Metabolomics meets lipidomics: Assessing the small molecule component of metabolism

Metabolomics, including lipidomics, is emerging as a quantitative biology approach for the assessment of energy flow through metabolism and information flow through metabolic signaling; thus, providing novel insights into metabolism and its regulation, in health, healthy ageing and disease. In this forward‐looking review we provide an overview on the origins of metabolomics, on its role in this postgenomic era of biochemistry and its application to investigate metabolite role and (bio)activity, from model systems to human population studies. We present the challenges inherent to this analytical science, and approaches and modes of analysis that are used to resolve, characterize and measure the infinite chemical diversity contained in the metabolome (including lipidome) of complex biological matrices. In the current outbreak of metabolic diseases such as cardiometabolic disorders, cancer and neurodegenerative diseases, metabolomics appears to be ideally situated for the investigation of disease pathophysiology from a metabolite perspective.

diseases (i.e., cancer, obesity, diabetes, cardiovascular disease and neurodegenerative diseases) from a metabolite perspective, in addition to gene perspective, is providing the necessary information on the metabolic activity that has taken place and thus the phenotype at the molecular level. [7][8][9][10][11][12] This information stored in metabolic signatures or metabolic profiles provides the additional insights into the functional status of a biological system and contributes to understanding of the pathophysiological mechanisms for more efficient diagnosis, treatment and ultimately the disease prevention. [2,13,14,15] Following significant advancements in technology, computing power and bioinformatic solutions, we can today measure not only the metabolites implicated in energy production and storage (i.e., highly abundant nutrients, energy currency metabolites, metabolic by-products, structural lipids and lipid reserves) but also the low abundant metabolites, present in trace amounts, that are responsible for the information flow through chemical signaling. These signaling molecules provide us the additional insights into the metabolic signaling and the regulation of essential biological processes, from cell growth, differentiation and activation to cell proliferation and apoptosis. Several recent metabolomics data-driven studies clearly demonstrate that "the metabolites are perhaps the body's most important signaling molecules" (David Wishart). [15] Following the paramount evidence on how metabolites act as signaling molecules and modulate protein activities, RNA metabolism and gene expression, and most importantly the disease phenotype, the metabolite role and activity presents a gold-mine yet to be explored. [17,18] In this review, we are providing an overview on MS-based technological platforms, as the most versatile systems to resolve and analyze the chemically highly diverse metabolome and lipidome. We discuss the advantages of different methodological approaches and provide a critical opinion on the future developments and requirements.
Finally, we highlight the importance of metabolomics application to reveal metabolite role and activity and provide us new means for the modulation of metabolic processes and health outcomes.

"ONE SIZE DOESN'T FIT ALL"-RESOLVING METABOLITE CHEMICAL DIVERSITY
The metabolome represents a small molecule complement (<1500 Da) of a cell, tissue or a biofluid. Its main constituents comprise polar metabolites (e.g., carbohydrates or sugars, amino acids and their derivatives, short peptides, other carboxylic acids, purines and pyrimidines and their nucleosides and nucleotides) and lipids and lipidlike metabolites, from free fatty acids, acylcarnitines, bile acids and steroids, to more complex glycerolipids, glycerophospholipids, sterol lipids, sphingolipids, oxylipins, etc. (see Figure 1B). These "primary" metabolites are highly conserved across different phyla and species and play essential role for organismal survival as fuels for cellular energetics, building blocks of structural components of cells and (bio)active andsignaling molecules. [15,16] More diverse and specific, exogenous metabolome encompasses the xenobiotics coming from diet or the environment, such as drugs, food additives, pollutants, toxicants, and natural products. The latter ones, also called "secondary" metabolites, are products of specialized metabolism in plants, fungi, microorganisms and animals (e.g., sessile marine invertebrates, insects), serving mainly as chemical defenses with the ecological role in improving the organismal fitness. [19,20] The tremendous chemical diversity and wide concentration ranges (spanning at least 11 orders of magnitude) [21] contained in complex biological matrices represent the aim and the challenge of 'omic scale metabolite analysis. [22] There are no limits on how metabolites can be assembled from the structural point of view and there is no universal technique or even combination of techniques that can be used to assess the entire metabolome, along with the lipidome, present in biofluids, cell and tissue lysates, etc. To resolve the chemical diversity, multiple metabolite extraction protocols, measurement technologies, approaches and modes of analysis are combined to make use of their complementarity and thus expand the coverage of polar metabolome and complex lipidome ( Figure 1A).
The metabolite extraction protocol, depending on the affinity of organic solvent and the reproducibility of the protocol itself, will determine the data quality with respect to the scope of extraction and introduced analytical bias (i.e., due to sample handling, spontaneous oxidation of certain metabolites, etc.). [23,16] The adequate protocol should efficiently quench the metabolism (to arrest the residual enzymatic activity), extract the broadest range of metabolites (of interest) and remove proteins. The metabolite extraction is sample type and analyte dependent although generic methods are used in untargeted assays, not to bias for or against specific classes of analytes. [23] The sample preparation also depends on the measurement technology used a posteriori. While derivatization is usually necessary to prepare the samples for GC-MS analysis (see Table 1), it is only rarely used prior to LC-MS analysis, mainly for the measurement of low abundant and poorly ionizable metabolites, such as oxysterols [24] and phosphoinositides, [25] for example.
Among the technological platforms used to resolve chemical diversity, nuclear magnetic resonance (NMR) spectroscopy and mass spectrometry (MS) constitute two main, widely applied technologies for broad-range metabolite and lipid analysis. [66] While NMR has the advantage of being quantitative, non-destructive and highly reproducible-mainly due to the lack of direct interaction between the sample and the instrument, its important drawbacks are the lack of sensitivity and spectral resolution or overlapping signals making unambiguous metabolite identification and quantification difficult when analyzing complex biological matrices. [27][28][29] NMR is widely used for the analysis of urinary metabolites, present in high concentrations, in the context of large-scale population studies, in view of its high robustness over time. [30] Although less standardized, it can also be efficiently used for the quantitative analysis of highly abundant cellular metabolites in cell lysates and their supernatants. [31] Despite of robustness and quantitative capacity of NMR, its application remains limited when compared to MS-based technologies that made the most significant imprint in metabolomics and lipidomics, mainly due to the most comprehensive breath of coverage, low instrumentation footprint as well as their cost-effectiveness. Major improvements in TA B L E 1 Technological platforms used to resolve polar and lipid metabolite diversity. The most common and successful applications are described  [67][68][69][70] F I G U R E 2 The scope of untargeted and targeted assays. The breadth of coverage of untargeted versus targeted assays is illustrated by the "top of the iceberg" illusion. Although the untargeted assays are unbiased in their approach to detect as many metabolites as possible, in a generic way, without an a priori hypothesis (sample preparation and analysis are performed in a way that should not favor any specific group of metabolites), their coverage is often limited to the most abundant metabolites present in biological samples in mM and μM concentrations (such as building blocks of cell membranes and "fuel" metabolites involved in energy production and storage); thus, constituting the "top of the iceberg". Low abundant, signaling metabolites (present in fM to nM concentrations) often need to be measured using targeted assays to significantly increase the measurement sensitivity and specificity. These assays need to be tailored specifically for metabolites belonging to one class or one pathway of interest sensitivity and specificity of MS-based techniques in the past decade allow us today to measure hundreds to thousands of polar metabolites and/or complex lipids from only minimal amounts of a biological sample (e.g., starting from 5 μL of blood plasma, 1-10 mg of tissue fresh weight, see Figure 1B). It is, however, important to highlight that the dynamic range of the state-of-the-art MS instrumentation remains limited to six orders of magnitude and therefore cannot englobe, in a single analysis, the wide range of concentrations in which small molecules are present in biological samples. For example, specific signaling metabolites are present in blood plasma at low picomolar levels compared to glucose and cholesterol at millimolar levels (see Figure 2).

Mass spectrometry (MS)-based technologies, approaches, and applications
How do we resolve chemical diversity and identify metabolites using mass spectrometry?
Mass spectrometer measures the mass-to-charge ratio (m/z) of a molecular ion. The metabolites must be ionized (i.e., charged) for their mass to be recorded and thus the ions need to be generated in the ion source of the mass spectrometer prior to the mass measurement by the mass analyzer ( Figure 1A). Several different types of mass analyzers can be used for metabolite analysis, depending on their resolution, accuracy, scan speed and dynamic range. Most commonly used analyzers comprise low-resolution triple quadrupoles (QqQ) and highresolution hybrid quadrupoles time-of-flight (Q-TOF) and quadrupole Orbitraps ( Figure 1A). [71,72] Mass spectrometer with high resolving power (Q-TOF and Q-Orbitrap) will allow for accurate m/z measurement, with errors <1 ppm. Although higher MS resolving power facilitates the resolution of chemical diversity (e.g., to distinguish between isobars-or metabolites of similar molecular mass), many , it is not sufficient to assign the unique metabolite identity. [73] Yet, the majority of metabolites have the highly specific fragmentation patterns that can be generated by collision-induced dissociation (CID). They represent the metabolite MS/MS signatures composed of product ions, which constitute the essential information to validate metabolite identity with high level of confidence. [74,75] The high-resolution instruments (HRMS) with high scanning speeds (from 20 Hz Orbitrap to 100 Hz Q-TOF) allow for subsequent collection of high amount and quality of MS (precursor ion) and MS/MS (product ion) spectra, thereby facilitating peak definition and integration for metabolite quantification, as well as metabolite identification using MS/MS data. [71,76] However, the current scanning speeds of HRMS instruments are still insufficient for a good definition of MS/MS peaks and the quantification at the MS/MS level. [77] The low-resolution tandem mass spectrometers (i.e., QqQ) remain the gold standard for this highly selective quantification at the MS/MS level with significantly lower signal-to-noise ratio and enhanced sensitivity.
The experimentally acquired MS/MS data, on pure standards or by data-dependent (DDA) or independent assays (DIA) on the sample extracts, are recorded and stored in a multitude of spectral databases and libraries and used for MS/MS matching. The content of spectral libraries maintained by the community is publically available and downloadable (e.g., MassBank [78] ) while the most comprehensive and well curated libraries remain proprietary and searchable only online (e.g., METLIN,79] mzCloud). It is worth emphasizing that the fragmentation patterns of polar metabolites are difficult to predict due to extremely diverse and species-specific physicochemical properties, whereas lipids have consistent and class-specific predictive fragmentation patterns. [80] Beyond mass, the additional chemical information can be acquired using hyphenated techniques, such as retention time (RT) and ion mobility (expressed as cross collisional section or CCS value) to facilitate the metabolite identification (see section below).

Direct injection-based mass spectrometry strategies
Most of the MS-based technologies are endowed with high-throughput capacity, thus facilitating the analysis of large batches of samples, up to thousands in human population studies. Direct injection strategies, such as flow injection analysis (FIA) are the most high-throughput, and of particular interest for in real time metabolite profiling; [37] however, they suffer from matrix effects and thus the poor reproducibility, as well as lack of specificity for metabolite identification. Contrary to its limited application for polar metabolite analysis, the direct infusion-based shotgun lipidomics (SL) is widely used in lipidomics community and has recently evolved towards multidimensional mass spectrometry-shotgun lipidomics (MDMS-SL), applying multiple acquisitions in full and MS/MS scan modes, for more robust lipid quantification and identification. [71,82] As such it has become relatively low-throughput and still remains subject to ion suppression, thus limiting the analysis to highly abundant lipid species.

Surface ionization techniques: mass spectrometry imaging
Mass spectrometry imaging applies direct surface ionization to gain insights into spatial distribution of endogenous lipids, polar metabolites, drugs and their metabolites, across profiled tissue sections. [83] Commonly used techniques for in situ tissue imaging include matrixassisted laser desorption ionization or MALDI-MS, desorption electrospray ionization or DESI-MS, matrix-free nanostructure initiated MS or NIMS and secondary ion -MS or SIMS. [84] While relying on differ-ent ionization principles, all these techniques, through ion detection as a function of position, yield spatially resolved chemical maps of tissue sections. The generated maps have been particularly useful for the metabolic characterization of heterogeneous tissues, such as tumor tissue. [85] MALDI is the most widely used tissue imaging technique with broad coverage of polar and lipid metabolites; however, the presence of organic matrices can interfere with the small molecule detection. [86] DESI is an ambient ionization technique that allows for the direct analysis of unaltered tissue sections with the improved signal in low mass range (<500 m/z) and has an important application in spatial mapping of tumor metabolism. [87] Also in the context of tumor tissue analysis, but with the foremost aim to facilitate intraoperative tumor tissue identification, as an alternative to frozen section histology, rapid evaporative ionization MS or REIMS has been recently developed for in vivo tissue analysis. It was straight off applied in clinical practice during surgical interventions (i.e., iKnife) for supervised cancerous tissue removal. [88,89] Although MS imaging offers this lateral resolution on local distribution of metabolites within tissue (vs. the analysis of a homogenized tissue extracts as a whole) and its spatial resolution was improved remarkably to allow for imaging at single-cell and even subcellular level, [90][91][92] several challenges, mainly regarding the low ionization yields of still broad-range of metabolites, identification and quantification issues remain to be addressed prior to its wider application in biomedical and clinical research.

Hyphenated MS techniques: why separation matters
Despite the resolving power of mass spectrometry, and notably highresolution mass analyzers (HRMS) to distinguish between metabolites based on m/z ratio, the structural diversity and dynamic concentration range of metabolites present in complex biological samples cannot be resolved uniquely by mass spectrometry. Therefore, the separation techniques such as gas chromatography (GC), liquid chromatography (LC) and capillary electrophoresis (CE) are commonly coupled to mass spectrometry to improve the sensitivity, selectivity and dynamic range of metabolite measurement ( Figure 1A, Table 1). In addition, a supercritical fluid chromatography (SFC) has recently received a lot of attention for lipid analysis, thanks to the development of robust commercial solutions and high separation efficiency and short analysis time compared to commonly used LC-MS. [93,94] In general, the specified separation techniques are powerful in reducing the matrix effects to minimize the ion-suppression and therefore maximize the MS signal intensity. They will also maximize the measurement specificity by providing the complementary data, such as retention time, for metabolite identification.

Gas Chromatography coupled to mass spectrometry (GC-MS)
Gas chromatography uses a gaseous phase to transport volatile metabolites that are separated along temperature gradient in very long columns (up to 60 m) depending on strength of chemical interaction with column's stationary phase. GC-MS is well-suited for the analysis of polar metabolites (e.g., sugars, amino acids and other organic acids, short and long chain fatty acids, cholesterol and its derivatives) that are made volatile by chemical derivatization. [95,96] In addition to improving the compound volatility, the derivatization will also reduce their polarity and increase their thermal stability, to facilitate the GC-MS analysis. [49] Gas chromatography provides high separation capacity (peak width < 5s) and is therefore fairly efficient for

Liquid Chromatography coupled to mass spectrometry (LC-MS)
Liquid chromatography coupled to mass spectrometry is the most versatile and commonly used separation technique in MS based metabolomics and lipidomics, for both targeted and untargeted assays.
It uses liquid phase to carry analytes for the separation along solvent gradient in columns (generally from 5 to 15 cm long) as a function of chemical interaction between the column's stationary phase and the mobile phase or solvent. The introduction of ultra-high-performance liquid chromatography (UHPLC) in the last decade, using sub-2 μm particle size columns and poroshell particles, [97] has improved the resolution and sensitivity of LC-MS analysis and significantly increased the number of metabolites detected in complex biological matrices.
In addition, the UHPLC methods allow for the operation at high flow rates (>400 μL min −1 ) with highly reproducible retention times hence increasing the throughput (i.e., number of samples analyzed per batch). [94,98] In general, UHPLC represents a good compromise with high (although still limiting) mass spectral acquisition (scanning) rates.
It significantly maximizes the measurement specificity by separation of isobars and isomers, and by providing the retention time identifiers.
Most commonly used stationary phases can be classified in the following three groups, depending on their chemistry and separation mechanisms: (i) reversed phase (RP), (ii) ion (cation or anion) exchange and (iii) hydrophilic interaction liquid chromatography (HILIC). Metabolite interaction with these different stationary phases and their retention is based on their hydrophobicity, positive and negative charges and hydrophilicity. While RPLC has the broadest applicability due to its robustness, HILIC separation mode is becoming increasingly popular as complementary method to RPLC for the analyses of polar and highly hydrophilic ionic compounds. [57][58][59] HILIC stationary phases are classified based on the functional groups (e.g., diol, amide, amino) present on their surface and their charge. They can be divided in unbounded and bonded phases, and the latter ones are classified as neutral and charged or zwitterionic phases. [99] The LC system is generally coupled to the MS system using electro-

Ion mobility coupled to mass spectrometry (IMS)
Even though the described separation techniques (i.e., LC and GC), maximize the sensitivity and specificity of MS analysis, they still cannot resolve the high chemical diversity of structural isoforms (e.g., sugars, complex lipids) present in complex biological matrices. The ion mobility spectrometry (IMS) has recently emerged as an additional dimension of separation offering an orthogonal resolving power to the existing analytical setup that further facilitates the metabolite identification and expands the metabolite coverage. [100][101][102][103] The technique is based on the separation of ions according to their shape, conformation and size, allowing for the measurement of the collisional cross section (CCS) values for each compound, depending on the IMS design (e.g., drift tube ion mobility MS or DTIMS and travelling wave ion mobility MS or TWIMS). [103] The CCS values can be obtained with high reproducibility, independently of analytical conditions used, thus providing a valuable additional chemical information to improve the annotation and identification of polar metabolites, lipids, glycans and proteins.
Besides, the separation of ions according to their size and conformation prior to MS/MS data acquisition, will also improve the quality of MS/MS data with respect to spectral clarity and fragmentation specificity. [23] In the past few years the IMS instrumentation, using different operation modes, has been introduced to the market, and the acquired CCS values are recorded in databases which growth will undoubtedly contribute to resolving the bottleneck of metabolite identification.

Targeted or untargeted, that is the question
To cope with metabolite chemical diversity and wide concentration range in which metabolites are present in complex biological matrices, two main approaches are used for metabolomic and lipidomic assays, the targeted approach that typically focuses on one pathway or class of metabolites of interest, and untargeted approach that tends to detect as many metabolites as possible without an a priori hypothesis ( Figure 2). Targeted approach has been employed in clinical chemistry since its beginnings in 1950s, dominated by GC-MS methodology, while the untargeted approach has evolved relatively recently (first untargeted experiment was performed in 1970s by Pauling and Robinson [107,108] ) following the advancement in MS technologies (i.e., enhancement in sensitivity with the introduction of ESI) and the systems biology concepts to complement the data acquired by other 'omic technologies. Both of these approaches, targeted and untargeted, have high-throughput phenotyping capacity and can be applied from model systems to human population studies.

Untargeted profiling without an a priori hypothesis
The concept of global untargeted analysis has evolved with the aim to assess the levels of the broadest range possible of polar metabolites and complex lipids to draw as complete picture possible of the metabolism as a whole. Metabolic diseases, such as type 2 diabetes and cancer are great examples why it is important to have an integrated look at the metabolism as a whole, including multiple pathways, rather than focusing on specific pathway out of its metabolic network context. For both of these disease states it is well-known that the origin of disease is not directly associated with abnormal glucose metabolism, which should be regarded as a consequence of deregulated lipid metabolism. [109,110] The untargeted profiling is a powerful discovery approach that may highlight yet unknown changes in (un)known metabolic pathways, associated with the investigated phenotype due to disease, drug treatment or environmental factors. It is considered "unbiased" because without an a priori hypothesis although the metabolite and/or lipid cov- extensively reviewed elsewhere. [23,111] The untargeted experiments are semi-quantitative, based on group comparison to evaluate the relative abundance (i.e., increase or decrease) of each metabolite (feature) to the average abundance measured in the control group. The untargeted experiment can also be coupled with the targeted quantification of specific class of relatively abundant metabolites, whose quantification can be performed in a full scan mode. [60] In this way the quantitative data can be generated while benefiting from the retrospective exploration of full scan data on other detected metabolites. While untargeted approach can reveal the unanticipated metabolic alterations, the obtained results should be considered as hypothetical and must be validated by targeted quantification that will allow for the normalization of the inherent analytical variability (related to matrix effects).

From single pathway to broad-coverage multiple pathway targeted assays
The relative lack of sensitivity of full scan instruments compared to tandem mass spectrometry (MS/MS), together with the bottleneck of (true) metabolite identification that it is vital for data interpretation have motivated the development of broad-coverage targeted assays to screen for small molecule intermediates in multiple pathways, including those present at low concentration levels in biological samples [112,113] (see Figure 2). These multiple pathway targeted assays with a comprehensive coverage of several hundred metabolites have therefore emerged as a surrogate to untargeted assays, taking the advantage of high sensitivity of triple quadrupole instruments operating in multiple reaction monitoring (MRM) or scheduled/dynamic MRM (sMRM or dMRM) mode. New generation of triple quadrupole instruments is endowed with high scan speeds and the capability of fast polarity switching while maintaining an excellent sensitivity. In the last decade, several high-throughput targeted methodologies have been developed to screen for hundreds of biologically relevant metabolites involved in central carbon metabolism, comprising glycolysis, TCA cycle, pentose phosphate pathway, oxidative phosphorylation, purine and pyrimidine metabolism, beta-oxidation, etc., if possible in a single run. [22,61,[114][115][116][117] These high-coverage assays also serve to bridge the gap between two extremes, the untargeted profiling of "as many metabolites as possible" and the absolute quantification of a small number of metabolites belonging to one specific chemical class.
The main drawback of such workflow is the time-consuming and costly method development primarily related to standard purchase and characterization, and its "targeted" character or data acquisition being limited to the selected set of known metabolites. Contrary to the untargeted analysis that demands a significant amount of work at the backend for data quality assessment, peak annotation and metabolite iden- have also emerged to further enhance the accuracy and precision of lipid quantification and offer even greater lipidome coverage. [119][120][121][122] Relative or absolute quantification?
Relative quantification of detected metabolites is based on group comparison or evaluation of relative metabolite abundance to control group using the peak areas or ion abundances expressed in arbitrary units (A.U.). In these semi-quantitative assays without an internal reference (due to difficulty to cover with internal standards all chemically diverse metabolites that behave differently in the same analytical conditions) the use of pooled samples as quality controls (QCs) is essential to monitor and correct for the analytical variability during the sample analysis. Pooled QC samples are used for system conditioning and to correct for within-and between-batch variation of signal (i.e., signal intensity drift). Guidelines and considerations applied for quality control and batch correction in semi-quantitative assays have been covered in many more details in a review by Ivanisevic and Want [23] and Broadhurst et al. [123] The results of these untargeted semi-quantitative assays lack the quantitative accuracy (i.e., the peak area in not proportional to metabolite concentration) and thus cannot be compared across different studies, laboratories, etc. Hence, we strongly recommend that the data generated by relative quantification approaches (untargeted and multiple pathway targeted profiling in the absence of internal standards) must be validated by an absolute targeted quantification approach as detailed below.
In quantitative mass spectrometry or absolute quantification approaches the calibration curves and stable isotope-labeled standards (deuterated, 13

Isotopic profiling or stable isotope-assisted (UN)targeted metabolite analysis
Untargeted and targeted metabolomic assays allow for the relative and absolute measurement of intracellular and extracellular metabolite levels. However, these measurements do not provide the information on relative pathway activities depending on investigated conditions.
The increase in level of specific metabolite can be due to the enhanced activity of metabolite producing enzymes or the decreased activity of metabolite consuming enzymes. The inability of standard untargeted and targeted assays to provide this information on pathway utilization in different conditions has led the development of stable F I G U R E 3 Targeted quantification assay. Internal standard spike or the addition of stable isotope-labeled standard (IS) mixture to the sample is necessary to translate the signal abundance (i.e., peak area) to metabolite concentration. Signal abundance is not an equivalent (or proportional) to metabolite concentration because it depends on the ionization efficiency of each metabolite. Metabolites producing higher signal intensities ionize well but do not necessarily represent the most abundant metabolites. The metabolite concentration will be calculated based on the calibration curve by reporting the area of detected endogenous metabolite peak to the area of its internal standard (IS) spiked at known concentration. This IS spike allows for the correction of analytical variability during the sample preparation (i.e., variation due to metabolism quenching and potential metabolite transformations) and analysis (i.e., due to matrix effects), and is mandatory for an accurate and precise quantification of metabolite concentrations and crosslaboratory comparisons F I G U R E 4 Isotopic profiling or stable isotope-assisted analysis. Isotopic profiling is applied to track the fate of labelled nutrient and understand the changes in its utilization in different conditions (e.g., WT vs. KO, CTRL vs. treated)-to identify the pathways that are actively used for labelled nutrient metabolism. The experiment can be performed in vitro (by the addition of labelled metabolite to cell culture media) or in vivo (by bolus injection or drinking water to mice) and is based on group comparison. Data can be acquired in several different analytical settings, in untargeted and in targeted fashion. The identification of 13 C or 15 N enriched metabolites (mainly based on carbon or nitrogen transfer) is based on the distribution calculation of isotopologue abundances (i.e., one, two, three, four, etc. carbon-labeled isotopologues), corrected for natural isotope abundances. In the example showed on the figure, the 13 C-labeled glucose was metabolized via pyruvate and TCA cycle and the succinate enriched in 13  analysis to determine metabolite production and consumption rates.
The latter requires computational data modeling based on known stoichiometry of the metabolic network. The review by Buescher et al. [124] describes the difference between qualitative 13 C tracer analysis and quantitative flux analysis with respect to experimental design and applications. Both types of these stable isotope-assisted analysis are essential to understand the mechanisms of metabolic regulations, like for example in reprogrammed cancer metabolism. [125] In these experiments a biological system is fed with one or more metabolic substrates or nutrients labeled with stable heavy isotopes F I G U R E 5 From model systems to human population studies. Due to its high-throughput and phenotyping capacity, different metabolomic and lipidomic approaches, from untargeted screening to targeted quantification and isotopic profiling, can be applied from different model systems, demanding lower number of independent biological replicates, to clinical research studies, demanding high number of participants due to high human inter-individual variability and multiple confounding factors . Estimated number of biological replicates necessary for the appropriate experimental design-to yield relevant conclusions with sufficient statistical power, [23] is indicated below each studied system (with the assumption of same genotype in model organisms). It is mandatory to validate the results of untargeted profiling by targeted quantification, particularly in human population studies, also to allow for the cross-study comparability ( 13 C, 15 N, 2 H, 18 O). These labeled nutrients can be metabolized by the studied system through different metabolic pathways (depending on different conditions of growth, genotypes, etc.) and the isotope labels spread in a time, reaction rate and pathway dependent manner will generate characteristic labelling patterns that can be measured and identified using MS or NMR techniques. In MS-based isotopic profiling the propagation of a label from the tracer to a given metabolite or label (e.g., 13 C) enrichment is quantified as the fractional abundance of isotopologues (relative to the total metabolite abundance). Isotopologues are molecules that have the same molecular formula and structure but differ in their isotopic composition through the substitution of one or more atoms with different isotopes (e.g., H 2 O/D 2 O, Glucose/ 13 C 6 -Glucose, Glutamine/ 13 C 5 -Glutamine). The measured isotopologue distribution needs to be corrected for the abundances of naturally occurring isotopes (that depend on the molecular composition). [126] The appropriate time-series experimental design is crucial to obtain solid conclusions from an isotopic tracer experiment. The isotopic profiling applications, tracers to use, metabolite readouts and biological data interpretation are extensively covered in two reviews, by Jang et al. on isotope tracing and metabolomics, [127] and by Belcells et al.
on metabolic flux analysis. [128] A very important aspect in isotopic tracer experiments is the choice of a tracer or labeled substrate. Generally, 13

CONCLUSIONS AND PERSPECTIVES
Metabolomics, including lipidomics, is focusing the metabolism research back to metabolites, in addition to genes and their function that were in the spotlight during the past era of biochemical genetics.

CONFLICT OF INTEREST
The authors declare no conflict of interest.

DATA AVAILABILITY STATEMENT
Data sharing is not applicable to this article as no new data were created or analyzed in this study.