HiTIME - American Chemical Society

---

Technical Note pubs.acs.org/ac

High-Resolution Twin-Ion Metabolite Extraction (HiTIME) Mass Spectrometry: Nontargeted Detection of Unknown Drug Metabolites by Isotope Labeling, Liquid Chromatography Mass Spectrometry, and Automated High-Performance Computing Michael G. Leeming,† Andrew P. Isaac,‡ Bernard J. Pope,‡,§ Noel Cranswick,∥,¶ Christine E. Wright,∥,⊥ James Ziogas,∥,⊥ Richard A. J. O’Hair,*,†,⊥ and William A. Donald*,# †

School of Chemistry and Bio21 Institute of Molecular Science and Biotechnology, University of Melbourne, 30 Flemington Road, Melbourne, Victoria 3010, Australia ‡ Victorian Life Sciences Computation Initiative, University of Melbourne, 187 Grattan Street, Carlton, Victoria 3010, Australia § Department of Computing and Information Systems, University of Melbourne, Parkville, Victoria 3010, Australia ∥ Department of Pharmacology and Therapeutics, University of Melbourne, Victoria 3010, Australia ⊥ ARC Centre of Excellence for Free Radical Chemistry and Biotechnology, University of Melbourne, Melbourne, Victoria 3010, Australia ¶ Royal Children's Hospital Melbourne, 50 Flemington Road, Victoria 3052, Australia # School of Chemistry, University of New South Wales, Sydney, New South Wales 2052, Australia S Supporting Information *

ABSTRACT: The metabolic fate of a compound can often determine the success of a new drug lead. Thus, significant effort is directed toward identifying the metabolites formed from a given molecule. Here, an automated and nontargeted procedure is introduced for detecting drug metabolites without authentic metabolite standards via the use of stable isotope labeling, liquid chromatography mass spectrometry (LC/ MS), and high-performance computing. LC/MS of blood plasma extracts from rats that were administered a 1:1 mixture of acetaminophen (APAP) and 13C6-APAP resulted in mass spectra that contained “twin” ions for drug metabolites that were not detected in control spectra (i.e., no APAP administered). Because of the development of a program (high-resolution twin-ion metabolite extraction; HiTIME) that can identify twin-ions in highresolution mass spectra without centroiding (i.e., reduction of mass spectral peaks to single data points), 9 doublets corresponding to APAP metabolites were identified. This is nearly twice that obtained by use of existing programs that make use of centroiding to reduce computational cost under these conditions with a quadrupole time-of-flight mass spectrometer. By a manual search for all reported APAP metabolite ions, no additional twin-ion signals were assigned. These data indicate that all the major metabolites of APAP and multiple low-abundance metabolites (e.g., acetaminophen hydroxy- and methoxysulfate) that are rarely reported were detected. This methodology can be used to detect drug metabolites without prior knowledge of their identity. HiTIME is freely available from https://github.com/bjpop/HiTIME.

L

radiolabeled drug compounds, fractionation of biological samples, and detection by scintillation counting.2 More recently, a range of LC/MS experiments and data treatments have been developed that aim to differentiate drug metabolites from endogenous compounds and eliminate the need for radiotracers,1a including those that attempt to differentiate the metabolites of drug molecules from endogenous compounds by use of mass defect filters (i.e., for drugs with relatively unusual mass defects)3 and tandem mass spectrometry (i.e., for metabolite ions with known/predicted fragmentation chem-

iquid chromatography mass spectrometry (LC/MS) is widely used for detecting and identifying drug metabolites at trace levels in highly complex biological fluids that can contain thousands of endogenous compounds.1 For metabolites that are unknown and unexpected, it can be challenging to determine which of the thousands of ions detected in an LC/ MS experiment correspond to the downstream metabolites of precursor compounds. This “needle-in-a-haystack” problem requires the development of methods to rapidly identify the products of compounds in complex mixtures, at trace levels, and without prior knowledge of the identity of the metabolites. Numerous methods have been developed that aim to detect metabolites with little or no prior information (i.e., nontargeted detection), which traditionally involve the administration of © XXXX American Chemical Society

Received: October 30, 2014 Accepted: March 28, 2015

A

DOI: 10.1021/ac504767d Anal. Chem. XXXX, XXX, XXX−XXX

Analytical Chemistry

Technical Note

Figure 1. (A) Profile LC/MS data in which doublet peaks assigned to APAP and [13C6]-APAP [M + H]+ are shown. (B) Representative “ideal” twinion peaks that were used to score the data. (C) Output data after scoring centered on the light isotope (HiTIME confidence of twin-ion “hit”). (D) Visualization of the results from HiTIME scoring as a two-dimensional heat map. RT corresponds to retention time.

istry).4 Although widely used, these LC/MS methods are usually less selective than radiotracers or require prior knowledge about metabolite identity and ion fragmentation patterns. An elegant approach to the rapid nontargeted detection of drug metabolites was demonstrated in the early 1970s by Knapp et al., who simultaneously administered a 1:1 mixture of nortriptyline and [2H3]-nortriptyline to human subjects and then analyzed urine samples using gas chromatography mass spectrometry (GC/MS).5 The 2H3-labeled and unlabeled metabolites of nortriptyline eluted from the GC column simultaneously in nearly the same abundances, and the m/z values differed by the mass of the isotopic label. These “twinions” provide a signature that is unique to drug metabolites, which can be used to differentiate metabolites that retain the isotope label from endogenous compounds without authentic standards or any information about the identity of the metabolites (i.e., nontargeted detection). Despite the obvious practical advantages of nontargeted twin-ion metabolite detection, the use of this method has been limited6 partly because (i) the resolution of many types of common mass spectrometers was insufficient for the detection of lowabundance twin-ions, and (ii) it was challenging to identify twin-ion signatures in large data sets.7 Given the significant advances in mass spectrometer performance, there has been renewed interest in using the twin-ion method for a range of biological applications,8 including the detection of trace level metabolites of insecticides in susceptible insects in vivo9 and the elucidation of metabolic pathways and mechanisms.6a The twin-ion method is complementary to other isotope-dilution based LC/MS experiments that are used to obtain pharmacokinetic parameters (e.g., bioavailability).10 Owing to the number of ions that can be detected in LC/MS experiments, it is important that twin-ion metabolite detection is automated to

alleviate the laborious and inefficient task of manually searching through such data sets. Several computer algorithms have been developed to identify twin-ions in LC/MS data.11 For example, MetExtract developed by Schuhmacher and co-workers11f initially considers each data point in each mass spectrum to be the unlabeled metabolite. If a twin-ion of comparable abundance is detected at a higher m/z value that is defined by the mass of the isotope label, a metabolite ion is assigned. However, all existing programs (to our knowledge) for automated twin-ion detection of unknown isotope-labeled metabolite ions are optimized for data sets that have been “centroided”, in which the number of data points corresponding to a given ion is reduced to a single point. While this significantly reduces computational cost, centroiding can increase the signal-to-noise (S/N) that is required to confidently detect a given ion. Ideally, all data reduction steps should be eliminated to enable trace level metabolites to be reliably detected. Here, we demonstrate a nontargeted approach to detect drug metabolites that combines the twin-ion method with automated data-mining software that is entitled high-resolution twin-ion metabolite extraction (HiTIME). We have selected acetaminophen (APAP), a clinically relevant and widely used analgesic, to test this approach because it has a metabolic profile that is wellestablished. With the elimination of centroiding, the number of APAP metabolites that can be detected in plasma extracts from rats (therapeutic dose) by use of HiTIME increased by nearly a factor of 2 under these conditions.

■

EXPERIMENTS A 1:1 mixture of APAP:13C6-APAP [5 mg kg−1 of each isotopologue, ca. therapeutic dose; 13C6-APAP corresponds to N-(4-hydroxy-13C6-phenyl)acetamide] was administered intravenously to adult male Sprague−Dawley rats (327−366 g). Blood samples were drawn after 30 min. Blood plasma was B

DOI: 10.1021/ac504767d Anal. Chem. XXXX, XXX, XXX−XXX

Analytical Chemistry

Technical Note

Numerous “bright spots” were assigned HiTIME scores of >5 indicating a strong correlation between the data in these regions and the model twin-ion signature. These signals likely arise from the elution of isotopically labeled drug metabolites. However, endogenous compounds unrelated to one another may by chance result in an LC/MS profile similar to that of the twin-ion signature. Cumulative intensity histograms depicting the distribution of weighted values are shown in Supporting Information Figure S3. Of the weighted values (representing tens of thousands of data points), 80% are assigned a very lowconfidence value (