An Overview of Lipidomic Analysis of Triglyceride Molecular Species in Biological Lipid Extracts

Abstract

Triglyceride (TG) is a class of neutral lipids, which functions as an energy storage depot and is important for cellular growth, metabolism, and function. The composition and content of TG molecular species are crucial factors for nutritional aspects in food chemistry, and are directly associated with several diseases including atherosclerosis, diabetes, obesity, stroke, etc. Due to the complexities of aliphatic moieties and their different connection/location to the glycerol backbone in TG molecules, accurate identification of individual TG molecular species and quantitative assessment of TG composition and content are particularly challenging, even at the current stage of lipidomics development. Herein, methods developed for analysis of TG species such as liquid chromatography-mass spectrometry with a variety of columns and different mass spectrometric techniques, shotgun lipidomics approaches, and ion mobility-based analysis are reviewed. Moreover, the potential limitations of the methods are discussed. It is our sincere hope that the overviews and discussions could provide some insights for researchers to select an appropriate approach for TG analysis and could serve as the basis for those who would like to establish a methodology for TG analysis or develop a new method when novel tools become available. Biologically, accurate analysis of TG species with an enabling method should lead us to improving the nutritional quality, revealing the effects of TG on diseases, and uncovering the underlying biochemical mechanisms related to these diseases.

Graphical Abstract

1. INTRODUCTION

1.1. Introduction to Triglycerides.

Triglyceride (TG) species are linear combinations of aliphatic chains covalently connected to the hydroxyl groups of glycerol. The most abundant aliphatic chain is the fatty acyl chain, although alkyl and alkenyl chains are present in low abundance in TG species¹. TG serves as essential energy depots of cells and caloric potential in living organisms. TG species from plant seeds and fruits, or vegetable oils, provide ~25% of dietary calories in developed countries and are widely utilized as industrial feedstocks and renewable biofuels^2–3. On the other hand, multiple lines of evidence have recently demonstrated unambiguously that both the levels and composition of TG species are associated with heart disease, stroke, obesity and diabetes, etc. collectively termed as metabolic syndrome^4–7. Moreover, such associations apparently make the quality of life significantly reduced even if not immediately fatal.

The major de novo synthesis pathway of TG species is via the Kennedy pathway^{3, 8} (see Section 6 for details), initiated from glycerol-3-phosphate (G3P) which is acylated to form 1-acyl lysophosphatidic acid (lysoPA) by glycerol-phosphate acyltransferase (GPAT) in endoplasmic reticulum (ER) and then phosphatidic acid (PA) by acylglycerol-phosphate acyltransferase (AGPAT). PA is dephosphated to yield diglyceride (DG) species through an activity of PA phosphatase (PAP). Finally, DG is reacylated to produce TG through a DG acyltransferase (DGAT) activity⁸. Ether-containing TG species are initiated from the acylation of dihydroxyacetone phosphate (DHAP). The formed acyl DHAP reacts with fatty alcohol species through catalysis by alkyl DHAP synthase to generate alkyl DHAP which is further reduced by an alkyl DHAP reductase activity to yield 1-O-alkylglycero-3-phosphate, an analog of 1-acyl lysoPA. All these reactions occur in peroxisome⁹. The remaining steps in TG synthesis of ether-containing species are identical to those of triacyl TG synthesis as described above.

In addition to the biosynthesis, in mammals, various molecular mechanisms have developed to regulate the TG content and composition in blood in order to correctly deliver fatty acids which are obtained from blood TG molecules to cells through diversified enzymes (e.g., hormone sensitive lipase, hepatic lipase, lipoprotein lipase, etc.) and the transport system (e.g., apolipoproteins, fatty acid transport proteins, and their receptors)¹⁰. In some cases, absorbed fatty acids by cells are intracellularly stored as TG molecular species by esterification to a glycerol backbone in addition to directly serving as energy substrates through fatty acid beta-oxidation in mitochondria. The optimal TG content and composition in blood and cells are also achieved through a futile cycle process¹¹, in which two metabolic pathways run simultaneously in opposite directions and have no overall effect other than to dissipate energy in the form of heat.

1.2. Complexity of TG Species.

As briefly discussed above, multiple acyltransferases are involved in the major pathway of TG biosynthesis. In theory, if these enzymes do not selectively acylated the fatty acyl chains, then all the fatty acyls presented in a system could be integrated into TG molecular species in an acyl-CoA concentration dependent manner, particularly after a futile cycle process. However, as we know, saturated and monounsaturated fatty acyl chains are usually present at the sn-1 position of glycerol in phospholipid species, including PA, whereas polyunsaturated fatty acyl chains are largely located at the sn-2 position^12–13. A combination of these biological factors makes the pool of TG molecules very complex (see below) and the analysis of the true identities of TG molecules very challenge and interesting.

Considering the existence of a large number of potential fatty acyl chains in the majority of biological systems¹⁴, we can predict that the number of molecular species in a TG pool is huge, i.e., N³, where N is the number of fatty acyl chains. Figure 1 illustrates the possible number of all TG species including positional isomers (i.e., regioisomers) and enantiomers (considering the possible chiral center of the glycerol carbon at the sn-2 position) as the number of fatty acyl chains change. Specifically, there only exists one TG species X-X-X with one fatty acyl X; a total of 8 TG isomers could be yielded from two fatty acyls of X and Y; and a total of 27 TG isomers could be counted from three fatty acyls of X, Y, and Z (Figure 1). This leads to a total of N³ possible TG isomers from a number of N fatty acyls (Table 1).

Figure 1.

Schematic illustration of TG isomers yielded from different numbers of fatty acyls of X, Y, and Z. The regioisomers due to different locations of fatty acids are highlighted with red and the enantiomers from their left side are highlighted with green.

TABLE 1.

A List of Potential TG Molecular Species Yielded from a Different Number of Fatty Acyls at Different Structural Levels.

Number of fatty acyls	Number of TG isomers
	Isomers not distinguished	No enantiomers considered	All isomers considered
1	1	1	1
2	4	6	8
3	10	18	27
5	35	75	125
10	220	550	1000
20	1540	4200	8000
N	N2 + C(N, 3)*	N(2N + 1) + 3C(N, 3)	N3

^*C denotes mathematic combination.

Owing to the limitation of different analysis techniques, the number of TG species analyzed are very different. If TG regioisomers and enantiomers (see Figure 1 for structural illustration) are not considered, but isomers generated from a combination of different fatty acyls having an identical number of total carbon atoms and double bonds, then the number of possible TG species are markedly reduced to N² + C(N, 3), where C denotes mathematic combination (Table 1). If regioisomers (excluded optical isomers) are considered, then the number of total TG species could be N(2N +1) + 3C(N, 3) (Table 1). Analysis of these isomers could be achieved to a certain degree using some techniques including tandem mass spectrometry (MS/MS) and ion mobility (IM) devices which will be discussed later in this review.

Accordingly, with the presence of a large number of TG isomers in a biological system, analysis of these TG species has already been the hot, but challenging topic in conventional lipidology¹⁵. Unfortunately, serious limitations are present in those conventional approaches which largely depend on chromatographic separation, as it is greatly demanded to be able to readily determine the structures of intact TG molecules along with their individual quantities occurred in any biological systems including living mammalian and plant samples to a desired degree.

1.3. Importance of Lipidomics Analysis of TG Composition and Content.

Owing to at least the aforementioned causes, quantitative analytical methods which can effectively analyze intact TG molecules in biological systems would become a key and integral component of research which yields novel findings of benefit to human beings in the biochemistry of plants and animals dealing with metabolic syndrome. Thus, an advanced technology for analysis of biological samples becomes very much desired to provide a detail atlas of TG molecules.

The complete TG content and composition could provide the information about nutritional and metabolic history of individual cell or organ, and their anticipated energy storage requirements, particularly if the research is conducted in combination with isotope labelling. It has been shown that changes in TG molecular metabolism play important roles in metabolic syndrome^{5, 8, 16–18}. Analysis of blood TG species could provide a window on diet and disease¹⁹. Moreover, numerous studies have shown that changes in intracellular TG content and composition serve as a potential mediator of diabetic cardiomyopathy^20–21.

Although the total amount of TG mass in different disease states can be readily estimated with a colorimetric assay (assuming the influence of coexisting DG and MG on the estimation is small), the information about altered TG molecular profiles during pathophysiological perturbation and/or progression has recently been caught great attention. For example, lipidomic analysis of TG species in diabetic rat myocardium have shown that marked changes of TG molecular composition could present without any substantial changes in total TG mass²⁰. This kind of findings indicates that changes in TG molecular composition in addition to its total content can also make contributions to some disease states to a certain degree.

With the great efforts of the scientists working in the lipidomics discipline, numerous MS-based methods in both with and without coupling to liquid chromatography (LC) have been developed for analysis of TG species in both content and composition. In this review, these approaches are briefly discussed including their advantages and limitations. Readers who are interested in technique details of these approaches are referred to the original studies cited.

1.4. Mass Spectrometry-Based Lipidomics.

The entire collection of chemically distinct lipid species in a cell, an organ, or a biological system has been referred to as a lipidome²². Lipidomics is the discipline applying analytical chemistry tools and principals to study lipidomes in a large scale and at the levels of intact molecular species in a high throughput manner^23–24. Lipids are metabolites. In this sense, lipidomics is under the umbrella of the general field of “metabolomics”. However, considering the uniqueness and functional specificity of lipids compared to other classes of metabolites, lipidomics is a distinct discipline from metabolomics. The uniqueness of lipids includes, but is not limited to, (1) largely extractable with organic solvents¹⁵, (2) readily form aggregates in all solvents essentially as their concentrations increase²⁵, and (3) structural similarity of the lipid species in a class¹⁴, which can be constructed with countable building blocks²⁶. The similarity of lipid species in a class leads to substantial difficulties for quantitative analysis of these species in their intact forms using LC separation only.

The term “lipidome” first appeared in literature in 2001²². In 2002, Rildfors and Lindblom²⁷ defined “the study of the role played by membrane lipids” as “functional lipidomics”. The lipidomics discipline suddenly began blooming with different definitions^{23, 28}, demonstrations of technologies^{23, 29}, and biological applications^{23, 30–31} in 2003, largely due to the development in MS. Since then, the entire field has emerged and become greatly expanded³².

Depending on whether a chromatography-based separation device is coupled to a mass spectrometer, lipidomics approaches are generally classified into two types: LC-MS based and shotgun lipidomics³³. LC-based lipidomics reduces the complexity utilizing separation science in addition to enriching the low abundance species^34–35. Retention time adds a valuable dimension to enhance identification of lipid species. However, the time restriction due to the “on-the-fly” chromatographic analysis markedly limits the number of the precursor ions to be fragmented during an LC run unless prolonged elution is practiced. In shotgun lipidomics, the complexity has been strategically reduced either during sample preparation or during MS analysis via intrasource separation/selective ionization^36–37 or chemical derivatization³⁸, or both^39–40. The “unlimited” time in this approach allows all abundant precursor ions of interest to be fragmented for identification as demonstrated with the development of MS/MS(ALL) technology⁴¹. Another key feature difference between these two approaches is that the lipid concentration during the ionization process is virtually constant in shotgun lipidomics, but always changing in the LC-MS based approach. The changed concentration of lipid solution may introduce difficulties for accurate quantification of individual lipid species and additional internal standards are requested to compensate this complexity as previously discussed³⁶. The advantages and limitations of both approaches have been extensively reviewed previously⁴².

It should be mentioned that IM based lipidomic analysis is emerging, which could become an independent one once it develops⁴³. Other type of classification of lipidomics approaches also exists in literature. For example, based on whether the analysis of lipids is focused on certain types of lipid classes and individual species, scientists classify lipidomics approaches into “targeted”, “untargeted”, and “pseudo-targeted” approaches⁴⁴.

2. LC-MS BASED APPROACHES FOR ANALYSIS OF TG SPECIES

2.1. Reversed Phase LC-MS for TG Analysis.

Reversed phase LC (RPLC) coupled with MS is commonly used for TG analysis^45–46. The separation process is mainly based on the interaction of the stationary phase (mostly C18) with TG molecular species. Although stationary phases with other chain length (e.g., C8 or C28) have also been used, it appears that there is no any significant improvement in the separation⁴⁷. This is likely due to the fact that the majority of TG species consist of fatty acyls with chain lengths containing 16 to 20 carbons, which has similar hydrophobicity to that of C18 stationary phase. To this end, separation of TG species can be improved with a prolonged gradient as demonstrated⁴⁸.

RPLC separates TG species based on the equivalent carbon number (ECN) with isocratic elution, which is defined as ECN = CN − (2 × DB), where CN is the number of carbon atoms and DB is the number of double bond(s) in a TG species⁴⁹. Therefore, separation of TG regioisomers which have an identical ECN is not easy unless using a long column and/or existing other contributory effects such as the position of double bonds (see analysis of regioisomers in Section 5). To this end, Holcapek’s group has conducted a series of studies on optimizing the separation of TG species through investigation of the suitable stationary phase, length of the column, and column temperature⁴⁷. The effects of mobile phase on separation efficiency of TG species have also been extensively examined by Kuksis’ group⁵⁰. These studies should be consulted if someone would like to adopt RPLC for analysis of TG species. An extensive list of applications with this and other LC-MS techniques can be found in a recent review article by Wei et al.⁵¹.

2.2. Argentation (Silver Ion) LC-MS for TG Analysis.

Argentation LC (Ag⁺-LC)-MS has also been widely used for analysis of TG species^52–55. The separation of TG species by Ag⁺-LC is mainly based on the interaction of π-electrons of the double bonds in fatty acyls of TG species with d-electrons of silver ion bonded to the stationary phase. Therefore, TG molecular species containing more double bonds have stronger interaction with the column and show a longer retention time. This method can separate regioisomers, since the double bond(s) in the acyl bound to the primary hydroxyl groups are less sterically hindered and therefore have greater affinity with silver ions^55–56 (Figure 2). It appears that the method is usable for a maximum of ~10 double bonds in TG species due to the sterically hindered saturation between the interactions of double bonds with silver ions. Thus, this method is unlikely suitable for analysis of TG species in fish oils or from algae products, in which TG species containing up to 18 double bonds (3x docosahexaenoyls) are common.

Figure 2.

Silver-ion HPLC/APCI-MS chromatogram of the randomization mixture prepared from triolein (OOO, Δ9-C18:1), trilinolenin (LnLnLn, Δ9,12,15-C18:3) and tri-γ-linolenin (γLnγLnγLn, Δ6,9,12-C18:3). Numbers correspond to the double bond number. (Modified from ref. 55 with permission from Elsevier, Copyright 2010)

2.3. Supercritical Fluid Chromatography (SFC)-MS for TG Analysis.

SFC has also caught attention for lipidomics analysis, including the class of TG (see citations in a recent review⁵⁷). Separation of TG species by SFC is based on the principle of behavior of supercritical carbon dioxide, which is often used together with a mixture of modifiers (usually organic solvents).

For example, Sandra et al. used SFC-MS in combination with argentation for characterization of TG species in vegetable oils⁵⁸. In the study, TG species are mainly separated based on the number of double bonds, whereas within each group, an additional separation is observed according to the total carbon number.

In another study, SFC-MS was applied for analysis of soybean lipids including TG⁵⁹. In the study, SFC-MS with three tandem monolith C18 columns was used to analyze the TG species in a mixture of soybean lipids as ammonium adducts. Individual TG molecules were separated effectively within 8 min and illustrated on the 2D map based on the retention time and m/z (Figure 3).

Figure 3.

Identification of TG species in soybean lipid extract using SFC-MS. The 2D map shows a magnified view of SFC-MS data obtained by tandem three Chromolith Performance RP-18e columns. Small circle: peak top of each TG species. There are two types of groups that have the pattern of TG arrangement as indicated with arrows. Line arrows: a group of TG species with an sn-1 fatty acyl changed (box), e.g., OLP, LLP, and LnLP (A). Dotted arrows: a group of TG species with an sn-2 fatty acyl changed (circle), e.g., POP, PLP, and PLnP (B). (Modified from ref. 59 with permission from Elsevier, Copyright 2011).

2.4. Two-Dimensional Liquid Chromatography (LC × LC)-MS for TG Analysis.

Obviously, it is very important to effectively separate the TG species in order to reduce the complexity of TG analysis in LC-MS. In addition to selecting an appropriate column matrix and/or mobile phase as well as using a prolonged gradient, combining different resolving powers from various column types has also been naturally considered and practiced inline or offline^{46, 60–63}. Any readers who are interested in this area of work could find comprehensive description of column selection, method setup, and other experimental practice, as well as its applications from the review articles written by Byrdwell^64–65.

2.5. Mass Spectrometric Acquisition Modes Applied for TG Analysis by LC-MS.

Previously, TG species have been popularly detected by atmospheric pressure chemical ionization (APCI)/MS⁶⁶. Nowadays, TG species are largely detected by electrospray ionization (ESI)/MS as ammonium (or similar modifiers) or alkaline adducts due to its sensitivity and informative fragmentation^67–69. The ammonium adducts are largely employed in association with LC-MS considering its accommodating with chromatographic conditions. The alkaline adducts are employed with many shotgun lipidomics approaches^{37, 70–73}. Detection of protonated or anionic adducts is seldomly used for analysis of TG species due to their weak interaction, thus low sensitivity.

During chromatographic separation, detection of TG ions can be achieved in multiple acquisition modes. It is common to acquire a large array of mass spectra in a full mass range for TG analysis. Once all the mass spectral scans are recorded, selected ion monitoring (SIM) can be used to extract individual m/z corresponding to a potential TG mass^{53, 74–75}, from which a putative TG species with a known total number of carbon atoms and of double bonds can be derived. In this case, a sensitive and high mass resolution instrument with a rapid scan rate is desirable. The advantage of this acquisition mode includes its readiness to set up, broad coverage, and high sensitivity. The limitations of this acquisition mode include, but are not limited to, (1) the detected ion only represents a putative TG species (i.e., the identity is not definitively identified since it is impossible to use isotope-labelled standards to match all the species in lipidomic studies), and (2) no information about fatty acyl chains is provided, thus, isomers due to differential fatty acyl chains cannot be assigned.

To improve the aforementioned limitations of SIM detection for analysis of TG species, data-dependent acquisition (DDA) can be implanted after each survey scan. Here, DDA is a method where a fixed number of the most abundant precursor ions which are not fragmented in the previous analysis point(s) are selected and analyzed by tandem MS to facilitate structural identification. Considering the time restriction between full mass spectral acquisitions in order to guarantee the accuracy of ion peak shape, the number of precursor ions in each DDA cycle is limited. Therefore, effective separation by chromatography is critical to reduce severe overlaps and achieve broad identification of complex TG species. As aforementioned, TG species possessing identical or similar ECN closely elute. The setting of DDA may need to be modified according to the separation conditions.

For many scientists who have low mass resolution instruments (e.g., QqQ type), multiple reaction monitoring (MRM) is preferably employed to assess the levels of TG species⁷⁶. Transitions from TG precursor ions to the diglyceride-like fragment ions (a loss of fatty acyl chain) can be readily set up for this purpose. It should be recognized that this approach only allows scientists to detect a few abundant TG species due to time restriction which does not allow for extensively determining all possible fatty acyl losses at any time point if a prolonged gradient is not used for separation of TG species. Moreover, due to the pre-setting of the MRM transitions, the flexibility to monitor the dynamic changes of TG species in biological systems is relatively moderate.

2.6. Potential Caveats for Quantification of TG Species by LC-MS.

Accurate quantification is a key component of lipidomics analysis in general, which is also challenging to the field the most^77–78. One of the key factors to achieve accurate quantification is to include appropriate numbers of internal standards³⁶. Basically, if there are no internal standards in MS-based lipidomics, then there is no accurate quantification. The minimal numbers of internal standards that need to be used depend on the variables of the methodology, whereas, unfortunately, there exist the more variables in LC-MS based methods than in shotgun lipidomics³⁶. Many researchers using LC-MS do not recognize this potential caveat and not include enough internal standards for accurate quantification.

The majority of the researchers using LC-MS employ ammonium adducts for analysis of TG species. They might not recognize that the neutral losses of different fatty acyls from an ammoniated TG ion depend on the fatty acyl identities as well as the position connected to the glycerol hydroxyl groups as previously demonstrated⁶⁸. This position-dependent differential loss of fatty acyls might be very useful for identification of fatty acyl position(s) in TG species (i.e., TG regioisomers). However, it should be mentioned that a larger difference between the neutral losses of fatty acyls from an ammoniated TG species may be not favored for any method using the neutral loss of fatty acyls for TG quantification such as in the MRM method. In this case, different standard calibration curves for individual fatty acyl chains need to be generated as demonstrated⁶⁸.

Similar to the last point, some researchers who quantify TG species based on MRM transitions of fatty acyl losses attempt to maximize the detection sensitivity of individual TG species through separate optimization of collision energy for individual MRM transition. As well recognized, the fragmentation pattern of individual lipid species largely depends on the conditions of collision-induced dissociation (CID) (i.e., collision energy (which is the energy used to fragment an ion in tandem MS) and pressure). Thus, this type of optimization likely leads to very different losses of different fatty acyls from different positions of TG glycerol hydroxyl groups and yields very different response factors. Apparently, it would be problematic for quantification of individual TG species with this kind of different response factors, unless standard calibration curves for individual fatty acyl chains of individual TG species can be established. As we know, establishment of all these curves is not practical for lipidomics analysis. In practice, it would be advised that we should attempt to determine an appropriate CID energy to achieve virtually equal losses of fatty acyls from individual TG species similar to what we have demonstrated below (Section 3.1).

Finally, anyone who would like to conduct quantitative analysis of TG species should recognize that ionization response factors of TG species, unlike the majority of the polar lipids, are very much dependent on individual TG species as previously demosntrated⁶⁷. Therefore, if not a standard calibration curve for individual TG species is established, then correction factors for the TG species should be applied⁶⁷.

3. SHOTGUN LIPIDOMICS FOR IDENTIFICATION AND QUANTIFICATION OF TG SPECIES

3.1. Multi-Dimensional Mass Spectrometry-Based Shotgun Lipidomics Analysis of TG Species.

As previously discussed⁷⁹, lithium adducts of lipids superior to the other alkaline and ammonium adducts enable providing more informative fragments for structural identification^{37, 67, 80}. TG molecules can be easily ionized by ESI-MS in the positive-ion mode as lithiated ions when lithium salt-methanol solution is used as a modifier^{67, 81}. It should be pointed out that the disadvantage of using lithium salt includes more frequent clearance of ion source and mass spectral complication resulted from the presence of Li-6 isotopomers. Fortunately, utilization of a microfluidic device (e.g., nanomate) could markedly reduce the ion source contamination⁸². The effects of Li-6 isotopomers on quantification are really minimal with the presence of an internal standard since the intensities of Li-6 isotopomers are proportionally presented to those of Li-7 isotopomers. Alternatively, if preferred, these effects could also be calculated with deisotoping correction⁷⁹.

Tandem MS analysis of [TG + Li]⁺ after CID in the product ion mode displays intensive and informative product ions, corresponding to a unique fragmentation pattern of TG species, i.e., neutral losses of both free fatty acid and its paired lithium salt from a specific fatty acyl chain present in individual TG species (Figure 4). For example, fragmentation of a lithiated TG (16:0/20:4/18:1) ion at m/z 887.7 yields three virtually equally intense ions at m/z 631.5, 605.4, and 583.4 (Figure 4A) with the selected CID energy. These fragments correspond to the neutral losses of 16:0, 18:1, and 20:4 fatty acids from the precursor, respectively (Figure 4A). Additionally, there exist three other abundant product ions at m/z 625.5, 599.4, and 577.4 (Figure 4A), corresponding to lithium 16:0, 18:1, and 20:4 carboxylates, respectively. The fragments after neutrally lost lithium salts are generally less intense than their counterparts resulted from losses of free fatty acids, but more sensitive to the positions of the fatty acyls. As demonstrated, the intensities of the set of fragments corresponding to the losses of lithium salts of fatty acyls are different. This set of fragments can be used to identify the regioisomers of TG species.

Figure 4.

Representative tandem MS analysis of lithium TG adducts in the product-ion mode. Tandem MS spectra of lithiated 16:0–20:4–18:1 (Panel A) and 18:1–20:4–18:1 TG species (Panel B) were obtained using a ThermoFisher TSQ Altis mass spectrometer with collision energy of 32 eV and collision gas pressure of 1 mTorr. The tandem MS spectral analyses demonstrated the abundant fragment ions yielded from neutral losses of either free fatty acids or lithium fatty acyl salts from corresponding TG lithium adducts.

Similarly, product ions at m/z 631.4 and 609.4, corresponding to the neutral losses of 18:1 and 20:4 fatty acids, respectively, are yielded from lithiated TG molecular ion at m/z 913.7. The intensities of these fragment are roughly in a ratio of 2:1 (18:1 vs. 20:4), reflecting the numbers of these fatty acyl chains in the TG species (Figure 4B). In contrast, the loss of lithium 20:4 salt is more intense than that of lithium 18:1 salt, indicating that 20:4 fatty acyl is located at the sn-2 position (see Section 5 for mechanistic discussion of this pattern).

These observations indicate that the number of intense fragments displayed in the product-ion mass spectrum of [TG + Li]⁺ ion is dependent on the types of different fatty acyl chains existing in the TG molecule. In detail, six intense fragments are displayed in the product ion mass spectrum if there exist three types of different fatty acyl chains in a TG species; four abundant fragments are yielded from a TG molecule with two types of different fatty acyls; and only two intense fragments, corresponding to neutral losses of both fatty acid and its paired salt are produced if the TG molecule only has three identical fatty acyls. This fragmentation pattern has been examined with different TG species under various experimental conditions^{67, 70, 81}. It is recognized that neutral losses of free fatty acids from TG species can be used for identification of their structures and then quantification of the identified TG species in biological samples with high sensitivity and relatively more stable thermodynamics (i.e., more comparable intensities of these fragment ions) than those of their corresponding salts^{37, 67}.

By employing this fragmentation pattern, all TGs existing in biological samples can be unambiguously identified and quantified by two-dimensional MS analysis in an multi-dimensional MS-based shotgun lipidomics (MDMS-SL) approach⁶⁷ (Other approaches are summarized below). This is a big advantage of MDMS-SL analysis of TG species. Specifically, in the approach, a survey MS spectrum in the range from m/z 750 to 1000 (which covers all the TG species in the majority of biological systems) and then sequential neutral loss (NL) scans of all naturally-occurring fatty acids (~30 kinds of fatty acyls) (i.e., NL228 for 14:0 fatty acid, NL256 for 16:0 fatty acid, NL282 for 18:1 fatty acid, etc.) as building blocks of TG species were acquired (Figure 5). Then, 2D mass spectrum can be built with the survey mass spectrum and all the neutral loss scans (Figure 6).

Figure 5.

A full mass spectrum and total ion current chromatogram of stepwise acquisition of neutral loss scans for identification and quantitation of TG molecular species in lipid extracts of rat superior mesenteric ganglia. The lipid extract from rat superior mesenteric ganglia was analyzed in the positive-ion mode after infusion of the diluted lipid extract in the presence of a small amount lithium chloride directly with a NanoMate device. A positive-ion full mass spectrum of lipid extracts of rat superior mesenteric ganglia was acquired in a survey scan mode by a QqQ-type mass spectrometer (Thermo Fisher Scientific TSQ Altis) (Panel A). The stepwise acquisition of neutral loss (NL) scans as indicated (Panel B) was conducted using a sequential and customized program operating under Xcalibur software. Each segment of individual NL scan was taken for 2 min in the profile mode. Panel C shows an example of NL scan averaged from all of scans acquired in the segment corresponding to NL310 (i.e., 20:1 fatty acid) which is usually present in very low abundance.

Figure 6.

A representative two-dimensional ESI/MS analysis of TG molecular species in rat superior mesenteric ganglion lipid extracts. The full mass spectrum acquired in the survey scan mode (the most top trace, the first dimension of the two-dimension mapping) was obtained in the positive-ion mode after direct infusion with a NanoMate device. Neutral loss (NL) scans of all naturally-occurring aliphatic chains of lipid extracts of rat superior mesenteric ganglia, serving as building blocks were acquired and utilized to identify TG molecular species, deconvolute isomeric molecular species, and quantify TG individual molecular species by comparisons with a selected internal standard. Each MS or MS/MS scan of two-dimensional ESI mass mapping was acquired by sequentially programmed custom scans operating under Xcalibur software as shown in Figure 5. For tandem mass spectrometry in the positive-ion neutral loss (NL) mode, both the first and third quadrupoles were coordinately scanned with a mass difference (i.e., neutral loss) corresponding to the neutral loss of a non-esterified fatty acid from TG molecular species, while collisional activation was performed in the second quadrupole. All mass spectral traces were displayed after normalization to the base peak in individual scan.

In the built 2D MS mapping, the cross peaks of a specific TG ion in the survey mass spectrum with those present in the neutral loss scans (i.e., the building blocks of TG species in the second dimension) represent the fatty acyl chains that constitute the isomeric TG species underlying the selected precursor ion. These isomeric TG species due to different types of fatty acyl chains, but having an identical total number of carbon atoms and of double bonds can thus be easily and accurately idnetified from the number and intensities of these crossing neutral loss fragment peaks along with the m/z value of TG ion⁶⁷.

For example, TG isomers from the molecular ion at m/z 865.7 can be manually identified as follows. This TG ion peak is crossed with neutral loss scans corresponding to 16:1, 16:0, 18:1, and 18:0 fatty acids at the modest to high abundance, and also crossed with neutral loss scans corresponding to 14:0, 18:2, and 20:2 fatty acids as well as a few others (not displayed in the two-dimensional MS mapping) at low abundance (as indicated with the broken line) (Figure 6). Based on the molecular mass of this lithiated TG species (i.e., 865.7 Da), this TG species should consist of a total of 52 carbon atoms and 2 double bonds or a total of 53 carbon atoms and 9 double bonds in three fatty acyls. Considering that the ion intensities resulting from the neutral losses of three fatty acyls from a given TG isomer are nearly equal (Figure 4), TG isomers of 16:0/18:1/18:1, 16:1/18:0/18:1, 14:0/16:0/20:2 (minor), 16:0/18:0/18:2 (minor), etc. can be identified. Other TG species displayed in the full MS scan can be identified similarly. Annotation of TG molecules herein follows the recommendation as previously described⁸³.

Unfortunately, it is time consuming and labor intense to manually assign individual TG isomers underlying a TG ion as aforementioned. Recently, a software program for identification of TG identities through simulation of TG ions as well as neutral loss scans of all naturally-occurring fatty acids, and based on TG synthesis pathways has been developed⁸⁴. This program enables us not only for assignment of individual TG isomers underlying each TG ion, but also identification of numerous low to very low abundance TG species (see Section 6 for details).

Similar to the MDMS-SL approach, analysis of TG isomers in plant seeds through acquisition of multiple neutral loss scans of fatty acids from ammoniated TG molecules was also conducted⁸⁵. In the study, the levels of 93 individual TG molecules from 13 TG molecular groups were determined in wild-type Arabidopsis seeds.

A few points need to be emphasized using the TG fragmentation pattern and two-dimensional MS mapping for identification and quantification of TG species in biological samples.

First, sodiated TG species are displayed in MS spectral analysis of TG to a certain extent even cautious measures are taken. Minimization of sodiated TG intensities is important since ionization of sodiated TG species is dependent on TG molecular structures. Specifically, the intensity of sodiated TG species depends on the number of double bonds existing in TG molecules as previously demonstrated⁶⁷.

Next, a concern with whether all the fatty acyl chains of TG molecules are determined is always present when the MDMS-SL technology is exploited to identify TG molecules in biological samples. In fact, scanning of 10 commonly natural fatty acids (even not counting any fatty acids with odd-numbered carbon atoms) should be good enough to detect the major mass of TGs (i.e., > 90 mol%) existing most of biological systems. However, monitoring all possibly neutral losses of naturally-occurring fatty acids in MDMS-SL analysis of TG species is always better if sample and resource permit.

Thirdly, further to the last point, while fatty acids containing less than 14 carbon atoms usually exist at the low content in TG pools, relatively short fatty acyls have been found abundantly present in some TGs (e.g., in milk)^86–87. Thus, it should always keep in mind that if the neutral loss of 14:0 fatty acid yields strong signals, neutral losses of shorter fatty acids should be assessed. A similar strategy should also be used for measuring the presence of fatty acyls containing odd-numbered carbon atoms in TG pools.

Fourthly, interfering identification and quantification TG species due to neutral losses of fatty acids from phosphatidylcholine (PC) molecules which are overlapped with TG species to a certain degree is usually negligible as previously discussed⁶⁷. Moreover, the fragment ions from PC precursors can be easily recognized from the even-numbered m/z whereas those from [TG + Li]⁺ ions are shown as odd-numbered m/z following the nitrogen rule (Figure 6).

Lastly, quantification of identified individual TG species can be readily conducted from MDMD-SL analysis as follows: first, the content of individual TG precursor ion is determined in comparison to the spiked internal standard (i.e., tri17:1 TG) after correction for differences of C-13 isotopologue distribution as well as the different number of total carbon atoms and of total double bonds from the standard as previously described^{67, 79}; then, the composition of individual TG isomers underlying each TG ion is determined according to the ion intensities of individual isomers as described above; and finally, the content of individual identified TG isomers can be readily derived from the total mass of the TG ion and the composition of individual TG isomers underlying the ion.

3.2. High Mass Resolution Mass Spectrometry-Based Shotgun Lipidomics for Analysis of TG Species.

Profiling of TG species by high mass resolution/accuracy shotgun lipidomics is conducted using a DDA-driven multiple neutral loss scanning approach⁶⁹ similar to that of MDMS-SL. In this approach, neutral losses of fatty acids present in TG pools are extracted from the DDA arrays of TG species.

For example, in a previous study⁶⁹, the band containing TG was scraped from the separation of total lipid extract of worms by preparative TLC. The extract was then subjected to DDA-driven profiling by a Q-ToF mass spectrometer. In the study, Q1 was operated under the unit mass resolution settings in order to accurately resolve the neighboring species. The subsequent automated data extraction of the dataset by LipidInspector was equivalent to the parallel acquisition of 69 quasi neutral loss spectra relevant to neutral losses of ammonia and fatty acids comprising 9 to 22 carbon atoms and zero to six double bonds. In each quasi neutral loss spectrum, the corresponding neutral loss fragment was identified. Individual TG species were assigned with the total number of carbon atoms and of double bonds in three fatty acid moieties. The relative intensities of individual TG peaks were summed up and normalized. In total, 35 TG species each with unique sum formula were identified.

The quasi neutral loss spectra indicated that the molecular composition of TG species was very heterogeneous, each precursor ion represented 5 to 15 isomeric species. The relative abundance of species with the unique fatty acyl composition was also estimated assuming that the yield of neutral loss products is approximately independent of the fatty acyls and their positions at the backbone, although their regioisomers (i.e., differing by their relative positions at the glycerol backbone) were not distinguished. The researchers composed a system of linear equations for each fragmented precursor and calculated the relative abundance of species with the unique fatty acyl composition after considering the relative intensities of detected neutral loss fragments.

3.3. Analysis of TG Species during Flow Injection.

According to our strict definition of shotgun lipidomics which classify the approaches having a constant concentration of the infusion solution³⁷, flow injection-based direct infusion methodology does not belong to this category. However, since this method represents a direct infusion-based approach, some scientists also call it shotgun lipidomics. Analysis of TG species after flow injection has also been performed^88–89. In fact, after flow injection, either MDMS-SL type or DDA-driven analysis as described above could be conducted to a certain degree depending on the coupled mass spectrometer. Moreover, similar to LC-MS MRM methodology, setting up different MRM transitions for detection of neutral losses of different fatty acids during the flow could also be conducted, although resolving isomeric TG species is difficult and inclusion of false positive identification of TG species is very likely in this case.

4. IM-MS FOR ANALYSIS OF TG SPECIES

IM-MS is a widely used and ‘well-known’ technology of ion separation in the gaseous phase based on the differences in ion mobilities under an electric field. The IM technology advances MS analysis to a new dimension since this technology allows scientists not only to separate analytes based on polarity, but also separate species with different configurations^{43, 90}. A variety of mass spectrometers integrated with an IM device with different separation principles and located at different stages of a MS separation system^{43, 90–91}. IM-MS can be readily and has been integrated with LC-MS or shotgun lipidomics for lipid analysis⁹². Therefore, we separately give an overview of the technology for TG analysis.

IM-MS is capable of separating analytes based on their configuration. Thus, applications of this technology for analysis of TG regioisomers naturally occur. For example, Sala et al.⁹³ used differential mobility mass spectrometry to test the feasibility of separation of TG regioisomers. In the study, they optimized various experimental parameters such as the separation and compensation voltages, the type and flow rate of chemical modifiers (e.g., 1-butanol and 1-propanol), the dwell time of analyte ions, etc. The researchers concluded that separation of TG regioisomers could get done in a much faster fashion (less than a minute) compared to other LC-MS settings such as silver-ion or RPLC-MS approaches.

Usually, distinguishing lipid isomers due to double bond position or steroconfiguration without chemical derivatization is very challenging in lipidomics. Luckily, these isomers can also be directly separated using IM devices. For example, IM-MS was combined with LC separation to determine the feasibility of various lipid isomers present in TG species including stereoisomers, isomers due to double bond locations, and cis/trans double bond isomers⁹⁴. The scientists employed a field asymmetric wave-form IM spectrometry (FAIMS) and found the resolution of these types of isomers can be readily achieved.

In other studies, IM-MS has been successfully applied for profiling TG species of biological samples. For example, LC-IM-MS was used for studying TG profiles and identified numerous novel TG species in human milk⁹⁵, although MS/MS analysis of these novel species to verify their structural identities is necessary. IM-MS after direct infusion was applied for examining the changes of TG species induced by nitrogen deprivation in green alga and demonstrated marked changes of TG content⁹⁶.

As aforementioned, in addition to the polarity of analytes, separation of analytes also depends on the configuration of individual compound, which can be evaluated with a collision cross-section value. Thus, scientists in the field have made efforts to determine or establish models to theoretically predict these values of lipids including TG species^97–98. These values are very useful not only for developing experimental parameters, but also for eliminating potential false positive identification of lipid species.

IM-MS technology offers great advantages for TG analysis as described above. However, there also exist some limitations which are varied with the type of IM devices as previously discussed⁴³. Moreover, for lipidomics in general and TG analysis in particular, accurate quantification of lipid species after IM is still rare and challenging. This area of research remains elusive.

5. IDENTIFICATION OF TG REGIOISOMERS AND ENANTIOMERS

Both shotgun lipidomics approaches described in Section 3 enables us to determine the isomers due to different fatty acyl substituents (e.g., X-Y-Z vs. V-W-Z, where the total numbers of carbon atoms and of double bonds in X and Y are identical to that in V and W, respectively) according to the different neutral losses of fatty acids from an identical precursor ion. However, there is a long-standing demand for identification of regioisomers (i.e., isomers due to different location of fatty acyl chains to the different position of the glycerol backbone) and enantiomers (chiral molecules that are mirror images of each other and possess different optical properties). Identification of regioisomers and enantiomers, which can provide informative insights into the synthesis pathways and lipase selective activities, is important for understanding TG metabolism and homeostasis.

As discussed in the introduction, various fatty acyls can connect to primary and secondary hydroxyls of the glycerol backbone in TG to form different regioisomers and enantiomers. Determination of these isomers, which depends on chemical and enzymatic methods in classical approaches, is time- and material-consuming without a clear outcome. However, employing LC-MS (e.g., a chiral column) and IM technology makes identification of individual TG regioisomers and enantiomers possible^{56, 93, 99}. The most commonly used stationary phases are cellulose-tris-(3,5-dimethylphenylcarbamate) or cellulose-tris-(3-chloro-4-methylphenyl carbamate) and mixtures of hexane/isopropanol or methanol as mobile phases. Examples of some regioisomer analyses have recently been reviewed⁴⁵. In addition to chiral chromatography, other types of chromatography coupled with MS such as reversed-phase or supercritical fluid chromatography are also frequently used for this purpose, but less accomplished, particularly for resolving regioisomers in TG pools from animal samples⁴⁵.

Identification of TG regioisomers can also be achieved by MS/MS analysis without any chromatographic separation as we discussed in the section above. The mechanistic basis of MS/MS identification of TG regioisomers has been well described and validated by Hsu and Turk⁸¹. They found that production of the fragments involves the initial elimination of fatty acyl salt together with an α-hydrogen atom from the neighboring fatty acyl chain, followed by formation of a cyclic intermediate that degrades to generate other characteristic fragments. According to this observation, the loss of fatty acyl salt at the sn-2 position is much easier than that from either sn-1 or -3 fatty acyls since both sn-1 and -3 fatty acyls can provide the α-hydrogen atom to facilitate the sn-2 fatty acyl salt elimination whereas only sn-2 fatty acyl can provide the α-hydrogen atom to facilitate either sn-1 (not shown) or -3 fatty acyl salt elimination (Figure 7). The relatively different intensities of the fragments resulting from neutral losses of fatty acyl salts in Figure 4 reflects such a differential loss. Thus, this character can be and has been employed to identify the regioisomers of fatty acyls on the glycerol backbone^{81, 100}. It should be recognized that the differences of peak intensities between the fragments resulting from the neutral losses of free fatty acids from lithium TG adducts are relatively smaller than those resulting from the neutral losses of their paired salts. In our studies, we purposely select an appropriate CID energy to eliminate this type of difference between the losses of fatty acids from different positions of glycerol backbone for quantification purposes^{67, 101}. Moreover, in samples from mammalian sources, the distribution of fatty acyls at different positions of glycerol backbone is relatively more homogeneous after extensive Land’s and futile cycles. These minimally-differential fragment ions resulted from lithium adducts make the quantification of individual TG isomers more convenient and accurate than those yielding significantly-differential fragments.

Figure 7.

Schematic illustration of fragmentation pathways yielding sn-2 vs sn-3 fatty acyl losses from a lithiated TG ion. Panels A and B indicate both sn-1 and -3 fatty acyls can provide the alpha-hydrogen atom to facilitate the sn-2 fatty acyl salt elimination. Panel C indicates only sn-2 fatty acyl can provide the alpha-hydrogen atom to facilitate either sn-1 (not shown) or sn-3 fatty acyl salt elimination. R_n = R_n’ = R_n”. (Modified from ref. 81 with permission from Elsevier, Copyright 1999).

6. SIMULATION OF LIPIDOMICS ANALYSIS OF TG SPECIES FOR DATA INTERPRETATION AND MOLECULAR SPECIES IDENTIFICATION

Bioinformatic tools and systems biology approaches to connect altered lipids to the changes of biological functions, such as enzymatic activities involving the de novo synthesis of the changed lipids remains further developing. This kind of development is very significant in advancing the understanding of the roles of lipids in biological systems and of the biochemical mechanisms underlying the changes of lipids¹⁰². Moreover, manual interpretation and assignment of lipidomics results for TG analysis in general is very complex and time-consuming. Therefore, it is critical to develop strategies for automatic assignment of TG species and interpretation of lipidomics data from TG analysis. To this end, we have attempted making some efforts in this area and the outcome is apparently very attractive⁸⁴.

We have developed a simulation strategy for processing the comprehensively-determined lipidomics data from TG analysis in combination with the known TG biosynthesis pathways in order to measure the levels of each TG biosynthesis pathways contributing to TG content as well as to determine the enzymatic activities related to TG de novo synthesis⁸⁴. The simulation strategy was explored according to the established TG de novo synthesis pathways of DG reacylation with acyl CoA¹⁰³, whereas we recognize that DG molecules might be yielded from various metabolic sources (Figure 8). It should be pointed out that there exist other additional pathways including lipase-mediated TG hydrolysis to yield DG as well as futile cycling. This simplified model only considers a steady state as the nature of lipidomics, thus futile cycling can be ignored. For the majority of the organs except adipocytes (e.g., the heart, the liver, and muscle) and seeds, TG hydrolysis through a lipase activity appears minimal as evidenced with the simulation⁸⁴.

Figure 8.

Schematic illustration of a TG de novo synthesis model for simulation of TG ion composition quantified using the MDMS-SL technology. TG molecules are biosynthesized with reacylation of diglyceride (DG) molecules of various sources yielded mainly through dephosphorylation of phosphatidic acid (PA) (DG_PA), reacylation of monoglyceride (MG) (DG_MG), and, to a less extent (as indicated with a broken line arrow), through hydrolysis of phosphatidylinositol (PI) with phospholipase C (PLC) activities (DG_PI). The levels of the metabolic pathways contributing to the TG content were calculated via simulation of each TG ion profile including all neutral loss scans with parameters of K₁, K₂, and K₃ using the equations (1) to (3) with the restriction of equation (4), respectively. These parameters represent the probabilities of each DG source being reacylated to TG. In addition, the parameters of k₁, k₂, and k₃ were used in the sn-1, 2, and 3 reacylation steps of TG species in the forms of exp(−k₁.x_j), exp(k₂.x_j), and exp(−k₃.x_j), respectively, where k₁ and k₃ represented a simulated decay constant whereas k₂ represented a simulated enhancing constant, and x_j is the number of double bonds present in the corresponding fatty acyl chain. MGAT and DGAT stand for MG and DG acyltransferases, respectively. The multiple arrows at the k₂ step indicate that MG molecules could be produced from various sources including lysoPA dephosphatation, DG hydrolysis, and glycerol acylation. (Modified from ref. 84 with permission from the American Society for biochemistry and molecular biology, Copyright 2013). Lipase-mediated hydrolysis of TG species is not considered under steady state conditions as discussed in the text.

The pathways yielding DG species include (i) the dephosphorylation of PA (DG_PA) through a PA phosphatase (PAP) activity, (ii) the reacylation of monoglyceride (MG) species (DG_MG), which might be generated from different origins, such as dephosphorylation of lysoPA, TG/DG hydrolysis with lipase activities, and reacylation from glycerol^103–104, and (iii) to a less degree, hydrolysis of phosphatidylinositol (PI) and PI polyphosphate, or even other glycerophospholipid (GPL) species through phospholipase C (PLC) activities (DG_PI). The developed simulation strategy was verified by comparing the content and composition of each simulated TG molecules with those determined from lipidomics analysis⁸⁴. The calculated K₁, K₂, and K₃ values (Figure 8) represent the corresponding metabolic pathways contributing to the TG biosynthesis. These contributions represent the enzymatic activities associated with individual metabolic pathways of TG biosynthesis.

Like the dynamic simulation of cardiolipin species¹⁰⁵, the bioinformatic simulation of TG molecular composition provides us a powerful approach for determining alterations in TG biosynthesis pathways under pathophysiological conditions. For example, we conducted a primary lipidomics study of TG profiling on liver samples from high fat diet-fed mice vs. chow diet-fed mice (unpublished data). From the simulation, we found that K1 and K2 contributions were changed from 0.493 ± 0.031 and 0.507 ± 0.031 in chow-fed mice to 0.446 ± 0.025* and 0.554 ± 0.030* (*p < 0.05) in high fat diet-fed mice, respectively. These results indicate that TG synthesis in the liver is switched from de novo synthesis through the PA dephosphorylation pathway to the MG reacylation pathway.

In addition to providing bioinformatic interpretation of lipidomics TG data, simulation of TG content and composition as well as all naturally-occurring fatty acid profiles determined by neutral loss scans can provide best prediction of TG molecules existing in the biological samples. For example, thousands of TG molecular species in the liver, heart, and muscle can be determined from simulation of lipidomics TG data⁸⁴, that are unaccomplishable using any currently available tools which usually analyze tens of TG species⁷⁶.

7. CONCLUSION

Collectively, in the current review, we summarized the methodology of both LC-MS and shotgun lipidomics for analysis of TG species, including analysis of regioisomers and enantiomers, present in biological samples such as plant seeds, human milk, and animal samples. We also discussed the limitations of individual methods for TG analysis to a certain degree. We believe that the reviews and discussions provided herein are helpful for researchers to select an appropriate approach for TG analysis and could serve as the basis for those who would like to establish novel methodology for TG analysis or develop a new method when novel tools become available. We also believe that bioinformatic development is crucial for determining the large number of TG species present in a biological system and providing molecular insights into alterations in metabolic pathways of TG biosynthesis. Finally, from the biochemical point of view, accurate analysis of TG species with an enabling method should allow us to improve the nutritional quality, reveal the effects of TG on diseases, and uncover the underlying biochemical mechanisms related to these diseases.

Abbreviations used:

AGPAT

acylglycerol-phosphate acyltransferase

APCI

atmospheric pressure chemical ionization

CID

collision-induced dissociation

DDA

data-dependent acquisition

DG

diglyceride

DGAT

DG acyltransferase

DHPA

dihydroxyacetone phosphate

ECN

equivalent carbon number

ER

endoplasmic reticulum

ESI

electrospray ionization

FAIMS

field asymmetric wave-form IM spectrometry

G3P

glycerol-3-phosphate

GPAT

glycerol-phosphate acyltransferase

GPL

glycerophospholipid

IM

ion mobility

LC

liquid chromatography

MDMS-SL

multi-dimensional MS-based shotgun lipidomics

MG

monoglyceride

MRM

multiple reaction monitoring

MS

mass spectrometry

MS/MS

tandem mass spectrometry

NL

neutral loss

PA

phosphatidic acid

PAP

PA phosphatase

PC

phosphatidylcholine

PI

phosphatidylinositol

PLC

phospholipase C

RPLC

Reversed phase LC

SFC

supercritical fluid chromatography

SIM

selected ion monitoring

TG

triglyceride