Ozone-enabled fatty acid discovery reveals unexpected diversity in the human lipidome

Abstract

Fatty acid isomers are responsible for an under-reported lipidome diversity across all kingdoms of life. Isomers of unsaturated fatty acids are often masked in contemporary analysis by incomplete separation and the absence of sufficiently diagnostic methods for structure elucidation. Here, we introduce a comprehensive workflow, to discover unsaturated fatty acids through coupling liquid chromatography and mass spectrometry with gas-phase ozonolysis of double bonds. The workflow encompasses semi-automated data analysis and enables de novo identification in complex media including human plasma, cancer cell lines and vernix caseosa. The targeted analysis including ozonolysis enables structural assignment over a dynamic range of five orders of magnitude, even in instances of incomplete chromatographic separation. Thereby we expand the number of identified plasma fatty acids two-fold, including non-methylene-interrupted fatty acids. Detection, without prior knowledge, allows discovery of non-canonical double bond positions. Changes in relative isomer abundances reflect underlying perturbations in lipid metabolism.

Fig. 1. OzFAD: A de novo workflow for semi-automated fatty acid analysis with isomer resolution.

a Lipids are extracted from human blood plasma, vernix caseosa or cell cultures. b After hydrolysis of lipids and addition of internal standards, fatty acids are derivatized with a fixed charge. c Liquid chromatography separates derivatized fatty acids, which undergo electrospray ionization (ESI) and are subjected to ozone-induced dissociation (OzID) with subsequent mass analysis (data-independent acquisition: DIA LC-OzID-MS). d Analysis of the DIA LC-OzID-MS dataset is initiated by a windows batch file that controls python scripts and instances of Skyline Runner. First, the retention times of precursors (unreacted, derivatized fatty acids) that can be identified in the dataset are saved. e Second, a target list (for a separate DDA LC-OzID-MS/MS run) is built based on precursor m/z values and retention times. After acquisition, an exhaustive search for all possible double bond positions (i.e., OzID product ions) and an automated filtering is carried out. Manual inspection in Skyline enables deletion of remaining false positives. Relative quantification is based on the DIA data including a manual correction of the deconvolution of extracted ion chromatograms. Finally, a python script formats the data, generates systematic names, retrieves LIPID MAPS IDs and common names from the LIPID MAPS database where available and generates a bar chart. The latter visualizes the relative abundance of fatty acid isomers and their double bond positions. f Direct infusion ESI-MS allows an estimation of abundance of fatty acid groups (no isomer resolution) and, combined with the relative abundance, an estimate of absolute quantities of each fatty acid.

Fig. 2. LC-OzID-MS(/MS) data for selected fatty acids in human plasma.

Data obtained by DIA UPLC-OzID-MS (a and c) and the associated DDA UPLC-OzID-MS/MS acquisition (b and d) from AMPP derivatized fatty acids from hydrolyzed human plasma (NIST 1950 SRM). Shown are extracted ion chromatograms and (tandem) mass spectra consistent with identification of FA 18:1n-9 (oleic acid), FA 18:1n−2, FA 18:1n−3, FA 20:3n−6,9,12 (dihomo-γ-linolenic acid), FA 20:3n−6,9,15 (sciadonic acid), and FA 20:3n−9,12,15 (Mead acid). The extracted ion chromatograms of the data-dependent acquisition contain only a few points across each chromatographic peak, as multiple precursors are mass selected consecutively according to a highly segmented target list. Source data are provided as a Source Data file.

Fig. 3. Analysis of fatty acids in human plasma (NIST 1950 standard reference material).

a Quantification of plasma fatty acids at the sum composition level based on direct infusion ESI-MS. Each bar shows the mean of three technical replicates ± SD, n = 3. Selected saturated fatty acid quantities are shown, for the full profile see the Supplementary Note 3, Table S4. b Color-coded bar chart showing the different fatty acid isomers and their relative abundance as determined by LC-OzID-MS. Each technical replicate is represented by a vertical segment of each bar. Non-methylene-interrupted (NMI) fatty acids are highlighted by patterns distinguishing between butylene-interrupted (Bu) fatty acids and others. c Identified monounsaturated fatty acid species in the NIST 1950 human plasma and their relation between structure and retention time. Each data point represents a fatty acid isomer with double bond position as shown on the y-axis, plotted against dECL, the difference of the observed equivalent chain length (ECL) to the chain length of the associated saturated fatty acid. d Comparison of the numbers of straight chain fatty acids (C12–24) that are reported in human plasma, detected by different analytical methods and summarized literature surveys. Colors highlight fatty acid species that are found in multiple contexts or only reported in one of the displayed contexts. Source data are provided as a Source Data file. *Docosenoic acids (FA 22:1) are excluded from the analysis due to unavoidable contamination from plastics additive erucamide (FA 22:1n−9).

Fig. 4. Discovery of non-methylene-interrupted ω-3 fatty acids in vernix caseosa.

a A proposed biosynthetic pathway rationalizing the discovery of the never previously reported fatty acids FA 16:2n−3,10 (6,13-hexadecadienoic acid); FA 18:2n−3,10 (8,15-octadecadienoic acid) and FA 18:3n−3,10,13 (5,8,15-octadecatrienoic acid). The double bond configuration is tentatively assigned as cis in each case. b Associated extracted ion chromatograms and OzID MS/MS spectra of the 4-I-AMPP derivatized fatty acids. For the complete fatty acid profile, refer to the Supplementary Note 5, Supplementary Fig. 31 and Supplementary Data 3. Source data are provided as a Source Data file.

Fig. 5. Fatty acid profiles of breast (MCF7) and prostate (LNCaP) cancer cell lines, also featuring inhibition of SCD1.

a Estimated fatty acid quantities by direct infusion ESI-MS. Shown are mean values ± SD of three biological replicates (n = 3). Due to interference of ion suppressing contaminants in two replicates of LNCaP, only one is shown here, instead of the mean of three replicates. b Relative quantification of fatty acid isomers in the three cell lines by LC-OzID-MS and LC-OzID-MS/MS. For each isomer group, a segmented bar is shown, where the segments on the left represent fatty acids in MCF7 cancer cell line extracts, the segments in the middle of each bar represent fatty acids in LNCaP cancer cell line extracts and segments on the right, respectively, represent LNCaP_SCD−1i cancer cell line extracts. Shown are mean values of relative abundances of three biological replicates of each cell line. For individual values and standard deviations, refer to the Supplementary Information. c Example eicosadienoic acids in either MCF7 or the LNCaP cell lines. d Volcano plots visualizing the fold changes (and associated p-values based on two-sided Welsh t-tests) in relative fatty acid isomer abundances as compared between MCF7 and LNCaP (left volcano plot) and between LNCaP and LNCaP_SCD1i (volcano plot on the right), see also Supplementary Figs. 43 and 44 and statistical details in Supplementary Data 8 and 9. Selected isomers are labelled to show examples of large changes (exceeding twofold changes, blue dashed lines) in relative isomer abundances within the respective fatty acid isomer groups with an associated p-value above 0.05 (red dashed line). Source data are provided as a Source Data file.