High-Throughput Measurement of Lipid Turnover Rates Using Partial

May 3, 2018 - The performance on lipid kinetics measurement of our methods was validated ... With HeLa cells cultured in 5% 2H2O media for 48 h, we co...
0 downloads 0 Views 3MB Size
Download PDF
Article Cite This: Anal. Chem. 2018, 90, 6509−6518

pubs.acs.org/ac

High-Throughput Measurement of Lipid Turnover Rates Using Partial Metabolic Heavy Water Labeling Byoungsook Goh,† Jinwoo Kim,‡ Seungwoo Seo,† and Tae-Young Kim*,†,§ †

Department of Chemistry, School of Physics and Chemistry, ‡School of Electrical Engineering and Computer Science, and §School of Earth Sciences and Environmental Engineering, Gwangju Institute of Science and Technology, 123 Cheomdangwagi-ro, Buk-gu, Gwangju 61005, South Korea

Anal. Chem. 2018.90:6509-6518. Downloaded from pubs.acs.org by UNIV OF SUNDERLAND on 10/20/18. For personal use only.

S Supporting Information *

ABSTRACT: Novel analytical platforms for high-throughput determination of lipid turnover in vivo have been developed based on partial metabolic 2H2O labeling. The performance on lipid kinetics measurement of our methods was validated in three different liquid chromatography−mass spectrometry (LC-MS) setups: MS-only, untargeted MS/MS, and targeted MS/MS. The MS-only scheme consisted of multiple LC-MS runs for quantification of lipid mass isotopomers and an extra LC-MS/ MS run for lipid identification. The untargeted MS/MS format utilized multiple data-dependent LC-MS/MS runs for both quantification of lipid mass isotopomers and lipid identification. An in-house software was also developed to streamline the data processing from peak area quantification of mass isotopomers to exponential curve fitting for extracting the turnover rate constant. With HeLa cells cultured in 5% 2H2O media for 48 h, we could deduce the species-level turnover rates of 108 and 94 lipids in the MS-only and untargeted MS/MS schemes, respectively, which covers 13 different subclasses and spans 3 orders of magnitude. Furthermore, the targeted MS/MS setup, which performs scheduled LC-MS/MS experiments for some targeted lipids, enabled differential measurement between the turnover rates of the head and tail groups of lipid. The reproducibility of our lipid kinetics measurement was also demonstrated with lipids that commonly detected in both positive and negative ion modes or in two different adduct forms. utilized to quantify lipid flux rates in recent years. Qi and coworkers used [13C3-2H5]-glycerol and [13C18]-oleic acid to distinguish between the two isoforms of diacylglycerol acyltransferase in terms of hepatic TG synthesis.6 McLaren et al. also administered [13C18]-oleic acid to mice to examine plasma TG synthesis.7 Additional stable-isotope labeled tracers including [2H9]-choline,8,9 [2H4]-ethanolamine, [2H3]-L-serine, and [2H6]-myo-inositol9 were applied to characterize the turnover of glycerophospholipids. Since these stable isotope tracers were designed for a specific class of lipid, they could not reveal comprehensive lipid turnover in a systematic manner. In order to overcome this limitation, [U−13C]-palmitic acid was utilized as a metabolic tracer to cover a wide range of lipid classes in human skeletal muscle cells in studying the dynamics of lipid metabolism.10 Whitelegge et al. pioneered the concept of 13C partial labeling as an alternative proteomics tool for relative quantification and flux measurement.11 They successfully proved the potential usefulness of subtle modification of

L

ipids in cells are continually synthesized and degraded to perform their biological roles under varying physiological conditions generating complex metabolites in dynamic transport within and between cells. The structural and functional integrity of lipids is closely connected to their intrinsic metabolic fluxes maintaining the lipid compositions and pool sizes. Lipid turnover can be defined as continuing degradation and replenishment of lipids in cells. Failure in precise regulation of lipid turnover is implicated in metabolic syndromes that significantly increase the chances for a number of diseases including atherosclerosis1 and diabetes.2 It has also been reported that accumulation of triacylglycerol (TG) by its low removal and high storage rates in human adipose tissue is highly related with obesity.3 Measurement of lipid turnover rates in vivo is a critical step for elucidating the mechanisms controlling pathogenesis of disorders of lipid metabolism. The kinetic information on lipid turnover can also be utilized to identify potential biomarkers for disease diagnosis. A variety of efforts have been made to analyze the kinetics of lipid metabolism. Radioactively labeled phosphate residues with 32 P have been used to measure the turnover of bacterial phospholipids.4,5 With the advent of high resolution mass spectrometry (MS), stable isotope tracers have been widely © 2018 American Chemical Society

Received: December 27, 2017 Accepted: May 3, 2018 Published: May 3, 2018 6509

DOI: 10.1021/acs.analchem.7b05428 Anal. Chem. 2018, 90, 6509−6518

Article

Analytical Chemistry

Figure 1. Metabolic 2H2O labeling of HeLa cell and three LC-MS and LC-MS/MS schemes for lipid turnover measurement. (A) HeLa cells were cultured in media enriched with 5% 2H2O and harvested at eight time points (0, 3, 6, 12, 18, 24, 36, and 48 h) after labeling. (B) MS-only method consists of eight MS runs with an additional data-dependent MS/MS analysis, while the untargeted MS/MS experiment acquires data-dependent MS/MS for lipids extracted at 8 time points. In the targeted MS/MS analysis, a list of target lipids generated from the MS-only experiment undergoes multiple MS/MS at each time point.

chance to distinguish the turnover rate of a lipid molecule itself from those of different moieties. The utility of our approach is demonstrated by characterizing the lipid turnover at specieslevel on a lipidomic scale in HeLa cell, which was cultured in media enriched with 5% 2H2O up to 2 d.

isotope ratio for systems biology studies. Hellerstein and coworkers have also explored the utility of partial heavy water (deuterium oxide, 2H2O) for quantifying metabolic fluxes of biomolecules based on mass isotopomer distribution analysis (MIDA).12 2H2O is an excellent alternative tracer for metabolic labeling of biomolecules for several reasons. First, intake of 2 H2O and subsequent 2H enrichment at low levels are physiologically safe for animals and human. Second, administration of 2H2O is convenient for a long-term labeling experiment.13 Third, 2H2O quickly equilibrates with body water and decays with a half-life of about 2 weeks maintaining a constant level during the labeling period.14 Fourth, 2H2O labeling is inexpensive and universally labels various biomolecules. Based on the advantages of 2H2O labeling, many dynamics studies of biomolecules such as nucleic acid,13,15 proteins,15−23 and lipids,12,15,24−26 have been performed. In the case of TG, 2H from 2H2O is incorporated into C−H bonds of glycerol-3-phosphate and fatty acids during de novo lipogenesis from glucose or pyruvate.12,24 Metabolic partial 2H2O labeling coupled with gas chromatography (GC)-MS has been widely employed to measure the rates of fatty acid,26,27 cholesterol,27 and total TG synthesis.12,24 However, turnover rates of individual lipids could not be measured by GC-MS. More recently, liquid chromatography (LC)-MS has been utilized to investigate the kinetics of individual phosphatidylcholines, TGs,28 and cholesterols29 by feeding 2H2O to mice, but the flux measurements were limited to a few lipid classes. In this work, we report novel LC-MS and LC-MS/MS platforms to measure individual lipid kinetics on a global scale with metabolic 2H2O labeling. To achieve a high-throughput calculation of fractional lipid synthesis, a data analysis program was developed in-house to automate data processing from quantifying the peak area of each mass isotopomer of an identified lipid to deducing the turnover rate from an exponential curve fitting. Unlike the metabolic 2H2O labeling methods combined with MIDA,30 our method does not need to know the exact precursor enrichment, the asymptotic isotopic ratio of lipid ions, or the number of labeling sites of lipids in order to determine their turnover rates, since we mathematically compute the parameters necessary for the kinetic analysis with our software. MS/MS analysis of lipid also provided a

■

EXPERIMENTAL SECTION Materials. Chloroform, methanol (MeOH), water (H2O), and acetonitrile (ACN) were from Thermo Fisher Scientific (Waltham, MA). Isopropanol (IPA) was from J.T. Baker (Center Valley, PA). Ammonium formate and ammonium acetate were from Fluka (Zwijndrecht, Netherland) and Merck (Darmstadt, Germany), respectively. HeLa cell line was obtained from Korean Collection for Type Cultures (Daejeon, Korea). Fetal bovine serum (FBS) and penicillin streptomycin were from Gibco (Grand Island, NY). Dulbecco’s modified eagle medium (DMEM) and trypsin-ethylenediaminetetraacetic acid were from Hyclone (South Logan, UT, U.S.A.) and Welgene (Daegu, Korea), respectively. 2H2O (2H, 99.9%, atom) was from Cambridge Isotope Laboratories (Andover, MA). Myristic acid, palmitic acid, oleic acid, and stearic acid were from Sigma-Aldrich (St. Louis, MO). Sample Preparation. DMEM supplemented with 10% FBS and penicillin streptomycin was used in HeLa cell culture. HeLa cells were adherently cultured in cell culture dishes prior to 2H2O labeling. Then, the cultured HeLa cells were separated into eight cell dishes in media enriched with 5% (mol/mol) 2 H2O, and each dish was used for cell harvest at 0, 3, 6, 12, 18, 24, 36, and 48 h after labeling (Figure 1A). In order to make the lipid pool size for the turnover measurement constant over the labeling times, which is necessary to interpret the measured fractional synthesis rate by deuterium incorporation into lipids as their turnover rates, the number of cells harvested at each time point was on the same order of magnitude. Lipids were extracted from the pellet by Folch method.31 Chloroform/ MeOH (2:1, v/v) was added to the pellet. The mixture was vortexed for 20 min and water was added. Then, the mixture was centrifuged at 500 × g for 10 min. The lower phase was recovered and dried in a vacuum evaporator. Finally, the dried lipid sample was reconstituted in methanol for LC-MS and LCMS/MS analyses. 6510

DOI: 10.1021/acs.analchem.7b05428 Anal. Chem. 2018, 90, 6509−6518

Article

Analytical Chemistry Analytical Scheme. Three analytical schemes were designed to evaluate the performance of LC-MS and LC-MS/ MS with respect to accuracy, precision, and the depth of coverage in the measurement of lipid turnover rates: MS-only, untargeted MS/MS, and targeted MS/MS experiments (Figure 1B). In the MS-only experiment, lipid extracts from the eight time points of 2H2O labeling were analyzed by LC-MS. For this mode, an additional set of HeLa cell was prepared in normal media without 2H2O and then the extracted lipids were subjected to data-dependent MS/MS analysis in order to obtain their tandem mass spectra for identification. The retention time (±20 s) and accurate mass (±30 ppm) of the lipids identified by tandem spectral matching in the extra MS/MS run were exploited to identify lipids at the eight labeling time points. In the untargeted MS/MS experiment, on the other hand, a list of identified lipids was generated by the data-dependent MS/MS analysis at each time point, respectively. Then, lipids identified in at least five different time points were selected for measurement of turnover rate. The main reason for designing both MS-only and untargeted MS/MS experiments is to investigate the effect of the number of MS scans on the accuracy of our turnover rate measurements, since the MS-only scheme provides more number of MS scans than the untargeted MS/MS mode. Lastly, the targeted MS/MS experiment was intended to examine differential turnovers between fatty acyl chains and the glycerol backbone on glycerolipid or glycerophospholipid. In the targeted MS/MS experiment, a list of eight target glycerophospholipids was made from the first MS-only mode and subjected to multiple MS/MS for each lipid to extract ion chromatograms of their fragment ions. LC-MS and MS/MS Conditions. Lipid extracts were analyzed with an Agilent 6520 quadrupole time-of-flight mass spectrometer coupled with an Agilent 1260 infinity HPLC system. Samples were separated using a C18 column (Agilent Zorbax Extend, 4.6 × 50 mm, 1.8 μm). A binary gradient was applied for the separation with mobile phase A (ACN/H2O, 60:40, v/v) and mobile phase B (IPA/ACN, 90:10, v/v). Solvent modifiers of 10 mM ammonium formate and 10 mM ammonium acetate were dissolved in both mobile phases for positive and negative ion mode, respectively. The mobile phase B was ramped to 30% for 1 min and linearly increased to 60% for 14 min, to 100% for 25 min and held for 5 min, and decreased to 0% for 3 min and finally equilibrated for 12 min. The LC flow rate was set to 0.4 mL/min. The ESI parameters were as follows: gas temperature, 350 °C; drying gas, 10 L/min; nebulizer, 35 psi; capillary voltage, 3.5 kV; fragmentor, 200 V. MS and MS/MS were performed in both positive and negative ion mode. Mass spectra were obtained in the m/z range of 200−1600. For MS/MS analysis, the top five most intense peaks in the MS scan were selected for auto collision induced dissociation with active exclusion of 0.4 min on the m/z range of 50−1600. The collision energy was set by the following 3 × (m / z) + 10(V ) linear formula: collision energy = 100 Lipid Identification and Nomenclature. All acquired LC-MS/MS data were processed with Agilent MassHunter qualitative analysis software to convert the raw data into MGF files. The resulting MGF files were searched against LipidBlast32 using NIST MS PepSearch platform. Precursor ion and fragment ion mass tolerances were set as 0.05 and 0.1 m/z, respectively. Putatively identified lipid species (reverse-dot product ≥ 800) were further analyzed for kinetic studies. Fatty

acids were identified by the retention times and accurate masses of four authentic standards including myristic acid, palmitic acid, oleic acid, and stearic acid. We adopted the lipid nomenclature system established by LIPID MAPS33 and the shorthand notation described by Liebisch et al.34 The following abbreviations were used in this paper: triacylglycerol (TG), diacylglycerol (DG), sphingomyelin (SM), phosphatidylcholine (PC), phosphatidylserine (PS), phosphatidylethanolamine (PE), phosphatidylinositol (PI), phosphatidylglycerol (PG), phosphatidic acid (PA), cardiolipin (CL), plasmalogen-PE (PE-P), plasmalogen-PC (PC-P), lysophosphatidylcholine (LPC), and lysophosphatidylethanolamine (LPE). Turnover Rate Calculation. Our approach for determining protein turnover rate reported elsewhere19 was adapted to calculate lipid turnover rate. Briefly, the kinetic information is extracted from a time-series change in the normalized peak intensity of each mass isotopomer corresponding to a lipid ion. The normalized peak intensity, Ai (t), at a given time point t was computed by the ratio of the peak intensity of the mass isotopomer i, Ii, to the summation of peak intensities of all mass isotopomers. N

Ai (t ) = Ii /∑ Ij

(1)

0

The rate constant of turnover, k, was determined by fitting the Ai (t) values at various labeling times into the first order decay eq 2 below. A(t ) = {A(∞) − A(0)}(1 − e−kt ) + A(0)

(2)

where A(0) and A(∞) are the normalized peak intensities at time 0 and infinity, respectively. In order to expedite the data processing from the peak quantification to the curve fitting, an in-house MS data analysis program was developed. Lipid identification data (TSV format) and MS data (mzData format) were imported to extract the extracted ion chromatograms (EICs) for mass isotopomers corresponding to an identified lipid with an m/z tolerance of ±30 ppm. Then, the acquired EICs were smoothed by the Savitzky-Golay algorithm35 with second order polynomial and 9 and 3 smoothing points for MS-only and untargeted MS/MS mode experiments, respectively. The smoothed chromatograms then underwent the peak area integration to obtain Ai(t). Finally, Ai(t) at each time point of labeling was fitted against the first-order kinetic eq 2 to deduce the turnover rate of lipid. The following operational conditions were applied to remove impractical data in the curve fitting: R2 > 0.7, −0.1 < A(0), A(∞) < 1.1, and A(t) ≠ 0. In addition, a filtering system that can systematically exclude the following three low quality EICs was implemented into our in-house program to ensure the fidelity of our measurement: EICs from low signal intensities, inconsistent peak picking due to coelution of isomeric lipids, and an unnatural isotopic distribution.

■

RESULTS AND DISCUSSION Lipid Identification. A lipid flux measurement based on LC-MS or LC-MS/MS combined with stable isotope labeling requires identification and quantification of lipids at multiple time points in order to extract the kinetic information from time-dependent changes in their profiles. Our approach also needs measurement of the peak area of each mass isotopomer corresponding to a lipid at several time points along with its 6511

DOI: 10.1021/acs.analchem.7b05428 Anal. Chem. 2018, 90, 6509−6518

Article

Analytical Chemistry

Figure 2. Venn diagrams and bar graphs for the number of lipids identified in the (A) MS-only and (B) untargeted MS/MS. The MS-only experiment identified 284 and 171 of individual lipid species from 13 lipid subclasses in positive and negative ion modes, respectively, while the untargeted MS/MS analysis identified 217 and 128 of individual lipid species from 13 lipid subclasses in positive and negative ion modes, respectively. The bar graphs display the number of identification for each lipid subclasses.

distribution of a lipid ion over the labeling time. In some cases, especially in the mass spectra of a lipid with a rapid turnover rate, the first mass isotopomer (the monoisotopic peak, M0) is no longer the most intense peak among the mass isotopomers of a lipid ion and the more intense second mass isotopomer peak (M1) is selected and subjected to MS/MS. The resulting tandem mass spectrum leads to wrong or no identification in the database search, which tends to decrease the number of lipids identified as the labeling time increases (Table S1). Second, the stochastic and irreproducible selection of precursor ions in a data-dependent MS/MS experiment results in inconsistent identification of lipids from one run to another. In our data-dependent experiment, only 99 and 76 lipids were commonly identified across 8 labeling time points in positive and negative ion modes, respectively (Table S2). Temporal Changes of Mass Isotopomer Distribution of Lipid Ions. Partial enrichment of 2H2O in the HeLa cell body pool induces deuterium incorporation into lipids during their biosyntheses. While the isotope labeling methods exploiting full atomic percent enrichment result in peak separation between light- and heavy-labeled forms, our partial enrichment approach causes some changes in the isotopic distribution of a peak due to incomplete separation between unlabeled and labeled samples. Figure 3A,B shows the sequential effects of partial 2H2O labeling on the profile of mass isotopomer distribution for two lipids, PG(16:0_18:1) and PE(16:0_20:4), at the labeling times of 0, 24, and 48 h. Without 2H2O enrichment (0 h), normal mass isotopic distributions mainly attributed to the natural abundance of 13 C were observed for the two lipids. After 48 h of labeling, the most intense peak for PG(16:0_18:1) became the second mass isotopomer peak (M1), which is accompanied by the third mass isotopomer peak (M2) with a comparable peak intensity and new features corresponding to M4, M5, and M6 (Figure 3A). On the other hand, PE(16:0_20:4) showed a relatively small degree of change in the profile of mass isotopomer distribution over the labeling time (Figure 3B). The monoisotopic peak (M0) still displayed the highest intensity after 48 h of labeling,

identification. Two different LC-MS and LC-MS/MS were designed to explore the analytical performance of lipid turnover measurement. In the MS-only experiment, LC-MS runs for lipid extracts from eight labeling time points were performed to quantify the isotopic distribution of a lipid ion and an additional datadependent LC-MS/MS for unlabeled lipids was executed to identify them by tandem mass spectral matching. Then, the retention time and accurate mass of the identified lipid in the additional LC-MS/MS were used for identification of lipids in the LC-MS runs corresponding to eight labeling time points. Thus, the number of lipids identified in the extra datadependent MS/MS analysis defines the maximum number of lipid identification in the LC-MS runs for eight labeling time points. As a result, 284 and 171 lipids were identified in positive and negative ion modes, respectively (Figure 2A). Among them, 37 lipids (PE: 26, LPE: 1, and PE-P: 10) were commonly observed in positive and negative ion modes since these contain both a basic amine and an acidic phosphate groups. In the untargeted MS/MS experiment, a data-dependent LC-MS/MS was run for the lipid extracts from each of eight labeling time points. Then, MS scans were utilized to quantify the relative mass isotopomer abundances of a lipid ion and MS/MS scans were exploited to identify lipids. Because of the nature of a data-dependent analysis, the number of identified lipids varied from one run to another. With a condition of at least 5 time identifications out of 8 labeling time points, a total of 217 and 128 lipids were identified in positive and negative ion modes, respectively, and 31 lipids were commonly detected in both ion modes (Figure 2B). In both MS-only and untargeted MS/MS experiments, 13 different lipid subclasses were identified and the most frequently identified lipid subclass was TG, followed by PE and PC. Approximately 100 more lipids were identified by the MSonly experiment than by the data-dependent experiment. The difference in the number of lipids identified can be attributed to the following two factors. First, partial 2H2O labeling exploited in this study induces changes in the relative isotopic 6512

DOI: 10.1021/acs.analchem.7b05428 Anal. Chem. 2018, 90, 6509−6518

Article

Analytical Chemistry

Figure 3. Lipid turnover measurement from temporal changes of mass isotopomer distribution of lipid ions. The relative abundances for mass isotopomers of (A) PG(16:0_18:1) and (B) PE(16:0_20:4) showed a decrease in I0 and the emergence of additional features corresponding to higher mass isotomomers across 48 h labeling of 2H2O. The turnover rates for (C) PG(16:0_18:1) and (D) PE(16:0_20:4), which were deduced by fitting A(t) at different labeling time points into an exponential curve, were 0.089 and 0.011 h−1, respectively.

matching a MS/MS spectrum to a mass spectral database, leading to a decrease in the number of lipid identification at longer labeling times in our untargeted MS/MS experiment (Table S1). Lipid Turnover Rate Measurement. In order to deduce the kinetic information from the acquired MS spectra, the intensity of each mass isotopomer belonging to a lipid ion was quantified by integrating the peak area in the EIC with our inhouse software. Then, the normalization of a peak intensity was calculated by the ratio of the peak intensity of a mass isotopomer to the summation of all quantifiable mass isotopomers (eq 1). Finally, the turnover rate of a lipid is determined by fitting the normalized peak abundances of the mass isotopomer over the labeling time into the first-order kinetics equation (eq 2). Figure 3C,D displays two examples for determination of lipid turnover rate in which time-resolved fractional peak intensities of m0 were fitted into an exponential curve. The turnover rates of PG(16:0_18:1) and PE(16:0_20:4) were determined at 0.089 and 0.011 h−1, respectively. Based on the data processing steps above, the

although the relative intensities of the high-mass isotopomer peaks (M1−M5) were increased. In partial metabolic labeling experiments, the enrichment level of the isotope-labeled precursor pool has a direct effect on the resulting isotopic distribution of a peak. Higher percentage of 2H2O would induce more change in the normalized peak intensities of mass isotopomers, which could in turn result in better accuracy and sensitivity of the turnover rate measurements, especially for lipids with slow kinetics. However, the use of a high concentration of 2H2O can cause biological and technical issues. In order to minimize the potential perturbation of normal physiological conditions in biological systems, an enrichment level of 5% 2H2O has been routinely employed for cell culture and rodent model experiments.15−17 Another drawback of a high degree of 2H2O enrichment is related to the MS/MS performance. As shown in Figure 3A, a relatively fast turnover of PG(16:0_18:1) resulted in a significant reduction of the monoisotopic peak intensity at 48 h of labeling. The decreased abundance of the monoisotopic peak was disadvantageous for triggering an MS/MS event and 6513

DOI: 10.1021/acs.analchem.7b05428 Anal. Chem. 2018, 90, 6509−6518

Article

Analytical Chemistry

Figure 4. Box and whisker plots for distribution of measured turnover rates by lipid subclass. A total of 108 and 94 individual lipid turnover rates were determined in the (A) MS-only and (B) untargeted MS/MS analyses, respectively. The lipid subclasses are plotted in decreasing order of the median turnover rate and the numbers of lipids characterized in each subclass are indicated in parentheses.

turnover rates of individual lipids in HeLa cell were characterized on a global scale. Figure 4 displays the distributions of measured turnover rates according to the lipid subclasses. For the lipids that were detected in both positive and negative ion modes or showed two adduct forms (e.g., [M + NH4]+ and [M + Na]+), the average of the two turnover rates is used. In the MS-only experiment, the turnover rates for 108 individual lipids covering 12 different lipid subclasses were determined in the range of 0.0058−0.23 h−1, which spans nearly 3 orders of magnitude in lipid kinetics range (Figure 4A). By contrast, we could measure turnover rates of 94 lipids from 13 subclasses in the untargeted MS/MS experiment (Figure 4B). The full lists of all measured lipid turnover rates in both experiments are provided in Supporting Information, Tables S3 and S4. Although most of lipid turnover rates were deduced at the fatty acyl/alkyl level, some were determined at the species level due to incomplete chromatographic separation between isomeric lipids. For example, the turnover rate of TG(48:1) was expressed at the species level for two isomeric compounds, TG(14:0_16:0_18:1) and TG(16:0_16:0_16:1) (Table S3). The two LC-MS and LC-MS/MS experimental schemes that we designed revealed different analytical performances in the lipid turnover characterization. The MS-only experiment performed quantification of mass isotopomer distributions of lipid ions based on their identification results acquired by an extra LC-MS/MS run. The separation of quantification of the normalized peak intensities and identification of lipids resulted in an increase in the number of lipid identification and gave a consistent lipid identification list across the labeling times, which is advantageous for extracting as many turnover rates as possible from the LC-MS data acquired at multiple time points. In addition, the MS-only run was beneficial for constructing a true EIC for each mass isotopomer of ions since it provided a large number of data points on a chromatogram for quantification. On the other hand, the MS-only approach required an extra data-dependent LC-MS/MS run and its quality was critical for the number of identifications and subsequently the number of turnover rates of lipids. In the untargeted MS/MS experiment, on the contrary, the number and classes of the identified lipids change from one labeling time point to another due to the stochastic nature of a datadependent MS/MS analysis (Table S2), which poses a constraint for securing the normalized peak intensities of mass isotopomers at all labeling time points. To alleviate this

limitation in the untargeted MS/MS experiment, dissimilar conditions were applied to the exponential fitting to deduce the lipid kinetics between the two experimental schemes. While the untargeted MS/MS experiment required the lipids identified at minimum 5 time points for the curve fitting, the MS-only mode needed lipid identification at all 8 labeling time points. This difference in the number of lipid identification necessary for the kinetic fitting led to the result that the numbers of lipid species for which the turnover rates were determined were comparable between the two experimental designs, although the MS-only experiment had about 100 more lipid identifications than the untargeted MS/MS one. Since the condition of at least five identifications out of eight time points in the untargeted MS/ MS scheme was arbitrary, the number of measured lipid turnover rates will change if the minimum requirement for the number of identification is varied. Indeed, the number of lipid turnover measurement became 117 and 81 when the minimum identification number was changed to 4 and 6, respectively (Figure S1). Although different MS schemes were applied, the lipid turnover rates determined commonly between the two experiments showed a good consistency with a Pearson’s correlation coefficient of 0.809 (Figure S2). Deuterium is known to show substantial kinetic isotope effect (KIE) that may introduce a systematic error in the measurement of lipid turnover rates, but we speculate that deuterium KIE would not be significant in our analytical platform. Although we have assumed, as an approximation, first-order kinetics for lipid turnover processes, a series of reactions are involved in the lipid synthesis from the precursor. For example, three fatty acyl chains are sequentially attached to glycerol-3-phosphate to form TG. Saturated fatty acids are synthesized by multiple cycles of two-carbon elongation and glycerol-3-phosphate is generated from pyruvate by glyceroneogenesis or from glucose by glycolysis.24 During these biosynthesis processes of fatty acid and glycerol-3-phosphate, deuterium is incorporated into their C−H bonds if 2H2O is available in the biological system. We can expect primary KIE for deuterium transfer from 2H2O to the carbon backbones of fatty acid and glycerol-3-phosphate, and secondary KIE for the rest of their synthetic reactions. The acetyl-CoA carboxylase step is the rate-determining step in fatty acid synthesis, and the isocitrate dehydrogenase step in the Krebs cycle and the phosphofructokinase step in glycolysis are the rate-determining steps for the synthesis of glycerol-3-phosphate.36 Also, the glycerol-3-phosphate acyltransferase is the rate-determining 6514

DOI: 10.1021/acs.analchem.7b05428 Anal. Chem. 2018, 90, 6509−6518

Article

Analytical Chemistry

Figure 5. Measurement of turnover rates of PG(16:0_18:1) fragment ions in the targeted MS/MS analysis. (A) MS/MS spectrum of PG(16:0_18:1) and (B) its structure. Two fragment ions corresponding to the loss of one fatty acyl chain (m/z 255.23 and 281.25) and the product ion generated from the loss of one acyl chain and glycerol group (m/z 391.23) were selected to measure the turnover rates at the MS/MS level. Exponential curve fitting results are shown for fragment ions at m/z (C) 255.23, (D) 281.25, and (E) 391.23. The fitting results for the fragment ion at m/z 255.23, 281.25, and 391.23 were 0.057, 0.018, and 0.086 h−1, respectively.

step of TG biosynthesis.37 None of the rate-determining steps involved in the biosynthesis of TG is associated with deuterium transfer to the carbon backbone of lipids, which indicates that no primary KIE is exhibited in the measurement of TG turnover rates based on metabolic 2H2O labeling. Given that secondary KIE is much smaller than primary KIE, deuterium KIE would not be a substantial factor in our kinetic studies. The research reported by Xu et al.38 in which [U−13C]-glucose was employed to measure the turnover of glycerophospholipids also implies that deuterium KIE causes little systematic shift of lipid turnover rates. In their study, the turnover rate of PI(38:2) in human embryonic kidney 293 cells was 0.014 h−1, which is comparable with the turnover rate of 0.019 h −1 for PI(18:0_20:2) in our MS-only experiment (Table S3). Turnover Rate Measurement Using Fragment Ions. The turnover rate of the head or tail group of a lipid can be extracted if a fragment ion containing the corresponding group is detected at multiple labeling time points in the untargeted MS/MS experiment. PG(16:0_18:1) was one of the several lipids for which the turnover rates of some fragment ions could be deduced. However, a single MS/MS scan for a lipid ion per labeling time was not enough to reconstruct the EICs for its fragment ions and led to poor fitting results (Figure S3). In order to obtain multiple MS/MS scans for lipid fragment ions, the targeted MS/MS experiment was conducted for several lipid ions. Figure 5 displays the measured turnover rates of fragment ions for PG(16:0_18:1) through the targeted MS/MS analysis. Three major fragment ions were observed in the MS/MS spectrum of PG(16:0_18:1) (Figure 5B). Two peaks at m/z 255.23 and 281.25 correspond to fragment ions from the loss of a single fatty acyl chain and the product ion at m/z 391.23 contains the glycerol moiety. While the two fragment ions at m/z 255.23 and 281.25 showed the slow turnover rates with 0.057 and 0.018 h−1, respectively (Figure 5C,D), the turnover

rate of the fragment ion at m/z 391.23 was measured as 0.086 h−1 (Figure 5E), which is comparable to the turnover rate (k = 0.089 h−1) of the entire lipid of PG(16:0_18:1). These results suggest that lipid turnover in PG occurs mainly through its headgroup and the fatty tail chains turn over more slowly. However, the lower deuterium incorporation observed in the fatty acyl chains than in the glycerol backbone can be attributed to recycling of fatty acids. Because deuterium labeling is involved only in de novo lipogenesis in metabolic 2H2O labeling, deuterium labeled fatty acids in the fatty acid pool could be diluted by unlabeled fatty acids recycled from lipolysis of TGs. In an attempt to assess the contribution of recycled free fatty acids to the observed turnover rates of the fatty acyl chains in glycerophospholipids, the turnover rates of four free fatty acids, including myristic acid, palmitic acid, oleic acid, and stearic acid were determined. Their turnover rates ranged from 0.025 to 0.040 h−1 (Table S5) were comparable to the average of measured turnover rates of glycerophospholipids in the MSonly experiment (kavg = 0.041 h−1), which indicates that turnover of free fatty acids is not significantly slower compared with other lipids. Interestingly, the palmitic acid acyl chains in PG(16:0_18:1) and PE(16:0_20:4) showed substantially different deuterium incorporation over the labeling time. The turnover rate of the palmitic acid chain in PG(16:0_18:1) was observed at 0.057 h−1, while that of PE(16:0_20:4) could not be determined due to low deuterium labeling (Table S6). Likewise, turnover rates of the oleic acid acyl chains in PI(18:1_20:4) and PI(18:1_20:3) (0.017 h−1 and 0.031 h−1, respectively) were measured quite differently. Their turnover rates, on the other hand, were comparable to those at the species level measured in the MS-only mode (0.014 h−1 for PI(18:1_20:4) and 0.026 h−1 for PI(18:1_20:3), Table S6). In the case of PI(18:1_20:4), the turnover rate of the glycerol moiety (0.016 h−1) was almost the same with those of the oleic 6515

DOI: 10.1021/acs.analchem.7b05428 Anal. Chem. 2018, 90, 6509−6518

Article

Analytical Chemistry

Figure 6. Comparison of turnover rates for lipids detected in both positive and negative ion modes or two different adduct forms. The two turnover rates determined from two ion modes in the (A) MS-only and (B) untargeted MS/MS modes, or (C) two salt adducts in the MS-only analysis showed good consistency, except PE(16:0_22:6) that suffered from a peak overlap of an isomeric compound in negative ion mode.

Figure 7. Lipid synthesis pathway with the mean turnover rates for TG, DG, PG, PI, and PE. The mean turnover rates were indicated for lipid subclasses that have at least five measured turnover rates.

acid acyl chain (0.017 h−1), which implies that the recycling of free fatty acids does not seem to contribute significantly to the synthesis of PI(18:1_20:4). In summary, the same type of fatty acid chain in different glycerophospholipids showed variable turnover rates depending on the specific type of lipid species. Given the complexity of lipid metabolism, the use of multiple differentially labeled tracers such as [U−13C]-palmitic acid for the fatty acid recycling and 2H2O for de novo lipogenesis would be a promising way to differentiate the two lipid synthetic pathways. Precision of Lipid Turnover Measurement. The lipid species that identified in both positive and negative ion modes or appeared as two different adduct forms were exploited to validate reproducibility of our analytical platforms for lipid turnover measurement. In the MS-only experiment, three lipids including PE(P-16:0/22:5), PE(P-16:0/22:6), and PE(P-18:0/ 22:4) were commonly detected and showed little difference in their turnover rates between the two ion modes (Figure 6A). In the untargeted MS/MS experiment, the turnover rates of PE(16:0_20:4), PE(P-18:0/22:4), and PE(16:0_22:6) were also measured in both ion modes, and the results were consistent with each other, except for PE(16:0_22:6) (Figure 6B). A large deviation for PE(16:0_22:6) could be explained by the fact that its third mass isotopomer (M2) overlapped with the monoisotopic peak (M0) of a coeluted lipid on the negative

ion mode chromatogram. This overlap overestimated I2 and underestimated I0 (eq 1) for PE(16:0_22:6), resulting in underestimated turnover rate as indicated in Figure 6B. TG species that appeared as both ammonium and sodium adducts in positive mode MS provided another evidence for reliability of our method. In the MS-only experiment, seven TGs showed consistent results between the turnover rates measured in the two different adducts (Figure 6C). In short, our new framework for lipid kinetics measurement revealed a reliable analytical precision as long as a good chromatographic separation is ensured. Mapping Lipid Dynamics into Biological Pathway. To better understand lipid dynamics, the relationship between lipids and their turnover rates needs be investigated from the viewpoint of their metabolic pathways. Figure 7 represents the biological pathways for glycerolipids and glycerophospholipids based on KEGG pathways.39 TG, DG, PI, PE, PG, and PE-P were the six lipid subclasses with at least 5 quantified lipid turnover rates in the MS-only experiment. TG resulted in the fastest mean turnover rates compared to other glycerophospholipids, which is consistent with the result obtained from human breast cancer tissues.40 As seen in the biological pathways of TG, it can further be hydrolyzed into DG and enter the biosynthetic pathways of glycerophospholipid. In addition, fatty acids can also be hydrolyzed from TG during 6516

DOI: 10.1021/acs.analchem.7b05428 Anal. Chem. 2018, 90, 6509−6518

Article

Analytical Chemistry

(2) Rivellese, A.; Vaccaro, O.; Patti, L. Int. J. Clin. Pract. 2004, 58, 32−35. (3) Arner, P.; Bernard, S.; Salehpour, M.; Possnert, G.; Liebl, J.; Steier, P.; Buchholz, B. A.; Eriksson, M.; Arner, E.; Hauner, H.; Skurk, T.; Ryden, M.; Frayn, K. N.; Spalding, K. L. Nature 2011, 478, 110− 113. (4) White, D. C.; Tucker, A. N. J. Lipid Res. 1969, 10, 220−233. (5) Kanfer, J.; Kennedy, E. P. J. Biol. Chem. 1963, 238, 2919−2922. (6) Qi, J.; Lang, W.; Geisler, J. G.; Wang, P.; Petrounia, I.; Mai, S.; Smith, C.; Askari, H.; Struble, G. T.; Williams, R.; Bhanot, S.; Monia, B. P.; Bayoumy, S.; Grant, E.; Caldwell, G. W.; Todd, M. J.; Liang, Y.; Gaul, M. D.; Demarest, K. T.; Connelly, M. A. J. Lipid Res. 2012, 53, 1106−1116. (7) McLaren, D. G.; Cardasis, H. L.; Stout, S. J.; Wang, S.-P.; Mendoza, V.; Castro-Perez, J. M.; Miller, P. L.; Murphy, B. A.; Cumiskey, A.-M.; Cleary, M. A. Anal. Chem. 2013, 85, 6287−6294. (8) Zhang, L.; Vertes, A. Anal. Chem. 2015, 87, 10397−10405. (9) Hermansson, M.; Hanninen, S.; Hokynar, K.; Somerharju, P. Biochim. Biophys. Acta, Mol. Cell Biol. Lipids 2016, 1861, 1058−1065. (10) Li, J.; Hoene, M.; Zhao, X.; Chen, S.; Wei, H.; Haring, H. U.; Lin, X.; Zeng, Z.; Weigert, C.; Lehmann, R.; Xu, G. Anal. Chem. 2013, 85, 4651−4657. (11) Whitelegge, J. P.; Katz, J. E.; Pihakari, K. A.; Hale, R.; Aguilera, R.; Gomez, S. M.; Faull, K. F.; Vavilin, D.; Vermaas, W. Phytochemistry 2004, 65, 1507−1515. (12) Strawford, A.; Antelo, F.; Christiansen, M.; Hellerstein, M. K. Am. J. Physiol. Endocrinol. Metab. 2004, 286, E577−588. (13) Busch, R.; Neese, R. A.; Awada, M.; Hayes, G. M.; Hellerstein, M. K. Nat. Protoc. 2007, 2, 3045−3057. (14) Messmer, B. T.; Messmer, D.; Allen, S. L.; Kolitz, J. E.; Kudalkar, P.; Cesar, D.; Murphy, E. J.; Koduru, P.; Ferrarini, M.; Zupo, S.; Cutrona, G.; Damle, R. N.; Wasil, T.; Rai, K. R.; Hellerstein, M. K.; Chiorazzi, N. J. Clin. Invest. 2005, 115, 755−764. (15) Foletta, V. C.; Palmieri, M.; Kloehn, J.; Mason, S.; Previs, S. F.; McConville, M. J.; Sieber, O. M.; Bruce, C. R.; Kowalski, G. M. Metabolites 2016, 6, 34. (16) De Riva, A.; Deery, M. J.; McDonald, S.; Lund, T.; Busch, R. Anal. Biochem. 2010, 403, 1−12. (17) Hesketh, S.; Srisawat, K.; Sutherland, H.; Jarvis, J.; Burniston, J. Proteomes 2016, 4, 12. (18) Li, L.; Willard, B.; Rachdaoui, N.; Kirwan, J. P.; Sadygov, R. G.; Stanley, W. C.; Previs, S.; McCullough, A. J.; Kasumov, T. Mol. Cell. Proteomics 2012, 11, M111.014209. (19) Kim, T. Y.; Wang, D.; Kim, A. K.; Lau, E.; Lin, A. J.; Liem, D. A.; Zhang, J.; Zong, N. C.; Lam, M. P.; Ping, P. Mol. Cell. Proteomics 2012, 11, 1586−1594. (20) Lau, E.; Cao, Q.; Ng, D. C.; Bleakley, B. J.; Dincer, T. U.; Bot, B. M.; Wang, D.; Liem, D. A.; Lam, M. P.; Ge, J.; Ping, P. Sci. Data 2016, 3, 160015. (21) Busch, R.; Kim, Y. K.; Neese, R. A.; Schade-Serin, V.; Collins, M.; Awada, M.; Gardner, J. L.; Beysen, C.; Marino, M. E.; Misell, L. M.; Hellerstein, M. K. Biochim. Biophys. Acta, Gen. Subj. 2006, 1760, 730−744. (22) Wang, D.; Liem, D. A.; Lau, E.; Ng, D.; Bleakley, B. J.; Cadeiras, M.; Deng, M. C.; Lam, M. P.; Ping, P. Proteomics: Clin. Appl. 2014, 8, 610−619. (23) Price, J. C.; Holmes, W. E.; Li, K. W.; Floreani, N. A.; Neese, R. A.; Turner, S. M.; Hellerstein, M. K. Anal. Biochem. 2012, 420, 73−83. (24) Turner, S. M.; Murphy, E. J.; Neese, R. A.; Antelo, F.; Thomas, T.; Agarwal, A.; Go, C.; Hellerstein, M. K. Am. J. Physiol. Endocrinol. Metab. 2003, 285, E790−803. (25) Kloehn, J.; Saunders, E. C.; O’Callaghan, S.; Dagley, M. J.; McConville, M. J. PLoS Pathog. 2015, 11, e1004683. (26) Wahjudi, P. N.; K. Yee, J.; Martinez, S. R.; Zhang, J.; Teitell, M.; Nikolaenko, L.; Swerdloff, R.; Wang, C.; Lee, W. N. J. Lipid Res. 2011, 52, 2226−2233. (27) Lee, W. N.; Bassilian, S.; Guo, Z.; Schoeller, D.; Edmond, J.; Bergner, E. A.; Byerley, L. O. Am. J. Physiol. 1994, 266, E372−383.

lipolysis, and they then can be reused for the synthesis of membrane glycerophospholipids. In particular, since the membrane remodeling is actively performed in proliferating cells, a storage lipid TG is degraded to offer fatty acids required for the remodeling process.41 With our new lipid kinetics analysis method, the fastest turnover rate of TG in HeLa cell could be understood from the viewpoint of the lipid metabolic pathways.

■

CONCLUSION A new analytical framework enables quantification of the in vivo turnover rate of an individual lipid on a global scale. Our experimental platforms combining partial metabolic 2H2O labeling with LC-MS and LC-MS/MS were adequate for high-throughput kinetics analysis of more than 100 HeLa lipids at the species level, and demonstrated good reproducibility irrespective of the charge state or adduct forms of lipid ions. In addition, the turnover rate measurement of fragment ions by MS/MS scans could distinguish the difference in kinetics between fatty acyl chain and glycerol backbone within a lipid. The economy of 2H2O labeling makes it possible to apply our methodology to a long-term kinetics study for higher organisms including human. The unique feature of 2H2O labeling to induce 2H incorporation into multiple biomolecules including nucleic acids, carbohydrates, and proteins, as well as lipids, will also be beneficial for measuring fluxes at the systems level.

■

ASSOCIATED CONTENT

S Supporting Information *

The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/acs.analchem.7b05428. Number of measured lipids in the untargeted MS/MS analysis by changing the number of time point filtering criteria (Figure S1); correlation of measured lipid turnover rates between the MS-only and untargeted MS/MS analyses (Figure S2); measured turnover rates of PG(16:0_18:1) fragment ions in the untargeted analysis (Figure S3); details of the number of identified lipids in the untargeted MS/MS analysis (Tables S1 and S2); lists of determined lipid turnover rates and fitted results in the MS-only and untargeted MS/MS analyses (Tables S3 and S4); turnover rates of free fatty acid (Table S5); and a list of lipids and their measured turnover rates in targeted MS/MS analysis (Table S6; PDF).

■

AUTHOR INFORMATION

Corresponding Author

*Tel.: +82-62-715-3647. E-mail: [email protected]. ORCID

Tae-Young Kim: 0000-0002-8846-3338 Notes

The authors declare no competing financial interest.

■ ■

ACKNOWLEDGMENTS The authors acknowledge support for this work by the National Research Foundation of Korea (NRF-2014R1A1A1003643). REFERENCES

(1) Choy, P. C.; Siow, Y. L.; Mymin, D.; O, K. Biochem. Cell Biol. 2004, 82, 212−224. 6517

DOI: 10.1021/acs.analchem.7b05428 Anal. Chem. 2018, 90, 6509−6518

Article

Analytical Chemistry (28) Castro-Perez, J. M.; Roddy, T. P.; Shah, V.; McLaren, D. G.; Wang, S. P.; Jensen, K.; Vreeken, R. J.; Hankemeier, T.; Johns, D. G.; Previs, S. F.; Hubbard, B. K. J. Proteome Res. 2011, 10, 4281−4290. (29) Mitsche, M. A.; McDonald, J. G.; Hobbs, H. H.; Cohen, J. C. eLife 2015, 4, e07999. (30) Hellerstein, M. K.; Neese, R. A. Am. J. Physiol. 1999, 276, E1146−1170. (31) Folch, J.; Lees, M.; Sloane Stanley, G. H. J. Biol. Chem. 1957, 226, 497−509. (32) Kind, T.; Liu, K. H.; Lee, D. Y.; DeFelice, B.; Meissen, J. K.; Fiehn, O. Nat. Methods 2013, 10, 755−758. (33) Fahy, E.; Subramaniam, S.; Murphy, R. C.; Nishijima, M.; Raetz, C. R.; Shimizu, T.; Spener, F.; van Meer, G.; Wakelam, M. J.; Dennis, E. A. J. Lipid Res. 2009, 50 (Suppl), S9−14. (34) Liebisch, G.; Vizcaino, J. A.; Kofeler, H.; Trotzmuller, M.; Griffiths, W. J.; Schmitz, G.; Spener, F.; Wakelam, M. J. J. Lipid Res. 2013, 54, 1523−1530. (35) Savitzky, A.; Golay, M. J. Anal. Chem. 1964, 36, 1627−1639. (36) Berg, J. M.; Stryer, L.; Tymoczko, J. L.; Gatto, G. J. Biochemistry; Macmillan Learning, 2015. (37) Wendel, A. A.; Lewin, T. M.; Coleman, R. A. Biochim. Biophys. Acta, Mol. Cell Biol. Lipids 2009, 1791, 501−506. (38) Xu, Y.; Schlame, M. Chem. Phys. Lipids 2014, 179, 17−24. (39) Kanehisa, M.; Goto, S. Nucleic Acids Res. 2000, 28, 27−30. (40) Sakai, K.; Okuyama, H.; Yura, J.; Takeyama, H.; Shinagawa, N.; Tsuruga, N.; Kato, K.; Miura, K.; Kawase, K.; Tsujimura, T.; Naruse, T.; Koike, A. Carcinogenesis 1992, 13, 579−584. (41) Natter, K.; Kohlwein, S. D. Biochim. Biophys. Acta, Mol. Cell Biol. Lipids 2013, 1831, 314−326.

6518

DOI: 10.1021/acs.analchem.7b05428 Anal. Chem. 2018, 90, 6509−6518