MS-DIAL 5 multimodal mass spectrometry data mining unveils lipidome complexities

Abstract

Lipidomics and metabolomics communities comprise various informatics tools; however, software programs handling multimodal mass spectrometry (MS) data with structural annotations guided by the Lipidomics Standards Initiative are limited. Here, we provide MS-DIAL 5 for in-depth lipidome structural elucidation through electron-activated dissociation (EAD)-based tandem MS and determining their molecular localization through MS imaging (MSI) data using a species/tissue-specific lipidome database containing the predicted collision-cross section values. With the optimized EAD settings using 14 eV kinetic energy, the program correctly delineated lipid structures for 96.4% of authentic standards, among which 78.0% had the sn-, OH-, and/or C = C positions correctly assigned at concentrations exceeding 1 μM. We showcased our workflow by annotating the sn- and double-bond positions of eye-specific phosphatidylcholines containing very-long-chain polyunsaturated fatty acids (VLC-PUFAs), characterized as PC n-3-VLC-PUFA/FA. Using MSI data from the eye and n-3-VLC-PUFA-supplemented HeLa cells, we identified glycerol 3-phosphate acyltransferase as an enzyme candidate responsible for incorporating n-3 VLC-PUFAs into the sn1 position of phospholipids in mammalian cells, which was confirmed using EAD-MS/MS and recombinant proteins in a cell-free system. Therefore, the MS-DIAL 5 environment, combined with optimized MS data acquisition methods, facilitates a better understanding of lipid structures and their localization, offering insights into lipid biology.

Fig. 1. Electron-activated dissociation (EAD)-based tandem mass spectrum facilitates efficient lipid structure elucidation.

a Spectrum entropy value distributions for 716 small molecules, with the x- and y-axes representing spectrum entropy and fragmentation conditions, respectively. b EAD-MS/MS spectra of 1,2-dioleoyl-sn-glycero-3-phosphocholine (DOPC) at kinetic energies (KE) of 10 and 14 eV, highlighting only the hydrogen (H) loss (blue), radical (black), and H-gain (red) fragment ions related to acyl chain properties. The proposed mechanism explaining the increased abundance of H-loss and radical fragments is also depicted. c EAD-MS/MS spectrum of 1,2-dioleoyl-sn-glycero-3-phosphatidylethanolamine (DOPE) at a 14 eV KE. d EAD-MS/MS spectra of 1,2-diarachidonoyl-sn-glycero-3-phosphocholine (DAPC) at KEs of 10, 14, and 18 eV. The mechanism behind the observed increase in H-gain fragment abundance at the delta-6 and 9 carbon positions is also illustrated. Numbers atop each fragment ion denote the carbon count remaining in a single acyl chain. Source data are provided as a Source Data file.

Fig. 2. Evaluation of dynamic ranges for in-depth lipid annotation with MS-DIAL 5.

a Dynamic range and limit of detection (LOD) required to confirm the presence of diagnostic ions for lipid structure elucidation. The x-axis represents the on-column volume of 1,2-dilinoleoyl-sn-glycero-3-phosphocholine (DLPC) and 1-palmitoyl-2-arachidonoyl-sn-phosphatidylcholine (PAPC), while the y-axis shows the peak heights of diagnostic ions from the centroided product ion spectrum obtained from the product ion scanning mode. The response values of important fragment ions were investigated; for instance, “C10 H loss_1 × 10³” and “SN1 18:2_5 × 10²” denote the LOD values of the H-loss fragment ion at the C10 position and the neutral loss (NL) of sn1-18:2 to characterize the sn-position at 1000 and 500 femtomoles (fmol) on-column volumes, respectively. Even in the authentic standard of PAPC, ions related to sn1-20:4 are detected due to chemical impurities or conformational changes during sample preparation. b Relationship between annotation level terminology and lipid description. The cases of PC and SM were described. c Validation of the MS-DIAL 5 environment for lipid structure description based on EAD-MS/MS spectra quality. Dilution series were analyzed three times at each concentration. The representative annotation was determined as follows: if the same lipid name was annotated in at least two of the three replicates, that name was used as the representative annotation. If the annotation results differed across all three replicates, the lipid with the highest score was adopted as the representative. For example, “x10” indicates a dilution 10 times less concentrated than the original, denoted as “x1.” For sphingolipids, green and red circles represent annotations where OH-positions and both OH- and C = C positions are resolved, respectively. For glycerophospholipids, green and red circles indicate annotations of sn-positions and both sn- and C = C-positions, respectively. Blue, orange, and yellow circles represent annotations at the C = C position resolved, molecular species, and species levels, respectively. If the MS/MS spectrum was not assigned to the precursor ion by DDA, a square shape is used. Incorrect annotations are shown as white fills with a border color indicating the source of the misannotation. Source data are provided as a Source Data file.

Fig. 3. MS-DIAL evaluation to decipher the mixed spectra of PC 16:0_18:1 lipid isomers.

a Validation of the MS-DIAL 5 environment for lipid annotation in the mixed spectra of co-eluted peaks PC 16:0/18:1(9) and PC 16:0/18:1(11). The “C = C high” peaks of the H-loss fragment ions for PC 16:0/18:1(9) and PC 16:0/18:1(11) were marked with blue and red colors, respectively. The “C = C high” peaks of the radical fragment ions for these were marked with sky-blue and orange colors, respectively. The V-shaped patterns were also described, where the valley peaks for PC 16:0/18:1(9) and PC 16:0/18:1(11) are detected at m/z 634.4442 and m/z 662.4755, respectively. The valley peak of PC 16:0/18:1(9) is completely consistent with the peak of the “C = C high” H-loss fragment ion of PC 16:0/18:1(11). When the annotations of MS-DIAL 5 were matched with the names of PC 16:0/18:1(9) or PC 16:0/18:1(11), the blue and orange check symbols were described in the upper right position of the MS/MS spectrum. The results of technical replicates 1, 2 and 3 described in Supplementary Data 7 are described by the left, middle and right symbols, respectively. b Validation of the MS-DIAL 5 environment for lipid annotation in the mixed spectra of co-eluted peaks PC 16:0/18:1(9) and PC 18:1(9)/16:0. The sn1-specific ions for PC 16:0/18:1(9) and PC 18:1(9)/16:0 are marked with blue and pink colors respectively. When the annotations of MS-DIAL 5 were matched with the names of PC 16:0/18:1(9) or PC 18:1(9)/16:0, the blue and pink check symbols were described at the top right position of the MS/MS spectrum. The definition of the symbol replicates is the same as in Fig. 3a. c Validation of the MS-DIAL 5 environment for lipid annotation in mouse brain and SRM 1950 NIST human plasma lipid extracts. Color definitions are the same as in Fig. 3a and Fig. 3b. d Validation of the MS-DIAL 5 environment for lipid annotations in mouse brain lipid extract spiked with PC 16:0/18:1(8) at different concentrations. The ‘X’ symbol indicates that the MS-DIAL annotation did not match any of the PC 16:0/18:1(9), PC 16:0/18:1(11), PC 18:1(9)/16:0 and PC 18:1(11)/16:0. Source data are provided as a Source Data file.

Fig. 4. Structural elucidations of very long-chain PUFA-containing phosphatidylcholine (PC).

The result of in-depth lipidome profiling is shown by the scatter plot of retention time- and m/z axis. The annotation results of molecular species (MSL), double-bond (DB) resolved, sn- or OH-positions (SN or OH) resolved, and both sn- and DB- or both OH- and DB-position (SN + DB or OH + DB) resolved levels are described by the same color charts used in Fig. 2b. The sn1-position determined or uncharacterized for VLC-PUFA is described by triangle and diamond symbols, respectively. The bottom-right panel describes the experimental spectrum of the lipid ion annotated as PC 34:6(16,19,22,25,28,31)/22:6(4,7,10,13,16,19), where the E/Z isomer definition in acyl chains is unsupported, yet a representative form is shown. The top panel displays a 50-fold zoomed experimental spectrum and a 10-fold zoomed in silico spectrum of the assigned lipid in the upper and lower panels, respectively. Brown, green, orange, and red spectral peaks represent ions related to homolytic cleavages in acyl chains, lyso PC substructures, neutral loss of sn1-34:6 moiety, and precursor- or polar head-specific fragments, respectively. Source data are provided as a Source Data file.

Fig. 5. Reanalysis of publicly available spatial- and untargeted lipidomics data for mouse eye tissues.

a A sunburst plot summarizing species/tissue-specific lipid database statistics containing collision-cross section (CCS) values. An eye-lipidome table with m/z and CCS values for 525 unique lipids was used to annotate lipids in MSI data analysis. A summary table of peak annotations in the analyzed MSI data is also provided. The abbreviations of FA, GL, GP, PR, SP, and ST mean fatty acyls, glycerolipids, glycerophospholipids, prenol lipids, sphingolipids, and sterol lipids, respectively. The colors in the sunburst plots were automatically generated and do not have any specific meaning. b Hematoxylin and eosin (HE) staining and MSI data in eye tissues from Acsl6^+/- and Acsl6^-/- mice. Ion distributions for five lipid molecules are shown. The reference m/z and CCS values for each lipid molecule are listed, with EAD-MS/MS-based annotations for each precursor m/z value in parentheses (n = 2 biologically independent samples available at the public repository). c Reanalysis of publicly available untargeted lipidomics data examining eye tissues from Acsl6^+/- and Acsl6^-/- mice at 10 weeks and 2 years of age (n = 3 biologically independent samples). Here, “22:6” denotes DHA, while “28:6,” “30:5,” “32:4,” “32:5,” “32:6,” “34:4,” “34:5,” “34:6,” “36:6,” and “38:6” are defined as VLC-PUFAs. An asterisk indicates acyl chains other than DHA and VLC-PUFA, with the sum of lipid molecules labeled ‘*_DHA’ or ‘*_VLC-PUFA’. ‘Other PCs’ refers to the total abundance of PC molecules not containing DHA or VLC-PUFA). Source data are provided as a Source Data file.

Fig. 6. Elucidation of VLC-PUFA PC metabolic pathway.

a HeLa cell lipid profiling with VLC-PUFA (FA n-3-32:6) supplementation (n = 4 biologically independent samples). The peak heights of PC 18:1_32:6, DG 18:1_32:6, LPA 32:6, and TG 18:1_18:1_32:6 at final concentrations of 1, 10, or 40 µM of FA n-3-32:6 supplementation are depicted. While LPA was analyzed by a derivatization method using trimethylsilyl-diazomethane, which converts LPA to bis-methyl LPA (BisMeLPA), other molecules were analyzed using conventional untargeted lipidomics methods. Acyl CoA (b) and LPA (c) profiling for the glycerol 3-phosphate acyltransferase 1 (GPAT1) recombinant enzyme assay (n = 4 independently prepared samples). The acyl CoAs and LPAs were analyzed with the vehicle, mock (native plasmid vector), active GPAT1^WT, and the inactive GPAT1 mutant (GPAT1^H230A), supplied with glycerol 3-phosphate and coenzyme in addition to ¹³C-uniformly labeled palmitic acid (FA 16:0 U-¹³C), docosahexaenoic acid (DHA, FA 22:6), or FA n-3-32:6, in the cell-free system enzymatic reaction. The fatty acid was supplied at a final concentration of 10 μM, and the same amount of 17:1 CoA (Fig. 6b) and LPA 17:1 (Fig. 6c) was supplied as the internal standards. The putative ratio between the converted product and the internal standard was used for the y-axis value of dot plots. The mean values were also described in dot plots. Significances were adjusted by false discovery rate in the student’s t-test (two-sided). Source data are provided as a Source Data file.