Quantitative Phosphorus Metabolomics Using Nanoflow Liquid

Apr 7, 2009 - method employing a stone-arch microcolumn with amino propyl silica gel achieved good separation of phosphorus metabolites with forty- to...
0 downloads 0 Views 339KB Size
Anal. Chem. 2009, 81, 3836–3842

Quantitative Phosphorus Metabolomics Using Nanoflow Liquid Chromatography-Tandem Mass Spectrometry and Culture-Derived Comprehensive Global Internal Standards Taisuke Uehara,†,‡ Akira Yokoi,† Ken Aoshima,†,‡ Satoshi Tanaka,‡ Tadashi Kadowaki,† Masayuki Tanaka,† and Yoshiya Oda*,†,‡ Eisai Co., Ltd., Tsukuba, Ibaraki 300-2635, Japan, and Core Research for Evolutional Science and Technology, Japan Science and Technology, Saitama 332-0012, Japan Highly sensitive and quantitative analytical methods are essential for metabolomics. In this report, we introduce an analytical method focused on endogenous phosphorus metabolites, using nanoflow liquid chromatographyelectrospray ionization tandem mass spectrometry (nanoLC-ESI-MS/MS) and culture-derived isotopetagged metabolites as global internal standards for quantitative metabolomics. The nanoLC-ESI-MS/MS method employing a stone-arch microcolumn with amino propyl silica gel achieved good separation of phosphorus metabolites with forty- to hundred-fold increase of sensitivity compared with semimicro flow LC-ESI-MS. The quantitative reproducibility of the nanoLC-ESI-MS has been improved to the point where it is useful for studies of cellular metabolism. Focused metabolomics using culture-derived internal standards was employed to monitor 184 phosphorus-related metabolic changes in cancer cells treated with metabolic enzyme inhibitors, methotrexate, fluorouracil, and gemcitabine. We found marked perturbations of cellular metabolism, of which many, though not all, were in line with the known biological activities of these drugs. Living organisms obtain chemical energy to maintain critical systems from diverse and complex physicochemical reactions, and also synthesize an enormous variety of molecules. Metabolomics, that is, the comprehensive analysis of endogenous metabolites in living cells, can reveal the physiological status of the cells at a given time point. Metabolomics is therefore expected to play an important role in bridging the gap between phenotype and genotype, and combining information from metabolomics, proteomics, transcriptomics, and genomics will help us to obtain a more complete picture of living organisms. However, quantitative metabolomics is practically difficult because of the diversity of structure and physicochemical properties of endogenous metabolites, which include amino acids, nucleosides, nucleotides, bases, vitamins, cofactors, lipids, carbohydrates, organic carboxylic acids, * To whom correspondence should be addressed. E-mail: [email protected]. Phone: +81-29-847-7084. Fax: +81-29-847-7614. † Eisai Co., Ltd. ‡ Core Research for Evolutional Science and Technology, Japan Science and Technology.

3836

Analytical Chemistry, Vol. 81, No. 10, May 15, 2009

and so on. Consequently, an appropriate method should be selected according to the nature of the analytes, or combinations of methods should be used to approach a comprehensive assay of metabolites.1 Recent developments in both nuclear magnetic resonance (NMR)2,3 and combined mass spectrometry technologies, such as gas chromatography-mass spectrometry (GC/MS),4,5 capillary electrophoresis-mass spectrometry (CE-MS),6,7 and liquid chromatography-mass spectrometry (LC/MS),8,9 have improved our ability to analyze the components of biological fluids. Although NMR is a fast and non-targeted technique, its sensitivity and resolution are lower than those of MS technologies. Highly sensitive and selective analysis can be achieved by using GC/MS, but tedious pretreatment and derivatization are required to obtain volatile and thermally stable compounds. Mass spectrometry in combination with capillary electrophoretic separation, CE-MS, offers excellent separation ability, but sample loading volumes are limited, and the technology is fragile. Liquid phase separation with LC/MS can robustly combine a variety of separation and detection methods, so it is a useful tool for the analysis of diverse endogenous metabolites. There are several thousands of different metabolites in mammalians, but current technologies only allow analysis of several hundred at a time and their sensitivity is too poor to detect low-abundance metabolites. We focus here on methodologies having sufficient sensitivity to detect low levels of endogenous metabolites. The sensitivity of ESI-MS is dependent (1) Van Der Greef, J.; Stroobant, P.; Van Der Heijden, R. Curr. Opin. Chem. Biol. 2004, 8, 559–565. (2) Brindle, J. T.; Antti, H.; Holmes, E.; Tranter, G. E.; Nicholson, J. K.; Bethell, H. W.; Clarke, S.; Schofield, P. M.; McKilligan, E.; Mosedale, D. E.; Grainger, D. J. Nat. Med. 2002, 8, 1439–1445. (3) Gavaghan, C. L.; Holmes, E.; Lenz, E. M.; Wilson, I. D.; Nicholson, J. K. FEBS Lett. 2000, 484, 169–174. (4) Fiehn, O.; Kopka, J.; Dormann, P.; Altmann, T.; Trethewey, R. N.; Willmitzer, L. Nat. Biotechnol. 2000, 18, 1157–1161. (5) Pasikanti, K. K.; Ho, P. C.; Chan, E. C. Y. Rapid Commun. Mass Spectrom. 2008, 22, 2984–92. (6) Soga, T.; Ueno, Y.; Ohashi, Y.; Naraoka, H.; Tomita, M.; Nishioka, T. J. Proteome Res. 2003, 2, 488–494. (7) Soga, T.; Ueno, Y.; Naraoka, H.; Matsuda, K.; Tomita, M.; Nishioka, T. Anal. Chem. 2002, 74, 6224–6229. (8) Ding, J.; Sorensen, C. M.; Zhang, Q.; Jiang, H.; Jaitly, N.; Livesay, E. A.; Shen, Y.; Smith, R. D.; Metz, T. O. Anal. Chem. 2007, 79, 6081–6093. (9) Wilson, I. D.; Plumb, R.; Granger, J.; Major, H.; Williams, R.; Lenz, E. M. J. Chromatogr. B 2005, 817, 67–76. 10.1021/ac9002062 CCC: $40.75  2009 American Chemical Society Published on Web 04/07/2009

on the concentration of analytes and is improved by column miniaturization, so nanoLC/MS has higher sensitivity10-12 and has been routinely applied in proteomics, though not metabolomics. Identification of metabolite structures from LC/MS data has not been automated because of the lack of a comprehensive metabolome database and appropriate search engines. Direct comparison of MS spectra and retention time with those of pure standard metabolites is currently used in metabolome identification, but quantitation of each metabolite is important, because metabolite concentrations in living cells vary under different conditions and at different times in the life cycle, so reliable analytical data are essential for proper interpretation of metabolic pathways. In quantitative metabolomics, the main sources of error are variations in the sample preparation procedures (e.g., metabolite extraction, derivatization, fractionation) and instability in MS analysis. These errors can be significantly reduced by using internal standards. An important characteristic of an internal standard is that its chemical-physical characteristics should be as similar as possible to those of the target analyte. For this reason, stable isotopes have been used for a few targeted analytes in quantitative combined MS analysis, and several stable isotope labeling approaches have been developed for quantitative proteomics. In in vivo labeling procedures for proteomics,13,14 internal standards are introduced early in the sample process, thus obviating the variations caused by differences in sample preparation and giving higher accuracy. In the study of metabolomics, Bajad et al. developed the differential analysis of Escherichia coli metabolism using bacteria cultured in a medium composed of 13Clabeled glucose.15 However, in vivo labeling of the metabolome in mammalian cells is difficult, because there are many sources of mammalian metabolites, including dietary components, and it is not practical to incorporate stable isotope into all of them. We have developed an approach based on the use of culturederived isotope tags (CDITs) for quantitative tissue proteome analysis.16 Stable isotope-enriched cultured cells were used as an internal standard to normalize tissue proteome levels. Here we apply CDIT to focused phosphorus metabolome analysis, mCDIT (Figure 1), to examine the modes of action of several metabolic enzyme inhibitors in cancer cells. MATERIALS AND METHODS Reagents and Standard Metabolites. HPLC-grade solvents, methotrexate, fluorouracil, gemcitabine, and the following standard metabolites; ribose 5-phosphate, pyridoxal phosphate, glucose 1-phosphate, fructose 1-phosphate, fructose 1,6-bisphosphate, phosphoenolpyruvate, phosphocreatine, AMP, CMP, dCMP, GMP, IMP, TMP, UMP, dUMP, 3′,5′-Cyclic AMP, 3′,5′-Cyclic GMP, ADP, dADP, CDP, GDP, IDP, dIDP, TDP, UDP, dUTP, TTP, UTP, ATP, dATP, GTP, ITP, dITP, UDP-glucose, UDP-glucuronate, UDP-N(10) Ishihama, Y. J. Chromatogr. A 2005, 1067, 73–83. (11) Wilm, M.; Mann, M. Int. J. Mass Spectrom. Ion Processes 1994, 136, 167– 80. (12) Wilm, M.; Mann, M. Anal. Chem. 1996, 68, 1–8. (13) Oda, Y.; Huang, K.; Cross, F. R.; Cowburn, D.; Chait, B. T. Proc. Natl. Acad. Sci. U.S.A. 1999, 96, 6591–6596. (14) Ong, S. E.; et al. Mol. Cell. Proteomics 2002, 1, 376–386. (15) Bajad, S. U.; Lu, W.; Kimball, E. H.; Yuan, J.; Peterson, C.; Rabinowitz, J. D. J. Chromatogr. A 2006, 1125, 76–88. (16) Ishihama, Y.; Sato, T.; Tabata, T.; Miyamoto, N.; Sagane, K.; Nagasu, T.; Oda, Y. Nat. Biotechnol. 2005, 23, 617–621.

Figure 1. Strategy of quantitative mammalian cell metabolomics using mCDIT. Quantitative cell metabolome analysis using stable isotope-labeled cultured E. coli as global standards. Mammalian cell extracts 1 and 2 are mixed with E. coli extracts, and the mixtures are analyzed by mass spectrometry to identify and quantify metabolites. The ratio between the two isotopic distributions (one from mammalian cells and one from E. coli labeled with isotopes) can then be determined from the mass chromatogram. Changes of metabolites level in two cell samples are estimated by calculating the ratio of the two ratios, ratio 1/ratio 2, a procedure which cancels out the internal standards (isotope-labeled E. coli metabolites).

acetyl-D-glucosamine, GDP-mannose, NAD+, NADH, NADP+, NADPH, riboflavin 5′-monophosphate, flavin adenine dinucleotide, CoA, acetyl CoA, malonyl-CoA, and succinyl-CoA, were purchased from Wako Pure Chemicals (Osaka, Japan). Ammonium carbonate, puratronic, was obtained from Alfa Aeser (Ward Hill, MA, U.S.A.). The BioExpress 1000 (13C,15N-labeled and nonlabeled) from Cambridge Isotope Laboratory (Andover, MA). Standard Sample Preparation. Standard stock solutions (1 mg/mL) were prepared in a mixture of methanol and water (50: 50 v/v). To prepare working sample solutions for LC/MS, individual metabolite stock solutions were diluted to suitable concentration, and a 2 µL aliquot was injected for the evaluation of chromatographic separation. Cell Sample Preparation. E. coli (ECOS-DH5a) single colony in an agar plate was isolated and grown in a shaker flask at 37 °C in 2 mL of BioExpress 1000 (unlabeled medium). After overnight incubation, 0.2 mL of cell culture was transferred to new flask with 5 mL of flesh medium and reincubated, and the absorbance of medium was measured at OD600 with a Thermo scientific NanoDrop ND-1000 spectrophotometer (Wyman Street Waltham, MA, U.S.A.) at the following time points; 0 h, 2 h, 3 h, 4 h, 6 h, 8 h, and 22 h. This growth curve data was shown in Supporting Information, Supplement 1. Metabolites of E. coli at 3 and 22 h were evaluated by LC-ESI-MS and were considered to represent the metabolic profiles of log phase and stationary phase cells, respectively. The extract of log phase bacteria grown in the same manner with a BioExpress 1000 (13C,15N-labeled) instead of unlabeled medium was used in the metabolome analysis as Analytical Chemistry, Vol. 81, No. 10, May 15, 2009

3837

Figure 2. MRM chromatograms of human HCT116, log phase E. coli, and stationary phase E. coli cell metabolites.

the stable isotope-labeled standards. Bacterial metabolites were extracted essentially as described in Lu et al. with a slight modification.17 Bacterial cells were collected by centrifugation for 3 min at 3000 rpm, and 2 mL of ice-cold 80% aqueous methanol was added to the cell pellet, which was then stored at -80 C until extraction of the metabolites. Cell lysate (0.7 mL) was mixed with 0.3 mL of water, and the mixture was sonicated in an ice bath for 10 min, then centrifuged in a Microcon 10 (Millipore, U.S.A.) at 13000 rpm for 40 min. A BioExpress kit allows highly efficient labeling of bacterial cells with stable isotopes. Indeed, the LC/MS peak ratio of nonlabeled/full-labeled metabolites obtained from log phase E. coli grown were 0.006 (fructose-1,6-bisphosphate), 0.011 (AMP), and 0.001 (NAD+). When labeling efficiency is unexpectedly low, our approach can still accept this sample as internal standard cells because we calculate ratio to ratio (sample A/internal standard versus sample B/internal standard), that is, the amounts of unlabeled metabolites derived from internal standard are canceled. However, unlabeled/partial labeled metabolites should be profiled before quantitation. HCT116 cells were grown to a density of 6 × 105 cells/well in RPMI-1640 medium (Sigma-Aldrich) supplemented with 10% fetal bovine serum and 0.5% antibiotics. Cell samples used for metabolome profiling were treated with methotrexate (MTX,), fluorouracil (5FU), or gemcitabine (GEM) for 4 h at a final concentration of 50 nM, 5 µM, or 350 nM, respectively. Before the cells were harvested, the cells were rinsed with 1 mL of ice-cold 150 mM ammonium acetate solution twice. Afterward, the samples were quenched with 0.7 mL of methanol and stored at -80 C until extraction of the metabolites. Cell lysate (0.7 mL) was vigorously mixed with 0.3 mL of water, and the mixture was sonicated in an ice bath for 10 min, then centrifuged in a Microcon 10 at 13000 rpm for 40 min. The filtrate was mixed with the bacterial extract, and the mixture was diluted with acetonitrile for nanoLC/MS analysis. (17) Lu, W.; Kimball, E.; Rabinowitz, J. D. J. Am. Soc. Mass Spectrom. 2006, 17, 37.

3838

Analytical Chemistry, Vol. 81, No. 10, May 15, 2009

Liquid Chromatography/Mass Spectrometric Analysis. An Applied Biosystems 4000 QTRAP mass spectrometer (Foster City, CA, U.S.A.) with a DIONEX UltiMate 3000 nanoLC pump (Sunnyvale, CA, U.S.A.) and a PAL autosample injector (CTC Analytics, Zwingen, Switzerland) were used as the nanoLC/MS system. For semimicro flow LC/MS analysis, a Shimadzu 10AD pump (Kyoto, Japan) was used instead of an UltiMate 3000. The Phenomenex LUNA amino column (Torrance, CA, U.S.A.) was selected as the stationary phase throughout this study. Linear gradients of B from 0-60% and 0-75% were run for nanoLC and semimicro-LC analysis, respectively, using mobile phase A of 95% acetonitrile in water and mobile phase B of 5% acetonitrile in 25 mM ammonium carbonate (pH 10.0) buffer for nano-LC/MS or mobile phase B of 5% acetonitrile in 15 mM ammonium carbonate (pH 10.0) buffer for semimicro LC/MS. LC/MS analysis using a “stone-arch” column was run according to Ishihama et al.18 Nitrogen gas pressure (at 7 MPa) was used to fill a 150 µm ID fused silica capillary (GL Sciences Inc., Tokyo, Japan) of approximately 15 cm length with the gel after the top of the capillary had been tapered with a laser capillary puller to a 0.008 mm opening. Detection of phosphorus metabolites in cells were carried out by negative ESI in multiple reaction monitoring (MRM) mode. MSMS parameters were automatically determined by using standard phosphorus metabolites according to the protocol supplied with the ABI 4000-Qtrap instrument. Other MSMS parameters for the phosphorus metabolome were predicted based on parameters of standard phosphorus metabolites and structural information registered in the Kyoto Encyclopedia of Genes and Genomes (KEGG).19 Data Processing. The mass spectrometric data were acquired using Analyst version 2.4, and the initial metabolomic profiling was performed using an in-house-developed Mass ++ data (18) Ishihama, Y.; Rappsilber, J.; Andersen, J. S.; Mann, M. J. Chromatogr. A 2002, 979, 233–239. (19) Kanehisa, M.; Goto, S.; Kawashima, S.; Nakaya, A. Nucleic Acids Res. 2002, 30, 42–46.

Figure 3. Validation of mCDIT strategy. (a) PLS-DA loadings S-plot comparing mammalian cell line HCT116 vs E. coli in log phase and stationary phase. (b) List of PLS-DA loading components lying in the upper right of the S-plot. Components depicted by color-coding line show the difference between HCT116 and E. coli samples. (c) List of PLS-DA loading components lying in the lower left of the S-plot. Components depicted by color-coding line show the difference between E. coli in the log phase and other samples. The identification of these components in the loading plot is summarized in the Supporting Information.

analysis software to obtain a peak list and align retention times (available at http://groups.google.com/group/massplusplus). Statistical analysis was performed using SIMCA-P+ version 11.0 (Umetrics AB, Umea, Sweden).

RESULTS AND DISCUSSION MS Detection of Phosphorus Metabolites. Multiple reaction monitoring (MRM) mode offers high selectivity, high sensitivity, and large dynamic range and is often used in quantitative Analytical Chemistry, Vol. 81, No. 10, May 15, 2009

3839

Table 1. CV Values of LC/MS Quantification of Cell Metabolites with or without mCDIT CV (%)a +

without mCDIT with mCDIT

P-valueb

NAD

UMP

UDP

UTP

NAD+

6% 3%

7% 9%

28% 11%

43% 8%

0.007 0.25

a CV (%) was obtained from data of 10 consecutive measurements. t test analysis of two data groups (n ) 3) obtained before and after continuous measurement overnight. The same samples were repeatedly analyzed. b

pharmacokinetic studies employing LC/MS analysis to monitor drug levels in body fluids. It is necessary to select product ions and to optimize collision energy for each target by testing corresponding pure standard compounds before real sample analysis. However, the number of endogenous metabolites is huge, and only a limited number of standard metabolites is available. This is one of the reasons why MRM analysis is not generally used for metabolomic studies. In spite of the structural diversity of endogenous phosphorus metabolites, there is a common feature of these compoundssthey give m/z 79 and m/z 97 fragment ions in the negative ESI-MS/ MS mode. Therefore, MRM can be applied to phosphorus metabolites based on limited experimental data combined with the extensive structural information registered in the KEGG database.19 The number of human metabolites is 1164 among the 2821 metabolites registered in the KEGG database, and 360 of them are phosphorus metabolites. After excluding inorganic compounds, RNAs, phospholipids, and long-alkyl-chain CoAs, the remaining number of human phosphorus metabolites becomes 240, and the number of MRM channels necessary to monitor those 240 phosphorus metabolites can be reduced to 162 because some precursors have the same values. Our MS apparatus currently allows simultaneous monitoring of 162 different channels (Supporting Information, Supplement 2). Development of nanoLC/MS. HPLC is one of the best established separation devices. Since the combination of MS with HPLC (LC/MS) offers robust, high-resolution separation, LC/ MS is probably the most suitable approach for comprehensive metabolome analysis. However, one of current bottlenecks is measurement of low-abundance cellular metabolites, so we developed an analytical tool based on nanoLC-ESI-MS technology to increase the sensitivity sufficiently to be able to detect endogenous polar metabolites. In this approach, the phosphorus metabolites are easily ionized in the negative ESI mode. However, highly polar phosphorus metabolites are not well retained in reverse phase columns, which are generally used in proteomics. In addition, HILIC mode LCs using silica gel, amide gel, diol gel, and ZIC-pHILIC gel with acidic mobile phase did not provide phosphorus metabolites good peak shapes. After screening of various HPLC conditions, we found that a ZIC-pHILIC and an amino-propyl silica gel column with basic ammonium carbonate buffer showed good retention and separation of polar phosphorus compounds because of both ionic and hydrophilic interaction between the compounds and the solid phase, and the amino-propyl silica gel showed stronger retention to weak acidic metabolites like cAMP and NAD+. Good MS chromatograms of phosphorus metabolites from the cell extracts were obtained by using the 3840

Analytical Chemistry, Vol. 81, No. 10, May 15, 2009

“stone-arch” microcolumn packed with amino-propyl gels (Supporting Information, Supplement 3). As we expected, the sensitivity of nanoLC/MS was forty to hundred fold better than that of semimicro LC/MS, and the limit of detection of four phosphorus metabolites, NAD+, UMP, UDP, and UTP were 0.78, 1.65, 2.36, and 7.10 fmole, respectively (Supporting Information, Supplement 4). The combination of nanoLC/MS and the predicted MRM monitoring described above yielded 184 LC/MS peaks from the HCT116 cell extract sample, and the peaks were observed on all MRM chromatograms. Among 184 LC/MS peaks, 33 peaks were confirmed by standard metabolites, and other 36 peaks were estimated by combination of precursor, product ions, and retention time. The remaining 115 peaks seemed to be phosphorus metabolites because of giving m/z 79 and m/z 97 fragment ions and reasonable retention as phosphorus compounds, but corresponding structures were not registered in the KEGG database and some of them might be in source fragment ions (Supporting Information, Supplement 2). Culture-Derived Stable Isotope-Tagged Standards (mCDIT) As Global Internal Standards. Although stable isotope-labeled peptides or proteins derived from cultured cell lines have been used as comprehensive internal standards for MS-based proteomics,13,14,16 labeling all cell metabolites in mammalian cells is difficult. It is well-known that many endogenous metabolites are common across species, for example, amino acids, nucleotides, carboxylic acids, cofactors, lipids, and so on. Phosphorylation of metabolites plays a key role in biological systems, and hence many common phosphoylated metabolites are utilized in diverse biological systems. Thus, bacterial cells grown in stable isotope-enriched medium are applicable as a biological source of comprehensive internal standards for mammalian cell metabolomics. Expression of E. coli metabolites at the log phase and stationary phase were compared with those of human-derived HCT116 cells (Figure 2). Interestingly, quite different profiles of metabolites were observed between the log phase and the stationary phase. Typically in stationary phase bacteria, the ratio of mono/triphospho-nucleotides increased drastically, and fructose-1,6-bisphosphate could not be observed. The most abundant cofactors of bacterial cells, NAD+/NADH and NADP+/NADPH, may indicate changes of redox state of bacteria between log phase and stationary phase. PLS-DA statistical analysis of the HCT116 and E. coli metabolome data uncovered two different factor groups (Figure 3a): one reflected the difference between human (HCT116) cells and bacterial cells (Figure 3b), and the other reflected the difference between proliferation phase cells (HCT116 cells and log phase bacterial cells) and stationary phase bacteria cells (Figure 3c). Both individual metabolome profiles and the statistical analysis showed that the log phase bacterial cells are a suitable source of internal standards to normalize quantitation of the mammalian metabolome. E. coli grown in 13C/15N stable isotope-enriched medium were collected at the log phase and mixed with HCT116 cells as global internal standard cells, and the peak ratio of each metabolite were measured using nanoLC/MS analysis. We repeated the analysis 10 times to evaluate the reproducibility (CV %) of our quantitative method (Table 1). Under our nanoLC/MS conditions, high concentrations of additive and high water content in the mobile phase to elute multiple phosphorus metabolites make ESI unstable, and nanoLC/MS is

Figure 4. Changes of cell nucleotide metabolism resulting from treatment with MTX, 5FU, and GEM. (a) Target pathways of anticancer agents. (b) Changes of cell nucleotide precursors/derivatives.

also more variable than semi-microLC/MS. Nevertheless, the internal standards for quantitation efficiently corrected the variation of peak areas of tri/diphosphorus nucleotides such as UDP/ UTP. The internal standard cells seemed to be less effective for other metabolites, such as UMP/NAD+, probably because we carefully set up the nanoLC/MS to optimize robustness. But when many samples have to be analyzed in a set of experiments, high reproducibility and stability are needed in all processes, including the nanoLC/MS system, to obtain reliable results. Thus, we repeatedly analyzed cell-extracted metabolites overnight, and the intensities of NAD+ in the first and last (each n ) 3) in the series of experiments without/with normalization by the internal standards are also given in Table 1. Absolute intensities (label-free quantitation) showed quite different values, but the normalized values using the internal standards succeeded

in decreasing the P value between the first and last measurements of the same samples. Thus, our sensitive and quantitative metabolome method appears to be applicable to real biological samples. Mode-of-Action Study of Anticancer Agents. HCT116 cells treated with anticancer drugs methotrexate (MTX),20,21 fluorouracil (5FU),22 and gemcitabine (GEM)23,24 for 4 h were quenched (20) Allegra, C. J.; Hoang, K.; Yeh, G. C.; Drake, J. C.; Baram, J. J. Biol. Chem. 1987, 262, 13520–13526. (21) Borsa, J.; Whitmore, G. F. Mol. Pharmacol. 1969, 5, 303–317. (22) Santi, D. V.; McHenry, C. S.; Sommer, H. Biochemistry 1974, 13, 471– 481. (23) Huang, P.; Chubb, S.; Hertel, L. W.; Grindey, G. B.; Plunkett, W. Cancer Res. 1991, 51, 6110–6117. (24) Heinemann, V.; Xu, Y.-Z.; Chubb, S.; Sen, A.; Hertel, L. W.; Grindey, G. B.; Plunkett, W. Mol. Pharmacol. 1990, 38, 567–572.

Analytical Chemistry, Vol. 81, No. 10, May 15, 2009

3841

and extracted in cold aqueous methanol. The cell extracts were mixed with internal standard cells obtained from bacteria grown in stable isotope-enriched medium. The nanoLC/MS analysis data of purine and pyrimidine nucleotides are depicted in Figure 4. Folic acid antagonist MTX inhibited purine de novo biosynthesis and thymidine biosynthesis,20,21 and we observed a decrease of dTDP and increases of 5′-phosphoribosylglycinamide (GAR), 5′phosphoribosyl-5-amino-4-imidazolecarboxamide (AICAR) and dUMP, which are substrates or products of the inhibited enzymes. But 5′-phosphoribosyl-4-(N-succinocarboxamide)-5-aminoimidazole (SAICAR) is a precursor of AICAR and a downstream product of GAR, and MTX inhibits two different enzymes in this pathway. Another pathway seems to cross this purine de novo biosynthesis pathway, otherwise the increase of SAICAR can not be not simply explained. Fluorouracil, a strong inhibitor of thymidylate synthetase,22 increased the level of dUMP and decreased dTDP, like methotrexate, but decreased intermediates of purine metabolism. Though gemcitabine is known to terminate DNA elongation by replacing deoxycytidine nucleotide with its phosphorylated metabolites,23 we also observed a decrease of dATP, which indicated ribonucleotide reductase inhibition24 by gemcitabine as another mode of action. Decreases of intermediates of purine metabolism were also observed after gemcitabine treatment. Although the changes of purine metabolism induced by fluorouracil and gemcitabine can not yet be interpreted in detail, and unknown pathways might also be inhibited by these compounds, the metabolome data clearly reflect the known biological activities of these anticancer drugs.

3842

Analytical Chemistry, Vol. 81, No. 10, May 15, 2009

CONCLUSIONS Combination of nanoLC/MS and hypothesis-driven MS/MS setting in the MRM mode allowed highly sensitive analysis, and the use of culture-derived stable isotope-labeled internal standards enabled comprehensive quantification of endogenous phosphorus metabolites. The sensitivity of the nano LC/MS/MS greatly exceeds that of conventional semimicro LC/MS by a factor of 40 to a 100-fold. The quantitative reproducibility of the analytical method has been improved sufficiently by using culture-derived internal standards to allow examination of changes in cellular metabolism. The application of the present nanoLC/MS method to examine phosphorus metabolome changes induced by anticancer drugs revealed marked changes of cellular metabolism, of which some, though not all, were in line with the known biological activities of these drugs. Thus, the approach described here is expected to be a powerful tool for mode-of-action studies in pharmacology, toxicology, and pathology. ACKNOWLEDGMENT This work was supported by a grant from Core Research for Evolution Science and Technology, Japan (CREST). T.U. thanks Dr. Naoki Asai for helpful support and fruitful discussions. 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 January 28, 2009. Accepted March 22, 2009. AC9002062