UHPLC-MS Setup

Analyses were performed using a Vanquish UHPLC system (Thermo Fisher Scientific, San Jose, CA, USA) coupled online to a Q Exactive mass spectrometer (Thermo Fisher Scientific, San Jose, CA, USA). Samples were resolved over a Kinetex C18 column, 2.1 × 150 mm, 1.7 μm particle size (Phenomenex, Torrance, CA, USA) equipped with a guard column (SecurityGuard^™ Ultracartridge – UHPLC C18 for 2.1 mm ID Columns – AJO-8782 – Phenomenex, Torrance, CA, USA) at 25°C using an isocratic condition of 5% acetonitrile, 95% water, and 0.1% formic acid flowed at 250 μl/min. The Q Exactive mass spectrometer (Thermo Fisher Scientific, San Jose, CA, USA) was operated independently in positive or negative ion mode, scanning in Full MS mode (2 μscans) from 60 to 900 m/z at 70,000 resolution, with 4 kV spray voltage, 15 shealth gas, 5 auxiliary gas. Calibration was performed prior to analysis using the PierceTM Positive and Negative Ion Calibration Solutions (Thermo Fisher Scientific). Acquired data was then converted from .raw to .mzXML file format using Mass Matrix (Cleveland, OH, USA). Metabolite assignments, isotopologue distributions, and correction for expected natural abundances of deuterium, ¹³C, and ¹⁵N isotopes were performed using MAVEN (Princeton, NJ, USA).^[16]

Graphs, heat maps and statistical analyses (either T-Test or ANOVA) were prepared with GraphPad Prism 5.0 (GraphPad Software, Inc, La Jolla, CA) and GENE-E (Broad Institute, MA).

Metabolite standard characterization

Selectivity and reproducibility

Over 400 pure analytical standards ranging from 72 to 666 Da were tested and characterized using an isocratic gradient, providing retention times for each compound (Supplementary Table 1). While aqueous dilution of standards into 0.1% formic acid improved chromatographic separation and detection over a four-minute window, dilution of the same compounds into the organic metabolite extraction solution (methanol/acetonitrile/water 5:3:2 v/v/v) ^{[10,15,17,18,21]} provided enhanced compound coverage, increased signal/noise ratios, and sensitivity within a shorter three-minute retention time window. Separation was achieved for key metabolites from pathways such as glycolysis, TCA cycle, glutaminolysis/glutathione homeostasis, and polyamine metabolism (Figure 1C). Despite the relatively short chromatographic resolution window, this method provided baseline separation of compound classes (Figure 1C, 1G). We observed resolution of structural and positional isomers including, but not limited, to 2,3-cAMP and 3,5-cAMP (Figure 1D); L-leucine, L-isoleucine, and norleucine; 3-aminoisobutanoate and 4-aminobutanoate; fumarate and maleic acid; betaine, L-valine, and norvaline; noradrenaline and pyridoxine; D-glucoronolactone and ascorbate; citrate and isocitrate (Supplementary Table 1).

Open in a separate window

Figure 1

UHPLC and MS characterization of metabolite standard compounds

Using a library of ~400 compounds, the retention times and peak areas were determined when diluted in aqueous and organic solutions (Supplementary Table 1). Results are depicted as 2-dimensional plots showing (A) neutral mass vs. ion intensity, or (B) neutral mass vs. chromatographic retention time. In C, metabolites critical to central carbon and nitrogen metabolism are resolved during the isochratic elution.. In D, resolution of isobaric isomers such as 2,3-cAMP and 3,5-cAMP was observed. Reproducibility of peak shape, retention time, and ion signal intensity is shown in 10 replicate injections of a technical mixture over the course of 1200 analytical runs spanning over 6 non-consecutive days (E). (F) The stability of system pressure indicates a lack of excessive analyte accumulation on the column during 1500 runs. In G, we provide a representative 3D chromatogram showing separation of different hydrophilic compound classes over a three minute run. Linearity of quantitation (r² > 0.98) for key metabolites in central carbon and nitrogen pathways (H) was assessed over 4 orders of magnitude using a 9-point standard curve. Representative metabolites are shown for glycolysis (red - I), TCA cycle (green - J), and glutaminolysis/glutathione homeostasis (GSH/GSSG – reduced/oxidized glutathione) (blue - K).

We assessed reproducibility of this method by analyzing the coefficients of variation (CV) for analytical parameters such as retention time, peak shape, and ion intensity (peak area top and peak area integration) of identified metabolites in a technical replicate mixture injected every 15 samples over the course of 1200 analytical runs during six non-consecutive days. An example is presented in Figure 1E, showing an inter day peak area CV = 2.7% for glutathione. Although this method does not incorporate a washing/re-equilibration segment, carry over of compounds were not observed in blank runs before and after 1500 sample runs spanning over 6 days. In support, the column operational pressure was maintained during this time period (Figure 1F) indicating that the guard column contributes to the stability of column performance by retaining relatively hydrophobic compounds. In addition, we determined retention time CVs of 0.28% and extremely stable system pressure (CV=0.37%) throughout the entire analysis (Figure 1E, 1F). MS signal was also stable during this extensive analysis as demonstrated by preservation of sub-5 ppm mass accuracy for all standard calibration mixtures analyzed both prior to and post-analysis (data not shown). We observed < 0.5% average deviation from the theoretical ion intensity for the natural abundance of the M+1 ion (¹³C₁) in glutathione (Figure 1E inset), indicating MS signal stability that is a key analytical feature for both the identification of compounds on the basis of the chemical formulas as well as for quantitative and tracing experiments.^[22]

Relative and absolute quantitative analysis with external calibration curves and heavy-labeled spiked in standards

We next characterized the performance of this method with respect to quantitative analyses. Relative levels of metabolites were assessed across samples by characterizing signal linearity in serial dilutions of select compounds from central pathways (Figure 1H). Signal linearity was assessed using linear regression analysis (r²), and was confirmed over four orders of magnitude that were tested for compounds reported in Figure 1I–K (Supplementary Table 2), with the lowest limit of detection (LLOD) = 100 fmol and the lowest limit of quantitation (LLOQ) = 500 fmol. By further validating signal reproducibility using internal standards such as the xenometabolite 5-fluorouracil (Figure 2A) and isotopically-labeled ¹³C₅¹⁵N₁-L-valine (Figure 2B), we then confirmed signal linearity in the presence of a complex matrix by determining an r² > 0.98 for ¹³C₆-glucose that was exogenously added to human plasma as an internal standard at four log orders of concentration (0.1 – 1000 μM) before extraction (Figure 2C and 2D). This experiment was performed over two non-consecutive days to demonstrate inter-day reproducibility of the linearity of quantifications (Figure 2D).

Open in a separate window

Figure 2

Robustness of relative and absolute quantitative analysis in complex matrices

To show reproducibility of internal standards, coefficients of variation (CV = average/standard deviation) for peak area and retention time were calculated for (A) the xenometabolite 5-fluorouracil and (B) isotopically-labeled ¹³C₅,¹⁵N₁-valine. (C) Interday variability, matrix effects on ionization, and signal linearity were determined by extracting plasma with the addition of 1 mM ¹³C₆-glucose. (D) Signal linearity is shown with a 5-point serial dilution in plasma over 4 orders of magnitude. Dilutions on two non-consecutive days and in extraction solution alone show interday reproducibility and matrix effects, respectively. (E) Dilution of plasma 1:1 (v/v) with extraction solution demonstrates linearity of fold change calculations for the relative quantitation for glucose, (F) and for additional compounds in central carbon and nitrogen pathways. Fold change values are depicted using a heat map, with blue representing a fold change of 0.5 and red representing a fold change of 1. (G) Red blood cell extracts sampled over a 42 day storage period were analyzed using the three-minute method and conventionl gradient-based methods using a C18 (9 minute method) and amide HILIC (15 minute method). Metabolite abundances are depicted as heat maps to facilitate comparison between methods of technical reproducibility and expected trends based on previously published results.

Human plasma, either undiluted or diluted two-fold in extraction solution (Figure 2E), was tested with the three-minute method to determine linearity of quantitation, recovery, and matrix-dependent ion suppression. Minor deviations from the expected fold changes of 2 were observed, with a CV < 2% for glucose and other representative metabolites of key pathways (Figure 2E and 2F).

Comparison to gradient-based methods: metabolomics of stored RBCs

Next, we tested the suitability of this method for applications requiring high-throughput relative and absolute quantitation of target compounds in clinically-relevant samples. Metabolic analyses of RBCs are clinically relevant in that blood biochemical testing is a hallmark of clinical biochemistry. Indeed, RBCs are by far the most abundant host cell in the human body (25 out of 30 trillion host cells in the human body are RBCs) and their metabolism mirrors states of health and disease. In addition, transfusion of packed RBCs is a life-saving intervention for patients with hematological malignancies and hemoglobinopathies, but also for surgical patients or those suffering from traumatic injuries with hemorrhage. While RBC storage is a logistic necessity to make approximately 108 million units available for millions of recipients worldwide every year, refrigerated storage of RBCs negatively affects RBC metabolism and physiology, the effects of which are referred to as the storage lesion. RBC metabolism during storage under blood bank conditions has been extensively characterized by our group using LC-MS, ^{[18,21,23–25]} and by others via NMR and gas chromatography-MS (GC-MS)-based approaches.^[26,27] Here, four units of RBCs were sampled during storage in Additive Solution (AS)-3 (Nutricel, Haemonetics) on a weekly basis until the expiration date of the units (42 days upon collection and processing). Samples were analyzed with the three-minute method and conventional gradient-based methods using either a C18 or HILIC column.^[19,20] It is worth noting that the HILIC method was validated in-house using the same IROA standard library (Supplementary Table 1). Samples were run in randomized order, while technical mixes (pool of day 0 samples from each of the four units) were run every 15 samples to ensure stability of the system under each analytical condition (Supplementary Table 3). Metabolite levels were graphed as a heat map in order to facilitate comparison with the existing literature that extensively describes the metabolic impairments associated with the storage lesion (Figure 2G, a vectorial version of the heat maps including metabolite names is provided in Supplementary Figure 1).^{[18,21,26,27]} The storage lesion is in part characterized by the progressive accumulation of lactate, increased purine catabolism, depletion of intermediates of the pentose phosphate pathway (PPP), and impaired glutathione synthesis. Decreased levels of citrate, adenine, urate, ribose phosphate, glutamine, reduced and oxidized glutathione (GSH and GSSG), and increased levels of hypoxanthine and 5-oxoproline are also observed in stored RBCs.^{[18,21,26,27]} Recently, we identified 8 metabolic markers of the RBC storage lesion that were validated by two independent laboratories with at least three different analytical setups.^[21] Comparison of the results for those biomarkers (three representative markers are shown in Supplementary Figure 2) generated by the three methods used here indicates that the methods are indeed comparable among each other and are consistent with previously observed RBC metabolic trends during storage ^{[18,21,26,27]}. For example, accumulation of the glycolytic catabolite lactate and purine oxidation product hypoxanthine has been shown using C18 and HILIC-based UHPLC-MS and NMR by independent laboratories.^{[18,21,26,27]} and here confirmed with all methods (Supplementary Figure 2). Analogously, all three methods highlighted comparable storage-dependent citrate consumption in RBCs stored in AS-3 (Supplementary Figure 2), consistent with recent reports identifying a role for citrate metabolism in stored RBCs and suggesting this metabolite as a marker of the RBC storage lesion^[21,24]. Furthermore, the three-minute method maintained superior reproducibility across technical replicates relative to the amide HILIC column despite lacking a conventional elution gradient, indicated by the analysis of technical mixes run every 15 samples during the entire analysis (Supplementary Table 3).

The Authors are grateful to Drs. Anthony Bacon, Annie Slaughter, Anirban Banerjee for the rat models of trauma/hemorrhagic shock; Dr. Tatsuro Yoshida for the red cell storage tracing experiments, as detailed in the paper. The authors would also like to thank Dr. Robert Hodges for reviewing the manuscript. Research reported in this publication was supported in part by funds from the National Blood Foundation Early career grant 2016 (AD), University of Colorado Comprehensive Cancer Center Core Support (P30 CA046934-17), the National Institute of General Medical Sciences of the National Institutes of Health under Award Numbers P50GM049222 and T32GM008315, and Grant #P50 GM049222 from NIGMS, NIH (CS). The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.