Use of Semiselective TOCSY and the Pearson Correlation for the Metabonomic Analysis of Biofluid Mixtures: Application to Urine Peter Sandusky † and Daniel Raftery* Department of Chemistry and the Bindley Biosciences Center/Discovery Park, Purdue University, West Lafayette, Indiana 47907 The authors recently proposed an approach to the meta- bonomic analysis of biofluid mixtures based on the use of the selective TOCSY experiment (Sandusky, P.; Raftery, D. Anal. Chem. 2005, 77, 2455). This method has some significant advantages over standard metabo- nomic analysis. However, when analyzing overlapped components, the selective TOCSY method can suffer from the relatively high likelihood of simultaneous excitation of several spin systems at once. This multiple excitation can cause problems both with the purity of the individual TOCSY peaks observed and with their assignment into specific spin systems. To address this problem, the possibility of using a more selective excitation is initially explored. Unfortunately, in most cases, greater spin system selectivity can only be gained at the expense of sensitivity. This is obviously an unacceptable tradeoff when dealing with biofluid samples. However, the ap- plication of the Pearson product moment correlation to the TOCSY peak integral intensities provides a test for individual TOCSY peak purity and allows for the assign- ment of the peaks into spin systems. The specific applica- tion of this two-stage “semiselective” TOCSY method to rat and human urine is presented. Significantly, it is also demonstrated that the use of semiselective TOCSY spectra as data inputs for PCA calculations provides a more sensitive and reliable method of distinguishing small differences in biofluid composition than the standard metabonomic approach using complete 1D proton NMR spectra of urine samples. The metabolomics approach, combining high-resolution NMR with multivariate statistical analysis, has been shown to be very powerful for distinguishing biofluid sample subpopulations based on subtle differences in the their spectra. 1,2 This approach can be widely applied to many types of samples, including urine, body fluids, and tissues. NMR-based approaches are attractive because they can look at essentially all of the components of a mixture simultaneously and thus avoid the sometimes difficult process of sample fractionation. These methods can also be rapid and quantitative. The authors have recently published a method employing the selective TOCSY experiment for the quantification of chemical species in honey and have proposed that this method may be used as the basis for an alternative approach to the metabonomic analysis of biofluids. 3 Briefly, a selective excitation pulse and TOCSY mixing period were used to focus the statistical analysis on a few preselected components. In the case of honey, amino acids were anticipated to vary significantly among the different honeys and, thus, were the focus of those experiments. This approach dramatically improved the discrimination of subpopu- lations in a set of samples, because the signals used came almost exclusively from components that varied significantly between samples. The new method’s key virtue is that it facilitates the accurate quantification of a predetermined set of chemical species, regardless of whether these species are major or minor compo- nents of the mixture. A set of chemical compounds to be studied in a metabonomic analysis may then be chosen based on their metabolic or pharmacological significance. For instance, a specific subset of chemical compounds present in a biofluid may be chosen for study because they are known to be metabolically related. One key limitation of the new method, and of other 1D NMR- based metabolomics approaches, results from the fact that in many biofluids of interest, such as urine and blood serum, only a fraction of the NMR spectral features of the different chemical species are resolved. The selective TOCSY experiment becomes “semise- lective” when applied to these mixtures, in that a single selective TOCSY spectrum will very often contain peaks from several different chemical species. In this paper, using human and rat urine as examples of typical biofluid samples, we examine and compare two different solutions to this problem. Longer shaped pulse durations in the selective TOCSY pulse sequence will narrow the selective excitation band, thus focusing the experiment more selectively on individual chemical species. Unfortunately, as reported here, increasing the shaped pulse duration results in a significant decrease in the intensity of the TOCSY peaks, and the resulting increase in selectivity is seldom worth the cost in terms of sensitivity. A better alternative is to use a relatively short excitation pulse duration to produce a spectrum composed of more * Corresponding author: (e-mail) raftery@purdue.edu. † Present address: Coordinated Instrumentation Facility, 605 Lindy Boggs, Tulane University, New Orleans, LA 70118. (1) Lindon, J. C.; Holmes, E.; Nicholson, J. K. Prog. Nucl. Magn. Reson. Spectrosc. 2001, 39,1-40. (2) Lindon, J. C.; Holmes, E.; Nicholson, J. K. Prog. Nucl. Magn. Reson. Spectrosc. 2004, 45, 109-143. (3) Sandusky, P.; Raftery, D. Anal. Chem. 2005, 77, 2455-2463. Anal. Chem. 2005, 77, 7717-7723 10.1021/ac0510890 CCC: $30.25 © 2005 American Chemical Society Analytical Chemistry, Vol. 77, No. 23, December 1, 2005 7717 Published on Web 10/25/2005