Preparative supercritical fluid chromatography for lipid class fractionation-a novel strategy in high-resolution mass spectrometry based lipidomics

Abstract

In this work, a lipidomics workflow based on offline semi-preparative lipid class-specific fractionation by supercritical fluid chromatography (SFC) followed by high-resolution mass spectrometry was introduced. The powerful SFC approach offered separation of a wide polarity range for lipids, enabled enrichment (up to 3 orders of magnitude) of lipids, selective fractionation of 14 lipid classes/subclasses, and increased dynamic range enabling in-depth characterization. A significantly increased coverage of low abundant lipids improving lipid identification by numbers and degree (species and molecular level) was obtained in Pichia pastoris when comparing high-resolution mass spectrometry based lipidomics with and without prior fractionation. Proof-of-principle experiments using a standard reference material (SRM 1950, NIST) for human plasma showed that the proposed strategy enabled quantitative lipidomics. Indeed, for 70 lipids, the consensus values available for this sample could be met. Thus, the novel workflow is ideally suited for lipid class-specific purification/isolation from milligram amounts of sample while not compromising on omics type of analysis (identification and quantification). Finally, compared with established fractionation/pre-concentration approaches, semi-preparative SFC is superior in terms of versatility, as it involved only volatile modifiers and salt additives facilitating any follow-up use such as qualitative or quantitate analysis or further purification down to the single lipid species level. Graphical Abstract.

Keywords: Human plasma; Lipid fractionation; Lipidomics; Pichia pastoris; Preparative supercritical fluid chromatography; SRM 1950.

Graphical Abstract

Fig. 1

Overview of the applied workflow including sample preparation, SFC separation, and MS analysis

Fig. 2

Supercritical fluid chromatogram of a a multi-lipid mix and b yeast (Pichia pastoris) detected with ELSD. Peak annotation: TG, triacylglycerols; CE, cholesteryl esters; SE, steryl esters; FA, fatty acids; DG, diacylglycerols; Chol, cholesterol; Ergo, ergosterol; MG, monoacylglycerols; Cer, ceramides (d,t,e, di,tri,tetra hydroxylated); SPH, sphingosine bases, HexCer, hexosyl ceramides; PG, phosphatidylglycerols; Hex2Cer, dihexosyl ceramides; PE, phosphatidylethanolamines; PA, phosphatidic acids; CL, cardiolipins; PI, phosphatidylinositols; LPE, lysophosphatidylethanolamines; PS, phosphatidylserines; PC, phosphatidylcholines; SM, sphingomyelins; LPC, lysophosphatidylcholines. “/” indicates co-eluting lipid classes

Fig. 3

Distribution of each yeast lipid class over the fractionated SFC run. Values are calculated by the summarized area values of the lipid species of each lipid class obtained by LipidSearch. The size of the points accounts for the relative area of each lipid class over all fractions (smaller points indicate a distribution over several fractions). Colors emphasize the different lipid classes on the y-axis

Fig. 4

Distribution of ceramides in the ceramide-containing fractions. Ceramides ordered after the number of hydroxyl groups in the three fractions, where ceramides were identified (fraction 7–9). Lipids marked with asterisk have a hydroxyl group on the fatty amid chain

Fig. 5

Improved lipid identification in yeast with SFC fractionation. The total number of identifications over all fractions of the fractionated yeast (whole bar) compared with identifications in the non-fractionated full extract (purple and white bar) was more than doubled (404 compared with 150). The white bar corresponds to the lipids identified in both fractionated and non-fractionated full extract, but the level of identification was improved in the fractionated yeast (molecular species level, e.g., PC 16:0_16:1 instead of lipid species level, e.g., PC 32:1). The blue bar shows the number of lipids only identified in the fractionated yeast

Fig. 6

Accuracy assessment for SRM 1950 - “Metabolites in Frozen Human Plasma.” Values are presented as normalized coverage equivalents at the mean (dots) and stdev (error bars, N = 2) of measurements, overlaid onto the consensus mean value (blue line) and uncertainty (95% coverage, green region; 99% coverage, red region). a Fractionated with SFC. b Non-fractionated full extract. See ESM Tables S3 and S4 for detailed results. Graphic is produced with LipidQC [58]