When using liquid chromatography/mass spectrometry (LC/MS) to perform untargeted metabolomics, it is now routine to detect tens of thousands of features from biological samples. Poor understanding of the data, however, has complicated interpretation and masked the number of unique metabolites actually being measured in an experiment. Here we place an upper bound on the number of unique metabolites detected in Escherichia coli samples analyzed with one untargeted metabolomics method. We first group multiple features arising from the same analyte, which we call “degenerate features”, using a context-driven annotation approach. Surprisingly, this analysis revealed thousands of previously unreported degeneracies that reduced the number of unique analytes to ∼2961. We then applied an orthogonal approach to remove nonbiological features from the data using the 13C-based credentialing technology. This further reduced the number of unique analytes to less than 1000. Our 90% reduction in data is 5-fold greater than previously published studies. On the basis of the results, we propose an alternative approach to untargeted metabolomics that relies on thoroughly annotated reference data sets. To this end, we introduce the creDBle database (http://creDBle.wustl.edu), which contains accurate mass, retention time, and MS/MS fragmentation data as well as annotations of all credentialed features.

Metabolite identifications are most frequently achieved in untargeted metabolomics by matching precursor mass and full, high-resolution MS2 spectra to metabolite databases and standards. Here we considered an alternative approach for establishing metabolite identifications that does not rely on full, high-resolution MS2 spectra. First, we select mass-to-charge regions containing the most informative metabolite fragments and designate them as bins. We then translate each metabolite fragmentation pattern into a binary code by assigning 1’s to bins containing fragments and 0’s to bins without fragments. With 20 bins, this binary-code system is capable of distinguishing 96% of the compounds in the METLIN MS2 library. A major advantage of the approach is that it extends untargeted metabolomics to low-resolution triple quadrupole (QqQ) instruments, which are typically less expensive and more robust than other types of mass spectrometers. We demonstrate a method of acquiring MS2 data in which the third quadrupole of a QqQ instrument cycles over 20 wide isolation windows (coinciding with the location and width of our bins) for each precursor mass selected by the first quadrupole. Operating the QqQ instrument in this mode yields diagnostic bar codes for each precursor mass that can be matched to the bar codes of metabolite standards. Furthermore, our data suggest that using low-resolution bar codes enables QqQ instruments to make MS2-based identifications in untargeted metabolomics with a specificity and sensitivity that is competitive to high-resolution time-of-flight technologies.

Analysis of a single analyte by mass spectrometry can result in the detection of more than 100 degenerate peaks. These degenerate peaks complicate spectral interpretation and are challenging to annotate. In mass spectrometry-based metabolomics, this degeneracy leads to inflated false discovery rates, data sets containing an order of magnitude more features than analytes, and an inefficient use of resources during data analysis. Although software has been introduced to annotate spectral degeneracy, current approaches are unable to represent several important classes of peak relationships. These include heterodimers and higher complex adducts, distal fragments, relationships between peaks in different polarities, and complex adducts between features and background peaks. Here we outline sources of peak degeneracy in mass spectra that are not annotated by current approaches and introduce a software package called mz.unity to detect these relationships in accurate mass data. Using mz.unity, we find that data sets contain many more complex relationships than we anticipated. Examples include the adduct of glutamate and nicotinamide adenine dinucleotide (NAD), fragments of NAD detected in the same or opposite polarities, and the adduct of glutamate and a background peak. Further, the complex relationships we identify show that several assumptions commonly made when interpreting mass spectral degeneracy do not hold in general. These contributions provide new tools and insight to aid in the annotation of complex spectral relationships and provide a foundation for improved data set identification. Mz.unity is an R package and is freely available at https://github.com/nathaniel-mahieu/mz.unity as well as our laboratory Web site http://pattilab.wustl.edu/software/.

An autonomous metabolomic workflow combining mass spectrometry analysis with tandem mass spectrometry data acquisition was designed to allow for simultaneous data processing and metabolite characterization. Although previously tandem mass spectrometry data have been generated on the fly, the experiments described herein combine this technology with the bioinformatic resources of XCMS and METLIN. As a result of this unique integration, we can analyze large profiling datasets and simultaneously obtain structural identifications. Validation of the workflow on bacterial samples allowed the profiling on the order of a thousand metabolite features with simultaneous tandem mass spectra data acquisition. The tandem mass spectrometry data acquisition enabled automatic search and matching against the METLIN tandem mass spectrometry database, shortening the current workflow from days to hours. Overall, the autonomous approach to untargeted metabolomics provides an efficient means of metabolomic profiling, and will ultimately allow the more rapid integration of comparative analyses, metabolite identification, and data analysis at a systems biology level.

Two-hydroxyglutarate (2HG) is present at low concentrations in healthy mammalian cells as both an L and D enantiomer. Both the L and D enantiomers have been implicated in regulating cellular physiology by mechanisms that are only partially characterized. In multiple human cancers, the D enantiomer accumulates due to gain-of-function mutations in the enzyme isocitrate dehydrogenase (IDH) and has been hypothesized to drive malignancy through mechanisms that remain incompletely understood. While much attention has been dedicated to identifying the route of 2HG synthesis, the metabolic fate of 2HG has not been studied in detail. Yet the metabolism of 2HG may have important mechanistic consequences influencing cell function and cancer pathogenesis, such as modulating redox potential or producing unknown products with unique modes of action.By applying our isotope-based metabolomic platform, we unbiasedly and comprehensively screened for products of L- and D-2HG in HCT116 colorectal carcinoma cells harboring a mutation in IDH1. After incubating HCT116 cells in uniformly 13C-labeled 2HG for 24 h, we used liquid chromatography/mass spectrometry to track the labeled carbons in small molecules. Strikingly, we did not identify any products of 2HG metabolism from the thousands of metabolomic features that we screened. Consistent with these results, we did not detect any significant changes in the labeling patterns of tricarboxylic acid cycle metabolites from wild type or IDH1 mutant cells cultured in 13C-labeled glucose upon the addition of L, D, or racemic mixtures of 2HG. A more sensitive, targeted analysis revealed trace levels of isotopic enrichment (<1 %) in some central carbon metabolites from 13C-labeled 2HG. However, we found that cells do not deplete 2HG from the media at levels above our detection limit over a 48 h time period.Taken together, we conclude that 2HG carbon is not readily transformed in the HCT116 cell line. These data indicate that the phenotypic alterations induced by 2HG are not a result of its metabolic products.

Studies of isotopically labeled compounds have been fundamental to understanding metabolic pathways and fluxes. They have traditionally, however, been used in conjunction with targeted analyses that identify and quantify a limited number of labeled downstream metabolites. Here we describe an alternative workflow that leverages recent advances in untargeted metabolomic technologies to track the fates of isotopically labeled metabolites in a global, unbiased manner. This untargeted approach can be applied to discover novel biochemical pathways and characterize changes in the fates of labeled metabolites as a function of altered biological conditions such as disease. To facilitate the data analysis, we introduce X13CMS, an extension of the widely used mass spectrometry-based metabolomic software package XCMS. X13CMS uses the XCMS platform to detect metabolite peaks and perform retention-time alignment in liquid chromatography/mass spectrometry (LC/MS) data. With the use of the XCMS output, the program then identifies isotopologue groups that correspond to isotopically labeled compounds. The retrieval of these groups is done without any a priori knowledge besides the following input parameters: (i) the mass difference between the unlabeled and labeled isotopes, (ii) the mass accuracy of the instrument used in the analysis, and (iii) the estimated retention-time reproducibility of the chromatographic method. Despite its name, X13CMS can be used to track any isotopic label. Additionally, it detects differential labeling patterns in biological samples collected from parallel control and experimental conditions. We validated the ability of X13CMS to accurately retrieve labeled metabolites from complex biological matrices both with targeted LC/MS/MS analysis of a subset of the hits identified by the program and with labeled standards spiked into cell extracts. We demonstrate the full functionality of X13CMS with an analysis of cultured rat astrocytes treated with uniformly labeled (U-)13C-glucose during lipopolysaccharide (LPS) challenge. Our results show that out of 223 isotopologue groups enriched from U-13C-glucose, 95 have statistically significant differential labeling patterns in astrocytes challenged with LPS compared to unchallenged control cells. Only two of these groups overlap with the 32 differentially regulated peaks identified by XCMS, indicating that X13CMS uncovers different and complementary information from untargeted metabolomic studies. Like XCMS, X13CMS is implemented in R. It is available from our laboratory website at http://pattilab.wustl.edu/x13cms.php.

The aim of untargeted metabolomics is to profile as many metabolites as possible, yet a major challenge is comparing experimental method performance on the basis of metabolome coverage. To date, most published approaches have compared experimental methods by counting the total number of features detected. Due to artifactual interference, however, this number is highly variable and therefore is a poor metric for comparing metabolomic methods. Here we introduce an alternative approach to benchmarking metabolome coverage which relies on mixed Escherichia coli extracts from cells cultured in regular and 13C-enriched media. After mass spectrometry-based metabolomic analysis of these extracts, we “credential” features arising from E. coli metabolites on the basis of isotope spacing and intensity. This credentialing platform enables us to accurately compare the number of nonartifactual features yielded by different experimental approaches. We highlight the value of our platform by reoptimizing a published untargeted metabolomic method for XCMS data processing. Compared to the published parameters, the new XCMS parameters decrease the total number of features by 15% (a reduction in noise features) while increasing the number of true metabolites detected and grouped by 20%. Our credentialing platform relies on easily generated E. coli samples and a simple software algorithm that is freely available on our laboratory Web site (http://pattilab.wustl.edu/software/credential/). We have validated the credentialing platform with reversed-phase and hydrophilic interaction liquid chromatography as well as Agilent, Thermo Scientific, AB SCIEX, and LECO mass spectrometers. Thus, the credentialing platform can readily be applied by any laboratory to optimize their untargeted metabolomic pipeline for metabolite extraction, chromatographic separation, mass spectrometric detection, and bioinformatic processing.

The METLIN metabolite database has become one of the most widely used resources in metabolomics for making metabolite identifications. However, METLIN is not designed to identify metabolites that have been isotopically labeled. As a result, unbiasedly tracking the transformation of labeled metabolites with isotope-based metabolomics is a challenge. Here, we introduce a new database, called isoMETLIN (http://isometlin.scripps.edu/), that has been developed specifically to identify metabolites incorporating isotopic labels. isoMETLIN enables users to search all computed isotopologues derived from METLIN on the basis of mass-to-charge values and specified isotopes of interest, such as 13C or 15N. Additionally, isoMETLIN contains experimental MS/MS data on hundreds of isotopomers. These data assist in localizing the position of isotopic labels within a metabolite. From these experimental MS/MS isotopomer spectra, precursor atoms can be mapped to fragments. The MS/MS spectra of additional isotopomers can then be computationally generated and included within isoMETLIN. Given that isobaric isotopomers cannot be separated chromatographically or by mass but are likely to occur simultaneously in a biological system, we have also implemented a spectral-mixing function in isoMETLIN. This functionality allows users to combine MS/MS spectra from various isotopomers in different ratios to obtain a theoretical MS/MS spectrum that matches the MS/MS spectrum from a biological sample. Thus, by searching MS and MS/MS experimental data, isoMETLIN facilitates the identification of isotopologues as well as isotopomers from biological samples and provides a platform to drive the next generation of isotope-based metabolomic studies.

XCMS Online (xcmsonline.scripps.edu) is a cloud-based informatic platform designed to process and visualize mass-spectrometry-based, untargeted metabolomic data. Initially, the platform was developed for two-group comparisons to match the independent, “control” versus “disease” experimental design. Here, we introduce an enhanced XCMS Online interface that enables users to perform dependent (paired) two-group comparisons, meta-analysis, and multigroup comparisons, with comprehensive statistical output and interactive visualization tools. Newly incorporated statistical tests cover a wide array of univariate analyses. Multigroup comparison allows for the identification of differentially expressed metabolite features across multiple classes of data while higher order meta-analysis facilitates the identification of shared metabolic patterns across multiple two-group comparisons. Given the complexity of these data sets, we have developed an interactive platform where users can monitor the statistical output of univariate (cloud plots) and multivariate (PCA plots) data analysis in real time by adjusting the threshold and range of various parameters. On the interactive cloud plot, metabolite features can be filtered out by their significance level (p-value), fold change, mass-to-charge ratio, retention time, and intensity. The variation pattern of each feature can be visualized on both extracted-ion chromatograms and box plots. The interactive principal component analysis includes scores, loadings, and scree plots that can be adjusted depending on scaling criteria. The utility of XCMS functionalities is demonstrated through the metabolomic analysis of bacterial stress response and the comparison of lymphoblastic leukemia cell lines.

It is well established that most cancer cells take up an increased amount of glucose relative to that taken up by normal differentiated cells. The majority of this glucose carbon is secreted from the cell as lactate. The fate of the remaining glucose carbon, however, has not been well-characterized. Here we apply a novel combination of metabolomic technologies to track uniformly labeled glucose in HeLa cancer cells. We provide a list of specific intracellular metabolites that become enriched after being labeled for 48 h and quantitate the fraction of consumed glucose that ends up in proteins, peptides, sugars/glycerol, and lipids.

The manipulation of distinct signaling pathways and transcription factors has been shown to influence life span in a cell-non-autonomous manner in multicellular model organisms such as Caenorhabditis elegans. These data suggest that coordination of whole-organism aging involves endocrine signaling, however, the molecular identities of such signals have not yet been determined and their potential relevance in humans is unknown. Here we describe a novel metabolomic approach to identify molecules directly associated with extended life span in C. elegans that represent candidate compounds for age-related endocrine signals. To identify metabolic perturbations directly linked to longevity, we developed metabolomic software for meta-analysis that enabled intelligent comparisons of multiple different mutants. Simple pairwise comparisons of long-lived glp-1, daf-2, and isp-1 mutants to their respective controls resulted in more than 11,000 dysregulated metabolite features of statistical significance. By using meta-analysis, we were able to reduce this number to six compounds most likely to be associated with life-span extension. Mass spectrometry-based imaging studies suggested that these metabolites might be localized to C. elegans muscle. We extended the metabolomic analysis to humans by comparing quadricep muscle tissue from young and old individuals and found that two of the same compounds associated with longevity in worms were also altered in human muscle with age. These findings provide candidate compounds that may serve as age-related endocrine signals and implicate muscle as a potential tissue regulating their levels in humans.

Global metabolomics describes the comprehensive analysis of small molecules in a biological system without bias. With mass spectrometry-based methods, global metabolomic data sets typically comprise thousands of peaks, each of which is associated with a mass-to-charge ratio, retention time, fold change, p-value, and relative intensity. Although several visualization schemes have been used for metabolomic data, most commonly used representations exclude important data dimensions and therefore limit interpretation of global data sets. Given that metabolite identification through tandem mass spectrometry data acquisition is a time-limiting step of the untargeted metabolomic workflow, simultaneous visualization of these parameters from large sets of data could facilitate compound identification and data interpretation. Here, we present such a visualization scheme of global metabolomic data using a so-called “cloud plot” to represent multidimensional data from septic mice. While much attention has been dedicated to lipid compounds as potential biomarkers for sepsis, the cloud plot shows that alterations in hydrophilic metabolites may provide an early signature of the disease prior to the onset of clinical symptoms. The cloud plot is an effective representation of global mass spectrometry-based metabolomic data, and we describe how to extract it as standard output from our XCMS metabolomic software.

Mass spectrometry-based metabolomics relies on MS2 data for structural characterization of metabolites. To obtain the high-quality MS2 data necessary to support metabolite identifications, ions of interest must be purely isolated for fragmentation. Here, we show that metabolomic MS2 data are frequently characterized by contaminating ions that prevent structural identification. Although using narrow-isolation windows can minimize contaminating MS2 fragments, even narrow windows are not always selective enough, and they can complicate data analysis by removing isotopic patterns from MS2 spectra. Moreover, narrow windows can significantly reduce sensitivity. In this work, we introduce a novel, two-part approach for performing metabolomic identifications that addresses these issues. First, we collect MS2 scans with less stringent isolation settings to obtain improved sensitivity at the expense of specificity. Then, by evaluating MS2 fragment intensities as a function of retention time and precursor mass targeted for MS2 analysis, we obtain deconvolved MS2 spectra that are consistent with pure standards and can therefore be used for metabolite identification. The value of our approach is highlighted with metabolic extracts from brain, liver, astrocytes, as well as nerve tissue, and performance is evaluated by using pure metabolite standards in combination with simulations based on raw MS2 data from the METLIN metabolite database. A R package implementing the algorithms used in our workflow is available on our laboratory website (http://pattilab.wustl.edu/decoms2.php).

Although the objective of any ‘omic science is broad measurement of its constituents, such coverage has been challenging in metabolomics because the metabolome is comprised of a chemically diverse set of small molecules with variable physical properties. While extensive studies have been performed to identify metabolite isolation and separation methods, these strategies introduce bias toward lipophilic or water-soluble metabolites depending on whether reversed-phase (RP) or hydrophilic interaction liquid chromatography (HILIC) is used, respectively. Here we extend our consideration of metabolome isolation and separation procedures to integrate RPLC/MS and HILIC/MS profiling. An aminopropyl-based HILIC/MS method was optimized on the basis of mobile-phase additives and pH, followed by evaluation of reproducibility. When applied to the untargeted study of perturbed bacterial metabolomes, the HILIC method enabled the accurate assessment of key, dysregulated metabolites in central carbon pathways (e.g., amino acids, organic acids, phosphorylated sugars, energy currency metabolites), which could not be retained by RPLC. To demonstrate the value of the integrative approach, bacterial cells, human plasma, and cancer cells were analyzed by combined RPLC/HILIC separation coupled to ESI positive/negative MS detection. The combined approach resulted in the observation of metabolites associated with lipid and central carbon metabolism from a single biological extract, using 80% organic solvent (ACN:MeOH:H2O 2:2:1). It enabled the detection of more than 30,000 features from each sample type, with the highest number of uniquely detected features by RPLC in ESI positive mode and by HILIC in ESI negative mode. Therefore, we conclude that when time and sample are limited, the maximum amount of biological information related to lipid and central carbon metabolism can be acquired by combining RPLC ESI positive and HILIC ESI negative mode analysis.

Metabolites, the chemical entities that are transformed during metabolism, provide a functional readout of cellular biochemistry. With emerging technologies in mass spectrometry, thousands of metabolites can now be quantitatively measured from minimal amounts of biological material, which has thereby enabled systems-level analyses. By performing global metabolite profiling, also known as untargeted metabolomics, new discoveries linking cellular pathways to biological mechanism are being revealed and are shaping our understanding of cell biology, physiology and medicine.

•Metabolomic data are information rich, but challenging to interpret biologically.•New software can guide hypothesis generation and direct further experiments.•Technologies for isotope analysis and metabolite imaging drive biological discovery.Liquid chromatography/mass spectrometry-based untargeted metabolomics is now an established experimental approach that is being broadly applied by many laboratories worldwide. Interpreting untargeted metabolomic data, however, remains a challenge and limits the translation of results into biologically relevant conclusions. Here we review emerging technologies that can be applied after untargeted profiling to extend biological interpretation of metabolomic data. These technologies include advances in bioinformatic software that enable identification of isotopes and adducts, comprehensive pathway mapping, deconvolution of MS2 data, and tracking of isotopically labeled compounds. There are also opportunities to gain additional biological insight by complementing the metabolomic analysis of homogenized samples with recently developed technologies for metabolite imaging of intact tissues. To maximize the value of these emerging technologies, a unified workflow is discussed that builds on the traditional untargeted metabolomic pipeline. Particularly when integrated together, the combination of the advances highlighted in this review helps transform lists of masses and fold changes characteristic of untargeted profiling results into structures, absolute concentrations, pathway fluxes, and localization patterns that are typically needed to understand biology.Download high-res image (483KB)Download full-size image

Renewed interest in metabolic research over the last two decades has inspired an explosion of technological developments for studying metabolism. At the forefront of methodological innovation is an approach referred to as “untargeted” or “discovery” metabolomics. The experimental objective of this technique is to comprehensively measure the entire metabolome, which constitutes a largely undefined set of molecules. Given its potential comprehensive coverage, untargeted metabolomics is often the first choice of experiments for investigators pursuing a metabolic research question. It is important to recognize, however, that untargeted metabolomics may not always be the optimal experimental approach. Conventionally, untargeted metabolomics only provides information about relative differences in metabolite pool sizes. Therefore, depending on the specific scientific question at hand, a complementary approach involving stable isotopes (such as metabolic flux analysis) may be better suited to provide biological insights. Unlike untargeted metabolomics, stable-isotope methods can provide information about differences in reaction rates.