MRM Method for Targeted


High-Throughput and Multiplexed LC/MS/MRM Method for Targeted...

0 downloads 169 Views 609KB Size

Anal. Chem. 2010, 82, 5527–5533

High-Throughput and Multiplexed LC/MS/MRM Method for Targeted Metabolomics Ru Wei,* Guodong Li, and Albert B. Seymour Metabolomics/Proteomics, Applied Quantitative Genotherapeutics, Pfizer Inc. 620 Memorial Drive, Cambridge, Massachusetts 02139 Target-based metabolomics, focused on a subset of metabolites representative of key pathways, is a valuable tool for assessing metabolic changes resulting from genetic mutation, altered gene expression, and protein dysfunction in a given disease state or as a consequence of an environmental perturbation, such as a pharmaceutical. However, simultaneously analyzing hundreds of endogenous metabolites presents a challenge because of their diverse chemical structures and properties. In this study, we report a high-throughput, sensitive, and reproducible method for target-based metabolomics studies. It combines different separation conditions, optimal ionization polarities, and the most sensitive triple-quadrupole MSbased data acquisition mode (MRM). In 10 min, 205 endogenous metabolites, divided into three subgroups (amino acids, sugar and nucleic acids, and organic acids), are sequentially analyzed on a LC/MS/MRM system. Low picogram sensitivity is achieved for more than half of the metabolites. A 3-4 order of linearity and assay coefficient of variation less than 15% are observed for ∼80% of the metabolites. In summary, we have established a multiplex LC/MS/MRM method for quantitatively profiling hundreds of known metabolites from complex biological samples. The methodology is generally applicable and easily expandable to include more endogenous or drug metabolites. The rapid growth in metabolite profiling research in recent years facilitates the characterization of a system from gene to metabolic activity. By targeting the smallest set of biochemically active molecules (metabolites), metabolite profiling reveals broad and dynamic insights into multiple metabolic pathways and networks that are the consequences of cellular activity. As we know today, the human metabolome is far smaller than proteome, making it more straight forwarding to characterize using high-content and -throughput approaches. With this set of biochemically active molecules, however, the challenges metabolomics face are very different. Unlike transcription profiling and proteomics, which are derived from 4 bases or 20 amino acids, metabolites have very diverse chemical structures, making analyte extraction, enrichment, separation, and analysis difficult. Two approaches have been used in metabolite profiling. The discovery based metabolomics approach, in which an attempt is * To whom correspondence should be addressed. E-mail: [email protected]. Phone: 617-551-3377. Fax: 617-551-3082. 10.1021/ac100331b  2010 American Chemical Society Published on Web 06/04/2010

made to detect all metabolite components by GCMS,1-7 LCMS,8-10 or NMR,11-15 has been adopted from early efforts for various research discovery studies such as gene function identification and toxicity studies. Today, several thousand metabolites have been characterized and advanced technology exists for evaluating them in multiple matrixes. However, target based metabolomics,16-19 which is detects and quantifies specifically selected metabolites, is growing slowly. Target-based approach is typically used in hypothesis-driven settings. It can analyze hundreds of analytes in minutes and hundreds of samples in hours with no structural/ identity ambiguity, but it is inefficient in detecting previously uncharacterized compounds because only selected analytes will be evaluated. Early efforts led by Fiehn,4 in GC/MS-based plant metabolomics and Nicholson and Lindon,12 in NMR-based body fluid or tissue metabolomics, are both discovery approaches. The (1) Thompson, J. A.; Markey, S. P. Anal. Chem. 1975, 478, 1313–21. (2) Ward, M. E.; Politzer, I. R.; Laseter, J. L.; Alam, S. Q. Biomed. Mass Spectrom. 1976, 3, 77–80. (3) Thompson, J. A.; Markey, S. P.; Miles, B. S.; Fennessey, P. V. Adv. Mass Spectrom. Biochem. Med. 1977, 2, 1–9. (4) Fiehn, O.; Kopka, J.; Dormann, P.; Altmann, T.; Trethewey, R. N.; Willmitzer, L. Nat. Biotechnol. 2000, 18, 1157–1161. (5) Fiehn, O. Comp. Funct. Genomics 2001, 2, 155–168. (6) Fiehn, O. Plant Mol. Biol. 2002, 48, 155–171. (7) Hall, R.; Beale, M.; Fiehn, O.; Hardy, N.; Sumner, L.; Bino, R. Plant Cell. 2002, 14, 1437–1440. (8) Want, E. J.; O’Maille, G.; Smith, C. A.; Brandon, T. R.; Uritboonthai, W.; Qin, C.; Trauger, S. A.; Siuzdak, G. Anal. Chem. 2006, 78, 743–52. (9) Paige, L. A.; Mitchell, M. W.; Krishnan, K. R.; Kaddurah-Daouk, R.; Steffens, D. C. Int. J. Geriatr. Psychiatry 2007, 22, 418–23. (10) Nordstro ¨m, A.; Want, E.; Northen, T.; Lehtio¨, J.; Siuzdak, G. Anal. Chem. 2008, 80, 421–9. (11) Dettmer, K.; Aronov, P. A.; Hammock, B. D. Mass Spectrom. Rev. 2007, 26, 51–78. (12) Lynch, M. J.; Masters, J.; Pryor, J. P.; Lindon, J. C.; Spraul, M.; Foxall, P. J.; Nicholson, J. K. J. Pharm. Biomed. Anal. 1994, 12, 5–19. (13) Nicholson, J. K.; Lindon, J. C.; Holmes, E. Xenobiotica 1999, 29, 1181– 1189. (14) Robertson, D. G.; Reily, M. D.; Sigler, R. E.; Wells, D. F.; Paterson, D. A.; Braden, T. K. Toxicol. Sci. 2000, 57, 326–337. (15) Lindon, J. C.; Nicholson, J. K.; Holmes, E.; Everett, J. R. Concepts Magn. Reson. 2000, 12, 289–320. (16) Munger, J.; Bajad, S. U.; Coller, H. A.; Shenk, T.; Rabinowitz, J. D PLoS Pathog. 2006, 2, 1165–1175. (17) Brauer, M. J.; Yuan, J.; Bennett, B. D.; Lu, W; Kimball, E. H.; Botstein, D.; Rabinowitz, J. D. Proc. Natl. Acad. Sci. U.S.A. 2006, 103, 19302–19307. (18) Shaham, O.; Wei, R.; Wang, T. J.; Ricciardi, C.; Lewis, G. D.; Vasan, R. S.; Carr, S. A.; Thadhani, R.; Gerszten, R. E.; Mootha1, V. K. Mol. Syst. Biol. 2008, 4, 214. (19) Lewis, G. D.; Wei, R.; Liu, E.; Yang, E.; Shi, X.; Martinovic, M.; Farrell, L.; Asnani, A.; Cyrille, M.; Ramanathan, A.; Shaham, O.; Berriz, G.; Lowry, P. A.; Palacios, I. F.; Tasan, M.; M.; Roth, F. P.; Min, J.; Baumgartner, C.; Keshishian, H.; Addona, T.; Mootha, V. K.; Rosenzweig, A.; Carr, S. A.; Fifer, M. A.; Sabatine, M; Gerszten, R. E J. Clin. Invest. 2008, 118, 3503– 3512.

Analytical Chemistry, Vol. 82, No. 13, July 1, 2010

5527

power of existing modern analytical techniques, such as LC/MS/ MRM, has been primarily used in pharmaceutical companies for drug metabolite quantitation, and has not been frequently applied in the metabolomics field. Because of the complexities of biological samples, a hyphenated technology is a common choice for metabolomics, among which, LC/MS has been widely used in recently years. Mass spectrometry (MS) has been a popular selection20,21 because of its high sensitivity, specificity (molecular weight detection), low cost, and coupling capability to separation technologies, such as liquid chromatography (LC), gas chromatography (GC), or capillary electrophoresis (CE). Among different types of mass spectrometers, such as ion trap, time-of-flight, orbitrap, and quadrupole mass spectrometers, a triple quadruple mass spectrometer is optimal for targeted metabolomics, based on its high sensitivity, high specificity, and excellent quantitation ability. Two stages of mass selection reside in a triple quadrupole mass spectrometer: precursor ion (MS1) and a fragment of the precursor ion (product ion, MS2) produce a molecular weight and structure specific sensitive measurement for a given analyte. The potential power of utilizing a triple quadrupole mass spectrometrybased quantitation technique, namely, multiple reaction monitoring (MRM), for metabolomics has not been fully recognized until recently. A triple quadrupole mass spectrometry-based quantitation for small molecules has been heavily used by analytical chemists for analyzing drug metabolites,22,23 hormones,24 pesticides,25,26 and herbicides27 with great precision (CV < 10%).28,29 To further increase selectivity and sensitivity of the triple quadrupole mass spectrometry-based quantitation assay, a front-end separation technique, such as LC, GC, or CE, is often added as the third dimension of separation. While a variety of separation techniques can be used to couple a triple quadrupole mass spectrometer, the polarity-based liquid chromatography (LC) stands out for its speed, simple sample pretreatment, and numerous choices of types of commercially available columns based on different separation mechanisms such as reverse phase, normal phase, or hydrophilic interaction, etc. Modern instrumentation advances, faster spectral scan rate, higher efficiency of ionization, ion introduction, and vacuum pumping system of a mass spectrometer, coupled with fast chromatography have added a multiplex and high-throughput feature to existing LC/MS, therefore, allowing analysis of tens to hundreds of endogenous or drug metabolites simultaneously.30-33 (20) Want, E. J.; Cravatt, B. F.; Siuzdak, G. ChemBioChem. 2005, 6, 1941–51. (21) Want, E. J.; Nordstro ¨m, A.; Morita, H.; Siuzdak, G. J. Proteome Res. 2007, 6, 459–68. (22) Zhong, D.; Chen, X.; Gu, J.; Li, X.; Guo, J. Clin. Chim. Acta 2001, 313, 147–150. (23) Kostiainen, R.; Kotiaho, T.; Kuuranne, T.; Auriola, S. J. Mass Spectrom. 2003, 38, 357–372. (24) Tait, A. D.; Abbs, D.; Teale, P.; Houghton, E. Biomed. Environ. Mass Spectrom. 1989, 18, 572–5. (25) Pico, Y.; Blasco, C.; Font, G. Mass Spectrom. Rev. 2004, 23, 45–85. (26) Hernandez, F.; Sancho, J. V.; Pozo, O. J. Anal. Bioanal. Chem. 2005, 382, 934–946. (27) William, L. B. Anal. Mass Spectrom. Herbic. 2004, 23, 1–24. (28) Paul, W.; Steinwedel, H. Z. Naturforsch. 1953, 8a, 448–450. (29) Yost, R. A.; Enke, C. G. J. Am. Chem. Soc. 1978, 100, 2274–2275. (30) Piraud, M.; Vianey-Saban, C.; Petritis, K.; Elfakir, C.; Steghens, J. P.; Morla, A.; Bouchu, D. Mass Spectrom. 2003, 17, 1297–1311. (31) Thieme, D.; Sachs, H. Anal. Chim. Acta 2003, 492, 171–186. (32) Gergov, M.; Ojanpera, I.; Vuori, E. J. Chromatogr. B 2003, 795, 41–53. (33) Mueller, C. A.; Weinmann, W.; Dresen1, S.; Schreiber, A.; Gergov, M. Rapid Commun. Mass Spectrom. 2005, 19, 1332–1338.

5528

Analytical Chemistry, Vol. 82, No. 13, July 1, 2010

Several target-based metabolomics methodology studies have been reported. Piraud et al30 reported analysis of 68 underivatized amino acids using a LC/MS/MRM technique. Using the same MRM techniques, Thieme et al31 improved screening capabilities in forensic toxicology studies; Gergov et al32 developed a MRM method for screening 238 drugs in blood, and Mueller et al33 used MRM as a survey scan in their assay for 301 drugs. More recently, Rabinowitz’s research group reported on a LC/MS/MRM method which analyzes 174 metabolites.34,35 In this method, two 45-min runs on one or two systems are required for each sample and the same LC condition were used for both positive and negative ESI MS analysis.34 These existing methods can only analyze similar classes of metabolites or drugs using one chromatography separation and one ionization polarity. The throughput was limited by column washing and equilibrium time, and the number of metabolite classes which can be simultaneously measured in one assay was limited by both ionization polarity and MS detection time (dwell time). In this study, we present a high throughput LC/MS/MRM method for target based metabolite profiling across different classes of metabolites (amino acids, sugars and nucleic acids, and organic acids). Method sensitivity, limits of detection, and dynamic range were assessed using a mixture of metabolite standards. By analyzing pooled rat plasma samples, we were able to test metabolite detection rate and establish method variations in the detected metabolites. This method allows simultaneous analysis of different class of small molecules, metabolites, or drugs in a short time, therefore enabling a quantitative profiling of hundreds to thousands of samples. EXPERIMENTAL SECTION Materials. Most metabolite standards and mobile phase additives (acids, bases, and salts) were purchased from Sigma Chemical Co. (St. Louis, MO). Fructose-1-phosphate was purchased from George Uhe Company, Inc. (Paramus, NJ). Isotope labeled metabolites were purchased from Cambridge Isotope, Inc. (Woburn, MA). Metabolite standards purchased are minimal 98% purity. All organic solvents and water used in sample or standard solution preparation and for LC/MS mobile phases were HPLC grade and purchased from J. T. Baker (Phillipsburg, NJ). Rat plasma samples were kindly provided by an internal project team. Individual Standard Solution and Standard Mixture Solution. Stock solutions were prepared, for each metabolite standard, at a concentration of 1.0 mg/mL or higher in 20/80 methanol/ water (v/v), 0.075% NH4OH. Final 2 µg/mL individual metabolite standard solutions were prepared, by serious dilution from stock, and were used for mass spectrometric tuning. A standard mixture of all metabolites at 10, 20, or 40 µg/mL for metabolites having higher, medium, and lower measuring sensitivities, respectively, under current experimental condition, was prepared, from individual standard stock solution, and then used to generate 11 additional lower concentration levels via series dilutions for linearity assessment. (34) Bajad, S. U.; Lu, W.; Kimball, E. H.; Yuan, J.; Peterson, C.; Rabinowitz, J. D. J. Chromatogr. A 2006, 1125, 76–88. (35) Lu, W.; Kimball, E. H.; Rabinowitz, J. D. J. Am. Soc. Mass Spectrom. 2006, 17, 37–50. (36) Roy, S. M.; Anderle, M.; Lin, H.; Becker, C. H. Int. J. Mass Spectrom. 2004, 238, 163–171. (37) Zhou, H.; Kantor, A. B.; Becker, C. H. Metabolome Anal. 2005, 137–157.

Figure 1. System configuration.

Metabolite Reconstitution Solution. A reconstitution stock solution, containing 50 µg/mL valine-d8, 300 µg/mL of glucose13 C6-d7, and 200 µg/mL of citrate-d4, was prepared, aliquotted in 60 µL, and stored in -20 °C freezer. Upon the analysis, an aliquot of reconstitution stock solution was diluted by 100 times with water; 50 µL of this metabolite reconstitution solution, containing 0.5 µg/mL valine-d8, 3 µg/mL of glucose-13C6-d7, and 2 µg/mL of citrate-d4, was used to reconstitute dry metabolite extract. Plasma Metabolite Extraction and Reconstitution. A pooled rat plasma sample, in five replicates, was used to assess metabolite detection rate and detection variation. The rat blood samples from experiment rats were immediately centrifuged at 3000 × g at 4 °C for 10 min after blood draw and plasma were collected, pooled, aliquotted in 1 mL size, and stored in -80 °C freezer. One vial (1 mL) of rat plasma was taken from -80 °C freezer and gradually thawed at -20 °C (2 h), 4 °C (1 h), and room temperature (30 min). The thawed plasma was then vortexed and equally divided into 5 vials to generate 5 replicates before metabolite extraction starts. For protein removal and metabolite extraction, 37.5 µL of plasma was slowly added into 75 µL of extraction buffer that contains 1 µg/mL of phenylalanine-d8, 2 µg/mL of thymine-d4, and 0.1% formic acid in 80/20 ethanol/water (v/v), vortex for 1 min, and incubated at 4 °C for 2 h. The solution is vortexed again and centrifuged at 14,000 × g at 4 °C for 15 min; 50 µL of the supernatant was transferred into a 96-well PCR plate (VWR, West Chester, PA) and dried down under N2 at 27 °C in TurboVap (Caliper, Hopkinton, MA). The dry extract is then stored at -20 °C for the next day LC/MS/MS analysis. After 60 min at 4 °C and 15 min at room temperature, the dry metabolite extract was reconstituted into 50 µL of metabolite reconstitution solution and placed at room temperature for 15 min and at 4 °C for 60 min before LC/MS/MS analysis. LC/MS/MS System Components. The system is composed of (1) three Agilent 1200 series pumps/degassers (Agilent, Santa Clara, CA), (2) a customized Leap HTS autosampler (Leap Technologies, Carrboro, NC) equipped with three injectors, one stream select valve, and a valve self-washing system, and (3) an Applied Biosystems API 4000 triple quadrupole mass spectrometer (Applied Biosystems Inc., Framingham, MA), equipped with a polarity interchangeable electrospray TurboIonSpray source. A system configuration is illustrated in Figure 1. Liquid Chromatography. HPLC separations were carried out on three different columns on reverse phase mode for 9.5 min. A Luna Phenyl-Hexyl column (1.0 × 50 mm, particle size 5 µm, Phenomenex, Torrance CA) was used to separate most of amino acids, and nucleic bases, in reverse phase mode. The mobile

phases used were 0.1% acetic acid in H2O (A) and acetonitrile (B). A 9.5-min cycle time starts from 0% B at 0.1 mL/min for 1 min, to a linear increase to 100% B in 0.5 min at 0.1 mL/min, hold at 100% B for 4 min at 0.2 mL/min, and then decrease to and hold at 0% B for 4 min at 0.1 mL/min. An Atlantis T3 OBD column (1.0 × 50 mm, particle size 5 µm, Waters, Milford, MA) was used to separate nucleotides, sugar, sugar phosphates, and sugar alcohols. The mobile phases used were 1% hexafluoroisopropanol (HFIP), 0.015% ammonium hydroxide in water (A) and acetonitrile (B). A 9.5-min cycle time starts from 0% B at 0.1 mL/min for 4.3 min, to a linear increase to 100% B in 0.4 min at 0.1 mL/min, hold at 100% B for 3.9 min at 0.15 mL/min, and then decrease to and hold at 0% B for 0.8 min at 0.1 mL/min. A Synergi Polar-RP, 80 Å column (1.0 × 50 mm, particle size 4 µm, Phenomenex, Torrance, CA), was used to separate most organic acids and some nucleotides. The mobile phases used were 5 mM ammonium acetate in 5/95 acetonitrile/water (A) and 95/5 acetonitrile/water (B). A 9.5-min cycle time starts from 95% B at 0.1 mL/min for 3 min, 0% B for 3.9 min at 0.1 mL/min, to a linear increase to 100% B in 0.5 min at 0.15 mL/min. Mass Spectrometry. All metabolites were automatically (flow injection) or manually (infusion) tuned for ionization polarity, optimal declustering potential (DP), product ion, and collision energy (CE) using metabolite standard solutions. Automation (Sciex, Toronto, ON, Canada) was used to control tuning in flow injection mode. 0.1% formic acid in 50/50 acetonitrile/water (v/ v); flowing at 10 or 30 µL/min was used to deliver compounds into electrospray ionization source for both positive and negative tuning in infusion or flow injection mode. Ion Spray (IS) potential was 5000 V for positive mode and 4500 V for negative mode. Nebulizer gas (GS1) and bath gas (GS2) were 5 psi, curtain gas (CUR) was 12 psi, and collision gas (CAD) was 5 psi. Source temperature (TEM) was set to zero and Interface Heater (ihe) was ON. House oil-free air, zero grade air, and UHP nitrogen were used as source exhaust gas, nebulizer/bath gas, and curtain/ collision gas respectively. The concentrations of metabolite standard solutions for tuning were 2 µg/mL or higher. During LC/MS/MRM run, the mass spectrometric analysis for each metabolite was performed at optimized polarity at multiple reaction monitoring (MRM) mode on an ESI triple quadrupole mass spectrometer. Total flow eluent from each column was sequentially directed into the TurboIonSpray source. The polarity of the ionization mode is switched from positive for phenyl-hexyl column eluent to negative for both Atlantis and Polar-RP columns. The mass spectrometric data acquisition time for three columns is 9 min, and the dwell time for each MRM channel is 12 ms. The Analytical Chemistry, Vol. 82, No. 13, July 1, 2010

5529

common mass spectrometric parameters are the same as tuning conditions described above, except: nebulizer gas (GS1) and bath gas (GS2) were 50 and 60 psi, respectively; curtain gas (CUR) was 20 psi, and collision gas (CAD) was 7 psi; source temperature (TEM) was 400 °C, for both positive and negative modes. Injection Sequence and System Clean. A customized injection program was developed to introduce a sample into three injectors and three different columns sequentially. A total volume of 15 µL was drawn, 5 µL of which was then dispensed into each injector/column before column flow was directed into mass spectrometer ionization source. After each injection, a valve selfwashing system is trigged (contact closure) to deliver 1 mL of 70% methanol, 29.9% water, 0.1% formic acid, and then 1 mL of water to clean injector. The same washing solutions were also delivered to clean syringe upon injection cycle completion. Data Processing. Extracted mass chromatogram peaks of metabolites were integrated using MultiQuant software from ABI (Sciex, Toronto, ON, Canada). Peak areas of corresponding metabolites are then used, as quantitative measurements, for assay performance assessments such as assay variation, linearity etc. RESULTS AND DISCUSSION We have developed a LC/MS/MRM method for quantitatively profiling hundreds of endogenous small molecules in a 9.5-min run with good reproducibility, as well as high sensitivity inherited from LC/MS/MRM based assay. Compared to other target based metabolomics methods, this method applies multiple and fast chromatographic separation conditions and different ionization polarities in a single LC/MS run in a high-throughput fashion. Metabolites Assayed. 205 known metabolites and 5 internal standards were included in this reported assay. The compound name, formula, monoisotopic molecular weight, KEGG, and METLIN ID are listed in Table S-1 (Supporting Information). Metabolites were included for their importance and/or relevance to metabolic pathways such as carbohydrate and energy metabolisms, and the assay feasibility with the extraction method and technology platform being utilized. On the basis of the KEGG metabolic pathways, this panel of 205 metabolites is representative of 7% of lipid metabolism, 20% of carbohydrate metabolism, 22% of metabolism of cofactors and vitamins, 51% of amino acid metabolism, including metabolism of other amino acids, 46% of nucleotide metabolism, 44% of energy metabolism, 33% of biosynthesis of polyketides and nonribosomal peptides, 4% of glycan biosynthesis and metabolism, and 7% of biodegradation of xenobiotics. Optimization of Mass Spectrometry and Combined LC/ MS/MRM Strategy. Experimental tuning was used to decide ionization polarity, to select the best product ion (Q3 ion) and optimize collision energy (CE), as well as declustering potential (DP). All tuning data were manually examined to ensure proper selection of ionization polarity and product ion. In this report, minimizing potential interference between MRM channels was also taken into consideration when selecting a product ion. We have optimized a total of 206 acquisition channels for 205 metabolites and 5 internal standards. The precursor and product ions, declustering potential and collision energy, for each MRM channel are listed in Table S-2 (Supporting Information); 175 metabolites have a single and unique acquisition MRM channel. Fifteen molecules including 4 internal standards indicated by 5530

Analytical Chemistry, Vol. 82, No. 13, July 1, 2010

comment code “c”, “e”, “f”, and “i” in Table S-2 (Supporting Information), have more than one channel. Among these, homeserine (M0040), and threonine (M0063) have unique channels in addition to a shared one, and they are used to increase specificity and assess its contribution in the response from primary shared channel. Each of fructose-2,6,-bisphosphate (F-26-BP, M0103), glucose-1,6-bisphosphate (G-16-BP, M0196), fructose-1-phosphate (F1P, M0104), fructose-6-phosphate (F6P, M0105), glucose-1phosphate (G1P, M0106), glucose-6-phosphate (G6P, M0107), has a second shared channel for better specificity. 37 metabolites, indicated by comment code “a”, “b”, “d” “e”, and “f” in Table S-2 (Supporting Information), have shared channels due to the same precursor and product ion. Among these metabolites, only glutamine (M0035) and lysine (M0045) can be chromatographically separated from each other; and hydroxyproline (M0041) can be chromatographically separated from leucine (M0042) and isoleucine (M0044). The rest of the metabolites having shared channels are assessed together with their shared metabolites for sensitivity, linearity, detection rate, and reproducibility, because they are not distinguishable from their shared metabolites under this current method. One of the biggest challenges in metabolomics studies is to assay diverse classes of metabolites. They cannot be all separated by a single chromatography method, and they cannot be all optimally ionized in a single ionization mode. Sacrifice in separation or sensitivity has to be made in existing methods, or samples have to be analyzed multiple times using one LC/MS system or multiple LC/MS systems.34 Our strategy is to divide the targeted molecules or their MRM channels into subgroups, each of which will then be separated by the most suitable chromatography method and ionized by an optimal ionization polarity in mass spectrometer. The column and ionization polarity information for each metabolite is listed in Table S-2 (Supporting Information). The 205 metabolites reported in this assay were divided into three subgroups running on three different chromatography columns. Five metabolites, adenosine (M0079), lactate (M0126), xanthine (M0152), ribose-5-phosphate (ribose-5-P, M0137), ribulose-5-phosphate (ribulose-5-P, M0138), and 4 isotope labeled internal standards, glucose-13C6-d7 (NM002), citrate-d4 (NM003), phenylalanine-d8 (Phe-d8, NM004), and thymine-d4 (NM005), are analyzed in at least two subgroups to provide experiment and LC/MS/MRM system performance quality control (QC). The first subgroup (79 molecules) is composed mainly of amino acids and, nucleic bases or nucleosides. They are separated on a reverse phase phenyl-hexyl based column, which provides excellent selectivity and stability, followed by positive electrospray ionization mass spectrometric analysis. The majority of the metabolites in the second subgroup (61) are nucleotides, sugars, and sugar phosphates. A 1.0 mm Atlantis T3 OBD column, which has excellent stability and retention for polar compounds, and negative electrospray ionization were selected for this subgroup of metabolites for optimal separation and sensitivity. Most of the organic acids and some nucleotides make up the third subgroup (67 molecules). These are separated on an ether-linked phenyl with polar end-cappingbased Synergi Polar-RP column, followed by negative electrospray ionization mass spectrometric detection.

Figure 2. Injection cycle scheme. Column 1: Luna phenyl-hexyl column (1.0 × 50 mm, 5 µm, Phenomenex). Column 2: Atlantis T3 OBD column (1.0 × 50 mm, 5 µm, Waters). Column 3: Synergi Polar-RP, 80 Å column (1.0 × 50 mm, 4 µm, Phenomenex).

To execute this strategy, we constructed our metabolomics platform with three HPLC units and an autosampler equipped with three injectors, to which three columns are connected. The time scheme of whole injection cycle is shown in Figure 2. A customized injection program was developed to introduce a sample into three injectors and three different columns sequentially. Sample draw from a vial or a plate well is made at the beginning of the injection cycle, and then a fraction (1/3) of the total volume is dispensed into each injector and loaded onto the corresponding column at the defined time interval. After a fraction of the sample is dispensed into an injector, the gradient is started to elute analytes, the eluent of the corresponding column flow is now directed into mass spectrometer via a stream select valve for mass spectral data acquisition, and the selected ionization polarity is switched before column flow switching. Meanwhile, the other two columns are in washing or equilibrium mode and column flows are directed into waste collection bottle. Washing and equilibrium are key components in gaining reproducibility in chromatography, particularly in fast chromatography. The sample injection time, chromatography gradient start times, column flow switching times, and ionization polarity switching times are programmed and synchronized to acquire an integrated mass chromatogram trace for all monitored MRM channels from three columns for a given sample. This strategy works very well to analyze all classes of metabolites reported in this assay. Examples of a total ion current chromatogram (TIC) from a standard mixture (concentrations are 1-4 µg/mL for metabolites in the mixture) and representative extracted mass chromatograms (XICs) are shown in Figure S-1 (Supporting Information). Sensitivity and Linearity. In addition to analyzing diverse classes of metabolites, another big challenge metabolomics faces is the analysis of hundreds of analytes, which could have very different measurement sensitivities under a given assay condition and exist at wide range of concentration levels in a given biological matrix. To test out sensitivity and linearity of the method, we assessed limits of detection (LODs) and dynamic ranges from metabolite standard mixtures. Standard mixture solutions containing all 205 metabolites at 12 different concentration levels, prepared as described in experimental section, were analyzed. LOD and the linear range for each metabolite are reported in Table S-3 (Supporting Information). The sorted LOD for all metabolites is shown in Figure 3 on a logarithmic scale. 53.3% of the measured metabolites have LOD of 200 pg/mL or lower, 27.4% have LOD

Figure 3. Distribution of limits of detection of assayed metabolites (y axis sets on a log scale).

in 200 pg/mL to 2 ng/mL range, 12.7% have LOD in 2-20 ng/ mL range, and less than 6.6% of metabolites have LOD above 20 ng/mL. A 3-4 order of magnitude shift in the dynamic range is seen for 79% of the measured metabolites. In this study, we did not perform standard spiking experiments to provide detection limits and dynamic ranges in a given biological matrix because detection limit and dynamic range could be optimized by using different optimal metabolite extraction protocol, using more or less material, depending upon the interest of the metabolites. We think it is time saving and efficient to perform this type of experiment after initial profiling experiments complete and result in only a few biological meaningful metabolite hits. Detection Rate and Variation Assessment by Rat Plasma. Method variation is a key component in detecting real concentration changes in profiling experiments. In contrast to traditional quantitation methods which are developed and optimized for one or few specific molecules, our method simultaneously analyzes hundreds endogenous molecules, which present at a variety of concentration levels in one sample. To reflect real sample situations and assess the method reproducibility in a biological matrix, we analyzed pooled rat plasma, to provide assessment of method variation. Five plasma replicates, in 37.5 µL size, were generated from 1 mL of pooled rat plasma and processed as described in plasma metabolite extraction section, 10% of the processed rat plasma (3.75 µL) was loaded onto each column and analyzed by LC/MS/ MRM assay as described in experiment section. When a metabolite is detected in at least 4 of the 5 replicates, it is counted as detected in rat plasma. A total of 160 out of 205 metabolites were detected in the rat plasma; this represents a detection rate of 78%. The percent coefficient of variation (CV%) was calculated, from Analytical Chemistry, Vol. 82, No. 13, July 1, 2010

5531

Figure 4. Distribution of coefficients of variation of detected metabolites in rat plasma.

Figure 5. Correlation of CV% and signal intensity.

five replicates, for each detected metabolite and listed in Table S-4 (supplement). The distribution of the CVs is shown in Figure 4. Among these detected metabolites, 80% have a CV less than 15%. There are about 20% of detected metabolites having a CV higher than 15%. Given the excellent reproducibility observed for most detected metabolites, the extra variation can not be explained by the classic and well-known variation-contributing factors such as sample preparation, data acquisition, and data processing (peak integration). We carefully examined the extracted mass chromatographic peaks (XICs) of those metabolites having larger CV (>20%), we found most of them to have weak XICs (low intensity or low signal-to-noise ratio). We further checked the correlation between CV% and signal intensity measured by the integrated peak area. The plot of CV% versus log2 peak area, in Figure 5, clearly shows the correlation between higher measured CV and low peak intensity. The extra CV observed could then be explained by the low measurement sensitivity resulted from current assay conditions, or low concentration levels presenting in rat plasma. Modified assay condition or more starting plasma can be used to reduce corresponding variations when these metabolites are particularly interested in a given metabolite profiling experiment from rat plasma. Specificity, Limitation, and Potential Improvement of the Method. LC/MS/MRM has previously proven to be high specific for many applications in measuring drug metabolites and other small molecules. When it is applied in a profiling experiment, where hundreds of analytes are measured simultaneously in a fast fashion to allow high throughput, specificity could be compromised because of insufficient chromatographic separation and limited number of MRM transition channels per analyte. In this study, fast chromatography was used and provided reasonable chromatographic separation for many metabolites prior to mass spectrometric analysis but was not sufficient to separate all 5532

Analytical Chemistry, Vol. 82, No. 13, July 1, 2010

analytes. For those metabolites that were not separated by chromatography, specificity is solely provided by a different precursor ion and different product ion. In other words, mass spectrometric separation via one optimal MRM channel plays a key role for these metabolites in this assay. Though interference was taken into consideration when selecting a product ion, and each MRM channel was manually and carefully selected to be specific to the targeted metabolite as described in previous section (optimization of mass spectrometry), each of 205 metabolites was individually analyzed to test the uniqueness of the selected MRM channels under the same experimental conditions (data not shown). There was no significant interference observed with the exception of high concentration of glucose (1 mg/mL, to simulate human plasma concentration, and it is 30-100× higher than other test metabolites) contributing to the glyceraldehyde (M0113) channel at the same retention time. This indicates adequate specificities are achieved for these 205 assayed metabolites in this assay via reasonable chromatographic separation and carefully selected MRM channels. Given the fast chromatography was applied and a single MRM channel was used for most of measured metabolites in this assay, we speculated that there exists the possibility of inadvertently combining signals from other similar metabolites in a complex biological matrix. We therefore examined in-house data from another 157 metabolites analyzed individually under similar experimental conditions (data not shown). About 60% of this set was selected from KEGG and METLIN and initially acquired as potential interfering metabolites to the current assay, and the rest were acquired for their importance or relevance to the metabolic pathways well represented by the current 205 assayed metabolites. The observed interferences from these individual metabolite analyses, against 206 MRM channels built in this assay, are reported in the Table S-5 (Supporting Information) along the retention times of the targeted metabolites. Most of the interferences come from isomeric or isobaric metabolites which are highly hydrophilic and similar in chromatography behavior, and potentially have the same primary fragmentation pathway. Separating these endogenous isomers, at the same time while sorting out a broad range of other metabolites, is an inherent metabolomics challenge in both targeted and discovery approach. Multiple MRM channels per metabolite and higher resolving power ultra performance liquid chromatography (UPLC) could be used to improve the specificity of these metabolites. The most recent instrumentation advances, in enabling scheduled MRM (sMRM) data acquisition mode, and much higher spectral scan rate (e.g., 2 ms dwell time) in a triple quadrupole mass spectrometer, will need to be incorporated in future method developments to allow more MRM channels per analyte. It is important to point out that validation assays are essential for following up metabolite hits identified from this high throughput assay. A long gradient, different type of chromatography column, and more MRM channels could be developed and used in metabolite hits validation. In addition to experimental approaches, a literature search on the molecules which have the same precursor ion, similar structure, and similar polarity or chromatography behavior, is proven to be an effective way in aiding the validation and confirming metabolite identity.

It is worth noting that while we can easily and efficiently combine different chromatography approaches to extend coverage for different classes of metabolites, a variety of mass spectrometry ionization techniques, such as atmospheric chemical ionization (APCI), atmospheric photonic ionization (APPI), etc., are difficult to be taken advantage of. Future instrumentation advances in switching between these ionization methods on the fly will certainly be a welcome addition to our platform. CONCLUSION Enabled by the enormous advances in analytical instrumentation over past two decades and by taking full advantage of broad knowledge on naturally occurring small molecules, metabolite profiling has been applied in identifying gene functions,4-6,38,39 studying drug toxicity,12,13,40 drug discovery and development,41-43 and discovering clinically relevant biomarkers for variety diseases.14,36,37,44,45 A LC/MS/MRM method for target-based metabolomics will be a valuable addition to the arsenal of tools (38) Eraly, S. A.; Vallon, V.; Rieg, T.; Gangoiti, J. A.; Wikoff, W. R.; Siuzdak, G.; Barshop, B. A.; Nigam, S. K. Physiol. Genomics 2008, 33, 180–92. (39) Vallon, V.; Eraly, S. A.; Wikoff, W. R.; Rieg, T.; Kaler, G.; Truong, D. M.; Ahn, S. Y.; Mahapatra, N. R.; Mahata, S. K.; Gangoiti, J. A.; Wu, W.; Barshop, B. A.; Siuzdak, G.; Nigam, S. K. J. Am. Soc. Nephrol. 2008, 19, 1732–40. (40) Coen, M.; Holmes, E.; Lindon, J. C.; Nicholson, J. K. Chem. Res. Toxicol. 2008, 21, 9–27. (41) Wishart, D. S. Drugs R&D 2007, 8, 349–62. (42) Wishart, D. S. Drugs R&D 2008, 9, 307–22. (43) Kaddurah-Daouk, R; Kristal, B. S.; Weinshilboum, R. M. Annu. Rev. Pharmacol. Toxicol. 2008, 48, 653–83. (44) Bowser, R.; Cudkowicz, M.; Kaddurah-Daouk, R. Expert Rev. Mol. Diagn. 2006, 6, 387–98. (45) Pendyala, G.; Want, E. J.; Webb, W.; Siuzdak, G.; Fox, H. S. J. Neuroimmune Pharmacol. 2007, 2, 72–80.

for modern drug discovery, human disease pathology studies, and clinical patient stratification. We have presented a high-throughput, highly sensitive, and reproducible method for target based metabolomics. It combines three different chromatography conditions and two different ionization polarities into one method to achieve optimal separation and sensitivity across a large array of metabolites. The time gaps between column switching are used for column wash and re-equilibrium for best reproducibility and separation in a fast fashion. We have analyzed rat plasma samples; result shows good performance in detected metabolites and excellent reproducibility. The transition of such an assay to a clinical setting will require consistent sample collection protocols to control for assay variation because of differences in collection time, processing, and shipping. However, the results of the assays presented here will enable the application of metabolic profiling in in vitro and in vivo settings, as well as in well-controlled clinical settings. ACKNOWLEDGMENT The authors gratefully thank Jennifer L. Colangelo and Cynthia A. Drupa for kindly providing us rat plasma samples. SUPPORTING INFORMATION AVAILABLE Additional information as noted in the text. This material is available free of charge via the Internet at http://pubs.acs.org.

Received for review February 5, 2010. Accepted May 17, 2010. AC100331B

Analytical Chemistry, Vol. 82, No. 13, July 1, 2010

5533