LC-MS-Based Targeted Metabolomics for FACS-Purified Rare Cells

Abstract

Metabolism plays a fundamental role in regulating cellular functions and fate decisions. Liquid chromatography-mass spectrometry (LC-MS)-based targeted metabolomic approaches provide high-resolution insights into the metabolic state of a cell. However, the typical sample size is in the order of 10⁵-10⁷ cells and thus not compatible with rare cell populations, especially in the case of a prior flow cytometry-based purification step. Here, we present a comprehensively optimized protocol for targeted metabolomics on rare cell types, such as hematopoietic stem cells and mast cells. Only 5000 cells per sample are required to detect up to 80 metabolites above background. The use of regular-flow liquid chromatography allows for robust data acquisition, and the omission of drying or chemical derivatization avoids potential sources of error. Cell-type-specific differences are preserved while the addition of internal standards, generation of relevant background control samples, and targeted metabolite with quantifiers and qualifiers ensure high data quality. This protocol could help numerous studies to gain thorough insights into cellular metabolic profiles and simultaneously reduce the number of laboratory animals and the time-consuming and costly experiments associated with rare cell-type purification.

Figure 1

General workflow of the optimized protocol.

Figure 2

(A) Shift in the retention time and (B) difference in signal intensity caused by the addition of 5 g/L NaCl or PBS compared to a water reference for 111 metabolites analyzed by HILIC chromatography.

Figure 3

Signal intensity over a number of sorted events for debris (black) and B cells from spleen (blue) for (A) aminoterephthalic acid (ATA, left panel), (B) S-adenosyl-methionine (SAM), and (C) 4-aminobutyric acid. Solid lines and dots indicate the mean, and the hatched area indicates the min-to-max range (n = 8). ATA had been spiked into the cell suspension before cell sorting and thus represented an example of an extracellular compound. A sizable fraction of SAM occurs outside the cells, but the amount detected in intact cells is still larger. 4-Aminoburyric acid occurs almost exclusively inside the cells.

Figure 4

Volcano Plot of differences between Jurkat cells harvested by centrifugation (5000 cells, n = 8) or by our FACS-based protocol (5000 events, n = 8). Vertical dashed lines indicate a fold change of ±2, and the horizontal dashed line indicates a p-value of 0.05 (two-sided t-test, unequal variance). The 46 metabolites with a fold change greater than ±1 and a p-value of less than 0.05 are indicated by blue solid circles, and the other 40 metabolites are indicated with gray open circles. Selected metabolites are labeled: 3: adenosine triphosphate (ATP), 2: adenosine diphosphate (ADP), 1: adenosine monophosphate (AMP), D: dihydroorotic acid, E: glutamic acid, F: flavin-adenosine-dinucleotide (FAD), G: glutathione (GSH), I: isoleucine, K: lysine, M: methionine, O: ornithine, P: proline, Q: glutamine, R: arginine, T: thiamine, X: oxidized nicotinamide-adenine-dinucleotide (NAD), and Y: reduced nicotinamide-adenine-dinucleotide (NADH).

Figure 5

Volcano plots of differences between etomoxir-treated (n = 8) and control (n = 8) Jurkat cells. (A) 5000 cells were harvested by centrifugation. (B) 5000 events were collected with our FACS-based protocol. Vertical dashed lines indicate a fold change of ±2, and the horizontal dashed line indicates a p-value of 0.05 (two-sided t-test, unequal variance). Metabolites with a fold change greater than ±1 and a p-value of less than 0.05 are indicated by blue solid circles, and all other metabolites are indicated with gray open circles. Selected metabolites are labeled: 1: proline, 2: carnitine, 3: propionyl-carnitine, 4: butyryl-carnitine, 5: S-adenosyl-homocysteine, 6: S-adenosyl-methionine, 7: guanosine-diphosphate, and 8: adenosine triphosphate.

Figure 6

Heatmap of series of extracts of B cells from spleen and matching debris samples with a varying number of sorted events. A total of 8 replicates per group were generated, but 3 outlier samples were detected by z-score filtering and removed. Above the black line are samples in which the signal intensity in cell extracts is significantly larger than in debris (FDR < 0.05 determined by a one-sided Wilcoxon rank sum test corrected for multiple testing with the Benjamini & Hochberg method). The internal standards ATA and CLF were included, but 28 metabolites that were detected but not above the blank level were omitted from the plot.

Figure 7

Principal component analysis of metabolic profiles obtained for 8 different cell types in 5 different experiments using an input of 5000 events per replicate. Only 10 metabolites that were detected above background in all cell types were included in this analysis. Colors indicate the cell type, and symbols indicate the experiment. Debris (background) samples from all experiments lie close together in the center of the plot, while all cell types are clearly separated. The number of replicates is 6 in experiments 1 and 2, 12 in experiments 3 and 4, and 8 in experiment 5.

Figure 8

Correlation of ratio of the median signal intensity between HSCs and B cells comparing the published data and this study. A total of 30 metabolites were measured in both studies.