Development and Evaluation of a Parallel ... - ACS Publications

Mar 22, 2016 - Development and Evaluation of a Parallel Reaction Monitoring ... In this study, we developed a quantitative metabolomics strategy in th...
0 downloads 0 Views 5MB Size
Article pubs.acs.org/ac

Development and Evaluation of a Parallel Reaction Monitoring Strategy for Large-Scale Targeted Metabolomics Quantification Juntuo Zhou,† Huiying Liu,† Yang Liu, Jia Liu, Xuyang Zhao, and Yuxin Yin* Institute of Systems Biomedicine, Department of Pathology, School of Basic Medical Sciences, Beijing Key Laboratory of Tumor Systems Biology, Peking-Tsinghua Center for Life Sciences, Peking University Health Science Center, No. 38 Xueyuan Road, Beijing 100191, China S Supporting Information *

ABSTRACT: Recent advances in mass spectrometers which have yielded higher resolution and faster scanning speeds have expanded their application in metabolomics of diverse diseases. Using a quadrupole-Orbitrap LC−MS system, we developed an efficient large-scale quantitative method targeting 237 metabolites involved in various metabolic pathways using scheduled, parallel reaction monitoring (PRM). We assessed the dynamic range, linearity, reproducibility, and system suitability of the PRM assay by measuring concentration curves, biological samples, and clinical serum samples. The quantification performances of PRM and MS1-based assays in Q-Exactive were compared, and the MRM assay in QTRAP 6500 was also compared. The PRM assay monitoring 237 polar metabolites showed greater reproducibility and quantitative accuracy than MS1-based quantification and also showed greater flexibility in postacquisition assay refinement than the MRM assay in QTRAP 6500. We present a workflow for convenient PRM data processing using Skyline software which is free of charge. In this study we have established a reliable PRM methodology on a quadrupole-Orbitrap platform for evaluation of large-scale targeted metabolomics, which provides a new choice for basic and clinical metabolomics study.

M

quantify over 339 metabolites from an extract of HCT8 cells with a linear range of over 4 orders of magnitude in a single chromatographic run, showing the power of high-resolution MS platforms in the study of metabolomics.16 Parallel reaction monitoring (PRM), originally described as high-resolution MRM or target MS/MS17,18 is another strategy that can be utilized on high-resolution MS platforms. A compound precursor ion is isolated in Q1, fragmented in Q2, and subsequently all generated MS/MS fragment ions are monitored in parallel on a high-resolution, accurate mass, full scan mass spectrometer. PRM has been used for targeted peptide quantification in high-resolution LC-MS platforms, and has shown reliable performance in several biological and clinical applications.19−22 However, to the best of our knowledge, there has been no report of application of PRM in large-scale targeted metabolite quantification. In this study, we developed a quantitative metabolomics strategy in the PRM mode. Initially, the dynamic range, linearity, reproducibility, and system suitability of the PRM assay were assessed in large-scale targeted metabolite quantification with various biological samples, and the perform-

etabolomics represents the collection, detection, and analysis of all kinds of small molecule metabolites, which are the end products of cellular processes.1,2 Common platforms used in metabolomics studies include nuclear magnetic resonance (NMR) and mass spectrometry (MS) coupled with gas chromatography (GC) or liquid chromatography (LC).3,4 Metabolomics has advanced rapidly in recent years on MS platforms with the development of a new generation of mass spectrometers that provide improved dynamic range, mass precision, and high resolution, and scan speed.5 For targeted metabolite quantification, multiple reaction monitoring (MRM) methodology performed on triple quadrupole instruments has been considered as the “gold standard”.6−8 For example, Min and colleagues were able to monitor 258 metabolites (289 Q1/Q3 transitions) from a single 15 min LC−MS acquisition in MRM mode.9 Nevertheless, MRM assays sometimes show nonspecific chromatographic peaks under a specific Q1/Q3, which may be caused by the resemblance of compounds with closely similar molecular weights and fragments, leading to difficulty in metabolite identification due to the lack of mass spectrum.10,11 Currently, technological improvements in high-resolution, full scanning MS instruments have led to new approaches for compound analysis, such as data-dependent acquisition (DDA), which is a powerful assay technique for nontargeted metabolite profiling.12−15 Xiaojing and colleagues, for example, were able to © 2016 American Chemical Society

Received: January 26, 2016 Accepted: March 22, 2016 Published: March 22, 2016 4478

DOI: 10.1021/acs.analchem.6b00355 Anal. Chem. 2016, 88, 4478−4486

Article

Analytical Chemistry

Figure 1. Overview of metabolomics strategy using the PRM mode. (A) Workflow for PRM method construction and sample analysis. (B) Schematic representation of PRM (upper) and MRM (lower) as performed on Q-Orbi and QQQ instrumentation, respectively. The precursor ions (shown as big colored circles) are selected in Q1 and fragmented in Q2 in a similar manner for both of these two kinds of assays. In MRM, each product ion transition (shown as small colored circles) is monitored as predesigned one at a time in distinct scans. In PRM, all possible product ion transitions are analyzed in one concerted high resolution and high mass accuracy mass analysis.

ance of PRM, MRM, and MS1-based quantification (used in DDA mode) were compared. In addition, software solutions for peak extraction and statistical analysis were explored using the Skyline software environment,23 which produces metabolomics data that can be analyzed on the MetaboAnalyst Web service.24



was coupled to a QTRAP 6500 MS (AB Sciex). The supplementary methods include more detailed information. Mass Spectrometry. PRM assays were performed using the Q-Exactive MS (Thermo Scientific). Typically, the mass resolution for MS1 scans and targeted metabolites MS/MS scans was 17 500. Final PRM acquisition monitoring 237 metabolite on schedule, consisted of 1 MS1 scan followed by targeted MS/MS scans in HCD mode. Acquisition was performed in positive ion mode (109 metabolites) and negative ion mode (128 metabolites). MRM analysis was performed using the QTRAP 6500 MS (AB Sciex) in a positive−negative ion switching mode without scheduling. Detailed information may be found in the supplementary methods section. Peak Extraction. Raw data collected from LC-QE-MS and LC-TQ-MS were processed on Skyline software according to the manufacturer protocols (https://skyline.gs.washington. edu/labkey/project/home/software/Skyline/begin.view). Briefly, raw MS data files were imported to the software and peak area of predesigned transitions corresponding to the 237 metabolites were autocalculated by the software. The ion match tolerance for peak extraction was set to 0.01 and 0.7 Da for PRM and MRM data, respectively. The results including metabolites names and peak area were exported in table format for further analysis. The supplementary methods include more detailed information about the peak extraction procedures. Statistical Analysis. Response curves were calculated with linear regression. The r2, CV (coefficient of variation)

EXPERIMENTAL SECTION

Sample Preparations. The human cell lines HEK293T and HCT116 and mouse embryonic fibroblasts (MEFs) were purchased from the American Type Culture Collection (ATCC). Cells were grown in RPMI 1640 with 10% FBS, 100 units/mL penicillin, and 100 μg/mL streptomycin at 37 °C in a humidified incubator with 5% CO2. Primary serum specimens from 19 T2DM (type 2 diabetes mellitus) patients or healthy people were obtained from the China-Japan Friendship Hospital. This study was approved by the Institute Research Ethics Committee for clinical studies. Only serum from individuals who agreed to give samples for the purpose of laboratory research was used. The supplementary methods include information about sample preparation. High-Performance Liquid Chromatography. For PRM assay, the Ultimate 3000 UHPLC system was coupled to QExactive MS (Thermo Scientific) for metabolite separation and detection. An Xbridge amide column (100 mm × 4.6 mm i.d., 3.5 μm; Waters) was employed for compound separation at 30 °C. For the MRM assay, the Nexera UHPLC system (Shimazu) 4479

DOI: 10.1021/acs.analchem.6b00355 Anal. Chem. 2016, 88, 4478−4486

Article

Analytical Chemistry

Figure 2. MS analysis result of 237 metabolites. (A) The extracted ion chromatogram (XIC) of 109 metabolites in positive ion mode. (B) The extracted ion chromatogram (XIC) of 129 metabolites in negative ion mode. (C) The XIC of the precursor-product ion pair 147.1 to 130.04958 for glutamine at a mass tolerance of 0.7 Da. (D) The XIC of the precursor-product ion pair 147.1 to 130.04958 for glutamine at a mass tolerance of 0.01 Da. (E) Full MS/MS spectra of glutamine acquired in PRM mode at a retention time of 12.5 min. (F) Full MS/MS spectra of lysine acquired in PRM mode at a retention time of 17.45 min.

about these metabolites is shown in detail in Table S1. These comprise the metabolites to be detected in our large-scale metabolite tests. The principle of PRM mode is selection of a particular mass in the first quadrupole of a mass spectrometer, and all fragments of the precursor ion are detected in parallel in the second stage of mass spectrometry at a high resolution (Figure 1B, upper). In our study, a set of metabolites were first identified according to the mass of their parent ions and fragments, and each retention time was aligned. Then an optimized and scheduled assay was generated to perform a large-scale metabolite analysis in the PRM mode. After acquisition of raw MS data, peak extraction and optimization were performed using Skyline software. The information thus acquired for these metabolites was then processed with multivariate statistical analysis using the MetaboAnalyst Web service.

distribution, and quantification of metabolites were represented as a histogram or box plot using GraphPad 6.0. Significance was determined by the Student’s t test. *p < 0.05; **p < 0.01; NS, not significant. Unsupervised hierarchical clustering (Pearson linkages), heat map generation, PCA and PLS-DA analysis were carried out with the MetaboAnalyst 3.0 Web service (http:// www.metaboanalyst.ca/).



RESULTS Conspectus of Metabolomics Strategy in the PRM Mode. In order to reliably apply the PRM assay to large-scale targeted metabolite quantification, we developed a workflow for PRM assay construction and data analysis. The workflow schematic is shown in Figure 1A. We selected a number of common metabolites from general metabolic pathways involved in various physiological and pathological processes such as bile acid biosynthesis, citrate cycle, and glycolysis. Information 4480

DOI: 10.1021/acs.analchem.6b00355 Anal. Chem. 2016, 88, 4478−4486

Article

Analytical Chemistry

Figure 3. Dynamic range and linearity evaluation of the PRM assay by concentration curves. (A) Two examples of the concentration curves in positive ion mode. (B) Two examples of the concentration curves in negative ion mode. (C) The r2 distribution of linear regression analysis of the metabolites. The r2 distribution is represented as a cumulate percentage distribution column graph; the percentage of metabolites that have an r2 smaller than the given r2 is shown. (D) The coefficient of variation (CV) distribution of triplicate analysis of each concentration point. The box plot shows the 75th/25th percentile, and the bar represents the median.

Spectral Library-Informed Metabolite Validation and Transition Selection for PRM Assay. Since it was difficult to fulfill an entire peak for each of the 237 compounds in a single run without time-scheduling, we divided the 237 targeted metabolites equally into seven segments according to molecular weight from low (60 Da) to high (900 Da). Analysis of these 237 metabolites in seven parts yielded sufficient information (MS/MS spectra and retention time) for further identification and method optimization. After peak detection and alignment, metabolites were identified and confirmed according to the MS/MS spectrum in public databases such as HMDB,25 METLIN,26 and Massbank27 with mass error lower than 0.01 Da. For method optimization, we chose the ideal fragment ions, adjusted CE values, and confirmed the exact elution time for each metabolite. Finally, a highly multiplexed PRM assay was constructed monitoring 237 metabolites in two 23 min assays

with a 4 min-RT scheduling window (positive ion mode monitoring of 109 metabolites, and negative ion mode monitoring of 128). Typical XIC chromatograms of the 237 metabolites acquired by the optimized PRM method and processed by Skyline are shown in Figure 2A (positive ion mode) and Figure 2B (negative ion mode). Evaluation of PRM Assay by Linearity and Dynamic Range of Concentration Curves. We acquired concentration curves in the PRM mode to evaluate linearity, dynamic range, and accuracy, which are important factors in quantification. PRM concentration curves were acquired for the set of 237 polar metabolites in serum sample, and 196 of the 237 metabolites were detected. After MS acquisition and peak extraction, the linear regression r2 and coefficient of variation (CV) of each detected metabolite were calculated. Figure 3A,B shows two examples of calculated concentration 4481

DOI: 10.1021/acs.analchem.6b00355 Anal. Chem. 2016, 88, 4478−4486

Article

Analytical Chemistry

Figure 4. Multiplexed PRM assay monitoring of 237 metabolites in common biological samples. (A) MS intensity (log2 transformed) distributions of eight biological samples including cells, medium, and serum. The box plot shows the 75th/25th percentile, and the bar represents the median. (B) The coefficient of variation (CV) distribution for triplicate analysis of eight biological samples. (C) Principal components analysis (PCA) score plot of these eight biological samples. (D) Clustering heat map by Pearson correlation of these eight biological samples. Rows represent the detected metabolites and columns represent the biological samples in triplicate, and the color represents metabolite intensity from low (blue) to high (red). For the PCA score plot and heat map, the colors represent different groups: red, 293T; green, 293T_M; deep blue, HCT116; blue, HCT116_M; pink, MEFs; yellow, MEFs_M; gray, Serum_Human; black, Serum_Mouse.

respectively. The MS intensity distributions of the metabolites are plotted in Figure 4A, and different kinds of samples were observed to have a different but small intensity range. The CVs across the triplicate acquisitions for each sample was less than 20% (Figure 4B). Metabolomics analysis was then performed. Principal components analysis (PCA) score plots showed that samples of biological triplicates clustered well, and different kinds of samples showed clear differences as shown in Figure 4C. The clustering heat map in Figure 4D shows similar features among biological triplicates but variations among different kinds of biological samples. These data represent a high level of performance of our assay strategy in biological metabolomics analysis. Comparisons of Large-Scale Targeted Metabolomics Quantification in PRM to MRM and MS1-Based Quantification Method. To understand the differences between PRM and other previously popular assay methods (MRM and DDA), two parallel detections of the 237 metabolites from biological samples were carried out in

curves for the positive ion mode (glycine and Malonyl-CoA) and negative ion mode (uric acid and dCTP), respectively. Figure 3C shows the r2 distribution of all the calculated concentration curves, in which nearly 50% of the metabolites have an r2 greater than 0.99, and more than 90% of the metabolites have an r2 greater than 0.85. This represents good linearity in the PRM assay. CV values calculated at different concentrations are shown in Figure 3D, and the majority of the concentrations have an average CV ≤ 20%, except the lowest concentration. Evaluation of PRM Assay by Monitoring 237 Metabolites in Common Biological Samples. To further evaluate the suitability and reproducibility of the PRM assay, PRM assays were performed for several biological samples. Three cell lines from different species (HCT116, 293T, MEFs), their corresponding culture media (HCT116-M, 293T-M, MEFs-M) and two serum samples (human and mouse) were evaluated in the PRM mode in triplicate. As a result 216, 196, and 180 metabolites were detected in cells, serum, and medium, 4482

DOI: 10.1021/acs.analchem.6b00355 Anal. Chem. 2016, 88, 4478−4486

Article

Analytical Chemistry

Figure 5. Comparison of large-scale targeted metabolomics quantification in PRM with MRM and MS1 based quantification methods. (A−C) The coefficient of variation (CV) distribution of triplicate analysis of eight biological samples using different assay methods. (D−F) Clustering heat map by Pearson correlation of the eight biological samples using different assay methods. Rows represent the detected metabolites and columns represent the biological samples in triplicate. The color represents metabolite intensity from low (blue) to high (red). (G−I) Principal components analysis (PCA) score plot of the eight biological samples using different assay methods. For the PCA score plot and heat map, the colors represent different groups: red, 293T; green, 293T_M; deep blue, HCT116; blue, HCT116_M; pink, MEFs; yellow, MEFs_M; gray, Serum_Human; black, Serum_Mouse.

synchrony with the MRM assay and MS1 based assay (full-MS scan in PRM mode). As a result, 230, 217, and 203 metabolites were detected by PRM, MRM, and MS1, respectively, and the total number of metabolites detected by the three kinds of assay method was 237. The reproducibility of these analytic measurements was evaluated by determining the CVs for triplicate injection of biological samples for each assay as shown in Figure 5A−C. Almost all of the samples yielded average CVs of less than 20%, although the MS1 method displayed a higher average CV. Metabolic analysis exhibited little obvious disparity

among these three methods, and this was supported by hierarchical analysis (Figure 5D−F) and PCA analysis (Figure 5G−I). However, in these experimental comparisons, the PRM assay showed advantages in metabolite identification. For example, lysine (146.1055 Da) and glutamine (146.0691 Da) could not be distinguished with QTRAP 6500, since these two molecules and their fragments have a mass difference of only 0.04 Da (Figure 2E,F). The extracted ion chromatogram (XIC) in MRM for precursor-product ion pair 147.1 to 130.04958 is 4483

DOI: 10.1021/acs.analchem.6b00355 Anal. Chem. 2016, 88, 4478−4486

Article

Analytical Chemistry

Figure 6. Metabolic phenotype validation of type 2 diabetes mellitus with the PRM assay. (A) Partial least-squares-discriminant analysis (PLS-DA) score plot in 3D view. Three groups (T2DM, blue; CTL, red; QC, green) are included. (B) Clustering heat map by Pearson correlation of the samples. Three groups (T2DM, blue; CTL, red; QC, green) are included. (C) The top 15 metabolites with high VIP scores and the trends of change among groups. (D) Quantification of the top 15 metabolites with high VIP scores in the PLS-DA model. Each bar represents the mean value and SEM (n = 19). Significance was determined by the Student’s t test. *p < 0.05; **p < 0.01; NS, not significant. See also Table S2.

because it is a metabolic disease with characteristic metabolic remodeling, suitable for the purpose of the study of metabolism. At the same time, metabolomics has been applied in T2DM studies for several years, and numerous references are available, making it easier to evaluate our PRM assay results. Figure 6A shows a 3D-PLS-DA score plot from 210 metabolites extracted from the serum of 19 T2DM patients and their healthy counterparts. Four QC samples were gathered, showing good stability of the assay method as well as the instrument performance. As shown in Figure 6A, the metabolic signatures of T2DM patient and normal samples showed noticeable separation, and no overfitting of the model was observed after cross validation and permutation testing (Figure S2). Figure 6B shows hierarchical clustering via Pearson correlation in MetaboAnalyst from these two sample sets, while the

shown in Figure 2C, and two peaks were extracted. However, using PRM mode in Q-Extractive, MS/MS fragmentation by HCD was obtained at a resolution of 17 500, and the XIC for the same ion pair showed only one peak (Figure 2D). In the MS/MS spectrum of the peak at 12.5 min, the fragment of m/z = 130.04970 was detected (Figure 2E), while in the peak at 17.5 min, the fragment of m/z = 130.08614 was detected (Figure 2F). Combining the MS/MS spectra with highresolution and standard spectra in public databases, these two peaks were confirmed as glutamine (12.5 min) and lysine (17.5 min). Evaluation of PRM Assay by Metabolic Phenotype Validation of Type 2 Diabetes Mellitus. To further evaluate the performance of the PRM assay in complex samples, a set of clinical samples were selected. We chose T2DM as a model 4484

DOI: 10.1021/acs.analchem.6b00355 Anal. Chem. 2016, 88, 4478−4486

Article

Analytical Chemistry

window of 0.7 Da. In contrast, PRM performs MS analysis at a resolution of 17 500−35 000 and identifies mass differences of less than 0.01 Da. In this context, accurate mass MS instruments offer significant advantages over conventional low mass resolution platforms, as ions from the metabolites of interest are more readily distinguished from those derived from the biological matrix. Another advantage of PRM is that it is less time-consuming for method development and shows more postacquisition flexibility. PRM is able to monitor the entire set of MS/MS fragment ions simultaneously without preselection in an assay development stage. When performing data processing, we deleted weak and interference-prone fragment ions and focused on stronger interference-free ions for final quantitative results. However, PRM still has disadvantages. Because of its relative slow scan speed, use of PRM was more time-consuming than MRM for analysis of large quantities of metabolites belonging to opposed ion modes. Considering the advantages of PRM and MRM assays, it is an ideal strategy for using PRM as a complementary tool for MRM when performing large-scale metabolite quantification assay. Once we have confirmed the MS/MS spectra and retention time of the metabolites by PRM, we can transfer the assay method to MRM assay to take advantage of the high scan speed. Another popular approach to metabolomics assay is DDA acquisition on a high-resolution Q-TOF MS or Q-Orbitrap MS.15,16 DDA acquisition has shown strong ability in nontargeted metabolomics profiling. For DDA assay, peaks extracted from MS1 scans are used for quantification, while MS/MS spectra are used for identification. On the basis of our data on targeted metabolites, MS1 based quantification is inferior to the two methods PRM and MRM, and this is likely due to the superior selectivity conferred on PRM and MRM by a second stage of MS. Actual quantification by MS1 is performed on peak areas of precursor ions, and no fragmentation is involved. So this mode is more likely to suffer from interference, especially when complex matrixes are involved.17

metabolite clustering is mostly consistent with disease status. Finally, the variable importance for the projection (VIP) parameter was used to select metabolites that most significantly contribute to discriminating between the T2DM patients and controls in this PLS-DA model. The top 15 metabolites with high VIP scores and the trends of change between these groups are shown in Figure 6C, and the quantification data is shown in Figure 6D. In accord with our result, many of these metabolites have been reported to be significantly changed in T2DM, such as the amino acids valine and proline,28 urea cycle metabolites urea and ornithine,29 deoxyuridine, and creatinine.30 These detected metabolites and their VIP scores are listed in detail in Table S2.



DISCUSSION For this study we constructed a PRM assay targeting 237 polar metabolites on a quadrupole-Orbitrap platform. Evaluation of a series of different kinds of biological samples showed excellent accuracy, sensitivity, and system suitability of the PRM assay for targeted metabolite quantification. At the same time we demonstrated the use of Skyline software for PRM data processing is convenient and accessible. The differences between the PRM assay, MRM assay, and the MS1-based assay were also illustrated. We conclude that the PRM assay is a powerful tool for targeted metabolite quantification as well as a method which perfectly complements and serves as an alternative for the MRM assay. PRM is a strategy that can be utilized on high-resolution MS platforms and has been used for targeted peptide quantification. Some studies suggest that PRM is more specific than SRM because more product ions can be used to confirm the identity of a peptide.17,22 However, no PRM assay for large-scale targeted metabolite quantification has been previously reported to our knowledge. On the basis of the metabolite concentration curves detected in serum (Figure 3), PRM exhibits ideal linearity, high precision at a range of over 3 orders of magnitude, and accurate measurement of relative mass intensity. Metabolic analysis of common biological samples (Figure 4) and clinical samples (T2DM patients and counterparts, Figure 6) show that PRM is an excellent approach for large-scale metabolite quantification in complex biological samples. In the present study we monitored the metabolites in separated ion modes due to the large amount of them (237 metabolites), but if the target metabolites are less or have separated elution times, a positive/negative switching mode in PRM would also be achieved well. The MRM assay on QQQ instruments has been used for evaluation of metabolites over a long period of time and is regarded as the gold standard for compound quantification.6 The advantages of QQQ instruments include faster scan speed (less than 5 ms dwell time for a single transition) and a positive−negative ion switching mode, making it ideal for targeted metabolite analysis. However, shortcomings of the MRM assay cannot be ignored. For general MRM assays, chemical standards are needed for compound identification and retention time alignment. For metabolomics study, hundreds of metabolites must be analyzed in a single assay, giving rise to difficulties in use of chemical standards for identification and retention time assignment. In performing PRM assays on a high-resolution instrument, the majority of metabolites were identified by high mass accuracy data. For example, the traditional MRM assay performed on QQQ instruments commonly uses unit resolution, representing an isolation



CONCLUSION The workflow demonstrated here is applicable with high confidence for quantitative analysis of 237 targeted metabolites using a Q-Exactive instrument in the PRM mode. Biological and clinical samples were used to validate the feasibility of this method. Comparisons of the performance of large-scale targeted metabolite analysis in the PRM, MRM, and MS1 modes strongly supports the feasibility and reliability of our strategy and provides a new perspective in metabolomics study, which is promising for future clinical metabolite determinations and metabolic analyses.



ASSOCIATED CONTENT

S Supporting Information *

The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/acs.analchem.6b00355. Supplementary methods, Figure S1, Tables S1 and S2 (PDF) Table S3 (XLSX)



AUTHOR INFORMATION

Corresponding Author

*Phone: +86-10-82805570. E-mail: [email protected]. 4485

DOI: 10.1021/acs.analchem.6b00355 Anal. Chem. 2016, 88, 4478−4486

Article

Analytical Chemistry Author Contributions

(22) Tang, H.; Fang, H.; Yin, E.; Brasier, A. R.; Sowers, L. C.; Zhang, K. Anal. Chem. 2014, 86, 5526−5534. (23) MacLean, B.; Tomazela, D. M.; Shulman, N.; Chambers, M.; Finney, G. L.; Frewen, B.; Kern, R.; Tabb, D. L.; Liebler, D. C.; MacCoss, M. J. Bioinformatics 2010, 26, 966−968. (24) Xia, J.; Sinelnikov, I. V.; Han, B.; Wishart, D. S. Nucleic Acids Res. 2015, 43, W251−257. (25) Wishart, D. S.; Tzur, D.; Knox, C.; Eisner, R.; Guo, A. C.; Young, N.; Cheng, D.; Jewell, K.; Arndt, D.; Sawhney, S.; Fung, C.; Nikolai, L.; Lewis, M.; Coutouly, M. A.; Forsythe, I.; Tang, P.; Shrivastava, S.; Jeroncic, K.; Stothard, P.; Amegbey, G.; Block, D.; Hau, D. D.; Wagner, J.; Miniaci, J.; Clements, M.; Gebremedhin, M.; Guo, N.; Zhang, Y.; Duggan, G. E.; Macinnis, G. D.; Weljie, A. M.; Dowlatabadi, R.; Bamforth, F.; Clive, D.; Greiner, R.; Li, L.; Marrie, T.; Sykes, B. D.; Vogel, H. J.; Querengesser, L. Nucleic Acids Res. 2007, 35, D521−526. (26) Tautenhahn, R.; Cho, K.; Uritboonthai, W.; Zhu, Z.; Patti, G. J.; Siuzdak, G. Nat. Biotechnol. 2012, 30, 826−828. (27) Horai, H.; Arita, M.; Kanaya, S.; Nihei, Y.; Ikeda, T.; Suwa, K.; Ojima, Y.; Tanaka, K.; Tanaka, S.; Aoshima, K.; Oda, Y.; Kakazu, Y.; Kusano, M.; Tohge, T.; Matsuda, F.; Sawada, Y.; Hirai, M. Y.; Nakanishi, H.; Ikeda, K.; Akimoto, N.; Maoka, T.; Takahashi, H.; Ara, T.; Sakurai, N.; Suzuki, H.; Shibata, D.; Neumann, S.; Iida, T.; Tanaka, K.; Funatsu, K.; Matsuura, F.; Soga, T.; Taguchi, R.; Saito, K.; Nishioka, T. J. Mass Spectrom. 2010, 45, 703−714. (28) Newgard, C. B.; An, J.; Bain, J. R.; Muehlbauer, M. J.; Stevens, R. D.; Lien, L. F.; Haqq, A. M.; Shah, S. H.; Arlotto, M.; Slentz, C. A.; Rochon, J.; Gallup, D.; Ilkayeva, O.; Wenner, B. R.; Yancy, W. S., Jr.; Eisenson, H.; Musante, G.; Surwit, R. S.; Millington, D. S.; Butler, M. D.; Svetkey, L. P. Cell Metab. 2009, 9, 311−326. (29) Shaham, O.; Wei, R.; Wang, T. J.; Ricciardi, C.; Lewis, G. D.; Vasan, R. S.; Carr, S. A.; Thadhani, R.; Gerszten, R. E.; Mootha, V. K. Mol. Syst. Biol. 2008, 4, 214. (30) Ng, D. P.; Salim, A.; Liu, Y.; Zou, L.; Xu, F. G.; Huang, S.; Leong, H.; Ong, C. N. Diabetologia 2012, 55, 499−508.



J.Z. and H. L. contributed equally to this work. J.Z., H.L., and Y.Y. designed the study; J.Z., H.L., Y.L., J.L., X.Z., and Y.Y. performed experiments; J.Z. and H.L. collected and analyzed data; J.Z., H.L., and Y.Y wrote the manuscript. Notes

The authors declare no competing financial interest.



ACKNOWLEDGMENTS The authors would like to acknowledge C. Jiang for help with instruction on metabolomics and H. Liang for providing cell samples. Research reported in this publication was supported by grants to Y. Yin from the National Natural Science Foundation of China (Grants 81430056, 31420103905, 21305005 and 31400695), the 111 Project (Grant B07001), and the Lam Chung Nin Foundation for Systems Biomedicine.



REFERENCES

(1) Griffin, J. L.; Shockcor, J. P. Nat. Rev. Cancer 2004, 4, 551−561. (2) Denkert, C.; Budczies, J.; Kind, T.; Weichert, W.; Tablack, P.; Sehouli, J.; Niesporek, S.; Konsgen, D.; Dietel, M.; Fiehn, O. Cancer Res. 2006, 66, 10795−10804. (3) Collins, F. S.; Green, E. D.; Guttmacher, A. E.; Guyer, M. S. Nature 2003, 422, 835−847. (4) German, J. B.; Hammock, B. D.; Watkins, S. M. Metabolomics 2005, 1, 3−9. (5) Michalski, A.; Damoc, E.; Hauschild, J.-P.; Lange, O.; Wieghaus, A.; Makarov, A.; Nagaraj, N.; Cox, J.; Mann, M.; Horning, S. Mol. Cell. Proteomics 2011, 10, M111.011015. (6) Kitteringham, N. R.; Jenkins, R. E.; Lane, C. S.; Elliott, V. L.; Park, B. K. J. Chromatogr. B: Anal. Technol. Biomed. Life Sci. 2009, 877, 1229−1239. (7) Ciccimaro, E.; Blair, I. A. Bioanalysis 2010, 2, 311−341. (8) Wei, R.; Li, G.; Seymour, A. B. Anal. Chem. 2010, 82, 5527− 5533. (9) Yuan, M.; Breitkopf, S. B.; Yang, X.; Asara, J. M. Nat. Protoc. 2012, 7, 872−881. (10) Tiller, P. R.; Yu, S.; Castro-Perez, J.; Fillgrove, K. L.; Baillie, T. A. Rapid Commun. Mass Spectrom. 2008, 22, 1053−1061. (11) Jemal, M.; Ouyang, Z.; Xia, Y. Q. Biomed. Chromatogr. 2010, 24, 2−19. (12) Wang, X.; Zhang, A.; Han, Y.; Wang, P.; Sun, H.; Song, G.; Dong, T.; Yuan, Y.; Yuan, X.; Zhang, M. Mol. Cell. Proteomics 2012, 11, 370−380. (13) Dunn, W. B.; Broadhurst, D.; Begley, P.; Zelena, E.; FrancisMcIntyre, S.; Anderson, N.; Brown, M.; Knowles, J. D.; Halsall, A.; Haselden, J. N.; Nicholls, A. W.; Wilson, I. D.; Kell, D. B.; Goodacre, R.; Human Serum Metabolome, C. Nat. Protoc. 2011, 6, 1060−1083. (14) Denery, J. R.; Nunes, A. A.; Dickerson, T. J. Anal. Chem. 2011, 83, 1040−1047. (15) Matsubara, T.; Tanaka, N.; Krausz, K. W.; Manna, S. K.; Kang, D. W.; Anderson, E. R.; Luecke, H.; Patterson, A. D.; Shah, Y. M.; Gonzalez, F. J. Cell Metab. 2012, 16, 634−644. (16) Liu, X.; Ser, Z.; Locasale, J. W. Anal. Chem. 2014, 86, 2175− 2184. (17) Ronsein, G. E.; Pamir, N.; von Haller, P. D.; Kim, D. S.; Oda, M. N.; Jarvik, G. P.; Vaisar, T.; Heinecke, J. W. J. Proteomics 2015, 113, 388−399. (18) Peterson, A. C.; Russell, J. D.; Bailey, D. J.; Westphall, M. S.; Coon, J. J. Mol. Cell. Proteomics 2012, 11, 1475−1488. (19) Kim, H.-J.; Lin, D.; Li, M.; Liebler, D. C. Mol. Cell. Proteomics 2015, mcp.O115.051813. (20) Schilling, B.; MacLean, B.; Held, J. M.; Sahu, A. K.; Rardin, M. J.; Sorensen, D. J.; Peters, T.; Wolfe, A. J.; Hunter, C. L.; MacCoss, M. J.; Gibson, B. W. Anal. Chem. 2015, 87, 10222−10229. (21) Gallien, S.; Bourmaud, A.; Kim, S. Y.; Domon, B. J. Proteomics 2014, 100, 147−159. 4486

DOI: 10.1021/acs.analchem.6b00355 Anal. Chem. 2016, 88, 4478−4486