Global strategies to integrate the proteome and metabolome Alan Saghatelian and Benjamin F Cravatt A fundamental goal of proteomics is to assign physiological nuria was the result of a build-up of homogentisic acid, functions to all proteins encoded by eukaryotic and prokaryotic which he attributed to an ‘inborn error of metabolism’ [4]. genomes. Of the many activities performed by proteins, the Of course, this discovery was predicated on having a chemical transformations catalyzed by enzymes form the basis straightforward bioassay (the color change associated with for most, if not all, metabolic and signaling pathways. oxidation of homogentisic acid), a methodological advan- Elucidation of these pathways and their integration into larger tage often not afforded by other changes in metabolism. cellular networks require new strategies to rapidly and Indeed, in the pursuit of global strategies to profile the systematically identify physiological substrates of enzymes. metabolome, contemporary researchers are confronted Here, we review emerging technologies that aim to assign with a daunting set of challenges that, in many ways, endogenous biochemical functions to enzymes by profiling the exceed those confronted by genomics or proteomics. For metabolome. example, unlike transcripts and proteins, metabolites share no direct link with the genetic code and are instead Addresses products of the concerted action of many networks of The Skaggs Institute for Chemical Biology and Departments of Cell enzymatic reactions in cells and tissues. Similarly, meta- Biology and Chemistry, The Scripps Research Institute, 10550 bolites are not linear polymers composed of a defined set North Torrey Pines Road, La Jolla, CA 92037, USA of monomeric units, but rather constitute a structurally Corresponding author: Cravatt, BF ([email protected]) diverse collection of molecules with widely varied che- mical and physical properties. As such, metabolites do not readily lend themselves to universal methods for analysis Current Opinion in Chemical Biology 2005, 9:62–68 and characterization. These distinguishing features make This review comes from a themed issue on the metabolome a unique portion of biomolecular space Proteomics and genomics that requires advanced methods for its analysis. Here, we Edited by Benjamin F Cravatt and Thomas Kodadek review contemporary approaches for global metabolite profiling and highlight recent applications. We focus in Available online 8th January 2005 particular on efforts to use metabolomic data to elucidate 1367-5931/$ – see front matter the endogenous functions of enzymes. Collectively, these # 2005 Elsevier Ltd. All rights reserved. studies underscore that the metabolome, by being both sensitive to changes in protein activity and, at the same DOI 10.1016/j.cbpa.2004.12.004 time, contributive to higher-order cellular phenotypes, offers an information-rich conduit between the proteome and the physiological processes it regulates. Introduction The metabolome, which refers to the full complement of The general analytical tools of metabolomics metabolites within a cell, tissue, or organism [1], encom- To date, metabolite profiling has primarily been per- passes a diverse array of molecular chemotypes, including formed using either NMR or MS techniques. NMR peptides, carbohydrates, lipids, nucleosides, and catabolic offers a rapid and non-invasive method to comparatively products of exogenous compounds. As a rich source of characterize metabolite expression patterns in vitro both signaling and structural molecules, the metabolome [5,6] or in vivo [7,8]. However, NMR exhibits limited and, in particular, its composition and regulation, are sensitivity and resolution and is therefore capable of understandably subjects of intense research interest. Akin detecting only the most abundant metabolites in com- to its more mature counterparts, genomics and proteom- plex samples. To increase the breadth and depth of ics, the burgeoning field of metabolomics aims to develop metabolite profiling, investigators have turned to MS and apply strategies for the global analysis of metabolites methods [9,10]. MS analyses can be performed either in cells, tissues and fluids [2,3]. By integrating these ‘small directly or in combination with GC or LC separation molecule’ profiles with transcript and protein expression/ steps. ‘Direct-inject’ MS experiments have the advan- activity patterns, researchers hope to discern the contri- tage of being fast (minutes per sample) [9], but, like bution that specific metabolites and metabolic pathways NMR, suffer from limited resolving power. By contrast, make to health and disease. GC- and LC-MS experiments are considerably slower (1–2 h/sample), but permit more than 1000 metabolites to An early example of the value of metabolite analysis can be analyzed per run, and thus sacrifice speed in exchange be attributed to Garrod, who discovered that the darken- for substantial increases in sensitivity and ‘metabolome ing of urine, upon standing, from patients with alkapto- coverage’. Current Opinion in Chemical Biology 2005, 9:62–68 www.sciencedirect.com Global strategies to integrate the proteome and metabolome Saghatelian and Cravatt 63 Table 1 Analytical tools of metabolomics. Author Model system Enzyme(s) disrupted Analysis method Metabolites identified (relationship to (targeted/untargeted) enzyme) Raamsdonk et al. [6] Yeast Multiple NMR (untargeted) Unknown Allen et al. [9] Yeast Multiple ESI-MS (untargeted) Unknown Rohde et al. [18] Arabidopsis PAL1 and PAL2 LC-UV-MS (targeted) Flavanols and amino acids (known substrates and novel 28 Metabolites) Wu, et al. [19] Mouse mBCAT Tandem MS (targeted) Amino acids (known substrates) Saghatelian, et al. [21] Mouse FAAH LC-MS (untargeted) Lipids (known and novel substrates) Each of the profiling methods described above can be the (patho)physiological functions of enzymes. By con- conducted in either a targeted or untargeted mode trast, equivalent experimental strategies for the global (Table 1). Targeted applications focus on the character- characterization of endogenous biochemical functions of ization of a specific class of metabolites, often exploiting enzymes do not yet exist, and, as a consequence, the their unique chemical properties — NMR-active atoms natural substrates for many enzymes remain unknown. (e.g. 31P), molecular masses and fragmentation patterns, etc. — to increase the sensitivity of detection and quan- Substrate selectivities of enzymes are typically deter- titation. By contrast, untargeted applications seek to mined in vitro using purified preparations of protein. broadly profile the metabolome by establishing condi- However, it is often difficult to ascertain the physiological tions for the concurrent analysis of as many metabolites as significance of such ‘test tube’ biochemistry experiments possible. As we highlight in the following sections, tar- for several reasons. First, many enzymes function as parts geted and untargeted methods offer complementary ways of large protein complexes and networks in vivo [16] that to interrogate the metabolome and assess the biochemical may be challenging to reconstruct or model in vitro. repercussions of specific genomic, proteomic and/or phy- Second, enzymes are often regulated by post-translational siologic perturbations. events in vivo (e.g. phosphorylation, proteolytic proces- sing) [17], which may alter substrate recognition and Biological applications of global metabolite catalysis. Finally, the characterization of enzyme- profiling substrate relationships in vitro is inherently limited by As with genomics and proteomics, metabolomic studies our current knowledge of cell metabolism and, therefore, typically prove most informative when performed in a ill-suited for the discovery of novel natural products comparative mode. Thus, while a static portrait of a cell regulated by enzymes in vivo. Metabolomics offers a or tissue metabolome may offer little in the way of potentially powerful strategy to address these limitations. hypothesis-testing or hypothesis-generating data, a com- Indeed, as highlighted below, several complementary parison of the metabolomes of cells or tissues that differ in analytical methods have recently been introduced for specific biological properties is likely to be much more both targeted and untargeted metabolite profiling that, enlightening. A significant portion of metabolomic by providing access to a portion of biomolecular space research is focused on generating small-molecule portraits (the metabolome) that is inaccessible to genomics and that distinguish health and disease (e.g. normal cell versus proteomics, enable the assignment of endogenous bio- cancer cell). These efforts, which hold great promise for chemical functions to a broad range of enzymes. the discovery of new biomarkers for the diagnosis of human pathologies, have been the subject of several Assignment of enzyme function by targeted recent reviews [2,3,11,12,13] and are not discussed metabolite profiling further here. Instead, we focus on an equally, if not more Two recent studies highlight the power of targeted enticing application of metabolomics — the assignment metabolite profiling for elucidating enzyme function in of endogenous functions to enzymes. vivo. Rohde and colleagues set out to distinguish the endogenous activities of phenylalanine ammonia lyase Assignment of enzyme function by global (PAL) enzymes in Arabidopsis thaliana [18], which par- metabolite profiling ticipate in the biosynthesis of the phenylpropanoid