Published: January 11, 2011 r2011 American Chemical Society 1116 dx.doi.org/10.1021/ac103011b | Anal. Chem. 2011, 83, 1116–1123 ARTICLE pubs.acs.org/ac Technical and Biological Variation in UPLC-MS-Based Untargeted Metabolic Profiling of Liver Extracts: Application in an Experimental Toxicity Study on Galactosamine Perrine Masson, † Konstantina Spagou, †,‡ Jeremy K. Nicholson, † and Elizabeth J. Want* ,† † Biomolecular Medicine, Department of Surgery and Cancer, Faculty of Medicine, Sir Alexander Fleming Building, Imperial College London, South Kensington, SW7 2AZ, U.K. ‡ Laboratory of Forensic Medicine and Toxicology, Faculty of Medicine, Aristotle University of Thessaloniki, 54124 Thessaloniki, Greece b S Supporting Information ABSTRACT: The relative importance of technical versus bio- logical variation in UPLC-MS liver metabolic profiling studies was assessed on liver samples collected as part of an in vivo hepato- toxicity study. Biological variability within and between two treat- ment groups (three rats treated with galactosamine and three with galactosamineþuridine) was compared with sampling/extraction variability (three portions extracted from each rat liver section) and UPLC-MS platform variability (triplicate injections of each extract) for aqueous and organic extracts. The impact of scaling on error measurement was investigated on replicate injections of a quality control sample, and consequently started log-transformation was used to stabilize the variance across the ion intensity range. For aqueous extracts, technical variability was two to four times lower than within group interanimal variability. Similar results were obtained for organic extracts for the galactosamine group, sampling/extraction variability being more elevated in the galactosamineþuridine group. For both extract types, di fferences between treatment groups were the principal source of observed variation, and triplicate injections clustered closely in PCA plots and in HCA dendrograms, indicating small instrument variability compared to observed biological variation. This protocol can be applied to investigate di fferences in liver metabolic profiles between animal groups in toxicology studies and clinical investigations of liver disease. U ntargeted metabolic profiling of biological samples aims to determine simultaneously and comprehensively the levels of all metabolites present and monitor their fluctuations over time due to stimuli such as diet, environment, diseases, or pharmaceutical interventions. 1,2 Exploratory metabolic profiling is commonly per- formed using nuclear magnetic resonance (NMR) spectroscopy and/or mass spectrometry (MS), the latter being generally coupled with gas chromatography (GC) or liquid chromatography (LC). 3 Interpretation of the complex resulting spectroscopic data sets is usually performed using chemometric mathematical modeling methods, such as principal components analysis (PCA) and partial least-squares discriminant analysis (PLS-DA). 4,5 Metabolic profiling has been applied to a broad range of environmental, preclinical, and clinical investigations, 6 e.g., monitoring exposure to environmental chemicals, 7 diagnosing and monitoring disease, 8,9 and identifying mechanisms and biomarkers of xenobiotic toxicity. 10-12 The liver plays a major role in the metabolism of many endogenous and exogenous compounds, such as drugs, food additives, bile acids, fatty acids, and steroids. 13-16 Thus, in the study of drug toxicity or liver disease, untargeted metabolic profiling of liver samples is likely to generate useful biological information, e.g., revealing affected metabolic pathways. This profiling can be per- formed on intact tissues by high-resolution magic angle spinning NMR (MAS NMR) spectroscopy 17-20 or on liquid metabolite extracts by conventional NMR spectroscopy 18,21,22 or by MS. 23,24 Among these techniques, Ultra Performance LC-MS (UPLC- MS) offers high sensitivity and resolution for a broad range of metabolites and has therefore been applied to various biological sample types. 3,25-31 To cover the large variability of metabolites in terms of chemical properties in the liver, polar and nonpolar metabolites are generally extracted into two distinct fractions and analyzed in two separate analytical runs, 23 allowing the use of optimized UPLC-MS conditions for each type of extract. The ability to distinguish biological, i.e., “true”, variation from technical variation is a key concern in metabolic profiling studies, regardless of the analytical platform and the sample type. 12,32-38 In the case of untargeted metabolic profiling of liver samples by UPLC-MS, technical variation can arise among others from sampling, metabolite extraction, and instrumental platform varia- bility. Concerning sampling, metabolite extracts are generally ob- tained from a small liver portion, cut from a section collected during in vivo studies at time of termination. These sections, although taken from a small, predefined part of the liver, are often not homogeneous and may lead to variable results depending on the sampled portion. 39 Received: November 18, 2010 Accepted: December 16, 2010