Experimental and Analytical Variation in Human Urine in 1 H NMR Spectroscopy-Based Metabolic Phenotyping Studies Anthony D. Maher, Se ´ verine F. M. Zirah, Elaine Holmes, and Jeremy K. Nicholson* Department of Biomolecular Medicine, Division of Surgery, Oncology, Reproductive Biology and Anaesthetics (SORA), Faculty of Medicine, Imperial College London, South Kensington SW7 2AZ, United Kingdom 1 H NMR spectroscopy potentially provides a robust ap- proach for high-throughput metabolic screening of bio- fluids such as urine and plasma, but sample handling and preparation need careful optimization to ensure that spectra accurately report biological status or disease state. We have investigated the effects of storage temperature and time on the 1 H NMR spectral profiles of human urine from two participants, collected three times a day on four different days. These were analyzed using modern chemo- metric methods. Analytical and preparation variation (tested between -40 °C and room temperature) and time of storage (to 24 h) were found to be much less influential than biological variation in sample classification. Statisti- cal total correlation spectroscopy and discriminant func- tion methods were used to identify the specific metabolites that were hypervariable due to preparation and biology. Significant intraindividual variation in metabolite profiles were observed even for urine collected on the same day and after at least 6 h fasting. The effect of long-term storage at different temperatures was also investigated, showing urine is stable if frozen for at least 3 months and that storage at room temperature for long periods (1-3 months) results in a metabolic profile explained by bacterial activity. Presampling (e.g., previous day) intake of food and medicine can also strongly influence the urinary metabolic profiles indicating that collective de- tailed participant historical meta data are important for interpretation of metabolic phenotypes and for avoiding false biomarker discovery. 1 H NMR spectroscopy has been used since the early 1980s to effect multivariate metabolic profiling in man and animals includ- ing diseases such as type II diabetes mellitus and hyperlipidemia. 1-3 Even these early studies showed good analytical reproducibility and quantitative accuracy of NMR measurements on defined compounds with reference to established HPLC or enzymatic assays for a variety of metabolites. 2,3 The applications of pattern recognition methods to analyze NMR data in the late 1980s 4 showed that effective classification of samples based on their NMR-derived metabolic properties could be achieved and that a wide range of toxic and disease states could be mapped and modeled. 5-10 Plasma and urine samples give complementary “global” representations of the integrated metabolic state of an organism, i.e., patterns related to metabolic processes under the control of many cell and tissue types. Plasma gives an “instanta- neous” (i.e., representing the metabolic content at the exact time of collection) readout of systematically controlled metabolites and lipoproteins, whereas urine gives a “time-averaged” pattern for polar metabolites that are excreted in variable amounts according to variations in whole-body homeostatic control. Metabonomic and metabolomic approaches to biomarker discovery are now wide- spread, and understanding analytical and biological variation is critically important in any metabolic profiling study. In particular, it is necessary to quantify and separate artifactual and analytical variation from the biological variations of interest and, at a higher level, to understand the combined sources of biological variation that may confuse the analysis and lead to erroneous biological conclusions. Sources of variation in metabolic studies that must be consid- ered include the following: (i) sample collection and storage; (ii) sample pretreatment prior to analysis; (iii) instrumental variation and calibration; (iv) intraindividual variation due to physiological factors and temporal patterns in exposure to environmental factors such as nutrients or stress; (v) interindividual variation due to genetic and environmental factors; (vi) interindividual variation related to biological hypothesis, e.g., presence of a particular disease state. Metabolites vary considerably in their chemical * To whom correspondence should be addressed. E-mail: j.nicholson@ imperial.ac.uk. (1) Bales, J. R.; Higham, D. P.; Howe, I.; Nicholson, J. K.; Sadler, P. J. Clin. Chem. 1984, 30, 426-432. (2) Nicholson, J. K.; Buckingham, M. J.; Sadler, P. J. Biochem. J. 1983, 211, 605-615. (3) Nicholson, J. K.; O’Flynn, M. P.; Sadler, P. J.; Macleod, A. F.; Juul, S. M.; Sonksen, P. H. Biochem. J. 1984, 217, 365-375. (4) Nicholson, J. K.; Wilson, I. D. Prog. Nucl. Magn. Reson. Spectrosc. 1989, 21, 449-501. (5) Gartland, K. P.; Beddell, C. R.; Lindon, J. C.; Nicholson, J. K. Mol. Pharmacol. 1991, 39, 629-642. (6) Gartland, K. P.; Sanins, S. M.; Nicholson, J. K.; Sweatman, B. C.; Beddell, C. R.; Lindon, J. C. NMR Biomed. 1990, 3, 166-172. (7) Holmes, E.; Bonner, F. W.; Sweatman, B. C.; Lindon, J. C.; Beddell, C. R.; Rahr, E.; Nicholson, J. K. Mol. Pharmacol. 1992, 42, 922-930. (8) Ghauri, F.; Nicholson, J. K.; Sweatman, B. C.; Wood, J.; Beddell, C. R.; Cairns, N. NMR Biomed. 1993, 6, 163-167. (9) Anthony, M. L.; Lindon, J. C.; Beddell, C. R.; Nicholson, J. K. Mol. Pharmacol. 1994, 46, 199-211. (10) Foxall, P. J.; Singer, J. M.; Hartley, J. M.; Neild, G. H.; Lapsley, M.; Nicholson, J. K.; Souhami, R. L. Clin. Cancer Res. 1997, 3, 1507-1518. Anal. Chem. 2007, 79, 5204-5211 5204 Analytical Chemistry, Vol. 79, No. 14, July 15, 2007 10.1021/ac070212f CCC: $37.00 © 2007 American Chemical Society Published on Web 06/08/2007