Development of a Data-Independent Targeted Metabolomics Method

tive method and traditional metabolomics method. EXPERIMENTAL SECTION. Chemicals. The three carnitine standards octanoyl-L-carnitine chloride (carniti...
0 downloads 0 Views 3MB Size
Download PDF
Article pubs.acs.org/ac

Development of a Data-Independent Targeted Metabolomics Method for Relative Quantification Using Liquid Chromatography Coupled with Tandem Mass Spectrometry Yanhua Chen,†,⊥ Zhi Zhou,†,⊥ Wei Yang,†,∥ Nan Bi,‡ Jing Xu,† Jiuming He,† Ruiping Zhang,† Lvhua Wang,‡ and Zeper Abliz*,†,§ †

State Key Laboratory of Bioactive Substance and Function of Natural Medicines, Institute of Materia Medica, Chinese Academy of Medical Sciences and Peking Union Medical College, Beijing 100050, P. R. China ‡ Cancer Institute and Cancer Hospital, Chinese Academy of Medical Sciences and Peking Union Medical College, Beijing 100021, P. R. China § Centre for Bioimaging & Systems Biology, Minzu University of China, Beijing 100081, P. R. China ∥ Center for DMPK Research of Herbal Medicines, Institute of Chinese Materia Medica, China Academy of Chinese Medical Sciences, Beijing 100700, P. R. China S Supporting Information *

ABSTRACT: Quantitative metabolomics approaches can significantly improve the repeatability and reliability of metabolomics investigations but face critical technical challenges, owing to the vast number of unknown endogenous metabolites and the lack of authentic standards. The present study contributes to the development of a novel method known as “data-independent targeted quantitative metabolomics” (DITQM), which was used to investigate the label-free quantitative metabolomics of multiple known and unknown metabolites in biofluid samples. This approach initially involved the acquisition of MS/MS data for all metabolites in biosamples using a sequentially stepped targeted MS/MS (sst-MS/MS) method, in which multiple product ion scans were performed by selecting all ions in the targeted mass ranges as the precursor ions. Subsequently, scheduled multiple reaction monitoring (MRM) by LC-MS/MS of the metabolome was established for 1658 characteristic ion pairs of 1324 metabolites. For sensitive and accurate quantification of these metabolites, mixed calibration curves were generated using sequentially diluted standard reference plasma samples using established MRM methods. Relative concentrations of all metabolites in each sample were calculated without using individual authentic standards. To evaluate the reliability and applicability of this new method, the performance of DITQM was validated by comparison to absolute quantification of 12 acylcarnitines using authentic standards and traditional metabolomics analysis for lung cancer. The results proved that the DITQM protocol is more reliable and can significantly improve clustering effects and repeatability in biomarker discovery. In this study, we established a novel methodology to standardize and quantify large-scale metabolome, providing a new choice for metabolomics research and its clinical applications.

S

analysis of several types of compounds, including amino acids, nucleosides, lipids, sugars, and hormones.11 This makes it difficult to establish an integrated quantitative method. Current targeted quantitative metabolomics methods entail quantification of identified known compounds using standard references12 or metabolites with specific chemical groups, such as amines and phenolic hydroxyl groups.13 Alternatively, metabolic profiling can be used to quantify specific compounds in

ince its introduction in 1999, metabolomics has been used to screen and identify disease biomarkers1−3 and has been applied to various pharmaco-metabolomics4,5 and environmental sciences.6 Nowadays, metabolomics is emerging to be developed as a quantitative and standardized tool, such as proteomics and transcriptomics, which is universally recognized and applied to a broader range of research areas.7−10 The development of a quantitative metabolomics approach to improve the repeatability, comparability, and reproducibility of data among different periods and different sample instruments and laboratories is critical. In contrast to other “omics” methods, which are used to analyze single types of molecules, metabolomics involves the © 2017 American Chemical Society

Received: November 29, 2016 Accepted: June 2, 2017 Published: June 2, 2017 6954

DOI: 10.1021/acs.analchem.6b04727 Anal. Chem. 2017, 89, 6954−6962

Article

Analytical Chemistry

All frozen plasma samples were initially thawed at 4 °C and vortexed on ice before use. Subsequently, 10 μL of methanol and 140 μL of acetonitrile containing 36.3 ng/mL octanoyl Lcarnitine-d3 (internal standard) were added to 50 μL of plasma samples in an ice−water bath. Mixtures were vortexed for 5 min in a vortex mixer (IKA, MS) at 3000 rpm and centrifuged at 9168g (Sigma 3-30k) for 10 min at 4 °C to remove protein. Supernatants were then analyzed. To test linearity, 500 μL of pooled samples were used as standard plasma mixtures with metabolite concentrations P and were deproteinized by adding 100 μL of methanol and 1400 μL of acetonitrile. Lyophilized supernatants were reconstructed with 200 μL of water to a concentration of 2.5-fold that of the standard plasma mixture (2.5P). Subsequently, mixed calibration curves were generated using various concentrations in water. Serial concentrations for the mixed calibration curves were 2.5P, 1.25P, 0.625P, 0.313P, 0.156P, 0.078P, 0.039P, 0.020P, 0.0098P, 0.0049P, 0.0024P, 0.0012P, and 0.0006P. Subsequently, 10 μL of methanol and 140 μL of acetonitrile containing 36.3 ng/mL octanoyl L-carnitine-d3 as the internal standard were added to 50 μL aliquots of serial dilutions for sample analyses. A 2-fold volume of 2.5P was then injected to simulate 5P conditions, and samples were injected in triplicate for LC-MS analyses in the MRM scan mode. Stock solutions of carnitine C8:0, carnitine C10:0, and carnitine C12:0 were individually prepared in methanol at concentrations of 188, 144, and 106 μg/mL, respectively, and stored at 4 °C prior to use. Acylcarnitines were freshly prepared at concentrations of 9.4, 7.2, and 5.3 μg/mL, respectively, by mixing 50 μL aliquots of each stock solution with 950 μL of methanol. Working analyte solutions were generated using appropriate dilutions of the above solutions in methanol (1− 1000-fold). Acylcarnitine-free plasma samples were obtained by vortexing plasma with activated charcoal to prepare calibration standards.20,21 Aliquots of working solution (10 μL) were spiked into 50 μL aliquots of acylcarnitine-free plasma and 140 μL of acetonitrile containing 36.3 ng/mL octanoyl L-carnitined3 as the internal standard to generate calibration standard curves and QCs at appropriate concentrations for each compound. The calibration standards for carnitine C8:0, carnitine C10:0, and carnitine C12:0 ranged from 470 pg/mL to 470 ng/mL, 360 pg/mL to 360 ng/mL, and 265 pg/mL to 265 ng/mL, respectively. Data Acquisition with sst-MS/MS Using LC-HR MS/MS. An Ultimate 3000 UHPLC system was coupled to a Q-Exactive MS instrument (Thermo Scientific) for global targeted metabolic profiling. Chromatographic separation was performed using a reverse-phase Waters HSS T3 column (1.8 μm, 100 mm × 2.1 mm) at 35 °C with mobile phases of water containing 0.1% formic acid (B) and acetonitrile (D), respectively. A 2 μL pretreated sample was injected. The flow rate was 0.25 mL/min, and gradient elution was performed as follows: the system was equilibrated with 2% D for 8 min followed by linear increases to 60% over 15 min. After maintenance at 60% for 15 min, the content of the mobile phase was increased to 100% D over 2 min and maintained at 100% for 3 min. The mass spectrometer was operated in positive ion mode with a spray voltage of 3500 V, a capillary temperature of 350 °C, sheath gas flow rate of 40 L/min, auxiliary gas flow rate of 11 L/min, and probe heater temperature of 220 °C. To simultaneously generate MS and MS/MS data for as many metabolites as possible in proposed plasma samples, the

biofluids.14 To this end, untargeted quantitative metabolomics is performed using stable isotope labeling or metabolic labeling.15,16 Previously, Engelsberger et al. used 15N-KNO3 as the sole nitrogen source in an Arabidopsis cell line and obtained full metabolic labeling of cells.16 However, although relative quantitative data for nitrogenous compounds in cells were acquired by comparing labeled and unlabeled cell extracts, this method was not suitable for clinical samples. Quantification of comprehensive metabolites in biosamples is indispensable and emerging in metabolomics research. However, standard compounds are not available for all metabolites, and isotope labeling is time-consuming and limited by the complexity of spectra and incomplete labeling.16−18 To address these issues, we developed an unbiased quantitative strategy that does not require standards or labels, named “dataindependent targeted quantitative metabolomics” (DITQM). Initially, the comprehensive metabolite information was acquired in plasma mixtures using sequentially stepped targeted MS/MS (sst-MS/MS). The scheduled multiple reaction monitoring (MRM) analysis by the LC-MS/MS method could subsequently be established. The quantitative method was developed using stepped diluted standard reference plasma to generate mixed calibration curves. To evaluate the suitability of this novel DITQM method, the dynamic range, linearity, and precision were assessed using 48 plasma metabolites. The performance of DITQM was compared to those of the absolute quantitative method and traditional metabolomics method.

■

EXPERIMENTAL SECTION Chemicals. The three carnitine standards octanoyl-Lcarnitine chloride (carnitine C8:0), decanoyl-L-carnitine chloride (Carnitine C10:0), and lauroyl-L-carnitine chloride (Carnitine C12:0) were purchased from Larodan AB (Malmö, Sweden). Octanoyl L-carnitine-d3 chloride was purchased from Cambridge Isotope Laboratories, Inc. (MA01810, USA). Acetonitrile (HPLC grade), methanol (HPLC grade), and formic acid were acquired from Merck (Darmstadt, Germany). All other chemicals and solvents were of analytical grade. Biological materials were stored at −80 °C until use. Sample Collection. Fasting plasma samples from 20 lung cancer male patients (55 ± 4.9 years old) and 20 age-matched healthy Chinese male volunteers (53.2 ± 9.1 years old) were collected from the Cancer Institute and Hospital of the Chinese Academy of Medical Sciences (Beijing, China). The study was approved by the hospital ethics committee and with the approval of corresponding regulatory agencies. Of the patients, six were diagnosed with squamous cell carcinoma, three with adenocarcinoma, six with small-cell carcinoma, and five with undifferentiated carcinoma. Blood samples were collected into EDTA-K2 vacutainer tubes and were immediately cooled to 4 °C. Samples were centrifuged within 2 h at 1000g for 10 min at 4 °C (Sigma 330k). Supernatant samples (plasma) were then separated, transferred into new vials, and immediately stored at −80 °C until preparation. Sample Preparation. Equal aliquots from lung cancer patients and healthy controls were mixed into pooled plasma samples for targeted profiling using UHPLC-Q orbitrap MS, method validation using UPLC-Qtrap MS, and quality control (QC) of batch analyses. All of the samples were injected in a random order. QC samples were placed in the queue for every 12 authentic samples.19 6955

DOI: 10.1021/acs.analchem.6b04727 Anal. Chem. 2017, 89, 6954−6962

Article

Analytical Chemistry

Figure 1. Overview of quantitative metabolomics strategy by the DITQM method.

mass spectrometer was operated in full positive scan mode followed by a target MS/MS scan. The resolution of the full scan was set at 70 000, and the scan range was m/z 70−1000. The conditions for targeted MS/MS were as follows: resolution, 17 500; AGC target, 1 × 106; MIT, 100 ms; isolation window, 1 m/z; stepped NCE at 25, 35, and 45. MRM Analysis by LC-MS/MS. An Agilent 1200 series RRLC (rapid resolution liquid chromatography; Agilent Technologies, Waldbronn, Germany) was coupled to a 5500 Q-Trap system (AB Sciex, Toronto, Canada) with an electrospray ion (ESI) source. Chromatographic separation was performed on a Waters HSS C18 column (2.1 × 100 mm, 1.8 μm) at 35 °C. Subsequently, a 2 μL pretreated sample was injected into the column and eluted with a linear gradient at a flow rate of 250 μL/min using 0.1% (v/v) formic acid in water as solvent A and acetonitrile as solvent B with the following gradient: isocratic 2% B (8 min) for balance, 2−60% B (0−15 min), 60% B (15−26 min), 60−100% B (26−27 min), and 100% B (27−30 min). The Q-Trap system was operated in positive ion mode using MRM, and source-dependent parameters were optimized as follows: ion spray voltage, 5 kV; vaporizer temperature, 500 °C; nebulizing gas (GS1), 40 psi; drying gas (GS2), 45 psi; curtain gas, 30 psi. Nitrogen gas was used in all analyses, and data acquisition and processing were performed using Analyst software version 1.6. Data Handling. Multivariate statistical analysis was performed using SIMCA-P (version 13.0, Umetrics, Umeå, Sweden), in which PCA model was used to present the group situation and OPLS-DA model was used to identify markers.

Then, using SPSS statistics (version 16.0, SPSS Inc.), the markers that were normally distributed and had homogeneity of variance were then analyzed by Student’s t test with sample type as the class variable, and for the markers not fulfilling the demand on normal distribution or homogeneity of variance, we used a nonparameter test to analyze for differences between groups. Unless specified, a value of p < 0.05 was selected for discriminating significant differences throughout.

■

RESULTS Conspectus of Quantitative Metabolomics Method by DITQM. The developed quantitative metabolomics strategy mainly comprises three procedures; its workflow scheme is shown in Figure 1. First, the MS/MS spectra of all the metabolites in the plasma mixture could be acquired using a sequentially stepped targeted MS/MS (sst-MS/MS) method using liquid chromatography coupled with high-resolution tandem mass spectrometry (LC-HR MS/MS). The fragment ions and retention times of each metabolite could then be aligned. Accordingly, the scheduled MRM method using LCMS/MS for comprehensive metabolomics was partially optimized and established by transferring corresponding molecular ions and characteristic product ions to Q1/ Q3MRM transitions. Second, using an established scheduled MRM method, the important quantification factors, such as concentration curves for linearity, dynamic range, and repeatability were acquired for the metabolome using a stepped diluted standard reference plasma to generate mixed calibration curves. Thus, the raw MS data could be calculated as relative 6956

DOI: 10.1021/acs.analchem.6b04727 Anal. Chem. 2017, 89, 6954−6962

Article

Analytical Chemistry

Figure 2. MS/MS spectra collection and selection of ion pairs of large numbers of metabolites from standard plasma mixtures using sst-MS/MS with LC-HR MS/MS.

identification. By the way, to generate “one feature for one peak” metabolomics data, the fragment ions, isotope ions, and adduct ions (eg., [M + Na]+, [M + K]+, etc.) were manually removed according to the annotations that were marked by Rpackage CAMERA (http://www.bioconductor.org/packages/ release/bioc/) and further confirmed though corresponding extracted ion chromatograms (XICs). When these metabolites were transferred from high resolution MS to Q1/Q3MRM transitions using Q-Trap MS, some metabolites were missed, and some metabolites with similar molecular weight in the same retention time were removed to avoid interference from the isotope ions. Finally, a total of 1324 metabolite molecular ions and corresponding characteristic product ions were generated. Their heterogeneous distribution of plasma metabolites was shown in Figure S1. To improve the specificity for some isomer metabolites, two MRM ion pairs were chosen and used for qualification and relative quantification, respectively. Subsequently, scheduled MRM methods for the metabolome were established, and 1658 peaks were detected. The corresponding extracted ion chromatograms (XIC) are presented in Figure 3. Quantitative Metabolomics Using Stepped Diluted Pooled Samples. To establish a quantitative metabolomics method, linearity, dynamic ranges, repeatability, and reliability

concentrations to the standard reference plasma. Finally, the resulting quantitative information was processed using multivariate statistical analysis to find reliable potential biomarkers with a higher fold change in concentration. sst-MS/MS and Scheduled MRM Methods for Comprehensive Metabolome Analysis. To develop a reliable, high-throughput metabolomics method for comprehensive relative quantification, the gold standard for quantification analysis, i.e., scheduled MRM, was chosen for its high sensitivity, high scan rate, and large linear range. MRM is generally based on data-dependent analysis of known metabolites or metabolites with high responses. However, many pathologically significant metabolites are missed, especially those with low response or low abundance. Hence, the data-independent method was used to obtain metabolome data for both high- and low-abundance metabolites. Multiple isomers with similar retention times are often present in biological samples, and these can be blended in product ion spectra by low-resolution MS. Thus, high-resolution mass spectra were first used to acquire precursor and corresponding characteristic product ions, and chromatography time was also extended to 35 min to maximize isomer separation and obtain more product information for the metabolome. Since many metabolites in plasma are unidentified, an sstMS/MS method using Q-Orbitrap MS was developed to obtain all metabolites’ MS/MS information in a data-independent manner. As illustrated in Figure 2, sst-MS/MS was complemented by full scan mode and targeted MS/MS; targeted precursor ions from m/z 80.1 to 850.3 were scanned with an m/z step of 1. However, the retention times of targeted precursor ions were unknown, and the scan rates of the mass spectrometer failed to accommodate the fragmenting of numerous metabolites in one circle. Thus, approximately 30 precursor ions with certain m/z intervals were grouped into an inclusion list, and the assay was completed in 26 injections. In these experiments, the first injection was targeted for ions at m/ z 80.1, 106.1, 132.1, etc., and the second injection was targeted for ions at m/z 81.1, 107.1, 133.1, etc. In this way, MS/MS information (including accurate m/z, retention time, corresponding characteristic product ions) of as many metabolites as possible could be obtained, independent of metabolite type, concentration, or response and yielded sufficient information for scheduled MRM method optimization and further

Figure 3. Extract ion chromatograms for 1658 ion pairs from 1324 metabolites in positive ion mode by scheduled MRM using LC-MS/ MS. 6957

DOI: 10.1021/acs.analchem.6b04727 Anal. Chem. 2017, 89, 6954−6962

Article

Analytical Chemistry should be evaluated. Because the quantification of the comprehensive metabolome comprising 1324 metabolites is a huge workload, specific software must be developed to meet automation requirements. Here, to assess suitability of this method, 48 metabolites were initially selected to assess the performance of the strategy in terms of dynamic range, linearity, and reproducibility, and comparisons were made with the absolute quantification method. Relative concentrations were fitted linearly to standard sample−response relationships, and the dynamic ranges of 48 metabolites were defined (Table S1). Of these, 43 yielded good linear functions (r2 > 0.99) with RSD lower than 20% and accuracy between 80% and 120%. Exponential and multivariable functions were obtained for four metabolites; no fitted curve was obtained for one of these because of its low response. Notably, we found that more accurate quantification results were obtained with higher similarities of MS intensities between analytes and internal standards (IS). Therefore, multiple ion transitions were used for the same IS to obtain responses with varying intensity. The IS octanoyl L-carnitine-d3 has main product ions at m/z 229 and 85, and m/z 85 is the main product ion with high intensity (peak area = 5.5 ± 0.33 × 105). Hence, the mass transition 291/85 was used as an IS for metabolites with peak areas greater than 5.5 × 104. Moreover, another product at m/z 229 had a relatively low intensity (peak area = 4.7 ± 0.39 × 104). Hence, the mass transition 291/229 was used for metabolites with peak areas less than 5.5 × 104. Accordingly, the peak area of metabolite 440/85 was within 2.5 × 104 and its linearity was poor (r2 = 0.93) when 291.1/85 was used as the calibration reference. However, with a calibration reference of 291.1/229.1, r2 reached 0.999 and the range of linearity was wider. Accordingly, the calibration references of seven metabolites with low concentrations were changed to 291.1/229.1 (Table S1). Not all plasma metabolites yield linear concentration/ response relationships, and nonlinear correlations were observed for four metabolites, potentially reflecting compound properties, oversaturation of the mass analyzer, or matrix effects. As illustrated in Figure 4, peak areas for the metabolites increased slowly with concentration. Traditional metabolomics methods using peak area differences to reflect concentration changes based on linear assumptions would have calculated incorrect concentrations for such metabolites, resulting in failure to identify metabolites with significant changes in concentration. In contrast, DITQM circumvents this limitation because the relative concentration of each metabolite can be obtained with respect to the standard plasma mixtures based on the corresponding standard curves. In further studies, the analytical performance of the developed quantitative metabolomics method was compared with absolute quantification to evaluate the accuracy of the method. In these analyses, all experimental parameters were standardized, and sample preparation methods and chromatographic conditions were the same for both methods. Three authentic compounds, carnitine C8:0, carnitine C10:0, and carnitine C12:0, were spiked into acylcarnitine-free plasma and analyzed separately using absolute quantification and DITQM to compare sensitivity, linearity, precision, and accuracy. As shown in Table S2, the novel DITQM method sufficiently met the requirements for quantitative analysis of complicated metabolites. Compared with the standard quantitative method, DITQM provided a much wider dynamic range (over 2 orders of magnitude), with correlation coefficients (r2) exceeding

Figure 4. Four metabolites with nonlinear fitting: (A) 140/96 with RT 1.21 min; (B) 144/58 with RT 1.34 min; (C) 153/110 with RT 1.51 min; (D) 524/506 with RT 16.6 min.

0.995 (Table S2). Since the standard quantitative method detected acylcarnitines (approximately 0.2−0.5 ng/mL) in acylcarnitine-free plasma obtained by vortexing plasma with activated charcoal, the lower limit of quantification was likely affected. However, for corresponding accuracy and intraday precision, DITQM performed slightly worse than standard quantification. In addition, we note that it is difficult to prepare a standard reference plasma at high concentrations for DITQM (the highest concentration was only 5P). Thus, samples out of the linear range of detection required dilution before detection. To validate the reliability and utility of the DITQM method, the metabolite concentration in each group and their fold change were compared to absolute quantification. Three metabolites have their own authentic reference standard (carnitine C8:0, 288.1/85; carnitine C10:0, 316/85; carnitine C12:0, 344/85). Another nine acylcarnitines without reference standards were also analyzed, and absolute concentrations were determined by substituting peak areas into equations for similar standards. Relative concentrations were also acquired using DITQM. As shown in Figure 5, fold changes and significant differences were identified using metabolomics analyses. Specifically, DITQM results were similar to those from absolute quantification, supporting the utility of this alternative quantitative metabolomics method for metabolites without reference standards and unknown metabolites. Application of DITQM to Plasma Biomarker Discovery for Lung Cancer Metabolomics. To assess the clinical utility of DITQM, plasma samples were collected from 20 lung cancer patients and 20 healthy volunteers and were subjected to metabolomics analyses. Figure 6A,B shows principle component analysis (PCA) score plots from DITQM and traditional metabolomics. Specifically, three outliers in PCA score plot were acquired using the traditional metabolomics method, indicating large data variability that could be eliminated using DITQM (Figure 6A).These comparisons show that DITQM is 6958

DOI: 10.1021/acs.analchem.6b04727 Anal. Chem. 2017, 89, 6954−6962

Article

Analytical Chemistry

Figure 5. Comparison of absolute quantification and the DITQM method by quantitative analysis of 12 acylcarnitines in lung cancer patients and healthy control plasma, respectively. (Fold changes of down regulation in lung cancer patients were displayed in blue font; *, **, and *** indicate confidence levels of 0.02, 0.005, and 0.001, respectively.)

Figure 6. PCA score plots acquired by DITQM (A) and traditional metabolomics (B). (C) Heat map representation of hierarchical clustering of molecular features found in each plasma sample using different methods. Each line of this graph represents a precursor fragment (Ql/Q3) ion pair ordered by retention time and colored according to relative concentrations and baseline medians and means across samples. The scale from blue to red represents normalized abundance in arbitrary units.

biomarkers. In traditional metabolomics analyses, discriminatory metabolites were selected from comparisons of peak areas between lung cancer patients and healthy volunteers, whereas DITQM comparisons were based on concentration. As illustrated in Figure 7, DITQM could effectively improve the uniformity of the data. Thirteen and 15 discriminated metabolites were found, respectively, using traditional methods and DITQM. Nine of these were identical (304/85, 288/85, 314/85, 360/85, 316/85, 384/85, 328/311, 356/339, and 286/ 85). However, some acylcarnitines (344/85, 440/85, 370/85, and 412/85) that differed significantly in absolute quantification (Figure 5) were missed by traditional metabolomics and

helpful during sample calibration and group clustering. Figure 6C shows a heat map of discriminated molecules in each plasma sample following DITQM, with clustering that reflected metabolite concentrations in each sample, their relative contributions to sample clustering, and all sample clustering situations. These data indicate that several metabolites are grouped well in lung cancer patients and healthy controls, although several metabolites displayed significant overlap. Hence, subsequent OPLS-DA was used to select most of the discriminated metabolites. Metabolites that differed significantly between cancer patients and healthy controls were further selected as potential 6959

DOI: 10.1021/acs.analchem.6b04727 Anal. Chem. 2017, 89, 6954−6962

Article

Analytical Chemistry

Figure 7. Discriminatory metabolites identified using different methods: (A) traditional metabolomics method; (B) DITQM. The discriminated metabolites in the red dashed box are identical in both methods. The others are different discriminated metabolites.

independent targeted quantitative metabolomics by combining the advantages of high-resolution Q-Orbitrap MS and QQQ MS systems. The scheduled MRM analysis of 1324 metabolites has been developed using an sst-MS/MS method. The method evaluated the selected 48 metabolites with excellent sensitivity and repeatability, demonstrating the suitability of the established DITQM. Sequentially stepped targeted MS/MS (sst-MS/MS) is a novel approach that can be utilized on a high-resolution MS platform to acquire sufficient information (MS/MS spectra and retention time) for all metabolites from different kinds of biological samples. It is somewhat similar to the newly proposed “pseudotargeted metabolomics method” and “globally optimized targeted mass spectrometry” (GOT-MS).22,23 In pseudotargeted metabolomics, MRM ion pairs are generated on the basis of data-dependent analysis (DDA), which can acquire only the five most intense precursor ions within one full scan cycle. Remarkably, many metabolites would be missed using this type of scan mode, especially the low-response or lowabundance metabolites that might be significant in diseases.

were only discriminated by DITQM. These results indicate that the traditional metabolomics method may omit important biomarkers and identify false biomarkers. Hence, the present DITQM could be effectively used to select more reliable potential biomarkers and significantly improve the uniformity of data distributions and repeatability of statistical analyses.

■

DISCUSSION One of the advantages of quantitative metabolomics is that data can be generated independently of platforms or technologies; uniformly, without requirements for spectral alignment or binning, the method leads to significantly improved metabolomics accuracy and reliability. However, current targeted quantitative metabolomics methods always quantify the identified compounds using standard references. In metabolomics analyses, the identification of a large number of metabolites is time-consuming, and this approach faces critical challenges owing to the vast quantities of unknown endogenous metabolites and the lack of authentic standards. To address these challenges, we proposed a strategy to perform data6960

DOI: 10.1021/acs.analchem.6b04727 Anal. Chem. 2017, 89, 6954−6962

Article

Analytical Chemistry

development, DITQM may ultimately replace traditional metabolomics methods for relative quantification of metabolomes of biological samples and will facilitate quantitative comparative analyses of data between batches and across independent laboratories.

The GOT-MS method is performed on a triple quadrupole mass spectrometer and is capable of detecting broad metabolites, including unknowns, but the generated data with unit resolution were not advantageous to the metabolite identification. In contrast, sst-MS/MS performs MS analysis at a resolution of 17 500−35 000 with identified mass differences of less than 0.01 Da. At the same time, to avoid the interference of isotope peaks or similar molecular weight, the precursor ions were set at 30 intervals. All detected metabolites could be fragmented and identified, including unknowns and lowabundance metabolites. The scheduled MRM on the QQQ MS system has been used for the quantification of all the metabolites. A newly improved pseudotargeted metabolomics approach was proposed by using high resolution MS in the multiple ion monitoring (MIM) mode with time-staggered ion lists (tsMIM).24 While its sensitivity and specificity is relatively lower, MRM is regarded as the gold standard for compound quantification because of its high sensitivity, high specificity, and fast scan speed. Thus, after we confirmed the MS/MS spectra and retention time of the comprehensive metabolome by sst-MS/MS, we transferred and established the scheduled MRM analysis. Moreover, for some isomers or similar precursor ions with similar product ions and retention times, we found two characteristic ion pairs: the product ion with the strongest response was selected for quantification and another specific product ion, for qualification. Thus, the comprehensive metabolome of 1658 characteristic ion pairs from 1324 metabolites could be simultaneously analyzed using scheduled MRM. To acquire quantitative data for all plasma metabolites, sequentially diluted plasma was used to prepare mixed calibration curves, and comparisons of individual metabolite concentrations in each group were performed independently of the authentic standards. Traditional metabolomics is based on the linear hypothesis (y = ax + b) between metabolite concentration x and MS response y, with an infinitesimal b value. However, for many endogenous metabolites, b is relatively large or there is nonlinear behavior between concentration and MS response, all of which cause inaccurate metabolomics results. Additionally, standardization and uniformity of the metabolomics data consolidated from different periods, equipment, and laboratories is difficult using traditional metabolomics. Here, the proposed DITQM strategy acquired quantitative standard curves for all the metabolites, including their nonlinear properties. All MS data could be transferred into standard curves to calculate their relative concentration to a standard reference plasma. Thus, the described method would greatly improve the repeatability and standardization of metabolomics research.

■

ASSOCIATED CONTENT

S Supporting Information *

The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/acs.analchem.6b04727. Distribution of detected metabolites; retention times, MRM transitions, and linear fitting of selected metabolites; comparison of calibration, precision, and accuracy for quantification of acylcarnitines using two methods (PDF)

■

AUTHOR INFORMATION

Corresponding Author

*E-mail: [email protected]; [email protected]. ORCID

Zeper Abliz: 0000-0002-4876-2392 Author Contributions ⊥

Y.C. and Z.Z. contributed equally. The manuscript was written through contributions of all authors. All authors have given approval to the final version of the manuscript.

Notes

The authors declare no competing financial interest.

■

ACKNOWLEDGMENTS The authors are thankful for the financial support from the National High Technology Research and Development Program of China (863 Program) (No. 2014AA021101) and National Natural Science Foundation of China (No. 21405179).

■

REFERENCES

(1) Nicholson, J. K.; Lindon, J. C.; Holmes, E. Xenobiotica 1999, 29, 1181−1189. (2) Fiehn, O. Plant Mol. Biol. 2002, 48, 155−171. (3) Sreekumar, A.; Poisson, L. M.; Rajendiran, T. M.; Khan, A. P.; Cao, Q.; Yu, J.; Laxman, B.; Mehra, R.; Lonigro, R. J.; Li, Y.; Nyati, M. K.; Ahsan, A.; Kalyana-Sundaram, S.; Han, B.; Cao, X.; Byun, J.; Omenn, G. S.; Ghosh, D.; Pennathur, S.; Alexander, D. C.; Berger, A.; Shuster, J. R.; Wei, J. T.; Varambally, S.; Beecher, C.; Chinnaiyan, A. M. Nature 2009, 457, 910−915. (4) Clayton, T. A.; Lindon, J. C.; Cloarec, O.; Antti, H.; Charuel, C.; Hanton, G.; Provost, J. P.; Le Net, J. L.; Baker, D.; Walley, R. J.; Everett, J. R.; Nicholson, J. K. Nature 2006, 440, 1073−1077. (5) Clayton, T. A.; Baker, D.; Lindon, J. C.; Everett, J. R.; Nicholson, J. Proc. Natl. Acad. Sci. U. S. A. 2009, 106, 14728−14733. (6) Mujahid, M.; Prasuna, M. L.; Sasikala, C.; Ramana, C. V. J. Proteome Res. 2015, 14, 711−727. (7) Colaert, N.; Van Huele, C.; Degroeve, S.; Staes, A.; Vandekerckhove, J.; Gevaert, K.; Martens, L. Nat. Methods 2011, 8, 481−483. (8) Bhargava, V.; Ko, P.; Willems, E.; Mercola, M.; Subramaniam, S. Sci. Rep. 2013, 3, 1740−1749. (9) Schicho, R.; Shaykhutdinov, R.; Ngo, J.; Nazyrova, A.; Schneider, C.; Panaccione, R.; Kaplan, G. G.; Vogel, H. J.; Storr, M. J. Proteome Res. 2012, 11, 3344−3357. (10) Chylla, R. A.; Hu, K.; Ellinger, J. J.; Markley, J. L. Anal. Chem. 2011, 83, 4871−4880.

■

CONCLUSION In the present study, a DITQM strategy was first proposed to perform label-free metabolomics for relative quantification. The corresponding characteristic ion pairs in the plasma metabolome comprising 1324 metabolites were acquired using an sstMS/MS method. The scheduled MRM quantitative method was then established accordingly. The present data from preliminary targeted analyses show that DITQM is an effective tool for metabolomics applications. In future studies, we aim to apply DITQM to large-scale quantitative metabolomics research. In addition, because metabolome quantification processes are complex, specific software will be developed to meet automation requirements. Following optimization and 6961

DOI: 10.1021/acs.analchem.6b04727 Anal. Chem. 2017, 89, 6954−6962

Article

Analytical Chemistry (11) Birkemeyer, C.; Luedemann, A.; Wagner, C.; Erban, A.; Kopka, J. Trends Biotechnol. 2005, 23, 28−33. (12) Zhou, J.; Liu, H.; Liu, Y.; Liu, J.; Zhao, X.; Yin, Y. Anal. Chem. 2016, 88, 4478−4486. (13) Guo, K.; Li, L. Anal. Chem. 2009, 81, 3919−3932. (14) Kaur, P.; Rizk, N.; Ibrahim, S.; Luo, Y.; Younes, N.; Perry, B.; Dennis, K.; Zirie, M.; Luta, G.; Cheema, A. K. Mol. BioSyst. 2013, 9, 307−317. (15) Kluger, B.; Bueschl, C.; Neumann, N.; Stückler, R.; Doppler, M.; Chassy, A. W.; Waterhouse, A. L.; Rechthaler, J.; Kampleitner, N.; Thallinger, G. G.; Adam, G.; Krska, R.; Schuhmacher, R. Anal. Chem. 2014, 86, 11533−11537. (16) Engelsberger, W. R.; Erban, A.; Kopka, J.; Schulze, W. X. Plant Methods 2006, 2, 14−25. (17) Bueschl, C.; Krska, R.; Kluger, B.; Schuhmacher, R. Anal. Bioanal. Chem. 2013, 405, 27−33. (18) Chokkathukalam, A.; Kim, D. H.; Barrett, M. P.; Breitling, R.; Creek, D. J. Bioanalysis 2014, 6, 511−524. (19) Gika, H. G.; Theodoridis, G. A.; Wingate, J. E.; Wilson, I. D. J. Proteome Res. 2007, 6, 3291−3303. (20) Liachenko, N.; Boulamery, A.; Simon, N. Fundam. Clin. Pharmacol. 2015, 29, 499−509. (21) Huang, J.; Sun, J.; Chen, Y.; Song, Y.; Dong, L.; Zhan, Q.; Zhang, R.; Abliz, Z. Anal. Chim. Acta 2012, 711, 60−68. (22) Chen, S.; Kong, H.; Lu, X.; Li, Y.; Yin, P.; Zeng, Z.; Xu, G. Anal. Chem. 2013, 85, 8326−8333. (23) Gu, H.; Zhang, P.; Zhu, J.; Raftery, D. Anal. Chem. 2015, 87, 12355−12362. (24) Wang, Y.; Liu, F.; Li, P.; He, C.; Wang, R.; Su, H.; Wan, J. B. Anal. Chim. Acta 2016, 927, 82−88.

6962

DOI: 10.1021/acs.analchem.6b04727 Anal. Chem. 2017, 89, 6954−6962