterol, 24,25-dihydrolanosterol and lanosterol are challenging to analyse by LC-MS, but readily separated by GC-MS. Total ion chromatograms (TIC) recorded on a GC-qTOF (Agilent).
1.3 Emerging Experimental Methods in Lipidomics
While DI-MS and LC-MS dominate the current lipidomic protocols a number of variants of these methods, and also new separation and data acquisition protocols, have the potential to become mainstream. These include the use of micro- and capillary-LC, ion mobility-MS (IM-MS) and MS imaging (MSI). There is also potential for the application for new derivatisation strategies.
1.3.1 Low-flow-rate-LC
Low-flow-rate-LC has been widely used in the proteomics field. The advantage provided by a reduced solvent content in a chromatographic peak is enhanced sensitivity when used with a concentration dependent ionisation method such as ESI. As discussed by Roberg-Larsen and Wilson in Chapter 3, low-flow-rate systems range from microbore columns (∼1–0.5 mm i.d., flow-rate ∼50–10 µL min−1), to capillary columns (∼0.5–0.2 mm i.d., flow-rate ∼20–2 µL) down to nano-columns (∼0.1–0.05 mm i.d., flow-rate ∼0.5–0.05 µL min−1). The separation of derivatised oxysterols extracted from plasma achieved on a 0.1 mm column at a flow-rate of 0.6 µL min−1 is shown in Figure 1.6. The oxysterols were derivatised with the EADSA method discussed in Section 1.4.4. While low-flow-rate-LC-ESI-MS offers advantages of sensitivity, potential disadvantages include elongated run times, reduced robustness and problems of carry over between injections. However, the inclusion of a trap column prior to the analytical column, allowing sample pre-concentration and washing when switched out of line, can speed up sample loading and minimise contamination of the analytical column. The trap column is switched in-line with the analytical column after the washing phase is complete and the analyte eluted from trap to analytical column and separated prior to MS analysis.37
Separation of different oxysterols from a plasma sample following EADSA derivatisation. The column was 100 mm × 0.1 mm (C18, 3 µm 100 Å) and a gradient delivered by the binary nanopump at 0.6 µL min−1.
Separation of different oxysterols from a plasma sample following EADSA derivatisation. The column was 100 mm × 0.1 mm (C18, 3 µm 100 Å) and a gradient delivered by the binary nanopump at 0.6 µL min−1.
1.3.2 IM-MS
IM separations provide an extra dimension to the MS analysis of lipids. This subject is discussed in detail by Paglia and Astarita in Chapter 4. IM separation is based on the time an ion takes to cross an IM cell containing buffer gas. This time will be dependent on the shape and size of the ion and the nature of the buffer gas contained within the IM cell. The more collisions an ion makes with the buffer gas the longer the drift time.38 IM separation is made on the millisecond timescale which makes it ideal for combination with fast scanning MS where detection can be at the microsecond scale. There are two main advantages of incorporating IM-MS in lipidomics experiments. Firstly, IM provides a separation dimension, potentially separating isomeric compounds indistinguishable by MS alone, secondly, the collisional cross section (CCS), a physicochemical property which is related to structure, can be used to facilitate compound identification.38,39 CCS can be calculated or experimentally determined for authentic compounds. IM separations can add an extra dimension to LC-MS experiments, being complementary to LC where separations are on the second timescale,40 but may offer most advantages for lipidomics experiments using MALDI, which is seldom coupled to chromatography, as the method of ionisation.41 This is illustrated in Figure 1.7 where a MALDI mass spectrum is shown alongside the accompanying IM separation in the analysis of a section of brain tissue. This combination can offer considerable advantages to MALDI-MSI experiments where chromatographic separation is not feasible.
MALDI mass spectrum of a mouse brain section which has been derivatised with the EADSA method to highlight cholesterol (m/z 518.4) and other sterols. (A) MALDI-MS, the most intense peak is at m/z 518.4. (B) IM separation showing the distribution of ions according to drift time and m/z. (C) MALDI-MSI of cholesterol exploiting IM separation for a sagittal section of mouse brain. Pixel size 45 µm. Signal normalised to sprayed-on [2H7] cholesterol quantitative standard. (D) Optical image of the brain slice. Note the cholesterol peak in (A) is so intense the detector is overloaded with the consequence of a loss of mass accuracy. The identity of the signal at m/z 518.4 was confirmed as cholesterol by MS/MS. The signal was attenuated in (C) to allow accurate profiling of cholesterol. Data was generated on a MALDI Synapt G2 Si (Waters Corporation).
MALDI mass spectrum of a mouse brain section which has been derivatised with the EADSA method to highlight cholesterol (m/z 518.4) and other sterols. (A) MALDI-MS, the most intense peak is at m/z 518.4. (B) IM separation showing the distribution of ions according to drift time and m/z. (C) MALDI-MSI of cholesterol exploiting IM separation for a sagittal section of mouse brain. Pixel size 45 µm. Signal normalised to sprayed-on [2H7] cholesterol quantitative standard. (D) Optical image of the brain slice. Note the cholesterol peak in (A) is so intense the detector is overloaded with the consequence of a loss of mass accuracy. The identity of the signal at m/z 518.4 was confirmed as cholesterol by MS/MS. The signal was attenuated in (C) to allow accurate profiling of cholesterol. Data was generated on a MALDI Synapt G2 Si (Waters Corporation).
1.3.3 MSI
MSI has great potential in lipidomics. Lipids tend to ionise particularly well in MALDI-MSI experiments allowing their location in different tissues.6 Khan and Andrew discuss MSI for lipidomics in detail in Chapter 5 and only a brief overview will be given here. Most MSI studies in lipidomics use MALDI as the ionisation method. MALDI may be performed under vacuum or at atmospheric pressure (AP); both regimes have their own advantages and disadvantages.42 A concern when using MALDI-MSI is that application of the matrix may result in lipid delocalisation and different matrix application methods and types may be most appropriate for different tissues. Nevertheless, MALDI-MSI of EADSA derivatised cholesterol (Figure 1.7) and of phospholipids (Figure 1.8) clearly reveal different structural features in mouse brain.
AP-MALDI-MSI of a sagittal section of mouse brain. Three different m/z values are imaged. By searching a window of m/z ± 0.005 against the Lipid Maps “LMSDB” database, a database restricted to biologically relevant lipids (https://www.lipidmaps.org/resources/tools/bulk_structure_searches.php?database=LMSD), m/z = 772.5274 reveals ten hits, reduced to four if only [M + H]+, [M + Na]+ and [M + K]+ hits are considered. As the brain tissue was exposed to a buffer containing K+ the hits corresponding to [M + K]+ ions of PC(32:0) and PE(35:0) are most likely. Odd numbered fatty acids are rare, so “biological intelligence” leads to an identification of PC(32:0). Data was generated on an Orbitrap Elite (Thermo Fisher Scientific) using an AP-MALDI source (MALDI-UHR from MassTech). Mass resolution was 60 000 at m/z 400, lateral resolution was 30 µm.
AP-MALDI-MSI of a sagittal section of mouse brain. Three different m/z values are imaged. By searching a window of m/z ± 0.005 against the Lipid Maps “LMSDB” database, a database restricted to biologically relevant lipids (https://www.lipidmaps.org/resources/tools/bulk_structure_searches.php?database=LMSD), m/z = 772.5274 reveals ten hits, reduced to four if only [M + H]+, [M + Na]+ and [M + K]+ hits are considered. As the brain tissue was exposed to a buffer containing K+ the hits corresponding to [M + K]+ ions of PC(32:0) and PE(35:0) are most likely. Odd numbered fatty acids are rare, so “biological intelligence” leads to an identification of PC(32:0). Data was generated on an Orbitrap Elite (Thermo Fisher Scientific) using an AP-MALDI source (MALDI-UHR from MassTech). Mass resolution was 60 000 at m/z 400, lateral resolution was 30 µm.
DESI represents an alternative ionisation method to MALDI and has the advantage that no matrix is required; however, the spatial resolution of routine DESI-MSI methods are lower than those that can be achieved by MALDI-MSI. DESI images of two phospholipid in mouse brain are shown in Figure 1.9.
DESI-MSI images of phospholipids from a sagittal section of mouse brain. m/z 810.5975 is annotated as the Na+ adduct of PC(36:1). PC(36:1), PE(39:1), PC(O-36:2(OH)) and PC(P-36:1(OH)) have an identical chemical formula and m/z. A search of m/z 810.5975 ± 0.005 against the Lipid Maps “Comp_DB” computationally generated database (http://www.lipidmaps.org/resources/tools/bulk_structure_searches.php?database=COMP_DB) reveals these four structures (as Na+ adducts) with a m/z error of 0.0008, and ten other structures with a greater error but within the window of m/z ± 0.005. However, if the search is restricted to the Lipid Maps “LMSDB”, a database restricted to biologically relevant lipids (http://www.lipidmaps.org/resources/tools/bulk_structure_searches.php?database=LMSD), the result is six hits within the window m/z ± 0.005, those with least m/z error being PC(36:1) and PE(39:1). Odd numbered fatty acid chains are not common, hence the PC(36:1) structure is more likely. Data is generated with a Synapt G2-Si mass spectrometer with a DESI source, pixel size 50 µm (Waters Corporation).
DESI-MSI images of phospholipids from a sagittal section of mouse brain. m/z 810.5975 is annotated as the Na+ adduct of PC(36:1). PC(36:1), PE(39:1), PC(O-36:2(OH)) and PC(P-36:1(OH)) have an identical chemical formula and m/z. A search of m/z 810.5975 ± 0.005 against the Lipid Maps “Comp_DB” computationally generated database (http://www.lipidmaps.org/resources/tools/bulk_structure_searches.php?database=COMP_DB) reveals these four structures (as Na+ adducts) with a m/z error of 0.0008, and ten other structures with a greater error but within the window of m/z ± 0.005. However, if the search is restricted to the Lipid Maps “LMSDB”, a database restricted to biologically relevant lipids (http://www.lipidmaps.org/resources/tools/bulk_structure_searches.php?database=LMSD), the result is six hits within the window m/z ± 0.005, those with least m/z error being PC(36:1) and PE(39:1). Odd numbered fatty acid chains are not common, hence the PC(36:1) structure is more likely. Data is generated with a Synapt G2-Si mass spectrometer with a DESI source, pixel size 50 µm (Waters Corporation).
Both MALDI-MSI and DESI-MSI can be extended to include MS/MS or MSn, the penalty of the added specificity this provides is a loss in ion current. We have investigated a combination of AP-MALDI-MS3 for the analysis of sterols in mouse brain. To enhance the signal, we exploited EADSA derivatisation. Besides observing a major ion signal corresponding to derivatised cholesterol, we also observed the dehydrogenated cholesterol derivative, presumably formed in the MALDI process. The MS3 image for the fragmentation of this ion from a sagittal section of mouse brain is shown in Figure 1.10, reflecting the distribution of cholesterol in the tissue.
AP-MALDI–MS3 images of EADSA derivatised cholesterol observed as the dehydrogenated ion. (A) Image for the TIC for the MS3 transition [M − 2]+→[M − 2 − Py]+→, where Py is the pyridine derivatisation group (see Section 1.5). (B) Image of the ion current for the MS3 transition [M − 2]+→[M − 2 − Py]+→m/z 163 ± 0.3. (C) [M − 2]+→[M − 2 − Py]+→ spectrum from a point in the corpus callosum. (D) Fragmentation pathway generating the ion at m/z 163. Data is generated by AP-MALDI (MassTech) on an Orbitrap Elite mass spectrometer (Thermo Fisher Scientific).
AP-MALDI–MS3 images of EADSA derivatised cholesterol observed as the dehydrogenated ion. (A) Image for the TIC for the MS3 transition [M − 2]+→[M − 2 − Py]+→, where Py is the pyridine derivatisation group (see Section 1.5). (B) Image of the ion current for the MS3 transition [M − 2]+→[M − 2 − Py]+→m/z 163 ± 0.3. (C) [M − 2]+→[M − 2 − Py]+→ spectrum from a point in the corpus callosum. (D) Fragmentation pathway generating the ion at m/z 163. Data is generated by AP-MALDI (MassTech) on an Orbitrap Elite mass spectrometer (Thermo Fisher Scientific).
1.4 Sample Preparation and Derivatisation
A general point to consider during sample preparation is that a method validated for one tissue may not be directly transferable for use on a different tissue in the absence of further validation. Other considerations during lipid extractions are the glass-ware and plastic-ware to be used, the purity of solvents and the batch to batch variation of standard compounds, solid phase extraction (SPE) columns and of solvents and plastic/glass-ware themselves.
1.4.1 Liquid and Solid Phase Extractions
Traditionally, lipids have been extracted into chloroform/methanol solvents using protocols based on the classic publications by Folch for extractions from tissue, and by Bligh and Dyer applicable to tissues and fluids.43,44 Both methods rely on the generation of a two-phase system where lipids are partitioned into the organic phase. Readers are recommended to consult the excellent website Cyberlipid http://cyberlipid.gerli.com/techniques-of-analysis/extraction-handling-of-extracts/ for extraction protocols. An alternative extraction solvent gaining popularity in lipidomics studies is methyl tert-butyl ether (MTBE).45 Extractions can also be made into alcohol-based solvents, and this is the basis of the BUME method where lipids are first extracted into a single phase by the addition of butanol and methanol, followed by a two-phase extraction into heptane/ethyl acetate using 1% acetic acid as a buffer.46,47
For more polar lipids SPE is often exploited.34,48–50 This is particularly the case for extractions of polar lipids from liquid samples.
1.4.2 Liquid Extraction for Surface Analysis (LESA)
Liquid extraction for surface analysis (LESA) is an extension of liquid micro-junction surface sampling pioneered by Kertesz and Van Berkel and commercialised by Advion Inc.51–54 The concept behind LESA is to form a liquid micro-junction between a solvent droplet held at the tip of a capillary, or pipette tip, positioned just above a surface, and the surface itself. Analyte at the surface is extracted by the solvent which after a suitable period of time is aspirated back to the capillary, or pipette, and subsequently injected via ESI into the MS. Most studies are performed with aqueous-based solvents where the surface tension is sufficient to stop the solvent droplet “falling” from the pipette tip onto the surface. However, for lipid analysis where solvents of high organic content (with low surface tension) are required this may be problematic. However, successful analysis of lipids in tissues, on TLC plates and from dried blood spots has been performed using LESA.55,56 Almeida et al. have developed a variant of LESA that they have termed pressurised-LESA or PLESA. Here the pipette tip comes into contact with the tissue surface, the solvent is dispensed towards the tip, surface extraction occurs and the solvent is aspirated back up the pipette tip and injected into the ESI-MS. The extraction surface area is limited to the inner diameter of the pipette tip and the method is applicable to organic solvents, and hence is appropriate for lipidomics analysis.57
1.4.3 LESA for MSI
Since inception, liquid micro-junction surface sampling and LESA have had the potential for application to image analysis;51,58 a major advantage of this sampling format is that it can be linked to HPLC with the capacity to simplify mass spectra by chromatographically separating the extracted sample.59,60 LESA-MSI has inferior spatial resolution compared to either DESI-MSI or MALDI-MSI (pixel size 400 µm cf. 30 µm in Figure 1.8 and 50 µm in Figure 1.9) but with the incorporation of an HPLC column can differentiate isomers. We have found this particularly useful for MSI of oxysterols in brain tissue, where there is the possibility that multiple isomers exist.61 The LESA-LC-ESI-MSI system currently utilised in our laboratory in Swansea for analysis of sterols derivatised by the EADSA method from tissue is shown in Figure 1.11. Derivatised sterols are extracted from tissue into an aqueous methanol solvent, loaded onto a trap column, the excess derivatisation reagent is washed away, and sterols eluted onto an analytical column for separation and ESI-MSI analysis. The extraction probe is then robotically moved from one pixel to another and an image is generated.
Schematic of the LESA-LC-ESI-MSI set-up in Swansea. The LESA system is commercialised as the TriVersa NanoMate by Advion Inc.
Schematic of the LESA-LC-ESI-MSI set-up in Swansea. The LESA system is commercialised as the TriVersa NanoMate by Advion Inc.
1.4.4 Derivatisation for Lipidomics
This subject is discussed in detail in Chapter 6 and only a few examples will be introduced here. As mentioned in Section 1.2.3, there are numerous reviews and texts on derivatisation for GC that will not be covered here, instead we will introduce methods applicable for DI-MS or LC-MS.
The idea behind derivatisation is to either enhance the ion current of the target analyte or to encourage its fragmentation in a predictable manner allowing structural determination. Preferably both goals are achieved with the same derivative. The derivatisation reaction will ideally go to completion with a minimum of side-products, the excess derivatisation reagent will be easily removed and be commercially available, cheap, and the derivatisation reaction will be safe and simple, i.e. click chemistry. With respect to MALDI-MS, if the derivatisation reagent can also act as the matrix this is a further benefit. Needless to say, there is no ideal derivatisation reagent, but there are many very good ones.
In Chapter 7, Ma and Xia introduce the Paternò–Büchi reaction for locating double bonds in lipids by MS/MS. The reaction is based on the photochemical cyclisation reaction between an alkene and carbonyl to give a 4-membered oxetane ring (see Chapter 7, Scheme 7.1), which acts as a marker for the location of the double bond, whose position can be determined by subsequent MS/MS.62 This reaction is often used with acetone and can be performed offline or within the ESI source. Lipids containing carbon–carbon double bond(s) can be derivatised in this way, ionised and when subjected to MS/MS their spectra reveal the original location of the double bond(s) but not their stereochemistry.
In Chapter 8 Liang and colleagues introduce the idea of using ozonolysis for determining carbon–carbon double bond locations in lipids. The reaction between ozone and an alkene will generate an ozonide at the original location of the double bond which upon MS/MS will reveal the position of the double bond (Figure 8.10).63 When used after silver-ion chromatography, which can separate isomers according to cis and trans geometry, and before MS/MS, the derivatisation method becomes particularly powerful for structure determination of unsaturated lipids.64 Ozonolysis can be performed offline prior to the mass spectrometer, or with ozone introduced into a collision cell of the mass spectrometer as in OzID.65
A favoured derivatisation method exploited by us is enzyme-assisted derivatisation for sterol analysis or EADSA.19 EADSA derivatisation is discussed in Chapter 6, in Chapter 3 by Roberg-Larsen and Wilson and by Abdel-Khalik in Chapter 9. The method is targeted at sterols with a 3β-hydroxy or oxo group. In brief, sterols with a 3β-hydroxy-5α-hydrogen or a 3β-hydroxy-5-ene group are specifically oxidised with cholesterol oxidase enzyme66 to their 3-oxo equivalents that are then reacted with the Girard reagent (Figure 1.12).67 The Girard T (GT) reagent has a charged quaternary nitrogen as part of a trimethylammonium group, while Girard P (GP) reagent has a charged quaternary nitrogen as part of a pyridine group. The reaction product is a charge-tagged sterol giving intense ions upon ESI or MALDI and informative MS/MS spectra. Some sterols naturally contain an oxo group, as in corticosterone, see Figure 5.3 in Chapter 5, and do not require any pre-treatment with cholesterol oxidase prior to reaction with the Girard reagent. The EADSA technology has been commercialised by Cayman Chemical Company and isotope-labelled Girard P reagent is available from Avanti Polar Lipids plc.
Derivatisation of sterols with Girard reagent. (A) Treatment of 24S-hydroxycholesterol with cholesterol oxidase followed by GP, [2H5]GP or GT reagent. (B) Treatment of 7-oxocholesterol with GP reagent in the absence of cholesterol oxidase.
Derivatisation of sterols with Girard reagent. (A) Treatment of 24S-hydroxycholesterol with cholesterol oxidase followed by GP, [2H5]GP or GT reagent. (B) Treatment of 7-oxocholesterol with GP reagent in the absence of cholesterol oxidase.
1.5 Databases Searches, Nomenclature and Reporting
In many lipidomic studies, identifications are made by searching measured m/z values with a defined m/z tolerance against a database, such as the Lipid Maps Comp_DB or LMSD databases. The links given in the caption of Figure 1.9 are appropriate to “bulk” searches of many m/z values, if a single m/z value is to be searched the http://www.lipidmaps.org/data/structure/LMSDSearch.php?Mode=SetupTextOntologySearch link is more appropriate. Users should be aware that such searches provide information on the number of carbons and rings/double bonds, but not stereochemistry or geometry. Lipid Maps highly recommend that identifications are validated using authentic lipid standards. This should be a general consideration whichever database is searched. As is noted in the captions of Figures 1.8 and 1.9, a given m/z value measured to an accuracy of 0.005 m/z reveals several hits in a database search. This leads to a reporting conundrum. If hundreds of lipid m/z values are searched for in a lipidomic study, how should all the identifications be reported? How much “biological intelligence” should be applied? A further issue that analysts should be aware of is isobaric interference. Lipids that differ merely in the number of double bonds generate substantial isotopic overlap, particularly resulting from 13C-atoms. The [M + H + 2]+ isotopic peak, where two 12C atoms are replaced by two 13C atoms, for a typical phospholipid, is above 10% of its monoisotopic peak [M + H]+ and overlaps with a species containing one less double bond. For example, [PC 32:1 + H + 2]+ (m/z 734.5605) differs from [PC 32:0 + H]+ (m/z 734.5694) by only 0.009 m/z and can only just be resolved at a resolving power of 120 000. These are issues that the community is now discussing via the Lipidomics Standard Initiative, https://lipidomics-standards-initiative.org/, a “communication platform for discussion and further development of lipidomics standards”.
What nomenclature should be used to report a lipid identification? In very few cases can structures be unequivocally defined by MS, and this uncertainty should be reflected in the nomenclature used for data reporting. Even for the apparently secure MS identification of cholesterol in animal tissue, is the analyst completely sure that cholesterol is being observed not its isomer lathosterol? Our biological knowledge tells us that cholesterol is about 1000 times more abundant than lathosterol in animal tissues and in body fluids, so application of “biological intelligence” makes the identification as cholesterol. But what if unbeknown to the analyst the sample is from a patient suffering from the rare disease lathosterolosis,68 where lathosterol is abundant due to a defect in a later enzyme of the cholesterol synthesis pathway, then “biological intelligence” will have led to a misidentification. Clearly, this uncertainty in identification by MS should be reflected in lipidomic nomenclature. Of course, the definitive identification of cholesterol or lathosterol could be made by using chromatography linked to MS, or by MS/MS and measurement of reference standards.69 In an attempt to link MS identifications to nomenclature at the level of which it is justified, Liebisch et al. introduced a shorthand notation.70 This is complementary to the Lipid Maps comprehensive classification system for lipids, which exactly defines a lipid with all stereochemistry.8 The nomenclature of Liebisch et al. is currently being updated. The problem of cholesterol and lathosterol will be solved in the absence of chromatography, or a definitive MS/MS spectrum, by using the shorthand notation ST 27:1;O, where ST refers to the sterol category, :1, to one double bond and ;O to one oxygen. It is recommended that an explicit lipid name should only be used if a validated method is adopted for its identification and any assumptions made are clearly stated. There is a need for transparency and a trail of decision-making to filter out non-validated information.
Reporting quantitative information is important for comparison between lipidomics studies in different laboratories. Reporting quantities in mole or gram (where molecular weight is known) units is preferable to the use of arbitrary units. The important studies by the Lipid Maps consortium on human plasma use these units.12 Quantification should now become easier with the ready availability of isotope-labelled deuterated standards for different categories of lipids and for individual lipid species. Important moves towards harmonizing lipidomics data with respect to plasm/serum are being made. Wenk and Shevchenko and colleagues are leading this in terms of general lipidomics71 while others are focusing on solving the same problem, but for defined lipid categories.72,73 A final important point to consider for reporting lipidomic data is depositing the data in a repository allowing other investigators to mine the data beyond the interests of the original investigators. Repositories are provided by Metabolomics Work Bench https://www.metabolomicsworkbench.org/ and Metabolights https://www.ebi.ac.uk/metabolights/index.
1.6 Conclusions
The field of lipidomics was reawakened by the creation of the Lipid Maps consortium in the USA in the early 2000s supported by NIH and of LipidomicNet supported by the European Union. Since then, the field has expanded hugely, and Lipid Maps is now supported by the Wellcome Trust in the UK. Lipidomics is now a field with many investigators covering all categories of lipids. Beside resources provided by Lipid Maps http://www.lipidmaps.org/, valuable information can be found on the site of Cyberlipid, http://cyberlipid.gerli.com/, at the Mass Bank of North America http://mona.fiehnlab.ucdavis.edu/, and for those with a passion for sphingolipids by Sphingomap http://www.sphingomap.org/. However, the field is not yet mature, and efforts are especially required to improve reporting standards to avoid misinformation populating the scientific literature. That said, the field offers great rewards to those who are immersed in its study.
Conflict of Interest Statement
Swansea Innovations, a wholly owned company of Swansea University has licenced EADSA technology as described in the patent US9851368B2 to Avanti Polar Lipids Inc and Cayman Chemical Company.
Work in Swansea is supported by the Biotechnology and Biological Sciences Research Council. A.D. is supported by a KESS2 fellowship provide by European Social Funds and the Welsh Government. RA is supported by a Marie Sklodowska-Curie Actions COFUND postdoctoral fellowship funded by European Regional Development Fund. Data in Figure 1.4 was generated by Dr Graham Mullard and data in Figures 1.7 and 1.9 was generated by Dr Wei Rao both of Waters Corporation.