Characterization of Isotopic Abundance Measurements in High

Apr 5, 2011 - Centre for Systems Biology,. ‡. School of Biosciences, and ... University of Birmingham, Edgbaston, Birmingham, B15 2TT, U.K.. bS Supp...
0 downloads 0 Views 1MB Size
ARTICLE pubs.acs.org/ac

Characterization of Isotopic Abundance Measurements in High Resolution FT-ICR and Orbitrap Mass Spectra for Improved Confidence of Metabolite Identification Ralf J. M. Weber,† Andrew D. Southam,‡ Ulf Sommer,§ and Mark R. Viant*,†,‡,§ †

Centre for Systems Biology, ‡School of Biosciences, and §NERC Biomolecular Analysis Facility - Metabolomics Node (NBAF-B), University of Birmingham, Edgbaston, Birmingham, B15 2TT, U.K.

bS Supporting Information ABSTRACT: Currently there is limited information available on the accuracy and precision of relative isotopic abundance (RIA) measurements using high-resolution direct-infusion mass spectrometry (HR DIMS), and it is unclear if this information can benefit automated peak annotation in metabolomics. Here we characterize the accuracy of RIA measurements on the Thermo LTQ FT Ultra (resolution of 100 000750 000) and LTQ Orbitrap (R = 100 000) mass spectrometers. This first involved reoptimizing the SIM-stitching method (Southam, A. D. Anal. Chem. 2007, 79, 45954602) for the LTQ FT Ultra, which achieved a ca. 3-fold sensitivity increase compared to the original method while maintaining a rootmean-squared mass error of 0.16 ppm. Using this method, we show the quality of RIA measurements is highly dependent on signalto-noise ratio (SNR), with RIA accuracy increasing with higher SNR. Furthermore, a negative offset between the theoretical and empirically calculated numbers of carbon atoms was observed for both mass spectrometers. Increasing the resolution of the LTQ FT Ultra lowered both the sensitivity and the quality of RIA measurements. Overall, although the errors in the empirically calculated number of carbons can be large (e.g., 10 carbons), we demonstrate that RIA measurements do improve automated peak annotation, increasing the number of single empirical formula assignments by >3-fold compared to using accurate mass alone.

M

etabolomics aims to detect and quantify all low molecular weight metabolites within a biological sample and has generated novel biochemical insight in many fields, particularly toxicology and disease diagnostics.2,3 High resolution Fourier transform mass spectrometry (HR FTMS), including Fourier transform ion cyclotron resonance (FT-ICR) MS4 and Orbitrap technologies,5 has become a leading analytical platform in metabolomics.6 HR FTMS analysis of a biological sample yields an extremely complex spectrum,7,8 typically containing thousands of metabolite signals, including multiple ion forms (e.g., [M þ H]þ, [M þ Na]þ and [M þ 39K]þ) and naturally occurring isotopes (e.g., 13C12Cn-1 and 18O16On-1). The high mass accuracy and resolution of HR FTMS instruments, together with strategies to improve their dynamic range, e.g., selected ion monitoring (SIM)-stitching,1 provide the critical combination of specifications to investigate the metabolome in detail.9 However, automated identification of the thousands of signals arguably remains the greatest challenge in metabolomics. This includes assigning one (or more) empirical formula(s) to each peak in the mass spectrum, assigning metabolite name(s) to these empirical formula(s) and, for definitive identification, further analytical measurements using, for example, liquid chromatography and/or fragmentation (MSn).1012 Here we focus on the primary step in metabolite identification, namely assigning empirical formula(s) to accurate mass measurements. High mass accuracy greatly facilitates this process by allowing a small mass error tolerance during searches, thus reducing the r 2011 American Chemical Society

number of potential empirical formula(s) found. Heuristic rules and restrictions can be applied to remove incorrect elemental compositions (e.g., the number and type of atoms, and atom ratios), ultimately yielding the most likely empirical formula(s).13 However, the number of potential empirical formulas per peak increases significantly with m/z; e.g., whereas m/z = 148.00389 can be assigned to a single formula, m/z = 450.01864 has 47 possible assignments (for e1 ppm mass error tolerance and [C034H072N015O019P07S08 þ H or Na or K]þ). Consequently, further strategies must be employed to reduce false-positive assignments and improve confidence in metabolite identification. Relative isotopic abundance (RIA) measurements (e.g., from isotope-pairs such as 12Cn and 13 12 C Cn-1, or 32Sn and 34S32Sn-1) can, in principle, be highly informative for assigning empirical formula(s) by providing an estimate of the numbers of atoms present. Fiehn et al. (2006) and others reported a considerable benefit of incorporating RIA into empirical formula(s) determination;1416 e.g., for a peak at m/z = 500.00000 (assuming 3 ppm mass accuracy), 33 potential empirical formulas can be assigned, yet on adding RIA measurements (assuming 2% accuracy) this decreases to only three possible assignments. This demonstrates the value of RIA measurements for cases where the intensity measurements are Received: January 21, 2011 Accepted: April 5, 2011 Published: April 05, 2011 3737

dx.doi.org/10.1021/ac2001803 | Anal. Chem. 2011, 83, 3737–3743

Analytical Chemistry accurate, but how appropriate is this assumption for HR FTMS instruments used in metabolomics? It has been demonstrated that FT-ICR MS RIA measurements of a polyethylene glycol (PEG) solution are relatively poor above m/z 500, with absolute errors in the number of carbons ranging from 0 to 96.17 Using the same type of instrument, a precision of (1.6 carbons was reported for peaks