Skip to main content
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2022 May 23.
Published in final edited form as: Rapid Commun Mass Spectrom. 2020 Nov 30;34(22):e8911. doi: 10.1002/rcm.8911

Evaluating a targeted multiple reaction monitoring approach to global untargeted lipidomic analyses of human plasma

Mostafa J Khan 1, Simona G Codreanu 1,2,3, Sandeep Goyal 1, Phillip A Wages 1, Santosh KK Gorti 5, Mackenzie J Pearson 5, Isabel Uribe 6, Stacy D Sherrod 1,2,3, John A McLean 1,2,3, Ned A Porter 1,3, Renã AS Robinson 1,3,4
PMCID: PMC9126483  NIHMSID: NIHMS1787685  PMID: 32738001

Abstract

Rationale:

The Lipidyzer platform was recently updated on a SCIEX QTRAP 6500+ mass spectrometer and offers a targeted lipidomics assay including 1150 different lipids. We evaluated this targeted approach using human plasma samples and compared the results against a global untargeted lipidomics method using a high-resolution Q Exactive HF Orbitrap mass spectrometer.

Methods:

Lipids from human plasma samples (N = 5) were extracted using a modified Bligh–Dyer approach. A global untargeted analysis was performed using a Thermo Orbitrap Q Exactive HF mass spectrometer, followed by data analysis using Progenesis QI software. Multiple reaction monitoring (MRM)-based targeted analysis was performed using a QTRAP 6500+ mass spectrometer, followed by data analysis using SCIEX OS software. The samples were injected on three separate days to assess reproducibility for both approaches.

Results:

Overall, 465 lipids were identified from 11 lipid classes in both approaches, of which 159 were similar between the methods, 168 lipids were unique to the MRM approach, and 138 lipids were unique to the untargeted approach. Phosphatidylcholine and phosphatidylethanolamine species were the most commonly identified using the untargeted approach, while triacylglycerol species were the most commonly identified using the targeted MRM approach. The targeted MRM approach had more consistent relative abundances across the three days than the untargeted approach. Overall, the coefficient of variation for inter-day comparisons across all lipid classes was ∼ 23% for the untargeted approach and ∼ 9% for the targeted MRM approach.

Conclusions:

The targeted MRM approach identified similar numbers of lipids to a conventional untargeted approach, but had better representation of 11 lipid classes commonly identified by both approaches. Based on the separation methods employed, the conventional untargeted approach could better detect phosphatidylcholine and sphingomyelin lipid classes. The targeted MRM approach had lower inter-day variability than the untargeted approach when tested using a small group of plasma samples. These studies highlight the advantages in using targeted MRM approaches for human plasma lipidomics analysis.

1 |. INTRODUCTION

Lipidomics has been one of the most rapidly developing branches of science over the last decade due to the potential of linking lipids to different human health issues. With the advancement in mass spectrometric instruments over the last few decades, more studies are being conducted focusing on lipids and their role in various disease pathologies, examples of which include diabetes,1,2 obesity,3 cystic fibrosis,4 and many types of cancers.59 Due to the high content of lipids in the brain and central nervous system, Alzheimer’s disease and many other neurological diseases such as multiple sclerosis, epilepsy, schizophrenia, and Parkinson’s disease have been associated with faulty lipid metabolism.1014 Numerous lipidomics studies have established a correlation between altered lipid metabolism and Alzheimer’s disease.1519 Although several studies have been conducted over the years, there are still many areas in the field that require improvement. For example, there is a major need for a universal method for the analysis of lipids from various classes.

Targeted and untargeted lipidomics approaches can be utilized to study single or multiple lipid species. In targeted approaches, a known lipid molecule and/or lipid class of interest is selectively chosen for mass spectrometry (MS) analysis, while, in an untargeted approach, all extracted lipid compounds from diverse classes are monitored simultaneously in a single MS assay. Relative and absolute quantification20 can be performed using internal standards in both these approaches. While targeted approaches have the advantages of being highly specific, selective, and accurate with regard to quantification, these approaches are based on a priori selection of species, leaving many unknowns undetected. On the other hand, untargeted approaches monitor all species in a putatively unbiased manner and have the potential to discover new lipid species that may be indicative of a disease state.

Recently, SCIEX introduced a targeted lipidomics method using a combination of hydrophilic interaction liquid chromatography (HILIC) separation and multiple reaction monitoring (MRM)-based assay to analyze ∼ 1150 different lipids from 19 different classes of lipids.21 This method has the advantage of being highly specific, with the ability to identify a broad array of lipids with high accuracy and precision, and with streamlined data analysis.21 The SCIEX QTRAP 6500+ mass spectrometer offers fast polarity switching between positive and negative ionization modes and high sensitivity at higher acquisition rates. These features, in conjunction with HILIC separation of lipids into their individual classes, make it easier to assign MRM measurements to individual lipid species within a narrow retention time window.21

In the study reported here, we evaluated the performance of this targeted MRM method against a conventional untargeted approach, using a reversed-phase (RP) separation in conjunction with a Thermo Q Exactive HF quadrupole-Orbitrap™ mass spectrometer. The untargeted approach used two separate injections for both positive and negative ionization modes, unlike the targeted approach, which utilized polarity switching within the same run. In these studies, we focused our evaluations on (1) the number of lipids identified by both approaches, (2) classes of lipids individually identified by each approach, (3) relative quantification of lipid classes, and (4) the overall ease of data acquisition and data processing. We also benchmarked these findings against previously reported studies.

2 |. EXPERIMENTAL

2.1 |. Plasma sample collection

Human plasma samples (N = 5) from healthy cognitively normal individuals were collected from the University of Pittsburgh Alzheimer’s Disease Research Center. Approval for the participation of human subjects was obtained by the Institutional Review Boards of the University of Pittsburgh and Vanderbilt University. Samples were collected in 2000–2015 from cognitively normal individuals. The average age of all the patients at the time of draw was ∼ 75 years and both male and female individuals were included in the study.

2.2 |. Untargeted lipidomics study

2.2.1 |. Lipid extraction

Lipids were extracted using a modified Bligh–Dyer extraction protocol.22 Briefly, plasma samples (30 μL) were transferred into a borosilicate glass tube followed by addition of 4 mL of solvent A (chloroform–methanol solution (1:1 v/v)) and 2 mL of 50mM LiCl. The tubes were vortexed for 20 s and centrifuged at 2700 g for 10 min. The bottom organic layer containing the lipids was carefully collected, and 2 mL of chloroform was added to the aqueous phase (upper layer) to re-extract the remaining lipids. The sample was vortexed and centrifuged again. The subsequent bottom layer was combined with the previously collected lipids and dried down using centrifugal evaporation. The whole procedure was repeated again with the dried-down sample as above except that 10mM LiCl was used instead of a 50mM LiCl solution. The lipid layer was collected and dried down before being reconstituted for injection into the instrument.

2.2.2 |. LC/MS/MS analyses

For the untargeted analysis, a reconstitution solution was prepared by adding internal standard solution (Splash Lipidomix® from Avanti Alabaster, AL, USA), which was constituted of a mixture of 14 different isotopically labelled lipids (Table S1, supporting information) at a 1:5 ratio of standard to solvent A (chloroform–methanol solution (1:1 v/v)).2325 The dried-down lipids were reconstituted with 100 μL of the reconstitution solution and vortexed to dissolve all the lipids. A quality control (QC) sample was prepared by adding an equal amount of each sample to generate a QC pool mixture. LC/MS/MS analysis was performed with a Vanquish high-performance liquid chromatography system (Thermo Fisher Scientific, Bremen, Germany) coupled to an Orbitrap Q Exactive HF hybrid quadrupole-Orbitrap™ mass spectrometer (Thermo Fisher Scientific, Bremen, Germany). Both positive and negative electrospray ionization modes were used for the analysis. In positive ion mode, 4 μL of sample was loaded onto a Hypersil Gold C18 column (3 μm, 2.1 mm × 100 mm; Thermo Fisher Scientific, Waltham, MA, USA); in negative ion mode, 6 μL of sample was injected. RP-LC separation was performed at a flow rate of 250 μ L/min using solvent A (water with 0.1% formic acid (FA)) and solvent B (isopropanol–acetonitrile–water at 60:36:4 ratio with 0.1% FA) with the following gradient: 40–70% B over 5 min, 70–100% B for 10 min, 100% B for 7 min, 90–20% B for 1 min, 20% B for 3 min, 40% B for 4 min for a total gradient time of 30 min. Full scan MS spectra were acquired over a mass range of m/z 100–1500 in both positive and negative ion mode. The Q Exactive-HF has the capability of performing polarity switching which has been used for untargeted lipidomics experiments previously26 and could have been used herein. However, in order to ensure comparable sampling across each lipid species, separate injections for positive and negative mode on the Q Exactive-HF were preferred. The source parameters were as follows: spray voltage 3 kV (both positive and negative modes); capillary temperature 280°C; sheath gas (N2) pressure 30 arbitrary units (a.u.) in positive ion mode and 40 a.u. in negative ion mode; auxiliary gas (N2) pressure 5 a.u. (positive) and 10 a.u. (negative); spare gas (N2) pressure 1 a.u.; probe heater temperature 300°C (positive) and 400°C (negative); and S lens level 40%. The resolution was set to 60 000 with the automatic gain control (AGC) target set at 1 × 106 ions and a maximum ion injection time of 100 ms. The top two most intense precursor ions were selected for tandem mass spectrometry (MS/MS) experiments. The MS/MS scans were acquired at a resolution of 15 000 using an isolation width of 1.5 m/z units, stepped collision energy (NCE 15, 20, 25), and a dynamic exclusion of 6 s. The AGC target was set at 2 × 105 ions and ion injection time of 100 ms.

2.2.3 |. Data analysis

RAW files were analyzed using Progenesis QI (Non-linear Dynamics, Newcastle, UK) following a previously described process.27 Briefly, all the data files (both sample and QC) were imported and aligned against a full MS QC pool reference and ions related to the intact molecule ([M + Na]+, [M + K]+, [M + Li]+, [M + H]+, [M + H − H2O]+, [M − H], [M – H − H2O], [M + Cl]) were selected for data processing and deconvolution. Peak picking was performed at a minimum threshold of 2.5 × 105 ion intensity. Unique ions (retention time and m/z pairs) were grouped (a sum of the abundances of unique ions) using both the above molecule-related ions and isotope deconvolution to generate unique “features” (retention time and m/z pairs) representative of each compound. Data were normalized using Progenesis QI for all compounds. Annotations were assigned within Progenesis QI using accurate mass measurements (<5 ppm error), isotope distribution similarity, and manual assessment of fragmentation spectral matching (when applicable) from LipidMaps,28 Lipidblast,29 and the Human Metabolome Database.30

2.3 |. MRM targeted study

2.3.1 |. LC/MS/MS analysis

Plasma samples were extracted using the same protocol as for the untargeted study, although prepared on a separate day using aliquots from the same sample set. The extracted lipids were reconstituted using ethanol with internal standards (Splash Lipidomix® mix and standard mix made by combining individual standards bought from Avanti) added at different concentrations (Table S1, supporting information). The amide-based LC/MS/MS analysis was performed using an ExionLC™ system (SCIEX, Framingham, MA, USA) consisting of a binary high-pressure mixing gradient pump with a degasser, a thermostatically controlled autosampler, and a column oven. Separation was achieved on an XBridge Amide column (4.6 × 150 mm, 3.5 μm; Waters, Milford, MA, USA). The LC method details were: column temperature, 35°C; flow rate, 0.7 mL/min; injection volume, 5 μL. The mobile phases were: solvent A (water–acetonitrile) at 5:95 ratio with 1mM ammonium acetate (adjusted to pH 8.4) and solvent B (water–acetonitrile) at 50:50 ratio with 1mM ammonium acetate (pH 8.2) with the following gradient: 0–6% B over 6 min, 6–25% B for 4 min, 25–98% B for 1 min, 98–100% B for 2 min, 100% B for 5.6 min, 100–0.1% B for 0.1 min, 0.1% B for 5.3 min for a total gradient time of 24 min. The QTRAP 6500+ mass spectrometer was equipped with an IonDrive™ Turbo V source and was operated in low mass and MRM mode with electrospray ionization polarity switching. The source and gas settings were: curtain gas (N2) pressure 35 a.u.; CAD gas (N2) pressure medium for positive mode and low for negative mode; ion spray voltage 5.2 kV in positive mode and −4.5 kV in negative mode; temperature 550°C; declustering potential 60 V (positive mode) and −80 V (negative mode); entrance potential 10 V (positive mode) and −10 V (negative mode).

2.2.3 |. Data processing

Data processing was performed using SCIEX OS™ software and Microsoft Excel for post data analysis. Analyte concentrations were calculated as follows:

Analyte concentration=Analyte area/IS area×IS concentration

where IS denotes the internal standard for a given lipid class.

3 |. RESULTS AND DISCUSSION

3.1 |. Lipid identification and class assignment

A general overview of the untargeted and targeted workflows is shown in Figure 1. To assign confident identification to the lipid classes in the untargeted study, the elution time profiles of the heavy-labeled lipid standard mixture were generated. The sample profile was referenced to the standards to assign elution order of lipid classes (Figure 2). Negative ion mode was observed to be favored by phosphatidylinositol (PI) species, while phosphatidylcholine (PC) and phosphatidylethanolamine (PE) species were more abundant in positive ion mode. The assigned annotations and lipid classes were confirmed by matching the MS/MS spectra of the lipid to the expected fragmentation patterns documented in the literature, when available, and also filtering based on the fragmentation score assigned by Progenesis QI for individual lipids.3134

FIGURE 1.

FIGURE 1

Experimental workflow. Lipids were extracted from 30 μL of plasma and reconstituted using assay-specific solvent with the addition of internal standards. Extracted lipids were analyzed using Thermo Orbitrap Q Exactive HF and SCIEX QTRAP 6500+ mass spectrometers and the resulting raw files were analyzed using software specific to the approach [Color figure can be viewed at wileyonlinelibrary.com]

FIGURE 2.

FIGURE 2

Total ion current chromatograms in positive and negative ionization mode showing regions where lipids of different classes elute in an untargeted approach. Lipid class elution in A, positive and B, negative ion mode. Elution order determined by analyzing internal standard by itself and comparing the elution order with that of the actual samples. (DG, diacylglycerol; LPC, lysophosphatidylcholine; LPE, lysophosphatidylethanolamine; PC, phosphatidylcholine; PE, phosphatidylethanolamine; PS, phosphatidylserine; PI, phosphatidylinositol; SM, sphingomyelin; Cer, ceramide; TG, triacylglycerol)

An example MS/MS spectrum for a PC 32:1 is shown in Figure 3A. For PCs, the most common adducts are either the protonated form [M + H]+ or alkali metal adduct [M + Na]+ or [M + Li]+. Here, we observed the characteristic PC product ion peak at m/z 184.073, for the [phosphocholine + H]+ species. There are also product ions present at m/z 695.462, [M + Na − 59]+, which corresponds to the loss of a trimethylamine, and m/z 571.469, [M + Na − 183]+, which corresponds to loss of a phosphocholine. Similarly, PE species can form either a protonated adduct [M + H]+ or alkali metal adduct [M + Na]+ or [M + Li]+. In addition, a product ion at m/z 599.503 [M + H − 141]+ was observed, corresponding to the loss of phosphoethanolamine (Figure 3B). For sphingomyelins (SMs), we observed [M + Na]+ species and product ion peaks formed by the neutral loss of trimethylamine at m/z 750.571 [M + Na − 59]+, neutral loss of phosphocholine at m/z 626.588 [M + Na − 183]+, and an ion at m/z 184.073 for the [phosphocholine + H]+ species (Figure 3C). Ceramides (Cer) have characteristic peaks for sphingoid base ions at m/z 264.268 and 282.279 and also product ions at m/z 502.499 [M + H − 2H2O]+ and 520.508 [M + H − H2O]+, corresponding to neutral losses of water (Figure 3D). Similar strategies for lipid characterization were applied across all detected lipid classes; these include: phosphatidylserine, lysophosphatidylcholine (LPC), lyso phosphatidylethanolamine (LPE), and others. After removing redundant annotations, the most confident annotation is selected on both the score output provided by Progenesis QI and the manual curation and verification of the MS/MS spectra available in external and internal databases. In this study, exact mass similarity, isotope similarity, and fragmentation score were used to determine annotation score.

FIGURE 3.

FIGURE 3

MS/MS fragmentation patterns of different lipids. Example MS/MS spectra from positive mode analysis of A, PC 32:1; B, PE 36:1; C, SM 40:1; and D, Cer 34:1 from the Thermo Orbitrap Q Exactive-HF

In the targeted lipidomics workflow, a previously established MRM method was used to target a large set of ∼ 1150 lipids using the SCIEX platform.35,36 The targeted analysis was accomplished using a “global MRM list” of ∼ 1150 lipids which constitute the most commonly identified lipids in human plasma from 19 different classes of lipids. Details of the MRM list are provided in Table S2 (supporting information). These lipids are analyzed using both positive and negative ion mode in the same run using the fast polarity switching mode of the QTRAP 6500+. This method has the advantage of identifying lipids at the molecular level, especially for phospholipids in which instead of using the loss of head group in the positive ion mode for lipid identification, it uses the loss of fatty acid chains from lipids in the negative ion mode.21 The samples were analyzed using a Scheduled MRM™ algorithm (SCIEX), which incorporates performing a series of unscheduled and scheduled analyses by injecting a QC sample, which represents similar complexity to all samples in the experiment. This is accomplished with a HILIC separation which separates the lipids on the basis of head group functionality (Figure 4) and recently has been shown to have better utility in separating lipid classes than RP separation.37 For example, TG, Cer, and CE elute at retention time (tR) 2.3 min; PS, PE, and PC at tR 9.8 min; LPS, LPI, SM, and LPG at tR 12.0–13.5 min; and LPC, PI, and LPE at tR 13.0 min (Figure 4A). One of the major benefits of using HILIC separation is its high reproducibility in elution times across triplicate injections (3 days) of the individual plasma samples (Figure 4B). Although there are differences in the intensities across some injections, these are mostly due to the intensity differences of the patient samples. In addition, as the endogenous lipids in the sample are co-eluting with the corresponding internal standard, differences in ionization efficiencies and ion suppression effects can be accounted. After ionization, the lipids are scanned in Q1, fragmented in Q2, and fatty acid specific scans are conducted in Q3. An example is shown in Figure 4C, in which the PE species elute at tR 9.8 min. Upon completion of fatty acid specific scans and data processing, the peaks at specific m/z values are annotated to PE species (Figure 4D). Similar strategies were used to annotate all the lipid classes in the targeted MRM method. A list of all identified lipids in both approaches is provided in Table S3 (supporting information). The assigned fatty acid composition of lipids in the untargeted analysis for PC, PE, PI, and TG species is based on fragmentation scores provided by Progenesis QI software, while LipidMaps naming convention was used for lipid annotations.38

FIGURE 4.

FIGURE 4

Chromatographic profiles of lipids from targeted MRM. A, Total ion chromatogram of all the lipid classes detected; B, overlayed chromatograms across 15 plasma injections; C, total ion chromatograms for all lipid classes with the highlighted portion for PE species which eluted at tR 9.8 min; and D, integrated MS spectra from (C) with peaks labeled according to annotations based on the fatty acid masses specific to the species matched from the MRM list with its corresponding Q1 and Q3 masses (listed as Q1/Q3). (TG, triacylglycerol; CE, cholesterol ester; Cer, ceramide; HexCer, hexosylceramide; LacCer, lactosylceramide; PG, phosphatidylglycerol; PS, phosphatidylserine; PE, phosphatidylethanolamine; PC, phosphatidylcholine; LPI, lysophosphatidylinositol; LPS, lysophosphatidylserine; LPC, lysophosphatidylcholine; PI, phosphatidylinositol; LPE, lysophosphatidylethanolamine)

3.2 |. Overall performance evaluation

We evaluated the performance of the untargeted and targeted approaches by comparing the overall variation in number of lipids identified against previous studies performed on human plasma samples.39,40 Specifically, we compared the coverage of the lipid classes common between the two approaches and also common with previous studies.39,40 Overall, 297 lipids were annotated using the untargeted approach, while 619 lipids were annotated using the targeted MRM approach from 11 classes of lipids (CE, cholesterol ester; Cer, ceramide; DG, diacylglycerol; LPC, lysophosphatidylcholine; LPE, lysophosphatidylethanolamine; PC, phosphatidylcholine; PE, phos phatidylethanolamine; PI, phosphatidylinositol; PS, phosphatidylserine; SM, sphingomyelin; TG, triacylglycerol). In the targeted MRM approach the lipids were annotated based on their individual fatty acid chain information, while, in the untargeted study, fatty acid chains were reported as a total number of carbons added together with the corresponding degree of unsaturation, identical to previous LipidMaps39 and NIST40 studies. To better compare the two approaches, the lipid annotations in the MRM approach were converted into the same annotations. In doing so, the total number of annotated lipids was reduced from 619 to 327. In total, 465 lipids were identified in both approaches; 159 lipids were similar between the two approaches (Figure 5A), 168 lipids were unique to the targeted MRM approach, and 138 lipids were unique to the untargeted approach. Figure 5B shows the distribution of lipid classes in both approaches. A majority of the lipids in the untargeted approach were from the PC and PE lipid classes. On the other hand, TG species were the major lipid class in the targeted method as would be expected from plasma samples, along with high numbers of PC and PE species. The numbers of LPC and LPE species identified in each approach were similar. A comparison of the different lipid classes annotated between the two approaches is shown in Figure 5C. The targeted approach had a higher number of TG species identified than the untargeted approach, due to lack of ammoniated buffer, while the numbers of SMs and PCs were higher in the untargeted approach. There were similar numbers of lipids from PE, PI, LPC, and LPE classes of lipids in both approaches, while there were higher numbers of lipids from Cer and PS lipid classes in the targeted approach. As for the common lipids identified between the two approaches, most identified species were from the PE, TG, and PC classes (Figure 5D). PI and PE had similar numbers of unique lipids for both approaches. As for other classes, CE had no unique lipids in the untargeted approach, and both LPE and LPC had a higher number of lipids that were common between the two approaches as opposed to those that were unique in each approach (Figure 5D).

FIGURE 5.

FIGURE 5

Distribution of lipid classes in both approaches. A, Overlap of the lipid identified between the two approaches; B, pie chart showing the different classes of lipids identified in both of the methods; C, comparison of the number of lipids from different classes in both approaches; and D, total number of lipids identified per lipid class in both approaches. (CE, cholesterol ester; Cer, ceramide; DG, diacylglycerol; LPC, lysophosphatidylcholine; LPE, lysophosphatidylethanolamine; PC, phosphatidylcholine; PE, phosphatidylethanolamine; PI, phosphatidylinositol; PS, phosphatidylserine; SM, sphingomyelin; TG, triacylglycerol)

Our results were compared against previous studies of human plasma and serum by LipidMaps39 and NIST.40 Both those studies are considered benchmarks for estimating concentrations of lipid species from several lipid classes and also for reporting the lipid species and lipid class complexity in these sample types. Although the total numbers of identified lipids in both those studies were higher (1527 lipids), especially in the NIST study,40 we focused our comparisons against the most commonly observed lipids in human serum or plasma. In addition, the lipid annotations herein were reformatted to match that of those studies. For example, in the LipidMaps study, Cer species were reported as a corresponding sphingoid base; this formatting is similar to our targeted approach. In the NIST study, however, they were reported as the total number of carbons in the fatty acid chains, similar to the annotations in the untargeted study. Once the necessary conversion was complete, 142 and 191 lipids were common in the untargeted approach compared with the LipidMaps and NIST studies, respectively (Figures 6A and 6C). In the targeted MRM approach, 189 and 208 lipids were common with the LipidMaps and NIST studies, respectively (Figures 6B and 6D). When comparing the numbers for individual lipid classes against the LipidMaps study, higher numbers of Cer, PS, and DG species were common with the targeted MRM study than in the untargeted study, while the number of SM species in common were higher in the untargeted study. The remaining classes had similar numbers for both approaches. On the other hand, higher numbers of PC and SM species were common with the NIST study for the untargeted approach, while there were higher numbers of TG, Cer, and PE species common with the NIST study for the targeted MRM approach (Figures 6E6H).

FIGURE 6.

FIGURE 6

Comparison of the lipid classes identified for the two approaches against previous studies. Overlap of the lipids identified between A, LipidMaps study and untargeted study; B, LipidMaps study and targeted study; C, NIST study and untargeted study; and D, NIST study and targeted study. (E, F) Comparison of lipids identified in the untargeted and targeted approach against the LipidMaps study; (G, H) comparison of lipids identified in the untargeted and targeted approach against the NIST study. (CE, cholesterol ester; Cer, ceramide; DG, diacylglycerol; LPC, lysophosphatidylcholine; LPE, lysophosphatidylethanolamine; PC, phosphatidylcholine; PE, phosphatidylethanolamine; PI, phosphatidylinositol; PS, phosphatidylserine; SM, sphingomyelin; TG, triacylglycerol)

We further compared the two approaches through evaluation of the inter-day relative abundances and concentrations of lipid species across three days for five individual patient plasma samples. The ratio of the MS signal for a given endogenous lipid to the corresponding spiked-in internal standard in the same lipid class was calculated. Because the - internal standard mixture was used in both the untargeted and targeted approaches, this accounted for differences related to separation and ionization of lipids. However, we note that matrix effects are complex between the HILIC and RP separation methods and may require more complex normalization approaches. Due to the higher injection amount of -internal standards for some - classes in the MRM approach, the final ratio values were much smaller than with the untargeted approach of the same species. Despite this, the ratios were more consistent for the MRM approach than the untargeted approach. Overall, the lipid classes had a higher percentage coefficient of variation (CV) across the three-day injections for the untargeted approach than for the targeted MRM approach. For example, the LPC species had a 17% CV for the untargeted approach, whereas the targeted MRM approach had a 6.4% CV. Similarly, the LPE species had a 16% CV and 3% CV for the untargeted and targeted MRM approach, respectively. Overall, the CV for the ratio of six most abundant lipid classes (PC, PE, LPC, LPE, TG, and SM) was ∼ 9% for the targeted MRM approach, whereas it was ∼ 23% for the untargeted approach. For plasma samples, there is inherent biological variation across patients, or in this study healthy volunteers. We assessed inter-day variation in plasma lipid concentrations (nmol/mL of plasma) for both the untargeted and the targeted approaches, and show results from four example lipid species across several classes. As the data were collected on patient samples, there was variation in the ratio values across the samples (Figure 7). This was evident in all lipid species identified in both approaches. Generally, the plasma concentrations determined from the MRM analyses were more stable across the three days than those determined from the untargeted approach. For example, PC 38:3 had average concentrations of 52.06, 47.2, and 68.1 nmol/mL plasma for days 1, 2, and 3, respectively, in the untargeted approach; for the MRM approach, the concentrations were 32.5, 31.4, and 34.2 nmoL/mL plasma for days 1, 2, and 3, respectively (Figure 7A). It should be noted that one patient sample had PC 38:3 concentrations that were noticeably higher than those of the other four patient samples. Other species such as PE 36:0, LPC 20:4, and LPE 20:3 also had more consistent average concentrations with the MRM approach than the untargeted approach across the three days (Figures 7B7D). Thus, these results are consistent with the lower CV values observed overall for the MRM approach.

FIGURE 7.

FIGURE 7

Concentrations of lipid species from different lipid classes demonstrating the inter-day variation of patient sample across the three days for both the targeted and untargeted platforms. Example of lipid species from A, phophatidylcholine (PC); B, phosphatidylethanolamine (PE); C, lysophophatidylcholine (LPC); and D, lysophasphotidylethanolamine (LPE) lipid classes. The five different color points represent individual patient samples, while the black points represent the average concentration among all the patient samples for that lipid species for that day

We also compared the relative total quantities of the six most abundant lipid classes in plasma (Figure 8A), which were calculated by multiplying the ratio of the abundance of the endogenous species and its corresponding internal standard by the amount of standard injected. In most cases the amounts (ng) were similar in both approaches, except for the SM species. In both approaches, LPE had the lowest amount (∼ 1 ng) of lipids, while TG had the highest amount (∼ 500–700 ng). We also compared the sum concentration of lipid species (nmol/mL plasma) in our study against that of the LipidMaps39 and NIST40 studies (Figure 8B) for the PC, PE, LPC, LPE, SM, and TG classes. Only the lipid species that were common among the studies were compared. For the LPC and SM classes, all studies had reasonably similar values. For the PC class the untargeted approach had similar values to the NIST study, while the targeted MRM approach and LipidMaps studies had similar outcomes. On the other hand, there were similarities between the untargeted and LipidMaps study for the PE class, while the targeted MRM approach and NIST study showed similar results. For the LPE class, LipidMaps had higher values. The relative concentrations of lipids common between the untargeted and MRM approaches are given in Figure S4 (supporting information).

FIGURE 8.

FIGURE 8

A, Box plot showing the sum of relative amount of lipids compared with the corresponding internal standard for phosphatidylcholine, phosphatidylethanolamine, lysophophatidylcholine, lysophosphatidylethanolamine, sphingomyelin, and triacylglycerol classes in both the untargeted and targeted (MRM) approaches. B, Bar chart of comparison of the sum of lipid species concentration of our study against the LipidMaps and NIST studies (N = lipids species common among the studies being compared) for the same classes as in (A)

While the untargeted and targeted MRM approaches were able to identify lipids from similar classes and had similar total numbers of observed lipids, the lipid assignments were less specific with the untargeted approach. In this untargeted approach, we were unable to distinguish the individual fatty acid chains. This was expected and could be in part due to limitations in database annotations and searching.20 For example, in the targeted MRM approach, PC, PE, PI, and PS species were identified up to their individual fatty acid chains, which was unavailable in the case of the untargeted approach, which reports it as a sum of the total carbons in the fatty acid chains. Similarly, in the targeted MRM approach, SM species were assumed to have an 18:1 sphingosine backbone due to its higher abundance in human plasma and reports the other fatty acid chain; however, the untargeted approach is able to identify both fatty acid chains. Similarly, TG species in the targeted MRM approach report only one fatty acid chain and the remaining chain as the total number of fatty acids, while the untargeted approach has the capability to identify all three fatty acid chains individually. We also acknowledge that the use of HILIC and RP separation in different approaches could have also influenced the number of lipids identified, especially for the untargeted approach where the number of identified lipids could have increased with the use of longer gradient times. Buffer composition can have an impact on the species detected and we note that a lower number of TG species was identified in the untargeted study although they have been previously reported to be higher in plasma samples. The incorporation of an ammoniated buffer could increase the observation of this lipid class. Despite all the positives in the targeted MRM approach, it does not account for any isotope correction, which is a limitation of the method. In addition, the targeted MRM approach has been developed for human plasma samples and its application for other samples requires further method development. Furthermore, the untargeted approach can identify more lipid classes in addition to the 11 lipid classes included here. Despite these limitations, the streamlined nature of the targeted MRM data analysis makes this approach very attractive. The major advantage of the targeted MRM method is its capability of reporting each lipid at its molecular species level, unlike the untargeted method, where the lipids are reported as the sum composition of their respective fatty acids. In addition, each lipid in the MRM approach has been pre-verified with standards and fragmentation, resulting in a higher confidence level for the identifications. Further development could include improvements in assays, such as use of the SelexION capabilities of the QTRAP 6500+ for the identification and quantification of lipid species; improvement in lipid annotations, validated identifications, and isotopic corrections are also necessary. Data analysis in the targeted MRM approach is much simpler (albeit this is subjective), straightforward, and less time consuming. Overall, these considerations make the targeted MRM approach highly attractive for plasma lipid analysis.

While an ideal study would include a direct comparison of both HILIC and RP separation methods on each MS analyzer used in this study, our focus was to determine the general pros and cons of these two entire platforms including different separation methods and MS analyzers. Others have recently reported direct comparisons of NIST human plasma standards on HILIC and RP on the same Q Exactive Plus mass spectrometer. Those studies provide further evidence that HILIC and RP separation can yield similar quantification for several lipid classes such as PE, LPE, and SM; however, overestimation of lipid concentrations for LPCs may occur with HILIC.41 The RP method offers higher separation power than HILIC, and is the most commonly used method in untargeted lipidomics studies.26,42,43 The separation observed using HILIC, having many species in a lipid class co-elute from the column, is particularly helpful for the analysis of phospholipids and sphingomyelins37 and is better for ensuring that similar ionization and matrix effects occur when deuterated internal standards are used.41

4 |. CONCLUSIONS

Overall, both the untargeted Q Exactive HF and the targeted MRM SCIEX QTRAP 6500+ approaches identified similar numbers of lipids across 11 lipid classes from human plasma. The targeted MRM approach had the advantage of identifying lipids at the molecular level with greater confidence than the untargeted approach. In addition, the targeted MRM approach had a much lower inter-day variability of lipid abundances and concentrations for the patient samples than the untargeted approach. Despite these positives, the targeted MRM approach is limited by the number of lipid transitions and lipid classes it can monitor to date, as it focuses on 1150 lipid transitions. The targeted MRM approach is specific for human plasma samples and would require further method development for other sample types. On the other hand, the untargeted approach identified more unique lipids and also lipids from other classes outside the 11 lipid classes mentioned. This is probably due to the separation power of RP liquid chromatography. In conclusion, the targeted MRM assay developed by SCIEX on the QTRAP 6500+ seems promising for characterizing the lipidome of human plasma samples.

Supplementary Material

RCM_supplemental_final

ACKNOWLEDGEMENTS

The authors acknowledge funding from the Alzheimer’s Association (AARGD-17–533405), the Vanderbilt University Start-Up Funds, pilot funds from the University of Pittsburgh Alzheimer Disease Research Center funded by the National Institutes of Health and National Institute on Aging (P50AG005133, RASR), NICHD (R01 HD064727, NAP), and the Vanderbilt Institute of Chemical Biology (fellowship, M.J.K.). This work was supported in part using the resources of the Center for Innovative Technology at Vanderbilt University. The authors also thank SCIEX for its partnership in the analysis of the targeted MRM samples. The author list includes two authors from SCIEX that contributed to the targeted MRM platforms on the 6500+ MS instrument.

Funding information

National Institutes of Health, Grant/Award Numbers: P50AG005133, R01HD064727, T32-GM065086; Vanderbilt Institute of Chemical Biology; NICHD, Grant/Award Number: R01 HD064727; National Institutes of Health and National Institute on Aging, Grant/Award Number: P50AG005133; Alzheimer’s Association, Grant/Award Number: AARGD-17–533405

Footnotes

SUPPORTING INFORMATION

Additional supporting information may be found online in the Supporting Information section at the end of this article.

REFERENCES

  • 1.Suarna C, Dean RT, May J, Stocker R. Human atherosclerotic plaque contains both oxidized lipids and relatively large amounts of alphatocopherol and ascorbate. Arterioscler Thromb Vasc Biol 1995;15(10): 1616–1624. [DOI] [PubMed] [Google Scholar]
  • 2.Han X, Yang J, Yang K, Zhao Z, Abendschein DR, Gross RW. Alterations in myocardial cardiolipin content and composition occur at the very earliest stages of diabetes: A shotgun lipidomics study. Biochemistry 2007;46(21):6417–6428. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 3.Miyoshi N, Iwasaki N, Tomono S, Higashi T, Ohshima H. Occurrence of cytotoxic 9-oxononanoyl secosterol aldehydes in human low-density lipoprotein. Free Radical Biol Med 2013;60:73–79. [DOI] [PubMed] [Google Scholar]
  • 4.Ollero M, Astarita G, Guerrera IC, et al. Plasma lipidomics reveals potential prognostic signatures within a cohort of cystic fibrosis patients. J Lipid Res 2011;52(5):1011–1022. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 5.Fhaner CJ, Liu S, Ji H, Simpson RJ, Reid GE. Comprehensive lipidome profiling of isogenic primary and metastatic colon adenocarcinoma cell lines. Anal Chem 2012;84(21):8917–8926. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 6.Gorke R, Meyer-Base A, Wagner D, He H, Emmett MR, Conrad CA. Determining and interpreting correlations in lipidomic networks found in glioblastoma cells. BMC Syst Biol 2010;4:126. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 7.Ishikawa S, Tateya I, Hayasaka T, et al. Increased expression of phosphatidylcholine (16:0/18:1) and (16:0/18:2) in thyroid papillary cancer. PLoS One 2012;7(11):e48873. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 8.Zhou X, Mao J, Ai J, et al. Identification of plasma lipid biomarkers for prostate cancer by lipidomics and bioinformatics. PLoS One 2012;7 (11):e48889. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 9.Yu Z, Chen H, Zhu Y, et al. Global lipidomics reveals two plasma lipids as novel biomarkers for the detection of squamous cell lung cancer: A pilot study. Oncol Lett 2018;16(1):761–768. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 10.Adibhatla RM, Hatcher JF. Role of lipids in brain injury and diseases. Future Lipidol 2007;2(4):403–422. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 11.Di Paolo G, Kim T-W. Linking lipids to Alzheimer’s disease: Cholesterol and beyond. Nat Rev Neurosci 2011;12(5):284–296. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 12.Adibhatla RM, Hatcher JF. Altered lipid metabolism in brain injury and disorders. Subcell Biochem 2008;49:241–268. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 13.Liu Q, Zhang J. Lipid metabolism in Alzheimer’s disease. Neurosci Bull 2014;30(2):331–345. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 14.Kolsch H, Heun R, Jessen F, et al. Alterations of cholesterol precursor levels in Alzheimer’s disease. Biochim Biophys Acta 2010;1801(8): 945–950. [DOI] [PubMed] [Google Scholar]
  • 15.Xianlin H Lipid alterations in the earliest clinically recognizable stage of Alzheimers disease: Implication of the role of lipids in the pathogenesis of Alzheimers disease. Curr Alzheimer Res 2005;2(1): 65–77. [DOI] [PubMed] [Google Scholar]
  • 16.Wood PL, Cebak JE, Woltjer RL. Diacylglycerols as biomarkers of sustained immune activation in proteinopathies associated with dementia. Clin Chim Acta 2018;476:107–110. [DOI] [PubMed] [Google Scholar]
  • 17.Wong MW, Braidy N, Poljak A, Sachdev PS. The application of lipidomics to biomarker research and pathomechanisms in Alzheimer’s disease. Curr Opin Psychiatry 2017;30(2):136–144. [DOI] [PubMed] [Google Scholar]
  • 18.Stepler KE, Robinson RAS. The potential of ‘omics to link lipid metabolism and genetic and comorbidity risk factors of Alzheimer’s disease in African Americans. In: Guest PC, ed. Reviews on Biomarker Studies in Psychiatric and Neurodegenerative Disorders Switzerland: Springer Nature; 2018. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 19.Wong MW, Braidy N, Poljak A, Pickford R, Thambisetty M, Sachdev PS. Dysregulation of lipids in Alzheimer’s disease and their role as potential biomarkers. Alzheimers Dement 2017;13(7): 810–827. [DOI] [PubMed] [Google Scholar]
  • 20.Bou Khalil M, Hou W, Zhou H, et al. Lipidomics era: Accomplishments and challenges. Mass Spectrom Rev 2010;29(6):877–929. [DOI] [PubMed] [Google Scholar]
  • 21.Mackenzie Pearson SKK, Wu S, Baker PR. Achieve broad lipid quantitation using a high-throughput targeted lipidomics method. SCIEX 2018.
  • 22.Wang M, Han X. Advanced shotgun lipidomics for characterization of altered lipid patterns in neurodegenerative diseases and brain injury. Methods Mol Biol 2016;1303:405–422. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 23.Bilgin M, Markgraf DF, Duchoslav E, et al. Quantitative profiling of PE, MMPE, DMPE, and PC lipid species by multiple precursor ion scanning: A tool for monitoring PE metabolism. Biochim Biophys Acta 2011;1811(12):1081–1089. [DOI] [PubMed] [Google Scholar]
  • 24.Ejsing CS, Sampaio JL, Surendranath V, et al. Global analysis of the yeast lipidome by quantitative shotgun mass spectrometry. Proc Natl Acad Sci 2009;106(7):2136–2141. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 25.Reis A, Rudnitskaya A, Blackburn GJ, Mohd Fauzi N, Pitt AR, Spickett CM. A comparison of five lipid extraction solvent systems for lipidomic studies of human LDL. J Lipid Res 2013;54(7):1812–1824. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 26.Breitkopf SB, Ricoult SJH, Yuan M, et al. A relative quantitative positive/negative ion switching method for untargeted lipidomics via high resolution LC-MS/MS from any biological source. Metabolomics 2017;13(3):30. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 27.Gibson CL, Codreanu SG, Schrimpe-Rutledge AC, et al. Global untargeted serum metabolomic analyses nominate metabolic pathways responsive to loss of expression of the orphan metallo beta-lactamase, MBLAC1. Mol Omics 2018;14(3):142–155. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 28.Sud M, Fahy E, Cotter D, et al. LMSD: LIPID MAPS structure database. Nucleic Acids Res 2007;35(suppl_1):D527–D532. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 29.Kind T, Liu K-H, Lee DY, DeFelice B, Meissen JK, Fiehn O. LipidBlast in silico tandem mass spectrometry database for lipid identification. Nat Methods 2013;10(8):755–758. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 30.Wishart DS, Tzur D, Knox C, et al. HMDB: The human metabolome database. Nucleic Acids Res 2007;35(Database issue): D521–D526. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 31.Pulfer M, Murphy RC. Electrospray mass spectrometry of phospholipids. Mass Spectrom Rev 2003;22(5):332–364. [DOI] [PubMed] [Google Scholar]
  • 32.Hsu F-F, Turk J Electrospray ionization with low-energy collisionally activated dissociation tandem mass spectrometry of glycerophospholipids: Mechanisms of fragmentation and structural characterization. J Chromatogr B 2009;877(26):2673–2695. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 33.Han X Lipidomics: Comprehensive Mass Spectrometry of Lipids New York: John Wiley & Sons; 2016. [Google Scholar]
  • 34.Murphy RC. Tandem Mass Spectrometry of Lipids: Molecular Analysis of Complex Lipids London: Royal Society of Chemistry; 2014. [Google Scholar]
  • 35.Wang J, Fu L, Gorti SKK, et al. Development of an amide column-based LC-MS/MS method for high-throughput targeted lipidomics ASMS Conference on Mass Spectrometry and Allied Topics; San Diego, CA; 2018. [Google Scholar]
  • 36.Gorti SKK, Merkel D, Fu L, et al. HILIC LC-MSMS assay for high-throughput targeted lipdomics analysis Lipidomics Forum; Dortmund, Germany; 2018. [Google Scholar]
  • 37.Ritchie M, Mal M, Wong S. UPLC BEH HILIC: The preferred separation chemistry for targeted analysis of phospholipids. Application Note Waters Corporation, Milford, MA, USA, 2012. [Google Scholar]
  • 38.Liebisch G, Vizcaino JA, Kofeler H, et al. Shorthand notation for lipid structures derived from mass spectrometry. J Lipid Res 2013;54(6): 1523–1530. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 39.Quehenberger O, Armando AM, Brown AH, et al. Lipidomics reveals a remarkable diversity of lipids in human plasma. J Lipid Res 2010;51 (11):3299–3305. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 40.Bowden JA, Heckert A, Ulmer CZ, et al. Harmonizing lipidomics: NIST interlaboratory comparison exercise for lipidomics using SRM 1950– metabolites in frozen human plasma. J Lipid Res 2017;58(12):2275–2288. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 41.Lange M, Fedorova M. Evaluation of lipid quantification accuracy using HILIC and RPLC MS on the example of NIST® SRM® 1950 metabolites in human plasma. Anal Bioanal Chem 2020;412(15): 3573–3584. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 42.Contrepois K, Mahmoudi S, Ubhi BK, et al. Cross-platform comparison of untargeted and targeted lipidomics approaches on aging mouse plasma. Sci Rep 2018;8(1):17747. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 43.Cajka T, Smilowitz JT, Fiehn O. Validating quantitative untargeted lipidomics across nine liquid chromatography–high-resolution mass spectrometry platforms. Anal Chem 2017;89(22):12360–12368. [DOI] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

RCM_supplemental_final

RESOURCES