Quantitative Profiling of Polar Cationic Metabolites in Human

Anal. Chem. 2009, 81, 1121–1129

Quantitative Profiling of Polar Cationic Metabolites in Human Cerebrospinal Fluid by Reversed-Phase Nanoliquid Chromatography/Mass Spectrometry Khin Than Myint, Ken Aoshima, Satoshi Tanaka, Tatsuji Nakamura, and Yoshiya Oda* Laboratory of Core Technology, Eisai Co., Ltd., Tsukuba, Ibaraki 300-2635, Japan, and Core Research for Evolutional Science and Technology, Japan Science and Technology, Saitama 332-0012, Japan Reversed-phase (RP) nanoliquid chromatography (LC)/ mass spectrometry (MS) is widely used for proteome analysis, but hydrophilic metabolites are poorly retained on RP columns. We describe here the development and application of an efficient, robust, and quantitative nanoLC/MS method for cationic metabolome analysis in the positive ionization mode without any derivatization of analytes. Various stationary phases for nano-LC, coating of the internal wall of the capillary column, and various mobile phases were evaluated in terms of separation and peak shapes for 33 hydrophilic metabolites, including nonderivatized amino acids. Polar cationic compounds were strongly bound to mixed-functional RP with cation exchange mode resin, and the best separation was obtained with hydrophilic internal wall coating and a twostep trifluoroacetic acid (TFA) gradient in methanol as the mobile phase. Simple, but optimized, sample processing and the use of a high content of methanol allowed robust nano-LC/MS analysis. Our developed method was applied for biomarker discovery in Alzheimer’s disease (AD). Several hundred peaks were detected from 10 µL of cerebrospinal fluid (CSF). In a principal component analysis (PCA) plot using peak intensities without normalization, peak separation depended on the experimental date, not disease state. Therefore, constant amounts of two stable isotope-labeled amino acids, Val and Lys, were added as internal standards (ISs) to each sample before processing. These ISs were eluted in different gradient slopes in the two-step gradient, and the normalized peak ratios using the corresponding ISs gave a unique group of PCA scores which could distinguish AD CSF samples from age-matched control CSF samples. Advanced analytical methods are required for quantitative metabolomics to find potential biomarkers that may exist in minute quantities in biological systems. Because of the high degree of chemical and physical diversity of metabolites, sensitive and highthroughput metabolomic analysis is a challenge, and many different analytical platforms have been developed for lipidomics,1,2 cationic metabolomics, and anionic metabolomics. Profiling of * To whom correspondence should be addressed. E-mail: [email protected]. Phone: +81-29-847-7084. Fax: +81-29-847-7614. (1) German, J. B.; Gillies, L. A.; Smilowitz, J. T.; Zivkovic, A. M.; Watkins, S. M. Curr. Opin. Lipidol. 2007, 18, 66–71. 10.1021/ac802259r CCC: $40.75  2009 American Chemical Society Published on Web 01/06/2009

Table 1. List of Standard Compounds and Their Limit of Detection (LOD) compound name

exact mass

LOD (pmol)

1,4-diaminobutane L-alanine L-proline cadaverine GABA L-serine creatinine L-norvaline valine L-homoserine L-threonine L-allothreonine cis-4-hydroxy-D-proline L-hydroxy-D-proline creatine L-norleucine L-asparagine L(+)ornithine glycylglycine urocanic acid L-glutamine L-lysine L-methionine 3,4-dihydroxy phenethylamine L-histidine L-phenylalanine L-arginine L-citrulline S-(carboxymethyl)-L-cysteine L-tyrosine dopa cystathionine S-(5′-adenosyl)-L-homocysteine

88.10 89.05 101.06 102.12 103.06 105.04 113.06 117.08 117.08 119.06 119.06 119.06 131.06 131.06 131.07 131.09 132.05 132.09 132.09 138.04 146.07 146.11 149.05 153.08 155.07 165.08 174.11 175.10 179.03 181.07 197.07 222.07 384.12

3.41 1.12 9.90 2.94 0.97 0.95 0.88 0.85 0.85 0.84 0.84 0.84 0.76 0.76 0.76 0.76 2.27 2.27 2.27 0.72 0.68 0.68 0.67 1.96 0.64 0.61 0.57 0.57 0.56 0.55 0.51 0.45 0.26

metabolites in biological fluids, such as cerebrospinal fluid (CSF),3,4 plasma,5 and urine,6,7 is useful in screening for disease (2) Roberts, L. D.; McCombie, G.; Titman, C. M.; Griffin, J. L. J. Chromatogr., B 2008, 871, 174–181. (3) David, W. S.; Michael, L. J.; Joshua, M. A.; Mitchel, F. D.; Kevin, J.; Xiong, Y.; Cheng, D.; Eisner, R.; Gautam, B.; Tzur, D.; Sawhney, S.; Bamforth, F.; Greiner, R.; Li, L. J. Chromatogr., B: Anal. Technol. Biomed. Life Sci. 2008, 871, 164–173. (4) Toczylowska, B.; Chalimoniuk, M.; Wodowska, M.; Mayzner-Zawadzka, E. Brain Res. 2006, 1104, 183–189. (5) Bogdanov, M.; Matson, W. R.; Wang, L.; Matson, T.; Saunders, P. R.; Bressman, S. S.; Beal, F. M. Brain 2008, 131, 389–396. (6) Kind, T.; Tolstikov, V.; Fiehn, O.; Weiss, R. H. Anal. Biochem. 2007, 363, 185–195. (7) Bjo ¨rkman, H. I.; Edlund, P. O.; Kvalheim, O. M.; Koistinen, I. S.; Jacobsson, S. P. Anal. Chem. 2003, 75, 4784–4792.

Analytical Chemistry, Vol. 81, No. 3, February 1, 2009

1121

Figure 1. Extracted nano-LC/MS chromatograms of polar compounds separated on Primesep 100 capillary column with noncoated internal wall (fused silica capillary) (a) and polar coated internal wall (TC-WAX capillary) (b). LC conditions are described in Table 2. Table 2. LC Conditions for Different Stationary Phases gradient name of stationary phase

particle size (µm)

pore diameter (Å)

mobile phase

time (min)

B (%)

YMC Pack Pro C18

5

120

A: 20 mM ammonium formate (pH 3.0)/ MeOH (8:2) B: 20 mM ammonium formate (pH 3.0)/MeOH (2: 8)

0 5 40 50 50.1

0 0 100 100 0

Synergi Fusion-RP

4

75

Obelisc N

5

100

Obelisc R

5

100

ZIC-HILIC

5

200

Primesep 100

5

100

Primesep A

5

100

0 5 20 20.1 50.0 50.1 60.0 60.1

0 0 20 40 70 100 100 0

A: B: A: B: A: B: A: B: A: B: A: B:

20 mM ammonium formate (pH 3.5)/ACN (1: 9) 20 mM ammonium formate (pH 3.5)/ACN (9: 1) 5 mM ammonium formate (pH 3.0)/ACN (2: 8) 100 mM ammonium formate (pH 3.0)/ACN (8: 2) 5 mM ammonium formate (pH 4.0)/ACN (8: 2) 100 mM ammonium formate (pH 4.0)/ACN (8: 2) 20 mM ammonium formate (pH 3.5)/ACN (1: 9) 20 mM ammonium formate (pH 3.5)/ACN (9: 1) MeOH/H2O/formic acid (50:50:0:1) MeOH/HO/formic acid/TFA (50:10:0.1:0.2) MeOH/H2O/formic acid (50:50:0:1) MeOH/HO/formic acid/TFA (90:10:0.1:2)

by comparing biomarkers with those of healthy controls. Nuclear magnetic resonance (NMR),8,9 gas chromatography/mass spectrometry (GC/MS),10,11 and LC/MS12,13 are commonly used for screening of metabolites in biological fluids. Although NMR has been accepted as a fast and information-rich technique, it has very (8) 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. (9) Gavaghan, C. L.; Holmes, E.; Lenz, E. M.; Wilson, I. D.; Nicholson, J. K. FEBS Lett. 2000, 484, 169–174. (10) Pasikanti, K. K.; Ho, P. C.; Chan, E. C. Y. Rapid Commun. Mass Spectrom. 2008, 22, 2984–92. (11) Fiehn, O. Trends Anal. Chem. 2008, 27, 261–269. (12) 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. (13) Wilson, I. D.; Plumb, R.; Granger, J.; Major, H.; Williams, R.; Lenz, E. M. J. Chromatogr., B 2005, 817, 67–76.

1122


low sensitivity, so that large sample amounts are required. GC/ MS offers high sensitivity and selectivity, but tedious pretreatment and derivatization are required to obtain volatile and thermally stable compounds. However, LC/MS can detect nonvolatile compounds after simple sample preparation. It is also important to down-scale LC to detect low-abundance analytes. Nano-LC/MS combined with reverse-phase (RP) chromatography is widely used as a highly sensitive and robust separation technique in proteomics.14,15 However, there are very few reports on the use of nano-RP columns for hydrophilic metabolome analysis.12 Highly polar metabolites found in biological samples are expected to be coeluted in the near-flow-through (14) Abian, J.; Oosterkamp, A. J.; Gelpı´, E. J. Mass Spectrom. 1999, 34, 244– 254. (15) Domon, B.; Aebersold, R. Science 2006, 312, 212–217.

amounts of proteins present in two different cell states in which a reference sample is labeled with a light isotope and the other sample with a heavy isotope. The reference and heavy-isotope labeled samples are then combined. Individual peptides labeled with light and heavy isotopes exhibit identical chemical properties (i.e., they behave identically throughout separation steps and in the ionization process in MS) but differ in mass by the difference in the weight of the isotopic label, so that the pairs of peaks can be easily observed in MS. The ratios of the peaks representing labeled and unlabeled species in the MS spectra then accurately reflect the relative quantities of the peptide in the samples from the two sources. In targeted metabolomics, on the other hand, the number of analytes is limited, so an internal standard approach is often used to normalize variations in both ionization suppression and sample processing.19 However, in untargeted metabolomics, the chemical structures show huge diversity, and metabolic labeling of mammalian samples, especially human biological fluids, is difficult. Thus, an efficient, robust and quantitative nano-LC/ MS method for polar metabolites would be an important advance in metabolomics to find biomarkers and understand disease mechanisms.

Figure 2. Extracted nano-LC/MS chromatograms at m/z 132 (cis4-hydroxy-D-proline, L-hydroxyproline, creatine, and leucine) and m/z 147 (glutamine and lysine) using YMC-packed pro C18, 5 µm (a), ZIC-HILIC, 3.5 µm (b), and Primesep A, 5 µm (c). LC conditions are described in Table 2.

fractions from RP columns. In addition, very-early eluted analytes from RP, such as highly polar metabolites, show very poor ionization efficiency in electrospray ionization (ESI)-MS due to very low organic solvent and high salt concentrations in the nearflow-through fractions. The matrix may suppress or enhance ionization of analytes, which would lead to a decreased or an increased MS response, respectively, and the level of the effect in the same matrix may vary depending on the amount/content of the matrix and the nature of the target. Unreliable analytical data can lead to an incorrect interpretation of molecular functions. For our purpose, it is necessary that polar metabolites should be sufficiently well retained on a nano-RP column to elute in a reasonable organic solvent concentration in the mobile phase and should be accurately quantitated. Label-free or standard-free LC/ MS based on ion intensities is therefore problematic, because the results are influenced by many variables. Relative expression levels of cellular proteins under different conditions (e.g., normal and disease states) have been measured by using stable isotope labeling strategies.16-18 A common feature of these techniques is that protein profiling can be performed by comparing the

EXPERIMENTAL SECTION Reagents and Standard Metabolites. Most of the reagents used in the experiments were of analytical grade and were purchased from Wako Pure Chemicals Co. (Osaka, Japan). Sequential grade trifluoroacetic acid (TFA) was obtained from Pierce (Woburn, MA). The standard metabolites (Table 1) were purchased from Sigma-Aldrich (Tokyo, Japan). 13C,15N-labeled L-lysine, and L-valine were purchased from Cambridge Isotope Laboratory (Andover, MA). Standard Sample Preparation. Standard stock solutions (1 mg/mL) were prepared in a mixture of methanol, water, and formic acid (50:50:0.1, v/v/v). To prepare working sample solutions for LC/MS, individual metabolite stock solutions were mixed to a final concentration at 30 ng/mL, and a 3 µL aliquot was injected for the evaluation of chromatographic separation. Detection limits were determined by injection of 3 µL of a known concentration of mixed standards. Cell Sample Preparation. HCT116 cells were grown in RPMI1640 medium, supplemented with 10% fetal bovine serum and 0.5% antibiotics. Before the cells were harvested, the plate containing 1 × 107 cells was rinsed with 5 mL of cold saline solution twice. Afterward, the samples were quenched with 3 mL of methanol and stored at -80 °C until extraction of the metabolites. Cell lysate (1 mL) was vigorously mixed with 0.7 mL of chloroform and 0.3 mL of water, and the mixture was centrifuged at 10 500g for 15 min. The aqueous upper fraction was transferred to a new vessel, evaporated under vacuum, and then reconstituted with 100 µL of 50% methanol for polar metabolome analysis. Cationic metabolites were extracted using Oasis MCX 96-well (16) Gygi, S. P.; Rist, B.; Gerber, S. A.; Turecek, F.; Gelb, M. H.; Aebersold, R. Nat. Biotechnol. 1999, 17, 994–999. (17) Oda, Y.; Huang, K.; Cross, F. R.; Cowburn, D.; Chait, B. T. Proc. Natl. Acad. Sci. U.S.A. 1999, 96, 6591–6596. (18) Ishihama, Y.; Sato, T.; Tabata, T.; Miyamoto, N.; Sagane, K.; Nagasu, T.; Oda, Y. Nat. Biotechnol. 2005, 23, 617–621. (19) Goto, T.; Myint, K. T.; Sato, K.; Wada, O.; Kakiyama, G.; Iida, T.; Hishinuma, T.; Mano, N.; Goto, J. J. Chromatogr., B 2007, 846, 69–77.


1123

Figure 3. Comparison of the total ion chromatograms of HCT116 cell lysate extracts of MCX without 1% formic acid washing (a) and with 1% formic acid washing (b).

plate cartridges (Waters, Tokyo) containing 10 mg of ion exchange sorbent. CSF Preparation. Cerebrospinal fluid samples of age-matched controls and subjects diagnosed with Alzheimer’s disease (AD) were purchased from PrecisionMed Inc., San Diego, CA. The deep-frozen CSF samples were thawed at room temperature and filtered with an Ultrafree-MC 5 000 NMWL centrifuge filter unit (Millipore Corporation, Bedford, MA) to remove proteins. Then 0.15 mL of protein-free CSF fraction was spiked with 30 µL of labeled internal standards (IS) solution (30 ng/mL of 13C,15Nlabeled L-lysine and L-valine in a mixture of methanol/water/ formic acid (50:50:0.1, v/v/v)). After addition of 700 µL of 0.1% formic acid to the sample, 8 µL of 5 M hydrochloric acid was added to adjust the pH to 1.0. The mixed solution was passed through an Oasis MCX 96-well plate cartridge preconditioned with 1 mL of methanol and 2 mL of deionized water. The cartridge was washed with 2 mL of water, 1 mL of 1% formic acid in 50% methanol, and 1 mL of methanol successively. Cationic metabolites in CSF were eluted with 0.8 mL of alkaline solution (25% ammonia solution/methanol (2:3)). The eluates were evaporated to dryness with a Turbo Vap evaporator at 35 °C and then reconstituted with 30 µL of 50% methanol. In order to remove solid particles, samples were filtered through a 0.2 µm PTFE filter unit before nano-LC/MS analysis. Liquid Chromatography/Mass Spectrometry. An Ultimate 3000 nano LC pump (Dionex Co., Sunnyvale, CA) was coupled with a LTQ-MS (Thermo Electron, San Jose, CA) and used in 1124


preliminary evaluation of capillary inner phases (Figure 1). The extracted CSF samples were analyzed using an Applied Biosystems QSTAR Pulsar i mass spectrometer equipped with a nanoflow Agilent 1100 solvent delivery system. Fused silica capillary and TC-WAX (150 µm i.d.) were purchased from GL Sciences Inc., Tokyo, Japan. Stationary-phase materials were a ZIC-HILIC column (SeQuant AB, Sweden), YMC Pack Pro C18 column (Europe GMBH), Obelisc N, Obelisc R, Primesep 100, and Primesep A from SIELC Technology (Prospect Heights, IL), and Synergi Fusion-RP material from Phenomenex. The mobile phases and gradient conditions are listed in Table 2. For the selection of the stationary phase for highly polar cationic metabolites, the materials were filled by nitrogen gas-pressure (at 7 MPa) into 0.15 mm internal diameter capillary tubes of approximately 15 cm length after the top of the fused silica capillary had been tapered with a laser capillary puller to a 0.008 mm opening. Mobile phases were prepared by mixing appropriate volumes of organic solvent and salt or acid solution. A polyethylene glycol-coated capillary column (TC-WAX, 150 µm i.d., ∼15 cm) filled with Primesep A (5 µm, 100 Å) was chosen for separation of both weak and strong cationic metabolites with a stepwise gradient. Mobile phase A was composed of methanol/ water/formic acid (50:50:0.1), and mobile phase B was composed of methanol/water/formic acid/trifluoroacetic acid (80:20:0.1:2.0, v/v/v). The gradient program was as follows: initial conditions were 100% A for 5 min, followed by a gradual increase to 20% B for 15 min, then a step to 40% for 0.1 min, and finally a gradual

Figure 4. Extracted ion chromatograms of metabolites in human cerebrospinal fluid (control). Experimental conditions are detailed in the Experimental Section.

increase to 70% for 30 min. Then, the column was washed with 100% B for 10 min. Before the next analysis, the column was conditioned with mobile phase A for 25 min. Data Analysis. The mass spectrometric data were acquired using Analyst version 1.1, and the initial metabolomic profiling was performed using in-house-developed Mass ++ data analysis software (http://groups.google.com/group/massplusplus) to obtain a peak list and align retention times. Principle component analysis (PCA) was performed using SIMCA-P+ version 11.0 (Umetrics AB, Umea, Sweden). Peak area subtraction between ADs and age matched controls was done from mass chromatograms obtained with Analyst QS 1.1 for peaks selected by multivariate analysis showing marked differences. In order to confirm that the same peak was selected from all samples, the product ions and the retention time of each peak were verified. RESULTS AND DISCUSSION Development of Nanoliquid Chromatography for Polar Cationic Metabolites. LC/MS for metabolomics typically uses RP columns with an inner diameter of more than 2.0 mm and a flow rate of several hundred microliters to a few milliliters per minute. However these RP columns are not suitable to retain and separate highly polar metabolites. Recently, several groups have used hydrophilic interaction chromatography (HILIC) columns,20-23 which involve a kind of normal-phase separation. Although highly (20) Rabinowitz, J. D.; Kimball, E. Anal. Chem. 2007, 79, 6167–6173. (21) Schlichtherle-Cerny, H.; Affolter, M.; Cerny, C. Anal. Chem. 2003, 75, 2349– 2354. (22) Tolstikov, V. V.; Fiehn, O. Anal. Biochem. 2002, 301, 298–307. (23) Alpert, A. J. J. Chromatogr. 1990, 499, 177–96.

polar metabolites are well retained on HILIC columns,24-27 the loading capacity is very much lower than that of an RP column and is not practical for real biological sample analysis to detect low-abundance metabolites by injection of large amounts of samples. The sensitivity of LC/MS should be improved as the flow rate at the ESI sprayer is decreased. It is feasible to separate nonderivatized highly polar metabolites (i.e., amino acids) by volatile ion pairing such as tridecafluoroheptanoic acid on RP C18 columns; however, it is not easy to get rid of lipophilic ion pairing reagents from LC/MS systems, when the systems are used for other purposes.28-30 Thus, we developed a new method for analysis of polar metabolites that utilizes nano-LC/MS in the positive ionization mode. First, we examined different types of coating for the internal wall of capillary columns and found that a hydrophilic coating, such as polyethylene glycol, sharpened the peaks (Figure 1) by reducing wall effects (analytes probably interact with the noncoated/hydrophobic internal wall). Thus, the hydrophilic coated TC-WAX capillary was selected for all experiments. For evaluation of stationary phases, we analyzed 33 commercially available hydrophilic standard metabolites listed in Table 1 with the various stationary phase materials presented in Table 2 and corresponding mobile phases. Generally, most of the standards are very polar and are poorly retained on RP columns, such as YMC Pack Pro C18 (Figure 2a). Several mixed functional RP columns were tested, such as a dual phase with a polar group embedded in C18 phase material (i.e., Synergi Fusion-RP). Theoretically, this enhances the retention of polar compounds and reduces the retention times, so that the total analytical time is shortened. Although the effects of several types of additives, pH, and organic solvents were examined, most of the polar standard metabolites were detected at the void volume. A mixed functional column with an ionic exchange mode might also be attractive for hydrophilic ionic metabolome analysis. A zwitterionic exchange mode in hydrophilic stationary phase materials, such as Obelisc-N, was expected to retain polar ionic analytes. After careful selection of the optimum pH of the mobile phase, Obelisc-N could retain only strongly cationic metabolites such as lysine, but not weakly cationic metabolites, using a mobile phase of pH 3. Alternatively, Obelisc-R has a zwitterionic exchange property in hydrophobic stationary phase materials. Unfortunately this column did not retain cationic polar compounds under any conditions examined. Currently, HILIC is widely used for the separation of polar compounds with micro-LC. One of the most widely used HILIC materials, ZIC-HILIC,31-33 which is zwitterionic with sulfobetaine groups covalently attached to the silica surface, was tested using a mobile phase containing a mixture of ammonium formate buffer (pH 3.5) and acetonitile. As shown in Figure 2b, we did not obtain reasonable peak shapes from nano-HILIC columns, even when the same amounts of standard samples as used with the other (24) Godejohann, M. J. Chromatogr., A 2007, 1156, 87–93. (25) Gika, H. G.; Theodoridis, G. A.; Wilson1, I. D. J. Sep. Sci. 2008, 31, 1598– 1608. (26) Hsieh, Y. J. Sep. Sci. 2008, 31, 1481–1491. (27) Kiefer, P.; Portais, J. C.; Vorholt, J. A. Anal. Biochem. 2008, 382, 94–100. (28) Armstrong, M.; Jonscher, K.; Reisdorph, N. A. Rapid Commun. Mass Spectrom. 2007, 21, 2717–2726. (29) Zoppa, M.; Gallo, L.; Zacchello, F.; Giordano, G. J. Chromatogr., B 2006, 831, 267–273. (30) Piraud, M.; Vianey-Saban, C.; Petritis, K.; Elfakir, C.; Steghens, J. P.; Bouchu, D. Rapid Commun. Mass Spectrom. 2005, 19, 1587–1602.


1125

Figure 5. PCA scores plots of peak intensity of before normalization to internal standard (a), after normalization to 13C5,15N-valine (b), after normalization to 13C6,15N2-lyine (c), and after normalization to 13C5,15N-valine and 13C6,15N2-lyine (d) obtained by nano-LC/MS analysis of agematched control and AD of all (female) CSF samples. Different experimental dates were represented by the triangles (4 for the control and 2 for AD) and the circles (O for the control and b for AD).

nanocolumns were injected. It is well-known that there are many limitations on HILIC columns in terms of sample amount, sample volume, and sample solvent, especially on nano-HILIC columns, and indeed, some of the peaks became symmetrical when a 10fold-diluted sample was injected onto the nano-ZIC-HILIC column. However, such a very low loading capacity of the nano HILIC column is not suitable for metabolome analysis. Another limitation of the nano-HILIC column is the sample solvent, since a high concentration of organic solvent is required. However, highly polar compounds are not well dissolved in small volumes of organicrich solvents. In general, ion exchange columns have a higher sample loading capacity than other types of columns, which is better in terms of detection of small amounts of materials, and polar ionic metabolites would be well retained on strong ion exchange columns. However, the peaks from ion exchange columns are (31) Kamleh, A.; Barrett, M. P.; Wildridge, D.; Burchmore, R. J. S.; Scheltema, R. A.; Watson, D. G. Rapid Commun. Mass Spectrom. 2008, 22, 1912– 1918. (32) Kamleh, M. A.; Hobanib, Y.; Dowb, J. A. T.; Watson, D. G. FEBS Lett. 2008, 582, 2916–2922. (33) Vikingsson, S.; Kronstrand, R.; Josefsson, M. J. Chromatogr., A 2008, 1187, 46–52.

1126


broadened and large amount of salts must be used in the mobile phase to elute bound compounds; this is not suitable for MS analysis. Therefore, we tested several mixed-functional RP with strong ionic exchange mode columns. Primesep 100 has a mixed mode of both RP and a strong cationic phase to separate basic polar compounds. The ionic groups on the solid phase have pKa of 1.0 and thus would hold weakly cationic compounds. Primesep 100 column retained weakly cationic compounds, such as creatine, leucine, and glutamine, but the peaks were broad. This peak broadening is due to the movement of analytes along the column during sample loading. Therefore, all analytes should be firmly retained at the top of a nano-LC column to minimize sample zone broadening. Primesep A stationary phase, which also has mixed RP with a strong cation exchange phase, has a stronger cationic character than Primesep 100. Therefore, it holds weaker ionic compounds more tightly at the top of the nanocolumn and, as a result, gave excellent peak shapes with good separation on elution with a TFA gradient in methanol (Figure 2c). The mobile phase also improved the stability of ESI due to the high concentration of organic modifier, methanol. Furthermore, Primesep A has a much greater loading capacity without tailing or leading than does HILIC. The

Figure 6. Reproducibility of quantitation before normalization (9) and after normalization to 13C5,15N-valine and 13C6,15N2-lysine (2) in six different analyses. The coefficients of variation (%) of each metabolite was plotted.

limit of detection (LOD) of this LC/MS setup with the full-scan mode was in the range of 0.3-9.9 pmol for 33 polar standard samples described in Table 1. Optimization of Sample Preparation. Sample cleanup is one of the keys to a robust analytical method. Biological samples contain a lot of matrixes such as proteins, DNA, and lipids as well as polar metabolites. There are many different types of real samples such as tissues, body fluids, and so on. Among them, it is relatively easy and low cost to obtain whole cell lysates from cell lines, and there is also no ethical issue to use cell line samples. Therefore we used HCT116 cells, which is a human colorectal carcinoma cell line, as the first step to optimize sample preparation. In this study, we focused on polar cationic metabolites, so hydrophobic, anionic, and neutral metabolites should be removed from samples. The content of hydrophobic materials, such as lipids, was reduced by organic solvent extraction in the first step, because lipids were distributed into the chloroform (lower) phase and polar metabolites into the upper phase. The amounts of anionic and neutral compounds could be reduced by cationic extraction cartridges such as Oasis MCX (strong cation exchange), Oasis WCX (weak cation exchange), and Bond elute SCX (strong cation exchange). After several attempts, we found that the Oasis MCX with 10% ammonia in methanol as the elution buffer provided better overall recoveries of highly polar metabolites, including from weak basic compounds to strong basic compounds such as polyamines, than the other two cartridges. When we analyzed cell lysate samples under the above conditions, the top of the nano-LC column always became clogged after 3-5 injections. We observed dirty black materials at the top of the nano-LC column through a microscope. During the washing step, acidic and neutral matrixes were supposed to have been removed from the final samples. Therefore, the black deposits might be materials such as phosphate-containing metabolites, which might interact with trace amounts of metals/silica gel in the solid phase

Table 3. Selected Peaks in PCA Score Plots in Figure 7 peak area peak area retention ratio increased retention ratio decreased time (min) m/z in AD (n ) 17) time (min) m/z in AD (n ) 17) 20.1 13.7 17.8 29.4 30.4 31.5 29.8 31.3 35.0 12.0 27.8 14.3 18.0 28.6 18.7

76.1 118.1 132.1 137.1 161.1 169.1 170.0 170.0 189.2 223.1 251.2 259.1 263.1 265.2 409.2

v v v v v v v v v v v v v v v

15.7 21.3 20.1 24.2 28.8 18.9 21.5 22.8 23.2 12.7 18.0 28.7 19.7 16.3 15.8 28.7 32.6 18.2 21.9 15.5 17.3 30.3 31.0 24.2 18.2 20.6 29.9 9.5 23.5 11.6 28.9 16.7 18.6 29.6 22.2 30.7 17.7 29.9

72.1 74.1 83.1 86.1 86.1 88.1 97.1 100.1 111.1 114.0 114.1 115.1 116.1 118.1 132.1 133.1 170.1 179.1 180.1 182.1 186.1 189.1 205.2 207.1 227.1 230.2 235.2 238.2 241.1 252.2 267.2 270.1 276.2 310.2 316.2 321.2 331.2 523.3

V V V V V V V V V V V V V V V V V V V V V V V V V V V V V V V V V V V V V V

extraction (SPE) cartridge and be burned by the high voltage at the top of the nano-LC column, where the internal wall coating Analytical Chemistry, Vol. 81, No. 3, February 1, 2009

1127

Figure 7. Total ion chromagrams of the male AD CSF sample (a) and the male age-matched control sample (b). Experimental conditions are detailed in the Experimental Section.

had been peeled off during the laser puller process. The use of 1% formic acid solution as the washing step to deionize lower pKa compounds, such as phosphates, and wash them away from the MCX cartridge was quite effective to prevent clogging. Figure 3 shows mass chromatograms obtained from the HCT116 cell lysate without (Figure 3a) and with (Figure 3b) washing with 1% formic acid, using an MCX cartridge. The background in the chromatogram was reduced in the case of the 1% formic acidwashed sample. Quantitation of Polar Metabolome by Nano-LC/MS Analysis. Reliability and reproducibility are also important parameters for a new analytical method for metabolomics, because the high chemical diversity of metabolites results in different recoveries during extraction and separation steps and different ionization responses in MS. Therefore, stable isotope-labeled internal standards (ISs) for each target compound have generally been used to normalize variations during sample processing and LC/ MS analysis, because such ISs have exactly the same physiochemical properties as their targets. Therefore, the recoveries and ionization efficiencies are the same, but the two can be separated by MS and their peak ratio measured. However this approach is not practical for metabolomics, because most metabolites are unknown and preparation of many ISs is expensive. From a practical point of view, the number of ISs should be minimized. The ionization efficiency in ESI is easily varied by changing the composition of the mobile phase, so we added constant amounts of two stable isotope-labeled amino acids, Val and Lys, which are 1128


eluted in different slopes of the two-step gradient, to each CSF sample at the beginning of sample processing. Analysis of CSF Samples. Biomarkers are very useful for diagnostics and monitoring of disease progress34 and are important for patient selection, avoidance of side-effects, aiding decisions about further medication of patients, and new drug discovery. In Alzheimer’s disease (AD), β amyloid 1-42 and phosphorylation of tau protein are currently considered the most useful biomarkers,35-37 though they cannot predict early stage AD, such as mild cognitive impairment, and are not useful for guiding drug treatment. A marker of AD progression is urgently required to assist early diagnosis and treatment of AD. Here, with the use of our optimized analytical procedure, several hundred peaks were detected from 10 µL of human CSF. Figure 4 shows the extracted chromatograms of some identified peaks. The overall principal component analysis (PCA) of peak intensities was performed after peak selection and retention time alignment of samples from 17 AD patients and 17 age-matched controls in order to discriminate AD patients from controls. (34) Ward, M. Expert Rev. Mol. Diagn. 2007, 7, 635–646. (35) Ibach, B.; Binder, H.; Dragon, M.; Poljansky, S.; Haen, E.; Schmitz, E.; Koch, H.; Putzhammer, A.; Kluenemann, H.; Wieland, W.; Hajak, G. Neurobiol. Aging 2006, 27, 1202–1211. (36) Kapakia, E. N.; Paraskevasa, G. P.; Tzerakisb, N. G.; Sfagosa, C.; Seretisb, A.; Kararizoua, E.; Vassilopoulosa, D. Eur. J. Neurol. 2007, 14, 168–173. (37) Bouwman, F. H.; Schoonenboom, N. S. M.; Verwey, N. A.; Elk, E. J.; Kok, A.; Blankenstein, M. A.; Scheltens, P.; van der Flier, W. M. Neurobiol. Aging, in press.

Figure 8. PCA scores plots of the peak area after normalization to 13C5,15N-valine and 13C6,15N2-lysine obtained by nano-LC/MS analysis of the age-matched control (4) and AD (2) of all CSF samples.

In a PCA plot of peak intensities (Figure 5a), the samples were clustered into two groups in independent experiments, but the patient group and control group were not differentiated. This indicates that the peak responses varied between different experimental dates, even though the sample preparation and instrumental parameters were the same. The PCA score plot using values normalized to a single IS, 13C,15N-lysine or valine, gave a unique group but with some outlying peaks (parts b and c of Figure 5). Thus, we normalized the peak intensities to the two ISs in different steps of the two-step gradient; peaks eluted at 0-20 min were normalized to 13C,15N-Val and peaks eluted at 20.1-60 min were normalized to 13C,15N-Lys. Interestingly, this procedure gave a compact group of PCA scores in the plot, but the patient group and control group were not separately clustered (Figure 5d). These results indicate that the developed analytical method with two ISs is reliable for quantitative metabolomics in biological samples, but it is difficult to identify putative biomarkers for AD. To obtain an uncorrelated feature between control and AD patients’ samples from several hundreds of peaks in CSF samples (Figure 6), we selected 53 peaks listed in Table 3, which appeared to have different peak intensities in normal and AD patients, by using multivariate analysis. Selected peaks are shown in PCA score plots in Figure 7. Each peak in all CSF samples was confirmed by retention time, m/z and MS/MS pattern to ensure exactly the same peaks were compared among all samples. As mentioned above, the experimental data for the selected peaks could not be distinguished by PCA analysis without ISs normalization. Normalization with either one of the ISs could separate controls from AD patients to some degree but without statistical significance. Finally, normalization using both ISs resulted in significant separation between AD and age-matched controls

(Figure 8). These results confirm the value of our refined nanoLC/MS method for efficient hydrophilic metabolome analysis. However, chemical structural determination remains an issue in untargeted metabolomics, and precise identification of these 55 candidate AD biomarkers among human CSF metabolites will require further study. CONCLUSIONS A sensitive nano-LC/MS method for identifying cationic polar metabolites in biological samples has been developed. Several cationic metabolites were evaluated as standards for LC/MS separation, and Primesep A stationary phase in a polyethylene glycol-coated capillary provided the best separation performance using a TFA and methanol two-step gradient. The simple sample preparation procedure can be applied in a high-throughput manner. The performance of the method is sufficient to detect subtle changes in metabolites in biological samples due to the use of two internal standards to normalize variation. We were able to distinguish AD CSF samples from age-matched control CSF samples with statistical significance by using a selected subset of peaks. This new cationic metabolome analysis method is expected to be useful for further metabolite profiling in biological samples and should also be effective for identifying novel biomarkers. ACKNOWLEDGMENT This work was supported by funds from Core Research for Evolutional Science and Technology (CREST). Received for review December 11, 2008.

October

27,

2008.

Accepted

AC802259R


1129

Quantitative Profiling of Polar Cationic Metabolites in Human

Recommend Documents