Epigenetics and Metabolomics

Epigenetics and Metabolomics, a new volume in the Translational Epigenetics series, offers a synthesized discussion of epigenetic control of metabolic activity, and systems-based approaches for better understanding these mechanisms. Over a dozen chapter authors provide an overview of epigenetics in translational medicine and metabolomics techniques, followed by analyses of epigenetic and metabolomic linkage mechanisms likely to result in effective identification of disease biomarkers, as well as new therapies targeting the removal of the inappropriate epigenetic alterations. Epigenetic interventions in cancer, brain damage, and neuroendocrine disease, among other disorders, are discussed in-depth, with an emphasis on exploring next steps for clinical translation and personalized healthcare.

• Offers a synthesized discussion of epigenetic regulation of metabolic activity and systems-based approaches to power new research
• Discusses epigenetic control of metabolic pathways and possible therapeutic targets for cancer, neurodegenerative, and neuroendocrine diseases, among others
• Provides guidance in epigenomics and metabolomic research methodology

• Language: English
• Publisher: Academic Press
• Release date: Aug 25, 2021
• ISBN: 9780323856539

Book preview

Preface

Paban K. Agrawala

Poonam Rana

In recent years, there is a tremendous increase in integrated systems biology toward the understanding of disease processes and personalized medicine approach. Epigenetic and metabolomic correlates are novel conception among the available techniques to comprehend the molecular mechanism, cell signaling mechanisms of the disease. Now, it is also one of the most promising strategies toward personalized medicine under the systems biology approach. This book entitled Epigenetics and Metabolomics under Translational Epigenetics Series is intended to focus on the linkage of epigenetics and metabolic dysregulation in diseases and other conditions with an aim to update on novel molecular targets whose intervention may lead to improvement in disease treatment.

Epigenetics is known to govern the phenotypic outcome of an organism; however, the metabolome is known to be closest to the phenotype. Metabolic rewiring in some diseases might have a fundamental role in the epigenetic regulation of gene expression. Likewise, epigenetic dysregulation may mediate to alter the expression of genes involved in cellular metabolism. Epigenetic and metabolic alterations are highly correlated. A thorough understanding of mechanisms of epigenetic and metabolomic linkage will likely result in the development of new effective therapeutic options targeted to the removal of the inappropriate epigenetic alterations and further could unravel novel molecular targets, whose intervention may lead to improvements in cancer or other disease treatment. A thorough understanding of mechanisms of epigenetic and metabolomic linkage will likely result in the development of new effective therapeutic options. The aim of this book is to provide the reader with the clinical and experimental evidence for epigenetic and metabolic phenotype correlates in disease processes and different stressful environmental conditions.

This book begins with an overview of metabolomics and epigenetics, and the techniques which have recently been used for data generation and analysis. Statistical modeling, data visualization, and efficient databases, which are all imperative for interpreting the complex data sets have been included in the chapters. Integration and interplay of epigenetics and metabolic regulations in different diseases, developmental processes, and stressful conditions have been elaborated in the next few chapters. Following these chapters, the contributions relate to applications of epigenetics in different disease processes and clinical conditions. There are also chapters on the concept of clinical phenomics and precision wellness, a translational approach of omics. The last chapter covers certain epigenetic aspects in plant biology making it different from the earlier and thus expanding the base of readers. We hope that this volume will be essential reading for readers, from graduate students to active researchers, who are in pursuit of understanding the mechanistic insights of diseased processes through integrated systems biology.

We are grateful to all authors who could find time from their busy schedules and agreed to write chapters for this book and thank them for their efforts. We are obliged to Peter Linsley for giving us an opportunity to contribute to the Translational Epigenetics series through this book. Finally, we would also like to thank the team at Elsevier and its associates (especially Megan, who handled the difficult stages of commissioning and author follow-up with great efficiency) who all contributed to bringing this book to publication.

Chapter 1: Metabolomics techniques: A brief update

Ritu Tyagi; Pawan Kumar; Uma Sharma    Department of NMR & MRI Facility, All India Institute of Medical Sciences, New Delhi, India

Abstract

Diagnosing, understanding, and elucidating the pathophysiological mechanism of any disease is the prime and the principal step for clinical research. The ever-evolving omic technology metabolomics is a dynamic and emerging field, downstream to genomics, transcriptomics, and proteomics that provide a comprehensive understanding of the metabolic phenotypes of biological systems. Metabolomics systematically study the metabolome of the cells, biofluids, tissues, or organisms so as to detect and quantify changes in low molecular weight exogenous and endogenous metabolites associated with disease processes using high-throughput analytical techniques. Due to the complexity of the metabolome and the broad range of physiochemical properties of the metabolites, more than one analytical technique is required. The most common analytical platforms employed are spectroscopic that is, nuclear magnetic resonance (NMR) and spectrometric that is, mass spectrometry (MS) coupled with separation techniques (liquid chromatography (LC), gas chromatography (GC), supercritical fluid chromatography (SFC)). These techniques enable extensive data generation and the advanced chemometric analysis provides information about the diverse group of metabolites. This information provides a deeper understanding of abnormal metabolic pathways associated with the disease processes and is also useful for identification of biomarkers. Several factors such as sample matrix, the properties and concentration of the metabolites, and the amount of sample determine the choice for a given technique. In this chapter, we have provided an update of the most commonly used analytical techniques, the statistical and bioinformatics tools used for analysis of the generated data and also few applications of metabolomics in human diseases.

Keywords

Metabolomics; NMR spectroscopy; Mass spectrometry; Chemometrics; Analytical techniques; GC; MS; LC-MS

Acknowledgments

The authors would like to acknowledge the funding (CRG/2019/002709) from Science and Engineering Research Board (SERB), Government of India.

Introduction

Metabolomics is a holistic approach towards the identification and discovery of diagnostic and prognostic biomarkers of a disease and also helps in understanding the complex mechanism of development of disease.¹ Metabolomics is the comprehensive study of metabolome that identify, quantify, and analyze both endogenous and exogenous metabolites in a high-throughput manner in a biological sample. The metabolome represents the full complement of low-molecular weight compounds, present in cell, tissue, or biological fluids that play vital roles in various metabolic pathways and cellular processes.² Since the biochemical changes precede anatomical changes; hence, these biomarkers may have the potential to predict a disease condition even in asymptomatic stage. This allows early and effective treatment of disease thus reducing complications and mortality rates.³ The metabolomics with its emerging role in biomarker discovery will help in fundamental understanding of disease pathophysiology and may also have an impetus on novel treatment approaches.⁴, ⁵ In addition to its applications in early disease detection and progression such as cancer,⁶ neurological diseases,⁷, ⁸ and diabetes,⁹ it has varied research domains such as toxicology,¹⁰ quality control (QC) of herbal extracts,¹¹ food and nutrition,¹² and environmental analysis.¹⁰

Metabolomic analysis is performed primarily using two approaches: (i) untargeted and (ii) targeted approach. The untargeted approach is regarded as unbiased because it is not limited to any specific metabolite and focuses on the comprehensive metabolic profiling of biological samples so as to classify the metabolic phenotypes. Untargeted metabolomics approach is hypothesis-generating and useful for discovering new biomarkers. Targeted metabolomics focuses on quantitative analysis of identified metabolites or specific chemical class involved in specific metabolic pathways and is therefore known as biased approach. An outline of the major steps involved in a metabolomic approach is shown in Fig. 1.1.

Fig. 1.1

Fig. 1.1 Workflow of the major steps involved in a metabolomics approach. No permission required.

The metabolomic analysis can be performed in various types of complex biological matrices including blood, saliva, sweat, cerebrospinal fluid, seminal fluid, cell extracts, feces as well as on herbal extracts.⁷ In order to have rapid and wide coverage of various compounds present in the biological fluids, appropriate analytical techniques are required. Obtaining a full spectrum of metabolome is indeed challenging due to the varied physiochemical properties and dynamic range of concentrations the metabolites present in each biofluids or tissue. In order to compensate this, several techniques are sometimes required to expand metabolite coverage.

To generate a metabolomic profile, the two main analytical platforms commonly used are: nuclear magnetic resonance (NMR) spectroscopy and spectrometric techniques like mass spectrometry (MS). The MS-based analysis requires separation of compounds using techniques such as high-performance liquid chromatography (HPLC) and gas chromatography (GC). Both of these platforms have their inherent advantages and disadvantages. Using NMR spectroscopy as an analytical tool for metabolomics has several advantages.¹³ NMR requires minimal sample preparation, does not require derivatization of compounds like MS, and allows the rapid analysis of individual samples. Another major advantage of NMR is that it provides simultaneous measurement of large number of compounds.¹⁴, ¹⁵ The drawback of NMR is its poor sensitivity compared to MS. However, with the development of high field magnets operating at ¹H resonance frequency up to 1.0 GHz and cryogenic probes technology, the sensitivity of NMR experiments has been markedly improved.¹⁶–¹⁸ Special probes allow detection of metabolites which are picomole in concentration.¹⁹ Table 1.1 provides the comparison of these two techniques in the analysis of metabolome. In the following section, we have briefly described the theory and major experimental aspects of these two techniques.

Table 1.1

NMR spectroscopy

The magnetic properties of the atomic nuclei as a result of their charge and spin form the basis of NMR spectroscopy. It is based on the absorption of radiofrequency (RF) pulse (energy) and its re-emission by nuclei in the presence of external applied magnetic field (B0).¹⁵ NMR spectroscopy has two important variants; solution state NMR spectroscopy, wherein samples are studied in liquid state and solid state NMR spectroscopy, which is used to study intact tissue specimens. Additionally, NMR spectroscopy is extensively used for structural elucidation of various types of drug molecules, nucleic acids, proteins, and interactions of molecules.²⁰

A wide range of experiments like one dimensional (1D), two-dimensional (2D), and even higher dimensional experiments are available to the researchers for identification and quantification of molecules for metabolomics studies. Proton (¹H) is most abundantly present nuclei in all the metabolites and has highest sensitivity, hence, 1D ¹H NMR spectra are mostly useful for NMR-based metabolomics studies. However, due to overlapping of various metabolite resonances, it is difficult to unambiguously assign all the metabolites in the 1D spectrum. Various 2D NMR experiments including homo- and hetero-nuclear 2D experiments are employed for unambiguous assignment of resonances. Following section briefly describes various 1D, 2D, and heteronuclear experiments especially useful in metabolomics studies.

1D NMR

Single pulse 1D ¹H NMR spectroscopy is a quick way to obtain the metabolic profile of the sample, acquisition time for this experiment is often as short as of few minutes. In a typical 1D pulse sequence, a single 90° pulse is applied on the sample that flips the z-magnetization into the x,y-plane. The magnetization is then allowed to freely evolve with time and the free induction decay (FID) is recorded. The Fourier Transformation (FT) of FID provides the 1D NMR spectrum of the sample. 1D spectrum represents the plot of intensity vs chemical shift positions of various resonances of metabolites present in the sample. Fig. 1.2 shows the 1D ¹H NMR spectrum of the urine sample obtained from a healthy control. It may be seen that a large number of metabolites have been identified in the spectrum. Single pulse 1D ¹H NMR is most commonly used in metabolomic profiling of tissue extracts, cell extracts, and biofluids like seminal plasma, urine, cerebrospinal fluid, etc.²¹

Fig. 1.2

Fig. 1.2 Representative 1D ¹ H NMR spectrum of urine obtained from a healthy control acquired at 700 MHz at 25°C, (A) aliphatic region (B) aromatic region. Abbreviations: Ala , alanine; Cho , choline; Crn , creatinine; DMA , dimethylalanine; DMG , dimethylglycine; For , formate; Gln , glutamine; Glu , glutamate; Gly , glycine; GPC , glycerophosphocholine; HA , hippurate; His , histidine; Ile , isoleucine; IS , indoxyl sulfate; Lac , lactate; Leu , leucine; NMN , N -methylnicotinamide; Phe , phenyalanine; Pyr , pyruvate; Tyr , tyrosine; Val , valine; βOHB , β-hydroxybutyarate. No permission required.

Serum or plasma contain massive amount of protein, lipoproteins, or phospholipids that results in peak broadening along with some narrow peaks from lower-molecular-weight metabolites. The analysis of such samples involves the suppression of resonances from macromolecules like lipoproteins and proteins, which are in general being suppressed using CPMG (Carr-Purcell-Meiboom-Gill) pulse sequence. Large molecules have short T2 relaxation and the total time delay of spin echo is chosen to match the T2 relaxation time of large molecules, thereby suppressing the signals.²² Solvent signals are saturated during relaxation delay before CPMG by low power RF pulse. This method helps in removing the broad resonances related with high molecular weight macromolecules and facilitates the detection of low molecular weight metabolites.

1D Nuclear Overhauser Spectroscopy (NOESY) is also a commonly used pulse sequence for biological samples.²¹ 1D version of selective total correlation spectroscopy (TOCSY) approach has also been used to quantitatively detect very low concentration (10–100 times below the major components) of metabolites. 1D TOCSY has been found to be highly useful for identifying targeted metabolites in biological samples.²³, ²⁴

Biofluids such as blood, urine, seminal plasma, are rich in water; therefore, the suppression of water signals is important to observe the NMR signal from the metabolites. A number of different solvent suppression schemes or NMR pulse sequences are available to suppress the water signal. Presaturation (PRESAT) is the most common technique which is used to suppress the water signal.²⁵ It basically consists of a long, low power RF pulse which is applied at the water frequency. This low power pulse excites the water proton signal continuously, rendering the water spins saturated and therefore unobservable. Solvent suppression is mainly affected by pulse sequence timings and power level. The power level used using presaturation pulse is critical and as small change in power level can have dramatic effects on nearby peaks and further downstream processing.²⁶–²⁸ Most metabolomics experiments are performed in aqueous solvent therefore, water suppression pulses are employed both in 1D and 2D NMR experiments.

The concentration of various metabolites can also be determined from 1D ¹H NMR spectrum since NMR spectral intensities are directly proportional to concentration of metabolites. The concentration of metabolite can be determined by comparing the intensity of resonance of the metabolite with the intensity of the standard compound added in known concentration to the sample using the formula given in the literature.²⁹, ³⁰ Nowadays, many offline software tools like Chenomx NMR Suit, Metaboanalyst, and Human Metabolome Database (HMDB) are available which can also be used for identification and quantification of metabolites.³¹–³⁶

Two dimensional NMR

The simplest 1D NMR spectrum provides the substantial information about the complex biological samples (e.g., blood, urine). However, high order spin coupling patterns complicate the 1D spectrum, making peak assignments difficult. To overcome this problem 2D NMR is carried out. In a simple 2D correlation spectroscopy (COSY) experiment, two 90° pulses are applied and the data are acquired by stepwise incrementing the time variable t1 (evolution period) and recording of data in t2 (detection period), thus generating, time domain data set (t1,t2). The data so obtained are Fourier transformed twice so as to attain a spectrum which is a function of two frequency variables.

Homonuclear 2D

The most commonly used 2D NMR technique is the homonuclear correlation spectroscopy (COSY) which provides correlation peaks between the ¹H nuclei which are J-coupled.³⁷, ³⁸ The COSY spectrum is a map indicating all the pairs of coupled nuclei and their frequencies, and thus is useful in unambiguous assignment of metabolite resonances.

Another widely used homonuclear 2D experiment is ¹H-¹H total correlation spectroscopy (TOCSY). TOCSY is a very useful experiment for assignment of metabolites, as it shows cross peaks between all the protons in the spin system.³⁹ Additionally, diffusion ordered spectroscopy (DOSY)⁴⁰, ⁴¹ and 2D J-resolved NMR spectroscopy (J-Res)⁴² are being used for NMR-based metabolomics. 2D J-Res NMR spectroscopy has also been found useful to assign overlapping metabolites resonances in biofluids.⁴³–⁴⁵

The 2D NMR spectrum has to be carried out with high resolution settings which take long experiment time. Such long experiment duration is not compatible with high-throughput character when analyzing large number of samples typically required in metabolomics studies. Favorably, novel pulse sequences have been developed to quicken multidimensional experiments.⁴⁶ These methods include fast repetition techniques,⁴⁷ spectral aliasing,⁴⁸ nonuniform sampling (NUS)⁴⁹ and Ultrafast (UF)⁵⁰ spectroscopic methods. Marchand et al. demonstrated applications of fast 2D approaches like NUS TOCSY and UF COSY to study the effect of growth promoter, ractopamine in the lipid extracts of pig serum.⁵¹ Apart from this fast metabolite quantification (FMQ) has also been developed for 2D ¹H-¹³C NMR and about 40 metabolites were quantified in about 12 min.⁵²

Heteronuclear 2D NMR

Additionally, heteronuclear 2D NMR experiments like ¹H,¹³C single quantum coherence (¹H-¹³C-HSQC) are useful in unambiguous assignment of metabolite resonances.⁵³ The 2D ¹³C-¹H HSQC, correlates the chemical shifts of ¹H and ¹³C bonded directly to each other thus revealing direct one-bond correlation between the ¹H and ¹³C spins. These experiments are very useful in assignments of backbone and NH signals from proteins.⁵⁴–⁵⁷ Bernini et al. have shown the utility of heteronuclear multiple bond correlation spectroscopy in metabolic profiling of urine.⁵⁸

Other nuclei used for metabolomics studies

As stated earlier that hydrogen is most commonly targeted nucleus with ¹H NMR spectroscopy for metabolomics studies. However, other than hydrogen less frequent atoms like carbon, phosphorus, and nitrogen are also targeted by NMR that is, ¹³C NMR, ³¹P NMR, and ¹⁵N to profile specific metabolites.⁵⁹ With respect to ¹H NMR spectroscopy that has narrow line width and chemical shift dispersion (~ 10 ppm), ¹³C NMR spectroscopy has broad (~ 200 ppm), chemical shift dispersion thus offering significant resolution advantage over ¹H NMR. However, the low sensitivity coupled with low natural abundance hinders its wide spectral width advantage and also substantially impede its application in metabolomics. Fortunately, different NMR approaches are available to enhance the ¹³C NMR signals and overcome the problem of low sensitivity. Keun et al. have used ¹³C direct-detect cryoprobe to examine the drug toxicity within acquisition time of 15 min. The ¹³C spectra so obtained had much greater spectral dispersion compared to ¹H spectra, providing sufficient signal to noise ratio to enable analysis.¹⁸, ⁶⁰ The ¹³C NMR is mostly useful for isotope tracing experiments⁶¹–⁶³ to study various biosynthetic pathways. Similarly, to ¹³C NMR, ¹⁵N NMR spectra also have broad chemical shift dispersion (~ 100 ppm) with narrow line widths. ¹⁵N NMR spectroscopy is predominately used for structural elucidation of proteins,⁶⁴–⁶⁶ RNA,⁶⁷–⁶⁹ and DNA,⁷⁰–⁷³ but not been commonly used in metabolomics studies due to its poor sensitivity. ³¹P is nearly 100% abundant with wide spectral dispersion but its utility in metabolomics is limited as number of metabolites containing phosphorus is less. It provides information on phospholipids and nucleotides.⁷⁴, ⁷⁵

High-resolution magic-angle spinning NMR spectroscopy

Apart from using the standard solution NMR methods, another nondestructive and real time NMR technique that can be used to analyze the semisolid tissue samples is high-resolution magic angle spinning (HRMAS). Using HRMAS spectroscopy, the intact tissue samples can be analyzed with a high resolution and is an efficient way to examine the metabolites without any prior need of sample extraction or prepreparation steps.⁷⁶–⁷⁸ The advantage with HRMAS is that it maintains tissue integrity and the microstructural information even after analysis. As the NMR analysis is carried out at low temperatures and examined within ~ 20 min, so no tissue decomposition occurs⁷⁹ and the tissue specimens can also be used for histopathological and other analyses for further validation. Few applications of HRMAS include development of database for biopsy specimens,⁸⁰ cancer research,⁷⁶, ⁸¹, ⁸² and biomarker discovery for drug-resistant epilepsy.⁸³ With recent development with microprobes for magic-angle spinning (μMAS) the analysis of submicrogram samples can be achieved.⁸⁴

Mass spectrometry

Mass spectrometry has been used as an important technique for the past few decades in many metabolomics studies due to high resolution, better sensitivity, and dynamic range of metabolites detected.⁸⁵ The recent advancements in MS have developed the potential to identify and quantify hundreds of metabolites in complex biological samples at a very low concentration up to attomolar.⁸⁵ The prerequisite four major steps involved in MS-based analyses include: (i) sample preparation/extraction; (ii) chromatographic separation of metabolites; (iii) ionization and separation of the ionized molecules by analyzer; (iv) detection of ionized molecules based on their mass-to-charge ratio (m/z). These steps are briefly described in the following section.

Sample preparation/extraction

Various types of biological samples like tissue, biofluids (blood, seminal fluid, urine) feces, bile, saliva, cerebrospinal fluid as well as cell culture can be analyzed through MS.⁸⁶ The appropriate sample storage, preparation and management are essential and crucial steps for high-throughput analysis and also for efficient extraction, recovery, and for profiling large number of metabolites. The efficient recovery of polar and nonpolar compounds, require use of different types of extraction procedures. Methanol-water-chloroform extraction is the most commonly used method for extraction of both types of compounds.

Chromatographic separation of metabolites

MS is coupled with different separation modalities like GC, LC, capillary electrophoresis (CE), and Ion mobility (IM) spectrometry that provide massive information for metabolomics studies.⁸⁷–⁹⁰

Gas chromatography-mass spectrometry

Gas chromatography-mass spectrometry (GC-MS) is one of the most established chromatographic-MS technique that is relatively inexpensive, convenient to use, stable, and reproducible.⁹¹ It is a technique of choice that is used to separate volatile metabolites. In order to increase the thermal stability and volatility of the analyte, various derivatization methods are used and among them methoximation and trimethylsilyation are mostly used for metabolomics studies.⁹² There are few drawbacks of GC-MS technique. The process of derivatization results in structural transformation of the analyte and various byproducts are formed, which may lead to difficulty in the interpretation of the data. Further, differences in the efficiency of derivatization affect the yield and reproducibility of data.⁹³, ⁹⁴ The chromatograms obtained after GC-MS technique are complex due to multiple derivatization products and take longer time to analyze, but the use of deconvolution software make the analysis shorter, but the resolution is not that good. The deconvolution approaches are appropriate for metabolomics data analysis, but the accuracy for untargeted analysis is lower compared with targeted analysis.⁹⁵

Liquid chromatography-mass spectrometry

Liquid chromatography-mass spectrometry (LC-MS) is one of the leading analytical technique used for metabolite profiling. The mobile phase here is liquid, thus there is no need for derivatizing the metabolites to volatile form. Therefore, a broader range of metabolites can be analyzed, detected, and quantitated. With the introduction of UPLC, high degree of separation, increased peak capacity and sensitivity can be achieved with the use of high flow rate and high pressure (range of 6000–19,000 psi) of solvent.⁹⁶, ⁹⁷ With the help of UPLC-MS better ionization with reduce mass spectral overlap can be accomplished with improved structural identification and conformation.⁹⁸ Presently, UPLC is the main technology used in metabolomic research. Based on the polarity of the metabolites different chromatographic columns are used.

Reversed-phase liquid chromatography-MS

Reversed-phase chromatographic columns are used to analyze lipid-soluble metabolites using C18 or C8 stationary phases. Reversed-phase liquid chromatography-mass spectrometry (RPLC) provides an opportunity for extensive range of selectivity values that helps in covering broad range of metabolites to be analyzed. During RPLC various buffer modifiers (e.g., acetic acid and formic acid) may be used with mobile phase so as to increase the separation efficiency.

For the analysis of highly polar and ionic compounds ion-pairing liquid chromatography (IPLC) an effective RPLC is used. For IPLC same type of stationary and mobile phase is used, the only difference is that for IPLC an ion pairing agent is added into the mobile phase.⁹⁹ The ion-pair reagent possess opposite charge than the target compound and is used to change the retention time of ionic analytes.¹⁰⁰ For anion analysis, ion pair reagent used is tetramethylammonium, tetraethylammonium, tetrapropylammonium, tetrabutylammonium, tributylamine, and hexylamine, while for cation analysis, the ion-pair reagent commonly includes are HCl, HClO4, perfuorocarboxylic acids, sulfonic acids (pentane, hexane, heptanes, and octane).¹⁰¹ Guo et al. used the IP-RPUHPLC-MS method for the analysis of metabolite (that contain phosphate and carboxylic acid) involved in cellular metabolism thus providing comprehensive knowledge of mechanisms related to metabolic adaptations.¹⁰²

Hydrophilic interaction chromatography-MS

Hydrophilic interaction chromatography (HILIC) is an alternative is an alternative HPLC mode used for separating polar compounds. It makes use of polar stationary phase and also proportion of organic mobile phase is high with increased percentage of eluent (water).¹⁰³ HILIC-MS is complementary to RPLC-MS in the sense that the metabolites that elute early in RPLC-MS are also retained. Nevertheless, it generates wider peak compared to RPLC resulting in lower peak capacity and high dependency of peak resolution on mass spectrometer.¹⁰¹ The HILIC-MS analysis provides a wider view for metabolic data analysis.

Supercritical fluid chromatography-MS

A supercritical fluid is known as a substance which is in a liquid state at a temperature and pressure above a critical point.¹⁰⁴ SFC uses supercritical fluids, mostly CO2, which is used as the mobile phase, as it is nontoxic and easy to handle.¹⁰⁵ Lisa and Holcapek using SFC-ESI-MS developed a novel analytical strategy to separate 30 nonpolar and polar lipids in a short analysis time of 6 min and also identify and quantify individual lipid species. The SFC-ESI-MS offered an improved quantitation of lipids due to complete separation preventing ion suppression effects in different lipid classes.¹⁰⁶

Capillary electrophoresis-MS

Capillary electrophoresis is an analytical technique that separate ions based on their electrophoretic mobility. It uses a capillary column filled with buffer in an electrical field (300–500 V/cm).¹⁰⁷ CE is a robust complementary technique for the analysis of polar and charged metabolites.⁸⁸ In CE-MS the metabolites are first separated based on their charge-to-size ratios and then based on mass-to-charge ratios.¹⁰⁸, ¹⁰⁹ CE-MS is a valuable analytical technique and offers few advantages over other analytical techniques, that is, short analysis time, low sample requirement, easy sample preparation, and limited consumption of reagents. The drawback with this technique is lower sensitivity as very small volume (nanoliter range) of analyte is introduced into the capillary.¹¹⁰

Ion mobility-MS

IM is regarded as a gas-phase electrophoresis and allows an effective means of separating gaseous ions in electric field based on their mass, shape, size, and charge and provides information both about chromatographic separation and ion mass spectrometry separation.⁸⁷ IM-MS can be used to distinguish isobars and isomers and has increased signal-to-noise ratio.¹¹¹ There are three major IMS techniques: traveling wave, drift-time IMS, and field asymmetric IMS.

Ionization

After the chromatographic separation, samples are subjected to ion source for the process of ionization. The ion source plays a very important role for the generation of ions in MS, which are subsequently analyzed based on their charge/mass ratio. The process of ionization converts the sample molecule into electrically charged positive or negative species in gaseous form. The charged species in the gaseous form are further exposed to heat and dry nitrogen to evaporate the droplets. The evaporated droplets then transfer the charge to compound via charge transfer and ionize them in the negative and positive form.¹¹² There are various methods of ion generation that includes protonation, deprotonation, cationization, transfer of charged molecules to gas phase, electron ejection, and electron capture. Protonation is used for basic species and often used for peptides. Deprotonation is very useful for acidic species like carboxylic acids, phenols, and sulfonic acids. Cationization is useful with molecules unstable to protonation and carbohydrates are excellent biomolecule to ionize by this process. There are various method that produce intact molecular ion with minimal fragmentation, under adequate experimental condition known as soft ionization. Soft ionization is useful for biological samples of large molecular weight as it doesn’t fragment the macromolecules into smaller charged particles, rather the macromolecule is being ionized into small droplets thus helps in accurate mass detection. Various types of ionization sources are used that include chemical ionization (CI), electron ionization (EI), atmospheric pressure chemical ionization (APCI), electrospray ionization source (ESI), and matrix-assisted laser desorption ionization (MALDI). The two most widely used soft ionization analytical technique for metabolomics research is ESI and MALDI and proven to be useful for large scale omics studies.¹¹³, ¹¹⁴ In the last few years, the ESI source has become more robust with wide applications as it offers a conventional method to generate gaseous ions of the analyte.¹¹⁵, ¹¹⁶ MALDI is primarily used for proteomic research and extensively used for the detection and quantification of small molecular metabolites.¹¹⁷

Detection

After the process of ionization, the ionized biomolecules are detected by the mass analyzer thereby generating a high-resolution mass spectrum. Different types of mass analyzers are used that include (i) time of flight (TOF), (ii) quadrupole time of flight (QTOF), (iii) quadrupole, (iv) ion trap, and (v) orbitrap.

Quadrupole

It is one of the most popular types of mass analyzer used. They are often used with triple quadrupole mass spectrometers or with other analyzers like TOF. A quadrupole mass analyzer is a type of mass filter as it can separate and filter ions of distinct m/z ratio.¹¹⁸ Quadrupoles consist of four parallel cylindrical rods and a radiofrequency potential (RF) potential is applied to these rods. The RF potential is superimposed by direct current (DC) potential. As the ions pass through the oscillator, both RF and DC potential are used to oscillate them. Based on DC potential and RF field frequency, only ions of a specific m/z would have stable trajectories while, the other ions with unstable trajectories are filtered out. By varying these, ions with different m/z are filtered.¹¹⁹ A quadrupole can also operate in RF-only mode and this allows all ions to pass through the multipole, thus transforming the quadrupole into a transmitting device that transmits ions from one area of the mass spectrometer to another. Therefore, RF-only multipoles can act as ion transmission guide. Apart from transmission guide, RF-only multipoles can also collision-induced dissociation (CID) by acting on collision cells.¹²⁰ The major advantage with quadrupole is the low cost, size, and stability and thus require less calibration. Since quadrupole analyzer have fast duty cycles so they can be easily combined with GC and LC; however, these are unsuitable for MALDI. As quadrupole has limited mass range and poor resolution so cannot be used for analyzing large molecular weight compounds or mixtures of compounds that have similar masses.¹²¹

Ion trap

The ion trap analyzer is an alternative form of quadrupole and there are two types of ion trap that is, 2D and 3D.¹²⁰ In the 3D ion trap, two hyperbolic electrode plates are placed facing each other and in between them a hyperbolic ring electrode is placed. Using an oscillating RF potential and a superimposed DC potential, ions are trapped between the electrodes. By applying a varying RF potential, ions with different m/z are ejected selectively and then detected. The 2D ion trap, is also called as linear trap, the RF potential is used at both ends to trap the ions. The selected ions are ejected either radially or axially or as per the design of trap. Since the ion trap accumulate ions thus improving the sensitivity. Similar to quadrupole they are small and compact and widely used in mass spectrometers and play a major role in proteomics.¹²², ¹²³ The disadvantage of ion trap analyzers is their low resolving power.

Time of flight

A TOF analyzer comprises of an acceleration grid and a flight tube, the acceleration grid that accelerate ions from the ionization source to the MS detector.¹²⁴ In TOF the m/z ratio is determined by time-of-flight measurement. An electric field of known strength is applied to accelerate ions. Due to acceleration, the ions with same charge will have same kinetic energy. However, the velocity of these ions depends on the m/z ratio of the ion, which will affect the time taken by the ions to reach the detector. Since, the time taken depends on the velocity of the ion, m/z ratio of the ion can be determined. Then using m/z ratio and experimental parameters, the unknown ions can be identified. The TOF analyzer allows the separation of ions with masses few daltons to 100 kDa thus making it the analyzer of choice for biomolecules such as proteins.¹²⁵ TOF analyzers are most popular for performing peptide mass fingerprinting.¹²⁶

Orbitrap

This is simple an ion trap and there is no RF and magnetic field. The orbitrap analyzer uses an oscillating electric field to store ions between external electrodes and also operates in MS/MS mode which combines a nonselective full-scan MS spectrum with high mass accuracy.¹²⁷ De Clercq et al. used the orbitrap analyzer to study fecal metabolomics fingerprints of the human gut to determine its phenotype.¹²⁸

Tandem mass analyzers

In tandem mass analysis, two or more analyzers are used successively.¹²⁹ In tandem MS analysis the first analyzer selects ions m/z value, and these ions are then subjected to CID and the resulting product ions are analyzed using a second mass analyzer.

Triple quadrupole (QqQ)

This is most common MS/MS instrument used for quantification. Quadrupole separates the sample ions based on m/z ratio. Separation of molecules in quadrupole depends on the stability of trajectories of their ions in the oscillating electrical fields. In this the first quadrupole measures precursor ion generated through ionization source, these isolated precursor ions are then transferred to the second quadrupole that is, collision cell where fragmentation of precursor ion occurs. The third quadrupole then selects one specific fragmented ion or several product ion(s).

Quadrupole TOF-MS

The quadrupole time-of-flight (Q-TOF) is the most commonly used mass spectrometer in untargeted metabolomics. Quadrupole TOF analyzers are composed of a quadrupole, a collision cell and a TOF tube. The TOF analyzer is positioned orthogonal to the quadrupole analyzer.¹³⁰ The ions are first filtered through the quadrupole analyzer and then injected into the TOF analyzer orthogonally with the help of puller and pusher plates placed between the two analyzers. Based on the m/z ratio the ion can be identified. Ning et al. has identified 23 potential biomarkers in plasma of patients with metabolic syndrome using LC-Q-TOF.¹³¹ Another study investigated the effect of a traditional Chinese medicine in urine samples of an acute blood stasis rat model.¹³²

TOF/TOF

In this scheme, two TOF analyzers are utilized and in between these analyzers, CID is performed.¹³³ In TOF/TOF analyzer, first, the ions of dissimilar m/z are being separated based on the velocity. In the second step, ions of a particular m/z are selected filtering the rest ions. In the third step, the selected precursor ions are deaccelerated by passing them to a set of ion optics. Then the ions pass through a collision cell for CID. These ions are then analyzed by accelerating into a second flight tube. The fast analysis through this analyzer in combination with MALDI ion sources is used for the analysis of peptides.

Chemometric analysis of metabolomics data

Data analysis is a decisive step in metabolomics due to high dimensionality and complexity of the raw data. The metabolomics data analysis involves following steps: data pretreatment, preprocessing step, processing, model validation, and postprocessing of the dataset.

Data pretreatment

The data pretreatment aims at improving the signal quality and also to reduce the possible biases. The preprocessing steps include baseline correction, correct peak assignment, peak detection, and peak alignment.

Baseline correction

In both, NMR and MS, baseline correction is carried out to remove low frequency artifacts that arise due to instrumental and experimental variations.¹³⁴

Peak detection

For NMR data mostly binning-based approaches are used to identify the feature peaks in complex biological samples. However, this method poorly performs whenever there is spectral unalignment and also when multiple bins are captured under one single bin. Due to this reason peak based methods are increasingly adopted for NMR studies.¹³⁵ For the MS data appropriate peak detection/assignment involves identification of matching m/z ratio and then assigning it to adduct appropriately.

Peak alignment

Peak alignment is one of the important preprocessing steps when multiple samples are involved and remove the undesired variability in the samples. In NMR, the peak position of the metabolites may be influenced by nonlinear shifts that may arise due to change in the chemical environment such as pH, ionic strength, or protein content of the sample.¹³⁶, ¹³⁷ In MS-based studies, the peak shifts across the retention time axis, may occur due to alteration in the stationary phase of the column used for chromatographic separation.¹³⁸ There are two types of spectral alignment algorithms (i) spectral alignment methods (ii) peak-based alignment methods. Apart from baseline correction, peak detection, and peak alignment the MS data preprocessing also include retention time correction and chromatogram alignment.

Following these steps, the data are then normalized, and the normalization steps include centering, scaling, and transformation.¹³⁹–¹⁴¹ The purpose of centering is to adjust the concentration differences present between the metabolites by converting all measurements to vary around zero instead of mean of the metabolite. Mostly mean centering is used for metabolomics data analysis,¹⁴² whereas scaling removes the variation present in the concentration of the metabolites which can influence the data analysis. Numerous scaling procedures are available for metabolomic data analysis but mean centering, autoscaling (standard deviation), pareto scaling (square root of the standard deviation) are commonly used. Transformations are the mathematical approaches that are applied to rectify the heteroscedasticity present in the dataset, thereby reducing the variability between variables. Log transformation is mostly used for metabolomics data¹⁴³; however, cubic root¹⁴⁴ and quantile¹⁴⁵ transformations can also be used.

Preprocessing steps for metabolomics data

The goal of the preprocessing method is to obtain a sketch of the variables of importance prior to processing by prediction model. The preprocessing involves mainly two types of approaches, that is, univariate and multivariate analysis.

Univariate analysis

Univariate statistical analysis is used to examine only one variable at a time. The commonly used methods when the data is parametric are t-tests and analysis of variance (ANOVA), whereas the nonparametric data analysis include Mann-Whitney test and Wilcoxon test. In addition to this, univariate analysis is also performed to determine correlations between variables. Pearson’s correlation is a choice of preference when the data is parametric while Spearman’s correlation is usually used in nonparametric datasets.

Multivariate analysis

Multivariate analysis is divided into two subgroups that is unsupervised and supervised data analysis. In unsupervised method of data analysis, the modeling is based only on the explanatory variables, without having any prior knowledge of the dataset. The most common are principal component analysis (PCA) and hierarchical cluster analysis (HCA).¹⁴³ PCA linearly transforms the metabolic features into linearly uncorrelated (orthogonal) variables into score vectors and loadings, called principal components.¹⁴⁶ PCA is also helpful in assessing the data quality by identifying sample outliers present in the study, thus making the data more robust for further analysis. HCA is also a type of clustering and visualization tool which is based on variables similarities/dissimilarities. It can be done in agglomerative mode (aggregation of samples into clusters) or divisive mode (division of complete dataset into clusters).¹⁴⁷, ¹⁴⁸

Processing approach for metabolomics

Once the variables of interest have been identified, the next step involves building up of a predictive model to classify new samples, identify biomarkers and to explore the mechanisms of metabolomics studies (i.e., metabolic pathways). At this stage, the supervised methods play an important role and use to identify the metabolic patterns that are correlated with the phenotype. These methods are also the foundation for building classifiers based on identified metabolomic features.¹⁴⁹ The most widely used supervised method in metabolomics is partial least squares (PLS). It can be used either as regression analysis or as binary classifier (PLS-DA i.e., binary variable of interest). PLS maximizes the covariance between the variable of interest and metabolomics data, thus producing score vectors and loading vectors. The loading vector measures how much a feature/metabolite contributes to demarcate between groups under analysis. In PLS, few metabolites that do not correlate with the variables present in the study and can influence the results, so in order to solve this problem, orthogonal PLS (OPLS) was developed.¹⁵⁰ OPLS is further expansion of PLS that includes an orthogonal signal correction. The other supervised analysis that uses nonlinear methods to build classifiers based on metabolomics data are Support vector machines (SVM), random forests (RF), and logistic regression analysis (LRA).

Model validation of the metabolomics data

Validation of a model is essential to estimate the performance of the model to find out how well the model will perform when applied to new dataset.

The performance and strength of the predictive models is a key and fundamental step for metabolomics data analysis and requires model validation methods. The aim of model validation is to accurately classify the hypothesized association between variables and responses.¹⁵¹ The coefficient of determination (R2) is the simplest method for choosing the optimal model and is indicated as the ratio between 0 and 1, where 1 suggests the perfect prediction. This validation method is used for small datasets, however, for large high-dimensional and complex datasets; cross validation (CV) methods are preferred. In CV, the data is split into two sets, training set and validation set.¹⁴⁶ The mostly used CV methods are k-fold, leave-one-out cross validation (LOOCV) and Monte Carlo cross validation (MCCV). However, nowadays the widely used standard method model validation is the receiver operating characteristic (ROC) curve.¹⁴⁹ The predictive ability of the model is validated in terms of specificity and sensitivity. The ROC curve is a plot of sensitivity versus (1-specificity) with a series of cut-off points. ROC is summarized as a single matric known as area under curve (AUC). AUC quantitatively measures the ability of predictive model and is expressed as ratio between 0 and 1, with value near to 1 indicates perfect classifier.¹⁴⁹

Postprocessing methods

The aim of postprocessing step is to recognize a useful biomarker or panel of biomarkers that helps in accurate prediction of particular phenotype. Usually, metabolic pathway analysis is the most useful approach that provides an outline of the relationship between different recognized metabolites, metabolic pathways, and other biological networks. Pathifier¹⁵² and Metaboanalyst¹⁴⁴, ¹⁵³ are the most commonly used software for metabolic pathway analysis of metabolomics data. Various metabolite spectral databases are available that helps in metabolite identification. Commonly used tools include Metaboanalyst,¹⁵⁴ XCMS¹⁵⁵ and 3Omics¹⁵⁶ with diverse analysis capabilities. Various software are available that are used to process the raw data, perform various statistical analyses for feature identification and for metabolite identifications and these software include Human Metabolome Database (HMDB),¹⁵⁷ LIPID MAPS,¹⁵⁸ Metabolite and Tandem MS Database (METLIN),¹⁵⁹ BiGG,¹⁶⁰ Madison Metabolomics Consortium Database (MMCD),³⁵ SetupX,¹⁶¹ MetaboLights,¹⁶² and KNApSAcK.¹⁶³ Compound-specific databases include Chemical Entities of Biological Interest (ChEBI)¹⁶⁴ and In Vivo/In Silico Metabolites Database (IIMDB).¹⁶⁵ The Metabolic pathway databases include BiKEGG,¹⁶⁶ MetaCyc,¹⁶⁷ BioCyc,¹⁶⁷ Model SEED,¹⁶⁸ and Reactome.¹⁶⁹

Applications of metabolomics in various diseases

Metabolomics approach has been applied to study, analyze, and characterize wide range of diseases, through the analysis of different kinds of biological samples, tissue biopsies, and tissue extracts. Metabolomics is known to play an important role in early disease diagnosis and progression. Following section describes the applications of NMR- and MRS-based metabolomics briefly. The readers may refer to extensive clinical diagnostic applications of metabolomics elsewhere.⁷, ⁸, ²¹, ³⁹, ¹⁴², ¹⁵³, ¹⁷⁰

Applications of NMR-based metabolomics

¹H NMR-based urinary metabolomic approach has been used to identify breast cancer (BC) specific metabolites that may be employed for diagnosis of BC. The metabolites such as creatine, glycine, trimethylamine N-oxide, and serine showed highest sensitivity and specificity and were able to discriminate BC patients from that of control.¹⁷⁰ Metabolomics has been proved to be a promising tool for illustrating the biochemical pathways effected by prostate cancer and for identification of new novel clinical biomarkers using biofluids.¹⁷¹ A ¹H NMR-based spectroscopic profiling of filtered serum demonstrated that glycine, sarcosine, alanine, creatine, xanthine, and hypoxanthine were able to differentiate abnormal prostate from that of healthy control.¹⁷² Apart from cancer, NMR-based metabolomics has also been used to study the metabolic disorders such as diabetes. In one of the recent study by Coco et al. identified the metabolic signature of serum associated with T2DM stages. The metabolic perturbation suggested that branched chain amino acids (BCAAs) were significantly altered in T2DM patients while perturbation in gluconeogenic amino acids other than BCAAs suggest both early and advanced T2DM stages.¹⁷³ Additionally, to study the metabolic disorders ¹H NMR spectroscopy in combination with chemometric analysis has been used for the diagnosis and for determining the severity of coronary heart disease. Metabolomic profiling technologies has the potential to differentiate between patients with different outcome after an acute myocardial infarction. Metabolomics depicted differential clustering between two cohorts with training set showing sensitivity of 76.9%, specificity of 79.5%, accuracy of 78.2% and AUC of 0.859. The results were reproduced in the validation set with sensitivity, specificity, and accuracy of 72.6% in each, respectively.¹⁷⁴ ¹H NMR-based metabolomics has also been used to study the autoimmune disorders. Among the various autoimmune disease, celiac disease (CeD) is chronic enteropathy caused by dietary gluten individuals with genetic predisposition. In the recent study by Upadhyay et al., metabolic profile of small intestinal mucosa, blood plasma, and urine demonstrated nine altered metabolic pathways contributing to the pathogenesis of CeD. Classification model was calculated using a combination of differentiating metabolites of blood plasma and urine with a sensitivity of 97.7%, specificity of 93.3%. The predictive accuracy was 95.1% for the model.¹⁵³

Applications of MS-based metabolomics

MS-based metabolomics plays a pioneer role in the field of health and medical science. GC-MS has wide application in the field of metabolomics from understanding the disease pathophysiology to microbial metabolism. Zhao et al. using GC-MS platform has developed automated high-throughput sample derivatization method to study more than 100 microbial metabolites in different biological samples that is human serum, urine, and feces in a very short span of 15 min for each sample. This study may provide a new avenue for metabolomics of microbiome.⁹⁰ Hadi et al. using GC-MS has identified serum metabolites that can act as predictive markers for diagnosis, grading, staging, and neoadjuvant status of breast cancer and thus may be helpful to improve and modify the treatment strategies for breast cancer. GC-MS was also found to be useful to study the autoimmune disease.¹⁷⁵ A study by Yan et al. has identified urinary metabolic markers for systemic lupus erythematosus. A total of 70 endogenous metabolites related to oxidative stress, energy metabolism, and metabolism of gut-microbiome. This study highlights the importance of GC-MS technique in understanding the pathophysiology of disease.¹⁷⁶ Liu et al. using LC-MS-based lipidomics and plasma metabolomics have identified panel potential biomarker that can differentiate bladder cancer and renal cell carcinoma patients from control. The eight metabolites were found to discriminate bladder cancer and renal cell carcinoma from that of controls. The AUC was found to be 0.985 and 0.993 for plasma metabolomics and lipidomics, respectively.¹⁷⁷

Boelaert et al. utilizes HILIC-MS to study novel biomarkers of chronic kidney disease.¹⁷⁸ Metabolites such as 2-hydroxyethane sulfonate, glycoursodeoxycholic acid, and pregnenolone sulfate were identified that demarcated chronic kidney disease patients from control thus opening a new perspective for future study.

Taguchi et al. used SFC-MS for target profiling of bile acids in rat serum. They simultaneously quantitated 24 bile acids without any extraction using solid-phase within in a short span of 13 min indicating the utility of SFC-MS in metabolomics.¹⁷⁹

Onjiko et al. developed microprobe single-cell CE-ESI-MS to carry out in situ analysis and characterization of the metabolic changes in single cells for the first time in a freely developing vertebrate embryo (Xenopus laevis). About 230 molecules are detected in as low as < 0.02% of the single cell content (positive ion mode). In situ microprobe single-cell CE-ESI-MS provides new avenues to understand the metabolic changes associated with cell differentiation process during impaired development.¹⁸⁰ Fujii et al. have used CE-MS to study the metabolic changes in the postmortem brain of patients with schizophrenia, and the findings revealed alteration in glucose metabolism and proteolysis in the brain of schizophrenia patients.¹⁸¹

IM-MS has numerous applications in metabolomics. Zhang et al. used ESI IM-MS to study the metabolomics of striatal rat model of Parkinson’s-like disease. Nine metabolites were identified using Human Metabolome Database (HMDB).¹⁸² The important finding of this work is that IM-MS identified a dopamine isomer, which has not been previously reported. The above study suggested that the reduction in dopamine in Parkinson’s disease may be miscalculated due to the presence of its isomer. Dwivedi et al. using IM-MS has done the profiling of human blood and identified 300 isomeric and 1100 metabolites ions. The metabolites present in low concentration that may not be detected using MS, can be detected in the mobility space. Also, the peak capacity of IM-MS is higher by sixfold more compared to MS analysis indicating its potential as promising analytical technique for