Fast and Accurate Quantitative Metabolic Profiling of Body Fluids by

Oct 12, 2012 - Nonlinear Sampling of 1H−13C Two-Dimensional Nuclear Magnetic ... Magnetic Resonance, SGPGIMS Campus, Raibarelly Road Lucknow, 226014...
0 downloads 6 Views 1MB Size
Article pubs.acs.org/ac

Fast and Accurate Quantitative Metabolic Profiling of Body Fluids by Nonlinear Sampling of 1H−13C Two-Dimensional Nuclear Magnetic Resonance Spectroscopy Ratan Kumar Rai and Neeraj Sinha* Centre of Biomedical Magnetic Resonance, SGPGIMS Campus, Raibarelly Road Lucknow, 226014 India S Supporting Information *

ABSTRACT: Two-dimensional (2D) nuclear magnetic resonance (NMR) methods have shown to be an excellent analytical tool for the identification and characterization of statistically relevant changes in low-abundance metabolites in body fluid. The advantage of 2D NMR in terms of minimized ambiguities in peak assignment, aided in metabolite identifications and comprehensive metabolic profiling comes with the cost of increased NMR data collection time; making it inconvenient choice for routine metabolic profiling. We present here a method for the reduction in NMR data collection time of 2D 1H−13C NMR spectroscopy for the purpose of quantitative metabolic profiling. Our method combines three techniques; which are nonlinear sampling (NLS), forward maximum (FM) entropy reconstruction, and J-compensated quantitative heteronuclear single quantum (HSQC) 1H−13C NMR spectra. We report here that approximately 22-fold reduction in 2D NMR data collection time for the body fluid samples can be achieved by this method, without any compromise in quantitative information recovery of various low abundance metabolites. The method has been demonstrated in standard mixture solution, native, and lyophilized human urine samples. Our proposed method has potential to make quantitative metabolic profiling by 2D NMR as a routine method for various metabonomic studies.

I

NMR for better metabolic profiling. By the application of 2D NMR spectroscopy along with multivariate statistical analysis, better metabolic profiling can be achieved even for the lowabundance metabolites. However, the quantification of various metabolites from 2D NMR spectroscopy is not straightforward. The cross peak intensity in 2D NMR is not directly proportional to concentration of metabolites. Cross peak intensity in 2D NMR are affected by several experimental factors such as relaxation times, mixing time, evolution time, and uneven excitation profiles, etc.14−17 Various studies have been reported in the literature to address these issues for the quantification by 2D heteronuclear single quantum coherence (HSQC)14,18−26 NMR spectrum. These methods are broadly classified on two categories, which are method based on generating calibration curves and compensation of experimental parameters by pulse sequence modifications. We have published a method based on calculation of the effect of experimental parameters on cross-peak intensity and measuring concentration by calculating correction factors.27 Another recent approach by Hu et al. uses a method based on recording HSQC experiment with different linear mixing time and extrapolating cross peak intensity to zero time for the measurements of metabolic concentration.22,23 All these approaches work well but require large experimental time for the NMR data collection, making it inappropriate for routine

dentification and quantification of various metabolites present in biological fluids, tissue extract, and complex mixtures is a challenging task. It has applications in many areas of chemistry, biology and medical research. Nuclear magnetic resonance (NMR) is one of the unique emerging analytical tools for providing wealth of metabolic information in a complex mixture without requiring any extensive sample preparation.1−6 One-dimensional (1D) 1H NMR is widely used technique for identification and quantification of various metabolites in biological fluids, tissue extract, and complex mixtures.7 Quantification of various metabolites from 1D 1H NMR spectrum can be achieved since NMR spectral intensities are directly proportional to concentration of metabolites. Effects of various experimental NMR parameters on 1D NMR spectral intensities are well understood and there are methods for extracting accurate quantitative information from such NMR spectrum.8 However, quantification from 1D 1H NMR suffers from spectral overlapping even at highest available magnetic field, which can affect accurate quantification.9 Twodimensional (2D) NMR spectroscopy can provide a better resolution in the chemical shift dimension. 2D NMR reduces signal overlap by spreading various resonance frequencies in two orthogonal dimensions. The advantage of 2D NMR spectroscopy has been utilized in various studies. In most of these studies, 2D NMR spectroscopy has been used for the purpose of identification of various metabolites.10−13 It has been demonstrated recently that metabolic profiling of urinary metabolites by 1D NMR spectroscopy will be biased toward high abundant metabolites,13 emphasizing importance of 2D © 2012 American Chemical Society

Received: August 25, 2012 Accepted: October 12, 2012 Published: October 12, 2012 10005

dx.doi.org/10.1021/ac302457s | Anal. Chem. 2012, 84, 10005−10011

Analytical Chemistry

Article

V0 C η0 = m × 0 Vm η Cm

metabonomic studies. The requirement of large sample size for getting statistically significant results is achievable only when NMR experiment time for each sample is as less as possible. There are methods for reducing multidimensional NMR data acquisition time. These methods are projection reconstruction,28,29 nonlinear sampling (NLS),30,31 single scan ultrafast methods,32 etc. These methods reduce 2D NMR data acquisition time to a large extent but have not been explored for quantitative metabolic profiling in body fluids. Recently, ultrafast 2D NMR methods have been explored for quantitative information recovery, and it is based on generating calibration curve.26 Further, this method may not be presently suitable for direct application to body fluid sample. Here, we present a quantitative 1H−13C HSQC method based on NLS with forward maximum entropy (FM) reconstruction.31,33−35 The modulation of cross peak intensity in the 2D HSQC experiment by heteronuclear J-coupling is compensated by the modulation of delays in (insensitive nuclei enhanced by polarization transfer) INEPT polarization transfer.19 The modulation of crosspeak intensity in HSQC due to relaxation parameters is calculated by earlier reported method.27 The method based on NLS with FM reconstruction is known to give linear correlation between linearly and nonlinearly sampled NMR cross-peak intensity with small offset. This offset may have an effect on accurate quantifications by NLS data.31 Other methods of quantifications by 2D HSQC have inherent limitations because of dependence on J-coupling dispersions19,27 and 13C offset.27 Here in our proposed method, the limitations of different methods are estimated for the quantification purpose. Our proposed method has resulted in substantial reduction in the NMR data collection time without compromising on the accuracy of quantitative information. The accuracy of the method is verified by standard solution of amino acid known concentration. We further explored the minimum time required for the 2D NMR data collection of human urine sample, which is an important body fluid used in various metabonomic studies.13,17,36−38 The accuracy for the measurement of the concentration of metabolites in human urine is verified by the spike-in experiments. We report here twenty two-fold reductions in NMR experimental time with NLS with FM-reconstruction in a single 2D 1H−13C HSQC experiment.

Here Cm and C0 are the molar concentration of metabolite and reference respectively and η=

(1 − E1)(1 − E2 cos ϕ) (1 − E1)(1 − E2 cos ϕ) − (E1)(E2 − cos ϕ)E2

where E1 = exp(−d/(T1)), E2 = exp(−d/(T2)), and ϕ = offset of the proton × d. Here d is the recycle delay used for recording the spectra. In above equation, the modulation due to heteronuclear J-coupling does not show up since we have used J-modulated INEPT. The details of the pulse sequence and method are given in Supporting Information. Sample Preparation. A standard solution of a mixture of amino acids was prepared for the verification of method. All amino acids were purchased from Merck Chemicals and SigmaAldrich with a minimum purity of 99% (AR grade). Standard amino acid solution was prepared using valine, alanine, glycine and methionine in D2O. For the NMR experiment we have used 350 μL of standard amino solution, 150 μL of D2O with 14.66 mM TSP, and 100 μL of [Cu(EDTA)] solution to shorten interscan delay. The final concentration of the solution was 24.5 (glycine), 40 (alanine), 52.42 (valine), and 77.89 mM (methionine), respectively. All standard amino acid solution was prepared in triplicates. A relaxation enhancing agent, [Cu(EDTA)] was used to a final concentration of 0.16 μM to shorten the inter scan delay.23,40 For human urine samples, 400 μL of native urine, 20 μL of D2O with 3.96 mM TSP, 40 μL of phosphate buffer (nondeuterated) of pH 7.4, and 100 μL of [Cu(EDTA)] solution to shorten interscan delay, was used for NMR experiment. All these samples were prepared in triplicates. One more sample was prepared by taking 3 mL of a different native urine sample, and then lyophilized and dissolved in 200 μL of D2O with 3.96 mM TSP, 250 μL of phosphate buffer (nondeuterated) of pH 7.4, and 100 μL of [Cu(EDTA)] solution to shorten interscan delay. The lyophilization of human urine sample was carried out by the freeze-drying method. This was done to ensure recovery of low boiling point metabolites. It has been reported previously that the 1H NMR spectrum of human urine does not show significant changes due to lyophilization.41 All NMR experiments were carried out at room temperature, that is, at 298 K. NMR Spectroscopy. All NMR spectra were recorded on Bruker 800-MHz NMR spectrometer equipped with a tripleresonance TCI (1H, 13C, 15N, and 2H lock) cryogenic probe. A series of 2D spectra were recorded with different recycle delay for the standard amino acid solution. For all 2D 1H/13C phase sensitive (echo/antiecho) spectra, with 2048 × 128 data points were collected using 4 scans per increment along with 16 dummy scans having acquisition time of 79 ms, with spectral widths of 16 and 165 ppm in 1H and 13C dimensions, respectively. For the NLS experiments, NLS sampling schedule was generated using Poisson gap sampling schedule generator as proposed by Hyberts et.al.34 We selected 42 and 32 t1 points out of 128 t1 point using above approach.34 For the lyophilized urine sample, NMR spectra were collected using 2048 × 256 points with 4 numbers of scans. 84 t1 were collected using Poisson gap sampling schedule with relaxation delay of 13 s (five times the longest T1 of interest) and 2 s. For the native urine sample, NMR spectra were



MATERIALS AND METHODS Details of Method. 2D HSQC spectrum of body fluid with sodium salt of trimethylsilyl propionic acid-d4 (TSP, internal standard) was recorded with NLS in the indirect dimension. Poisson Gap sampling schedule was used for recording NMR data in indirect dimension.34 The relaxation time used for the NMR experiment was as low as possible. Acquired NMR data was converted to NMRPipe format.39 FM reconstructed 2D spectra was generated. Cross peak intensity for various peaks were calculated from generated data along with intensity of TSP signal. The concentration of various peaks was calculated based on relaxation parameters and relaxation delays by the following relationship. By measuring ratio of cross peak intensity from the 2D NMR spectra between standard peak of known concentration V0, and test metabolite Vm, along with calculation of various correction factors, we can estimate the molar concentration of the metabolite, 10006

dx.doi.org/10.1021/ac302457s | Anal. Chem. 2012, 84, 10005−10011

Analytical Chemistry

Article

Figure 1. Linear regression curve obtained between gravimetric concentrations and concentration of various metabolites (glycine 1H = 3.57 ppm, 13 C = 44.37 ppm; alanine 1H = 1.48 ppm, 13C = 19.00 ppm, and 1H = 3.79 ppm, 13C = 53.47 ppm; valine 1H = 0.99 ppm, 13C = 19.5 ppm; 1H = 1.01 ppm 13C = 20.77 ppm; 1H = 3.61 ppm, 13C = 63.23 ppm; methionine 1H = 2.65 ppm, 13C = 31.7 ppm; 1H = 3.87 ppm, 13C = 57.72 ppm) measured by 2D 1H−13C HSQC. For multiple peaks, concentration was calculated for each peak before taking average value. The best possible fit equation and R2 values are shown in the figure. The concentrations of the four metabolites: alanine, methionine, glycine, and valine were considered. Total experimental time were (a) 7 h, (b) 2 h 24 min, (c) 1 h 15 min, and (d) 24 min. Various experimental details along with experimental time for recording the NMR data are shown in the figure.

collected using 2048 × 256 points with 12 numbers of scans. 84 t1 were collected using Poisson gap sampling schedule with relaxation delay of 2 s. For the estimation of the longitudinal relaxation parameter T1 of proton resonances, inversion recovery experiments were performed with presaturation of the water signal. The transverse relaxation rate (T2) of the proton resonances was determined by the Carr−Purcell− Meiboom−Gill (CPMG) pulse sequence with a variable echo time using presaturation during relaxation delay. The relaxation parameters T1 and T2 were measured for all standard mixture solutions and urine samples. Values of relaxation parameters are given in the Supporting Information (Table S1). To test the accuracy of the 2D method for lyophilized human urine samples, spike-in experiments were performed with the addition of different concentrations (300−700 μmol) of alanine, glycine, and hippurate. Verification in native urine sample, spike-in experiments were performed with the addition of different concentrations (400−700 μmol) of alanine and glycine. A correlation analysis was carried out between the recovered and added concentration values. CARA has been utilized for the peak volume calculation.42 For CARA details visit http://cara.nmr-software.org/portal/.

Data Processing and FM-Reconstruction. We used FMreconstruction program for construction of 2D HSQC data.31 We have used NMRPipe format for data handling.35



RESULT AND DISCUSSIONS Verification of Method. The quantification method was verified with a triplicate solution containing a mixture of amino acids such glycine, alanine, valine, and methionine. The final concentration of the various amino acids in the solution was 24.5 (glycine), 40 (alanine), 52.42 (valine), and 77.89 mM (methionine), respectively. The typical 2D HSQC spectrum of a test sample is shown in Supporting Information Figure 2a. The 2D NMR spectra recoded with NLS and FM reconstruction having different number of t1 points and relaxation delays are given in Supporting Information (Supporting Information Figure 2). The quantification method was verified in three steps in standard solution of amino acid. First 1H−13C HSQC spectra were recorded with large relaxation delay (12 s, approximately 5 times of the largest T1 value). In the second step we have used FM-reconstruction with Poisson gap sampling schedule in t1 dimension and relaxation delay was large. And finally we recorded spectrum 10007

dx.doi.org/10.1021/ac302457s | Anal. Chem. 2012, 84, 10005−10011

Analytical Chemistry

Article

Figure 2. Two-dimensional 1H−13C HSQC spectrum of native human urine sample recorded with NLS and FM reconstruction with relaxation delay 2 s. (b) Cross sections at different 1H chemical shift. (c) Results of spike-in experiments performed in the native human urine sample. Regression curves for various peaks, the best-fit straight-line equation and R2 values are shown in the plot.

human urine sample is a good system to verify the utility of this method in which dynamic range is too high. This system can be ideal to check if the metabolites resonances having high as well as low concentration are constructed faithfully for quantification purpose. The 2D 1H−13C HSQC spectra of human urine samples are shown in Figure 2. Various resonances can be unambiguously assigned on the basis of 1H and 13C chemical shifts. The number of resolved cross peak in 2D 1H−13C HSQC spectra is far better than 1D 1H spectra. Numerous metabolites such as hippurate, glycine and alanine cross peaks can be identified easily and their concentrations can be measured from 2D spectrum. For the verification of method, we used spike-in experiments for adding small amount of alanine (1H = 1.48 ppm, 13C = 19.33 ppm), glycine (1H = 3.56 ppm, 13C = 44.4 ppm), and hippurate (1H = 3.94 ppm, 13C = 46.50 ppm). The corresponding plots between added and recovered concentration are shown in Figure 2c. This shows almost linear variation. In routine NMR-based metabonomic studies, native, as well as lyophilized, human urine43−49samples are used. It has been shown earlier that lyophilization41 of human urine sample does not affect the quantification of various metabolites. It has advantages of reduced NMR experimental time because of high concentration of the metabolites. We have performed spike-in experiments in both kind of human urine sample. Native urine samples were prepared in triplicates to validate analytical procedure and to show that NLS sampling along with FM reconstruction can be verified for quantification purpose in both cases. Results of the spike-in experiments in native urine along with 2D data collected with uniform sampling and with NLS sampling along with FM reconstruction with standard deviation are presented in Figure 2. It can be seen that good correlation is observed even for the spectra recorded with NLS and FM reconstruction. Again we performed the spike-in experiment in lyophilized urine sample samples and results are shown in Figure 3. For the lyophilized urine sample, we performed the experiment at relaxation delay at 13 and 2 s with

with different relaxation delay and with different sampling schedule. We first verified the quantification of various metabolites from the spectrum recorded in completely relaxed state. The result is shown in Figure 1a, which shows the correlation between measured and gravimetric values of concentration of different metabolites. The corresponding resonances of different metabolites are given in figure caption. Regression analysis between the gravimetric value and measured from 2D 1H−13C HSQC show good correlation with R2 value of 0.99. Then we used FM-reconstruction with total relaxation delay (12 s) (42 t1 point out of 128) and again perform the regression analysis between the gravimetric concentration and measured concentration from 2D 1H−13C HSQC spectra. This shows good correlation with R2 value 0.99 (Figure 1b). This further confirms that quantification from 2D spectra recorded with NLS and FM reconstruction is similar to the spectra recorded with uniform sampling. Also, in this case, the NMR experimental time reduces 3-fold without any change in the concentration measurement. We then tried to find least possible relaxation delay that can give good measurement of concentration. For this we used relaxation delay of 2 s for recording NMR spectrum and reconstructed the data (Supporting Information Tables S8 and S9). The quantitative analysis was performed on 2D data acquired with reduced relaxation delay and NLS with FM reconstruction. The resulting NMR spectra were verified for quantitative information recovery and shows good correlation with gravimetric value, Figure 1d (Supporting Information Tables S8 and S9). With NLS schedule (32 t1 points out of 128) and recycle delay 2 s we find that NMR experimental time reduces dramatically (about 18-fold) and R2 value 0.99 was observed between experimental concentration measurements and gravimetric values (Figure 1d). Application to a Human Urine Sample. Verification of the Method by Spike-In Experiments. We verified the method on metabonomic sample of human urine in which we have large number of metabolites having varied concentration range. The 10008

dx.doi.org/10.1021/ac302457s | Anal. Chem. 2012, 84, 10005−10011

Analytical Chemistry

Article

Figure 3. Two-dimensional 1H−13C HSQC spectrum of lyophilized human urine sample recorded with (a) relaxation delay 13 s and (b) cross sections at different 1H chemical shift (c) Two-dimensional 1H−13C HSQC spectrum of human urine sample recorded with NLS and FM reconstruction with relaxation delay 2 s.(d) cross sections at different 1H chemical shift. Results of spike-in experiments performed in the lyophilized human urine sample. Regression curves for various peaks by (e) 1H−13C HSQC spectroscopy recorded with relaxation delay of 13 s (f) 1H−13C HSQC spectroscopy recorded with relaxation delay 2 s and NLS with FM reconstruction. The best-fit straight-line equation and R2 values are shown in the plot.

NLS (84 t1 point) with FM reconstruction. It can be seen that good correlation is observed even for the NMR spectra recorded with NLS and FM reconstruction (Figure 3). However, the experimental time of the NMR spectra recorded with NLS is very less compared to the spectra recorded with uniform sampling. We found approximately 22-fold reduction in the NMR experimental time can be achieved. The NMR recording data time for the native urine sample is higher than the lyophilized sample. This is expected because of the difference in the concentration of various metabolites in two samples. However, the reduction in NMR data collection time with NLS and FM reconstruction is similar. Sensitivity and Limitation of the Method. Sensitivity of the method will depend upon how faithful is our 2D NMR data reconstruction in indirect dimension. This will depend upon number of t1 points in indirect dimension. We have found that 1/3 reduction in the t1 (from t1 of 256) values is good enough for the NMR data reconstruction in human urine samples.

Further reduction in the t1 values for NLS does not faithfully reconstruct 2D NMR data. The lower limit of detection (LOD) of this method will depend upon several experimental parameters, which govern cross-peak volumes observed in the HSQC spectrum. In the present example, we could measure approximately few tens of micromolar concentration of metabolites in human urine samples. This lower limit will further reduce depending upon sensitivity and maximum t1 points.



CONCLUSIONS We have shown that present method of measuring concentration of various metabolites with 2D 1H−13C HSQC NMR spectra with NLS and FM reconstruction is efficient for human urine sample. The time reduction in the case of human urine sample is 22-fold and does not affect the concentration measurement. The efficiency of this method can be improved if we incorporate “fast maximum likelihood reconstruction” 10009

dx.doi.org/10.1021/ac302457s | Anal. Chem. 2012, 84, 10005−10011

Analytical Chemistry

Article

(FMLR) for the peak intensity calculation.22 Use of two internal standards for the higher and lower abundance metabolite separately will further improve quantification results. The method has been demonstrated for native, as well as lyophilized human urine samples, which has applicability in various metabonomic studies. The method is also general in nature and can be used for other mixture analysis.



(19) Koskela, H.; Heikkilä, O.; Kilpeläinen, I.; Heikkinen, S. J. Magn. Reson. 2010, 202, 24−33. (20) Heikkinen, S.; Toikka, M. M.; Karhunen, P. T.; Kilpelainen, I. A. J. Am. Chem. Soc. 2003, 125, 4362−4367. (21) H. Koskela, T. V. Magn. Reson. Chem. 2002, 40, 705−715. (22) Hu, K.; Ellinger, J. J.; Chylla, R. A.; Markley, J. L. Anal. Chem. 2011, 83, 9352−9360. (23) Hu, K.; Westler, W. M.; Markley, J. L. J. Am. Chem. Soc. 2011, 133, 1662−1665. (24) Bingol, K.; Brüschweiler, R. Anal. Chem. 2011, 83, 7412−7417. (25) Gronwald, W.; Klein, M. S.; Kaspar, H.; Fagerer, S. R.; Nrnberger, N.; Dettmer, K.; Bertsch, T.; Oefner, P. J. Anal. Chem. 2008, 80, 9288−9297. (26) Giraudeau, P.; Remaud, G. S.; Akoka, S. Anal. Chem. 2009, 81, 479−484. (27) Rai, R. K.; Tripathi, P.; Sinha, N. Anal. Chem. 2009, 81, 10232− 10238. (28) Kupce, E.; Freeman, R. J. Am. Chem. Soc. 2004, 126, 6429− 6440. (29) Kupce, E.; Freeman, R. J. Am. Chem. Soc. 2003, 125, 13958− 13959. (30) Jaravine, V. A.; Zhuravleva, A. V.; Permi, P.; Ibraghimov, I.; Orekhov, V. Y. J. Am. Chem. Soc. 2008, 130, 3927−3936. (31) Hyberts, S. G.; Heffron, G. J.; Tarragona, N. G.; Solanky, K.; Edmonds, K. A.; Luithardt, H.; Fejzo, J.; Chorev, M.; Aktas, H.; Colson, K.; Falchuk, K. H.; Halperin, J. A.; Wagner, G. J. Am. Chem. Soc. 2007, 129, 5108−5116. (32) Frydman, L.; Scherf, T.; Lupulescu, A. Proc. Natl. Acad. Sci. U. S. A. 2002, 99, 15858−15862. (33) Hyberts, S.; Arthanari, H.; Wagner, G.; Billeter, M., Orekhov, V., Eds.; Topics in Current Chemistry, Springer: Berlin, 2012; Vol. 316, p 125−148. (34) Hyberts, S. G.; Takeuchi, K.; Wagner, G. J. Am. Chem. Soc. 2010, 132, 2145−2147. (35) Hyberts, S.; Milbradt, A.; Wagner, A.; Arthanari, H.; Wagner, G. J. Biomol. NMR 2012, 52 (4), 315−327. (36) Holmes, E.; Foxall, P. J. D.; Spraul, M.; Farrant, R. D.; Nicholson, J. K.; Lindon, J. C. J. Pharmaceut. Biomed. Anal. 1997, 15, 1647−1659. (37) Keun, H. C.; Beckonert, O.; Griffin, J. L.; Richter, C.; Moskau, D.; Lindon, J. C.; Nicholson, J. K. Anal. Chem. 2002, 74, 4588−4593. (38) Shanaiah, N.; Desilva, M. A.; Nagana Gowda, G. A.; Raftery, M. A.; Hainline, B. E.; Raftery, D. Proc. Natl. Acad. Sci. U. S. A. 2007, 104, 11540−11544. (39) Delaglio, F.; Grzesiek, S.; Vuister, G. W.; Zhu, G.; Pfeifer, J.; Bax, A. J. Biomol. NMR 1995, 6, 277−293. (40) Wickramasinghe, N. P.; Kotecha, M.; Samoson, A.; Past, J.; Ishii, Y. J. Magn. Reson. 2007, 184, 350−356. (41) Lenz, E. M.; Bright, J.; Wilson, I. D.; Hughes, A.; Morrisson, J.; Lindberg, H.; Lockton, A. J. Pharm. Biomed. Anal. 2004, 36, 841−849. (42) Keller, R. L. J. Thèse de doctorat, Thesis No. 15947, ETH Zurich, Switzerland, 2004. (43) Ekman, D. R.; Teng, Q.; Jensen, K. M.; Martinovic, D.; Villeneuve, D. L.; Ankley, G. T.; Collette, T. W. Aquat. Toxicol. 2007, 85, 104−112. (44) Schoonen, W. G. E. J.; Kloks, C. P. A. M.; Ploemen, J.-P. H. T. M.; Smit, M. J.; Zandberg, P.; Horbach, G. J.; Mellema, J.-R.; Thijssenvan Zuylen, C.; Tas, A. C.; van Nesselrooij, J. H. J.; Vogels, J. T. W. E. Toxicol. Sci. 2007, 98, 286−297. (45) Schoonen, W. G. E. J.; Kloks, C. P. A. M.; Ploemen, J.-P. H. T. M.; Horbach, G. J.; Smit, M. J.; Zandberg, P.; Mellema, J.-R.; Zuylen, C. T.-v.; Tas, A. C.; van Nesselrooij, J. H. J.; Vogels, J. T. W. E. Toxicol. Sci. 2007, 98, 271−285. (46) Sewell, A. C.; Murphy, H. C.; Iles, R. A. Clin. Chem. 2002, 48, 357−359. (47) Lamers, R. J.; DeGroot, J.; Spies-Faber, E. J.; Jellema, R. H.; Kraus, V. B.; Verzijl, N.; TeKoppele, J. M.; Spijksma, G. K.; Vogels, J. T.; van der Greef, J.; van Nesselrooij, J. H. J. Nutr. 2003, 133, 1776− 1780.

ASSOCIATED CONTENT

S Supporting Information *

Additional material as described in the text. This material is available free of charge via the Internet at http://pubs.acs.org.



AUTHOR INFORMATION

Corresponding Author

*E-mail: [email protected], [email protected]. Fax: +91-522-2668215. Notes

The authors declare no competing financial interest.



ACKNOWLEDGMENTS Financial support from Department of Biotechnology, INDIA (Grant number BT/PR12700/BRB/10/719/2009) is gratefully acknowledged. R.K.R. acknowledges financial support from CSIR, INDIA.



REFERENCES

(1) Nicholson, J. K.; Wilson, I. D. Nat. Rev. Drug Discovery 2003, 2, 668−676. (2) David S, W. TrAC, Trends Anal. Chem. 2008, 27, 228−237. (3) van der Greef, J.; Stroobant, P.; van der Heijden, R. Curr. Opin. Chem. Biol. 2004, 8, 559−565. (4) MacKinnon, N.; Ge, W.; Khan, A. P.; Somashekar, B. S.; Tripathi, P.; Siddiqui, J.; Wei, J. T.; Chinnaiyan, A. M.; Rajendiran, T. M.; Ramamoorthy, A. Anal. Chem. 2012, 84, 5372−5379. (5) Somashekar, B. S.; Kamarajan, P.; Danciu, T.; Kapila, Y. L.; Chinnaiyan, A. M.; Rajendiran, T. M.; Ramamoorthy, A. J. Proteome. Res. 2011, 10, 5232−5241. (6) MacKinnon, N.; Khan, A.; Chinnaiyan, A.; Rajendiran, T.; Ramamoorthy, A. Metabolomics 2012, DOI: 10.1007/s11306-0120398-4. (7) Malz, F.; Jancke, H. J. Pharm. Biomed. Anal. 2005, 38, 813−823. (8) Bharti, S. K.; Roy, R. TrAC, Trends Anal. Chem. 2012, 35, 5−26. (9) Holmes, E.; Foxall, P. J.; Spraul, M.; Farrant, R. D.; Nicholson, J. K.; Lindon, J. C. J. Pharm. Biomed. Anal. 1997, 15, 1647−1659. (10) Zheng, M.; Lu, P.; Liu, Y.; Pease, J.; Usuka, J.; Liao, G.; Peltz, G. Bioinformatics 2007, 23, 2926−2933. (11) Tang, H.; Wang, Y.; Nicholson, J. K.; Lindon, J. C. Anal. Biochem. 2004, 325, 260−272. (12) Adosraku, R. K.; Choi, G. T.; Constantinou-Kokotos, V.; Anderson, M. M.; Gibbons, W. A. J. Lipid Res. 1994, 35, 1925−1931. (13) Van, Q. N.; Issaq, H. J.; Jiang, Q.; Li, Q.; Muschik, G. M.; Waybright, T. J.; Lou, H.; Dean, M.; Uitto, J.; Veenstra, T. D. J. Proteome. Res. 2008, 7, 630−639. (14) Koskela, H.; Kilpelainen, I.; Heikkinen, S. J. Magn. Reson. 2005, 174, 237−244. (15) Lewis, I. A.; Schommer, S. C.; Hodis, B.; Robb, K. A.; Tonelli, M.; Westler, W. M.; Sussman, M. R.; Markley, J. L. Anal. Chem. 2007, 79, 9385−9390. (16) Hu, F.; Furihata, K.; Kato, Y.; Tanokura, M. J. Agric. Food. Chem. 2007, 55, 4307−4311. (17) Gronwald, W.; Klein, M. S.; Kaspar, H.; Fagerer, S. R.; Nurnberger, N.; Dettmer, K.; Bertsch, T.; Oefner, P. J. Anal. Chem. 2008, 80, 9288−9297. (18) Zhang, L.; Gellerstedt, G. Magn. Reson. Chem. 2007, 45, 37−45. 10010

dx.doi.org/10.1021/ac302457s | Anal. Chem. 2012, 84, 10005−10011

Analytical Chemistry

Article

(48) Kumar, S.; Ghoshal, U.; Jayalakshmi, K.; Roy, R.; Misra, A.; Khetrapal, C. Dig. Dis. Sci. 2011, 56, 161−169. (49) Jayalakshmi, K.; Ghoshal, U. C.; Kumar, S.; Misra, A.; Roy, R.; Khetrapal, C. L. J. Gastrointest. Liver Dis. 2009, 18, 27−32.

10011

dx.doi.org/10.1021/ac302457s | Anal. Chem. 2012, 84, 10005−10011