Handbook of Experimental Pharmacology 277

Veronica Ghini
Kathleen A. Stringer
Claudio Luchinat Editors

Metabolomics
and Its Impact
on Health and
Diseases
Handbook of Experimental Pharmacology

Volume 277

Editor-in-Chief
Martin C. Michel, Dept of Pharmacology, Johannes Gutenberg Universität, Mainz,
Germany

Editorial Board Members

James E. Barrett, Center for Substance Abuse Research, Lewis Katz School of
Medicine at Temple University, Philadelphia, PA, USA
David Centurión, Dept. of Pharmabiology, Center for Research and Advanced
Studies, Col. Granjas-Coapa, Mexico
Veit Flockerzi, Institute for Experimental and Clinical Pharmacology and Toxicol-
ogy, Saarland University, Homburg, Germany
Pierangelo Geppetti, Headache Center, University of Florence, Florence, Italy
Franz B. Hofmann, Forschergruppe 923 Carvas, Technical University, München,
Germany
Kathryn Elaine Meier, Dept. of Pharmaceutical Sciences, Washington State Univer-
sity Spokane, Spokane, USA
Clive P. Page, SIPP, Kings College London, London, UK
KeWei Wang, School of Pharmacy, Qingdao University, Qingdao, China
The Handbook of Experimental Pharmacology is one of the most authoritative and
influential book series in pharmacology. It provides critical and comprehensive
discussions of the most significant areas of pharmacological research, written by
leading international authorities. Each volume in the series represents the most
informative and contemporary account of its subject available, making it an unri-
valled reference source.
HEP is indexed in PubMed and Scopus.
Veronica Ghini • Kathleen A. Stringer •
Claudio Luchinat
Editors

Metabolomics and Its

Impact on Health
and Diseases
Editors
Veronica Ghini Kathleen A. Stringer
Department of Chemistry Department of Clinical Pharmacy
University of Florence University of Michigan–Ann Arbor
Sesto Fiorentino, Florence, Italy Ann Arbor, MI, USA

Claudio Luchinat
Department of Chemistry
University of Florence
Sesto Fiorentino, Florence, Italy

ISSN 0171-2004 ISSN 1865-0325 (electronic)

Handbook of Experimental Pharmacology
ISBN 978-3-031-26858-8 ISBN 978-3-031-26859-5 (eBook)
https://doi.org/10.1007/978-3-031-26859-5

# The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature Switzerland
AG 2023
This work is subject to copyright. All rights are solely and exclusively licensed by the Publisher, whether
the whole or part of the material is concerned, specifically the rights of translation, reprinting, reuse of
illustrations, recitation, broadcasting, reproduction on microfilms or in any other physical way, and
transmission or information storage and retrieval, electronic adaptation, computer software, or by
similar or dissimilar methodology now known or hereafter developed.
The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication
does not imply, even in the absence of a specific statement, that such names are exempt from the relevant
protective laws and regulations and therefore free for general use.
The publisher, the authors, and the editors are safe to assume that the advice and information in this book
are believed to be true and accurate at the date of publication. Neither the publisher nor the authors or the
editors give a warranty, expressed or implied, with respect to the material contained herein or for any
errors or omissions that may have been made. The publisher remains neutral with regard to jurisdictional
claims in published maps and institutional affiliations.

This Springer imprint is published by the registered company Springer Nature Switzerland AG
The registered company address is: Gewerbestrasse 11, 6330 Cham, Switzerland
Preface

This volume of the Handbook of Experimental Pharmacology, which celebrated its

100th anniversary in 2019, addresses the rapidly growing and evolving field of
metabolomics. It has been compiled and designed to broaden and enrich your
understanding as well as simplify a complicated picture of the diverse field of
metabolomics. This is accomplished by chapters from experts in the field on basic
principles as well as reviews and updates of analytical techniques. The variety and
different perspectives of the nuclear magnetic resonance approaches are described in
the chapters authored by David S. Wishart and co-authors, G. A. Nagana Gowda and
Daniel Raftery, and Ryan T. McKay. Advances in mass spectrometry are covered by
Charles R. Evans and co-authors and Stefan Kempa and co-authors. This book also
reflects the state of the art in the application of metabolomics to cell biology (Ulrich
L. Günther and co-authors) and chapters that share insights into the application of
metabolomics. These include the assessment of treatment response (Paola Turano
and co-authors) and the phenotyping of various diseases (Rachel S. Kelly and co-
authors, Paige Lacy and co-authors, and Angela J. Rogers and co-author).
Relationships of metabolomics and drugs are highlighted by Robert Verpoorte and
co-authors, Oscar Millet and co-authors, and Daniel L. Hertz and co-author. Given
the diverse topics addressed, we believe this book has interdisciplinary appeal and
scholars with an interest in the role of metabolomics in achieving precision medicine
will find it of particular or special interest.
We want to thank the authors for their contributions as this Handbook would not
be possible without them. We also express our appreciation to the many
investigators who work in the field of metabolomics and strive to advance the
science. Its analytical advancement as well as its translation to the clinic is of vital
importance to the field. It is our belief that metabolomics will continue to provide
new and novel insights into complex illnesses and enable more accurate and precise
therapies in the future. We would also like to express our sincere appreciation to
Susanne Dathe, Springer Editor for Neurosciences/Pharmaceutical Sciences/
Protocols, whose commitment and competence have helped to continue the tradition

v
vi Preface

of this remarkable series, and to the past and current editorial board members who
have dedicated time and effort into establishing this series as one of the most
recognized publications in pharmacology.

Florence, Italy Veronica Ghini

Ann Arbor, MI, USA Kathleen A. Stringer
Florence, Italy Claudio Luchinat
Contents

Practical Aspects of NMR-Based Metabolomics . . . . . . . . . . . . . . . . . . . 1

David S. Wishart, Manoj Rout, Brian L. Lee, Mark Berjanskii,
Marcia LeVatte, and Matthias Lipfert
Compound Identification Strategies in Mass Spectrometry-Based
Metabolomics and Pharmacometabolomics . . . . . . . . . . . . . . . . . . . . . . 43
Rylan Hissong, Kendra R. Evans, and Charles R. Evans
Metabolomics and NMR . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
Ryan T. McKay
Natural Products Drug Discovery: On Silica or In-Silico? . . . . . . . . . . . 117
Hye Kyong Kim, Young Hae Choi, and Robert Verpoorte
Quantitative NMR Methods in Metabolomics . . . . . . . . . . . . . . . . . . . . . 143
G. A. Nagana Gowda and Daniel Raftery
Advancements in Pulsed Stable Isotope-Resolved Metabolomics . . . . . . . 165
Martin Forbes, Sabrina Geisberger, Matthias Pietzke, Guido Mastrobuoni,
and Stefan Kempa
Metabolomics in Cell Biology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 181
Zuhal Eraslan, Marta Cascante, and Ulrich L. Günther
NMR-Based Metabolomics to Evaluate Individual Response
to Treatments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 209
Alessia Vignoli, Gaia Meoni, Veronica Ghini, Francesca Di Cesare,
Leonardo Tenori, Claudio Luchinat, and Paola Turano
Pharmacometabolomics of Asthma as a Road Map to Precision
Medicine . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 247
Rachel S. Kelly, Margaret F. Cote, Sofina Begum, and Jessica Lasky-Su
Prospective Metabolomic Studies in Precision Medicine:
The AKRIBEA Project . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 275
Maider Bizkarguenaga, Rubén Gil-Redondo, Chiara Bruzzone,
Ganeko Bernardo-Seisdedos, Ana Laín, Beatriz González-Valle,
Nieves Embade, José M. Mato, and Oscar Millet
vii
viii Contents

Chemotherapy-Induced Peripheral Neuropathy . . . . . . . . . . . . . . . . . . . 299

Ciao-Sin Chen and Daniel L. Hertz
Metabolomics of Respiratory Diseases . . . . . . . . . . . . . . . . . . . . . . . . . . 339
Subhabrata Moitra, Arghya Bandyopadhyay, and Paige Lacy
The Metabolomics of Critical Illness . . . . . . . . . . . . . . . . . . . . . . . . . . . . 367
Ana E. Pacheco-Navarro and Angela J. Rogers
Practical Aspects of NMR-Based
Metabolomics

David S. Wishart, Manoj Rout, Brian L. Lee, Mark Berjanskii,

Marcia LeVatte, and Matthias Lipfert

Contents
1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
2 A Short History of NMR-Based Metabolomics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
3 Practical Aspects of NMR-Based Metabolomics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
3.1 Sample Preparation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
3.2 Choosing the Right Pulse Sequence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
3.3 Spectral Acquisition and Processing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
4 Data Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
4.1 Data Analysis for Targeted Metabolomics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
4.2 Data Analysis for Untargeted Metabolomics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
4.3 Targeted or Untargeted? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
5 Biological Interpretation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
6 Conclusion and Future Prospects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31

D. S. Wishart (✉)
Department of Biological Sciences, University of Alberta, Edmonton, AB, Canada
Department of Computing Science, University of Alberta, Edmonton, AB, Canada
Department of Laboratory Medicine and Pathology, University of Alberta, Edmonton, AB, Canada
Faculty of Pharmacy and Pharmaceutical Sciences, University of Alberta, Edmonton, AB, Canada
e-mail: [email protected]
M. Rout · B. L. Lee · M. Berjanskii · M. LeVatte
Department of Biological Sciences, University of Alberta, Edmonton, AB, Canada
e-mail: [email protected]; [email protected]; [email protected]; [email protected]
M. Lipfert
Department of Biological Sciences, University of Alberta, Edmonton, AB, Canada
Reference Standard Management & NMR QC, Lonza Group AG, Visp, Switzerland
e-mail: [email protected]

# The Author(s), under exclusive license to Springer Nature Switzerland AG 2022 1

V. Ghini et al. (eds.), Metabolomics and Its Impact on Health and Diseases,
Handbook of Experimental Pharmacology 277, https://doi.org/10.1007/164_2022_613
2 D. S. Wishart et al.

Abstract

While NMR-based metabolomics is only about 20 years old, NMR has been a key
part of metabolic and metabolism studies for >40 years. Historically, metabolic
researchers used NMR because of its high level of reproducibility, superb instru-
ment stability, facile sample preparation protocols, inherently quantitative char-
acter, non-destructive nature, and amenability to automation. In this chapter, we
provide a short history of NMR-based metabolomics. We then provide a detailed
description of some of the practical aspects of performing NMR-based
metabolomics studies including sample preparation, pulse sequence selection,
and spectral acquisition and processing. The two different approaches to
metabolomics data analysis, targeted vs. untargeted, are briefly outlined. We
also describe several software packages to help users process NMR spectra
obtained via these two different approaches. We then give several examples of
useful or interesting applications of NMR-based metabolomics, ranging from
applications to drug toxicology, to identifying inborn errors of metabolism to
analyzing the contents of biofluids from dairy cattle. Throughout this chapter, we
will highlight the strengths and limitations of NMR-based metabolomics. Addi-
tionally, we will conclude with descriptions of recent advances in NMR hard-
ware, methodology, and software and speculate about where NMR-based
metabolomics is going in the next 5–10 years.

Keywords
Applications · Experimental methods · NMR spectroscopy · Targeted
metabolomics · Untargeted metabolomics

1 Introduction

Metabolomics is a branch of analytical chemistry that comprehensively characterizes

the molecules in various biofluids and tissues. Metabolites are the chemical
constituents of the metabolome. The metabolome, therefore, can be defined as the
complete collection of all chemicals or metabolites found within cells, biofluids,
organs, or organisms (Oliver et al. 1998; Wishart 2005). These chemicals (most of
which have a molecular weight <1,500 Da) include endogenously derived
compounds (amino acids, nucleic acids, organic acids, carbohydrates, lipids, and
minerals) and exogenously acquired compounds (vitamins, food additives, plant
phytochemicals, drugs, cosmetic chemicals, dyes, detergents) or just about any
other chemical that an organism can consume or to which it can be exposed.
Metabolites are essential building blocks for all life processes. They serve as the
bricks and mortar for cells, being the small molecule constituents (proteins, RNA,
DNA) for all basic cellular functions. Furthermore, metabolites are the fuel for
cellular processes, the barriers to maintain cellular integrity, and messengers for
signaling processes. Metabolites are the end-products of complex processes which
Practical Aspects of NMR-Based Metabolomics 3

are encoded for and controlled by genes. Therefore, metabolites are exquisitely
sensitive indicators of problems in the genome. Indeed, a single base change in a
gene can lead to a 10,000-fold change in the concentrations of certain metabolites
(Wishart et al. 2007). This remarkable sensitivity of metabolite levels to genetic
variations led to one of the most common and widespread medical testing
initiatives – newborn screening. For more than 100 years, metabolite testing has
been used to identify and detect genetic diseases and inborn errors of metabolism or
IEMs, such as phenylketonuria or alkaptonuria (Levy 2010). Metabolites are not
only sensitive to genome-related processes, but also to what happens in the environ-
ment. In particular, metabolite concentrations are influenced by nutrition, exposure
to workplace or household chemicals, physical activity, the time of day, or even the
outside temperature (Bassini and Cameron 2014; Brown 2016).
Because metabolites are affected by what happens intra-cellularly (via the
genome) and extra-cellularly (via the environment), metabolomics provides a
detailed view of the gene–environment interactions. Metabolomics is therefore an
ideal route for scientists to access and measure an organism’s “chemical phenotype”
(Fiehn 2002). This represents an important advantage of metabolomics over geno-
mics analyses. While the genome can suggest what might happen, the metabolome
actually indicates what is happening.
Continued advances in analytical chemistry and computational data analysis have
made the study of metabolomics more accessible to a wider range of scientific
disciplines. These advances have led to metabolomics being routinely used in
disease screening, drug discovery, food and nutritional analysis, veterinary studies,
crop assessment, biomaterial production, and environmental monitoring (Holmes
et al. 2008; Viant 2008; Wishart 2008a, 2016; Kim et al. 2016). Indeed,
metabolomics research has grown exponentially since 1999 which reported just
two metabolomics papers to nearly 9,000 papers published in 2020.
Metabolomics experiments are relatively simple to perform. The general
workflow to collect metabolomics data is shown in Fig. 1. The experiments begin
with a biological sample which can be a biofluid or a tissue. For tissues, the
metabolites must first be extracted or homogenized to produce a fluid. Once an
appropriate metabolite extract or biofluid has been obtained, the liquid sample must
be analyzed by one or more analytical chemistry platforms. The most popular
platforms are liquid chromatography mass spectrometry (LC-MS) and nuclear
magnetic resonance (NMR) spectroscopy. These analytical platforms are ultimately
responsible for helping to identify and/or quantify the chemicals in the different
biological mixtures. With the help of specialized software and carefully developed
databases of compounds, the data generated from these platforms can be used to
identify hundreds of compounds in the biological samples.
While LC-MS methods account for >70% of published metabolomics studies to
date, NMR-based methods still garner considerable interest among metabolomics
researchers. For instance, more than 1,200 NMR-based metabolomics papers were
published in 2020, the most ever published in any given year. This suggests that
NMR-based metabolomics is growing, and it still has plenty to offer to the
metabolomics community. NMR has some unique advantages compared to other
4 D. S. Wishart et al.

Fig. 1 A simplified workflow for metabolomics. Tissue samples may be obtained and
homogenized, or biofluids such as urine or blood may be collected. Spectra of the liquid portion
of the samples can be acquired using NMR or mass spectrometry, and the resulting data used for
analysis

platforms such as LC-MS or gas chromatography mass spectrometry (GC-MS). The

most obvious advantage is its non-destructive nature. Moreover, NMR is
non-biased, supports accurate metabolite quantification, requires little or no com-
pound separation, allows the identification of novel compounds, and does not require
chemical derivatization. Furthermore, NMR is highly automatable and is exception-
ally reproducible, making automated, high-throughput metabolomics studies with
NMR spectroscopy much more feasible than with LC-MS or GC-MS. Moreover,
NMR can detect and characterize compounds that are difficult for LC-MS, such as
sugars, organic acids, alcohols, polyols, and other highly polar or low molecular
weight compounds.
In this chapter, we will provide an overview of NMR-based metabolomics with
an emphasis on the practical aspects of NMR-based applications. First, we will
discuss a brief history of NMR-based metabolomics followed by a short discussion
of the two different approaches to metabolomics, targeted vs. untargeted. Next, a
discussion of the practical aspects of NMR sample preparation, as well as NMR
spectral recording and acquisition will be provided. Then, we will explain how to
interpret NMR spectra collected from biological samples using targeted or
untargeted data analyses and provide references to several software packages to
Practical Aspects of NMR-Based Metabolomics 5

help users process NMR spectra obtained via these two different approaches. We
also illustrate several examples of where and how NMR-based metabolomics has
been successfully implemented. Finally, we speculate on the future of NMR-based
metabolomics and the potential areas of growth for this field.

2 A Short History of NMR-Based Metabolomics

For the past 40 years, NMR has played a central role in the understanding of
metabolism and metabolomic processes. The first example demonstrating how
NMR could be used in metabolic studies was published in 1974 when a 13C
isotope-tracer analysis was combined with NMR studies to decipher specific details
of ethanol metabolism (Wilson and Burlingame 1974). Since then, steady
improvements in NMR technology, NMR field strength, and other advancements
have increased the popularity of NMR for many applications in biochemistry and
metabolism. Indeed, NMR quickly became the tool of choice for many metabolism
research areas, particularly those studying drug metabolism. During the 1970s and
1980s, NMR spectroscopy was used in studies to explore drug kinetics, drug
metabolism, and the identification of drug metabolites (Midgley and Hawkins
1978; Williams et al. 1979). These studies were complemented by more traditional
metabolic studies, focusing on cellular, microbial, plant, and animal metabolism
using in vivo or in vitro 1H, 13C, and 31P NMR techniques (Cohen et al. 1979;
Weiner et al. 1989; Rothman et al. 2003). The widespread use of NMR for “classi-
cal” metabolic studies combined with its exceptional capacity to handle complex
metabolomic mixtures made NMR the preferred analytical chemistry platform for
launching the new field of metabolomics. Indeed, the very first metabolomics
(or metabonomics) studies were conducted using NMR (Bock 1982; Yoshikawa
et al. 1982; Bales et al. 1984a, b). By the late 1980s, NMR-based metabolomics
studies of human plasma led to the identification of several putative biomarkers for
cancer and coronary artery diseases (Fossel et al. 1986; Otvos et al. 1991). These
studies were complemented by pioneering studies by Nicholson and colleagues who
used NMR spectra from urine to characterize inborn errors of metabolism (IEM) and
drug toxicity. These early studies proved that urine, a very complex biofluid, can be
successfully analyzed by NMR (Bales et al. 1984a, b; Nicholson et al. 1984a, b).
The greatest challenge facing these early metabolomics researchers was the sheer
complexity of the NMR spectra they were collecting from biofluids and tissue
extracts. A metabolically rich biofluid such as urine can contain up to 5,000
detectable proton resonances in countless variations depending on the pH or
concentrations of salts in the sample (Nicholson and Wilson 2003). This complexity
of the spectral output led researchers to develop two different approaches for
collecting, processing, analyzing, and interpreting metabolomics NMR data. One
approach called “targeted metabolomics” uses spectral deconvolution software to
identify and quantify fluid-specific or known metabolites in individual NMR spectra.
The second approach called “untargeted metabolomics,” or statistical spectroscopy,
uses spectral alignment, spectral binning, and multivariate statistical analysis to
6 D. S. Wishart et al.

identify spectral features of interest. Once the key features are identified, the
corresponding compounds and metabolites may or may not be identified. Both
targeted and untargeted approaches have their advantages and disadvantages.
While targeted metabolomics is more precise, highly quantitative, and much more
reproducible, it is more time-consuming, more limited in scope and it does not allow
one to identify novel chemicals. Untargeted metabolomics is faster, relatively open
ended, and more useful for identification of novel compounds but it is not quantita-
tive nor is it particularly reproducible. Given the importance of targeted and
untargeted approaches in NMR-based metabolomics and given their requirements
for different types of data analysis techniques, we will discuss them in more detail in
Sect. 4.

3 Practical Aspects of NMR-Based Metabolomics

This section will provide an overview of the practical aspects and consensus
recommendations for conducting liquid-state NMR metabolomics studies with a
primary focus on 1H NMR of biofluids or fluidized tissue extracts (McKay,
“Metabolomics using NMR – avoiding the black box”; Raftery, “Quantitative
NMR methods in metabolomics”). Readers interested in obtaining precise protocols
or information about solid-state NMR or magic angle sample spinning (MAS) NMR
should refer to other excellent reviews and book chapters (Weber et al. 2012; Wolak
et al. 2012; Nagana Gowda and Raftery 2014; Nagana Gowda et al. 2015; Zhang
et al. 2016; Mazzei and Piccolo 2017; Tilgner et al. 2019).

3.1 Sample Preparation

Every metabolomics experiment starts with a biological sample. Regardless of

whether the sample is a tissue or a biofluid, it is important to remember that the
sample is “alive.” Unless the sample is frozen, dried, sterilized, or otherwise devoid
of cells or enzymes, the sample is metabolically active. If not handled properly, this
underlying metabolic activity can lead to deceptive results with large variations in
metabolite composition and concentrations. Therefore, all metabolomics studies
must include a metabolic quenching step. Metabolomic quenching uses either
physical or chemical means to arrest all metabolic activities (Teng et al. 2009).
The easiest method to quench a tissue sample is rapid freezing (using liquid
nitrogen). For blood, the red and white blood cells should be first removed before
freezing to prevent cell lysis upon thawing. For most other biofluids, moderate to
rapid freezing is generally sufficient. In the frozen state, the biological sample may
be stored for months or even years at -80°C (Sellick et al. 2009; Vuckovic 2012).
Prior to analysis, the sample must be thawed carefully and further extracted or
purified before NMR analysis. Typically, different extraction methods or purifica-
tion protocols are required for different samples. Tissue samples are often ground
into a powder in a still-frozen state using a pestle and mortar and then the metabolites
Practical Aspects of NMR-Based Metabolomics 7

are extracted with cold methanol or cold methanol/water or with chloroform (Wolak
et al. 2012; Nagana Gowda and Raftery 2014). The use of an organic solvent, in
addition to extracting the metabolites of interest, quenches unwanted metabolism by
denaturing and precipitating almost all proteins/enzymes within the sample. The
organic solvent extract is then centrifuged to separate the precipitated proteins from
the sample. The sample then must be dried (by freeze-drying/lyophilization) and
then re-dissolved in water or appropriate NMR solvents prior to NMR analysis.
For biofluids, the extraction protocols are much simpler. Cell-free biofluids such
as serum, plasma, saliva, growth media, cell extracts, plant sap, or fecal water
typically require ultrafiltration through a 3–5 kDa molecular weight cut-off filter
that removes higher molecular weight proteins and enzymes (Psychogios et al. 2011;
Nagana Gowda and Raftery 2014). Since no organic solvent is used, no lyophiliza-
tion or evaporation is required, and volatile metabolites, such as formate, acetate,
methanol and ethanol, can be easily detected by NMR. However, ultrafiltration can
cause chemical contamination as the filters contain glycerol or other humectants as
preservatives. These filters must be washed multiple times to remove any traces of
these agents. Ultrafiltration may also artificially lower the concentrations of some
metabolites, such as benzoic acid or tryptophan, which appear to bind to the filter or
are bound to proteins that are removed by the filter (Psychogios et al. 2011; Nagana
Gowda and Raftery 2014). Urine and cerebrospinal fluid samples (as well as other
fluids, such as juice, wine, or beer, that are largely sterile and protein-free) are often
simply filter-sterilized by passing the fluid through a 0.22 μm filter to remove any
cells or organic debris. Often researchers also add a small amount of sodium azide
(an NMR-invisible salt) to the sample to destroy any residual microbial activity.
An alternative method to ultrafiltration or filter-sterilization is organic solvent
extraction. In general, methanol or acetonitrile (for hydrophilic compounds) and
chloroform or methyl-tert-butyl ether (for hydrophobic compounds) can be added to
a sample and used to simultaneously extract metabolites, sterilize the sample, and
precipitate proteins. This method may be applied to serum, plasma, or urine
(Beckonert et al. 2007; Nagana Gowda and Raftery 2014). Solvent extraction may
also enrich for a particular class of chemical compounds (hydrophobic or hydro-
philic compounds) depending on the choice of solvent and/or extraction protocol. As
a general rule, methanol extraction works best for most NMR samples (Lin et al.
2007). However, as organic solvent extraction involves an evaporation step to
remove the organic solvent (which is time consuming), volatile metabolites such
as formic acid, ethanol, or acetic acid may be volatilized and no longer detectable.
Another method to enrich or concentrate particular classes of metabolites uses
stable isotope chemical derivatization. Although more commonly used in MS-based
metabolomics (Gowda et al. 2010), this technique is less widely utilized in
NMR-based metabolomics. Stable isotope labeling can enhance the sensitivity and
resolution by enabling heteronuclear NMR spectroscopy. It can also help to enrich
certain classes of metabolites that contain a specific reactive chemical group (i.e.,
amines or carboxylate groups). Two isotopic tags have gained some popularity in the
NMR community, 15N-ethanolamine and 15N-cholamine (see Fig. 2). Both react
selectively with carboxyl groups (Ye et al. 2009; Tayyari et al. 2013). These isotope
8 D. S. Wishart et al.

Fig. 2 The chemical structures of 15N-labeled ethanolamine and cholamine (top). The amine group
can react with carboxylic acid groups on metabolites, using DMT-MM (4-(4,6-dimethoxy[1,3,5]
triazin-2-yl)-4-methylmorpholinium-chloride) as a catalyst (bottom). 2D 15N-1H HSQC spectra can
then be acquired to detect the tagged metabolites

tags can be used to enhance the detection of organic acids and amino acids and can
also be used in MS-based metabolomics studies.
After the extraction and/or enrichment step is complete, it is critical to use the
right buffer system to adjust the salt concentrations and pH of the sample. For
untargeted approaches, sample uniformity is absolutely required. Ideally all samples
should have identical pH values and identical salt concentrations to ensure uniform
chemical shifts among all metabolites. This uniformity greatly enhances the spectral
alignment. For targeted metabolomics studies, pH and salt concentrations are not as
critical. Nonetheless, buffering the sample with a 50–150 mM potassium phosphate
buffer, maintaining the sample temperature at a constant value (say 25°C), adding a
small amount of D2O as a lock solvent, and ensuring the pH is near 7.0 are all
recommended for targeted profiling with software tools such as Bayesil or Chenomx
(to be discussed later) which have well-defined sample collection conditions
(Mercier et al. 2011; Ravanbakhsh et al. 2015).
Sample preparation for NMR analysis always requires the addition of a chemical
shift reference compound. The International Union of Pure and Applied Chemistry
(IUPAC) and the International Union of Biochemistry and Molecular Biology
(IUBMB) recommend the use of 4,4-dimethyl-4-silapentane-1-sulfonic acid (DSS)
as an internal chemical shift standard for aqueous samples (Wishart et al. 1995;
Harris et al. 2002). Usually DSS is used in its deuterated form DSS-d6. As the DSS
peak at 0.00 ppm is usually well resolved and easily detected, it may also be used as
an internal quantification reference. Trimethylsilylpropanoic acid (TSP) is also
commonly used as a chemical shift standard but is not recommended for
NMR-based metabolomics applications as the chemical shift varies substantially
with pH. Both DSS and TSP can bind to macromolecules such as lipids or proteins
(found in unfiltered serum or plasma), resulting in a broadened signal if these
molecules are present in large quantities. If this occurs, the signal cannot be used
for quantification. An alternative to the internal standard is an electronic reference
signal, often called ERETIC (Electronic REference To access In vivo
Concentrations) (Akoka et al. 1999). The ERETIC method is particularly appealing
since the signal is electronically generated and can be placed in any position in the
Practical Aspects of NMR-Based Metabolomics 9

spectrum. It can also be used to determine absolute concentrations (Watanabe et al.

2016).
After the sample preparation step is complete, the sample must be transferred to
an NMR sample tube. Commonly, 5 mm borosilicate glass tubes (requiring volumes
of 500 to 600 μL) are used for NMR experiments. More recently, many labs have
converted to using 3 mm tubes as less volume (150 to 200 μL) is required while
yielding almost the same signal-to-noise (S/N) ratio as 5 mm tubes. These narrower
tubes also reduce the dielectric loss due to high salt concentrations, making them
more amenable for use with cryo-probes. Shigemi tubes with susceptibility matching
glass plugs can also be used to reduce the sample volume (~250 μL) while retaining
the 5 mm tube diameter that is optimal for 5 mm NMR probes and 5 mm spinners.
Microprobes are also available for volume-limited samples. These microprobes are
optimized for use with 1.0 or 1.7 mm NMR tubes. However, 1.0 and 1.7 mm tubes
are very delicate and difficult to clean (the 1.0 mm tubes are actually disposable since
they are essentially impossible to clean). Nevertheless, with these tubes it is possible
to work with volumes as low as 10 μL for a 1.0 mm tube and 35 μL when using a
1.7 mm tube.
Given the variety of ways biological samples can be collected, prepared, and
processed prior to NMR-based metabolomics analysis, it has long been recognized
that standardized protocols are needed to facilitate proper comparison of data
between samples and between studies (Beckonert et al. 2007; Bernini et al. 2011;
Emwas et al. 2015). As outlined in these papers, there are a relatively small number
of preferred or optimal methods for sample collection, storage, and preparation.
Some of these methods are quite specific to certain biosamples or biofluids, while
others are nearly universal. Regardless of the methods used, it is vital that detailed,
complete, and appropriate information about study design, sample types, sample
collection methods, sample handling, sample processing, and sample storage be
provided. Without this information, it can be very difficult for others to reproduce or
interpret reported results. These factors are especially important if studies involve
multiple locations and multiple laboratories. Toward this end, a number of initiatives
have been launched to improve and standardize sample collection, analysis, and
reporting. These include the Metabolomics Standards Initiative (Fiehn et al. 2007)
along with recent updates (Spicer et al. 2017), COSMOS (Salek et al. 2015), and
more recently SPIDIA and SPIDIA4P (Ghini et al. 2019). Such initiatives should
help guide and encourage members of the metabolomics community to employ
standard protocols and fully report standard procedures to acquire, store, prepare,
process, and report metabolomics data. This kind of standardization helps ensure that
NMR-based metabolomics experiments can be repeated, the collected data can be
re-analyzed, and comparisons can be consistently made between samples, studies,
and laboratories.
10 D. S. Wishart et al.

3.2 Choosing the Right Pulse Sequence

Once the sample is loaded into the NMR spectrometer, the NMR spectrum
(or spectra) can be acquired. Most NMR-based metabolomics studies use 1D 1H
NMR because of its speed and simplicity with which spectra can be acquired,
processed, and interpreted. Two types of pulse sequences or experiments are used
for 1D NMR-based metabolomics: the metnoesy or 1D NOESY (Nuclear
Overhauser Effect Spectroscopy) experiment and the Carr-Purcell-Meiboom-Gill
(CPMG) experiment. The metnoesy experiment is a simple 1D NOESY pulse
sequence that provides solvent suppression before the experiment and during the
mixing time without the use of gradients (see Fig. 3). This simple pulse sequence can
be used with almost any NMR probe (Mckay 2011). The metnoesy pulse sequence
can be modified with more advanced water suppression techniques that use gradients
(i.e., watergate, excitation sculpting) for more robust and effective solvent removal.
With such modifications, more dilute samples can be recorded in less time (McKay
2009). Another advantage of the metnoesy sequence is that shaped pulses can be
used to suppress not only water but additional other strong signals such as ethanol in
wine and beer samples or organic solvents used in the extraction process. As most
databases and deconvolution programs for NMR-based metabolomics were built
using this simplistic metnoesy pulse sequence, this pulse sequence predominates
most metabolomics studies with 1D 1H-NMR.
The second most commonly used 1D NMR pulse sequence is the CPMG experi-
ment (Fig. 4). This pulse sequence can spectroscopically remove signals of large
molecules, such as proteins or lipoproteins, from the spectrum without the need for
ultrafiltration or solvent extraction (Beckonert et al. 2007). The CPMG experiment
takes advantage of the fact that small molecule metabolites and macromolecules
(such as proteins) have different T2 relaxation times. The T2 relaxation time of
macromolecules is very short (milliseconds) while the T2 of metabolites is longer
(seconds). With the CPMG sequence, all molecules with a short T2 are suppressed,
whereas those with a large T2 are unaffected. Under ideal conditions, the CPMG
pulse sequence would eliminate the need for solvent extraction and ultrafiltration,
reducing time and resources required for sample preparation. However, the CPMG

Fig. 3 The pulse sequence for the metnoesy or 1D NOESY experiment. This pulse sequence has a
delay (D1) with a low power presaturation pulse, followed by two successive 90° pulses followed
by a defined mixing time. The pulse sequence concludes with a final 90° pulse followed by the
acquisition period
Practical Aspects of NMR-Based Metabolomics 11

Fig. 4 The pulse sequence for the Carr-Purcell-Meiboom-Gill (CPMG) experiment. This experi-
ment uses repeated pulses of 180° to select signals with long T2 relaxation times and remove signals
with short T2 times

experiment is not perfect and all protein signals are not suppressed, increasing the
time required to properly phase, process, and compare multiple CPMG spectra. As
very few CPMG reference spectra are available in the NMR databases, it makes the
CMPG pulse sequence very difficult to use for spectral deconvolution in targeted
NMR-based metabolomics studies.
More recent developments in NMR-based metabolomics use “pure-shift” pulse
sequences (Moutzouri et al. 2017; Lopez et al. 2019) to simplify overcrowded 1D
1
H-NMR spectra. Pure-shift NMR aims to convert all signals into singlets by
refocusing homonuclear couplings and collapsing multiplet peaks into singlets,
thereby reducing the overlap between compound peaks. However, this class of
experiments are much less sensitive than standard 1H experiments due to the use
of spatial or frequency selective techniques. They also contain artifacts that result
from the pseudo-2D method of data acquisition. The reduced sensitivity, the
increased complexity, and the greater hardware requirements over more conven-
tional 1D NMR experiments have likely prevented their widespread use in
metabolomics. Despite these drawbacks, the improved resolution provided by
these methods has seen their successful application in metabolite profiling (Lopez
et al. 2019).
1D NMR pulse sequences are not the only experiments available to metabolomics
researchers. With 2D NMR, multidimensional data can be recorded from the same
type of nuclei (homonuclear) or different nuclei (heteronuclear). For metabolically
complex samples such as urine with hundreds of different and variable metabolites,
it can be advantageous to use 2D NMR experiments. Indeed homonuclear 2D
experiments, such as 2D 1H COSY (COrrelated SpectroscopY), 2D 1H INADE-
QUATE (Incredible Natural Abundance DoublE QUAntum Transfer Experiment),
or 2D J-resolved experiments have been widely used in NMR-based metabolomics
studies (Martineau et al. 2011; Bingol and Brüschweiler 2014). 2D NMR spectra
allow researchers to more easily identify unknown compounds, characterize novel
compounds, and deconvolute overlapping peaks, which could be problematic for 1D
spectral deconvolution programs. Several tools and databases are available to inter-
pret 2D homonuclear metabolomics data (Bingol et al. 2014, 2016). For complex
mixtures such as those found in metabolomics studies, the 2D J-resolved (JRES)
experiment is particularly attractive due to its simplicity and relatively short
12 D. S. Wishart et al.

acquisition time relative to other 2D experiments (Ludwig and Viant 2010). The
JRES experiment results in singlets in the 1H dimension of the spectrum, effectively
providing a robust, broadband decoupled 1H spectrum, similar to the “pure-shift”
experiments mentioned above. This provides a way to resolve the complex
overlapping signals from a 1D 1H spectrum. The multiplet patterns, however, are
retained in the second dimension of the JRES spectrum, which can further aid in the
identification of the metabolites. The JRES experiment can also be used for metabo-
lite quantification. Further enhancement of metabolite identification can be achieved
using heteronuclear 2D experiments, such as 2D 1H-15N HSQC (Heteronuclear
Single Quantum Coherence Spectroscopy) and 1H-13C HSQC experiments. These
experiments provide additional chemical shift information in the second dimension
(15N/13C) and offer important structural detail and connectivity information (Lewis
et al. 2007; Ye et al. 2009). Similar to the homonuclear 2D metabolomics data,
spectral databases and programs are also available to facilitate the interpretation of
2D heteronuclear metabolomics data (Bingol et al. 2015, 2016).
While 2D NMR experiments offer tremendous advantages, there are at least three
major disadvantages to using 2D or multidimensional NMR for metabolomics. The
first disadvantage is the time required to collect, process, and interpret the data
(hours compared to minutes for the 1D experiments). Second, 2D NMR experiments
are less sensitive, with a lower limit of metabolite detection that is 5-10X higher than
1D NMR (often >50 μM compared to 10 μM). And lastly, obtaining robust absolute
quantification of metabolites from 2D spectra is particularly challenging – although
improvements are being made (Martineau and Giraudeau 2019; Martineau et al.
2020; Hansen et al. 2021).
Since the informational density in 2D NMR experiments is higher than 1D
experiments, there is a considerable interest in developing techniques to reduce the
time required to acquire 2D spectra. A number of different approaches are being
investigated to reduce 2D spectral scanning time (Le Guennec et al. 2014). The first
approach utilizes spectral folding or aliasing to reduce the spectral width. With this
technique, peaks can be folded in empty spaces of the 2D spectrum. The reduction of
the spectral width means less data points are required, shortening the overall
acquisition time. However, additional NMR experiments have to be performed to
determine the true chemical shift of folded peaks (Foroozandeh and Jeannerat 2010).
If the sample pool is the same (such as urine), this only needs to be performed once
with a single representative sample.
A second approach shortens the acquisition time by reducing the delay between
scans or shortening the relaxation time. These experiments are known as band-
selective optimized flip angle short transient (SOFAST) and band-selective excita-
tion short transient (BEST) methods. However, as these experiments rely on spin
diffusion as an effective relaxation mechanism (which is only true for
macromolecules or small molecules in viscous solvents), SOFAST and BEST
experiments can only be used in studies involving living cells, which have a viscous
cytoplasm (Motta et al. 2010).
A third approach is available that dramatically shortens 2D NMR data acquisition
time by combining multiple pulse sequences (such as COSY, NOESY, HSQC) into
Practical Aspects of NMR-Based Metabolomics 13

one supersequence (Kupče and Claridge 2017; Hansen et al. 2021). An example of
such a supersequence is called NOAH (NMR by Ordered Acquisition using 1H
detection). This pulse sequence has enabled 2D data collection in a single measure-
ment and has been used to comprehensively characterize multiple metabolites within
metabolically complex biofluids such as mouse urine.
A fourth method to shorten the experimental time of 2D NMR experiments uses
non-uniform sampling (NUS). With this method, all data points in the indirect
dimension are not recorded. Instead, the recorded data points are randomized and
differentially weighted across the indirect dimension(s). The missing data points are
reconstructed after the data is collected. With the NUS technique, the acquisition
time of 2D and 3D experiments can be reduced by up to 75% while still retaining the
same spectral resolution as a full-time multidimensional experiment. The quality of
the spectrum mainly depends on the algorithm used to reconstruct it (Kazimierczuk
et al. 2010).
The last technique is the ultrafast (UF) 2D NMR. This technique utilizes the
sample height (or length) and generates slices which correlate with different time
points in the second dimension (called spatial encoding). With the UF-NMR tech-
nique, a 2D spectrum can be acquired in a single scan, making it the fastest 2D NMR
technique available. However, using this experiment, a compromise between spec-
tral widths, resolution, and sensitivity is often necessary. Fortunately, the spectral
width limitations of UF-NMR can be addressed by the folding/aliasing method
(mentioned above) and the sensitivity can be improved by increasing the number
of scans (Shrot and Frydman 2009; Tal and Frydman 2010; Pathan et al. 2011).
Compared to 1D NMR, 2D NMR techniques have a number of limitations with
respect to metabolite quantification. Because 2D NMR techniques use more pulses,
they are more sensitive to pulse imperfections. This may result in inconsistencies in
peak intensity and peak volume, limiting the reliability of quantification. Further-
more, for heteronuclear NMR, the 13C or 15N nuclei for most metabolites have a very
broad chemical shift range. An equal excitation of the 13C or 15N complete spectral
region is difficult for higher field (>500 MHz) spectrometers. To achieve equal
excitation, one could use complex pulses or pulse sequences, such as adiabatic
pulses or shaped pulses. However, the use of these pulses leads to inconsistencies
in spectral peak intensities and volumes, which makes quantification challenging. In
addition, highly variable coupling and relaxation times also alter the peak volume in
hard-to-predict ways, thereby limiting the use of 2D techniques for quantitative
analysis. To overcome these limitations, large numbers of 2D spectral calibration
curves must be collected for each type of 2D NMR experiment in order to use them
for accurate metabolite quantification. However, gathering this kind of data is
tedious, time-consuming, and difficult due to the long experimental acquisition
times (Lewis et al. 2007).
In recent years, it has been shown that 13C-HSQC data can be recorded in a way
that is inherently quantitative. This can be done by extrapolation of the signal back to
the initial excitation (time point zero), known as the HSQC0 experiment (which
requires three separate HSQC experiments) or with a quantitative sequence or the
Q-HSQC experiment (which requires four times more scans to achieve the same
14 D. S. Wishart et al.

sensitivity as normal HSQC) or the quicker variant of the Q-HSQC experiment

called the QQ-HSQC experiment (Peterson and Loening 2007; Hu et al. 2011;
Martineau et al. 2013; Sette et al. 2013). These developments are encouraging and
given the rapid progress in recent years, 2D homonuclear and heteronuclear NMR
may become more appealing and more widely used in the near future.

3.3 Spectral Acquisition and Processing

As described above, the selection of the appropriate NMR experiment is obviously

very important for obtaining high-quality NMR data. However, other aspects of
spectral acquisition and data processing also have a significant impact on spectral
quality. These aspects include the ability to obtain sharp, well-shimmed peaks; the
level of digital resolution; the S/N ratio or signal quality; the presence of well-phased
signals; and the ability to obtain flat baselines.
To obtain good quality NMR data, both the instrument and the sample must be
appropriately locked, tuned, matched, and carefully shimmed. Most instruments are
tuned for salt-free solvents (such as deuterochloroform or pure D2O), but
metabolomic samples usually contain relatively high salt concentrations
(50–200 mM NaCl). If the spectrometer probe is improperly tuned or matched,
these high salts could lead to poor performance, noticeable by a low (S/N) ratio, long
excitation pulses, poor solvent suppression, etc. These effects become more signifi-
cant with cryogenically cooled probes compared to room-temperature probes. Thus,
with each NMR experiment, due care and effort must be applied to ensure that the
sample is well locked, and optimal tuning and matching have been achieved. Often
tuning and matching are performed manually. However, newer instruments support
rapid automatic tuning and matching. This improved auto-tune/auto-match makes it
possible to run multiple sample types with different solvents or salt concentrations in
the same run.
Another prerequisite for obtaining good quality NMR spectra is good shimming.
During the shimming process, small electromagnets (so-called shims) are adjusted to
compensate for magnetic inhomogeneity in the superconducting magnet field.
Inhomogeneities in the magnetic field can lead to distorted peaks in the spectrum
(see Fig. 5). Most modern NMR spectrometers support automated shimming and can
reliably adjust the magnet shims to achieve excellent line shapes. For both manual
and automated shimming, the signal of the internal chemical shift standard (such as
DSS and/or the solvent signal) is used to optimize the shims. After the shimming
procedure, the line width of the reference standard (DSS) should be well below 1 Hz
(0.5 Hz is typical). Adjustment of the field and the lock position during the shimming
process can also improve peak shapes and the overall quality of the spectrum.
To obtain high-quality NMR spectra, good digital resolution and a large number
of data points are required. Modern NMR spectrometers are equipped with high-
speed, high memory computers that support rapid and high bandwidth analog to
digital conversion (ADC). With digital oversampling techniques, 64,000 data points
or even 128,000 data points (np) can be collected on modern high field NMR
Practical Aspects of NMR-Based Metabolomics 15

Fig. 5 A simple illustration of the effects of shimming on an NMR signal on the DSS reference
peak. Poor shimming can cause peaks to appear unsymmetric or even appear like multiple peaks

spectrometers. This ensures that the digital resolution will typically be less than
0.25 Hz in the 1H dimension on most high-resolution instruments (see Eq. 1; sweep
widths (sw) range from 6,000 to 10,000 Hz):

sw
Resolution = 2 × ð1Þ
np
Good digital resolution ensures sharp resonances, but good S/N ensures good
sensitivity. NMR is not known to be a particular sensitive technique. Thus,
metabolomics researchers are always looking for ways to improve NMR sensitivity
to extend the lowest detectable metabolite concentration. Increased sensitivity can be
achieved by increasing the magnetic field strength (the higher the better), using
cryogenically cooled probes (which have 2–4 times better S/N compared to room-
temperature probes via the reduction of electronic noise), concentrating the sample,
increasing sample volume, or optimizing the excitation flip angle. However, one of
the simplest approaches to increasing the S/N and lowering the limit of detection is
increasing the number of scans (ns) (as shown in Eq. 2):

S pffiffiffiffiffi
= ns ð2Þ
N
Using Eq. 2, one can see that increasing the scans does not increase the S/N
linearly. With four scans, the sensitivity (S/N) increases just twofold. With
100 scans, the sensitivity increases only 10-fold. The number of scans collected in
a given NMR experiment must be tempered by the time it takes to collect each of
those scans. That time is determined by the repetition period between scans, also
known as relaxation delay. This delay is defined as the sum of the acquisition time
and acquisition delay prior to the next scan. The relaxation delay should be five times
longer than the longitudinal relaxation time (also known as T1). For metabolites, T1
is typically 2–3 s (Bloembergen et al. 1948). Therefore, a relaxation delay of about
15 s is usually enough time for a complete relaxation of all resonances (and
consequently full recovery of signal intensity) between scans. However, employing
such long delays would make data collection incredibly inefficient and severely limit
the number of scans that could be collected. Therefore, shorter repetition times of
16 D. S. Wishart et al.

2–4 s are often used for the majority of NMR-based metabolomics studies
(Beckonert et al. 2007). These shorter relaxation delays represent a reasonable
compromise between trying to maximize signal recovery and maximizing the
number of scans. However, this compromise means that spectral deconvolution
algorithms used in targeted metabolomics cannot use “idealized” or theoretical
reference NMR spectra but must, instead, use reference spectra that have been
experimentally recorded using exactly the same type of acquisition parameters and
short relaxation delays as used in the actual metabolomics experiment (Mercier et al.
2011; Worley and Powers 2014; Ravanbakhsh et al. 2015).
After an NMR spectrum has been collected, it must be properly phased. Phasing
is an NMR spectral adjustment process that is designed to maximize the absorptive
character of NMR peaks over all regions of an NMR spectrum. There are two types
of phasing: zero-order phase correction (frequency-independent) and first-order
(frequency-dependent) phase correction. While zero-order phase correction is inde-
pendent of the position of the peaks, first-order phasing increases linearly with the
offset from the carrier frequency. Both types of phase correction are usually needed
to obtain symmetric, purely absorptive peaks. Accurate phase correction is an
important step in spectral processing of metabolomics data as even small phase
errors can lead to a cascade of problems in downstream spectral processing and post-
spectral analysis affecting targeted or untargeted metabolomics techniques (Emwas
et al. 2018). Several algorithms have been published for automated phasing of NMR
spectra (Chen et al. 2002; de Brouwer 2009; Binczyk et al. 2015; Zorin et al. 2017;
Steimers et al. 2020). Some of these routines have already been implemented in the
operating systems of many modern NMR spectrometers. However, additional man-
ual phasing is often required in NMR-based metabolomics studies since auto-
phasing routines may have difficulty with more spectrally crowded metabolite
spectra. Furthermore, auto-phasing programs can sometimes end up distorting the
entire NMR spectrum while attempting to correct for the residual water signal.
Despite these caveats, auto-phasing is still widely used in the NMR metabolomics
community because it is fast, reasonably reliable and it avoids operator bias.
Baseline correction is another important step in NMR spectral data processing.
Baseline correction yields a more pleasant looking NMR spectrum where signal-free
regions appear as completely flat lines with zero intensity. While baseline correction
is relatively easy for simpler NMR spectra with just a few peaks, it is much more
difficult for NMR spectra containing thousands of peaks with large differences in
peak intensities and peak widths. High-quality baseline correction is critical for
proper spectral alignment (in untargeted metabolomics) and proper quantification
or peak integration (in targeted metabolomics). Like phase correction, small errors in
the baseline correction can lead to significant errors in the quantification of low
abundance metabolites.
There are two general approaches to baseline correction: one involves correction
in the time domain and the other involves correction in the frequency domain.
Baseline correction through the time domain removes corrupted data in the free
induction decay (FID) to decrease the effect of low frequencies. This can be done by
discarding some of the initial data points and recreating them by a technique called
Practical Aspects of NMR-Based Metabolomics 17

Fig. 6 An example of how good shimming, water suppression, phasing, baseline correction, and
chemical shift referencing can make a significant difference to the quality and usability of a 1D
NMR spectrum of a biofluid. The same sample and experimental parameters were used for both
spectra, except shimming and water suppression were adjusted away from optimal for the left
spectrum. The receiver gain was adjusted automatically

back-prediction (Heuer and Haeberlen 1989; Halamek et al. 1994). Baseline correc-
tion in the frequency domain involves selecting the valley or signal-absent regions of
the spectra (either automatically or manually) and fitting these regions with a
polynomial spline function. The corresponding baseline offset values based on this
spline function are then subtracted from the spectrum to yield a corrected baseline
(Golotvin and Williams 2000; Xi and Rocke 2008). Combining both frequency and
time domain methods further improves the quality of the baseline. Baseline correc-
tion routines are available in the operating systems of most modern NMR
instruments, although often the best baseline correction routines for complex spectra
are found in spectral deconvolution tools designed specifically for NMR-based
metabolomics (Weljie et al. 2006; Mercier et al. 2011; Worley and Powers 2014;
Ravanbakhsh et al. 2015).
The final step of NMR data processing is spectra alignment. In this step, the ppm
scale of all the experiments is calibrated by adjusting the position of the internal
standard (DSS) to 0 ppm. Then, the intensities of all the peaks are normalized using
the height of the internal standard peak. An example of a 1D NMR spectrum before
and after proper chemical shift referencing, shimming, phasing, solvent removal, and
baseline correction is shown in Fig. 6.
18 D. S. Wishart et al.

4 Data Analysis

After collecting a set of NMR spectra for a metabolomics study, the next step is data
analysis. The type of data analysis one undertakes depends on the experimental
design, the choice of the metabolomics experiment (targeted vs. untargeted), and the
type of multivariate statistical techniques or software available to the researcher.
Many excellent reviews have been written on experimental design and multivari-
ate statistics for metabolomics and readers who are interested in these topics should
refer to these publications (Madsen et al. 2010; Smolinska et al. 2012; Saccenti et al.
2014; Ebbels et al. 2019; Percival et al. 2020). We would encourage new users or
readers to explore a software tool called MetaboAnalyst to better understand and
visualize multivariate statistics as used in metabolomics (Xia et al. 2009).
MetaboAnalyst is an easy-to-use, web-based tool that is routinely used by approxi-
mately half of the global metabolomics community. It supports a wide range of
downstream data analysis applications with extensive graphics support including
principal component analysis (PCA), analysis of variance (ANOVA), partial least-
squares discriminant analysis (PLS-DA), heat mapping and clustering, biomarker
identification, pathway analysis, power analysis, and time series analysis (Verpoorte,
“Natural products drug discovery: on silica or in-silico?”; Millet, “Prospective
metabolomic studies in precision medicine. The AKRIBEA project”). A screenshot
of the MetaboAnalyst package is shown in Fig. 7. MetaboAnalyst is configured to
work with both targeted and untargeted metabolomics data, as well as with NMR and
MS data. It is also quite unique in that it has a number of freely available NMR
metabolomic data sets that users can test or explore.

4.1 Data Analysis for Targeted Metabolomics

For targeted metabolomics studies, the NMR spectra need to be analyzed using
specialized peak fitting software. This software fits a reference set of NMR spectra
obtained from pure compounds to the NMR spectra of the mixture of compounds
found in the biofluid or extract. By matching the positions and intensities (or area) of
peak clusters, the software identifies each reference spectrum and accurately
determines their concentrations in the biofluid. This approach is often called spectral
deconvolution. An illustration of how spectral deconvolution works is shown in
Fig. 8. Once the compounds are identified and their concentrations have been
precisely quantified, this information can then be compared against known or normal
concentrations for that biofluid or further analyzed to detect significant differences
between one sample (or group of samples) and the next.
The concepts underlying targeted NMR metabolomics emerged independently on
three separate occasions – for three different applications. Targeted metabolomics
was first described in 1991, as a novel approach to perform serum lipid and
lipoprotein particle profiling (Otvos et al. 1991). The protocol uses a reference
spectral library of different lipoprotein subclasses, to identify and quantify 15 differ-
ent subcategories of VLDL (very low density lipoprotein), LDL (low density
Practical Aspects of NMR-Based Metabolomics 19

Fig. 7 A screenshot montage of the MetaboAnalyst website. More than a dozen statistical modules
are available for analyzing metabolomics data

lipoprotein), and HDL (high density lipoprotein) particles (Jeyarajah et al. 2006).
Several variations of the method have been described in the literature, with some
methods using only 11 lipoprotein categories or others using more sophisticated
wavelet deconvolution algorithms (Serrai et al. 1998; Ala-Korpela et al. 2007).
Nevertheless, the method has proven to be particularly simple, fast, and robust
compared to traditional methods of lipoprotein profiling which are less accurate,
provide less information, and require time-consuming, manually intensive, multi-
step separations (McNamara et al. 2006).
Targeted NMR metabolomics emerged for a second time in 1993, when it was
used to identify brain metabolites from localized in vivo NMR spectroscopy
(Provencher 1993). However it was not until 2001 that the method, known as
LCModel, was made generally available (Provencher 2001). The central concept
behind the LCModel approach for metabolite identification is to fit the broad peaks
20 D. S. Wishart et al.

Fig. 8 A schematic illustration of the principles behind spectral deconvolution. The biofluid
spectrum at the top contains a combination of the three individual reference spectra below it. The
challenge in spectral deconvolution is to determine which combination of reference spectra, their
scaling, and positioning, best produces the observed spectrum

obtained from in vivo NMR data to spectral libraries of pure metabolites (called
in vitro basis sets) in which the spectra are artificially broadened and otherwise
mathematically modified to look like those seen for in vivo NMR spectra. LCModel
uses a constrained regularization method to handle differences in phase, baseline,
and line shapes between the in vitro and in vivo spectra. It is able to identify between
10 and 15 metabolites and can accurately estimate the metabolite concentrations and
their uncertainties (Simister et al. 2003; Marliani et al. 2007).
The third occasion in which targeted metabolomics emerged was in 2001, when it
was specifically implemented for small molecule identification and quantification in
biofluids and tissue extracts (i.e., traditional metabolomics) (Wishart et al. 2001).
While similar in principle to the LCModel and lipoprotein characterization software
mentioned above, the actual implementation is somewhat different. This is because
spectral deconvolution of high-resolution NMR spectra consisting of dozens of
small molecules requires the fitting of not just 15–20 broad peaks, but the fitting
of hundreds to thousands of very sharp peaks. Small variations in position, line
width, shape, or intensity due to pH or matrix effects can make the spectral fitting
problem particularly challenging and “ill-conditioned.” Likewise, because of spin-
coupling effects, the NMR spectrum of a small molecule collected at 500 MHz often
differs substantially from one collected at 800 MHz. To address these problems,
thousands of reference NMR spectra must be collected for hundreds of reference
compounds at different pH values and at different NMR spectrometer frequencies.
Additionally, very sophisticated curve fitting programs that robustly handle sparse
matrices and potential singularities must be employed.
Fortunately, a variety of software tools for small molecule NMR spectral
deconvolution have been developed over the past 10 years that make this spectral
fitting process relatively painless. These include commercial tools, such as the
Chenomx NMR Suite (Mercier et al. 2011), Bruker’s JuiceScreener (Monakhova
et al. 2014), WineScreener (Spraul et al. 2015), FoodScreener (https://www.bruker.
Practical Aspects of NMR-Based Metabolomics 21

com/en/products-and-solutions/mr/nmr-food-solutions/food-screener.html), and
in vitro diagnostic research system (IVDr). Bruker’s IVDr system expands the utility
of automated metabolite quantification to several biofluids (urine, cerebrospinal
fluid, plasma, and serum) by standardizing sample analyses through the
incorporation of standard hardware, standard operating procedures, and automated
sample handling and analysis (Bruker Corporation-Bruker 2013). The Bruker “X”-
-Screener and IVDr systems are normally bundled with specific NMR instruments or
licensed on a per-sample basis, making them relatively expensive. A cheaper
alternative to the commercial deconvolution and automated quantification tools are
a number of freely available, non-commercial tools, such as Batman (Hao et al.
2014), Bayesil (Ravanbakhsh et al. 2015), an automated quantification algorithm
(AQuA) (Röhnisch et al. 2018), an automatic method for identification and quantifi-
cation of metabolites (ASICS) (Tardivel et al. 2017) and rDolphin (Cañueto et al.
2018). A screenshot of the Bayesil web server and its standard output is shown in
Fig. 9. As can be seen in this figure, most deconvolution programs provide lists of
compound identities and estimated concentrations along with an interactive display
showing the fit between the observed NMR spectrum and the reference library NMR
spectra. Some of these deconvolution tools are almost fully automated (such as
Bayesil, Chenomx, and the Bruker products) while others require a fair bit of manual
manipulation.
In addition to tools such as Bayesil, Chenomx, and the Bruker “X”-Screener and
IVDr products, which identify and quantify organic compounds, there are now
NMR-based tools for identifying and quantifying inorganic compounds. In many
fields of metabolomics, metal ions are considered to be important metabolites and
this subdiscipline of studying metal ions is called “metallomics.” Normally metal
ions are measured via inductively coupled plasma mass spectrometry (ICP-MS).
However, it is also possible to identify and quantify metal ions and other inorganic
ions from 1H-NMR spectra. This can be done by taking advantage of the effect of
dissolved inorganic ions (invisible by NMR) on organic compound chemical shifts
(visible by NMR). This approach was described in detail by Takis et al. (2017). The
technique required the measurement of 4,000 variable synthetic urine mixtures
containing differing amounts of the most common organic compounds (> 90%
occurrence) and differing amounts of the most abundant inorganic ions in urine
and acquiring 1H-NMR spectra of each of these mixtures. From these spectra, Takis
et al. created an algorithm, called the Urine Shift Predictor, which is able to predict
the concentration-dependent, inorganic ion-induced chemical shifts on different
organic compounds. By measuring these chemical shift changes it is possible to
estimate inorganic ion and metal ion concentrations from 1H urine spectra.
Regardless of whether one measures organic or inorganic metabolite data
(or both), once the list of metabolites and their concentrations has been obtained
from a targeted metabolomics study, it is relatively easy to apply standard multivari-
ate statistics such as PCA or PLS-DA (and other tools found in packages such as
MetaboAnalyst) to identify significantly changed metabolites or to detect specific
sets of metabolites as biomarkers.
22 D. S. Wishart et al.

Fig. 9 A series of screenshots taken from the Bayesil NMR server. Bayesil is a freely available
spectral deconvolution package that can take high-resolution NMR spectra of biofluids, such as
serum, and automatically identify and quantify the compounds in that biofluid. The top image (a)
shows the Bayesil home page. The middle image (b) shows the interactive spectral viewer. The
bottom image (c) shows a selection of the compounds identified from the spectrum

4.2 Data Analysis for Untargeted Metabolomics

Untargeted NMR metabolomics first emerged in the early 1990s when the complex-
ity of high-resolution NMR spectra collected on biofluids seemed to be too daunting
to allow routine compound identification (Gartland et al. 1990). Rather than
attempting to identify compounds via spectral deconvolution as is done with targeted
metabolomics, untargeted metabolomics exploits a field of science known as
chemometrics to simplify, classify, and interpret groups of NMR spectra. More
specifically, chemometrics is a branch of information science that uses mathematical
and statistical methods to identify patterns and extract information from large data
Practical Aspects of NMR-Based Metabolomics 23

Fig. 10 An example of how spectral alignment of NMR spectra can be performed. Typically,
multiple NMR spectra are collected, compared, and then aligned. The top figure shows multiple
superimposed 1D 1H NMR spectra of human serum. The glutamine peak at 2.427 ppm shows a
larger sensitivity to sample conditions between spectra compared to other nearby peaks. The bottom
figure shows the same spectra aligned using a spectral alignment algorithm (Savorani et al. 2010).
The glutamine peak is now better aligned between the spectra

sets collected on analytical instruments, such as UV, IR, and NMR spectrometers.
When chemometrics is applied to NMR data, it is essential to have many NMR
spectra already collected (generally dozens to hundreds of spectra, including both
cases and controls). These spectra must then be aligned and binned using specially
developed statistical approaches. A simple illustration of how spectral alignment is
done is shown in Fig. 10. After the alignment step has been completed, the spectra
must be scaled or normalized so that they can be easily compared. Once the scaling
and normalization are complete, multivariate statistical techniques such as PCA and
PLS-DA can be used to identify interesting spectral regions or clusters of peaks that
differentiate one group of spectra from another (Beckonert et al. 2007; Lindon et al.
2007; Barton et al. 2008).
A method called statistical total correlation spectroscopy or STOCSY has also
been developed which generates a pseudo-2D NMR spectrum representing the
correlation among the peaks in a set of NMR spectra (Cloarec et al. 2005). STOCSY
allows for the identification of peaks from the same compound or sets of compounds
that co-vary among the spectra. In an untargeted analysis, sometimes only a set of
statistically important spectral peaks or features is presented as the final result,
without compound identification. In other cases, compound identification occurs
only on the peaks which show the most significant changes in a particular study. The
final peak identification step may use spectral deconvolution, compound spike-in
methods, or peak look-up tables (Martínez-Arranz et al. 2015). A variety of software
packages for NMR statistical spectroscopy have been developed over the past
24 D. S. Wishart et al.

10–15 years to make this kind of analysis more routine. These include MVAPack
(Worley and Powers 2014), Automics (Wang et al. 2009), and KIMBLE (Verhoeven
et al. 2018).

4.3 Targeted or Untargeted?

For relatively simple biofluids with fewer than 60–70 NMR-detectable compounds
(such as serum, plasma, cerebrospinal fluid, fecal water, juice, or other fruit extracts)
targeted NMR metabolomics techniques that use spectral deconvolution appear to
work very well (Ravanbakhsh et al. 2015). Most of these methods focus on
analyzing 1D 1H NMR spectra, although methods have also been developed to
analyze 2D 1H NMR spectra and 2D heteronuclear spectra (Lewis et al. 2009; Bingol
et al. 2014, 2015). Extensive spectral libraries now exist for essentially all the
NMR-detectable compounds found in these biofluids and a number of the
deconvolution software tools are becoming almost fully automated. Indeed, some
software packages can be extremely fast and robust with >95% of the known
compounds in a given biofluid being identified and accurately quantified within a
few minutes (Mercier et al. 2011; Zheng et al. 2011; Hao et al. 2014; Ravanbakhsh
et al. 2015).
On the other hand, for very complex biofluids, such as cell growth media, cell
lysates, and urine, the corresponding NMR spectra are often too complex for any
existing spectral deconvolution packages. These biofluids can contain between
70–150 NMR-detectable compounds and often less than 50% of the known
compounds can be confidently identified or quantified using spectral deconvolution.
Furthermore, the quality of the annotation is often highly dependent on the skill or
experience of the operator (Sokolenko et al. 2013). Therefore, for complex biofluids,
such as urine, statistical spectroscopy techniques or untargeted NMR approaches
appear to offer the best option for spectral interpretation. These approaches allow
useful results to be obtained with relatively little manual effort.

5 Biological Interpretation

The ultimate goal of acquiring and analyzing NMR metabolomics data is to use the
results to reveal something about the biology of the system under study. This may
involve the identification of biomarkers or biomarker panels, the tabulation of
reference metabolite concentrations for specific samples or sample types, the char-
acterization of activated or deactivated biological pathways, or the acquisition of
insights into the underlying molecular metabolic or biochemical mechanisms
associated with a particular biological condition. Rather than describing the software
or methods used to perform biological interpretation, here we will focus on some of
the more successful applications of NMR-based metabolomics toward biological
interpretation.
Practical Aspects of NMR-Based Metabolomics 25

One of the earliest applications of NMR in the field of metabolomics focused on

drug toxicology (Midgley and Hawkins 1978; Cohen et al. 1979; Williams et al.
1979; Bock 1982; Nicholson et al. 1984a, b, 2002; Weiner et al. 1989; Rothman
et al. 2003). This work led to the formation of the COnsoritum for MEtabonomic
Toxicology (COMET) (Lindon et al. 2005) and became one of the main drivers for
the development and growth of metabolomics (and especially NMR-based
metabolomics) as a field. COMET helped establish baseline concentrations of
metabolites in urine and blood in both human and lab animals and it helped to
develop techniques and biomarker panels for high-throughput NMR-based toxicol-
ogy screening. These pioneering studies demonstrated the potential of NMR and
NMR-derived biomarkers to non-invasively identify and diagnose liver, brain, and
kidney toxicity arising from specific drugs or drug metabolites.
NMR-based metabolomics has long been used to aid in the diagnosis, interpreta-
tion, and monitoring of inborn errors of metabolism (IEMs) (Iles et al. 1984;
Yamaguchi et al. 1984; Griffiths and Edwards 1987; Lutz et al. 2013; Kostidis and
Mikros 2015; Embade et al. 2019). IEMs are rare genetic disorders characterized by
significant changes (several-fold increases or decreases) in the concentration of
specific metabolites that arise from genetic disturbances in normal metabolism.
According to the Human Metabolome Database (Wishart et al. 2007, 2018), there
are more than 400 different metabolites associated with IEMs, of which at least
90 are detectable by NMR (Kostidis and Mikros 2015). While individually rare, as a
group, IEMs have been estimated to occur at a rate of up to 1 in 800 births (Mak et al.
2013). Early detection and identification of IEM disorders is critical, as is the need to
understand which metabolic pathways are affected. So too is the need to iteratively
monitor and adjust treatments throughout a patient’s life. Because NMR is so
quantitative and reproducible, NMR-based metabolomics has been particularly
useful for IEM detection, IEM pathway analysis, and disease monitoring.
More recently, NMR, in combination with mass spectrometry, has been used to
determine the baseline concentrations of dozens to hundreds of metabolites in human
biofluids including cerebrospinal fluid (Wishart et al. 2008), serum (Psychogios et al.
2011), urine (Bouatra et al. 2013), saliva (Dame et al. 2015), and feces (Karu et al.
2018). NMR has also been used to characterize the blood, urine, milk, and ruminal
fluid of dairy cattle as part of a long-term project to characterize the metabolomes of
cows (Saleem et al. 2013; Sundekilde et al. 2014; Foroutan et al. 2019, 2020) and
other livestock (Chapinal et al. 2012; Hailemariam et al. 2014; Goldansaz et al.
2017). These referential values for each of the major metabolomes are being used by
researchers around the world to help interpret metabolite measurements and to
identify important biomarkers of diet, health, and disease.
One notable success story for NMR-based metabolomics has been in the area of
medical diagnostics. Perhaps the first NMR-based metabolomics-related medical
diagnostic test was the NMR LipoProfile test developed by LipoMed (now owned by
LabCorp) in the 1990s. The LipoProfile test uses 1D 1H NMR to measure the
concentrations of lipoprotein particles in blood plasma samples (Otvos et al.
1991). Each lipoprotein class has a specific chemical shift and line shape, which
can be used in a least-squares fit of the plasma samples to calculate the
26 D. S. Wishart et al.

concentrations of each lipoprotein class. Because of the success of the LipoProfile

test, a number of other NMR-based lipid-profiling companies have emerged. These
include Biosfer Teslab (https://biosferteslab.com/), Nightingale Health (https://
nightingalehealth.com), and numares Health (https://www.numares.com/). Biosfer
Teslab uses 2D diffusion ordered spectroscopy (DOSY) NMR to quantify lipopro-
tein concentrations. In a similar manner, numares Health combines NMR with
machine learning to perform lipid-profiling as well as additional NMR-based
renal, oncological, and neurological biomarker tests. Nightingale Health is the
largest and most successful of these NMR-based metabolomics/diagnostic
companies. It offers low-cost, high-throughput NMR-based metabolomics analysis
of blood samples directly to customers or to healthcare service providers. Nightin-
gale currently identifies and quantifies over 200 blood biomarkers including small
molecule metabolites as well as lipids and lipoprotein particles (Soininen et al. 2015;
Würtz et al. 2017). These biomarkers are further combined into “health indicators”
that provide an individual with a quick summary of their overall health and their
risks for particular diseases such as diabetes, hypertension, obesity, and
atherosclerosis.
Another important application of NMR-based metabolomics is the tracking of
small molecules as they travel through various metabolic pathways. The ability of
NMR to detect isotopically labeled molecules, and to determine the position of these
labels at atomic-resolution, makes NMR a particularly valuable tool in the study of
metabolic pathways, their regulation, and the effects of disease on their integrity
(Lane et al. 2011; Fan and Lane 2011a, b, 2016; Saborano et al. 2019). Isotopic
labeling and the quantitative nature of NMR have also led to the development of
NMR-based “fluxomics” which focused on determining the rates of metabolic
reactions within a biological system. The unique needs of fluxomics have also
spurred the development of rapid and quantitative 2D NMR experiments (Massou
et al. 2007).
NMR-based metabolomics has been widely used in the study of both cancer and
neurological diseases. Cancer is now widely regarded as a metabolic disease
(Warburg 1956; Hanahan and Weinberg 2000; Seyfried and Shelton 2010; Wishart
2015). As a result, NMR-based metabolomics has become the technique of choice in
many cancer studies (Bathe et al. 2011; Carrola et al. 2011; Namer et al. 2011;
Teahan et al. 2011; Weljie et al. 2011; Cao et al. 2012; Farshidfar et al. 2012; Eisner
et al. 2013; Wishart 2015; Kim et al. 2019). Other NMR-based studies have been
used to classify tumors or to follow the efficacy of radiation or chemotherapy
treatment (Blankenberg et al. 1997; Chan et al. 2009; Fong et al. 2011; Palmnas
and Vogel 2013). NMR has also been used to identify potential biomarkers for early
stages of Alzheimer’s disease (Kork et al. 2012; Karamanos et al. 2015),
amyotrophic lateral sclerosis (Blasco et al. 2010), or Parkinson’s disease
(Wu et al. 2016), or for early detection of schizophrenia (Kaddurah-Daouk 2006;
Tasic et al. 2017). Non-invasive techniques such as NMR offer the promise of
detecting and halting the progression of these diseases at an early stage.
NMR-based metabolomics is not limited to biomedical studies or biomedical
applications. It has been widely applied to food safety and food origin studies
Practical Aspects of NMR-Based Metabolomics 27

(Capitani et al. 2017; Sobolev et al. 2019), and to the composition and quality of
foods, such as wine, beer, oil, juice, milk, honey, and fruit (Wishart 2008b; Melzer
et al. 2013; Kim et al. 2016). NMR-based methods have also been applied to
environmental metabolomics. In particular, NMR-based metabolomics has been
used to explore how organisms respond to environmental changes, pollution, and
climate change (Williams et al. 2009; Simpson and Bearden 2013; Sumner et al.
2015).

6 Conclusion and Future Prospects

The field of metabolomics has been around for 20 years. However, NMR has been a
key part of metabolic studies for more than 40 years. The popularity of NMR as a
metabolomics platform is largely due to its high instrument stability, reproducibility,
simple non-destructive sample preparation, ease of quantification, and its amenabil-
ity to automation. However, NMR is not without its limitations. Compared to MS,
NMR has relatively poor sensitivity, a large instrument footprint, and high up-front
and maintenance costs. NMR also lags behind MS in terms of available
metabolomics libraries, easy-to-use data processing and analysis tools, and easy-
to-use sample preparation kits.
However, progress is being made on almost all the fronts where NMR lags behind
MS. To address the cost and space issues of NMR instrumentation,
non-superconducting, bench-top NMR instruments with field strengths of up to
60 MHz are coming into routine use (Percival et al. 2019; Izquierdo-Garcia et al.
2020) and ones approaching or exceeding 200 MHz (Blümich and Singh 2018) are
being developed. However, these lower field instruments have lower sensitivity and
less spectral resolution than higher field instruments. These disadvantages could be
compensated by adopting 2D UF-NMR techniques (Giraudeau and Frydman 2014)
on bench-top instruments.
At the other extreme, the recent introduction of 1.2 GHz NMR spectrometers
promises to greatly improve the sensitivity and resolution of many NMR-based
metabolomics studies (Schwalbe 2017; Luchinat et al. 2021). These super-high field
instruments will lower the limits of metabolite detection to the high nanomolar range
(compared to 5 μM with lower field NMR instruments) and will likely double the
number of metabolites detectable in biofluids such as serum and urine.
Developments in NMR probe technology are also leading to some exciting
improvements in sensitivity and compound structural elucidation. For instance, the
recently developed 13C-optimized 1.5-mm cryoprobe and a 1H-13C dual-optimized
NMR probe permit 2D 1H-13C HSQC experiments to be collected at natural abun-
dance (Ramaswamy et al. 2013, 2016; Clendinen et al. 2014, 2015).
Another route to improving the sensitivity of NMR is hyperpolarization where
sensitivity enhancements of several thousand-fold (allowing detection of metabolites
in the nM range) are technically possible. While hyperpolarization has been used for
13
C-labeled metabolites (Ardenkjaer-Larsen et al. 2003; Keshari et al. 2010; Lumata
et al. 2015; Dey et al. 2020), the need for specialized equipment and sample
28 D. S. Wishart et al.

preparation have limited its uptake. Another hyperpolarization method called

SABRE-SHEATH (Signal Amplification by Reversible Exchange in SHield Enables
Alignment Transfer to Heteronuclei) hyperpolarizes 15N spins at room temperature
(Truong et al. 2015; Theis et al. 2016). SABRE-SHEATH could revolutionize
metabolomics studies if it could be adapted to NMR-based metabolomics.
Advances in NMR software and NMR data libraries are also helping to move
NMR-based metabolomics forward. The recent development of nmrML (Schober
et al. 2018) as a universal standard for the exchange of NMR spectra could help to
address the issue of limited or incompatible NMR data libraries and modest numbers
of reference NMR spectra by providing a common, easily readable, and shareable
file format. The release of a file format for mass spectrometry in 2011 (Martens et al.
2011) led to an explosion in the number of publicly available MS spectra. The
release and adoption of nmrML by the NMR metabolomics community could
potentially lead to the same result. Surprisingly, NMR has not had a modern data
exchange standard for more than 30 years. Continued development of open-source
or open-access software tools for spectral processing and automated or semi-
automated NMR spectral convolution such as Bayesil (Ravanbakhsh et al. 2015),
BATMAN (Hao et al. 2014), AquA (Röhnisch et al. 2018) and rDolphin (Cañueto
et al. 2018) should also make NMR-based metabolomics more efficient, more user-
friendly, and more appealing to non-NMR specialists.
Another area where software development is expected to have an impact on
NMR-based metabolomics is in the application of machine learning techniques to
analyzing and processing NMR spectra. With the availability of more NMR data and
greater computing power, there has been significant growth in the applications of
machine learning techniques to NMR. Deep-learning techniques, in particular, could
greatly increase the sensitivity and accelerate the analysis of NMR-based
metabolomics workflows. For example, deep neural networks have been used to
denoise NMR spectra using a program called DN-Unet (Wu et al. 2021) and to
efficiently reconstruct NUS NMR spectra (Hansen 2019). This has allowed the
acquisition of very high S/N spectra in a fraction of the time. Bruker has developed
a deep-learning algorithm called sigreg that enables signal recognition detection in
1
H-NMR spectra across multiple complex data sets (Paruzzo et al. 2020). Different
deep-learning approaches have also been used for chemical shift prediction and
molecular structure elucidation of small molecules and related metabolites (Cobas
2020). These deep-learning applications typically require massive training data sets
to yield useful results. Given the current paucity of large NMR data sets, much work
still needs to be done to maximize the benefits offered by machine learning to
NMR-based metabolomics research.
One continuing challenge in NMR-based metabolomics lies in the identification
of unknown compounds. As a rule, unknown identification for NMR-based
metabolomics is not as difficult as it is for MS-based metabolomics. This is because
most of the unknowns detected by NMR must have relatively high concentrations
(>5 mM) and consequently are more likely to be “known unknowns” rather than
“unknown unknowns.” The term “known unknowns” refers to compounds that have
been previously characterized or which already exist in chemical databases, but for
Practical Aspects of NMR-Based Metabolomics 29

which no reference NMR (or MS) spectra exist. On the other hand, “unknown
unknowns” are compounds that are completely novel and have never been described
before. To identify “known unknowns,” a targeted NMR approach is often possible.
For instance, one can use standard 2D NMR approaches to determine the unknown’s
molecular constituents or atomic connectivities which might lead to a possible match
to an existing/known structure. Alternately, one can use a “suspect screening”
approach where intelligent guesses (guided by chemical shift similarities) and the
spiking in of pure standard of the suspected compound into the sample can help
confirm its identity.
Obviously, the most appropriate solution to identifying “known unknowns” is
adding more high-quality, experimentally collected, fully assigned reference NMR
data for many more metabolites to NMR databases such as the HMDB, BMRB, or
nmrshiftdb (McAlpine et al. 2019). Unfortunately, the pace at which reference
experimental NMR spectra are being added to these databases has slowed consider-
ably. In the absence of reference experimental NMR data, the use of predicted NMR
spectra to identify “known unknowns” may be possible. Certainly, continued
advancements in computing power along with improvements to the accuracy of
quantum mechanical and machine learning based predictions of NMR chemical
shifts and NMR coupling constants suggest that computational approaches could
be an appealing new route (Borges et al. 2021).
To identify truly novel compounds (i.e., “unknown unknowns”), de novo struc-
ture determination would be required. De novo structure determination often
requires partial or complete purification of the compound of interest, followed by a
combination of mass spectrometry and 2D NMR analysis of the purified compound
(Garcia-Perez et al. 2020). The process of characterizing novel compounds, how-
ever, can be particularly time consuming and also often limited by the low
concentrations of many unknown compounds. Computational approaches for the
analysis of spectral data and the proposal of candidate structures (through techniques
such as computer-aided structure elucidation – also known as CASE) can accelerate
the de novo structure determination process (Boiteau et al. 2018; Leggett et al.
2019).
The computational and experimental innovations being applied to the identifica-
tion of unknowns by NMR is typical of the remarkable inventiveness and impressive
creativity often seen in the NMR community. No doubt many other innovations in
NMR instrumentation, in NMR spectral collection, or in NMR data analysis that will
benefit NMR-based metabolomics are on the horizon. Certainly, for anyone who has
been in the NMR field for more than a few years, one quickly learns to never
underestimate the potential of NMR spectroscopists to come up with some remark-
able, paradigm-changing innovations. It is probably fair to expect that the next
20 years of NMR-based metabolomics will be as interesting and fruitful as the
past 20 years.

Acknowledgments The authors wish to thank Genome Alberta (a division of Genome Canada),
The Canadian Institutes of Health Research (CIHR), Western Economic Diversification (WED),
and Alberta Innovates Health Solutions (AIHS) for financial support.
30 D. S. Wishart et al.

Glossary
Baseline correction A spectral processing technique that yields a more pleasant looking
NMR spectrum where signal-free regions appear as completely flat
lines with zero intensity
BEST Band-Selective Excitation Transient
Chemometrics A branch of information science that uses mathematical and statistical
methods to identify patterns and extract information from large data
sets collected on analytical instruments, such as UV, IR, and NMR
spectrometers
CPMG Carr-Purcell-Meiboom-Gill
CPMG experiment A pulse that can filter out the signals arising from large molecules,
such as proteins or lipoproteins, from the spectrum (without the need
for ultrafiltration or solvent extraction)
COSY COrrelated SpectroscopY
DSS 4,4-dimethyl-4-silapentane-1-sulfonic acid, a chemical shift reference
compound
ERETIC Electronic REference To access In vivo Concentrations, an electronic
reference signal
HSQC Heteronuclear Single Quantum Coherence Spectroscopy
IEM Inborn error of metabolism. They are rare genetic disorders
characterized by significant changes (several-fold increase or absence)
in the concentration of specific metabolites that result from
disturbances in normal metabolism
INADEQUATE Incredible Natural Abundance DoublE QUAntum Transfer
Experiment
IUPAC The International Union of Pure and Applied Chemistry
IUBMB The International Union of Biochemistry and Molecular Biology
Metabolomics A branch of analytical chemistry that comprehensively characterizes
the molecules in various biofluids and tissues
Metabolites The chemical constituents of the metabolome
Metabolome The complete collection of all chemicals or metabolites found within
cells, biofluids, organs, or organisms
Metnoesy experiment A simple 1D NOESY pulse sequence that provides solvent suppres-
sion before the experiment and during the mixing time without the use
of gradients
NOESY Nuclear Overhauser Effect Spectroscopy
Phasing An NMR spectral adjustment process that is designed to maximize the
absorptive character of NMR peaks over all regions of an NMR
spectrum
Relaxation delay The sum of the acquisition time and acquisition delay prior to the
next scan
SABRE-SHEATH Signal Amplification by Reversible Exchange in SHield Enables
Alignment Transfer to Heteronuclei
SOFAST Band-Selective Optimized Flip Angle Short Transient
STOCSY Statistical total correlation spectroscopy
T1 Longitudinal relaxation time
T2 Transverse relaxation time
Targeted metabolomics A metabolomics technique that uses spectral deconvolution software
to identify and quantify fluid-specific or targeted metabolites in indi-
vidual spectra
Practical Aspects of NMR-Based Metabolomics 31

TSP Trimethylsilylpropanoic acid, a chemical shift reference compound

Untargeted metabolomics A metabolomics technique that uses spectral alignment, spectral bin-
ning, and multivariate statistical analysis to identify spectral features
of interest

References
Akoka S, Barantin L, Trierweiler M (1999) Concentration measurement by proton NMR using the
ERETIC method. Anal Chem 71(13):2554–2557. https://doi.org/10.1021/ac981422i
Ala-Korpela M et al (2007) The inherent accuracy of 1H NMR spectroscopy to quantify plasma
lipoproteins is subclass dependent. Atherosclerosis 190(2):352–358. https://doi.org/10.1016/j.
atherosclerosis.2006.04.020
Ardenkjaer-Larsen JH et al (2003) Increase in signal-to-noise ratio of >10,000 times in liquid-state
NMR. Proc Natl Acad Sci 100(18):10158–10163. https://doi.org/10.1073/pnas.1733835100
Bales JR, Higham DP et al (1984a) Use of high-resolution proton nuclear magnetic resonance
spectroscopy for rapid multi-component analysis of urine. Clin Chem 30(3):426–432
Bales JR, Sadler PJ et al (1984b) Urinary excretion of acetaminophen and its metabolites as studied
by proton NMR spectroscopy. Clin Chem 30(10):1631–1636
Barton RH et al (2008) High-throughput 1H NMR-based metabolic analysis of human serum and
urine for large-scale epidemiological studies: validation study. Int J Epidemiol 37(Suppl 1):i31–
i40. https://doi.org/10.1093/ije/dym284
Bassini A, Cameron LC (2014) Sportomics: building a new concept in metabolic studies and
exercise science. Biochem Biophys Res Commun 445(4):708–716. https://doi.org/10.1016/j.
bbrc.2013.12.137
Bathe OF et al (2011) Feasibility of identifying pancreatic cancer based on serum metabolomics.
Cancer Epidemiol Biomarkers Prev 20(1):140–147. https://doi.org/10.1158/1055-9965.EPI-
10-0712
Beckonert O et al (2007) Metabolic profiling, metabolomic and metabonomic procedures for NMR
spectroscopy of urine, plasma, serum and tissue extracts. Nat Protoc 2(11):2692–2703. https://
doi.org/10.1038/nprot.2007.376
Bernini P et al (2011) Standard operating procedures for pre-analytical handling of blood and urine
for metabolomic studies and biobanks. J Biomol NMR 49(3):231–243. https://doi.org/10.1007/
S10858-011-9489-1
Binczyk F, Tarnawski R, Polanska J (2015) Strategies for optimizing the phase correction
algorithms in nuclear magnetic resonance spectroscopy. Biomed Eng Online 14(2):S5. https://
doi.org/10.1186/1475-925X-14-S2-S5
Bingol K, Brüschweiler R (2014) Multidimensional approaches to NMR-based metabolomics. Anal
Chem 86(1):47–57. https://doi.org/10.1021/ac403520j
Bingol K et al (2014) Customized metabolomics database for the analysis of NMR 1H-1H TOCSY
and 13C-1H HSQC-TOCSY spectra of complex mixtures. Anal Chem 86(11):5494–5501.
https://doi.org/10.1021/ac500979g
Bingol K et al (2015) Unified and isomer-specific NMR metabolomics database for the accurate
analysis of 13 C – 1 H HSQC spectra. ACS Chem Biol 10(2):452–459. https://doi.org/10.1021/
cb5006382
Bingol K et al (2016) Comprehensive metabolite identification strategy using multiple
two-dimensional NMR spectra of a complex mixture implemented in the COLMARm web
server. Anal Chem 88(24):12411–12418. https://doi.org/10.1021/acs.analchem.6b03724
Blankenberg FG et al (1997) Quantitative analysis of apoptotic cell death using proton nuclear
magnetic resonance spectroscopy. Blood 89(10):3778–3786
32 D. S. Wishart et al.

Blasco H et al (2010) 1H-NMR-based metabolomic profiling of CSF in early amyotrophic lateral

sclerosis. PLoS One 5(10):e13223. https://doi.org/10.1371/journal.pone.0013223
Bloembergen N, Purcell EM, Pound RV (1948) Relaxation effects in nuclear magnetic resonance
absorption. Phys Rev 73(7):679–712. https://doi.org/10.1103/PhysRev.73.679
Blümich B, Singh K (2018) Desktop NMR and its applications from materials science to organic
chemistry. Angew Chem Int Ed 57(24):6996–7010. https://doi.org/10.1002/anie.201707084
Bock JL (1982) Analysis of serum by high-field proton nuclear magnetic resonance. Clin Chem
28(9):1873–1877
Boiteau R et al (2018) Structure elucidation of unknown metabolites in metabolomics by combined
NMR and MS/MS prediction. Metabolites 8(1):8. https://doi.org/10.3390/METABO8010008
Borges R et al (2021) Quantum chemistry calculations for cetabolomics. Chem Rev 121(10):
5633–5670. https://doi.org/10.1021/ACS.CHEMREV.0C00901
Bouatra S et al (2013) The human urine metabolome. PLoS One 8(9):e73076. https://doi.org/10.
1371/journal.pone.0073076
Brown SA (2016) Circadian metabolism: from mechanisms to metabolomics and medicine. Trends
Endocrinol Metab 27(6):415–426. https://doi.org/10.1016/j.tem.2016.03.015
Bruker Corporation-Bruker (2013) Bruker announces AVANCE™-IVDr as a standardized NMR
platform for clinical screening and in vitro diagnostics (IVD) discovery and validation. https://ir.
bruker.com/press-releases/press-release-details/2013/Bruker-Announces-AVANCE-IVDr-as-a-
Standardized-NMR-Platform-for-Clinical-Screening-and-In-Vitro-Diagnostics-IVD-Discov
ery-and-Validation/default.aspx. Accessed 21 Sept 2021
Cañueto D et al (2018) rDolphin: a GUI R package for proficient automatic profiling of 1D
1 H-NMR spectra of study datasets. Metabolomics 14(3):24. https://doi.org/10.1007/s11306-
018-1319-y
Cao M et al (2012) NMR-based metabolomic analysis of human bladder cancer. Anal Sci 28(5):
451–456. https://doi.org/10.2116/analsci.28.451
Capitani D, Sobolev AP, Mannina L (2017) Nuclear magnetic resonance – metabolomics. In:
Georgiou CA, Danezis GP (eds) Food authentication: management, analysis and regulation.
Wiley, Chichester, pp 177–197. https://doi.org/10.1002/9781118810224.ch6
Carrola J et al (2011) Metabolic signatures of lung cancer in biofluids: NMR-based metabonomics
of urine. J Proteome Res 10(1):221–230. https://doi.org/10.1021/pr100899x
Chan ECY et al (2009) Metabolic profiling of human colorectal cancer using high-resolution magic
angle spinning nuclear magnetic resonance (HR-MAS NMR) spectroscopy and gas chromatog-
raphy mass spectrometry (GC/MS). J Proteome Res 8(1):352–361. https://doi.org/10.1021/
pr8006232
Chapinal N et al (2012) The association of serum metabolites in the transition period with milk
production and early-lactation reproductive performance. J Dairy Sci 95(3):1301–1309. https://
doi.org/10.3168/jds.2011-4724
Chen L et al (2002) An efficient algorithm for automatic phase correction of NMR spectra based on
entropy minimization. J Magn Reson 158(1–2):164–168. https://doi.org/10.1016/S1090-7807
(02)00069-1
Clendinen CS et al (2014) 13C NMR metabolomics: applications at natural abundance. Anal Chem
86(18):9242–9250. https://doi.org/10.1021/ac502346h
Clendinen CS et al (2015) An overview of methods using (13)C for improved compound identifi-
cation in metabolomics and natural products. Front Plant Sci 6:611. https://doi.org/10.3389/fpls.
2015.00611
Cloarec O et al (2005) Statistical total correlation spectroscopy: an exploratory approach for latent
biomarker identification from metabolic 1 H NMR data sets. Anal Chem 77(5):1282–1289.
https://doi.org/10.1021/ac048630x
Cobas C (2020) NMR signal processing, prediction, and structure verification with machine
learning techniques. Magn Reson Chem 58(6):512–519. https://doi.org/10.1002/mrc.4989
Practical Aspects of NMR-Based Metabolomics 33

Cohen SM, Ogawa S, Shulman RG (1979) 13C NMR studies of gluconeogenesis in rat liver cells:
utilization of labeled glycerol by cells from euthyroid and hyperthyroid rats. Proc Natl Acad Sci
U S A 76(4):1603–1609. https://doi.org/10.1073/pnas.76.4.1603
Dame ZT et al (2015) The human saliva metabolome. Metabolomics 11(6):1864–1883. https://doi.
org/10.1007/s11306-015-0840-5
de Brouwer H (2009) Evaluation of algorithms for automated phase correction of NMR spectra. J
Magn Reson 201(2):230–238. https://doi.org/10.1016/j.jmr.2009.09.017
Dey A et al (2020) Hyperpolarized NMR metabolomics at natural 13C abundance. Anal Chem
92(22):14867–14871. https://doi.org/10.1021/acs.analchem.0c03510
Ebbels TMD, De Iorio M, Stephens DA (2019) Statistical methods in metabolomics. In: Balding D,
Moltke I, Marioni J (eds) Handbook of statistical genomics. Wiley, pp 949–975. https://doi.org/
10.1002/9781119487845.ch34
Eisner R et al (2013) A machine-learned predictor of colonic polyps based on urinary
metabolomics. Biomed Res Int 2013:1–11. https://doi.org/10.1155/2013/303982
Embade N et al (2019) NMR-based newborn urine screening for optimized detection of inherited
errors of metabolism. Sci Rep 9(1):13067. https://doi.org/10.1038/s41598-019-49685-x
Emwas A-H et al (2015) Standardizing the experimental conditions for using urine in NMR-based
metabolomic studies with a particular focus on diagnostic studies: a review. Metabolomics
11(4):872–894. https://doi.org/10.1007/S11306-014-0746-7
Emwas A-H et al (2018) Recommended strategies for spectral processing and post-processing of
1D 1H-NMR data of biofluids with a particular focus on urine. Metabolomics 14(3):31. https://
doi.org/10.1007/s11306-018-1321-4
Fan TW-M, Lane AN (2011a) Erratum to: NMR-based stable isotope resolved metabolomics in
systems biochemistry. J Biomol NMR 49:325. https://doi.org/10.1007/s10858-011-9503-7
Fan TWM, Lane AN (2011b) NMR-based stable isotope resolved metabolomics in systems
biochemistry. J Biomol NMR 49(3–4):267–280. https://doi.org/10.1007/s10858-011-9484-6
Fan TWM, Lane AN (2016) Applications of NMR spectroscopy to systems biochemistry. Prog
Nucl Magn Reson Spectrosc 92–93:18–53. https://doi.org/10.1016/j.pnmrs.2016.01.005
Farshidfar F et al (2012) Serum metabolomic profile as a means to distinguish stage of colorectal
cancer. Genome Med 4(5):42. https://doi.org/10.1186/gm341
Fiehn O (2002) Metabolomics – the link between genotypes and phenotypes. Plant Mol Biol
48(1–2):155–171. https://doi.org/10.1023/A:1013713905833
Fiehn O et al (2007) The metabolomics standards initiative (MSI). Metabolomics 3(3):175–178.
https://doi.org/10.1007/S11306-007-0070-6
Fong MY, McDunn J, Kakar SS (2011) Identification of metabolites in the normal ovary and their
transformation in primary and metastatic ovarian cancer. PLoS One 6(5):e19963. https://doi.
org/10.1371/journal.pone.0019963
Foroozandeh M, Jeannerat D (2010) Deciphered chemical shifts in aliased spectra recorded with
two slightly different narrow windows or differential chemical shift evolution. ChemPhysChem
11(12):2503–2505. https://doi.org/10.1002/cphc.201000421
Foroutan A et al (2019) Chemical composition of commercial cow’s milk. J Agric Food Chem
67(17):4897–4914. https://doi.org/10.1021/acs.jafc.9b00204
Foroutan A et al (2020) The bovine metabolome. Metabolites 10(6):233. https://doi.org/10.3390/
metabo10060233
Fossel ET, Carr JM, McDonagh J (1986) Detection of malignant tumors. Water-suppressed proton
nuclear magnetic resonance spectroscopy of plasma. N Engl J Med 315(22):1369–1376. https://
doi.org/10.1056/NEJM198611273152201
Garcia-Perez I et al (2020) Identifying unknown metabolites using NMR-based metabolic profiling
techniques. Nat Protoc 15(8):2538–2567. https://doi.org/10.1038/s41596-020-0343-3
Gartland KP et al (1990) Pattern recognition analysis of high resolution 1H NMR spectra of urine. A
nonlinear mapping approach to the classification of toxicological data. NMR Biomed 3(4):
166–172. https://doi.org/10.1002/nbm.1940030404
34 D. S. Wishart et al.

Ghini V et al (2019) NMR for sample quality assessment in metabolomics. N Biotechnol 52:25–34.
https://doi.org/10.1016/J.NBT.2019.04.004
Giraudeau P, Frydman L (2014) Ultrafast 2D NMR: an emerging tool in analytical spectroscopy.
Annu Rev Anal Chem 7(1):129–161. https://doi.org/10.1146/annurev-anchem-071213-020208
Goldansaz SA et al (2017) Livestock metabolomics and the livestock metabolome: a systematic
review. PLoS One 12(5):e0177675. https://doi.org/10.1371/journal.pone.0177675
Golotvin S, Williams A (2000) Improved baseline recognition and modeling of FT NMR spectra. J
Magn Reson 146(1):122–125. https://doi.org/10.1006/jmre.2000.2121
Gowda GAN et al (2010) Quantitative analysis of blood plasma metabolites using isotope enhanced
NMR methods. Anal Chem 82(21):8983–8990. https://doi.org/10.1021/ac101938w
Griffiths RD, Edwards RH (1987) Magnetic resonance spectroscopy in the recognition of metabolic
disease. J Inherit Metab Dis 10(Suppl 1):147–158
Hailemariam D et al (2014) Identification of predictive biomarkers of disease state in transition
dairy cows. J Dairy Sci 97(5):2680–2693. https://doi.org/10.3168/jds.2013-6803
Halamek J, Vondra V, Kasal M (1994) The elimination of baseline distortions induced by audio
filters. J Magn Reson Ser 110(2):194–197. https://doi.org/10.1006/jmra.1994.1204
Hanahan D, Weinberg R (2000) The hallmarks of cancer. Cell 100(1):57–70. https://doi.org/10.
1016/S0092-8674(00)81683-9
Hansen DF (2019) Using deep neural networks to reconstruct non-uniformly sampled NMR
spectra. J Biomol NMR 73(10–11):577–585. https://doi.org/10.1007/s10858-019-00265-1
Hansen AL et al (2021) 2D NMR-based metabolomics with HSQC/TOCSY NOAH
supersequences. Anal Chem 93(15):6112–6119. https://doi.org/10.1021/acs.analchem.0c05205
Hao J et al (2014) Bayesian deconvolution and quantification of metabolites in complex 1D NMR
spectra using BATMAN. Nat Protoc 9(6):1416–1427. https://doi.org/10.1038/nprot.2014.090
Harris RK et al (2002) NMR nomenclature: nuclear spin properties and conventions for chemical
shifts. IUPAC recommendations 2001. Solid State Nucl Magn Reson 22(4):458–483. https://
doi.org/10.1006/snmr.2002.0063
Heuer A, Haeberlen U (1989) A new method for suppressing baseline distortions in FT NMR. J
Magn Reson (1969) 85(1):79–94. https://doi.org/10.1016/0022-2364(89)90322-3
Holmes E, Wilson ID, Nicholson JK (2008) Metabolic phenotyping in health and disease. Cell
134(5):714–717. https://doi.org/10.1016/j.cell.2008.08.026
Hu K, Westler WM, Markley JL (2011) Simultaneous quantification and identification of individual
chemicals in metabolite mixtures by two-dimensional extrapolated time-zero 1H-13C HSQC
(HSQC 0). J Am Chem Soc 133(6):1662–1665. https://doi.org/10.1021/ja1095304
Iles RA et al (1984) Rapid screening of metabolic diseases by proton NMR. Lancet 2(8413):
1221–1222. https://doi.org/10.1016/S0140-6736(84)92783-1
Izquierdo-Garcia JL et al (2020) Discovery and validation of an NMR-based metabolomic profile in
urine as TB biomarker. Sci Rep 10(1):22317. https://doi.org/10.1038/s41598-020-78999-4
Jeyarajah EJ, Cromwell WC, Otvos JD (2006) Lipoprotein particle analysis by nuclear magnetic
resonance spectroscopy. Clin Lab Med 26(4):847–870. https://doi.org/10.1016/j.cll.2006.
07.006
Kaddurah-Daouk R (2006) Metabolic profiling of patients with schizophrenia. PLoS Med 3(8):
e363. https://doi.org/10.1371/journal.pmed.0030363
Karamanos TK et al (2015) Mechanisms of amyloid formation revealed by solution NMR. Prog
Nucl Magn Reson Spectrosc 88–89:86–104. https://doi.org/10.1016/j.pnmrs.2015.05.002
Karu N et al (2018) A review on human fecal metabolomics: methods, applications and the human
fecal metabolome database. Anal Chim Acta 1030:1–24. https://doi.org/10.1016/j.aca.2018.
05.031
Kazimierczuk K et al (2010) Random sampling in multidimensional NMR spectroscopy. Prog Nucl
Magn Reson Spectrosc 57(4):420–434. https://doi.org/10.1016/j.pnmrs.2010.07.002
Keshari KR et al (2010) Hyperpolarized (13)C spectroscopy and an NMR-compatible bioreactor
system for the investigation of real-time cellular metabolism. Magn Reson Med 63(2):322–329.
https://doi.org/10.1002/mrm.22225
Practical Aspects of NMR-Based Metabolomics 35

Kim S et al (2016) Food metabolomics: from farm to human. Curr Opin Biotechnol 37:16–23.
https://doi.org/10.1016/j.copbio.2015.09.004
Kim E et al (2019) Urine-NMR metabolomics for screening of advanced colorectal adenoma and
early stage colorectal cancer. Sci Rep 9(1):4786. https://doi.org/10.1038/s41598-019-41216-y
Kork F et al (2012) A biomarker for severity of Alzheimer’s disease: 1H-NMR resonances in
cerebrospinal fluid correlate with performance in mini-mental-state-exam. Biomarkers 17(1):
36–42. https://doi.org/10.3109/1354750X.2011.635806
Kostidis S, Mikros E (2015) NMR studies of inborn errors of metabolism. In: eMagRes. Wiley,
Chichester, pp 57–68. https://doi.org/10.1002/9780470034590.emrstm1400
Kupče Ē, Claridge TDW (2017) NOAH: NMR supersequences for small molecule analysis and
structure elucidation. Angew Chem Int Ed 56(39):11779–11783. https://doi.org/10.1002/anie.
201705506
Lane AN et al (2011) Stable isotope-resolved metabolomics (SIRM) in cancer research with clinical
application to nonsmall cell lung cancer. OMICS 15(3):173–182. https://doi.org/10.1089/omi.
2010.0088
Le Guennec A, Giraudeau P, Caldarelli S (2014) Evaluation of fast 2D NMR for metabolomics.
Anal Chem 86(12):5946–5954. https://doi.org/10.1021/ac500966e
Leggett A et al (2019) Identification of unknown metabolomics mixture compounds by combining
NMR, MS, and cheminformatics. Methods Enzymol 615:407–422. https://doi.org/10.1016/BS.
MIE.2018.09.003
Levy PA (2010) An overview of newborn screening. J Dev Behav Pediatr 31(7):622–631. https://
doi.org/10.1097/DBP.0b013e3181eedf01
Lewis IA et al (2007) Method for determining molar concentrations of metabolites in complex
solutions from two-dimensional 1H-13C NMR spectra. Anal Chem 79(24):9385–9390. https://
doi.org/10.1021/ac071583z
Lewis IA, Schommer SC, Markley JL (2009) rNMR: open source software for identifying and
quantifying metabolites in NMR spectra. Magn Reson Chem 47(Suppl 1):S123. https://doi.org/
10.1002/mrc.2526
Lin CY et al (2007) Evaluation of metabolite extraction strategies from tissue samples using NMR
metabolomics. Metabolomics 3(1):55–67. https://doi.org/10.1007/s11306-006-0043-1
Lindon JC et al (2005) The consortium for metabonomic toxicology (COMET): aims, activities and
achievements. Pharmacogenomics 6(7):691–699. https://doi.org/10.2217/14622416.6.7.691
Lindon JC, Holmes E, Nicholson JK (2007) Metabonomics in pharmaceutical R & D. FEBS J
274(5):1140–1151. https://doi.org/10.1111/j.1742-4658.2007.05673.x
Lopez JM, Cabrera R, Maruenda H (2019) Ultra-clean pure shift 1 H-NMR applied to
metabolomics profiling. Sci Rep 9(1):6900. https://doi.org/10.1038/s41598-019-43374-5
Luchinat E et al (2021) Protein in-cell NMR spectroscopy at 1.2 GHz. J Biomol 75(2–3):97–107.
https://doi.org/10.1007/s10858-021-00358-w
Ludwig C, Viant M (2010) Two-dimensional J-resolved NMR spectroscopy: review of a key
methodology in the metabolomics toolbox. Phytochem Anal 21(1):22–32. https://doi.org/10.
1002/PCA.1186
Lumata L et al (2015) Hyperpolarized 13C magnetic resonance and its use in metabolic assessment
of cultured cells and perfused organs. Methods Enzymol 561:73–106. https://doi.org/10.1016/
bs.mie.2015.04.006
Lutz N, Sweedler J, Wevers R (eds) (2013) Methodologies for metabolomics: experimental
strategies and techniques. Cambridge University Press, Cambridge. https://doi.org/10.1017/
CBO9780511996634
Madsen R, Lundstedt T, Trygg J (2010) Chemometrics in metabolomics – a review in human
disease diagnosis. Anal Chim Acta 659(1–2):23–33. https://doi.org/10.1016/j.aca.2009.11.042
Mak CM et al (2013) Inborn errors of metabolism and expanded newborn screening: review and
update. Crit Rev Clin Lab Sci 50(6):142–162. https://doi.org/10.3109/10408363.2013.847896
36 D. S. Wishart et al.

Marliani AF et al (2007) Quantitative proton magnetic resonance spectroscopy of the human

cervical spinal cord at 3 tesla. Magn Reson Med 57(1):160–163. https://doi.org/10.1002/mrm.
21113
Martens L et al (2011) mzML – a community standard form mass spectrometry data. Mol Cell
Proteomics 10(1):R110.000133. https://doi.org/10.1074/mcp.R110.000133
Martineau E, Giraudeau P (2019) Fast quantitative 2D NMR for untargeted and targeted
metabolomics. Methods Mol Biol 2037:365–383. https://doi.org/10.1007/978-1-4939-9690-2_
20
Martineau E et al (2011) Fast and precise quantitative analysis of metabolic mixtures by 2D 1H
INADEQUATE NMR. J Pharm Biomed Anal 54(1):252–257. https://doi.org/10.1016/j.jpba.
2010.07.046
Martineau E et al (2013) Fast quantitative 1H-13C two-dimensional NMR with very high precision.
Anal Chem 85(9):4777–4783. https://doi.org/10.1021/ac4005309
Martineau E, Dumez JN, Giraudeau P (2020) Fast quantitative 2D NMR for metabolomics and
lipidomics: a tutorial. Magn Reson Chem 58(5):390–403. https://doi.org/10.1002/mrc.4899
Martínez-Arranz I et al (2015) Enhancing metabolomics research through data mining. J Proteomics
127(Pt B):275–288. https://doi.org/10.1016/j.jprot.2015.01.019
Massou S et al (2007) NMR-based fluxomics: quantitative 2D NMR methods for isotopomers
analysis. Phytochemistry 68(16–18):2330–2340. https://doi.org/10.1016/j.phytochem.2007.
03.011
Mazzei P, Piccolo A (2017) HRMAS NMR spectroscopy applications in agriculture. Chem Biol
Technol Agric 4(1):11. https://doi.org/10.1186/S40538-017-0093-9
McAlpine J et al (2019) The value of universally available raw NMR data for transparency,
reproducibility, and integrity in natural product research. Nat Prod Rep 36(1):35–107. https://
doi.org/10.1039/C7NP00064B
McKay RT (2009) Chapter 2 recent advances in solvent suppression for solution NMR: a practical
reference. In: Annual reports on NMR spectroscopy. Elsevier, pp 33–76. https://doi.org/10.
1016/S0066-4103(08)00402-X
Mckay RT (2011) How the 1D-NOESY suppresses solvent signal in metabonomics NMR spec-
troscopy: an examination of the pulse sequence components and evolution. Concepts Magn
Reson A 38(5):197–220. https://doi.org/10.1002/cmr.a.20223
McNamara JR, Warnick GR, Cooper GR (2006) A brief history of lipid and lipoprotein
measurements and their contribution to clinical chemistry. Clin Chim Acta 369(2):158–167.
https://doi.org/10.1016/j.cca.2006.02.041
Melzer N, Wittenburg D, Repsilber D (2013) Integrating milk metabolite profile information for the
prediction of traditional milk traits based on SNP information for Holstein cows. PLoS One
8(8):e70256. https://doi.org/10.1371/journal.pone.0070256
Mercier P et al (2011) Towards automatic metabolomic profiling of high-resolution
one-dimensional proton NMR spectra. J Biomol NMR 49(3–4):307–323. https://doi.org/10.
1007/s10858-011-9480-x
Midgley I, Hawkins DR (1978) The use of 13C-nmr spectroscopy for the detection and identifica-
tion of metabolites of carbon- 13 labelled amitriptyline. J Pharm Pharmacol 30(9):547–553.
https://doi.org/10.1111/j.2042-7158.1978.tb13321.x
Monakhova YB et al (2014) Validation studies for multicomponent quantitative NMR analysis: the
example of apple fruit juice. Accred Qual Assur 19(1):17–29. https://doi.org/10.1007/s00769-
013-1026-3
Motta A, Paris D, Melck D (2010) Monitoring real-time metabolism of living cells by fast
two-dimensional NMR spectroscopy. Anal Chem 82(6):2405–2411. https://doi.org/10.1021/
ac9026934
Moutzouri P et al (2017) Ultraclean pure shift NMR. Chem Commun (Camb) 53(73):10188–10191.
https://doi.org/10.1039/c7cc04423b
Nagana Gowda GA, Raftery D (2014) Quantitating metabolites in protein precipitated serum using
NMR spectroscopy. Anal Chem 86(11):5433–5440. https://doi.org/10.1021/ac5005103
Practical Aspects of NMR-Based Metabolomics 37

Nagana Gowda GA, Gowda YN, Raftery D (2015) Expanding the limits of human blood metabolite
quantitation using NMR spectroscopy. Anal Chem 87(1):706–715. https://doi.org/10.1021/
ac503651e
Namer IJ et al (2011) Metabolomic characterization of ovarian epithelial carcinomas by HRMAS-
NMR spectroscopy. J Oncol 2011:174019. https://doi.org/10.1155/2011/174019
Nicholson JK, Wilson ID (2003) Opinion: understanding “global” systems biology: metabonomics
and the continuum of metabolism. Nat Rev Drug Discov 2(8):668–676. https://doi.org/10.1038/
nrd1157
Nicholson JK et al (1984a) Monitoring metabolic disease by proton NMR of urine. Lancet 2(8405):
751–752. https://doi.org/10.1016/S0140-6736(84)92656-4
Nicholson JK et al (1984b) Proton-nuclear-magnetic-resonance studies of serum, plasma and urine
from fasting normal and diabetic subjects. Biochem J 217(2):365–375. https://doi.org/10.1042/
bj2170365
Nicholson JK et al (2002) Metabonomics: a platform for studying drug toxicity and gene function.
Nat Rev Drug Discov 1(2):153–161. https://doi.org/10.1038/nrd728
Oliver S et al (1998) Systematic functional analysis of the yeast genome. Trends Biotechnol 16(9):
373–378. https://doi.org/10.1016/S0167-7799(98)01214-1
Otvos JD, Jeyarajah EJ, Bennett DW (1991) Quantification of plasma lipoproteins by proton
nuclear magnetic resonance spectroscopy. Clin Chem 37(3):377–386
Palmnas MSA, Vogel HJ (2013) The future of NMR metabolomics in cancer therapy: towards
personalizing treatment and developing targeted drugs? Metabolites 3(2):373–396. https://doi.
org/10.3390/metabo3020373
Paruzzo F, Bruderer S, Janjar Y, Bjoern Heitmann B, Bolliger C (2020) Automatic signal region
detection in 1H NMR spectra using deep learning [White paper]. Switzerland
Pathan M et al (2011) “Multi-scan single shot” quantitative 2D NMR: a valuable alternative to fast
conventional quantitative 2D NMR. Analyst 136(15):3157–3163. https://doi.org/10.1039/
c1an15278e
Percival BC et al (2019) Low-field, benchtop NMR spectroscopy as a potential tool for point-of-
care diagnostics of metabolic conditions: validation, protocols and computational models. High-
Throughput 8(1):2. https://doi.org/10.3390/ht8010002
Percival B et al (2020) Chapter 1: univariate and multivariate statistical approaches to the analysis
and interpretation of NMR-based metabolomics datasets of increasing complexity. In: Compu-
tational techniques for analytical chemistry and bioanalysis. Royal Society of Chemistry, pp
1–40. https://doi.org/10.1039/9781788015882-00001
Peterson DJ, Loening NM (2007) QQ-HSQC: a quick, quantitative heteronuclear correlation
experiment for NMR spectroscopy. Magn Reson Chem 45(11):937–941. https://doi.org/10.
1002/mrc.2073
Provencher SW (1993) Estimation of metabolite concentrations from localized in vivo proton NMR
spectra. Magn Reson Med 30(6):672–679. https://doi.org/10.1002/mrm.1910300604
Provencher SW (2001) Automatic quantitation of localized in vivo 1H spectra with LCModel.
NMR Biomed 14(4):260–264. https://doi.org/10.1002/nbm.698
Psychogios N et al (2011) The human serum metabolome. PLoS One 6(2):e16957. https://doi.org/
10.1371/journal.pone.0016957
Ramaswamy V et al (2013) Development of a 13C-optimized 1.5-mm high temperature
superconducting NMR probe. J Magn Reson 235:58–65. https://doi.org/10.1016/j.jmr.2013.
07.012
Ramaswamy V et al (2016) Development of a 1H-13C dual-optimized NMR probe based on
double-tuned high temperature superconducting resonators. IEEE Trans Appl Supercond
26(3):1–5. https://doi.org/10.1109/TASC.2016.2522302
Ravanbakhsh S et al (2015) Accurate, fully-automated NMR spectral profiling for metabolomics.
PLoS One 10(5):1–15. https://doi.org/10.1371/journal.pone.0124219
38 D. S. Wishart et al.

Röhnisch HE et al (2018) AQuA: an automated quantification algorithm for high-throughput

NMR-based metabolomics and its application in human plasma. Anal Chem 90(3):
2095–2102. https://doi.org/10.1021/acs.analchem.7b04324
Rothman DL et al (2003) In vivo NMR studies of the glutamate neurotransmitter flux and
neuroenergetics: implications for brain function. Annu Rev Physiol 65(1):401–427. https://
doi.org/10.1146/annurev.physiol.65.092101.142131
Saborano R et al (2019) A framework for tracer-based metabolism in mammalian cells by NMR. Sci
Rep 9(1):2520. https://doi.org/10.1038/s41598-018-37525-3
Saccenti E et al (2014) Reflections on univariate and multivariate analysis of metabolomics data.
Metabolomics 10(3):361–374. https://doi.org/10.1007/s11306-013-0598-6
Saleem F et al (2013) The bovine ruminal fluid metabolome. Metabolomics 9(2):360–378. https://
doi.org/10.1007/s11306-012-0458-9
Salek R et al (2015) COordination of standards in MetabOlomicS (COSMOS): facilitating
integrated metabolomics data access. Metabolomics 11(6):1587–1597. https://doi.org/10.
1007/S11306-015-0810-Y
Savorani F, Tomasi G, Engelsen SB (2010) Icoshift: a versatile tool for the rapid alignment of 1D
NMR spectra. J Magn Reson 202(2):190–202. https://doi.org/10.1016/j.jmr.2009.11.012
Schober D et al (2018) NmrML: a community supported open data standard for the description,
storage, and exchange of NMR data. Anal Chem 90(1):649–656. https://doi.org/10.1021/acs.
analchem.7b02795
Schwalbe H (2017) Editorial: new 1.2 GHz NMR spectrometers- new horizons? Angew Chem Int
Ed 56(35):10252–10253. https://doi.org/10.1002/anie.201705936
Sellick CA et al (2009) Effective quenching processes for physiologically valid metabolite profiling
of suspension cultured mammalian cells. Anal Chem 81(1):174–183. https://doi.org/10.1021/
ac8016899
Serrai H et al (1998) Quantification of plasma lipoprotein fractions by wavelet transform time-
domain data processing of the proton nuclear magnetic resonance methylene spectral region.
NMR Biomed 11(6):273–280. https://doi.org/10.1002/(SICI)1099-1492(199810)11:6<273::
AID-NBM523>3.0.CO;2-J
Sette M, Lange H, Crestini C (2013) Quantitative HSQC analyses of lignin: a practical comparison.
Comput Struct Biotechnol J 6(7):e201303016. https://doi.org/10.5936/csbj.201303016
Seyfried T, Shelton L (2010) Cancer as a metabolic disease. Nutr Metab 7(1):1–22. https://doi.org/
10.1186/1743-7075-7-7
Shrot Y, Frydman L (2009) Spatial/spectral encoding of the spin interactions in ultrafast multidi-
mensional NMR. J Chem Phys 131(22):224516. https://doi.org/10.1063/1.3266422
Simister RJ et al (2003) A proton magnetic resonance spectroscopy study of metabolites in the
occipital lobes in epilepsy. Epilepsia 44(4):550–558. https://doi.org/10.1046/j.1528-1157.2003.
19102.x
Simpson MJ, Bearden DW (2013) Environmental metabolomics: NMR techniques. In: Harris RK,
Wasylishen RL (eds) eMagRes. Wiley, pp 549–560. https://doi.org/10.1002/9780470034590.
emrstm1352
Smolinska A et al (2012) NMR and pattern recognition methods in metabolomics: from data
acquisition to biomarker discovery: a review. Anal Chim Acta 750:82–97. https://doi.org/10.
1016/j.aca.2012.05.049
Sobolev AP et al (2019) Use of NMR applications to tackle future food fraud issues. Trends Food
Sci Technol 91:347–353. https://doi.org/10.1016/j.tifs.2019.07.035
Soininen P et al (2015) Quantitative serum nuclear magnetic resonance metabolomics in cardiovas-
cular epidemiology and genetics. Circ Cardiovasc Genet 8(1):192–206. https://doi.org/10.1161/
CIRCGENETICS.114.000216
Sokolenko S et al (2013) Understanding the variability of compound quantification from targeted
profiling metabolomics of 1D-1H-NMR spectra in synthetic mixtures and urine with additional
insights on choice of pulse sequences and robotic sampling. Metabolomics 9(4):887–903.
https://doi.org/10.1007/s11306-013-0503-3
Practical Aspects of NMR-Based Metabolomics 39

Spicer R, Salek R, Steinbeck C (2017) A decade after the metabolomics standards initiative it’s time
for a revision. Sci Data 4(1):170138. https://doi.org/10.1038/sdata.2017.138
Spraul M et al (2015) Wine analysis to check quality and authenticity by fully-automated
1 H-NMR. BIO Web Conf 5(23):02022. https://doi.org/10.1051/bioconf/20150502022
Steimers E et al (2020) Application of a new method for simultaneous phase and baseline correction
of NMR signals (SINC). Magn Reson Chem 58(3):260–270. https://doi.org/10.1002/mrc.4964
Sumner LW et al (2015) Modern plant metabolomics: advanced natural product gene discoveries,
improved technologies, and future prospects. Nat Prod Rep 32(2):212–229. https://doi.org/10.
1039/C4NP00072B
Sundekilde UK et al (2014) Association between the bovine milk metabolome and rennet-induced
coagulation properties of milk. J Dairy Sci 97(10):6076–6084. https://doi.org/10.3168/jds.
2014-8304
Takis PG et al (2017) Deconvoluting interrelationships between concentrations and chemical shifts
in urine provides a powerful analysis tool. Nat Commun 8(1):1662. https://doi.org/10.1038/
s41467-017-01587-0
Tal A, Frydman L (2010) Single-scan multidimensional magnetic resonance. Prog Nucl Magn
Reson Spectrosc 57(3):241–292. https://doi.org/10.1016/j.pnmrs.2010.04.001
Tardivel P et al (2017) ASICS: an automatic method for identification and quantification of
metabolites in complex 1D 1H NMR spectra. Metabolomics 13(10):109. https://doi.org/10.
1007/s11306-017-1244-5
Tasic L et al (2017) Metabolomics and lipidomics analyses by1H nuclear magnetic resonance of
schizophrenia patient serum reveal potential peripheral biomarkers for diagnosis. Schizophr Res
185:182–189. https://doi.org/10.1016/j.schres.2016.12.024
Tayyari F et al (2013) 15N-Cholamine – a smart isotope tag for combining NMR- and MS-based
metabolite profiling. Anal Chem 85(18):8715–8721. https://doi.org/10.1021/ac401712a
Teahan O et al (2011) Metabolic signatures of malignant progression in prostate epithelial cells. Int
J Biochem Cell Biol 43(7):1002–1009. https://doi.org/10.1016/j.biocel.2010.07.003
Teng Q et al (2009) A direct cell quenching method for cell-culture based metabolomics.
Metabolomics 5(2):199–208. https://doi.org/10.1007/s11306-008-0137-z
Theis T et al (2016) Direct and cost-efficient hyperpolarization of long-lived nuclear spin states on
universal 15N2-diazirine molecular tags. Sci Adv 2(3):e1501438. https://doi.org/10.1126/
sciadv.1501438
Tilgner M et al (2019) High-resolution magic angle spinning (HRMAS) NMR methods in
metabolomics. Methods Mol Biol 2037:49–67. https://doi.org/10.1007/978-1-4939-9690-2_4
Truong ML et al (2015) 15N hyperpolarization by reversible exchange using SABRE-SHEATH. J
Phys Chem 119(16):8786–8797. https://doi.org/10.1021/acs.jpcc.5b01799
Verhoeven A, Giera M, Mayboroda OA (2018) KIMBLE: a versatile visual NMR metabolomics
workbench in KNIME. Anal Chim Acta 1044:66–76. https://doi.org/10.1016/j.aca.2018.07.070
Viant MR (2008) Recent developments in environmental metabolomics. Mol Biosyst 4(10):
980–986. https://doi.org/10.1039/b805354e
Vuckovic D (2012) Current trends and challenges in sample preparation for global metabolomics
using liquid chromatography-mass spectrometry. Anal Bioanal Chem 403(6):1523–1548.
https://doi.org/10.1007/s00216-012-6039-y
Wang T et al (2009) Automics: an integrated platform for NMR-based metabonomics spectral
processing and data analysis. BMC Bioinformatics 10(1):83. https://doi.org/10.1186/1471-
2105-10-83
Warburg O (1956) On the origin of cancer cells. Science 123(3191):309–314. https://doi.org/10.
1126/SCIENCE.123.3191.309
Watanabe R et al (2016) Quantitative nuclear magnetic resonance spectroscopy based on PULCON
methodology: application to quantification of invaluable marine toxin, okadaic acid. Toxins
8(10):294. https://doi.org/10.3390/toxins8100294
40 D. S. Wishart et al.

Weber F et al (2012) 31P and 13C solid-state NMR spectroscopy to study collagen synthesis and
biomineralization in polymer-based bone implants. NMR Biomed 25(3):464–475. https://doi.
org/10.1002/nbm.1649
Weiner MW et al (1989) Clinical magnetic resonance spectroscopy of brain, heart, liver, kidney,
and cancer. A quantitative approach. NMR Biomed 2(5–6):290–297. https://doi.org/10.1002/
nbm.1940020519
Weljie AM et al (2006) Targeted profiling: quantitative analysis of 1H NMR metabolomics data.
Anal Chem 78(13):4430–4442. https://doi.org/10.1021/ac060209g
Weljie AM et al (2011) 1H NMR metabolomics identification of markers of hypoxia-induced
metabolic shifts in a breast cancer model system. J Biomol NMR 49(3–4):185–193. https://doi.
org/10.1007/s10858-011-9486-4
Williams TH et al (1979) Novel application of proton nuclear magnetic resonance spectroscopy in
the identification of 2’-chloronordiazepam metabolites in the dog. J Med Chem 22(4):436–440.
https://doi.org/10.1021/jm00190a016
Williams TD et al (2009) Hepatic transcriptomic and metabolomic responses in the stickleback
(Gasterosteus aculeatus) exposed to environmentally relevant concentrations of
dibenzanthracene. Environ Sci Technol 43(16):6341–6348. https://doi.org/10.1021/es9008689
Wilson DM, Burlingame AL (1974) Deuterium and carbon-13 tracer studies of ethanol metabolism
in the rat by 2H, 1H-decoupled 13C nuclear magnetic resonance. Biochem Biophys Res
Commun 56(3):828–835. https://doi.org/10.1016/0006-291X(74)90680-9
Wishart DS (2005) Metabolomics: the principles and potential applications to transplantation. Am J
Transplant 5(12):2814–2820. https://doi.org/10.1111/j.1600-6143.2005.01119.x
Wishart DS (2008a) Applications of metabolomics in drug discovery and development. Drugs R D
9(5):307–322. https://doi.org/10.2165/00126839-200809050-00002
Wishart DS (2008b) Metabolomics: applications to food science and nutrition research. Trends
Food Sci Technol 19(9):482–493. https://doi.org/10.1016/j.tifs.2008.03.003
Wishart DS (2015) Is cancer a genetic disease or a metabolic disease? EBioMedicine 2(6):478–479.
https://doi.org/10.1016/j.ebiom.2015.05.022
Wishart D (2016) Emerging applications of metabolomics in drug discovery and precision medi-
cine. Nat Rev Drug Discov 15(7):473–484. https://doi.org/10.1038/nrd.2016.32
Wishart DS et al (1995) 1H, 13C and 15N chemical shift referencing in biomolecular NMR. J
Biomol NMR 6(2):135–140. https://doi.org/10.1007/BF00211777
Wishart DS et al (2001) Magnetic resonance diagnostics: a new technology for high-throughput
clinical diagnostics. Clin Chem 47(10):1918–1921
Wishart DS et al (2007) HMDB: the human metabolome database. Nucleic Acids Res 35(Suppl 1):
D521–D526. https://doi.org/10.1093/nar/gkl923
Wishart DS et al (2008) The human cerebrospinal fluid metabolome. J Chromatogr B Analyt
Technol Biomed Life Sci 871(2):164–173. https://doi.org/10.1016/j.jchromb.2008.05.001
Wishart DS et al (2018) HMDB 4.0: the human metabolome database for 2018. Nucleic Acids Res
46(D1):D608–D617. https://doi.org/10.1093/nar/gkx1089
Wolak J et al (2012) Noninvasive fluxomics in mammals by nuclear magnetic resonance
spectroscopy. In: Fan TW-M, Lane A, Higashi R (eds) The handbook of metabolomics.
Methods in pharmacology and toxicology. Humana Press, Totowa, pp 321–392. https://doi.
org/10.1007/978-1-61779-618-0_11
Worley B, Powers R (2014) MVAPACK: a complete data handling package for NMR
metabolomics. ACS Chem Biol 9(5):1138–1144. https://doi.org/10.1021/cb4008937
Wu J et al (2016) NMR analysis of the CSF and plasma metabolome of rigorously matched
amyotrophic lateral sclerosis, Parkinson’s disease and control subjects. Metabolomics 12(6):
101. https://doi.org/10.1007/s11306-016-1041-6
Wu K et al (2021) Improvement in signal-to-noise ratio of liquid-state NMR spectroscopy via a
deep neural network DN-Unet. Anal Chem 93(3):1377–1382. https://doi.org/10.1021/acs.
analchem.0c03087
Practical Aspects of NMR-Based Metabolomics 41

Würtz P et al (2017) Quantitative serum nuclear magnetic resonance metabolomics in large-scale

epidemiology: a primer on -Omic technologies. Am J Epidemiol 186(9):1084–1096. https://doi.
org/10.1093/aje/kwx016
Xi Y, Rocke DM (2008) Baseline correction for NMR spectroscopic metabolomics data analysis.
BMC Bioinform 9(1):324. https://doi.org/10.1186/1471-2105-9-324
Xia J et al (2009) MetaboAnalyst: a web server for metabolomic data analysis and interpretation.
Nucleic Acids Res 37(Web Server issue):W652–W660. https://doi.org/10.1093/nar/gkp356
Yamaguchi S et al (1984) Rapid screening of metabolic disease by proton NMR urinalysis. Lancet
2(8397):284
Ye T et al (2009) Chemoselective 15N tag for sensitive and high-resolution nuclear magnetic
resonance profiling of the carboxyl-containing metabolome. Anal Chem 81(12):4882–4888.
https://doi.org/10.1021/ac900539y
Yoshikawa K, Matsushita K, Ohsaka A (1982) 1H-NMR spectroscopy in aqueous mediums.
Examination of experimental conditions with human urine as a model sample. Physiol Chem
Phys 14(4):385–389
Zhang B et al (2016) Nanoparticle-assisted removal of protein in human serum for metabolomics
studies. Anal Chem 88(1):1003–1007. https://doi.org/10.1021/acs.analchem.5b03889
Zheng C et al (2011) Identification and quantification of metabolites in 1H NMR spectra by
Bayesian model selection. Bioinformatics 27(12):1637–1644. https://doi.org/10.1093/bioinfor
matics/btr118
Zorin V, Bernstein MA, Cobas C (2017) A robust, general automatic phase correction algorithm for
high-resolution NMR data. Magn Reson Chem 55(8):738–746. https://doi.org/10.1002/mrc.
4586
Compound Identification Strategies in Mass
Spectrometry-Based Metabolomics
and Pharmacometabolomics

Rylan Hissong, Kendra R. Evans, and Charles R. Evans

Contents
1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
2 Step 1: Study Design: Define Compound Identification Goals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
3 Step 2: Acquire Data – Experimental Strategies to Identify Compounds
in Metabolomics Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
4 Step 3: Computational Strategies for Data Cleaning and Feature Annotation . . . . . . . . . . . . . . 54
5 Step 4: MS/MS Libraries and Compound Identification Using Library Search . . . . . . . . . . . . 57
6 Step 5: Assess and Report Identification Confidence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
7 Conclusion: Pharmacology-Focused Compound Identification . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64

Abstract
The metabolome is composed of a vast array of molecules, including endogenous
metabolites and lipids, diet- and microbiome-derived substances,
pharmaceuticals and supplements, and exposome chemicals. Correct identifica-
tion of compounds from this diversity of classes is essential to derive biologically
relevant insights from metabolomics data. In this chapter, we aim to provide a
practical overview of compound identification strategies for mass spectrometry-
based metabolomics, with a particular eye toward pharmacologically-relevant
studies. First, we describe routine compound identification strategies applicable
to targeted metabolomics. Next, we discuss both experimental (data

Rylan Hissong and Kendra R. Evans contributed equally to this work.

R. Hissong · C. R. Evans (✉)

University of Michigan, Ann Arbor, MI, USA
e-mail: [email protected]
K. R. Evans
University of Detroit Mercy, Detroit, MI, USA

# The Author(s), under exclusive license to Springer Nature Switzerland AG 2022 43

V. Ghini et al. (eds.), Metabolomics and Its Impact on Health and Diseases,
Handbook of Experimental Pharmacology 277, https://doi.org/10.1007/164_2022_617
44 R. Hissong et al.

acquisition-focused) and computational (software-focused) strategies used to

identify unknown compounds in untargeted metabolomics data. We then discuss
the importance of, and methods for, assessing and reporting the level of confi-
dence of compound identifications. Throughout the chapter, we discuss how these
steps can be implemented using today’s technology, but also highlight research
underway to further improve accuracy and certainty of compound identification.
For readers interested in interpreting metabolomics data already collected, this
chapter will supply important context regarding the origin of the metabolite
names assigned to features in the data and help them assess the certainty of the
identifications. For those planning new data acquisition, the chapter supplies
guidance for designing experiments and selecting analysis methods to enable
accurate compound identification, and it will point the reader toward best-practice
data analysis and reporting strategies to allow sound biological and pharmaco-
logical interpretation.

Keywords
Compound identification · Identification confidence · LC-MS · Metabolomics ·
Molecular formula assignment · MS/MS search

1 Introduction

Metabolomics is a technique designed to provide a window into the small-molecule

composition of a biological sample. In the context of pharmacological research,
metabolomics can give insight into the uptake, metabolism, and clearance of a drug,
delineate its impact on endogenous metabolism, or be used to discover biomarkers
that predict which individuals will have a favorable response when it is administered.
In all cases, metabolomics data is most meaningful when compounds in the sample
can be accurately quantitated, confidently identified to the level of a unique chemical
structure, and mapped to a biochemical or pharmacological pathway. Mass
spectrometry-based metabolomics can be used to detect anywhere from a single
compound of interest using a targeted method to tens of thousands of features using
an untargeted workflow (Verpoorte, “Natural products drug discovery: on silica or
in-silico?”; Günther, “Metabolomics in cell biology”). However, in a typical
untargeted metabolomics study, a substantial portion of the features detected are
artifacts and only a fraction of the presumed unique features can be readily identified
(Mahieu and Patti 2017). For those features that can be identified, the degree of
confidence with which chemical structures can be assigned varies and is determined
by multiple factors. Therefore, selection of proper study design, data acquisition, and
data analysis methods is essential to achieve the desired compound identification
goals, and should be considered in the context of all the steps of a metabolomics
study.
Compound Identification Strategies in Mass Spectrometry-Based. . . 45

Most studies that use pharmacometabolomics rely on accurate compound identi-

fication, but many manuscripts provide limited detail about how compound
identifications are made or the confidence with which they were assigned. It is
therefore important for researchers in the field of pharmacology seeking to generate
or interpret metabolomics data to be aware of strategies for compound identification
and limitations of the methods. Numerous excellent reviews have already been
published about compound identification strategies for metabolomics (Wishart
2009; Watson 2013; Blaženović et al. 2018; Chaleckis et al. 2019). In this chapter,
we will summarize both basic strategies and innovative techniques that can be used
to facilitate reliable compound identification when using metabolomics. We also
include some discussion of compound identification strategies relevant to mass
spectrometry-based lipidomics, which focuses on study of non-polar lipid and
lipid-like molecules as opposed to polar metabolites. Whenever possible, we will
note examples of studies that demonstrate varied compound identification strategies
in the context of pharmacology and pharmacometabolomics. However, even though
most of the literature focused on small molecule compound identification techniques
does not specifically describe application to pharmacological research, the
techniques apply to all classes of small molecules, whether endogenous, drug-
derived, or of other origin.
This chapter is sub-divided into “steps” structured around a typical metabolomics
(or pharmacometabolomics) study, focusing on how compound identification is
relevant to or implemented at every stage in the workflow. In “Step 1,” we begin
by discussing how to select a workflow to achieve the investigator’s compound
identification-related goals, contrasting compound identification in targeted and
untargeted metabolomics. “Step 2” focuses on experimental data collection
strategies, describing the preparations needed to attempt to identify unknown
features in untargeted metabolomics data. “Step 3” and “Step 4” turn our attention
to data analysis, describing computational strategies developed to aid in identifying
unknown features. The inter-connectedness and inter-dependency of computational
and experimental methods will be highlighted. “Step 5” discusses strategies to assess
confidence in the accuracy of compound identifications and best practices for
reporting compound identifications in publications. The chapter concludes by
summarizing considerations relating to compound identification of particular rele-
vance to pharmacology. Throughout the chapter, we describe some of the major
challenges that prevent universal and complete identification of all features that can
be detected in metabolomics data, and we highlight current research and potential
future developments that may bring this goal closer to fruition.

2 Step 1: Study Design: Define Compound

Identification Goals

The first step of any metabolomics experiment is to define its goals; this also applies
to compound identification. Is the project intended to perform routine identification
and quantitation of a moderate number of well-known metabolic intermediates or a
46 R. Hissong et al.

drug and its known metabolites in human urine? Or, is it to detect and identify
biomarkers of a disease in human plasma, including unknown compounds? Or, to
determine if a particular metabolite profile predicts responsiveness to a treatment?
All these scenarios, and many other uses of metabolomics, rely on robust compound
identification. However, the way compound identification is best achieved is likely
to differ depending on study design. One of the first choices that must be made is
whether the metabolomics workflow to be used should be targeted, untargeted, or
use a combination of both approaches. The distinctions between targeted and
untargeted metabolomics are described in other chapters of this book and have
been reviewed extensively (Patti et al. 2012; Cajka and Fiehn 2016; Fiehn 2016a;
Schrimpe-Rutledge et al. 2016); Saigusa et al. describe advantages and
disadvantages of targeted and untargeted methods in the context of
pharmacometabolomics and drug discovery (Saigusa et al. 2021). Here, we briefly
consider commonalities and distinctions of compound identification in targeted and
untargeted metabolomics.

Compound Identification in Targeted Metabolomics In targeted metabolomics

studies, most of the effort for compound identification is carried out up-front, before
data from biological samples are collected. Methods are set up using authentic
standards that allow detection of a pre-defined set of compounds, based on their
mass, retention time, and in the case of MS/MS-based methods, specific fragment
ions. Once targeted metabolomics methods are created, in principle no additional
steps beyond routine instrumental and data analysis are needed to identify
compounds. The tradeoff for the simplicity of compound identification in targeted
metabolomics is the fact that the number of compounds that can be identified in a
typical targeted analysis is limited. Most targeted assays are used to quantitate from
one to a few hundred compounds, though recent targeted lipidomics methods have
been reported to have the capability to detect and quantitate over 1,000 species using
three separate LC–MS runs per sample (Contrepois et al. 2018). It is also important
to note that identifications in targeted methods are not always free from interference
or error. Structural isomers share the same molecular mass, are often indistinguish-
able by MS/MS fragmentation pattern, and may be difficult to resolve by chroma-
tography. These factors may prevent unambiguous identification in targeted and
untargeted methods alike.
Several guides and reviews have been written that describe how to develop
targeted metabolomics methods (Griffiths et al. 2010; Parker et al. 2014; Zhou and
Yin 2016; Roberts et al. 2012); but a few notes with relevance to compound
identification merit mention. First, it is wise to investigate possible structural isomers
of compounds of interest by using metabolite databases, described in “Step 3” in this
chapter. In some cases, it may be prudent to purchase standards for potential
interfering compounds and confirm they are chromatographically or spectrally
resolved from targeted analytes. Secondly, if a method is to be adopted from the
literature, it is not always safe to assume that provided parameters will produce
results free from interference from isobaric (same nominal mass) species. Authentic
standards should be used to validate methods, preferably by spiking a biological
Compound Identification Strategies in Mass Spectrometry-Based. . . 47

sample with the expected compound at an appropriate concentration to observe an

expected increase in peak area, with no evidence of alteration in peak shape or
retention time that could indicate presence of an interferent. Finally, in the case of
selected reaction monitoring or multiple reaction monitoring methods, it is beneficial
to use both a primary product ion for quantification and secondary ion as a qualifier,
and to monitor the abundance ratio between the two to ensure no deviation from that
observed with an authentic standard.
Targeted metabolomics, or more broadly, targeted mass spectrometry-based
small molecule assays, are widely used for pharmacological applications. Such
applications include quantitation of drug molecules themselves or their direct
metabolites within biological samples. Targeted approaches find use in pharmacoki-
netics studies (Kantae et al. 2017), forensic and clinical toxicology (Maurer 1998),
assays for drugs of abuse (Zaitsu et al. 2016), along with many other applications.
Prakash et al. reviewed strategies for identification of drug metabolites using mass
spectrometry as the primary approach while defining the utility of complementary
methods such as NMR and chemical derivatization (Prakash et al. 2007). Targeted
metabolomics in pharmacological studies also frequently focuses on the study of
endogenous metabolites, which can be used to assess the effect of a drug on specific
endogenous metabolites or metabolic pathways of interest (Kantae et al. 2017;
McCann et al. 2021).

Compound Identification in Untargeted Metabolomics and Lipidomics In con-

trast to targeted metabolomics, which focuses on analysis of specific pre-selected
compounds, untargeted methods aim to quantitate as broad a swath of the
metabolome or lipidome as possible using unbiased data acquisition methods.
Thus, untargeted metabolomics is designed to detect all features in raw LC–MS or
GC–MS data that may originate from a molecule in the sample. Initially, these
features have no assigned chemical identity; they are represented by their measured
physical or chemical properties such as their mass/charge ratio (m/z), retention time
(RT), fragmentation pattern, and/or other measured data. Some features in
untargeted metabolomics data may be identified by matching their m/z and RT,
and when available their MS/MS spectrum or measured ion mobility drift time, to
standards analyzed under identical conditions (preferably in the same lab, on the
same instrument using the same chromatographic and mass spectrometric method).
However, a substantial proportion of features detected in untargeted metabolomics
data typically cannot be identified using this “targeted” strategy, even when many
standards have been catalogued. Some portion of these unidentified features repre-
sent contaminants or degenerate signals; removal of as many of these artifacts as
possible from the data is an important preliminary computational step in compound
identification efforts and is described “Step 3” in this chapter.
After artifact removal, at least a portion of the unknown features may represent
biologically or pharmacologically relevant compounds. Identification of these
features is central to the discovery-driven nature of untargeted metabolomics, and
is a primary goal of many pharmacometabolomics studies (Steuer et al. 2019).
Strategies for unknown feature identification are the focus of the remainder of this
48 R. Hissong et al.

Fig. 1 Generalized workflow for compound identification in untargeted metabolomics illustrating

both experimental and computational approaches

chapter. A generalized workflow for compound identification in untargeted

metabolomics is illustrated in Fig. 1. It contains both experimental and computa-
tional components, which will be described in detail in “Steps 2–4” of this chapter.

3 Step 2: Acquire Data – Experimental Strategies to Identify

Compounds in Metabolomics Data

Once a study design has been selected and goals for compound identification have
been established, the next step is to perform experimental analysis. Here, we
describe major steps of a metabolomics experiment relevant to compound identifi-
cation, pointing out specific strategies that can be employed to improve the number
and quality of identifications that can be obtained from the data.

Sample Preparation Sample preparation for metabolomics plays a substantial role

in determining metabolite recovery, and by extension which metabolites can be
detected and identified (Lu et al. 2008). Solvent extraction is a near-universal
component of metabolomics sample preparation protocols and has been studied
extensively in the context of recovery and quantitation of metabolites (Bruce et al.
2009; Lorenz et al. 2011), though some studies have also investigated how sample
preparation methods modulate the number and chemical class of features that can be
identified (Anderson et al. 2021; Lenz et al. 2007; Koek et al. 2008). One key choice
that can impact compound identification is selecting between a single-phase or
Compound Identification Strategies in Mass Spectrometry-Based. . . 49

biphasic extraction method. While single-phase extracts are simpler to prepare,

multi-phase extracts separate polar and non-polar metabolites into separate
solutions, allowing MS data acquisition parameters and database selection to be
tailored to the polarity of each phase, potentially yielding more and more accurate
identifications (Blaženović et al. 2018; Matyash et al. 2008). Other sample prepara-
tion methods such as solid phase extraction can be used to concentrate
low-abundance compounds and potentially improve identification performance,
though these methods’ impact on compound identification has been less thoroughly
studied (Wu et al. 2019). The simpler strategy of increasing column loading can also
be used to enhance detection and identification of low-abundance compounds.
Anderson et al. demonstrated that increasing sample concentration and injection
volume to achieve column loading over 10-fold higher than typical loading for a
reversed-phase LC–MS metabolomics experiment yielded more and higher-quality
compound identifications, at the expense of a modest loss of chromatographic
resolution (Anderson et al. 2021).

Chemical Derivatization The primary goal of chemical derivatization, a step used

in some but not all metabolomics workflows, is usually to make compounds more
amenable to detection and quantitation by GC–MS or LC–MS. However, it can also
play a role in compound identification. GC–MS-based metabolomics relies on
derivatization to convert non-volatile compounds to species that can be vaporized
and chromatographed in the gas phase. The most widely used approach employs
methoximation followed by addition of trimethylsilyl (TMS) or tert-
butyldimethylsilyl (TBDMS) groups to derivatize ketones, alcohols, amines, and
amides (Lisec et al. 2006; Lee et al. 2005). Fragmentation of TMS- or TBDMS-
derivatized metabolites follow consistent patterns that can be interpreted to help
identify unknown features. Lai et al. reviewed decades of publications describing
GC–MS analysis of TMS-derivatized small molecules and compiled a series of rules
and fragmentation trees useful for assigning substructures and facilitating unknown
identification (Lai and Fiehn 2018). Silylated derivatives are also amenable to
spectral library search, described in “Step 4” below; many GC–MS libraries contain
spectra of an extensive catalog of silylated metabolite derivatives (Kind and Fiehn
2010; Halket et al. 2005). Silylated metabolites, like all compounds amenable to
GC–MS analysis, can also be characterized by their retention index relative to a
series of standards such as n-alkanes or fatty acid methyl esters (Kind et al. 2009;
Strehmel et al. 2008). This allows chromatographic retention information to be
included in GC–MS databases like FiehnLib and the NIST EI-MS library, both
described in more detail below, adding orthogonal information for confirmation of
compound identification (Kind and Fiehn 2010).
Chemical derivatization also has potential application to compound identification
in LC–MS-based metabolomics (Zhao and Li 2020; Han et al. 2015). Multiple
reagents with different selectivity are available. Phenylisothiocyanate (PITC) is a
classic derivatization reagent commonly used for amino acid analysis by LC with
ultraviolet absorbance or fluorescence detection, but has also been used for LC–MS
analysis of species such as catecholamines (Zheng et al. 2018). Similarly, dansyl
50 R. Hissong et al.

chloride and benzoyl chloride react quantitatively with amines, improving ionization
efficiency, chromatographic retention, and MS/MS fragmentation (Guo and Li 2009;
Wong et al. 2016). The compound 3-nitrophenylhydrazine can be used to derivatize
fatty acids, other carboxylic acids, and phosphoryl metabolites (Meng et al. 2021).
Using a combination of derivatization techniques has been proposed as a strategy to
facilitate characterization of the chemical composition of unknowns by revealing the
presence of key functional groups and allowing focused analysis of these subsets of
the metabolome (Zhao and Li 2020). Derivatization also has potential disadvantages:
(1) derivatization of unknowns may not necessarily result in easier characterization
of their structure, and (2) derivatization reactions may not always proceed to
completion, resulting in a mixture of partially derivatized molecules and hence a
more complex sample to analyze, and (3) some metabolites may not derivatize at all.

Stable Isotopes Compounds enriched with one or more stable isotopes of common
elements (e.g., 2H,13C, 15N, or 18O) can be easily distinguished from their unlabeled
counterparts by mass spectrometry, while their chemical properties remain essen-
tially identical to the unlabeled compound. In addition to their applications as
internal standards and as tracers to help quantitate metabolite flux, stable isotopes
can also be used to aid compound identification in metabolomics. One approach
using stable isotopes is termed isotope ratio outlier analysis (IROA), which uses
paired samples isotopically labeled with either 5% or 95% 13C and uses characteris-
tic isotope patterns to help differentiate biological signals from artifacts and con-
strain potential molecular formulas assigned to unknowns (Clendinen et al. 2015).
Another application of stable isotopes is to investigate the metabolic fate of individ-
ual metabolites or drugs (Kempa, “Advancements in pulsed stable isotope re-solved.
Metabolomics”; Günther, “Metabolomics in cell biology”). In one example, Chen
et al. injected mice with 400 mg/kg of [acetyl-2H3]- or [2,3,5,6-2H4]-acetaminophen
and used untargeted feature detection, followed by MS/MS-based structural elucida-
tion, to identify three novel urinary acetaminophen metabolites potentially
associated with its toxicity in overdose (Beyoğlu et al. 2018). Another potential
use of stable isotopes is the use of hydrogen-deuterium exchange to help determine
the chemical structure of unknown compounds (Majuta et al. 2019).

Chromatography Chromatographic methods for metabolomics are often

optimized for quantitation and speed but can also be tailored to improve compound
identification. One important choice is whether to use GC, LC, or both methods to
characterize a sample (Gowda and Djukovic 2014). GC offers fast and reproducible
chromatography and can be coupled to both electron impact (EI) and chemical
ionization (CI) methods to obtain complementary fragmentation and molecular ion
data, both of which contribute to compound identification (Misra and Olivier 2020),
but samples typically require derivatization as described above. LC with electrospray
ionization is more versatile for larger molecules or non-derivatizable species, and
high-resolution accurate mass LC–MS/MS systems are more common than GC–MS
systems with similar capabilities (Shackleton et al. 2018; Theodoridis et al. 2012). In
both GC–MS and LC–MS methods, chromatographic conditions should be
Compound Identification Strategies in Mass Spectrometry-Based. . . 51

evaluated in terms of their potential impact on compound identification, in addition

to considerations regarding quantitation and throughput (Rainville et al. 2017).
Combined optimization of gradient length, sample loading, and use of iterative
precursor ion exclusion for LC–MS/MS analysis of the human plasma metabolome
was determined to generate over a 10-fold improvement of probable unique com-
pound identifications, including numerous trace-level drug compounds, compared to
typical high-throughput methods (Anderson et al. 2021). Multi-hour gradients
(Wang et al. 2015) and extreme column loading are not practical for quantitative
analysis of large numbers of samples; however, features identified using such high-
resolution methods can be aligned with features detected but not identified using
faster run conditions (Habra et al. 2021).

Authentic Standard Libraries and Retention Time Databases As in targeted

metabolomics, authentic standards can be used to help unambiguously identify
features detected in untargeted metabolomics data by matching m/z, RT, and
MS/MS spectrum to features detected in biological samples. Effective use of authen-
tic standard libraries reduces the burden of unknown identification for the remainder
of the features. An authentic standard library can be assembled by purchasing
individual chemical standards from vendors or by purchasing pre-assembled kits
or libraries. Of relevance to pharmacological research, several libraries containing
hundreds to tens of thousands of known bioactive, drug-like, or drug candidate
molecules can be obtained through commercial suppliers and/or government-
supported repositories (Health in Northern Ireland 2022). Disadvantages of authentic
standard libraries include the initial expense of purchasing, solvating, and/or
analyzing the chemical standards, as well as the fact that a library may not be
representative of a biological sample’s constituents (Mahieu and Patti 2017). An
alternative to in-house standard libraries are digital databases containing retention
time or retention index information compiled on a specific analytical method. These
provide some of the benefits of standard libraries without the need to purchase and
analyze all the standards experimentally. One example is the FiehnLib GC–MS
method, which uses retention indices to aid in compound identification (Kind and
Fiehn 2010). Some online metabolite databases also contain RT information for LC
methods, but inter-laboratory use of these data is less frequent as retention time
alignment or indexing across labs is viewed as more challenging than for GC–MS.
Spectral databases are described in more detail in “Step 3” of this chapter.

Mass Spectrometry: High Resolution, Accurate Mass Measurement While

targeted metabolomics is dominated by single-quadrupole GC–MS and tandem-
quadrupole LC–MS, untargeted analysis is more optimally performed using high-
resolution accurate mass (HRAM) analyzers, including quadrupole-time of flight
(QToF), Orbital Ion Trap (Orbitrap), and Fourier-Transform Ion Cyclotron Reso-
nance (FT-ICR) instruments. Measurements performed on a well-calibrated QToF
typically have a mass accuracy in the range of 3–5 ppm when internal calibration is
used; Orbitrap instruments may achieve 0.5–1 ppm mass accuracy, whereas FT-ICR
analyzers may achieve between 0.1 and 1 ppm (Balogh 2004; Fiehn 2016b). A
52 R. Hissong et al.

principal advantage of higher mass accuracy is that it reduces the number of possible
molecular formulas that can be assigned to a feature; although, as demonstrated by
Fiehn et al. even 0.1 ppm mass accuracy is not sufficient to unambiguously assign a
molecular formula for species <500 Da, much less for larger compounds (Kind and
Fiehn 2007). Spectral resolution does not directly determine the accuracy of molec-
ular formula determination. Instead, higher resolution reduces the possibility of
interference from co-eluting nearly isobaric species. The isotope distribution of an
unknown is another feature of MS data that aids compound identification; if
measured accurately, it is very effective in constraining molecular formula assign-
ment (Böcker et al. 2009).

Mass Spectrometry: In-Source Fragmentation In GC–MS studies, electron

impact ionization is performed at a standardized energy of 70 eV, which causes
extensive in-source fragmentation in a manner that is reproducible from instrument
to instrument (Taylor 2015). The fragmentation pattern reveals structural informa-
tion about the compound and allows library searching to facilitate identification
(as described in “Step 4” of this chapter). Co-eluting compounds may complicate
interpretation by generating mass spectra that contain fragment ions from two or
more compounds; one strategy to compensate for this is spectral deconvolution as
implemented in NIST AMDIS software (Davies 1998) or GC–MS/MS analysis
(Kvitvang et al. 2011). In-source fragmentation also occurs to a lesser extent in
LC–ESI–MS but is usually considered undesirable and source settings are chosen to
minimize it, although it is occasionally used to perform pseudo-MS/MS or pseudo-
MS3 for instruments lacking those capabilities. (Xue et al. 2020; Abdelhameed et al.
2014)

Mass Spectrometry: MS/MS and MSn In LC–ESI–MS, tandem mass spectrome-

try (MS/MS or MSn) is an important aspect of data collection in untargeted
metabolomics studies, especially in the context of compound identification.
MS/MS has also seen increasing use in GC–MS, though mainly for targeted analy-
sis. Tandem MS involves selection and fragmentation of specific ions within the
mass spectrometer; fragment ion spectra can be interpreted to discern information
regarding the precursor ion’s structure or used as a “fingerprint” for database
searching. Multiple methods of fragmentation are available, depending on instru-
ment capabilities. These include collision-induced dissociation (CID), higher-energy
collisional dissociation (HCD), electron transfer dissociation (ETD) and infrared
multi-photon dissociation (IRMPD) (Ichou et al. 2014; Alley et al. 2009; Yoo et al.
2007). Recently, a strategy to obtain more extensive fragmentation of ESI-generated
ions that uses an electron impact-type mechanism has been devised; one mode of
operation has been termed electron impact excitation of ions from organics (EIEIO)
(Baba et al. 2018; Ducati et al. 2021). There are two main methods that can be used
to perform tandem MS data acquisition: data-dependent analysis (DDA), in which
MS/MS is automatically triggered for the most abundant or otherwise selected
features detected in the preceding MS1 scan, and data-independent analysis (DIA),
in which MS/MS acquisition is performed according to pre-defined criteria not
Compound Identification Strategies in Mass Spectrometry-Based. . . 53

influenced by the MS1 data. Details regarding these approaches and their advantages
and disadvantages have been described in detail elsewhere (Guo and Huan
2020a, b).
It is good practice to acquire MS/MS data during any substantial untargeted
metabolomics project. This allows both automated database searching and follow-
up analysis to attempt identification or classification of features of interest, without
returning to the instrument to acquire new data. In many cases, MS/MS data need not
be acquired for each individual sample, but can be generated using representative
pooled samples from the study. To acquire high-quality MS/MS spectra of as many
features as possible, recent data acquisition software packages enable automatic
generation of precursor ion exclusion lists that prevent serial LC–MS/MS runs of
the same sample from acquiring MS/MS data on ions already fragmented in previous
runs. This method, sometimes termed iterative DDA, is a powerful strategy to obtain
deeper MS/MS coverage of detected features than is possible in a standard DDA
workflow (Anderson et al. 2021; Koelmel et al. 2017).
Certain mass analyzers are capable of multiple stages of fragmentation, in which
fragment ions are further fragmented. This is termed MSn analysis; it enables
construction of fragmentation trees for unknown compounds (Vaniya and Fiehn
2015). MSn data are less well cataloged and interpretation of MSn data is less widely
supported by databases and software tools compared to MS/MS data. Nevertheless,
under many circumstances MSn data provide useful information to characterize and
annotate unknown compounds (Vinaixa et al. 2016; Ridder et al. 2012).

Ion Mobility Spectrometry and Collisional Cross-Section Measurement Ion

mobility spectrometry (IMS) separates gas phase ions based on differential mobility
through low-pressure buffer gas (Kanu et al. 2008). Ions with a larger cross-sectional
area experience more resistance and move more slowly than smaller, more compact
ions. In the context of metabolomics, ion mobility is performed inside a mass
spectrometer directly preceding the standard mass analyzer(s). It offers a semi-
orthogonal separation to MS and resolves some isobaric species that are not separa-
ble by LC–MS, allowing generation of MS/MS spectra from a single precursor ion
rather than from multiple co-eluting isobars (Rainville et al. 2017). With appropriate
calibration, IMS data can also be used to compute collisional cross-section values
(CCS), which are considered an intrinsic property of an ion and can be used to
facilitate compound identification (Zhou et al. 2020). Various manufacturers have
produced instruments with ion mobility spectrometry capabilities, and collisional
cross-section values are beginning to be included with metabolite databases (Wishart
et al. 2022). CCS measurements have been evaluated as a means of characterizing
drug compound structure (Hines et al. 2017) and to predict their ability to cross the
blood-brain barrier (Guntner et al. 2019), demonstrating the potential pharmacolog-
ical relevance of this rapidly developing technology.

Nuclear Magnetic Resonance Spectroscopy for Metabolite

Identification Although this chapter focuses on mass spectrometry-based methods,
it is important to acknowledge the utility of nuclear magnetic resonance
54 R. Hissong et al.

spectroscopy (NMR)-based methods in identifying unknown compounds. NMR is

an important and widely used technique for metabolomics and
pharmacometabolomics in its own right; its strengths, limitations, and applications
are thoroughly described in the literature (Emwas et al. 2019) and in other chapters
of this book. In the context of compound identification, NMR is a touchstone method
for the fields of synthetic chemistry, natural products chemistry, and drug develop-
ment. For a pure sample of a small molecule compound analyzed in a modern high-
field instrument using 1H and 13C NMR, it is typically possible to assign a definitive
structure via computational modeling and/or manual interpretation (Willoughby
et al. 2014). The primary challenge associated with applying the NMR to compound
identification in metabolomics is its sensitivity. For higher-concentration
compounds, NMR is effective at both quantification and identification, and detailed
guides describing compound identification strategies in NMR metabolomics have
been written (Dona et al. 2016). However, compounds present at lower
concentrations (mid-low micromolar and below) are not amenable to identification
without purification and concentration. Techniques to purify unknowns using chro-
matographic fractionation have been devised (van der Laan et al. 2021; Whiley et al.
2019) but are still not practical for lower-abundance features. While scale-up to
semi-preparative or preparative chromatography or other refinements are possible, it
remains challenging to obtain high-quality NMR spectra of a substantial portion of
features detectable by mass spectrometry.

4 Step 3: Computational Strategies for Data Cleaning and

Feature Annotation

Acquisition of high-quality data is not sufficient to identify unknown metabolites.

Datasets are far too large to manually review, much less interpret, every spectrum
collected in an experiment. Fortunately, numerous computational strategies can be
employed to aid in identifying or annotating unknowns; when necessary, they can
also help guide acquisition of additional data. Here, we present an overview of major
computational strategies to aid the reader in finding and understanding available
tools. Readers interested in a more detailed discussion of computational compound
identification strategies are directed to one of several excellent reviews that have
been published on the topic (Wishart 2009; Watson 2013; Blaženović et al. 2018) or
to primary sources cited below.

Degeneracy Removal In the simplest scenario, each compound present in a sample

would be represented by one feature in experimentally acquired untargeted
metabolomics data. However, this is seldom the case. In GC-EI-MS, extensive
in-source fragmentation is expected for each compound, but when peaks co-elute,
it can be difficult to determine which fragment ions originate from which parent ion.
In LC-ESI-MS, using positive ion mode as an example, singly-protonated [M + H]+
ions are usually the most common form of ion, but other adducts such as [M + Na]+,
[M + NH4]+, etc. or in-source fragment ions such as [M-H2O + H]+ are often formed
Compound Identification Strategies in Mass Spectrometry-Based. . . 55

and detected. More complex ion formation processes are frequently observed,
including multiply charged ions, ions with multiple charge carriers, solvent adducts,
adducted fragment ions, and dimers, multimers and heterodimers. (Mahieu and Patti
2017; Mahieu et al. 2016; Kachman et al. 2019; Nelson et al. 2022) Collectively,
such features can be termed “degenerate” signals.
To avoid investing effort to identify features that are degenerate signals produced
by other already-identified compounds, data clean-up steps including adduct and
fragment annotation and/or removal are necessary. For GC-EI-MS, spectral
deconvolution software such as AMDIS (Davies 1998) or MS-DIAL (Tsugawa
et al. 2015) can help separate spectra of features that are only partially resolved by
chromatography. For LC–MS, the process of adduct annotation can be at least
partially automated by either instrument vendor or open-source data analysis soft-
ware such as CAMERA, MZmine, or MS-DIAL (Tsugawa et al. 2015; Kuhl et al.
2012; Pluskal et al. 2010). The most rigorous approaches for degeneracy annotation
also consider correlation of intensity for co-eluting features; those with high correla-
tion are more likely to represent degenerate features (Kachman et al. 2019;
Broeckling et al. 2014). Application of these clean-up steps reduces compound
identification workload by decreasing the number of features that must be subjected
to additional computational analysis and/or manual review.

Molecular Formula Assignment A fundamental step in identifying a compound,

whenever possible, is to determine its molecular formula. The approach for this
process differs depending on whether data are acquired on a low-resolution instru-
ment, such as a GC–MS with a quadrupole mass analyzer, or a high-resolution
accurate mass instrument, such as a QToF, orbital ion trap, or ion cyclotron
resonance instrument. In the former case, molecular formula assignment is
performed using a probability-based strategy (Scott 1992) such as that implemented
in NIST MS Search software (Stein 1999). In the latter case, an effective strategy for
molecular formula assignment from accurate mass data was developed by Kind et al.
(Kind and Fiehn 2007) It uses “seven golden rules” developed to constrain potential
chemical formulas based on characteristics shared by almost all common biological
molecules containing some or all of the elements C, H, O, N, P, and S. In addition to
the accurate m/z value measured for a compound, the natural isotope distribution of
the compound is required to sufficiently constrain candidate formulas for all but the
smallest metabolites. Fortunately, software tools to assign molecular formulas auto-
matically using metabolomics data have been devised (Tolić et al. 2017; Ludwig
et al. 2019; Dührkop et al. 2019) and are implemented in various instrument vendor
software packages. These molecular formula assignment methods have also proven
highly applicable to pharmaceuticals, returning a correct molecular formula as the
top candidate with 88–99% probability for thousands of database spectra of drug
molecules (Kind and Fiehn 2007). While these algorithms have become increasingly
reliable, de-novo assignment of formula for larger compounds (MW >500) remains
challenging and are best supplemented by support from compound database and
spectral search.
56 R. Hissong et al.

Table 1 Selected small molecule compound databases. Table adapted from Blaženović et al.
(2018)
Free web
Additional compound ID-relevant access/free
Database Contents features download?
PubChem (Bolton All small Structure similarity search tool Y/Y
et al. 2008) molecules
ChemSpider All small Y/N
(Pence and molecules
Williams 2010)
ChEBI (Hastings Small Focus on compounds of biological Y/ Y
et al. 2016) molecules interest
KEGG (Kanehisa Metabolites Curated pathway maps Y/N
et al. 2006)
MetaCyc (Caspi Metabolites Curated pathway maps Y/
et al. 2008) noncommercial
HMDB (Wishart Human Physiological concentration data, Y/Y
et al. 2022) metabolites/ MS/MS spectra, text-mined literature
exposome context
Metlin (Guijas Metabolites Y (MS1 only)/
et al. 2018) N
RefMet (Fahy and Metabolites Name conversion tool Y/Y
Subramaniam
2020)
ChEMBL (Davies Bioactive Y/Y
et al. 2015) drug-like
molecules
DrugBank Known drugs Y/
(Wishart et al. noncommercial
2006)

Compound Databases Once degeneracy removal has been performed, and prefer-
ably after a molecular formula has been assigned, a feature’s m/z or neutral mass can
be searched against a compound database. Although m/z alone is by no means
sufficient to identify an unknown metabolite, databases help provide a list of
candidate compounds that can be evaluated using other means. Small molecule
compound databases vary in their scope and application; major examples are listed
in Table 1.
Databases range from those which focus on known endogenous metabolites to
those which attempt to cover all known and plausible small molecule chemical
structures. It is desirable to begin by querying the most specific applicable database
first. Small molecule databases focused on pharmaceutical compounds, such as
DrugBank (Wishart et al. 2018) and ChEMBL (Papadatos and Overington 2014;
Mendez et al. 2019), are of particular use for putative annotation of drugs and their
metabolites in pharmacometabolomics data. Likewise, using organism-specific
databases or constraining taxonomy to the sample type being analyzed will generate
Compound Identification Strategies in Mass Spectrometry-Based. . . 57

a narrower list of candidate compounds than an open-ended search against

PubChem. Annotation of unusual or novel compounds may require use of broader
databases.
Regardless of the database used, database hits against unknowns should not be
treated or reported as confident identifications, even if the hit list contains only a
single compound. Additional confirmation from other data sources is required, as
described in “Steps 4 and 5” in this chapter.

Retention Time Prediction As already noted, some compound databases contain

retention time information for specific compounds collected using defined chro-
matographic methods. However, in the frequent event that no experimentally deter-
mined RT information is available for these compounds, retention time prediction
strategies and software tools can help fill the gap (Bonini et al. 2020; Stanstrup et al.
2015). These approaches require training a computational model using predicted
chemical properties by inputting retention times and structures of known
compounds. Once the model is complete, retention times can be predicted for
candidate spectral matches and used to help confirm or refute compound
assignments made by MS/MS or other data. No retention time prediction software
yields precision comparable to experimental data; often, predicted retention times
deviate 10% or more from the experimental value. Nevertheless, in many cases, this
level of precision is enough to rule out a substantial portion of incorrect
identifications.

Collisional Cross-Section Prediction Analogous to RT prediction, collisional

cross-section values can be predicted based on compound structure and machine
learning (Plante et al. 2019) or quantum chemistry-based (Colby et al. 2019) models.
These values can then be compared against values experimentally determined for
unknown features, which are beginning to appear in major databases like HMDB
and Metlin. The desired result is either confirmation or refutation of the assigned
compound identity. Computational collisional cross-section prediction is in the early
phases of development but promises to impact compound identification as adoption
of ion mobility spectrometry increases.

5 Step 4: MS/MS Libraries and Compound Identification

Using Library Search

To move beyond candidate screening to true compound identification, the first-line

strategy is to search the MS and/or MS/MS spectrum of the unknown against a
spectral library. Mass spectral libraries contain experimentally collected or compu-
tationally predicted spectra of a database of small molecule compounds. For GC–
MS, the most extensive libraries consist of EI-MS spectra acquired at a standard
70 eV ionization energy. For LC-ESI, databases often consist of MS/MS spectra that
have been acquired for authentic standards on several different instrument types at a
range of collision energy values. Certain libraries may focus on a particular class of
58 R. Hissong et al.

Table 2 Widely used spectral libraries. Table adapted from Blaženović et al. (2018)
Free web
Additional compound access/free
Library Data type Type ID-relevant features download?
NIST EI-MS, Experimental Highly curated, includes N/N
(Stein CID-MS/ search software, available from
1999) MS multiple vendors
Wiley EI-MS, Experimental Largest collection of EI-MS N/N
(Solutions CID-MS/ data, available from multiple
2022) MS vendors
METLIN CID-MS/ Experimental Developed for Q-ToF N/N
(Guijas MS instruments, licensed annually
et al. 2018)
MoNA EI, Experimental, Community database, Y/Y
(Fiehn MS/MS, user- automated curation
2016b) MSn contributed
MassBank Metabolites Experimental, Community database Y/Y
(Horai et al. user-
2010) contributed
mzCloud MS/MS, Experimental Most complete MSn database Y/N
(LLC, H MSn
2022)
GNPS MS/MS Experimental, Integrated with molecular Y/Y
(Wang user- networking tools
et al. 2016) contributed
LipidBlast Bioactive Computational Fully computational Y/Y
(Kind et al. drug-like lipidomics database
2013) molecules

molecule, be acquired on a specific instrument, or contain MSn spectra (where

n > 2). Several major spectral libraries exist (Table 2). The National Institutes of
Standards and Testing (NIST) EI-MS and ESI-MS/MS databases contain carefully
curated, periodically updated libraries of spectra of authentic standards; this and
several other libraries are available from commercial distributors. Other libraries are
public; some of these are derived from “crowdsourced” compilations of spectra and
their assigned chemical structures, which typically receive community review to
assess quality. Still other spectral libraries are generated by computationally
predicting fragmentation of a list of chemical structures from a database, yielding
an “in silico” spectral library.
To perform a library search, the mass spectrum of an unknown feature is
computationally compared to entries in the spectral library. After an initial candidate
screening step, each potential spectral match is ranked by a scoring function that
generates a match score reflecting the similarity between the spectrum of
the unknown compound and the library spectrum. Score ranges differ depending
on the function that is used, but in general a higher score corresponds to a closer
spectral match. The chemical structure of the compound associated with the highest-
scoring hit is assigned as the most probable identification; additional lower-scoring
Compound Identification Strategies in Mass Spectrometry-Based. . . 59

hits may be retained for user review. Many different scoring functions exist; among
these, probability-based matching (PBM) was one of the first introduced scoring
functions for EI-MS spectra (Stauffer et al. 1985), and the classic “dot-product” and
“reverse dot-product” scoring algorithms have been widely used for MS/MS
searching (Stein and Scott 1994). Many other scoring functions have also been
evaluated, including a recent spectral entropy-based algorithm that demonstrated
superior performance to dot-product scores (Li et al. 2021). Additional constraints,
such as a narrow mass window for precursor ion match and a precursor ion isotope
pattern match, can serve to further shorten a list of candidates or even select a single
probable compound identity.
Many software tools exist to automate spectral library search. Most instrument
manufacturers incorporate library search tools into their data analysis software; some
also provide access to proprietary libraries searchable only using these tools. Among
cross-platform tools, NIST MS Search is one of the best known (Stein 1999); it
allows searching individual spectra against the NIST library and other user-loaded
libraries with visualization of spectral matches using head-to-tail or difference plots.
To enable much more rapid search of many spectra, a useful alternative is
MSPepSearch (Zhang et al. 2018). Originally designed for peptide spectra, it has
been adopted and extensively used for small molecule analysis; it uses a similar
scoring algorithm but generates output in tabular form. Progenesis QI is a commer-
cial data analysis tool that enables cross-vendor MS/MS search in addition to general
data analysis. MS-DIAL is a free, open-source alternative that integrates feature
finding, alignment, and spectral search in a unified workflow (Tsugawa et al. 2015).
Other widely used untargeted metabolomics data analysis tools, including XCMS
(Smith et al. 2006) and MZmine (Pluskal et al. 2010), also have some MS/MS search
capabilities, though they are implemented in a less visual manner than in MS-DIAL.
SIRIUS (Böcker et al. 2009; Dührkop et al. 2019) and GNPS (Wang et al. 2016),
described later in this chapter, also use spectral search heavily in their workflow but
due to other features are classified separately from typical library search tools.
It is important to note that while spectral searching provides a fast and often
accurate means of assigning chemical structures to features in metabolomics data, it
does not provide an objective means to assess the probability that these
identifications are correct. Manual review of spectra, including matches beyond
the top hit, can sometimes help clarify ambiguous assignments and determine
appropriate score thresholds for identification. Well-defined strategies for assessing
and reporting identification confidence are important to data analysis and are
described in “Step 5” of our workflow.

Spectral Similarity Searching Searching an unknown feature against a library for

a precise match is not likely to produce informative results when the compound in
question is not in the library. As an alternative, it is possible to search for spectra that
are not a direct match to the unknown but share some of its features. This approach is
termed similarity searching and can be performed using several software tools. One
of the most robust is “hybrid search,” implemented in NIST MS Search and
MSPepSearch (Cooper et al. 2019). This search method allows both direct peak
60 R. Hissong et al.

matching (as in identity searching) as well as matching of masses shifted by a neutral

loss within a single spectrum. This accounts for cases in which the m/z of some
fragments of an unknown are identical to those of a library compound, while others
are shifted by a structural difference typically confined to a single region of the
molecule. A second similarity search strategy is implemented in the software tool
DeepMASS (Tiwary et al. 2019), which uses a machine deep-learning approach to
perform structural similarity scoring. All similarity searches aim to generate matches
that are likely to be structurally related to the unknown compound. The resulting
matches can be used for compound class assignment as well as to help elucidate the
structure of the unknown, with the aid of other techniques described below.

In Silico Libraries and In Silico Fragmentation Software To generate more

extensive spectral libraries than is possible using authentic standards, several
methods have been used to generate in silico spectra libraries. These predict how
compounds will fragment, using techniques ranging from quantum-chemistry-based
calculations (Wang et al. 2020) to rule-based methods (Tsugawa et al. 2016), and
generate a theoretical fragmentation spectrum for each compound. Each strategy has
advantages and limitations; these have been described and reviewed in detail
elsewhere (Borges et al. 2021). In silico libraries are most widely used in lipidomics.
The LipidBlast MS/MS spectral library enables identification of thousands of lipid
species, and stands out in its accuracy and widespread use, because MS/MS frag-
mentation patterns are typically reproducible for all lipids within a lipid class (Kind
et al. 2013). In silico spectral libraries of other small molecule classes are also
available, but due to limited fragmentation, the high diversity of chemical structures,
and the difficulty of predicting relative abundance of fragment ions, most non-lipid
in silico spectral databases are of limited utility for automated spectral search.
An alternative strategy is to use the technique of in silico fragmentation to help
interpret spectra of unknown compounds. Many software packages exist to help
perform this task; among the most prominent are MetFrag (Ruttkies et al. 2016)
and CSI:FingerID (Dührkop et al. 2015), which is now implemented in the SIRIUS
data analysis package (Dührkop et al. 2019). To interpret an unknown spectrum, a
user inputs information into the software regarding the unknown compound, includ-
ing the observed precursor ion, adduct type or molecular formula if known, the
obtained MS/MS spectrum of the unknown, and in some cases, additional data such
as the isotope distribution of the precursor ion and the taxonomy of the organism
from which the sample that generated the spectrum was obtained. The software then
selects candidate compounds from a large compound database (PubChem or similar)
that matches the precursor ion mass and any other metadata and then predicts
fragments that would be formed from these precursors using rule-based, machine-
learning, fragmentation tree generation, or other strategies. The predicted fragments
are matched against the unknown spectrum, and the data are used to predict the
structure, or at least key structural elements, of the unknown. In silico fragmentation
and structure prediction does not always yield a definite compound identification for
an unknown but nevertheless is one of the simplest strategies to help predict structure
based on spectral data.
Compound Identification Strategies in Mass Spectrometry-Based. . . 61

Spectral Networking Analysis Another technique that has gained prominence as a

means of interpreting MS/MS spectra in untargeted metabolomics data is spectral
networking analysis. The most widely used software tool in this area is the Global
Natural Products Database (Wang et al. 2016). The MS/MS spectra of identified and
unknown features in a dataset are searched against each other, in addition to against
library spectra. The most similar spectra are grouped, and clustering methods are
used to generate spectral networks that can be examined visually. Since the most
strongly associated spectra are grouped together, neighboring features on the net-
work may represent similar molecules with small structural differences. By calculat-
ing the precise mass shift difference between precursor ions, a molecular formula
“difference” can often be assigned, which can help highlight functional groups that
differentiate the compounds. If one of the two features is identified with high
confidence, it is sometimes possible to predict the structure of the unknown feature
from these data. Even when this is not possible, the networking analysis can reveal
useful information about the structure of the unknowns.

6 Step 5: Assess and Report Identification Confidence

As evidenced by the range of techniques described in this chapter, compound

identification in metabolomics ranges from routine to extremely challenging. Like-
wise, the evidence supporting an identification can range from unequivocal to
uncertain. In the interest of scientific transparency, it is important that researchers
report not only the assigned identity of a compound but also the method and data
used to make identifications and an estimate of their confidence that the identifica-
tion is correct at any level of information (structure, formula, compound class). As
illustrated in Fig. 2, currently recommended methods for reporting compound
identifications involve semi-subjective classification by the analyst, while methods
under development have potential to improve accuracy, speed, and transparency of
the compound identification and reporting process.

Reporting Identification Rigor Using Identification Levels The importance of

data-reporting standards has long been clear to the metabolomics community. In
2007 an international consortium termed the Metabolomics Standards Initiative
(MSI) published a set of recommended minimum data-reporting standards, which
included four “identification levels.” (Sumner et al. 2007) Level 1 represents the
highest degree of rigor for compound identification, in which compound identity is
established at the chemical structure level by matching at least two orthogonal forms
of experimental data to reference data collected in the researcher’s own laboratory
using an authentic standard. These could include, for instance, accurate mass and
retention time, or retention time and MS/MS spectrum. Level 2 is a putative
annotation at the structure level based on data collected outside the researcher’s
laboratory, such as an MS/MS spectral match with a library, or a literature-based
retention time or retention index. Level 3 signifies a putative compound class
assignment based on spectral similarity searching and/or physiochemical property
62 R. Hissong et al.

Typical current workflow Improved reporng strategy

Level 1:
IDENTIFIED Time (min)
Time (min)
Spectral
Level 2: Search Decoy
PUTATIVE ID m/z Score(s) Library search for
m/z
Feature data Feature data FDR
esmaon
Level 3: ID with
Manual COMPOUND RT
Predicon confidence
Review CLASS ID esmate
m/z Score
Time (min)

Level 4: E. coli
Spectral
UNKNOWN In silico ? ? ?
Search
Fragmentaon Biological

Dri me
Score(s)
Score Probability
Assessment
RT
Collisional Cross Secon Score

Fig. 2 Current typical current metabolite identification and reporting strategy compared to a
hypothetical multi-input method for improving identification accuracy and assessing and reporting
identification confidence

assessment. Level 4 represents unknown compounds. Further refinements to com-

pound identification levels have been proposed and adopted by various groups or for
particular applications (Schymanski et al. 2014). When applied, the MSI compound
identification levels have served the community by allowing scientists to communi-
cate about identifications using consistent terminology. One challenge is that MSI
levels (or their equivalent) have not been universally adopted in the literature; many
publications report metabolite identifications with minimal information about how
they were made. Another key limitation is that acceptance or rejection of an
identification at any identification level is, in large part, at the discretion of the
analyst. For instance, no minimum spectral library search score is given for accep-
tance of a level 2 identification, nor is it easily possible to establish one since an
appropriate score threshold may vary from one study or spectral library to the next.
Manual review of compound identifications can help confirm correct and refute
incorrect identifications, but this approach is too time consuming for experiments
with tens of thousands of compounds detected.

False Discovery Estimation Using Decoy Metabolite Libraries An objective and

automatable approach to assess and report compound identification confidence in the
form of an estimated false discovery rate (FDR) would be useful to improve
consistency and inter-lab comparability of metabolomics data (Scheubert et al.
2017). In proteomics, FDR estimation is performed by searching peptide MS/MS
spectra against both normal and “decoy” in silico spectral libraries, the latter of
which is generated by scrambling the amino acid sequence of all proteins in the
library (Elias and Gygi 2010). By comparing the number of hits to the decoy library
relative to the true library, an FDR can be estimated and reported with the data.
Although no direct equivalent to amino acid sequence scrambling exists in
metabolomics, several approaches for generation of decoy libraries have been
Compound Identification Strategies in Mass Spectrometry-Based. . . 63

proposed and tested for small molecules, ranging from randomizing all peaks found
in a typical library to approaches based on rearrangement of fragmentation trees
(Scheubert et al. 2017; Wang et al. 2018; Alka et al. 2022). Decoy libraries have
been used to help select spectral search score thresholds and other criteria to achieve
a desired FDR. However, these attempts have demonstrated that decoy libraries are
not very helpful to assess FDR for some classes of metabolites, particularly for
compounds that produce very few fragment ions and those that have multiple
structural isomers that produce similar fragmentation spectra. Thus, FDR estimation
strategies for metabolomics need further refinement and will certainly require sup-
plementation with strategies beyond decoy library searching before compound
identification in metabolomics can achieve the level of automation and accuracy
enjoyed by other omics sciences.

Integrating Multiple Strategies for Confident Compound Identification In this

chapter, we have described a variety of experimental and computational strategies to
help annotate and identify features in metabolomics data. Using present
technologies, it is sometimes possible to arrive at a single correct structure-level
identification for features of interest in the data, while in other cases, more limited
annotation is all that is possible. Currently, integrating available information from
the methods and tools we have described is the task of the analyst, who uses their
judgment to set thresholds, finalize identifications and report methods used. Moving
forward, a priority for the metabolomics community is to develop a strategy to
integrate all available information in a consistent and automated manner to make
compound identification assignments, with support from objective data that estimate
identification confidence. A compound identification “meta-analysis” approach may
prove useful (Fig. 2), but no comprehensive strategy has yet been devised. For the
time being, compound identification in metabolomics is an exercise in defining aims
appropriately, selecting methods and collecting sufficient relevant data, and applying
available tools to help with interpretation. As the metabolomics research community
repeats and refines this process, it continues to work toward more unifying methods
for compound identification.

7 Conclusion: Pharmacology-Focused
Compound Identification

As described throughout the chapter, considerations relevant to identification of

small molecules in metabolomics data are, in general, fully applicable to
pharmacology-focused metabolomics studies. A key reminder is to tailor strategies
to study design. When only specific drug-derived or endogenous metabolites are of
interest, a suitable targeted metabolomics workflow reduces the burden of assigning
compound identities to the large number of features that would be detected in an
untargeted study design. When untargeted analysis is desired, selection of a com-
pound database or library to focus on the organism being studied or a drug class of
interest can yield more meaningful results than when a broad, nonspecific database is
64 R. Hissong et al.

used. Researchers should also consider biological factors that may affect both
compound identification and quantitation, such as sex-specific responses to drug
metabolism (Soldin et al. 2011; Chary et al. 2022). Finally, the application of
metabolomics to pharmacological studies is still a developing area of study. While
at present most of the data relevant to compound identification in metabolomics is
found in the analytical chemistry and bioinformatics literature, resources specific to
the challenges of pharmacology can be expected to develop and strengthen over
future years.

References
Abdelhameed A et al (2014) Pseudo-MS3 approach using electrospray mass spectrometry
(ESI-MS/MS) to characterize certain (2E)-2-[3-(1H-Imidazol-1-yl)-1-phenylpropylidene]
hydrazinecarboxamide derivatives. J Chem 10 p. https://doi.org/10.1155/2014/386301
Alka O et al (2022) DIAMetAlyzer allows automated false-discovery rate-controlled analysis for
data-independent acquisition in metabolomics. Nat Commun 13:1347. https://doi.org/10.1038/
s41467-022-29006-z
Alley WR, Mechref Y, Novotny MV (2009) Characterization of glycopeptides by combining
collision-induced dissociation and electron-transfer dissociation mass spectrometry data.
Rapid Commun Mass Spectrom 23:161–170. https://doi.org/10.1002/rcm.3850
Anderson BG, Raskind A, Habra H, Kennedy RT, Evans CR (2021) Modifying chromatography
conditions for improved unknown feature identification in untargeted metabolomics. Anal
Chem 93:15840–15849. https://doi.org/10.1021/acs.analchem.1c02149
Baba T, Campbell J, Le Blanc Y, Baker P, Ikeda K (2018) Quantitative structural multi-class
lipidomics using differential mobility-electron impact excitation of ions from organics (EIEIO)
mass spectrometry. J Lipid Res 59:jlr.D083261. https://doi.org/10.1194/jlr.D083261
Balogh MP (2004) Debating resolution and mass accuracy in mass spectrometry. Spectroscopy
(Santa Monica) 19:34–39
Beyoğlu D, Zhou Y, Chen C, Idle JR (2018) Mass isotopomer-guided decluttering of metabolomic
data to visualize endogenous biomarkers of drug toxicity. Biochem Pharmacol 156:491–500.
https://doi.org/10.1016/j.bcp.2018.09.022
Blaženović I, Kind T, Ji J, Fiehn O (2018) Software tools and approaches for compound identifica-
tion of LC-MS/MS data in metabolomics. Metabolites 8. https://doi.org/10.3390/
metabo8020031
Böcker S, Letzel MC, Lipták Z, Pervukhin A (2009) SIRIUS: decomposing isotope patterns for
metabolite identification. Bioinformatics 25:218–224
Bolton EE, Wang Y, Thiessen PA, Bryant SH (2008) Annual reports in computational chemistry,
vol 4. Elsevier, pp 217–241
Bonini P, Kind T, Tsugawa H, Barupal DK, Fiehn O (2020) Retip: retention time prediction for
compound annotation in untargeted metabolomics. Anal Chem 92:7515–7522. https://doi.org/
10.1021/acs.analchem.9b05765
Borges RM et al (2021) Quantum chemistry calculations for metabolomics. Chem Rev 121:5633–
5670. https://doi.org/10.1021/acs.chemrev.0c00901
Broeckling CD, Afsar FA, Neumann S, Ben-Hur A, Prenni JE (2014) RAMClust: a novel feature
clustering method enables spectral-matching-based annotation for metabolomics data. Anal
Chem 86:6812–6817. https://doi.org/10.1021/ac501530d
Bruce SJ et al (2009) Investigation of human blood plasma sample preparation for performing
metabolomics using ultrahigh performance liquid chromatography/mass spectrometry. Anal
Chem 81:3285–3296. https://doi.org/10.1021/ac8024569
Compound Identification Strategies in Mass Spectrometry-Based. . . 65

Cajka T, Fiehn O (2016) Toward merging untargeted and targeted methods in mass spectrometry-
based metabolomics and Lipidomics. Anal Chem 88:524–545. https://doi.org/10.1021/acs.
analchem.5b04491
Caspi R et al (2008) The MetaCyc database of metabolic pathways and enzymes and the BioCyc
collection of pathway/genome databases. Nucleic Acids Res 36:D623–D631. https://doi.org/10.
1093/nar/gkm900
Chaleckis R, Meister I, Zhang P, Wheelock CE (2019) Challenges, progress and promises of
metabolite annotation for LC-MS-based metabolomics. Curr Opin Biotechnol 55:44–50.
https://doi.org/10.1016/j.copbio.2018.07.010
Chary S, Amrein K, Mahmoud SH, Lasky-Su JA, Christopher KB (2022) Sex-specific catabolic
metabolism alterations in the critically ill following high dose vitamin D. Metabolites 12. https://
doi.org/10.3390/metabo12030207
Clendinen CS et al (2015) An overview of methods using 13C for improved compound identifica-
tion in metabolomics and natural products. Front Plant Sci 6. https://doi.org/10.3389/fpls.2015.
00611
Colby SM et al (2019) ISiCLE: a quantum chemistry pipeline for establishing in silico collision
cross section libraries. Anal Chem 91:4346–4356. https://doi.org/10.1021/acs.analchem.
8b04567
Contrepois K et al (2018) Cross-platform comparison of untargeted and targeted Lipidomics
approaches on aging mouse plasma. Sci Rep 8:17747. https://doi.org/10.1038/s41598-018-
35807-4
Cooper BT et al (2019) Hybrid search: a method for identifying metabolites absent from tandem
mass spectrometry libraries. Anal Chem 91:13924–13932. https://doi.org/10.1021/acs.
analchem.9b03415
Davies AN (1998) The new Automated Mass Spectrometry Deconvolution and Identification
System (AMDIS). Spectrosc Eur 10(3):22–26
Davies M et al (2015) ChEMBL web services: streamlining access to drug discovery data and
utilities. Nucleic Acids Res 43:W612–W620. https://doi.org/10.1093/nar/gkv352
Dona AC et al (2016) A guide to the identification of metabolites in NMR-based metabonomics/
metabolomics experiments. Comput Struct Biotechnol J 14:135–153. https://doi.org/10.1016/j.
csbj.2016.02.005
Ducati AO et al (2021) Improved metabolite characterization by liquid chromatography – Tandem
mass spectrometry through electron impact type fragments from adduct ions. Anal Chim Acta
1150:338207. https://doi.org/10.1016/j.aca.2021.338207
Dührkop K, Shen H, Meusel M, Rousu J, Böcker S (2015) Searching molecular structure databases
with tandem mass spectra using CSI:FingerID. Proc Natl Acad Sci 112:12580–12585. https://
doi.org/10.1073/pnas.1509788112
Dührkop K et al (2019) SIRIUS 4: a rapid tool for turning tandem mass spectra into metabolite
structure information. Nat Methods 16:299–302. https://doi.org/10.1038/s41592-019-0344-8
Elias JE, Gygi SP (2010) Target-decoy search strategy for mass spectrometry-based proteomics.
Methods Mol Biol 604:55–71. https://doi.org/10.1007/978-1-60761-444-9_5
Emwas A-H et al (2019) NMR spectroscopy for metabolomics research. Metabolites 9. https://doi.
org/10.3390/metabo9070123
Fahy E, Subramaniam S (2020) RefMet: a reference nomenclature for metabolomics. Nat Methods
17:1173–1174. https://doi.org/10.1038/s41592-020-01009-y
Fiehn O (2016a) Metabolomics by gas chromatography-mass spectrometry: combined targeted and
untargeted profiling. Curr Protoc Mol Biol 114:30.34.31–30.34.32. https://doi.org/10.1002/
0471142727.mb3004s114
Fiehn O (2016b) Mass resolution and resolving power. https://fiehnlab.ucdavis.edu/projects/seven-
golden-rules/mass-resolution
Gowda GAN, Djukovic D (2014) Overview of mass spectrometry-based metabolomics:
opportunities and challenges. Methods Mol Biol 1198:3–12. https://doi.org/10.1007/978-1-
4939-1258-2_1
66 R. Hissong et al.

Griffiths WJ et al (2010) Targeted metabolomics for biomarker discovery. Angew Chem Int Ed 49:
5426–5445. https://doi.org/10.1002/anie.200905579
Guijas C et al (2018) METLIN: a technology platform for identifying knowns and unknowns. Anal
Chem 90:3156–3164. https://doi.org/10.1021/acs.analchem.7b04424
Guntner AS, Thalhamer B, Klampfl C, Buchberger W (2019) Collision cross sections obtained with
ion mobility mass spectrometry as new descriptor to predict blood-brain barrier permeation by
drugs. Sci Rep 9:19182. https://doi.org/10.1038/s41598-019-55856-7
Guo J, Huan T (2020a) Comparison of full-scan, data-dependent, and data-independent acquisition
modes in liquid chromatography–mass spectrometry based untargeted metabolomics. Anal
Chem 92:8072–8080. https://doi.org/10.1021/acs.analchem.9b05135
Guo J, Huan T (2020b) Evaluation of significant features discovered from different data acquisition
modes in mass spectrometry-based untargeted metabolomics. Anal Chim Acta 1137:37–46.
https://doi.org/10.1016/j.aca.2020.08.065
Guo K, Li L (2009) Differential (12)C/(13)C-isotope dansylation labeling and fast liquid chroma-
tography/mass spectrometry for absolute and relative quantification of the metabolome. Anal
Chem 81:3919–3932. https://doi.org/10.1021/ac900166a
Habra H et al (2021) metabCombiner: paired untargeted LC-HRMS metabolomics feature matching
and concatenation of disparately acquired data sets. Anal Chem 93:5028–5036. https://doi.org/
10.1021/acs.analchem.0c03693
Halket JM et al (2005) Chemical derivatization and mass spectral libraries in metabolic profiling by
GC/MS and LC/MS/MS. J Exp Bot 56:219–243. https://doi.org/10.1093/jxb/eri069
Han J, Lin K, Sequeira C, Borchers CH (2015) An isotope-labeled chemical derivatization method
for the quantitation of short-chain fatty acids in human feces by liquid chromatography-tandem
mass spectrometry. Anal Chim Acta 854:86–94. https://doi.org/10.1016/j.aca.2014.11.015
Hastings J et al (2016) ChEBI in 2016: improved services and an expanding collection of
metabolites. Nucleic Acids Res 44:D1214–D1219. https://doi.org/10.1093/nar/gkv1031
Health in Northern Ireland (2022) National Center for Advancing Translational Sciences Com-
pound Management. https://ncats.nih.gov/preclinical/core/compound.
Hines KM, Ross DH, Davidson KL, Bush MF, Xu L (2017) Large-scale structural characterization
of drug and drug-like compounds by high-throughput ion mobility-mass spectrometry. Anal
Chem 89:9023–9030. https://doi.org/10.1021/acs.analchem.7b01709
Horai H et al (2010) MassBank: a public repository for sharing mass spectral data for life sciences. J
Mass Spectrom 45:703–714. https://doi.org/10.1002/jms.1777
Ichou F et al (2014) Comparison of the activation time effects and the internal energy distributions
for the CID, PQD and HCD excitation modes. J Mass Spectrom 49:498–508. https://doi.org/10.
1002/jms.3365
Kachman M et al (2019) Deep annotation of untargeted LC-MS metabolomics data with Binner.
Bioinformatics 36:1801–1806. https://doi.org/10.1093/bioinformatics/btz798
Kanehisa M et al (2006) From genomics to chemical genomics: new developments in KEGG.
Nucleic Acids Res 34:D354–D357
Kantae V et al (2017) Integration of pharmacometabolomics with pharmacokinetics and pharma-
codynamics: towards personalized drug therapy. Metabolomics 13. https://doi.org/10.1007/
s11306-016-1143-1
Kanu AB, Dwivedi P, Tam M, Matz L, Hill HH (2008) Ion mobility-mass spectrometry. J Mass
Spectrom 43:1–22. https://doi.org/10.1002/jms.1383
Kind T, Fiehn O (2007) Seven golden rules for heuristic filtering of molecular formulas obtained by
accurate mass spectrometry. BMC Bioinformatics 8:105. https://doi.org/10.1186/1471-2105-
8-105
Kind T, Fiehn O (2010) Advances in structure elucidation of small molecules using mass spec-
trometry. Bioanal Rev 2:23–60. https://doi.org/10.1007/s12566-010-0015-9
Kind T et al (2009) FiehnLib: mass spectral and retention index libraries for metabolomics based on
quadrupole and time-of-flight gas chromatography/mass spectrometry. Anal Chem 81:10038–
10048. https://doi.org/10.1021/ac9019522
Compound Identification Strategies in Mass Spectrometry-Based. . . 67

Kind T et al (2013) LipidBlast in silico tandem mass spectrometry database for lipid identification.
Nat Methods 10:755–758. https://doi.org/10.1038/nmeth.2551
Koek MM, Muilwijk B, van Stee LL, Hankemeier T (2008) Higher mass loadability in comprehen-
sive two-dimensional gas chromatography-mass spectrometry for improved analytical perfor-
mance in metabolomics analysis. J Chromatogr A 1186:420–429. https://doi.org/10.1016/j.
chroma.2007.11.107
Koelmel JP et al (2017) Expanding lipidome coverage using LC-MS/MS data-dependent acquisi-
tion with automated exclusion list generation. J Am Soc Mass Spectrom 28:908–917
Kuhl C, Tautenhahn R, Böttcher C, Larson TR, Neumann S (2012) CAMERA: an integrated
strategy for compound spectra extraction and annotation of LC/MS data sets. Anal Chem 84:
283–289. https://doi.org/10.1021/ac202450g
Kvitvang HFN, Andreassen T, Adam T, Villas-Bôas SG, Bruheim P (2011) Highly sensitive
GC/MS/MS method for quantitation of amino and nonamino organic acids. Anal Chem 83:
2705–2711. https://doi.org/10.1021/ac103245b
Lai Z, Fiehn O (2018) Mass spectral fragmentation of trimethylsilylated small molecules. Mass
Spectrom Rev 37:245–257. https://doi.org/10.1002/mas.21518
Lee HB, Peart TE, Svoboda ML (2005) Determination of endocrine-disrupting phenols, acidic
pharmaceuticals, and personal-care products in sewage by solid-phase extraction and gas
chromatography-mass spectrometry. J Chromatogr A 1094:122–129. https://doi.org/10.1016/j.
chroma.2005.07.070
Lenz EM et al (2007) HPLC-NMR with severe column overloading: fast-track metabolite identifi-
cation in urine and bile samples from rat and dog treated with [14C]-ZD6126. J Pharm Biomed
Anal 43:1065–1077. https://doi.org/10.1016/j.jpba.2006.09.010
Li Y et al (2021) Spectral entropy outperforms MS/MS dot product similarity for small-molecule
compound identification. Nat Methods 18:1524–1531. https://doi.org/10.1038/s41592-021-
01331-z
Lisec J, Schauer N, Kopka J, Willmitzer L, Fernie AR (2006) Gas chromatography mass
spectrometry-based metabolite profiling in plants. Nat Protoc 1:387–396. https://doi.org/10.
1038/nprot.2006.59
LLC, H (2022) mzCloud advanced mass spectral database. https://www.mzcloud.org/
Lorenz MA, Burant CF, Kennedy RT (2011) Reducing time and increasing sensitivity in sample
preparation for adherent mammalian cell metabolomics. Anal Chem 83:3406–3414. https://doi.
org/10.1021/ac103313x
Lu W, Bennett BD, Rabinowitz JD (2008) Analytical strategies for LC-MS-based targeted
metabolomics. J Chromatogr B Analyt Technol Biomed Life Sci 871:236–242. https://doi.
org/10.1016/j.jchromb.2008.04.031
Ludwig M et al (2019) ZODIAC: database-independent molecular formula annotation using Gibbs
sampling reveals unknown small molecules
Mahieu NG, Patti GJ (2017) Systems-level annotation of a metabolomics data set reduces 25,000
features to fewer than 1000 unique metabolites. Anal Chem 89:10397–10406. https://doi.org/
10.1021/acs.analchem.7b02380
Mahieu NG, Spalding JL, Gelman SJ, Patti GJ (2016) Defining and detecting complex peak
relationships in mass spectral data: the Mz.Unity algorithm. Anal Chem 88:9037–9046.
https://doi.org/10.1021/acs.analchem.6b01702
Majuta SN et al (2019) Rapid solution-phase hydrogen/deuterium exchange for metabolite com-
pound identification. J Am Soc Mass Spectrom 30:1102–1114. https://doi.org/10.1007/s13361-
019-02163-0
Matyash V, Liebisch G, Kurzchalia TV, Shevchenko A, Schwudke D (2008) Lipid extraction by
methyl-tert-butyl ether for high-throughput lipidomics. J Lipid Res 49:1137–1146. https://doi.
org/10.1194/jlr.D700041-JLR200
Maurer HH (1998) Liquid chromatography mass spectrometry in forensic and clinical toxicology. J
Chromatogr B 713:3–25. https://doi.org/10.1016/s0378-4347(97)00514-8
68 R. Hissong et al.

McCann MR, George De la Rosa MV, Rosania GR, Stringer KA (2021) L-carnitine and
acylcarnitines: mitochondrial biomarkers for precision medicine. Metabolites 11. https://doi.
org/10.3390/metabo11010051
Mendez D et al (2019) ChEMBL: towards direct deposition of bioassay data. Nucleic Acids Res 47:
D930–d940. https://doi.org/10.1093/nar/gky1075
Meng X et al (2021) Simultaneous 3-nitrophenylhydrazine derivatization strategy of carbonyl,
carboxyl and phosphoryl submetabolome for LC-MS/MS-based targeted metabolomics with
improved sensitivity and coverage. Anal Chem 93:10075–10083. https://doi.org/10.1021/acs.
analchem.1c00767
Misra BB, Olivier M (2020) High resolution GC-Orbitrap-MS metabolomics using both electron
ionization and chemical ionization for analysis of human plasma. J Proteome Res 19:2717–
2731. https://doi.org/10.1021/acs.jproteome.9b00774
Nelson AB, Chow LS, Hughey CC, Crawford PA, Puchalska P (2022) Artifactual fatty acid dimers
mimic FAHFA signals in untargeted metabolomics pipelines. J Lipid Res 100201. https://doi.
org/10.1016/j.jlr.2022.100201
Papadatos G, Overington JP (2014) The ChEMBL database: a taster for medicinal chemists. Future
Med Chem 6:361–364. https://doi.org/10.4155/fmc.14.8
Parker CE et al (2014) In: Tang NLS, Poon T (eds) Chemical diagnostics: from bench to bedside
vol. 336 topics in current chemistry-series. pp 117–137
Patti GJ, Yanes O, Siuzdak G (2012) Metabolomics: the apogee of the omics trilogy. Nat Rev Mol
Cell Biol 13:263–269. https://doi.org/10.1038/nrm3314
Pence H, Williams A (2010) ChemSpider: an online chemical information resource. J Chem Educ
87:1123–1124. https://doi.org/10.1021/ED100697W
Plante P-L et al (2019) Predicting ion mobility collision cross-sections using a deep neural network:
DeepCCS. Anal Chem 91:5191–5199. https://doi.org/10.1021/acs.analchem.8b05821
Pluskal T, Castillo S, Villar-Briones A, Oresic M (2010) MZmine 2: modular framework for
processing, visualizing, and analyzing mass spectrometry-based molecular profile data. BMC
Bioinformatics 11:395. https://doi.org/10.1186/1471-2105-11-395
Prakash C, Shaffer CL, Nedderman A (2007) Analytical strategies for identifying drug metabolites.
Mass Spectrom Rev 26:340–369. https://doi.org/10.1002/mas.20128
Rainville PD et al (2017) Ion mobility spectrometry combined with ultra performance liquid
chromatography/mass spectrometry for metabolic phenotyping of urine: effects of column
length, gradient duration and ion mobility spectrometry on metabolite detection. Anal Chim
Acta 982:1–8. https://doi.org/10.1016/j.aca.2017.06.020
Ridder L et al (2012) Substructure-based annotation of high-resolution multistage MSn spectral
trees. Rapid Commun Mass Spectrom 26:2461–2471. https://doi.org/10.1002/rcm.6364
Roberts LD, Souza AL, Gerszten RE, Clish CB (2012) Targeted metabolomics. Curr Protoc Mol
Biol. Chapter 30, Unit 30.32.31-24. https://doi.org/10.1002/0471142727.mb3002s98
Ruttkies C, Schymanski EL, Wolf S, Hollender J, Neumann S (2016) MetFrag relaunched:
incorporating strategies beyond in silico fragmentation. J Chem 8:3. https://doi.org/10.1186/
s13321-016-0115-9
Saigusa D, Matsukawa N, Hishinuma E, Koshiba S (2021) Identification of biomarkers to diagnose
diseases and find adverse drug reactions by metabolomics. Drug Metab Pharmacokinet 37:
100373. https://doi.org/10.1016/j.dmpk.2020.11.008
Scheubert K et al (2017) Significance estimation for large scale metabolomics annotations by
spectral matching. Nat Commun 8:1494. https://doi.org/10.1038/s41467-017-01318-5
Schrimpe-Rutledge AC, Codreanu SG, Sherrod SD, McLean JA (2016) Untargeted metabolomics
strategies – challenges and emerging directions. J Am Soc Mass Spectrom 27:1897–1905.
https://doi.org/10.1007/s13361-016-1469-y
Schymanski EL et al (2014) Identifying small molecules via high resolution mass spectrometry:
communicating confidence. Environ Sci Technol 48:2097–2098. https://doi.org/10.1021/
es5002105
Compound Identification Strategies in Mass Spectrometry-Based. . . 69

Scott DR (1992) Rapid and accurate method for estimating molecular weights of organic
compounds from low resolution mass spectra. Chemom Intel Lab Syst 16:193–202. https://
doi.org/10.1016/0169-7439(92)80037-5
Shackleton C, Pozo OJ, Marcos J (2018) GC/MS in recent years has defined the Normal and
clinically disordered Steroidome: will it soon be surpassed by LC/tandem MS in this role? J
Endocr Soc 2:974–996. https://doi.org/10.1210/js.2018-00135
Smith CA, Want EJ, O'Maille G, Abagyan R, Siuzdak G (2006) XCMS: processing mass spec-
trometry data for metabolite profiling using nonlinear peak alignment, matching, and identifica-
tion. Anal Chem 78:779–787. https://doi.org/10.1021/ac051437y
Soldin OP, Chung SH, Mattison DR (2011) Sex differences in drug disposition. J Biomed
Biotechnol 2011:187103. https://doi.org/10.1155/2011/187103
Solutions WS (2022) Wiley registry 12th edition/NIST 2020 mass spectral library. https://
sciencesolutions.wiley.com/solutions/technique/gc-ms/wiley-registry-12th-edition-nist-2020/
Stanstrup J, Neumann S, Vrhovšek U (2015) PredRet: prediction of retention time by direct
mapping between multiple chromatographic systems. Anal Chem 87:9421–9428. https://doi.
org/10.1021/acs.analchem.5b02287
Stauffer DB, McLafferty FW, Ellis RD, Peterson DW (1985) Probability-based-matching algorithm
with forward searching capabilities for matching unknown mass spectra of mixtures. Anal Chem
57:1056–1060. https://doi.org/10.1021/ac00283a021
Stein SE (1999) An integrated method for spectrum extraction and compound identification from
gas chromatography/mass spectrometry data. J Am Soc Mass Spectrom 10:770–781
Stein SE, Scott DR (1994) Optimization and testing of mass spectral library search algorithms for
compound identification. J Am Soc Mass Spectrom 5:859–866. https://doi.org/10.1016/1044-
0305(94)87009-8
Steuer AE, Brockbals L, Kraemer T (2019) Metabolomic strategies in biomarker research-new
approach for indirect identification of drug consumption and sample manipulation in clinical
and forensic toxicology? Front Chem 7:319. https://doi.org/10.3389/fchem.2019.00319
Strehmel N, Hummel J, Erban A, Strassburg K, Kopka J (2008) Retention index thresholds for
compound matching in GC-MS metabolite profiling. J Chromatogr B Analyt Technol Biomed
Life Sci 871:182–190. https://doi.org/10.1016/j.jchromb.2008.04.042
Sumner LW et al (2007) Proposed minimum reporting standards for chemical analysis chemical
analysis working group (CAWG) metabolomics standards initiative (MSI). Metabolomics 3:
211–221. https://doi.org/10.1007/s11306-007-0082-2
Taylor T (2015) Understanding electron ionization processes for GC–MS. LCGC N Am 33:290
Theodoridis GA, Gika HG, Want EJ, Wilson ID (2012) Liquid chromatography-mass spectrometry
based global metabolite profiling: a review. Anal Chim Acta 711:7–16. https://doi.org/10.1016/
j.aca.2011.09.042
Tiwary S et al (2019) High-quality MS/MS spectrum prediction for data-dependent and data-
independent acquisition data analysis. Nat Methods 16:519–525. https://doi.org/10.1038/
s41592-019-0427-6
Tolić N et al (2017) Formularity: software for automated formula assignment of natural and other
organic matter from ultrahigh-resolution mass spectra. Anal Chem 89:12659–12665. https://doi.
org/10.1021/acs.analchem.7b03318
Tsugawa H et al (2015) MS-DIAL: data-independent MS/MS deconvolution for comprehensive
metabolome analysis. Nat Methods 12:523–526. https://doi.org/10.1038/nmeth.3393
Tsugawa H et al (2016) Hydrogen rearrangement rules: computational MS/MS fragmentation and
structure elucidation using MS-FINDER software. Anal Chem 88:7946–7958. https://doi.org/
10.1021/acs.analchem.6b00770
van der Laan T et al (2021) Fractionation platform for target identification using off-line directed
two-dimensional chromatography, mass spectrometry and nuclear magnetic resonance. Anal
Chim Acta 1142:28–37. https://doi.org/10.1016/j.aca.2020.10.054
70 R. Hissong et al.

Vaniya A, Fiehn O (2015) Using fragmentation trees and mass spectral trees for identifying
unknown compounds in metabolomics. Trends Anal Chem 69:52–61. https://doi.org/10.1016/
j.trac.2015.04.002
Vinaixa M et al (2016) Mass spectral databases for LC/MS- and GC/MS-based metabolomics: state
of the field and future prospects. TrAC Trends Anal Chem 78:23–35. https://doi.org/10.1016/j.
trac.2015.09.005
Wang H et al (2015) Systematic optimization of long gradient chromatography mass spectrometry
for deep analysis of brain proteome. J Proteome Res 14:829–838. https://doi.org/10.1021/
pr500882h
Wang M et al (2016) Sharing and community curation of mass spectrometry data with global
natural products social molecular networking. Nat Biotechnol 34:828–837. https://doi.org/10.
1038/nbt.3597
Wang X et al (2018) Target-decoy-based false discovery rate estimation for large-scale metabolite
identification. J Proteome Res 17:2328–2334. https://doi.org/10.1021/acs.jproteome.8b00019
Wang S, Kind T, Tantillo DJ, Fiehn O (2020) Predicting in silico electron ionization mass spectra
using quantum chemistry. J Chem 12:63. https://doi.org/10.1186/s13321-020-00470-3
Watson DG (2013) A rough guide to metabolite identification using high resolution liquid chroma-
tography mass spectrometry in metabolomic profiling in metazoans. Comput Struct Biotechnol J
4:e201301005. https://doi.org/10.5936/csbj.201301005
Whiley L et al (2019) Systematic isolation and structure elucidation of urinary metabolites
optimized for the analytical-scale molecular profiling laboratory. Anal Chem 91:8873–8882.
https://doi.org/10.1021/acs.analchem.9b00241
Willoughby PH, Jansma MJ, Hoye TR (2014) A guide to small-molecule structure assignment
through computation of (1H and 13C) NMR chemical shifts. Nat Protoc 9:643–660. https://doi.
org/10.1038/nprot.2014.042
Wishart DS (2009) Computational strategies for metabolite identification in metabolomics.
Bioanalysis 1:1579–1596. https://doi.org/10.4155/bio.09.138
Wishart DS et al (2006) DrugBank: a comprehensive resource for in silico drug discovery and
exploration. Nucleic Acids Res 34:D668–D672. https://doi.org/10.1093/nar/gkj067
Wishart DS et al (2018) DrugBank 5.0: a major update to the DrugBank database for 2018. Nucleic
Acids Res 46:D1074–d1082. https://doi.org/10.1093/nar/gkx1037
Wishart DS et al (2022) HMDB 5.0: the human metabolome database for 2022. Nucleic Acids Res
50:D622–D631. https://doi.org/10.1093/nar/gkab1062
Wong JM et al (2016) Benzoyl chloride derivatization with liquid chromatography-mass spectrom-
etry for targeted metabolomics of neurochemicals in biological samples. J Chromatogr A 1446:
78–90. https://doi.org/10.1016/j.chroma.2016.04.006
Wu Q et al (2019) Enhancing coverage in LC-MS-based untargeted metabolomics by a new sample
preparation procedure using mixed-mode solid-phase extraction and two derivatizations. Anal
Bioanal Chem 411:6189–6202. https://doi.org/10.1007/s00216-019-02010-x
Xue J et al (2020) Enhanced in-source fragmentation annotation enables novel data independent
acquisition and autonomous METLIN molecular identification. Anal Chem 92:6051–6059.
https://doi.org/10.1021/acs.analchem.0c00409
Yoo HJ, Liu HC, Hakansson K (2007) Infrared multiphoton dissociation and electron-induced
dissociation as alternative MS/MS strategies for metabolite identification. Anal Chem 79:7858–
7866. https://doi.org/10.1021/ac071139w
Zaitsu K, Hayashi Y, Kusano M, Tsuchihashi H, Ishii A (2016) Application of metabolomics to
toxicology of drugs of abuse: a mini review of metabolomics approach to acute and chronic
toxicity studies. Drug Metab Pharmacokinet 31:21–26. https://doi.org/10.1016/j.dmpk.2015.
10.002
Zhang Z et al (2018) Reverse and random decoy methods for false discovery rate estimation in high
mass accuracy peptide spectral library searches. J Proteome Res 17:846–857. https://doi.org/10.
1021/acs.jproteome.7b00614
Compound Identification Strategies in Mass Spectrometry-Based. . . 71

Zhao S, Li L (2020) Chemical derivatization in LC-MS-based metabolomics study. TrAC Trends

Anal Chem 131:115988. https://doi.org/10.1016/j.trac.2020.115988
Zheng J, Mandal R, Wishart DS (2018) A sensitive, high-throughput LC-MS/MS method for
measuring catecholamines in low volume serum. Anal Chim Acta 1037:159–167. https://doi.
org/10.1016/j.aca.2018.01.021
Zhou JT, Yin YX (2016) Strategies for large-scale targeted metabolomics quantification by liquid
chromatography-mass spectrometry. Analyst 141:6362–6373. https://doi.org/10.1039/
c6an01753c
Zhou Z et al (2020) Ion mobility collision cross-section atlas for known and unknown metabolite
annotation in untargeted metabolomics. Nat Commun 11:4334. https://doi.org/10.1038/s41467-
020-18171-8
Metabolomics and NMR

Ryan T. McKay

Contents
1 Objectives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
2 Brief History . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
2.1 Potential . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
2.1.1 A Warning . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
2.2 Definitions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
2.2.1 Metabolome . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
2.2.2 Metabo -Nomics or -Lomics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
2.2.3 Metabolites . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
3 Expectations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
3.1 Reality . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
3.1.1 Key Considerations for NMR . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
4 NMR and MS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
4.1 Complementation: Can We Just Skip to the End? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
4.2 Metabolomics via Mass Spectrometry . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
4.3 Metabolomics by NMR . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
5 Nuclear Magnetic Resonance Spectroscopy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
5.1 NMR Experiment for Metabolomics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
5.1.1 First Impressions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82
5.2 Liquids NMR . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82
5.2.1 Progression to Metabolomics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
5.3 Samples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
5.3.1 Preparation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
5.3.2 NMR Tubes, Spinners, and Side-Bands . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
5.3.3 Gathering and Handling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86
5.3.4 In the Tube . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89
5.4 NMR Spectrometer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90
5.4.1 Probe . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90
5.4.2 Console . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92

R. T. McKay (✉)
Department Chemistry, College of Natural and Applied Sciences, University of Alberta, Edmonton,
AB, Canada
e-mail: [email protected]

# The Author(s), under exclusive license to Springer Nature Switzerland AG 2022 73

V. Ghini et al. (eds.), Metabolomics and Its Impact on Health and Diseases,
Handbook of Experimental Pharmacology 277, https://doi.org/10.1007/164_2022_616
74 R. T. McKay

5.4.3 Host Computer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92

5.4.4 Experiment Pulse Sequence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
5.4.5 Parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94
5.4.6 Phase Cycle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95
5.4.7 Temperature Calibration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95
5.4.8 Tune/Match . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96
5.4.9 Excitation Pulse . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97
5.4.10 Gain . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98
5.4.11 Pulse Field Gradients . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99
5.5 Manuscript Reporting Parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100
5.5.1 Pulse Width . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100
5.5.2 Delays: Relaxation, Equilibrium . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101
5.5.3 Post-Acquisition: Weighting Function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102
5.5.4 Corrections: Linear Prediction, Baseline, Phasing . . . . . . . . . . . . . . . . . . . . . . . . . . . 102
5.5.5 Referencing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103
6 Future Ideas . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103
6.1 Solids NMR Metabolomics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103
6.2 Working with Raw FID . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104
6.3 PureShift 1H-1H J-Coupling Removal . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104
7 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105
Appendix A: Example Pulse Sequence: Metabolomics 1D-1H “Metnoesy” . . . . . . . . . . . . . . . . . . 106
Appendix B: Example Experimental Section . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 108

Abstract
The purpose of this manuscript will be to convince the reader to dive deeper into
NMR spectroscopy and prevent the technique from being just another “black-
box” in the lab. We will try to concisely highlight interesting topics and supply
additional references for further exploration at each stage. The advantages of
delving into the technique will be shown. The secondary objective, i.e., avoiding
common problems before starting, will hopefully then become clear. Lastly, we
will emphasize the spectrometer information needed for manuscript reporting to
allow reproduction of results and confirm findings.

Keywords
Automation · Biochemistry · Liquids · Metabolites · Metabolomics · NMR ·
Nuclear magnetic resonance · Quantitation · Small molecule · Solvent
suppression · Spectrometry

1 Objectives

The purpose of this manuscript will be to convince the reader to dive deeper into
NMR spectroscopy and prevent the technique from being just another “black-box”
in the lab. We will try to concisely highlight interesting topics and supply additional
references for further exploration at each stage. The advantages of delving into the
technique will be shown. The secondary objective, i.e., avoiding common problems
Metabolomics and NMR 75

before starting, will hopefully then become clear. Lastly, we will emphasize the
spectrometer information needed for manuscript reporting to allow reproduction of
results and confirm findings.

2 Brief History

Metabolomics involves a few key steps: hypothesis of perturbation observation by

metabolomic measurement (e.g., identifying a disease), sample selection and
handling, instrumentation setup/management, and finally processing and analysis.
In any living system any number of molecules regularly and/or responsively change
their prevalence and activity, and thus cellular function requires constant feedback/
control of metabolites through kinetic and energetics. Therefore, at any stage,
metabolites certainly change their presence and/or abundance, but can we
detect them?
Seeking to elucidate, understand, and be able to apply our knowledge about these
changes is the foundation of metabolomics, and there has been an astounding
amount of development regarding metabolomics (Lindon et al. 1999; Wang and Li
2020; Tenori et al. 2020; Giraudeau 2020) spanning the last 20 years. Nuclear
magnetic resonance (NMR) spectroscopy has been one of two primary tools from
the start, with the second being mass spectrometry (MS), and while there are
certainly other techniques, NMR and MS are arguably the most widely applied.
Both NMR and MS are certainly powerful analytical techniques, each with
advantages and disadvantages when compared directly or to other instrumental
methods. This will be detailed later (e.g., see section NMR and MS Competition
and Complementation).
Dramatic NMR spectrometer improvements including common access to higher
magnetic fields, cryogenically cooled probes, new robotic sample handling with
temperature control, automated software calibration/optimization/acquisition/
processing and analysis have made the instrument “black-box” mode ever more
seductive. There are many of these aspects that (with some interesting background
and a little emphasis) can be efficiently optimized. Specifically understanding the
fundamentals of sample management, instrument preparation, and analysis
expectations can make project planning and setup easier with a concomitantly higher
probability of success with a more reliable, comparable, and efficient study before
even starting. Please note that by comparable, we mean the ability to validate and
take data from other studies, instruments, and/or facilities for inclusion into your data
analysis (or vice versa), not just the data collected in one location and/or one
dedicated instrument (see (Lacy et al. 2014; Sokolenko et al. 2013) and references
therein). It is often assumed that collecting data on one dedicated instrument
removes multi-facility/multi-instrument complications. However while using a sin-
gle dedicated instrument should achieve consistent precision, there is no guarantee
of accuracy (all samples may be equally inaccurate) nor does this assumption
consider unavoidable changes in instrumentation over time, e.g. repairs,
replacements, updates, etc. (see Sect. 5.4 and other sections below).
76 R. T. McKay

2.1 Potential

Metabolomics has been utilized now for several decades (see (Emwas et al. 2020;
Finco et al. 2016; Heather et al. 2013; Kenny et al. 2010; Li et al. 2012; Psychogios
et al. 2011; Rasmussen et al. 2012) and references therein) but has not reached the
promised potential for applicable personalized/precision medicine. While there are
certainly specifically applied findings, there has not been the explosion of the
predicted novel medical treatments. Very recently mRNA vaccine developments
have introduced new concerns regarding the possible over-production of proteins/
metabolites (see review (Pardi et al. 2018)) and highlighted the application of
metabolomics for monitoring.

2.1.1 A Warning
It is important to remind the reader that the promise of marketability drove much of
the initial science. Metabolomics starting in the early 2000s, was envisioned as a
novel tool with exceedingly high profitability that would be: rapid, reliable, repro-
ducible, using easily acquired samples (e.g., urine) without extensive sample manip-
ulation (Bingol et al. 2016; Tayyari et al. 2013), and contain directly interpretable,
widely applicable, and useful results. Companies quickly jockeyed to be the first to
sell and inhabiting their NMR spectrometers (and now perhaps benchtop units
(Izquierdo-Garcia et al. 2020)) throughout clinical testing facilities across the
globe. For any new marketable technique to survive it must either be unique and
advantageous (i.e., provide a novel result) or be extremely competitively priced
when compared to existing technology (i.e., faster, cheaper, etc.). The corporate
emphasis was on finding unique biomarkers for dramatic high-profile diseases (e.g.,
cardiovascular disease, cancers, etc.) as quickly as possible and thus establish
lucrative patents and contracts. This did not encourage systematic, calm evaluation,
nor retesting, and that rush may have inadvertently hurt the entire field. Subsequent
validation studies (Emwas et al. 2020; Lacy et al. 2014; Sokolenko et al. 2013;
Markley et al. 2017) have begun to fill in the gaps, however questions regarding the
cross-validation of NMR data persists, e.g., site to site and/or study to study along
with assessment (Rocca-Serra et al. 2016; Cassiède et al. 2017).

2.2 Definitions

There has been ambiguity in the literature regarding key definitions and to avoid
confusion we will quickly define our working interpretation below.

2.2.1 Metabolome
The metabolome is commonly defined as all small molecules in the mass range of
50 to 1,500 Da (not a hard limit as certainly lipids can quickly exceed), associated
with a particular organism (Dunn et al. 2011; Psychogios et al. 2011; Wolfender
et al. 2013; Zulyniak and Mutch 2011). This includes all the various complexes,
sizes, and repetitive units of amino acids, lipids, carbohydrates, and other organic
Metabolomics and NMR 77

molecules collectively termed “metabolites” and are involved in all stages of metab-
olism both from naturally internal (endogenous) starting points, and those
originating from external (exogenous) sources. External include ingested foods or
pharmaceuticals, and the gastrointestinal metagenome along with subsequent
by-products, more colloquially referred to as the “gut flora and fauna” with these
microbes apparently playing a diverse role as our understanding expands.

2.2.2 Metabo -Nomics or -Lomics

Here are two important terms that, depending on which manuscript you reference,
can have subtle but important differences. Metabonomics was generally given a
wider definition of studying all relevant interactions over an enter species (i.e., the
subject’s metabonome). Metabolomics was mostly considered to be a more focused
subset in a defined case. Also NMR and MS manuscripts tended to use one of the
terms, but not the other. Metabolomics has become the most common and to avoid
confusion we will use only metabolomics, defining it as studying the metabolites in a
defined situation (e.g., human urine metabolites detectable by NMR). Readers
interested in the distinctions are directed to excellent reviews such as (Bouatra
et al. 2013; Psychogios et al. 2011; Macel et al. 2010; Gibney et al. 2005) and
papers therein.

More “-Omics”?
For a review, one cannot simply ignore the “-omics” flood as it now also includes
subtopics of metabolomics. Logically we start with the genome, then onto the
regulation of transcription with subsequent modifications. Then we consider effec-
tive production/regulation/degradation (e.g., protein turn-over), and finally the basic
building block level resulting in functional metabolism. These levels have an
expanding and/or encompassing “-omics” (Ragguett and McIntyre 2020) associated
with their study (i.e., genomics, transcriptomics, proteomics, metabolomics), and
even more recent approaches such as pharmacogenomics or pharmacometabolomics
(Van Der Wouden et al. 2020; Emwas et al. 2021; Lasky-Su et al. 2021; Vignoli
et al. 2019; Sherlock and Mok 2019).
There are many new subgroupings of of “omics” including Lipidomics (see a
recent full Nature review1), Foodomics (Balkir et al. 2021; Picone et al. 2022; Valdés
et al. 2021), and Elemental Metabolomics (Andersson et al. 2021; Edison et al. 2020;
Niziol et al. 2021; Zhang et al. 2018). In case readers have not had enough “-omics,”
there is a relatively new though well established and pertinent sub-group focusing on
the changes of metabolites over time called Fluxomics, see the comprehensive
review by Giraudeau (2020) and also (Emwas et al. 2020) for additional references.
Of key interest is the use of specific NMR labelling (Xu et al. 1999) in metabolites.

1
https://www.nature.com/subjects/lipidomics.
78 R. T. McKay

2.2.3 Metabolites
To properly address the term metabolites, we will use the excellent definition
provided by Tenori et al. in their concise review (Tenori et al. 2020).

Metabolites are the small molecules produced by, or taking part in, the chemical reactions
due to biochemical activities (i.e., the metabolism) in living organisms, and their levels can
change according to pathophysiological or environmental factors.

3 Expectations

As metabolites encompass the end stage of cellular chemical regulation/function, it

is postulated that this should be the easiest way to detect and ascertain the cause of
fundamental upstream cellular changes. Essentially, we are hoping to observe an
amplification of small changes from further up the “-omic” ladder therefore making
disease detection and analysis faster and easier.

3.1 Reality

While DNA is a single chemical class comprising only four nucleotides, immense
molecular stability, and a second strand as an immediately available backup copy for
error correction, moving downstream to the metabolites suddenly expands to
thousands if not tens of thousands of arguably critical molecules. These downstream
molecules can be quickly and substantially impacted by even small changes at the
DNA or transcription level. Metabolite concentrations can normally and
dynamically range by several orders of magnitude (Bouatra et al. 2013). Immedi-
ately the reader can see that while there may indeed be amplification, i.e., a
downstream effect any single “snap-shot” acquisition of metabolites (regardless of
how detailed and accurate) may have limited practical applicability to any disease
detection. This gets even more complicated regarding any practical translation to
personalized-medicine (“bedside”) treatment. Also metabolites have an immense
range of molecular variability in terms of primary/secondary structure, function,
modification, and lifespan.

3.1.1 Key Considerations for NMR

For NMR, concentration is perhaps the most crucial aspect due to inherent detection
limitations. Any low concentration biomarkers exist in a literal sea of much higher
concentration compounds, complexes, and aggregates. Essentially the detected
signal’s “dynamic range” (stealing an electrical engineer and audiophile term) is
incredibly large spanning many orders of magnitude regardless of the technique
selected. This is also directly relevant to the source/type of tissue involved that can
result in vastly different concentrations of metabolites. Then there are complications
regarding simple collection, e.g., considering human-based samples: the time of day,
method of isolation/collection, fasting level of donor, female or male, age, activity,
Metabolomics and NMR 79

genetic background, environmental influences, etc. We may also need to consider

not a single biomarker, but instead a group or family of inter-related metabolites
responding directly or indirectly to other metabolites. On top of all those
considerations, when then have the metabolomic influence of the “gut metabolome.”
Specifically the natural beneficial and negative opportunistic organisms existing in
our bodies consuming and producing metabolites as part of their normal functions
(Visconti et al. 2019).
The key point is substantial planning, and experimental preparation will likely be
required to minimize all these complications. We will either address these directly
and/or provide references to assist the reader.

4 NMR and MS

4.1 Complementation: Can We Just Skip to the End?

Metabolomics by any method seeks a statistically relevant and consistent change

(s) in a single or multiple observed variables. Ideally, we would like to see the novel
presence or absence of a recognizable and preferably unique signal associated to a
specific molecule which can be related to a disease, i.e., literally the lightbulb on/off.
A disease “marker” that could be regularly, rapidly, and easily followed.2
There is far more likely a dynamic and complex series of cascading and/or
interlocking molecular pathways responding to change(s) with dependency on the
complexity of the organism being monitored. Feedback loops, changes of enzymatic
expression levels, regulation of genetic expression, alterations of metabolism rates,
changes in diet and/or activity due to the subject’s phenotypic expression (e.g., how
bad they may feel), self-medicating, the body actively trying to re-establish equilib-
rium; all should contribute to an interdependence if not system wide response.
So is there any single technique that can identify all possible metabolites in all
samples regardless of origin? No, but there are several techniques offering different
strengths and weaknesses, and there is always the added potential of linking together
different techniques. The so-called “hyphenated” methods where one runs the
sample consecutively through multiple instruments and gathers the acquired data
to make a far more powerful determination.
The problems with “hyphenation,” i.e., the myth of interconnected instruments
with streamlined throughput, such as mass spectroscopy (MS), high-performance
liquid chromatography (HPLC), fast protein liquid chromatography (FPLC), gas
chromatography (GC), circular dichroism (CD) is that they may not always work
well together, or the sequence is important to the result, or one needs multiple
identical samples. An example is obtaining NMR data first, which then MS
contaminates the sample with deuterium (for NMR “lock” see below), or capillary
electrophoresis (CE) then MS (Qiu et al. 2020). There are always the concerns about

2
This overly simplified one-molecule/one-marker situation is highly unlikely.
80 R. T. McKay

sample cross-contamination, how to adequately resolve signals in mixtures, incor-

rect assignment, or obscured/missing assignment. Then there is the practical problem
of maintaining and constantly checking that the inter-related instruments are work-
ing consistently. Striving for any technique that provides fully automated sample
preparation, instrument optimization, sample handling, data acquisition, sample
storage, data processing and finally analysis is presently impossibly ambitious, but
a semi-automated approach might be easier, faster, and more reliable. Like NMR
structure-function protein backbone assignments – semi-automated, i.e., get the
computer to do the routine work and then present the challenges to the experienced
instrument operator who can use visual pattern recognition and experience to make
the more difficult decisions seems to be the most efficient and reliable. As the title of
this review implies, “black-boxing” the problem leads to (at best) consistent errors
and more likely unusable data. At worst, false information makes it into the peer-
reviewed literature taking a huge amount of work to recognize and correct later.
Unfortunately, we cannot skip to the end and we will need to discuss and understand
the instrumentation.

4.2 Metabolomics via Mass Spectrometry

The author approaches mass spectrometry (MS) with admittedly little practical
experience surrendering any in-depth evaluation of MS to the experts, e.g., see
highly recommended reviews (Wang and Li 2020; Alseekh et al. 2021)and
references therein. While mass spectrometry has undeniably superior solute sensi-
tivity (e.g., microlitre or sub-microlitre volumes with relatively low concentrations
(Li et al. 2020) per unit instrument time, it does come with a literal financial cost and
a cost in terms of total experimental length, monitoring/compensation for separation
technique(s), necessity of quality/control sampling, and finally the destruction of the
sample. It is important to note that MS instrumentation has a substantially smaller
initial instrument cost, along with smaller maintenance costs (e.g., cryogens). A
critical distinction is that MS usually requires some form of sample separation
(Petrović et al. 2005; Korfmacher 2005; Alseekh et al. 2021): chromatographic
(liquid or gas), electrophoretic, or based on ion mobility that can perturb the types
and quantities of measured metabolites. The separation efficacy changes over time
requiring continuous monitoring, evaluation, and correction during analysis, e.g.,
separatory columns degrade nonuniformly over time requiring a calibration of
resulting metabolites. Sample separation and detection limitations (Wang and Li
2020) can also be linked to the particular detector such as3: Fourier transform ion
cyclotron resonance (Marshall et al. 1998, 2007; Nikolaev et al. 2016), ion trap
(Todd and March 1999), Orbitrap and linear ion (Perry et al. 2008), time-of-flight
(Boesl 2017), and quadrupole (Linge and Jarvis 2009) to name some of the most
common. Evaluation of the potential and realized application has been extensively

3
Detector order is alphabetical only, not intended to imply frequency nor capabilities.
Metabolomics and NMR 81

covered and readers are directed to representative articles (Trifonova et al. 2021;
Alseekh et al. 2021; Lasky-Su et al. 2021; Wang and Li 2020). In the end MS has
become an extremely successful method accurately detecting thousands of
molecules and at much lower concentrations that can be practicably detected
by NMR.

4.3 Metabolomics by NMR

Metabolomics by NMR has been vastly reviewed (Tenori et al. 2020; Giraudeau
2020; Giraudeau et al. 2014, 2015; Tavares et al. 2015; Halabalaki et al. 2014;
Bingol and Brüschweiler 2014; Bouatra et al. 2013; Wolfender et al. 2013; Lubbe
et al. 2013; Ellinger et al. 2013; Heather et al. 2013; Smolinska et al. 2012; Dunn
et al. 2011; Macel et al. 2010; Lindon et al. 2007; Beckonert et al. 2007; Emwas et al.
2018, 2019; Stringer et al. 2016). There is also a highly recommended NMR book
with an entire chapter dedicated to the practical aspects of metabolomics and NMR
sample handling/data/processing (Teng 2012).
Metabolomics by NMR contains all the common instrumental concerns (e.g.,
consistency, detection, assignment of signals, etc.), but a unique aspect of NMR is
that each atom, even in the same molecule, can have a specific resonance. This is
both the strength, i.e., the ability to resolve an atom based on its magnetic environ-
ment, and weakness. Every atom resonating by itself provides little cumulative
signal. Each relevant atomic signal must be detected, assigned, and then analyzed.
All of this with the massive assumption that no unexpected external change(s), other
than the central hypothesis of the study will perturb the measurement.
We will address that perturbation assumption and more below.

5 Nuclear Magnetic Resonance Spectroscopy

We will highlight key points in the subsections below to detail important information
from our experience to hopefully improve the reader’s future research studies. The
points will use a special format indicated by Note.

5.1 NMR Experiment for Metabolomics

To start, a metabolomics NMR experiment (i.e., the “pulse sequence” itself) ideally
must be easy to acquire, i.e., any required hardware is commonplace, minimal setup,
robust, reliable, and reproducible. The instrument is assumed to be constantly
maintained, tested, and repaired be experienced operators to ensure predictable
performance. Next there are literally hundreds of NMR experiments (Berger and
Braun 2004; Braun et al. 1998), and the number of individual and inter-related NMR
parameters for each experiment is often overwhelming. The initial selection, testing,
and maintenance can be a full-time occupation for facilities (Reynolds and Enriquez
82 R. T. McKay

2002). The person contemplating a metabolomics NMR project must either acquire
equipment and become this, or already have the equipment and technical personnel
available.

5.1.1 First Impressions

Users commonly first experience NMR through introductory organic chemistry
courses (Derome 2013). This exposure can expand into some inorganic chemistry
and certainly into biomolecular NMR (Ban et al. 2017; Göbl et al. 2014; Gardner and
Kay 1998; Ziarek et al. 2018), but very rarely into the hardware itself. NMR is also
commonly involved in natural products identification, analytical food
ID/confirmation (e.g., wine, beer, scotch, juice, honey) (Spraul et al. 2015; Esslinger
et al. 2015; Link et al. 2014; Kew et al. 2017; Link et al. 2014; Sandusky and Raftery
2005) and legal evaluation (i.e., spiking) (Lesar et al. 2011). Other experiences with
NMR could perhaps involve academic spin-off pharmaceutical/commercial
applications (Lindon et al. 2007; Duarte et al. 2014; Lepre 2011; Shuker et al.
1996). This type of experience lends itself well to metabolomics as most metabolites
are small and will yield NMR spectra similar to what users have experienced.
However, the sheer number of compounds in a common biological sample will
certainly not be typical of an organic chemistry problem, nor does this address how
the instrument operates.
NMR originally started as an interesting physics phenomena and expensive pH
meter (Bull et al. 1964). The field has certainly grown since the 1950s and 1960s in
terms of distribution/application of instruments, stable/achievable magnetic field
strength, versatility, sensitivity, consistency of equipment performance, and reduced
cost of operation. Instrument consoles have gotten smaller while magnets have
become increasingly powerful and shielded thus reducing the laboratory footprints
for installation and/or the number of instruments required.

Note
While the accepted “NMR standard” is 0.1% ethylbenzene with 0.01% TMS
in deuterated chloroform, Benchtop NMR manufacturers have created their
own standard using 1% ethylbenzene instead. Therefore, the reported bench-
top 1H NMR signal-to-noise ratios must be either divided by a factor of 10 to
compare to previously published standards and/or the user must recognize a
factor of 100 for required experiment time to achieve comparable results.

5.2 Liquids NMR

While it is fun to teach and delve into the world of NMR theory (e.g., providing
useful analogies describing each atomic nuclei as a little bar magnet spinning at its
own frequency depending on the magnetic field strength etc.), we simply do not have
the space in this review; especially considering the wealth of previously published
Metabolomics and NMR 83

information by extremely articulate lectures/authors/presenters on the subject. The

interested reader is first guided to very well-established books4 on the subject, for
example, Hore 2015; Hore et al. 2015; Vögeli 2014; Silverstein et al. 2014; Derome
2013; Keeler 2010; Bakhmutov 2005; Harris 1986; Freeman 1997; Brown 2016;
Cavanagh et al. 2006; Levitt 2001; Zerbe and Jurt 2013; Wüthrich 1986 aimed at
audiences ranging from the interested to the specialist/expert. These resources will
then escort the reader into the vast fount of peer-reviewed manuscripts and reviews.
It is worth mentioning that NMR can directly cover/monitor processes time scales
covering many orders of magnitude (Ziarek et al. 2018). From the medical/biochem-
istry aspect protein structure function is one of the most recognized applications of
NMR, but protein bioNMR is usually most effective with a solitary type of molecule
and at relatively high concentrations (e.g.,>1 mM) and there is also the need
(Amoureux et al. 2008) for isotopic enrichment, i.e., 13C, 15N, and even 2H (Gardner
and Kay 1998; Hiroaki 2013) or selectively enriched at key positions. For reviews
see (Kay and Frydman 2014; Kay 2016).

Note
It is important to note that much of the background NMR material does not
address complex or “strong” 1H-1H coupling regarding metabolomics
assignments (see sections below).

Once the reader goes beyond the simple spin-1/2-coupling “tree” diagrams into
the higher order (also called second order, complex, or strong coupling depending on
the terminology used), we discover spectrometer/magnetic field-dependent patterns
(Foroozandeh et al. 2014; Bain et al. 1994). This is important for software selection,
analysis, and anyone aiming to assign biomarkers (Mercier et al. 2011; Tredwell
et al. 2011; Weljie et al. 2006). Using a reference database that does not take this into
account and/or have information for the magnetic field used will be more difficult to
use and require operators to be more experienced (Lacy et al. 2014; Mercier et al.
2011; Tredwell et al. 2011; Weljie et al. 2006).

5.2.1 Progression to Metabolomics

Metabolomics evolved in earnest in the early 2000s (Lindon et al. 2007; Lauridsen
et al. 2007; Wang et al. 2010; McGrath et al. 2007; Weljie et al. 2006; Saude et al.
2006; Tilgner et al. 2019). There were earlier endeavors seeking insight into what
was then referred to as “in-born errors,” but the technology needed time to catch up
to the intellectual concepts (Lehnert and Hunkler 1986). The field5 is now well
established with typical superconducting magnets now allowing observation of

4
No implied order, and naming a few of the author’s personal favourites. There are many others
certainly worth the reader’s attention and the provided list is not exhaustive.
5
Please forgive the puns.
84 R. T. McKay

hydrogen from 400 to 800 MHz (i.e., 9.39 to 18.8 Tesla static magnetic field
strength) and offered from a range of vendors.6
While concerns about data processing and analysis are equally important, there
are many publications and reviews focusing on issues involving time domain to
frequency signals such as signal enhancement (apodization), deconvolution, integra-
tion, and then reliably identifying and assigning signals or signal components
(Schönberger et al. 2015) to particular metabolite resonances. Readers seeking
even more information are directed to a representative set of papers and references
therein (Bartel et al. 2013; Beckonert et al. 2007; Bingol et al. 2016; Dudzik et al.
2018; Dunn et al. 2011; Eghbalnia et al. 2017; Ellinger et al. 2013; Emwas et al.
2016, 2018, 2019; Wang and Li 2020; Kohl et al. 2012; Krishnamurthy 2013; Lacy
et al. 2014; Parsons et al. 2009). We will now focus on the practical aspects of
sample and NMR instrumentation, e.g., see Chap. 3 of reference (Cavanagh et al.
2006) and Chap. 2 of reference (Zerbe and Jurt 2013).

Minimizing Problems
The detection and elimination of “confounders,” i.e., signals or influences on signals
that are not dependent on the hypothesized change (e.g., disease vs. healthy) but
instead are artifacts and/or errors in sample preparation and/or instrumentation, must
be a major focus for everyone involved in metabolomics. We wish to remove (or at
least minimize) as many variables as possible prior to the acquisition of NMR data
(Athersuch et al. 2013; Gibney et al. 2005; Meissner et al. 2014; Staab et al. 2010;
Zulyniak and Mutch 2011). This requires prior awareness, continued attention,
precise planning throughout, and finally careful preparation, and this leads us
directly to the sections below regarding sample and instrumentation details and
recommendations.

5.3 Samples

5.3.1 Preparation
Sample selection, preparation, consideration for repetitive sampling, and handling
consistency are all incredibly important. This is only the first step, but the easiest to
get wrong, and sometimes without being determinable until very late in the study.

Note
The Brian Sykes’ First rule of NMR applies here, i.e., “Garbage” in equals
“Garbage” out. Essentially no matter how good the operator nor expensive the
instrumentation, if your sample is poor, your results are poor.

6
Unfortunately in 2014, Agilent (who had purchased Varian Inc. in 2010) exited the NMR market.
This has added a level of uncertainty for those still needing repair/support/parts for their massive
equipment investments, especially in the ever-expanding austerity environments facing publicly
funded academic institutions.
Metabolomics and NMR 85

Any complication such as contamination in the sample solution, contamination

on the outside of the tube,7 material not dissolved and floating, i.e., hair, glass wool,
debris etc., precipitated solute, a scratched NMR tube (see Sects. 3.2.3–3.2.5 in
(Derome 2013)), and even just improperly positioned in the NMR spinner, can all
have dramatic impacts on the overall quality and therefore the consistency of the
recorded data. Without everyone involved in the study agreeing and following
precise protocols (i.e., user/handling confounders), the study is doomed to problems.

Note
The key point here is absolute consistency, from everyone. Despite careful
planning all the group members that will be handling samples must understand
the stringent necessity of consistent sample treatment.

5.3.2 NMR Tubes, Spinners, and Side-Bands

Traditionally NMR tubes are placed in carriers called “spinners.” The spinners allow
the insertion/removal of a sample via compressed air shuttling the sample up to
the top and down into the magnet core for observation. The spinner can also suspend
the tube while inside the magnet on a slight cushion of air to facilitate rotation of the
sample (i.e., parallel axis to the NMR magnet bore tube)(Harris 1986). The rotation
of ~15–20 Hz was used to average or “spin-out” inhomogeneities in the magnetic
field which were difficult to compensate using early NMR systems. Poorly
manufactured tubes would cost less but have lower tolerances in concentricity
(i.e., centering of inner bore of the tube properly in the glass) and/or camber (i.e.,
straightness of the tube). These imperfections during spinning would result in NMR
spectral artifacts called spinning side-bands due to vibrations and imperfections in
the magnetic field. High-quality NMR tubes can be purchased albeit with much
higher prices.8 For further information on shims (i.e., the small electro-magnets)
used to optimize the magnetic field, please see references Liu et al. 2014; Maudsley
et al. 1984; Van Zijl et al. 1994; van Zijl 1987. The basic results were narrower/taller
peaks above the baseline noise with better resolution (less overlap). One would then
logically wonder why we do not routinely “spin” metabolomics samples (or all NMR
samples)? First the cost of the high precision tubes is a major barrier to large-scale
studies, and even the best tubes would still experience some spinning-induced “side-
bands” (i.e., individual or sets of symmetric artifactual peaks at specific distances
from the real signal). These artifacts cannot be completely removed, resulting in
additional peaks in the spectra. This causes confusion during interpretation, espe-
cially in heavily overlapped regions.

7
https://blogs.umass.edu/weiguoh/?cat=81121.
8
X and Y based shims (termed the non-spin shims) could be optimized more easily and the overall
spectral line shape was normally much improved.
86 R. T. McKay

Modern instruments are now usually housed in excellent laboratory environments

with exceptionally homogeneous magnetic fields due to refined manufacturing and
extended shim sets, therefore spinning is no longer necessary. This has been
published by Zerbe and Jurt (2013) and we confirm from our personal experiences.
Lastly, many NMR experiments using pulse field gradients for coherence selection
suffer dramatically from sample spinning.

Note
In our experience, the performance gains from spinning samples do not make
up for the variability and problems associated with spinning side-bands. We
do not use spinning nor recommend for metabolomics.

Additionally
The NMR tube manufacturing common practice of putting a magnetic field
strength recommendation (e.g., eco tube for 400 MHz or below, or precision
NMR tube for 600–800 MHz) assumes spinning of samples and is not
necessary for non-spun and high throughput studies. In our experience, basic
economical NMR tubes are sufficient in nearly all sample applications.

5.3.3 Gathering and Handling

Long before we can begin acquiring NMR data, we have to trust that all samples
were handled uniformly prior to their arrival at the NMR preparation stage (Barton
et al. 2008; Bernini et al. 2011; Dumas et al. 2006; Lauridsen et al. 2007; Pinto et al.
2014; Rist et al. 2013; van der Sar et al. 2015). Unfortunately, this is more difficult
that may be initially considered, and we will present examples from our first human
urine study.

Volunteers
First, can enough volunteers be found? How does one know how many or even how
to appropriately attract volunteers for a study? While seemingly a trivial point it is no
easy task with modern privacy laws, e.g., how do you legally even approach people
without prior permission? A definite catch-22 situation. Our experience was this
severely slowed down our study almost to the point of cancellation regardless of
successful funding.
Metabolomics and NMR 87

Note
How to get study volunteers? This is under jurisdictional control so carefully
check your local requirements. We had to have research assistants set up tables
on campus with signs and wait for individuals to approach. Privacy rules
prevented pretty much every other idea we came up with (e.g., emails, offering
prizes/incentives, telephone, etc.).

Another example of complication, volunteers needed to be pre-selected based on

a predefined and often limited range of age, sex, weight, health, diet, medical history,
fasting prior to sampling, a sufficient number of volunteers had to be selected for
statistically relevance, and lastly enough extra volunteers needed to be found so that
a sufficient number will successfully complete the study (there are always
complications/withdrawing - see references above). All of these are no small feat.
In our case physical activity was involved in the hypothesis, therefore volunteers
needed to be selected for fitness/risk with full disclosure, etc. The question of
number of samples per volunteer needs to be statistically addressed starting with
power and sample size calculations (Jones et al. 2003) and we recommend a
collaboration with someone specializing in this area along with analysis/interpreta-
tion of principle component data.

Long-Term Studies and Storage

For longer term studies, sufficient9 samples must be acquired from each volunteer
over the course of the study to properly address the hypothesis(es), e.g., establish
baselines, and see statistically significant results. Additionally, appropriately con-
firmed preservatives and/or spiking using internal standards may be necessary for
determination of time-dependent degradation. In our case, we needed to acquire
samples to establish a minimal baseline prior to a physical stress-test, acquire
samples during the test, and then re-establish return to equilibrium. This amplified
the number of samples dramatically. Larger numbers of samples will require more
long-term storage space, with suitable and stable temperature (i.e., -80°C), and
complications on uniform handling/preparation thereafter. So many issues and we
have not even gotten started yet with the NMR data.
If these storage requirements can be met, archived samples will still need absolute
consistency in transport (time/conditions), preparation handling conditions, thawing
temperature/time/exposure, addition of NMR referencing/internal concentration ref-
erence, pH balancing, addition of deuterated lock solvent, and then finally waiting
time to acquire at the instrument. Extensive planning must be established prior to all
stages. A full-time person handling all these aspects is ideal, but we know not always
possible. Part-time or time-shared individuals must be even better organized in-order
to coordinate their combined work. Sample preparation planning needs to include

9
This is a loaded word, i.e., depending on study type, number of participants, metabolic reaction
time, etc. just to name a few likely variables.
88 R. T. McKay

very precise details, e.g., how long does it take (and what temperature do samples
reach?) for samples to transfer from long-term storage to “in the instrument.” Prior
agreement between individuals is key for consistency.

Note
While the final data acquisition and analysis (Schönberger et al. 2015) is often
strongly emphasized in publications, we recommend focusing more on the
initial preparation, handling, and organization as this will facilitate analysis.

Examples from practical experience; does one take all samples out of -80°C
storage at the same time? We did choose to remove all samples at once, but
immediately stored them in a normal +5°C refrigerator to thaw slowly and consis-
tently (i.e., samples on the outside of the group did not thaw first etc.). We also had to
address the preparation time difference between the first and last handled sample.
Specifically, how long does the last sample sit in the spectrometer robot sample
handling system waiting while all the previous samples are acquired, e.g., 7–8 min
per sample with 100 submitted samples therefore the last sample sits ~800 min
longer at room temperature than the first sample. Assuming bacterial inhibitors
(Bernini et al. 2011; Lauridsen et al. 2007) are used, any chemical reaction and/or
oxidation within the sample will undoubtedly be temperature and time sensitivity.
So there are really three key areas to consider. The first, what is the long-term
storage available? Second, conditions (time, temp, etc.) from thawing to analysis.
The last, how many times can the sample be frozen/re-frozen either for re-analysis or
due to unforeseen delays (Pinto et al. 2014; Saude et al. 2006; Saude and Sykes
2007; Rist et al. 2013). The first concern is usually dealt with by storing at a
consistent -80°C, and this has become the standard.

Note
Do not forget -80°C ultra-low temperature freezers are expensive, can have
additional electrical requirements, produce a lot of residual room heating, and
do fail. This means extensive infrastructure pre-planning and monitoring.

The second issue has a relatively simple solution, i.e., prep samples just prior to
acquiring data and use lower ambient temperatures during prep. Specifically we
recommend keeping samples at 5°C until just before instrument acquisition (and
depending on your latitude/location you may also need to include humidity control).
This includes appropriately selected robotic sample controls, and a consideration of
total time waiting for acquisition (Saude and Sykes 2007) such as limiting the size
and using a batch method for sample preparation. Preparing 40 samples at a time
instead of 400 reduces the time differences, however this substantially increases the
labor involved as more frequent/smaller batches must be prepared reducing lab
efficiency and increasing costs. It was our experience that 20–30 samples per
Metabolomics and NMR 89

batch, with two batches a day worked well allowing a single spectrometer to be
utilized efficiently. This was especially important as we were restricted to weekend
access and therefore had to maximize usage.

Note
While lowered temperature control while waiting on the instrument itself is
beneficial, it adds complications for the NMR instrumentation. Thermal equi-
librium (usually at room temperature) must be reached for the entire NMR
sample prior to data acquisition in the spectrometer receiver coil.

Thermal equilibrium can take seconds to minutes depending on the instrument

variable temperature controls. Pre-warming of samples may be necessary prior to
insertion into the spectrometer to optimize instrument time (see Sect. 5.4.7 below).

5.3.4 In the Tube

Assuming the samples have been uniformly: selected, acquired, stored, and handled;
the next step is preparation for NMR data acquisition. Liquid NMR samples are
usually in 5 mm diameter NMR tubes, between 500 and 600 μL in volume, and have
a concentration of somewhere in the 10’s mM to μM range. Concentrations are
approximate providing enough material to conclude 1H NMR experiments in con-
veniently short periods (minutes to seconds, respectively). For a detailed review of
relative and absolute NMR sensitivity see (Sanders and Hunter 1988). As we can add
scans of the same sample together building up the signal-to-noise (Hoult and
Richards 1976) ratio over time NMR therefore has no theoretical detection limit.
However, the S/N is proportional to the square of the number of scans taken (i.e., to
double the S/N requires 22 times the number of scans). We cannot realistically expect
anything resembling high throughput for samples if each requires multiple hours for
a 1D experiment. NMR detection becomes even more limiting when considering
multi-nuclear and/or multi-dimensional spectroscopy. Interested readers are referred
to Bingol and Brüschweiler (2014), Dumez (2018), Gardner and Kay (1998),
Hyberts et al. (2007), Silverstein et al. (2014), Zerbe and Jurt (2013), Ziessow
(1990) and references therein.
Regarding the standard 5 mm NMR tube, various investigators have tried smaller
tube diameters to reduce the sample volume. For example, Bruker Inc. has
introduced a 1.7 mm NMR tube diameter probe (see NMR Probes section below)
designed to maximize NMR signal detection from low volume samples. One of the
purposes is to reduce sample storage/volume/preparation costs and increase “mass
sensitivity.” Other studies have adopted shorter and thinner 3 mm NMR tubes to
reduce glass costs while optimizing throughput. Mass sensitivity is a bit of a tricky
definition and assumes that the solute’s solubility is not a limiting factor, not always
the case.
For some samples, solubility is the key limiting factor, i.e., one cannot concen-
trate the solute further without precipitation and/or molecular changes occurring.
90 R. T. McKay

Here one does not gain “mass sensitivity” through concentrating into smaller
diameter tubes. In these cases, a wider diameter sample tube physically allows
more atoms (albeit at a lower “concentration” via a greater number of overall
molecules) to be placed inside the observing receiver coils, and results in an
improved signal. Conversely if solubility is not a limiting factor and the same
molar mass of material can be concentrated into a smaller volume and smaller
NMR tube, e.g., a 1.7 mm diameter NMR tube versus the larger 5 mm tube, then
the atoms are physically closer to the receiver NMR coils and produce a larger
relative signal. Smaller tubes also become advantageous with high salt samples and
cryogenically cooled (i.e., cold) NMR probes. The S/N improvements of cold-
probes are quickly lost (Kelly et al. 2002; Nausner et al. 2010; Xiao et al. 2009) as
the ionic concentrations rise, but this phenomenon can be compensated for with
narrower or specially shaped tubes.

Note
One substantial advantage for 5 mm NMR tubes is the increased ease of
removing/recovering samples for further study and/or storage also the
increased ability to wash and re-use the NMR tubes.

Groups may opt to dispose of 3 mm or thinner NMR tubes after each use due to
cleaning/recovery difficulties and therefore the costs must be taken into consider-
ation for large-scale samples studies.

5.4 NMR Spectrometer

5.4.1 Probe
The NMR probe is critical. The probe inductively delivers relatively high-power
electromagnetic energy to the sample and influences the sample’s atomic state(s),
therefore performing the NMR experiment (called a “pulse sequence”). The probe is
also responsible for detecting the sample’s subsequent response(s), i.e., very low
power inductive signals from the precessing atoms in the sample, due to the presence
of a strong static external magnetic field (i.e., the NMR magnet).
One can easily imagine that the probe experiences a light impact from each
sample and spinner as the air pressure is decreased and sample raised/lowered.
Also contamination from any materials passed from the operator’s hands to the
NMR sample, material (e.g., dust) from the air as the sample lowers, NMR tube
breakage (e.g., multiple samples inserted without removal of the previous10) and
wear due to high-power electromagnetic induction during pulse sequences to name
but a few challenges. There are also physical moving electronic components (see

10
For high throughout facilities under manual (i.e., non-robotic) sample handling, this problem
occurs more often than you might think.
Metabolomics and NMR 91

Sect. 5.4.8 below) (Halliday et al. 2013; Derome 2013) with the temperature, ionic
strength, and dielectric constant of the solvent all coming into play. These aspects are
important and cannot simply be considered “constant.” Even if the instrument is
solely dedicated to the metabolomics user and not a multi-user facility (more
common), the NMR instrumentation experiences wear-and-tear over time.
An improperly setup probe will not yield the expected signal amplitudes and may
even rapidly devolve into relaxation and off-resonance effects, and far more compli-
cated spectra outcomes.11 On the receiver side, the resulting signal intensity will also
suffer at best decreasing certainty of measurement, or worse preventing observation
of the signal. Early metabolomics studies may not have had access to robotic sample
handling that included automated tune/match (nor even regulated temperatures for
samples awaiting data acquisition). Early studies assumed that the tune and match
could be optimized for the first sample and would not deviate perceptively from
sample to sample. In our experience, this assumption is not correct especially with
human urine samples where the salt concentrations and therefore the dielectric
matching conditions can change substantially with every sample.

Note
The practical result of improper tune/match (incorrect sample impedance
matching) is inconsistent and inefficient energy transfer/excitation to the
sample. Remember that metabolomics is all about consistency. It is our
recommendation that any study include infrastructure capable of optimizing
the tune/match for every entering NMR data acquisition.

Cryogenically Cooled Probes

Among some of the many recent NMR developments (Kovacs et al. 2005; Kupče
2007; Matsuki et al. 2015; Rovnyak et al. 2004; Webb 2006), one that has been
particularly applicable to metabolomics is cryogenically cooled NMR probes. The
cooling of the electronics has dramatically decreased the noise, and thereby
improved the signal-to-noise ratio. Interestingly the same cooling technology
utilized for the cryogenically cooled probes has also been leveraged into “cryogen-
free” magnets, i.e., self-helium re-liquifying included inside the magnet, but unfor-
tunately the electrical and annual maintenance costs of the cooling technologies
(depending on size, manufacturer, and type of system) can make them financially
impractical. The systems are variations on a compressor and/or gas expansion, each
having advantages/disadvantages including vibration introduction into the spectrom-
eter, and these vibrations can often be detrimental to the spectrometer performance.
For metabolomics the use of helium cryogenically cooled NMR probes (Webb
2006) and the associated increase in signal-to-noise (Schönberger et al. 2014) is
often worth the complexity (Shishmarev and Otting 2011) and additional costly

11
For those wishing a more detailed and mathematical description, see Chap. 23 especially Sect.
5 in (Brown 2016).
92 R. T. McKay

upkeep as the signal-to-noise ratio increases threefold–fourfold with a concomitant

reduction in experimental time of 9–16 times. This is a substantial time savings and
has been a major advance (Kovacs et al. 2005).

Note
Cryogenically cooled probes are often a logical and cost-effective upgrade to
existing systems. However, probes and the application to magnets along with
annual maintenance costs and supporting infrastructure requirements are not
trivial so beware.

5.4.2 Console
The NMR console is the heart of the NMR spectrometer containing frequency
generation, timing control, power supply, input/output controls, amplifiers, band
selective equipment, relays, pneumatic controllers, thermal regulation, pulse field
gradient generation, shim controls, etc. all the components necessary to perform the
desired NMR experiments. The console, just like the NMR probe, requires constant
monitoring, calibration, and inevitably some repair and/or replacement of parts.
These must be done with full understanding of all consequences regarding instru-
ment performance for metabolomics studies.

Note
We strongly recommend a professional staff person be utilized to initially
calibrate and monitor the equipment at regular intervals to ensure consistency
in the console performance.

5.4.3 Host Computer

The last supporting piece of equipment is the computer used by the operator to
control and interact with the spectrometer. We will consider the communication
hardware between the computer, console, and probe/magnet/supporting-
infrastructure to be included. Like the previously mentioned equipment, the com-
puter and software running the operating system and console are not always static.
Anything from software updates incorporated for security and/or feature inclusion to
spectrometer controlling software/firmware can go through changes. Any of these
changes must also be taken into consideration by the metabolomics user as the
spectrometer performance and/or saved NMR data may not be consistent and
therefore introduce problems during processing and analysis.
Metabolomics and NMR 93

5.4.4 Experiment Pulse Sequence

The order, number, repetition, carrier position, duration, magnitude (i.e., power),
dynamics, and phase of any induced electromagnetic fields (called pulses), and any
incorporated delays before/between/after these pulses along with the final read
period to collect the observable data is collectively called the “pulse sequence” or
“pulse program” (see example in Appendix A). This is complex and incredibly
important regarding the consistency of the experiment used for data acquisition.
Interested readers should see reference (Hore et al. 2015) and Chap. 13 of reference
(Zerbe and Jurt 2013) for an excellent starting description. Then additionally
references (Freeman 1997; Cavanagh et al. 2006) for further details. It is important
to note that any change of the aforementioned settings, termed parameters (see this
excellent review (Reynolds and Enriquez 2002) and our section below), can dramat-
ically alter the experimental performance of the instrument (McKay 2011; Potts et al.
2001). This was also addressed by Saude et al. (2006) in 2006 when they carefully
examined which pulse sequence to use for metabolomics. Types of pulse sequences
for metabolomics were also included in the review by Beckonert et al. (2007) and
more directly by others as interest in multi-dimensional and multi-nuclear
experiments were explored (Van et al. 2003; Potts et al. 2001).
Pulse sequences can go through versions, and this can be a problem when
software is updated as small changes can go unnoticed. Especially while the study
data is being collected. It can be necessary to freeze a system in place and prevent
software updates to ensure data collection integrity. For example, the most common
metabolomics 1D-1H NMR pulse sequence is the first increment of a
2D-1H,1H-NOESY (Bain et al. 1994; Blake and Summers 1990; Kumar et al.
1980), i.e., a one-dimensional data collection with no indirectly detected dimension.
During one of our studies, the pulse sequence phase cycle (McClung 1999; Kingsley
1995; Kay 1995) (see also Sect. 5.4.6 below) was slightly altered, which dramati-
cally changed the water solvent suppression (McKay 2009) and resulted in analog-
to-digital overload errors, but only after the collection of the 16th free induction
decay. Nowhere else in the data collection experienced a problem, and if one
collected fewer than 16 scans (1/2 the phase cycle), there was no error and the
data looked normal. This took quite a bit of time to discover and test the source of the
issue. Eventually we had to retrieve an older pulse sequence version and rename it to
prevent modification. This older sequence is now commonly known as the Chenomx
“metabolomics-1D” (Lacy et al. 2014).

Note
Cursory comparisons of pulse sequence basics between versions or vendors
will often miss internal phase cycle details. Even experienced NMR users can
be caught unaware of the complexities involved.
94 R. T. McKay

5.4.5 Parameters
There are dozens to hundreds of parameters involved various pulse sequences. For
example, even the simplest 1D “Pulse-Read” experiment (see Appendix A) requires
the carrier position, power level, and pulse duration to be properly set. In addition,
the length of the acquisition period and dwell time (i.e., time between sample-
induced voltage readings in the receiver coil) will need to be properly selected so
that required information is properly represented. Even this is a huge oversimplifi-
cation. The more complex the manipulation of spins during the sequence the more
parameters are involved, and the incorporation of multiple nuclei and/or multiple
dimensions of indirectly detected dimensions will amplify the number of parameters
exponentially.
We will focus on the most concise list of NMR parameters that need to be
addressed for each sample involved in metabolomics NMR experiments. These
are also necessary for testing and reproduction and should be included for publica-
tion of manuscript data.
We will first briefly address the practical outcome of parameters.

Error Bars and Statistical Relevance

One of the first metabolomics questions asked about NMR data is regarding error
bars. This leads to a second question; how can a single NMR experiment be
considered statistically relevant? For instrument error, we direct interested readers
to reference (Sokolenko et al. 2013), where we attempted to address this question. In
summary, the instrumental error is many orders of magnitude smaller than the error
introduced during the assignment/integration/analysis phase depending on the oper-
ator experience and interest.
The second question is far more difficult. A single NMR experiment consists of
tens or hundreds of thousands (or millions, e.g., solids NMR) of individual voltage
measurements over several milliseconds or seconds of acquisition called the free
induction decay or FID. These individual voltage observations are then repeated for
each subsequent scan of the sample and added together. Example, a 500 MHz
spectrometer with a 6,000 Hz sweep width, 4 s acquisition time, and 1 s recovery
delay with presaturation of the solvent peak will typically have something on the
order of 24 k complex acquire points (i.e., 24 k “real” and simultaneously 24 k
“imaginary” collected at 90° offset from the first set making it even more complex)
for quadrature detection. We will ignore the digital oversampling architecture of
modern NMR consoles. Couple this with recording the experiment over-and-over on
the same sample (~128 times) and then adding all those individual measurements
together makes a statistical analysis surprisingly difficult. This does not even begin
to touch on the self-artifact cancelling nature of phase cycling (see below) involved
in the repetitive acquisition and how it relates to metabolomics (McKay 2009, 2011).
Suffice to say that even a single NMR acquisition is statistically relevant and
experimental error is extremely low (Sokolenko et al. 2013).
Metabolomics and NMR 95

5.4.6 Phase Cycle

Though we have mentioned phase cycling before, it is important enough to further
detail. While overall two NMR pulse sequences may look identical, the applied
orientation (i.e., angle for each of the induced pulse or pulses and the receiver) can be
individually controlled. This orientation of each component is termed the
“phase”(Odedra and Wimperis 2012; McClung 1999; Kingsley 1995; Kay 1995;
Cavanagh et al. 2006), and the total list of changes to any angle that each pulse/
receiver is used for each FID over the course of the entire experiments is called the
“phase cycle.” This can also change with each repetition of the same experiment.
Phase cycles are easily displayed in a table for each pulse/receiver on one axis with
the scan number on the other axis. The phase cycle can be quite complex, e.g.,
32 step phase cycle meaning that the experiment must be repeated 32 times to
complete all phase changes. Experiments can usually be performed with far fewer
(e.g., 4 acquired transients or scans) repeats of the data acquisition than the entire
cycle requires, however there will be compromises and may result in instability in
the resulting data. The larger phase cycle usually has more complicated and effective
artifact suppression, e.g., collecting four scans of the sample and trying to compare
to 32 scans of another sample with immediately have obvious S/N differences. There
may also be differences due to partial phase cycle completion (McKay 2011).

Note
The phase cycle is used to cancel any general imperfections in the spectrome-
ter receiver path, i.e., cyclops basic phase cycle. The higher orders of the phase
cycle will usually attempt to remove smaller and smaller artifacts so one can
usually use a subset or portion of the entire cycle. Consistency is again key,
and one must determine the best selection prior to study start.

5.4.7 Temperature Calibration

Users often assume that the spectrometer displayed temperature is both accurate and
precise for the sample once set via the spectrometer software. However, it has been
our experience that this is almost never the case. Some instruments are certainly
close, but there is always a benefit and often a substantial need for calibration (for
examples, see Bernard et al. 2017; Raiford et al. 1979 and references therein). In our
experience over the decades of instruments tested, once calibrated we see deviation
of the set point to the actual temperature ranging from 1.6 to a few tenths of a degree
Celsius in regulated temperature. This comparison is between what is measured via a
calibrated sample and the software display. As metabolomics requires absolute
consistency to identify small changes in sample composition and then equate those
changes back to usable phenotypes, any variation in temperature will immediately
create a problem.
96 R. T. McKay

Note
We are concerned with comparing data from instrument to instrument, but
even if a single unchanging/dedicated instrument is used for the entire study,
one would only have precision (i.e., consistent error). This may not be
reproducible later even with the same spectrometer so relying on precision
may not be sufficient. We highly recommend directly calibrating each instru-
ment prior to study start, and after any hardware changes using referenced
techniques (Holz et al. 2000; Karschin et al. 2022; Raiford et al. 1979).

Regarding sample temperature, Saude et al. (2006, 2007; Saude and Sykes 2007)
published detailed tests of human urine stability versus storage time and temperature
and we highly recommend interested readers consider this type of preparation for
their samples to establish utmost consistency.

Note
To speed up data acquisition, we used a two-stage sample cooler. The first
stage was long-term storage at ~5°C. The second stage was (using a golf
analogy) termed “on-deck” and was a slightly higher temperature than the
equilibrium temperature. See below for details.

Robotic Sample Handling

As the robot acquired the first sample, the otherwise idle robot would move the
second sample from cold storage (e.g., 5°C) to the warmer on-deck region. The
on-deck sample was set to slightly warmer than the NMR spectrometer (e.g., 30.4°C
if samples were to be run at 30°C) so that as the sample was moved from the waiting
on-deck position to the spectrometer, the slight cooling caused by transporting
through the room air would be offset. The sample would arrive in the spectrometer
receiver coil at the perfect temperature ready for equilibrium saving several minutes
per sample. This could save hours of spectrometer data time per batch. The slightly
warmer on-deck temperature was empirically calibrated based on average travel
times and current room temperature so that the arriving sample temperature inside
the spectrometer would be as close to ideal as possible. The on-deck temperature
would vary throughout the year/seasons as the room temperature also changed.

5.4.8 Tune/Match
Frequency “tuning” and impedance “matching” of the sample and the spectrometer
transmit/receive electronic circuit is critical for optimization. This ensures that the
inductive moment or energy transfer going to, and the subsequent signal coming
from the sample is as efficient as possible.
The concepts involved in impedance involve the resistors, frequencies, inductors,
capacitance, mutual or self-inductance, etc. and goes far beyond this chapter (e.g.,
Metabolomics and NMR 97

see Chaps. 32 and 33 dealing with these topics (Halliday et al. 2013)) and the authors
capabilities. However in terms of practical NMR tuning and matching are still vitally
important, and it is usually sufficient to use the common analogy that impedance
matching involves adjusting the spectrometer circuitry to “match” that of the sample
and the sample’s dielectric constant of the solvent. Each solvent/sample can have
different impendence characteristics (depending on temperature, salts, solvent(s),
type of glass, ions in solution, etc.), so the probe must have an electronic range of
capabilities for the nuclei to be observed and appropriate for the sample itself. Probes
are usually designed with either manually manipulated, or automated movement of
physical rods connecting deep inside the probe head. These rods move specific
electronic components changing the transmitter and receiver’s electronic circuit
configuration.12 A maximum transfer of power occurs when the two circuits (i.e.,
probe and sample) are identical in resistance, or in this case, due to alternating
current, their impedance (Wilson 2007).

Note
With a poor “matching” condition, there can be substantial energy produced
by the NMR console that does not enter the sample, but instead reflects,
returning up the pathway and potentially damaging equipment.

This situation is commonly recognized by observing abnormally long NMR pulse

widths that are inefficient along with poor S/N.
Interested readers are directed to excellent article from 1978 by Prof. David Hoult
(1978) covering many aspects including the electronics of the probe and console
design/function.13 Other articles focus more on the practical implications and
applications of matching (Bendet-Taicher et al. 2014; Nausner et al. 2010; Torchia
2009). An electronic-based description of impedance matching can be found in the
ARRL handbook, e.g., Wilson 2007 or physics textbooks (Halliday et al. 2013;
Feynman et al. 1965).

5.4.9 Excitation Pulse

Readers looking for further background are reminded of the previously reference
books (Hoult and Richards 1976; Hore 2015; Silverstein et al. 2014; Keeler 2010;
Levitt 2001; Freeman 1997; Harris 1986) and articles describing NMR and NMR
theory. We will assume bulk coherence and transverse magnetization are
understood.

12
Capacitors and inductors (trying not to get confused with inductance nor impedance) of different
sizes, types, and changeable physical positions are used to control analog frequency band widths
and the subsequent matching of the circuitry.
13
I had the pleasure of attending one of Dr. Hoult’s talks as he was an invited speaker for the Alberta
Cancer Foundation at the University of Alberta in 2015. His talk was amazing and the first time I
really started to understand the difference between true NMR magnetic inductance of signal
transmission versus the standard NMR analogy of radio signals and antennas.
98 R. T. McKay

For the simplest 1D-1H direct detection NMR pulse sequence experiment (aka
pulse-read) formed from a recovery delay, excitation pulse, and observation/data
acquisition period, the only pulse power/duration consideration is the excitation
pulse. As the reader may be aware, especially from the referenced literature, most
spectroscopists use the maximum power (or close to it) that the probe can withstand.
Therefore only the duration of the pulse application can be altered safely by the user.
Consistent determination and application of the optimal excitation (i.e., movement
of the bulk magnetization vector fully into the XY plane for detection by the receiver
NMR coil in the probe) are critical for multi-pulse experiments, but the story can be
more complex for the simplest pulse-read experiment. Examples of optimization are
provided in these references (Bodenhausen et al. 1984; Breton and Reynolds 2013;
Burrow et al. 2014; Reynolds and Enriquez 2002; Reynolds and Burns 2012). While
some articles (Schönberger et al. 2014) recommend using only a 90° excitation
pulse, there are complications due to relaxation. The “Ernst Angle”(Cavanagh et al.
2006; Freeman 1997; Keeler 2010; Levitt 2001) can be applied for simple
experiments or more complex ones (Ogg et al. 1994; Zhang et al. 2000). The Ernst
angle was originally developed to achieve the greatest S/N per unit time available. In
our facility, the primary purpose of using less than a 90° pulse is to improve the
accuracy of integration over all regions of the molecule. Different regions/atoms/
functional-groups may have different relaxation rates, and therefore additional scans
(to improve the S/N) will result in different signal intensities/integrations based not
just on the number of nuclei, but also their relative atomic and molecular mobility.
Using a smaller pulse angle (e.g., 45°) reduces the time needed to fully relax, thus
restoring the integrity of integration but with an S/N cost. For multi-pulse, e.g.,
multi-dimensional/multi-nuclear experiments involving subsequent magnetization
precession and evolution with time, the optimal full 90° pulse is critical. Otherwise,
there is a cumulative and compounding loss for each “imperfect” pulse.

Note
For metabolomics using the first increment of the 2D-1H,1H-NOESY experi-
ment (aka metnoesy (McKay 2011)) which is standard for the Chenomx
metabolomics software and database, it is essential that the calibrated 90°
pulse is used consistently for all samples. Unless the solvent is highly consis-
tent, unlikely with human urine/metabolomics samples, the 90° pulse needs to
be determined for each sample prior to data acquisition.

5.4.10 Gain
Gain is essentially the volume control on the signal receiver. While we do not have
time for the fundamental radio/receiver electronic aspects, suffice to say that too low
a gain negatively impacts the experiment with poor S/N, while setting the gain too
high risks overloading the receiver coil and damaging the FID data. Damaged data is
not something that can be corrected later, and the experiment is forfeited. Overloads
Metabolomics and NMR 99

occur when the voltage reading from the transverse recorded magnetization exceeds
the electronic capability, and this is most often caused by a disproportionately strong
solvent signal (e.g., water) or sample component (e.g., urea). Careful and consistent
setting must be used to insure proper acquisition over a long-term metabolomics
study.

Note
While it is not relevant to report the instrumental value of gain in the manu-
script, it is important to note for the manuscript reviewer and readers how an
appropriate gain was determined and if consistently used for all samples
ensuring that receiver and/or analog-to-digital converter (ADC) errors were
avoided.

5.4.11 Pulse Field Gradients

Pulse field gradients (PFGs) are an extraordinarily useful tool in NMR spectroscopy
(Sakhaii et al. 2013; van Zijl and Hurd 2011; Zangger et al. 2001; Kay 1995; Keeler
et al. 1994). The easiest analogy for PFGs is that we are slicing the length of the
sample horizontally (think about slicing a sausage into thin circular cross-sections).
Each cross-section experiences a unique magnetic field depending on the vertical
position (i.e., isocromats). This allows spectroscopists to manipulate spins based on
their relaxation, diffusion, or chemical shifts independently depending on where
they started in the vertical sample.
While incredibly useful and often considered for solvent suppression in
metabolomics, a common error occurs when excitation pulses and/or pulsed field
gradient “eddy currents” have not subsided prior to data acquisition. If the “ringing”
from the last pulse or gradient remains, the recording of the early data points will be
distorted essentially by extra signal that does not originate from the sample, but
instead the hardware. This is seen as baseline distortion and overall noise in the
spectrum depending on how much hardware distortion leaks into the real data.14
We have found that PFGs disrupt the deuterium resonance peak used for the
spectrometer deuterium “lock” (i.e., the automatic magnetic field compensation/
stabilization). The instrument attempts to try and follow the deuterium signal during
the initial application for PFGs, and then afterwards to re-gain the lock resonance
causing field distortions. For many peaks, these distortions are small enough to
ignore, but for the strongest/sharpest of the NMR peaks, the distortion appears as a
dispersive NMR component. Unfortunately for DSS (Sheedy et al. 2010; Harris et al.
2008; Markley et al. 1998) commonly used as an internal reference standard and a
peak shape normalization factor for the Chenomx software (Weljie et al. 2006), it

14
A possible solution is to apply backwards linear prediction to the first few points, thus replacing
the damaged data with realistic predictions based on the later non-distorted information. Not ideal
but sometimes useful.
100 R. T. McKay

sees a large distortion on peak shape. Chenomx software uses the shape of the DSS
methyl group to back-calculate corrections for any small shim imperfections. A
distortion of the DSS methyl reference peak, however, can dramatically affect
resonance integrations and the subsequent metabolite identification stage. There is
no way to correct for PFG dispersive changes in the reference signal once collected
(i.e., you cannot phase the distortion out without negatively impacting other
resonances).

Note
For metabolomics, it is our experience that PFGs cause far more problems than
they solve, and we highly recommend not utilizing pulse sequences
with PFGs.

The easiest and therefore most common method is to remove all PFGs from the
metabolomics NMR pulse sequence. This has been the case for nearly two decades
now. While the use of alternating directional gradients can be used to reduce the
artifacts (Nguyen et al. 2007; Sokolenko et al. 2013), validation of NMR
metabolomics databases has not been done with any of the proposed pulse sequence
changes.

5.5 Manuscript Reporting Parameters

To our knowledge, there is no standardized (e.g., IUPAC) recommendation for

appropriately reporting NMR experimental parameters. Far too often when
reviewing manuscripts, we find limited or nonexistent NMR experiment parameters
(see Appendix B). An analogy would be stating you went on a trip in a “truck” and
only specified the destination thus lacking all crucial information, e.g., what route
was taken, how long were you gone, what type of truck, did you stop for fuel, if so
how often, what speed, etc. to reproduce and confirm the journey. Stating in an
experimental section, “Data was collected on a 500 MHz NMR spectrometer from
manufacturer X” provides almost no usable information.
The pulse sequence/program used is essential and type/manufacturer of probe/
spectrometer is essential for the experimental section. Temperature, sweep/dwell
time, etc. are all needed. The reporting should be considered as if a new group
member would reproduce the exact experiment after reviewing the experimental
write up. We detail certain key aspects below.

5.5.1 Pulse Width

In terms of reporting, it is easy to report the 90° pulse duration (usually in
microseconds) however depending on the application (e.g., shaped pulses (Freeman
1998; Kupce and Freeman 1995; Morris and Freeman 1978; Prost et al. 2002))
Metabolomics and NMR 101

Note
Without the 90° excitation pulse duration (often termed pulse width) or that a
90° pulse was used and stating the induced field strength(s) for the various
pulses, we cannot reproduce the experiment. No reviewer should allow a paper
into publication that does not have at least that minimum information.

5.5.2 Delays: Relaxation, Equilibrium

Relaxation, i.e., the return of signal to equilibrium between scans effects
integrations. This makes molecule identification difficult and quantitation inaccu-
rate. For metabolomics it will result in misassignment of molecules and/or missing
the biomarker entirely.
Metabolomics relaxation has been studied, e.g., see Saude et al. 2006,
Bakhmutov 2005 and references therein. Different pulse sequences will likely
have different total relaxation times between scans. It is also important to understand
that under these circumstances a single acquisition (i.e., 1 scan) will not necessarily
give the same result when compared to the entire experiment (e.g., 32 scans) in terms
of integration. The first point cannot be solely trusted to represent the final data that
will be collected as the first scan usually has many seconds or minutes to establish
spin equilibrium in the magnetic field.

Note
The key point here is consistency of relaxation times within the study. The
number of scans, and other experimental parameters must be maintained for all
the samples in the study.

The pulse sequence cannot be swapped out nor altered halfway through the study
(e.g., choice of water suppression), as any change will affect the relaxation, water
suppression, degree of excitation, etc. which will certainly be detected later in the
processing/analysis phase.
Of equal importance is that one must conform to the pulse sequence delays
required by the database(s) being used to determine sample content (see Sect.
5.4.6). For example, Chenomx Inc. (Mercier et al. 2011; Tredwell et al. 2011;
Weljie et al. 2006) relies on the user running a 10 ms recovery time, with 990 ms
solvent saturation at a controlled level (i.e., ~80 Hz gammaB1 effective induced field
strength see previous Sect. 5.4.9) precisely on resonance, followed by the excitation
pulses, a 100 ms “mix” where saturation is turned back on, and finally the last 90°
pulse and 4 s of acquisition. Total experiment time for a single scan is 5 s. Deviation
from this will move the user away from accurate/useful integrations when comparing
to the Chenomx database (Lacy et al. 2014; Mercier et al. 2011; Tredwell et al. 2011;
Weljie et al. 2006).
102 R. T. McKay

5.5.3 Post-Acquisition: Weighting Function

The specific weighting function used should not be an issue, with the caveat that it is
uniformly applied to all acquired data sets. We will recommend the selection of a
weighting function matched to the observed decay of the acquired signal, and in our
experience, a 0.5 Hz line broadening function works reasonably well, especially with
the typical 1D-NOESY (metnoesy) NMR pulse sequence between 400 and 800 MHz
(1H). However, the selection of an appropriate apodization function will be left to the
experts and undoubtedly some users will have strong reason for their selection.

Note
We have found that a great deal of problems can be avoided by using line
broadening, e.g., 0.25–0.5 Hz.

Small shimming errors can be easily averaged or approximated out by the slight
change. This must be included in the reported parameters for comparison and
evaluation.

5.5.4 Corrections: Linear Prediction, Baseline, Phasing

We have found linear prediction to be of little value in 1D spectra, unless under
specific hardware issues/circumstances usually outside metabolomics (e.g., long ring
down times on cold-probes, etc.). We do not recommend, however, if used it must be
explicitly stated in the manuscript.
Baseline correction can again be useful however as always in metabolomics,
consistency is the key. We have found a general drift correction on a well-phased
spectra to be relatively harmless. We typically avoid other corrections or try to
correct the source problem either in the pulse sequence or spectrometer setup. For
example, the pre-acquisition delays between the last excitation pulse and the opening
of the receiver gate electronics.15 Again any method must be included in the
manuscript.
Phase correction (i.e., phasing) is commonly required, especially in 1D-NMR.
There is a temptation to use automated phase corrections, but auto-phasing can be
unreliable. It can also over emphasize the largest peak in the spectra which is most
likely the solvent, and certainly not the most interesting and important resonance.
Manual correction is more accurate, but introduces user error (precision), especially
for inexperienced users.

15
e.g. alfa and rof2 on older Varian/Agilent spectrometers with ddrtc coming into play on newer.
Metabolomics and NMR 103

Note
Utmost effort should be taken to optimize NMR pulse sequence delays prior to
the start of NMR acquisition. This care along with knowledgeable lock
settings/setup should minimize the necessity of phasing and the introduction
of possible user error.

5.5.5 Referencing
Referencing the zero point on the spectra is incredibly important for metabolomics.
The databases typically use the chemical shift position of a peak to begin all auto-
assignments. Even peak pattern recognition usually starts with the chemical shift of a
unique resonance and tries to build out from there the possible coupling profiles. If
the sample is referenced incorrectly the software will have infinitely more difficulty
assigning peaks, let alone correctly.
As mentioned, the Chenomx software has several internal reference standards
available checking the reference peak line shape for application of corrections to the
rest of the spectra.

Note
The known amount of reference intensity also provides quantitative informa-
tion for the integrations.

6 Future Ideas

6.1 Solids NMR Metabolomics

While metabolomics has been predominantly only liquids likely due to a focus on
easily acquired samples (e.g., blood, urine, sputum, etc.), there is a growing amount
of research into solids NMR utilizing high-resolution magic angle spinning
(HRMAS) for metabolomics (Cheng 2007) and specifically Chap. 4 (Tilgner et al.
2019). We direct interested readers to the many excellent reviews of solid-state
NMR (Laws et al. 2002; Ashbrook and Sneddon 2014; Reif et al. 2021) and
references therein, solid-state cold-probes (Matsuki et al. 2015), and the relatively
new field of dynamic nuclear polarization (DNP) enhancement (Albert et al. 2017;
van Bentum et al. 2016; Matsuki et al. 2015). This is an exciting new area, and we
look forward to developments.
104 R. T. McKay

6.2 Working with Raw FID

Dr. Krishnamurthy16 has suggested a different route with Complete Reduction to

Amplitude Frequency Table (CRAFT) NMR (Krishnamurthy 2013). As the author
details, complexes and/or mixtures are very common and pose a difficulty for
traditional NMR analysis. An example would be tailings “ponds” water used to
deposit the remaining materials after oil extraction. We have seen spectra contain
tens of thousands of compounds and those are only the ones in the lower molecular
weight categories distinguishing themselves as sharp resonances.
Having a method that could identify and quantitate even complex mixtures would
be amazingly beneficial and that is what Dr. Krishnamurthy has proposed in a 2013
publication. While risking oversimplifying, CRAFT does not use Fourier transfor-
mation to convert the time domain data into the commonly recognized NMR spectra.
Instead it identifies the frequencies and amplitude of each of the raw components and
reports them in a simple table ready for statistical analysis. Specifically they examine
a fermentation broth for quantitative analysis and a spiked human blood plasma. The
benefits of removing FT and operator error are certainly enticing. The method has
yet to become prevalent with the majority of metabolomics analysis, likely due to
users not being familiar with the technique nor confirmed when compared to
traditional approaches.

6.3 PureShift 1H-1H J-Coupling Removal

One of the biggest potential new developments involves the PureShift style/family of
NMR pulse sequences (Kiraly et al. 2018, 2021; Dumez 2018; Foroozandeh et al.
2018; Moutzouri et al. 2017; Castañar 2017; Mishra and Suryaprakash 2017; Kew
et al. 2017; Zangger 2015; Aguilar et al. 2015; Foroozandeh et al. 2014, 2015;
Mauhart et al. 2015; Reinsperger and Luy 2014; Kaltschnee et al. 2014; Paudel et al.
2013; Meyer and Zangger 2013; Aguilar et al. 2012; Aguilar et al. 2010; Zangger
et al. 2001). There are many variations, iterations, and subsequent improvements,
but all focus on removing the dipolar “coupling” or crosstalk between neighboring
hydrogen atoms (as previously mentioned). PureShift sequences attempt to remove
(or at least minimize) the couplings and complexity of the spectra. The variations
attempt to compensate for the inherent weaknesses of the pulse sequence, i.e., signal-
to-noise, and distortions in tightly coupled systems.
While these couplings are often crucial for organic chemistry and basic molecular
identification in traditional NMR usage, for metabolomics they create greater spec-
tral complexity with a multitude of overlapping congested spectral information.
Chenomx software checks each peak position, relative integration, and coupling
pattern when it attempts identification, and these have been validated for each

16
“Krish” to friends and colleagues is an amazing resource for the NMR community and we
gratefully acknowledge years of benefiting from this person’s contributions.
Metabolomics and NMR 105

molecule in question, in isolation, for each common spectrometer frequency.

PureShift sequences could dramatically reduce the complexity and therefore the
overlap of the NMR spectra. However, to use the Chenomx software with PureShift
NMR, the molecule database would have to be entirely re-acquired. This would take
a great deal of invested money and time.
The pulse sequences also require PFGs making the implementation more difficult
for inexperienced users. Potentially more variability as well as upkeep and mainte-
nance would be increased (see previous Sect. 5.4.11).
Lastly there is a large S/N reduction reducing the confidence in the data gathered
and potentially missing weak signals. It is possible to compensate, but studies would
be forced to substantially increase the number of scans (time) acquired for each
sample. This would also re-introduce problems with sample storage/handling men-
tioned earlier as each sample would take more time on the instrument and robotic
sample handling system.

7 Conclusion

The most important point we can make is that there are many aspects requiring
attention prior to initiating a metabolomics study via NMR. The information and
experiences detailed above will hopefully help the reader avoid some of the common
pitfalls that we encountered. We also hope that we have convinced new users that
every single metabolomics study can benefit from having experienced knowledge-
able spectroscopists involved, especially in the earliest planning stages. The chances
of acquiring reliable/reproducible data increase exponentially when including the
NMR operators. Unfortunately, these problems are often unrealized until the statis-
tical analysis stage, when the errors due to confounders can exceed the actual data.
The entire study may thus be ruined.
Consistency in every aspect is paramount. From sample acquisition, storage,
preparation, handling, data acquisition, storage and/or retesting, to processing and
analysis. There are many steps that can go astray, making the data more difficult to
assess. However, organized from the start, carefully monitored throughout, and
diligently systematically checking while acquiring the data will give the research
group their best chance of having a useful data set at the end. Spectrometer
consistency is paramount and mentioned above extensively, therefore users may
want to explore the possibility of establishing quality control checks are regular
intervals in their study.
Finally, we hope that the excellent referenced publications and books referenced
will be as useful to your study as they have been to our education, though we are sure
there are many others we have not yet discovered. Best of luck.

Acknowledgments The author would like to thank Professor Vladimir Michaelis (Department of
Chemistry, Univ. of Alberta) for critical reading and suggestions on the writing of this document.
106 R. T. McKay

Appendix A: Example Pulse Sequence: Metabolomics 1D-1H

“Metnoesy”

12.5us 12.5us
12.5us 100.0ms

3.999sec

990.0ms

Tx 10.0ms 0.0
H1 d1 satdly pw pw hst mix-hst
pw

A B C D

Appendix B: Example Experimental Section

Experimental Section

NMR Samples
NMR samples were dissolved in 90% D2O (~600 μL in volume) and sample data
acquired in 5 mm XXXmodel NMR tubes purchased from XXXmanufacturer. All
NMR solvents were purchased from XXX. NMR tubes were washed using three
rinses of 95% ethanol with a single final D2O rinse, and then inverted to air dry
overnight.

NMR Spectroscopy
NMR experiments were collected on either a 14.1T (600 MHz) Varian/Agilent
VNMRS with an Agilent 7,620 automatic sample handling system, or a 9.39T
Metabolomics and NMR 107

(400 MHz) Varian Inova NMR spectrometer, both at 27°C (calibrated using metha-
nol17). On the 600 MHz spectrometer, an HCN BioNMR probe (i.e., direct detect 1H
inner coil) was used, while on the 400 MHz instrument an AutoDB broadband (i.e.,
broadband inner coil) 5 mm probe was available. All spectra were run “locked” on
the 2H resonance signal and chemical shifts were referenced using the residual
proton 1HOD signal position18 (i.e., 4.7 ppm) prior to saturation. One dimensional
1
H data was acquired using either presaturation19 followed by a single excitation
pulse/acquire spectrometer sequence, or the first dimension of a -2D-1H,1H-NOESY
(metnoesy). The metnoesy (see Appendix A) uses a recovery delay of 10 ms, then
990 ms of presaturation followed by two 90° pulses, a mixing time of 100 ms with
saturation, a final 90° pulse and lastly a 4 s acquisition period. The saturation pulse
and carrier position were manually optimized and placed on the water resonance.
Saturation was applied with a gammaB1 induced field strength of 100 Hz (600) or
30 Hz (400) depending on water suppression efficacy and to avoid receiver
overloads. For the simple 1D-1H experiments, the duration of the saturation pulse
was 2 s on both instruments. Parameter settings for all experiments were: sweep
width of 7,183 Hz, acquisition time 2 s, with 28,736 real plus imaginary acquired
(600), or a sweep width of 4,801 Hz. For the 1D-1H, an acquisition time of 3 s with
28,812 real and imaginary points for the 400 MHz NMR was utilized, and an ~30°
(i.e., 3.4 μs) excitation pulse angle (applied at ~24 kHz gammaB1) was used
following the concept of the improved integration and reduced relaxation times
(commonly known as the “Ernst Angle”),20 however on some of the initial highly
concentrated samples an extremely short pulse length was needed to avoid receiver
overflows (e.g., 1us excitation pulse).
For processing of all NMR data, the acquired points were zero-filled to twice the
number of acquired points, and a line-broadening apodization function of 0.5 Hz was

17
D. S. Raiford, C. L. Fisk, E. D. Becker, Anal. Chem. 51, 2050 (2002).
18
Wishart, D., Bigam, C., Yao, J., Abildgaard, F., Dyson, H. J., Oldfield, E., Markley, J., and Sykes,
B. (1995) 1H, 13C and 15N chemical shift referencing in biomolecular NMR, J Biomol NMR 6,
135–140.
Trainor, K., Palumbo, J. A., MacKenzie, D. W. S., and Meiering, E. M. (2020) Temperature
dependence of NMR chemical shifts: Tracking and statistical analysis, Protein Science 29,
306–314.
19
Hoult, D. I. (1976). Solvent Peak Saturation with SIngle Phase and Quadrature Fourier Transfor-
mation. Journal of Magnetic Resonance, 21, 337–347.
Campbell, I. D., Dobson, C. M., Jeminet, G., & Williams, R. J. P. (1974). Pulsed NMR methods
for the observation and assignment of exchangeable hydrogens: Application to bacitracin. FEBS
Letters, 49(1), 115–119.
20
Lauridsen, M., Maher, A. D., Keun, H., Lindon, J. C., Nicholson, J. K., Nyberg, N. T. et al.
(2008). Application of the FLIPSY pulse sequence for increased sensitivity in 1H NMR-based
metabolic profiling studies. Anal Chem, 80(9), 3365-3371.
Waugh, J. S. (1970). Sensitivity in Fourier transform NMR spectroscopy of slowly relaxing
systems. Journal of Molecular Spectroscopy, 35(2), 298–305.
Ernst, R. R., & Anderson, W. A. (1966). Application of Fourier Transform Spectroscopy to
Magnetic Resonance. Review of Scientific Instruments, 37(1), 93–102.
108 R. T. McKay

then applied. Manual phasing and referencing to the solvent peak were used to
confirm referencing based on the lock solvent (when available) and previously
determined carrier position. Spectra were analyzed using VNMRJ 4.2 patch110
software.

References
Aguilar JA et al (2010) Pure shift 1H NMR: a resolution of the resolution problem. Angew Chem
Int Ed Engl 49(23):3901–3903
Aguilar JA et al (2012) Decoupling two-dimensional NMR spectroscopy in both dimensions: pure
shift NOESY and COSY. Angew Chem Int Ed Engl 51(26):6460–6463
Aguilar JA et al (2015) Minimising research bottlenecks by decluttering NMR spectra. Chemistry
21(17):6623–6630
Albert BJ et al (2017) Instrumentation for cryogenic magic angle spinning dynamic nuclear
polarization using 90L of liquid nitrogen per day. J Magn Reson 283:71–78
Alseekh S et al (2021) Mass spectrometry-based metabolomics: a guide for annotation, quantifica-
tion and best reporting practices. Nat Methods 18(7):747–756
Amoureux JP et al (2008) Rapid analysis of isotopically unmodified amino acids by high-resolution
(14)N-edited (1)H-(13)C correlation NMR spectroscopy. Chem Commun (Camb) 48:6525–
6527
Andersson ER et al (2021) Identifying metabolic alterations associated with coral growth anomalies
using 1H NMR metabolomics. Coral Reefs 40(4):1195–1209
Ashbrook SE, Sneddon S (2014) New methods and applications in solid-state NMR spectroscopy
of quadrupolar nuclei. J Am Chem Soc 136(44):15440–15456
Athersuch TJ et al (2013) Evaluation of 1H NMR metabolic profiling using biofluid mixture design.
Anal Chem 85(14):6674–6681
Bain AD, Burton IW, Reynolds WF (1994) Artifacts in two-dimensional NMR. Prog Nucl Magn
Reson Spectrosc 26:59–89
Bakhmutov VI (2005) Practical nuclear magnetic resonance relaxation for chemists. Wiley, p 216
Balkir P, Kemahlioglu K, Yucel U (2021) Foodomics: a new approach in food quality and safety.
Trends Food Sci Technol 108:49–57
Ban D et al (2017) Recent advances in measuring the kinetics of biomolecules by NMR relaxation
dispersion spectroscopy. Arch Biochem Biophys 628:81–91
Bartel J, Krumsiek J, Theis FJ (2013) Statistical methods for the analysis of high-throughput
metabolomics data. Comput Struct Biotechnol J 4:e201301009
Barton RH et al (2008) High-throughput 1H NMR-based metabolic analysis of human serum and
urine for large-scale epidemiological studies: validation study. Int J Epidemiol 37 Suppl 1:i31–
i40
Beckonert O et al (2007) Metabolic profiling, metabolomic and metabonomic procedures for NMR
spectroscopy of urine, plasma, serum and tissue extracts. Nat Protoc 2(11):2692–2703
Bendet-Taicher E, Müller N, Jerschow A (2014) Dependence of NMR noise line shapes on tuning,
matching, and transmission line properties. Concepts Magn Reson B Magn Reson Eng 44(1):
1–11
Berger S, Braun S (2004) 200 and more NMR experiments: a practical course. Wiley-VCH, p 838
Bernard GM et al (2017) Methylammonium lead chloride: a sensitive sample for an accurate NMR
thermometer. J Magn Reson 283:14–21
Bernini P et al (2011) Standard operating procedures for pre-analytical handling of blood and urine
for metabolomic studies and biobanks. J Biomol NMR 49(3–4):231–243
Bingol K, Brüschweiler R (2014) Multidimensional approaches to NMR-based metabolomics. Anal
Chem 86(1):47–57
Metabolomics and NMR 109

Bingol K et al (2016) Emerging new strategies for successful metabolite identification in

metabolomics. Bioanalysis 8(6):557–573
Blake PR, Summers MF (1990) NOESY-1-1-Echo spectroscopy with eliminated radiation
damping. J Magn Reson 86:622–625
Bodenhausen G, Kogler H, Ernst RR (1984) Selection of coherence-transfer pathways in NMR
pulse experiments. J Magn Reson 58(3):370–388
Boesl U (2017) Time-of-flight mass spectrometry: introduction to the basics. Mass Spectrom Rev
36(1):86–109
Bouatra S et al (2013) The human urine metabolome. PLoS One 8(9):e73076
Braun S et al (1998) 150 and more basic NMR experiments: a practical course. Wiley-VCH, p 596
Breton RC, Reynolds WF (2013) Using NMR to identify and characterize natural products. Nat
Prod Rep 30(4):501–524
Brown KC (2016) Essential mathematics for NMR and MRI Spectroscopists. Royal Society of
Chemistry
Bull HB et al (1964) The pH of urea solutions. Arch Biochem Biophys 104(2):297–304
Burrow TE et al (2014) CRAPT: an improved version of APT with compensation for variations in
JCH. Magn Reson Chem 52(5):195–201
Cassiède M et al (2017) Assessment of 1H NMR-based metabolomics analysis for normalization of
urinary metals against creatinine. Clin Chim Acta 464:37–43
Castañar L (2017) Pure shift 1 H NMR: what is next. Magn Reson Chem 55(1):47–53
Cavanagh J et al (2006) Protein NMR spectroscopy: principles and practice, 2nd edn. Academic
Pres, Toronto, pp 1–885
Cheng LL (2007) Tissue and cell samples by HRMAS NMR. eMagRes
Der Wouden V, Cathelijne H et al (2020) Generating evidence for precision medicine:
considerations made by the ubiquitous pharmacogenomics consortium when designing and
operationalizing the PREPARE study. Pharmacogenet Genomics 30(6):131
Derome AE (2013) Modern NMR techniques for chemistry research. Elsevier, p 299
Duarte IF, Diaz SO, Gil AM (2014) NMR metabolomics of human blood and urine in disease
research. J Pharm Biomed Anal 93:17–26
Dudzik D et al (2018) Quality assurance procedures for mass spectrometry untargeted
metabolomics. A review. J Pharm Biomed Anal 147:149–173
Dumas ME et al (2006) Assessment of analytical reproducibility of 1H NMR spectroscopy based
metabonomics for large-scale epidemiological research: the INTERMAP study. Anal Chem
78(7):2199–2208
Dumez JN (2018) Spatial encoding and spatial selection methods in high-resolution NMR spec-
troscopy. Prog Nucl Magn Reson Spectrosc 109:101–134
Dunn WB et al (2011) Systems level studies of mammalian metabolomes: the roles of mass
spectrometry and nuclear magnetic resonance spectroscopy. Chem Soc Rev 40(1):387–426
Edison AS et al (2020) NMR: unique strengths that enhance modern metabolomics research. Anal
Chem 93(1):478–499
Eghbalnia HR et al (2017) Increasing rigor in NMR-based metabolomics through validated and
open source tools. Curr Opin Biotechnol 43:56–61
Ellinger JJ et al (2013) Databases and Software for NMR-based metabolomics. Curr
Metabolomics 1
Emwas AH et al (2016) Recommendations and standardization of biomarker quantification using
NMR-based metabolomics with particular focus on urinary analysis. J Proteome Res 15(2):
360–373
Emwas AH et al (2018) Recommended strategies for spectral processing and post-processing of 1D
1
H-NMR data of biofluids with a particular focus on urine. Metabolomics 14(3):31
Emwas AH et al (2019) NMR spectroscopy for metabolomics research. Metabolites 9(7)
Emwas AH et al (2020) NMR as a “Gold Standard” method in drug design and discovery.
Molecules 25(20):E4597
110 R. T. McKay

Emwas A-H et al (2021) Pharmacometabolomics: a new horizon in personalized medicine. In:

Metabolomics – methodology and applications in medical sciences and life sciences.
IntechOpen
Esslinger S, Fauhl-Hassek C, Wittkowski R (2015) Authentication of wine by1H-NMR spectros-
copy: opportunities and challenges. In: Ebeler SB et al (eds) ACS symposium series: advances
in wine research, ACS symposium series. American Chemical Society, Washington, pp 85–108
Feynman RP et al (1965) The Feynman lectures on physics; vol I. Am J Physiol 33(9):750–752
Finco G et al (2016) Can urine metabolomics be helpful in differentiating neuropathic and
nociceptive pain? A proof-of-concept study. PLoS One 11(3):e0150476
Foroozandeh M et al (2014) Ultrahigh-resolution total correlation NMR spectroscopy. J Am Chem
Soc 136(34):11867–11869
Foroozandeh M et al (2015) Measuring couplings in crowded NMR spectra: pure shift NMR with
multiplet analysis. Chem Commun 51(84):15410–15413
Foroozandeh M, Morris GA, Nilsson M (2018) PSYCHE pure shift NMR spectroscopy. Chemistry
24(53):13988–14000
Freeman R (1997) Spin choreography: basic steps in high resolution NMR. Oxford University
Press, Oxford, pp 1–391
Freeman R (1998) Shaped radiofrequency pulses in high resolution NMR. Prog Nucl Magn Reson
Spectrosc 32(1):59–106
Gardner KH, Kay LE (1998) The use of 2H, 13C, 15N multidimensional NMR to study the
structure and dynamics of proteins. Annu Rev Biophys Biomol Struct 27:357–406
Gibney MJ et al (2005) Metabolomics in human nutrition: opportunities and challenges. Am J Clin
Nutr 82(3):497–503
Giraudeau P (2020) NMR-based metabolomics and fluxomics: developments and future prospects.
Analyst 145(7):2457–2472
Giraudeau P et al (2014) Reference and normalization methods: essential tools for the intercompar-
ison of NMR spectra. J Pharm Biomed Anal 93:3–16
Giraudeau P, Silvestre V, Akoka S (2015) Optimizing water suppression for quantitative
NMR-based metabolomics: a tutorial review. Metabolomics 11(5):1041–1055
Göbl C et al (2014) NMR approaches for structural analysis of multidomain proteins and complexes
in solution. Prog Nucl Magn Reson Spectrosc 80:26–63
Halabalaki M et al (2014) Recent advances and new strategies in the NMR-based identification of
natural products. Curr Opin Biotechnol 25:1–7
Halliday D, Resnick R, Walker J (2013) Fundamentals of physics. Wiley
Harris RK (1986) Nuclear magnetic resonance spectroscopy
Harris RK et al (2008) Further conventions for NMR shielding and chemical shifts (IUPAC
recommendations 2008). Magn Reson Chem 46(6):582–598
Heather LC et al (2013) A practical guide to metabolomic profiling as a discovery tool for human
heart disease. J Mol Cell Cardiol 55:2–11
Hiroaki H (2013) Recent applications of isotopic labeling for protein NMR in drug discovery.
Expert Opin Drug Discovery 8(5):523–536
Holz M, Heil SR, Sacco A (2000) Temperature-dependent self-diffusion coefficients of water and
six selected molecular liquids for calibration in accurate 1H NMR PFG measurements. Phys
Chem Chem Phys 2(20):4740–4742
Hore PJ (2015) Nuclear magnetic resonance. Oxford University Press, p 112
Hore P, Jones J, Wimperis S (2015) NMR – the toolkit - how pulse sequences work. Oxford
University Press, p 120
Hoult DI (1978) The NMR receiver: a description and analysis of design. Prog Nucl Magn Reson
Spectrosc 12(1):41–77
Hoult DI, Richards RE (1976) The signal-to-noise ratio of the nuclear magnetic resonance experi-
ment. Journal of Magnetic Resonance (1969) 24(1):71–85
Metabolomics and NMR 111

Hyberts SG et al (2007) Ultrahigh-resolution (1)H-(13)C HSQC spectra of metabolite mixtures

using nonlinear sampling and forward maximum entropy reconstruction. J Am Chem Soc
129(16):5108–5116
Izquierdo-Garcia JL et al (2020) Discovery and validation of an NMR-based metabolomic profile in
urine as TB biomarker. Sci Rep 10(1):1–13
Jones S, Carley S, Harrison M (2003) An introduction to power and sample size estimation. Emerg
Med J 20(5):453
Kaltschnee L et al (2014) “Perfecting” pure shift HSQC: full homodecoupling for accurate and
precise determination of heteronuclear couplings. Chem Commun (Camb) 50(99):15702–15705
Karschin N et al (2022) Extension and improvement of the methanol-d4 NMR thermometer
calibration. Magn Reson Chem 60(2):203–209
Kay LE (1995) Field gradient techniques in NMR spectroscopy. Curr Opin Struct Biol 5:674–681
Kay LE (2016) New views of functionally dynamic proteins by solution NMR spectroscopy. J Mol
Biol 428(2 Pt A):323–331
Kay LE, Frydman L (2014) A special “JMR perspectives” issue: foresights in biomolecular
solution-state NMR spectroscopy – from spin gymnastics to structure and dynamics. J Magn
Reson 241:1–2
Keeler J (2010) Understanding NMR spectroscopy. Wiley
Keeler J et al (1994) Pulsed-field gradients: theory and practice. Methods Enzymol 239:145–207
Kelly AE et al (2002) Low-conductivity buffers for high-sensitivity NMR measurements. J Am
Chem Soc 124(40):12013–12019
Kenny LC et al (2010) Robust early pregnancy prediction of later preeclampsia using metabolomic
biomarkers. Hypertension 56(4):741–749
Kew W et al (2017) Advanced solvent signal suppression for the acquisition of 1D and 2D NMR
spectra of Scotch Whisky. Magn Reson Chem 55(9):785–796
Kingsley PB (1995) Product operators, coherence pathways, and phase cycling. Part III: phase
cycling. Concepts Magn Reson 7(3):167–192
Kiraly P, Nilsson M, Morris GA (2018) Semi-real-time acquisition for fast pure shift NMR at
maximum resolution. J Magn Reson
Kiraly P et al (2021) Single-scan selective excitation of individual NMR signals in overlapping
multiplets. Angew Chem 133(2):676–679
Kohl SM et al (2012) State-of-the art data normalization methods improve NMR-based
metabolomic analysis. Metabolomics 8(Suppl 1):146–160
Korfmacher WA (2005) Foundation review: principles and applications of LC-MS in new drug
discovery. Drug Discov Today 10(20):1357–1367
Kovacs H, Moskau D, Spraul M (2005) Cryogenically cooled probes – a leap in NMR technology.
Prog Nucl Magn Reson Spectrosc 46(2–3):131–155
Krishnamurthy K (2013) CRAFT (complete reduction to amplitude frequency table) – robust and
time-efficient Bayesian approach for quantitative mixture analysis by NMR. Magn Reson Chem
51(12):821–829
Kumar A, Ernst RR, Wüthrich K (1980) A two-dimensional nuclear Overhauser enhancement
(2D NOE) experiment for the elucidation of complete proton-proton cross-relaxation networks
in biological macromolecules. Biochem Biophys Res Commun 95(1):1–6
Kupče Ē (2007) NMR with multiple receivers. eMagRes:721–732
Kupce E, Freeman R (1995) Close encounters between soft pulses. J Magn Reson 112A:216–264
Lacy P et al (2014) Signal intensities derived from different NMR probes and parameters contribute
to variations in quantification of metabolites. PLoS One 9(1):e85732
Lasky-Su J et al (2021) A strategy for advancing for population-based scientific discovery using the
metabolome: the establishment of the metabolomics society metabolomic epidemiology task
group. Metabolomics 17(5):1–6
Lauridsen M et al (2007) Human urine as test material in 1H NMR-based metabonomics:
recommendations for sample preparation and storage. Anal Chem 79(3):1181–1186
112 R. T. McKay

Laws DD, Bitter HM, Jerschow A (2002) Solid-state NMR spectroscopic methods in chemistry.
Angew Chem Int Ed Engl 41(17):3096–3129
Lehnert W, Hunkler D (1986) Possibilities of selective screening for inborn errors of metabolism
using high-resolution 1H-FT-NMR spectrometry. Eur J Pediatr 145(4):260–266
Lepre CA (2011) Practical aspects of NMR-based fragment screening. Methods Enzymol 493:219–
239
Lesar CT et al (2011) Report on the analysis of common beverages spiked with gamma-
hydroxybutyric acid (GHB) and gamma-butyrolactone (GBL) using NMR and the PURGE
solvent-suppression technique. Forensic Sci Int 212(1–3):e40–e45
Levitt MH (2001) Spin dynamics: basics of nuclear magnetic resonance
Li K et al (2012) Establishment of a blunt impact-induced brain injury model in rabbits. Chin J
Traumatol 15(2):100–104
Li Y et al (2020) Sub-nanoliter metabolomics via mass spectrometry to characterize volume-limited
samples. Nat Commun 11(1):1–16
Lindon JC, Nicholson JK, Everett JR (1999) NMR spectroscopy of biofluids. In: Annual reports on
NMR spectroscopy: annual reports on NMR spectroscopy volume 38. Elsevier, pp 1–88
Lindon JC, Holmes E, Nicholson JK (2007) Metabonomics in pharmaceutical R&D. FEBS J
274(5):1140–1151
Linge KL, Jarvis KE (2009) Quadrupole ICP-MS: introduction to instrumentation, measurement
techniques and analytical capabilities. Geostand Geoanal Res 33(4):445–467
Link M et al (2014) Food analysis to check quality, safety and authenticity by full-automated
1H-NMR. ll1.workcast.net
Liu G et al (2014) Fast 3D gradient shimming by only 2×2 pixels in XY plane for NMR-solution
samples. J Magn Reson 248:13–18
Lubbe A et al (2013) NMR-based metabolomics analysis. In: Lämmerhofer M, Weckwerth W (eds)
Metabolomics in practice, Metabolomics in practice: successful strategies to generate and
analyze metabolic data. Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim, pp 209–238
Macel M, Van Dam NM, Keurentjes JJ (2010) Metabolomics: the chemistry between ecology and
genetics. Mol Ecol Resour 10(4):583–593
Markley JL et al (1998) Recommendations for the presentation of NMR structures of proteins and
nucleic acids. IUPAC-IUBMB-IUPAB inter-union task group on the standardization of data
bases of protein and nucleic acid structures determined by NMR spectroscopy. J Biomol NMR
12(1):1–23
Markley JL et al (2017) The future of NMR-based metabolomics. Curr Opin Biotechnol 43:34–40
Marshall AG, Hendrickson CL, Jackson GS (1998) Fourier transform ion cyclotron resonance mass
spectrometry: a primer. Mass Spectrom Rev 17(1):1–35
Marshall AG et al (2007) Fourier transform ion cyclotron resonance: state of the art. Eur J Mass
Spectrom 13(1):57–59
Matsuki Y et al (2015) Closed-cycle cold helium magic-angle spinning for sensitivity-enhanced
multi-dimensional solid-state NMR. J Magn Reson 259:76–81
Maudsley AA, Simon HE, Hilal SK (1984) Magnetic field measurement by NMR imaging. J Phys E
Sci Instrum 17(3):216–220
Mauhart J et al (2015) Faster and cleaner real-time pure shift NMR experiments. J Magn Reson 259:
207–215
McClung RED (1999) Coherence transfer pathways and phase cycles: the decoding of a pulse
sequence. Concepts Magn Reson 11(1):1–28
McGrath BM et al (2007) Unlike lithium, anticonvulsants and antidepressants do not alter rat brain
myo-inositol. Neuroreport 18(15):1595–1598
McKay RT (2009) Recent advances in solvent suppression for solution NMR: a practical
reference. In: Webb G (ed) Annual reports on NMR spectroscopy. Elsevier, pp 33–76
McKay RT (2011) How the 1D-NOESY suppresses solvent signal in metabonomics NMR spec-
troscopy: an examination of the pulse sequence components and evolution. Concepts Magn
Reson A 38A(5):197–220
Metabolomics and NMR 113

Meissner A et al (2014) 1H-NMR metabolic profiling of cerebrospinal fluid in patients with

complex regional pain syndrome-related dystonia. Pain 155(1):190–196
Mercier P et al (2011) Towards automatic metabolomic profiling of high-resolution
one-dimensional proton NMR spectra. J Biomol NMR 49(3–4):307–323
Meyer NH, Zangger K (2013) Simplifying proton NMR spectra by instant homonuclear broadband
decoupling. Angew Chem Int Ed Engl 52(28):7143–7146
Mishra SK, Suryaprakash N (2017) Pure shift edited ultra high resolution NMR spectrum with
complete eradication of axial peaks and unwanted couplings. J Magn Reson 279:74–80
Morris GA, Freeman R (1978) Selective excitation in Fourier transform nuclear magnetic reso-
nance. J Magn Reson 29:433–462
Moutzouri P et al (2017) Ultraclean pure shift NMR. Chem Commun (Camb) 53(73):10188–10191
Nausner M et al (2010) Signal enhancement in protein NMR using the spin-noise tuning optimum. J
Biomol NMR 48(3):157–167
Nguyen BD et al (2007) SOGGY: solvent-optimized double gradient spectroscopy for water
suppression. A comparison with some existing techniques. J Magn Reson 184(2):263–274
Nikolaev EN, Kostyukevich YI, Vladimirov GN (2016) Fourier transform ion cyclotron resonance
(FT ICR) mass spectrometry: theory and simulations. Mass Spectrom Rev 35(2):219–258
Niziol J et al (2021) Metabolomic and elemental profiling of human tissue in kidney cancer.
Metabolomics 17(3):1–15
Odedra S, Wimperis S (2012) Use of composite refocusing pulses to form spin echoes. J Magn
Reson 214(1):68–75
Ogg RJ, Kingsley PB, Taylor JS (1994) WET, a T1- and B1-insensitive water-suppression method
for in vivo localized 1H NMR spectroscopy. J Magn Reson B 104(1):1–10
Pardi N et al (2018) mRNA vaccines – a new era in vaccinology. Nat Rev Drug Discov 17(4):261
Parsons HM, Ludwig C, Viant MR (2009) Line-shape analysis of J-resolved NMR spectra:
application to metabolomics and quantification of intensity errors from signal processing and
high signal congestion. Magn Reson Chem 47 Suppl 1:S86–S95
Paudel L et al (2013) Simultaneously enhancing spectral resolution and sensitivity in heteronuclear
correlation NMR spectroscopy. Angew Chem Int Ed Engl 52(44):11616–11619
Perry RH, Graham Cooks R, Noll RJ (2008) Orbitrap mass spectrometry: instrumentation, ion
motion and applications. Mass Spectrom Rev 27(6):661–699
Petrović M et al (2005) Liquid chromatography–tandem mass spectrometry for the analysis of
pharmaceutical residues in environmental samples: a review. J Chromatogr A 1067(1–2):1–14
Picone G, Mengucci C, Capozzi F (2022) The NMR added value to the green foodomics perspec-
tive: advances by machine learning to the holistic view on food and nutrition. Magn
Reson Chem
Pinto J et al (2014) Human plasma stability during handling and storage: impact on NMR
metabolomics. Analyst 139(5):1168–1177
Potts BC et al (2001) NMR of biofluids and pattern recognition: assessing the impact of NMR
parameters on the principal component analysis of urine from rat and mouse. J Pharm Biomed
Anal 26(3):463–476
Prost E et al (2002) A simple scheme for the design of solvent-suppression pulses. J Magn Reson
159(1):76–81
Psychogios N et al (2011) The human serum metabolome. PLoS One 6(2):e16957
Qiu D et al (2020) Analysis of inositol phosphate metabolism by capillary electrophoresis
electrospray ionization mass spectrometry (CE-ESI-MS). bioRxiv
Ragguett R-M, McIntyre RS (2020) Metabolomics in psychiatry. In: Personalized psychiatry.
Elsevier, pp 459–464
Raiford DS, Fisk CL, Becker ED (1979) Calibration of methanol and ethylene glycol nuclear
magnetic resonance thermometers. Anal Chem 51(12):2050–2051
Rasmussen LG et al (2012) Assessment of the effect of high or low protein diet on the human urine
metabolome as measured by NMR. Nutrients 4(2):112–131
Reif B et al (2021) Solid-state NMR spectroscopy. Nat Rev Methods Primers 1(2):1–23
114 R. T. McKay

Reinsperger T, Luy B (2014) Homonuclear BIRD-decoupled spectra for measuring one-bond

couplings with highest resolution: CLIP/CLAP-RESET and constant-time-CLIP/CLAP-
RESET. J Magn Reson 239:110–120
Reynolds WF, Burns DC (2012) Getting the most out of HSQC and HMBC spectra. In: Annual
reports on NMR spectroscopy: annual reports on NMR spectroscopy volume 76. Elsevier, pp
1–21
Reynolds WF, Enriquez RG (2002) Choosing the best pulse sequences, acquisition parameters,
postacquisition processing strategies, and probes for natural product structure elucidation by
NMR spectroscopy. J Nat Prod 65(2):221–244
Rist MJ et al (2013) Influence of freezing and storage procedure on human urine samples in
NMR-based metabolomics. Metabolites 3(2):243–258
Rocca-Serra P et al (2016) Data standards can boost metabolomics research, and if there is a will,
there is a way. Metabolomics:12–14
Rovnyak D et al (2004) Resolution and sensitivity of high field nuclear magnetic resonance
spectroscopy. J Biomol NMR 30(1):1–10
Sakhaii P et al (2013) Broadband homodecoupled NMR spectroscopy with enhanced sensitivity. J
Magn Reson 233:92–95
Sanders JKM, Hunter BK (1988) Modern NMR spectroscopy: a guide for chemists
Sandusky P, Raftery D (2005) Use of selective TOCSY NMR experiments for quantifying minor
components in complex mixtures: application to the metabonomics of amino acids in honey.
Anal Chem 77(8):2455–2463
Saude EJ, Sykes BD (2007) Urine stability for metabolomic studies: effects of preparation and
storage. Metabolomics 3(1):19–27
Saude EJ, Slupsky CM, Sykes BD (2006) Optimization of NMR analysis of biological fluids for
quantitative accuracy. Metabolomics 2(3):113–123
Saude EJ et al (2007) Variation of metabolites in normal human urine. Metabolomics 3(4):439–451
Schönberger T et al (2014) Guide to NMR method development and validation-part I: identification
and quantification. Eurolab Tech Rep. 01/2014
Schönberger T et al (2015) Guide to NMR method development and validation–part II: multivariate
data analysis
Sheedy JR et al (2010) A sample preparation protocol for 1H nuclear magnetic resonance studies of
water-soluble metabolites in blood and urine. Anal Biochem 398(2):263–265
Sherlock L, Mok KH (2019) Metabolomics and its applications to personalized medicine. Europe-
Korea Conf Sci Technol:25–42
Shishmarev D, Otting G (2011) Radiation damping on cryoprobes. J Magn Reson 213(1):76–81
Shuker SB et al (1996) Discovering high-affinity ligands for proteins: SAR by NMR. Science
274(5292):1531–1534
Silverstein RM et al (2014) Spectrometric identification of organic compounds, 8th edn. Wiley,
Hoboken, p 464
Smolinska A et al (2012) NMR and pattern recognition methods in metabolomics: from data
acquisition to biomarker discovery: a review. Anal Chim Acta 750:82–97
Sokolenko S et al (2013) Understanding the variability of compound quantification from targeted
profiling metabolomics of 1D-1H-NMR spectra in synthetic mixtures and urine with additional
insights on choice of pulse sequences and robotic sampling. Metabolomics 9(4):887–903
Spraul M et al (2015) Wine analysis to check quality and authenticity by fully-automated1H-NMR.
BIO Web Conf 5:02022
Staab JM, O’Connell TM, Gomez SM (2010) Enhancing metabolomic data analysis with progres-
sive consensus alignment of NMR Spectra (PCANS). BMC Bioinform 11 123
Stringer KA et al (2016) Metabolomics and its application to acute lung diseases. Front
Immunol:7–44
Tavares LC et al (2015) Metabolic evaluations of cancer metabolism by NMR-based stable isotope
tracer methodologies. Eur J Clin Invest 45 Suppl 1:37–43
Metabolomics and NMR 115

Tayyari F et al (2013) 15N-cholamine – a smart isotope tag for combining NMR- and MS-based
metabolite profiling. Anal Chem 85(18):8715–8721
Teng Q (2012) Structural biology: practical NMR applications. Springer Science & Business Media
Tenori L, Turano P, Luchinat C (2020) Metabolic profiling by NMR. eMagRes 9:199–204
Tilgner M et al (2019) High-resolution magic angle spinning (HRMAS) NMR methods in
metabolomics. In: NMR-based metabolomics. Springer, pp 49–67
Todd JFJ, March RE (1999) A retrospective review of the development and application of the
quadrupole ion trap prior to the appearance of commercial instruments. Int J Mass Spectrom
190:9–35
Torchia DA (2009) Slight mistuning of a cryogenic probe significantly perturbs the water 1H
precession frequency. J Biomol NMR 45(3):241–244
Tredwell GD et al (2011) Between-person comparison of metabolite fitting for NMR-based
quantitative metabolomics. Anal Chem 83(22):8683–8687
Trifonova OP et al (2021) Mass spectrometry-based metabolomics diagnostics–myth or reality.
Expert Rev Proteomics 18(1):7–12
Valdés A et al (2021) Foodomics: analytical opportunities and challenges. Anal Chem 94(1):
366–381
van Bentum J et al (2016) Perspectives on DNP-enhanced NMR spectroscopy in solutions. J Magn
Reson 264:59–67
van der Sar SA et al (2015) Ethanol contamination of cerebrospinal fluid during standardized
sampling and its effect on (1)H-NMR metabolomics. Anal Bioanal Chem
van Zijl PCM (1987) The use of deuterium as a nucleus for locking, shimming, and measuring
NMR at high magnetic fields. J Magn Reson (1969) 75(2):335–344
van Zijl PC, Hurd RE (2011) Gradient enhanced spectroscopy. J Magn Reson 213(2):474–476
Van Zijl PCM et al (1994) Optimized shimming for high-resolution NMR using three-dimensional
image-based field mapping. J Magn Reson 111:203–207
Van QN, Chmurny GN, Veenstra TD (2003) The depletion of protein signals in metabonomics
analysis with the WET-CPMG pulse sequence. Biochem Biophys Res Commun 301(4):
952–959
Vignoli A et al (2019) High-throughput metabolomics by 1D NMR. Angew Chem Int Ed 58(4):
968–994
Visconti A et al (2019) Interplay between the human gut microbiome and host metabolism. Nat
Commun 10(1):1–10
Vögeli B (2014) The nuclear Overhauser effect from a quantitative perspective. Prog Nucl Magn
Reson Spectrosc 78:1–46
Wang X, Li L (2020) Mass spectrometry for metabolome analysis. Mass Spectrom Lett 11(2):
17–24
Wang H et al (2010) Metabolomics and detection of colorectal cancer in humans: a systematic
review. Future Oncol 6(9):1395–1406
Webb AG (2006) Advances in probe design for protein NMR. In: Annual reports on NMR
spectroscopy. Elsevier, pp 1–50
Weljie AM et al (2006) Targeted profiling: quantitative analysis of 1H NMR metabolomics data.
Anal Chem 78(13):4430–4442
Wilson MJ (2007) The ARRL operating manual for radio amateurs. American Radio Relay League
Wolfender JL et al (2013) Plant metabolomics: from holistic data to relevant biomarkers. Curr Med
Chem 20(8):1056–1090
Wüthrich K (1986) NMR of proteins and nucleic acids, vol 17. Wiley, New York, pp 1–292
Xiao C et al (2009) An optimized buffer system for NMR-based urinary metabonomics with
effective pH control, chemical shift consistency and dilution minimization. Analyst 134(5):
916–925
Xu R et al (1999) Chemical ligation of folded recombinant proteins: segmental isotopic labeling of
domains for NMR studies. Proc Natl Acad Sci 96(2):388–393
Zangger K (2015) Pure shift NMR. Prog Nucl Magn Reson Spectrosc 86–87:1–20
116 R. T. McKay

Zangger K, Oberer M, Sterk H (2001) Pure-phase selective excitation in fast-relaxing systems. J

Magn Reson 152(1):48–56
Zerbe O, Jurt S (2013) Applied NMR spectroscopy for chemists and life scientists. Wiley, p 548
Zhang S, Yang X, Gorenstein DG (2000) Enhanced suppression of residual water in a “270” WET
sequence. J Magn Reson 143(2):382–386
Zhang P, Georgiou CA, Brusic V (2018) Elemental metabolomics. Brief Bioinform 19(3):524–536
Ziarek JJ, Baptista D, Wagner G (2018) Recent developments in solution nuclear magnetic
resonance (NMR)-based molecular biology. J Mol Med (Berl) 96(1):1–8
Ziessow D (1990) Understanding multiple-pulse experiments – an introduction to the product-
operator description I. Starting with the vector model. Concepts in Magnetic Resonance 2:81–
100
Zulyniak MA, Mutch DM (2011) Harnessing metabolomics for nutrition research. Curr Pharm
Biotechnol 12(7):1005–1015
Natural Products Drug Discovery: On Silica
or In-Silico?

Hye Kyong Kim, Young Hae Choi, and Robert Verpoorte

Contents
1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 118
2 Chemical Profiling Techniques in Metabolomics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 122
3 Statistical Methods to Correlate Between Chemical Profiles and Bioactivity . . . . . . . . . . . . . 125
3.1 Correlation with a Simplified Dataset: Discriminant Analysis . . . . . . . . . . . . . . . . . . . . . . . 127
3.2 Correlation Analysis with Non-Discriminant Variables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 128
4 Application of Metabolomics-Based Bioactivity to Various Disease Models . . . . . . . . . . . . . 130
4.1 Antimicrobial Activity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 130
4.2 Anticancer Activity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 131
4.3 Anti-inflammatory Activity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 135
5 Conclusions and Perspective . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 136
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 138

Abstract

Natural products have been the most important source for drug development
throughout the human history. Over time, the formulation of drugs has evolved
from crude drugs to refined chemicals. In modern drug discovery, conventional
natural products lead-finding usually uses a top-down approach, namely
bio-guided fractionation. In this approach, the crude extracts are separated by

H. K. Kim · R. Verpoorte (✉)

Natural Products Laboratory, Institute of Biology Leiden, Leiden University, Leiden, The
Netherlands
e-mail: [email protected]; [email protected]
Y. H. Choi
Natural Products Laboratory, Institute of Biology Leiden, Leiden University, Leiden, The
Netherlands
College of Pharmacy, Kyung Hee University, Seoul, Republic of Korea
e-mail: [email protected]

# The Author(s), under exclusive license to Springer Nature Switzerland AG 2022 117
V. Ghini et al. (eds.), Metabolomics and Its Impact on Health and Diseases,
Handbook of Experimental Pharmacology 277, https://doi.org/10.1007/164_2022_611
118 H. K. Kim et al.

chromatography and resulting fractions are tested for activity. Subsequently,

active fractions are further refined until a single active compound is obtained.
However, this is a painstakingly slow and expensive process. Among the
alternatives that have been developed to improve this situation, metabolomics
has proved to yield interesting results having been applied successfully to drug
discovery in the last two decades. The metabolomics-based approach in lead-
finding comprises two steps: (1) in-depth chemical profiling of target samples,
e.g. plant extracts, and bioactivity assessment, (2) correlation of the chemical and
biological data by chemometrics. In the first step of this approach, the target
samples are chemically profiled in an untargeted manner to detect as many
compounds as possible. So far, NMR spectroscopy, LC-MS, GC-MS, and
MS/MS spectrometry are the most common profiling tools. The profile data are
correlated with the biological activity with the help of various chemometric
methods such as multivariate data analysis. This in-silico analysis has a high
potential to replace or complement conventional on-silica bioassay-guided frac-
tionation as it will greatly reduce the number of bioassays, and thus time and
costs. Moreover, it may reveal synergistic mechanisms, when present, something
for which the classical top-down approach is clearly not suited. This chapter aims
to give an overview of successful approaches based on the application of chemi-
cal profiling with chemometrics in natural products drug discovery.

Keywords

Antibiotics · anticancer · anti-inflammatory · Chemometrics · Correlation

analysis · Discriminant analysis · In-silico · Metabolomics

1 Introduction

Since ancient times mankind has searched for medicinal plants in nature, in early
times very likely by simple trial and error which eventually evolved into complex
systems that included the documentation of the accrued knowledge which allowed
its organized transmission from generation to generation. This is known as tradi-
tional medicine, some outstanding examples of which are the traditional Chinese
Medicine (TCM) and Ayurveda in India. The traditional medicine of the Mediterra-
nean region became the basis of Western medicine (Leonti and Verpoorte 2017).
Thanks to the technological advances in processing and the accumulation of clinical
information, the application of natural products developed from the direct use of a
medicinal plant to the use of extracts, and then to pure compounds. Processing steps
including grinding, drying, fermentation, extraction, heating, and the addition of
other plants or materials (e.g., milk, butter, honey, sugar) have been, and still are,
worldwide common practices in traditional medicine. This information is important
for the transformation of the transmitted traditional knowledge into novel leads for
medicines.
Natural Products Drug Discovery: On Silica or In-Silico? 119

In Medieval times alchemists started to experiment with chemistry, looking for

the essence of medicines, but it was not until the beginning of the nineteenth century
that the first pure active compounds such as morphine, strychnine, quinine, caffeine,
nicotine, atropine, and cocaine were isolated (Samuelsson and Bohlin 2009). How-
ever, the elucidation of the structures of these alkaloids took almost 150 years. At the
end of the nineteenth century, the first synthetic medicines were produced, mostly
based on models from nature, acetylsalicylate being the best known example.
Thanks to the rapid advancement of technology in the twentieth century, many
biologically active natural products have been isolated, some of which became
important therapeutic aids per se, while many others served as a scaffold for the
design of (semi)synthetic drugs.
No matter how we exploit nature for medicinal drugs, testing biological activity
in various stages is an inevitable step for drug development, both for mixtures and
isolated active components. In the twentieth century, pharmacologists developed
various screening tools aimed at the detection of certain biological effects. Among
these, the Hippocratic screening (Malone and Robichaud 1962) is still applied to
screen plant extracts and pure compounds for a wide range of ailments in a systemic
way directly on mice or rats. At some stage, the alleged pharmacological activity
must be proven on animals and human beings. In the past 50 years, simple and fast
in-vitro cell-, enzyme-, or receptor-based assays that can follow the activity during
isolation or synthesis of novel active compounds have been developed. In the later
stages of the evaluation of leads for drug development, detailed pharmacological
studies of the mode of action are needed. This also includes in-silico docking of
compounds in models for various receptors or enzymes.
The importance of bioactivity screening lies in the rapid activity mapping of large
numbers of extracts or fractions that are highly complex mixtures of compounds.
This can be done using bioassay-guided fractionation as an experimental design. The
immanent paradigm of screening tests is a single target-single compound approach,
e.g., measuring the degree of binding of a drug to a receptor or an enzyme. This
approach was particularly successful in novel anticancer drug screening (Cragg and
Newman 2005; Newman and Cragg 2007, 2020). This classical bioprospecting
approach was one of the themes of the Nobel Prize in Medicine and Physiology in
2015, related to the random screening of soil bacteria and the resulting discovery of
avermectin. This antibiotic in turn was the basis for a novel medication for river
blindness. The other part of this Prize went to Chinese research that led to the
identification of artemisinin, a novel antimalarial drug extracted from a millennial
antimalarial traditional Chinese medicinal plant.
These conventional biological screening tests usually follow a top-down
approach, starting from a mixture and narrowing down to a single compound
which is eventually responsible for the alleged activity. However, a multitude of
limitations persist. Among others, the long time it takes to find an active compound
due to the number of steps involved in the identification of the features associated to
activities. Moreover, in many cases the isolated active compound turns out to be an
already-known compound. Furthermore, the long procedure required by conven-
tional screening methods often leads to the loss of activity caused by chemical
120 H. K. Kim et al.

degradation. But the greatest limitation of this method is its intrinsic ignorance of
synergistic effects and prodrugs. The single target-single compound paradigm
appears to contradict the base of traditional medicines that often relies on complex
mixtures of plants in which every plant has a different function. That means that the
multitarget-multicomponent approach is the key difference between traditional and
modern drug development. It is obvious that such a systemic approach to drug
discovery demands different tools. In the case of therapeutic activity it means a
multitarget-testing system, i.e. the use of living organisms. For the identification of
active compounds, a holistic approach such as metabolomics can provide the type of
information required to identify correlations between activities and metabolites,
thanks to the application of chemometric tools (see below).
Metabolomics aims at profiling all the metabolites in an organism. To these ends,
it provides an overview of the metabolic profiles of the subject of interest in an
untargeted manner, that is, ultimately, a systems biology approach (Wang et al.
2005). This systems biology approach using omics tools gained importance in the
quest for biomarkers for diseases and other situations that require the identification
of distinctive markers, e.g. quality standards. The experimental design for this
approach includes processing a large number of representative samples (e.g., plant
extracts or fractions) to obtain both metabolomics and activity data of each one.
Numerous chemometric methods such as multivariate data analysis have been
developed to identify potential correlations between all observations and identify
indicators (e.g., NMR signals, LC-MS or GC-MS peaks) that correlate with a given
activity. In this approach, any variable that has been measured can be used to test for
possible correlations. Metabolomics, eventually in combination with proteomics,
transcriptomics, and/or genomics, can also be used to identify potential targets in the
tested organisms used in bioassays, e.g. cell lines or test organisms such as mice
(Parng et al. 2002), zebra fish (Danio rerio) (Mushtaq et al. 2013; Akhtar et al.
2016), Caenorhabditis elegans (Salzer and Witting 2021), and brine shrimp
(Artemia salina) (Ntungwe et al. 2020), it is possible to gain better insight into the
possible mode(s) of action of an extract or a pure compound. This would be a
molecular follow-up of the classic Hippocratic screening. The method must be
validated by measuring the effect of various classes of drugs on the metabolome
and a database of the changes in the metabolome can then be used to compare with
the effects of extracts or novel compounds.
Finally, there is also the possibility of coupling a bioassay with the separation of
extracts. For example, inhibition of acetylcholine esterase activity can be coupled
with HPLC (Ingkaninan et al. 2000) or TLC (Rhee et al. 2003). Particularly in the
case of TLC, its coupling with various bioactivity screening methods has shown to
be a powerful tool for nano-scale identification of biologically active compounds
(Klingelhofer et al. 2021; Morlock 2021; Morlock et al. 2021; Schreiner and
Morlock 2021). In this case, though, the use of a single compound, single target
approach, makes it somewhat less attractive for the study of complex traditional
medicines, where synergy most likely plays a predominant role (Verpoorte et al.
2018). As seen in the study of Artemisia annua, the antimalarial activity observed in
tea extract cannot be explained only by the already-known active compound
Natural Products Drug Discovery: On Silica or In-Silico? 121

Translational Medicinal Chemical Traditional

Bioprospecting
research chemistry synthesis medicine

In-vitro and in-vivo Bioassays

M

Lead

Clinical studies

Production
Synthesis
New medicine Formulation
Quality control
Biotechnology

Network pharmacology

Fig. 1 Different steps in drug discovery and examples of metabolomics applied in this process.
Adapted from Cuperlovic-Culf and Culf (2016)

artemisinin, as its concentration is too low in the tea extract to be responsible for the
activity. The possible explanation is the presence of other active compounds such as
flavonoids or saponins which could have a synergistic effect with artemisinin
(De Donno et al. 2012). The limitation of this fully “on-silica” approach is that it
is based on the single target-single compound approach, so any activity involving
synergy and other interactions between molecules will not be detected, whereas the
“in-silico” approach can reveal synergistic effects on other types of interaction-based
mechanisms. The “in-silico” tools require the building of databases for fast identifi-
cation of known compounds in extracts, and metabolomics databases of the changes
caused by a standard set of major drugs in the in-vivo test systems. Network
pharmacology would be useful to support such databases. Network pharmacology
brings together all information on the effect of known drugs on the network of
metabolism, signaling and diseases, on the level of genome, transcriptome, prote-
ome, and metabolome. It can be used to obtain an insight into the mode of action of
medicines and evaluate the differences and similarities of new drugs with existing
ones (Boezio et al. 2017; Hopkins 2007, 2008; Jiang et al. 2021; Li and Zhang 2013;
Ye et al. 2016). In Fig. 1 the various aspects of drug development are summarized,
and the applications of metabolomics in these fields are highlighted. There are,
basically, two types of applications. One applies metabolomics to various test
organisms and patients aiming to identify markers for diseases and for the effects
of (novel) medicines on their evolution. In the other case, metabolomics is used to
identify active compounds in complex mixtures, e.g. traditional medicines. The
systems biology approach opens new windows into the interactions between
mixtures of active molecules and the test organisms providing even new insights
122 H. K. Kim et al.

in the diseases themselves. This chapter describes the important features of

metabolomics-based bioactivity screening using diverse activity models including
antimicrobial, anticancer, and anti-inflammatory activities. The analysis of the
chosen examples allows a discussion of the potential and the unresolved limitations
of the method. Apart from the systemic approach chosen for the experimental
design, we will also refer to new ideas for the fast dereplication of active compounds.

2 Chemical Profiling Techniques in Metabolomics

Chromatography became a major game changer in life sciences in the 1960s (TLC
and GC) and 1970s (HPLC). An example of metabolomics avant-la-lettre is a paper
by Baerheim Svendsen and Karlsen (1967) on the GC analysis of the essential oils of
three different plants. However, in the 1980s this was considered to be like
“collecting stamps.” Fifteen years later, this sort of analysis was called metabolomics
and included under the umbrella of “omics” technologies, in which the understand-
ing of the function of genes was studied by the integration of genomics (DNA),
transcriptomics (RNA), proteomics (proteins), and metabolomics data. The DNA
and RNA molecules share similar physicochemical properties, enabling highly
reproducible analyses based on strict robust extraction protocols. This is not the
case of proteins that have a wide range of properties. In proteomics analysis, proteins
are divided into two major classes, according to their solubility in water in hydro-
philic and hydrophobic proteins and have thus different extraction protocols. In
metabolomics, the situation is even more complex due to the great difference in
physicochemical properties of metabolites in general. These can stem from their
molecular size (e.g., polysaccharides, lignin) or even hydro/lipophilicity in the case
of the small molecules. Among the “small” molecules there are primary metabolites,
ubiquitous compounds that are found in all living cells but there are also specialized
metabolites that are generally species-specific and related to the survival of an
organism in its ecosystem. Altogether some 350,000 natural products are known
from various sources (e.g., plants, microbes, insects) (Banerjee et al. 2015).
Estimations of the number of specialized metabolites are mere speculations, but
knowing that the number of species of living organisms in the world is somewhere
between 10 and 100 million (Pimm et al. 1995) and assuming that every species
produces one unique compound, there should be between 10 and100 million
metabolites in the metabolome of the Earth. The total chemical space of natural
products is thus very much larger than that covered by our present knowledge. There
is still so much to be discovered!
This brings us to the problem of how to define a metabolome. We have just
referred to the metabolome of the planet earth, i.e., the sum of all the metabolomes of
all the organisms (dead or alive) and their environment. Each plant species has its
own metabolome, which is the sum of the metabolomes of all its organs, i.e., roots,
stem, and leaves, etc. These metabolomes correspond to that of the many different
cells in the plant. Even within the cell the different organelles have different
metabolomes. The metabolome of a species might be best defined as the measurable
Natural Products Drug Discovery: On Silica or In-Silico? 123

set of metabolites that an organism is able to produce, i.e. some compounds might
not always be present in certain plant parts, but their presence depends on the
existence of certain circumstances. The metabolome of a species depends on the
species’ genome. Obviously the single cell analysis at all omics levels is the holy
grail of the omics.
Clearly, any reference to metabolomics requires a clear documentation of the
source of the “metabolome.” The ultimate goal of metabolomics is to describe the
changes in the metabolome under well-defined conditions. Eventually the fluxes in
the organism through the various metabolic pathways should be analyzed
(fluxomics), as changes in the homeostasis of an organism are connected with
changes in fluxes.
To attain the goal of metabolomics, all applications have the same requirements,
i.e. both qualitative and quantitative information about all metabolites in the
biological sample must be generated. However, as per the discussion above it is
clear that the requirements vary according to the application and so will the method
needed to obtain the information. Firstly, when searching for biomarkers of diseases
or bioactivities, it is essential to be able to count on a large database with
metabolomic data of the test organisms to be used or material that has the desired
activity, e.g. extracts from a specific plant species with an interesting activity. This
database should ideally provide information on the effect of all biological variables
on the studied material, for example, the diurnal variation of the metabolome.
Existing databases with all kinds of information from previous experiments that
provide abundant background information can help in the identification of
compounds from their spectra and/or chromatographic data. However,
bioprospecting requires the screening of large numbers of plants that are
characterized by a highly variable and species-specific metabolome, particularly in
the case of specialized metabolism, a situation that is clearly different from the case
of mammalian cells, for example, in which most compounds are well-known
primary metabolites.
It is now generally accepted that no single analytical tool can reveal the real
metabolome (Wishart 2008; Emwas et al. 2019). At present, analytical chemists are
moving in two possible directions to circumvent this limitation, i.e., improving the
quality of individual techniques and/or integrating data from multiple methods.
In the case of NMR, the main limiting issues are its low sensitivity (μmol for
detection limit) and complexity of signals. Sensitivity has been increased using cryo
(cold)-probes and reducing the diameter of NMR tubes (capillary-, micro-, or nano-
tubes). Nevertheless, the sensitivity is still far below that of MS-based methods
(up to 10 times for conventional NMR probe). Generally speaking, NMR is consid-
ered to be the most powerful analytical tool for structure elucidation of pure
compounds but not necessarily in the case of mixture analysis. The most attractive
advantage of NMR as a metabolomics tool is its ease of quantitation. The height or
area of 1H NMR signals is directly proportional to the molar concentration of
analytes, i.e. with a single internal standard all signals can be easily integrated and
quantified. It is in this aspect that NMR has a clear edge over all other analytical
techniques.
124 H. K. Kim et al.

It should also be noted that NMR is currently the only metabolomics method that
does not include a separation step.
In MS-MS, the separation is in the first MS dimension, based on molecular mass,
after which the individual signals are subject to a second MS step in which the
molecules are subject to fragmentation. In hyphenated chromatography-MS
techniques, the retention time, molecular mass, and the fragmentation pattern can
be used to search databases for already-known compounds. For example, it is
possible to identify unknown peaks with high-resolution mass measurements in
combination with the “seven golden rules” or other “dereplication” approaches.
Though (stereo) isomers will be difficult to identify with 100% certainty, unfortu-
nately too often the identity of a compound is accepted ignoring the inherent level of
uncertainty of the used method. For a full identification of new or rare compounds,
their UV, MS, 1H NMR, 13C NMR, and various multidimensional NMR spectral
information must be analyzed. The 1D-NMR data alone are insufficient. Most cases
require multidimensional NMR methods such as COSY, TOCSY, HSQC, and
HMBC spectroscopy to determine distances and interactions between protons,
e.g. to confirm stereochemistry. Known compounds can be identified with a reason-
able certainty if isolated, but the full structural elucidation of new specialized
metabolites from natural products mixtures is still enormously challenging. The
technical limitations of the instrumental NMR analysis were partially solved by a
statistical signal correlation of 1H NMR resonances, the so-called statistical total
correlation spectroscopy (STOCSY) that has been used to identify novel metabolites
in urine, enabling the selection of signals characteristic of one molecule; there are, as
well, a multitude of deconvolution methods for NMR data (Cloarec et al. 2005). As
mentioned before, most of the chemical diversity of metabolites includes specialized
metabolites of which there are no reliable comprehensive NMR databases. However,
there are a number of available NMR spectra databases of primary metabolites
(Human Metabolome Database (HMDB, http://www.hmdb.ca), Biological Mag-
netic Resonance Bank (BMRB, http://www.bmrb.wisc.edu/metabolomics/),
NMRshiftDB (http://nmrshiftdb.ice.mpg.de/).
Mass spectrometry (MS) is a highly sensitive method (pmol level) and is also
selective due to the high resolution and level of accuracy of the determined molecu-
lar mass and the different fragmentation patterns of the metabolites. The number of
detected signals in MS-based platforms is 10–100-fold that of NMR. In terms of
identification, however, it has inherent limitations. The mass spectrometer only
detects ions formed in the ion source, but there can be a large variation in the
sensitivity for the formation of ions. That means that absolute quantitation is only
possible by running calibration curves for every single compound within certain
ranges of concentrations. Identification is possible by comparison of the exact
molecular mass of the compounds with databases and comparison of fragmentation
patterns with possible candidates and closely related compounds. The MS data is
insufficient for the structural elucidation of new compounds. For low molecular
mass values there are scores of isomers, thus for the elucidation of the full structure
of a molecule, including its stereochemistry, further spectral data are needed (see
above).
Natural Products Drug Discovery: On Silica or In-Silico? 125

Recent improvements of MS in metabolomics have risen from signal

deconvolution of MS data rather than sensitivity or accuracy which is already
remarkably high. This means that the focus has moved to the identification of
compounds by increasing resolution and developing algorithms that can deduce
molecular formulae from adduct ions based on a classification of molecules. Increas-
ing resolution is mostly associated with improved mass analyzers, such as a quadru-
pole, TOF, Orbitrap, or iontrap. In the early times, the key issue was to improve
accuracy of ion mass by suppressing fragmented ions. But the trend is changing. The
goal is now to keep fragment signals together with molecular ones to use them for
fingerprinting. This additionally offers the possibility of comparing the spectra to
reveal the shared fragments of compounds which may point to a basic structure that
is similar to all. Recently, many manufacturers offer improved ion trap analyzers to
produce robust MS/MS signals that are the key for the identification with matching
techniques. Unlike NMR signals, MS data do not provide any indication of similar-
ity between signals. Jeffryes et al. (2015) have suggested the use of Metabolic
In-silico Network Expansions (MINEs) as a new tool for identifying metabolites
from an LC-MS dataset. MINEs used generalized biochemical transformations to
propose structures, leading thus to the suggestion of putative metabolite structures.
Another dereplication approach is the use of molecular networking (MN) (Yang
et al. 2013). This statistical correlation of MS/MS data is based on a generated
MS/MS database of metabolites which can be searched for fragments of known
metabolites in the biosynthetic network. The construction of a molecular network is
based on the analysis of MS/MS spectra of compounds, presuming that molecules
with similar structures should display similar fragmentation patterns. Allard et al.
(2016) demonstrated the use of MN combined with other in-silico MS/MS fragmen-
tation database as a dereplication strategy of the metabolites from natural sources.
The comparison of different profiling tools is summarized in Table 1.

3 Statistical Methods to Correlate Between Chemical Profiles

and Bioactivity

The key step of OMICS-based chemical profiling is data mining using statistical
analysis. The metabolomics datasets generated by NMR or MS are vast, requiring
chemometrics to extract any useful information from the data. The chemometrics
methods usually include a statistics-focused approach and/or a computer-dominant
approach such as machine learning (Wishart 2008). Paul and de Boves Harrington
(2021) summarized the basic concepts and applications of those methods very
clearly.
In this section, we will focus on the statistical approach which is generally used to
detect the active compounds in natural products extracts. This approach uses two
different methods to link signals of compounds with activity: discriminant analysis
or correlation analysis.
126 H. K. Kim et al.

Table 1 Comparison of different technologies used in metabolomics

Technology Advantages Disadvantages
NMR • Non-destructive • Low sensitive
spectroscopy • Rapid (ca. 5 min per • Requires large sample size
sample) • Cannot detect inorganic ions and
• Simple sample non-protonated compounds
preparation • Expensive equipment
– Requires no
derivatization
– Requires no
extraction in case of biologic
fluids
• Detects broad range of
compounds
• Quantitative
• Strong structural
elucidation and identification
power
• Robust and reproducible
• Compatible with liquids
and solids
GC-mass • Robust • Destructive (Sample not recoverable)
spectroscopy • High resolution and • Requires sample derivatization
sensitivity • Requires individual calibration curve
• Requires modest sample for absolute quantitation
size • Takes longer time (ca 30 min per
• Detects most organic and sample)
some inorganic molecules • Limitation on the novel compound
• Excellent database for identification
identification
LC-mass • High sensitivity • Destructive
spectroscopy • Requires minimal sample • Requires individual calibration curve
size for absolute quantitation
• Relatively easy sample • Takes longer time (ca 30 min per
preparation sample)
• Deconvolution • Limitation on the novel compound
overlapping features identification
• Limited software available and
databases for (secondary) metabolite
identification
MS and • High resolution & • Destructive
MS/MS reproducibility • Requires individual calibration curve
• Relative quantitation for absolute quantitation
• Large databases for • Limited software available and
metabolite identification databases for (secondary) metabolite
• Can be used in metabolite identification
imaging single cells (MALDI)
NMR nuclear magnetic resonance, GC gas chromatography, LC liquid chromatography, MALDI
matrix-assisted laser desorption/ionization
Natural Products Drug Discovery: On Silica or In-Silico? 127

3.1 Correlation with a Simplified Dataset: Discriminant Analysis

Discriminant analysis is nowadays the most popular approach for the identification
of active compounds in complex mixtures. Partial least squares discriminant analysis
(PLS-DA) is frequently applied to detect differences by making qualitative
classifications, such as active and non-active samples; it is based on the definition
of preferably two classes, and any type of classes can be formulated. PLS-DA, a
supervised method, models the variance within the dataset by statistically discrimi-
nating groups of observations. Orthogonal PLS (OPLS) was developed to improve
the correlation power of PLS by orthogonalizing non-related variables through
orthogonal signal correction (OSC filtering). This facilitates the interpretation of
the model because the variables (metabolome) are related to the targeted property
(e.g., high or low bioactivity).
A good example of this application, published by Cardoso-Taketa et al. (2008), is
the correlation of the sedative effect of Galphimia glauca Cav. and galphimine, a
triterpenoid already known to occur in this plant. Working on six different
collections of G. glauca they found two collections that were highly active, while
the rest did not show much activity. Applying PLS-DA to the NMR metabolomic
data and the sedative effects found in the animal model, they were able to correlate
the biological activity with galphimine.
A similar approach was applied to identify biologically active metabolites
obtained with different extraction methods and solvents from Ocotea odorifera
using MS-based metabolomics. Alcântara et al. (2021) investigated the anti-
inflammatory activity of a decoction of Ocotea odorifera and different fractions
from its ethanolic extracts using dual inhibition of edema and neurophil recruitment.
The chemical profiling data obtained by UPLC-HRMS or GC-MS were correlated
with the anti-inflammatory activity by PLS-DA, resulting in the identification of
S-(+)-reticuline as an active principle. Its activity was confirmed by testing the pure
compound after isolation.
Other examples of the successful application of this approach are the studies of
the antimicrobial activity of essential oils (Maree et al. 2014) and the identification of
an antitussive active compound from Tussilago farfara L. (Lia et al. 2013).
The work of Maree et al. (2014) showcases the strength of the chemometrics and
in particular of the OPLS-DA method. Based on the data obtained from the analysis
of 158 different essential oils by GC-MS and their antimicrobial activities on several
different strains of microorganisms, two classes were defined: active (MIC <2 mg/
ml) and non-active (MIC >2 mg/ml). This allowed the detection of eugenol as a
putative marker for activity, and an apparent synergistic effect with geraniol.
α-Pinene, limonene, and sabinene, other components of the essential oils correlated
with no activity. The study also revealed both antagonistic and synergistic antimi-
crobial effects between monoterpenes and eugenol on some microbes.
The correlation approach is clearly more efficient than conventional bioactivity-
guided fractionation, to identify active compounds in the initial stage of drug
development. The limitation of this application, however, lies in a tendency to
overfit if the number of variables considerably exceeds the number of samples
128 H. K. Kim et al.

(Gromski et al. 2015). In other words, in many cases there is a risk of a

model showing a significant separation mainly by chance. This problem can be
partially solved by increasing the number of samples which in turn increases the
demand on the robustness of the analytical methods to, i.e. their ability to provide
highly reproducible results.
The possibility of overfitting in PLS- and OPLSD-DA can be tested by proper
validation methods. Commonly used validation tests are the permutation test or the
CV-ANOVA test (Szymańska et al. 2012). Once the model is validated,
corresponding discriminating components can be further investigated to consider
the responsible metabolites for the activity, preferably at single compound level. The
validation methods for the correlation tests have been well reviewed by Westerhuis
et al. (2008) providing details of their advantages and limitations.
Other limitations are associated with the quality of biological data. Ideally, the
number of data processing steps should be reduced as it has been shown that there is
an inverse relationship between the number of steps and closeness to an intact
correlation. However, it is not always possible to have two clearly activity-
distinguished groups since most biological data have a much lower degree of
resolution than chemical data. Thus, original data are often grouped into simple
classes, for example, active and non-active groups that severely misrepresent reality.
For example, Maree et al. (2014) considered the samples with MIC values below
2 mg/ml as active and non-active above this value. It is challenging to set up the
proper criteria, in fact, in most cases it is based on trial and error. Although the
simplified variables can reduce the ambiguity in the correlation between two
datasets, there is clearly some degree of overfitting. Instead of dividing groups by
different criteria, it could be better to work with the quantitative data obtained from
the biological activity tests and correlate the activity directly without the input of
supervised group information.

3.2 Correlation Analysis with Non-Discriminant Variables

In a correlation analysis, the relationship between two different datasets,

e.g. metabolome data (mostly X-variables) and biological data (Y-variables, quanti-
tative), is determined without simplifying original variables. Among the available
methods, partial least squares to latent structures (PLS) modeling is the most
popular. In this method, biological data are the quantitative and continuous
variables, if necessary with some post-processing, e.g. logarithmic transformation
of the biological data. Many studies have been done using this approach to identify
active metabolites.
One example is the study published by Yuliana et al. (2011) who identified two
flavonoids as responsible for the adenosine A1 receptor binding activity of
Orthosiphon aristatus (Blume) Miq. (synonym Orthosiphon stamineus) based on
the correlation of its NMR-based metabolome and the tested activity. In their study, a
novel gradient extraction method was applied. In this method, instead of a single
solvent, a stepwise polarity gradient from, e.g., ethyl acetate via methanol to water
Natural Products Drug Discovery: On Silica or In-Silico? 129

was pumped through the ground dry plant material mixed with kieselguhr (7:1),
yielding between 20 and 30 fractions. Each fraction was profiled by 1H NMR and its
potential anti-obesity activity was assessed based on an adenosine A1 receptor
binding assay. Examination of the loading plots of the PLS and OPLS models led
to the identification of two active methoxyflavonoids, 4′,5,6,7-tetramethoxyflavone
(tetramethyl scutellarein) and 3′,4′,5,6,7- pentamethoxyflavone (sinensetin).
Another example is the work of Ali et al. (2013), who applied PLS and OPLS
analyses to correlate the metabolome of sponges obtained using NMR spectroscopy
with their effect in the adenosine A1 receptor binding activity assay. The
metabolome of over 200 sponge samples was obtained using 1H NMR, and their
adenosine A1 receptor binding activity was measured. OPLS analysis was
performed with two datasets. From the loading plots several signals were selected
as activity-associated. Most signals were unknown or difficult to identify only by
1
H-NMR, requiring the isolation of compounds. Having isolated several compounds
which matched well with activity-associated signals, these active metabolites were
then identified as the sesterterpenes, halisulfate -1, -3, -4, -5, and suvanine.
De Melo et al. (2020) studied the spasmolytic activity of Cissampelos
sympodialis Eichler leaf extract in a trachea preparation. While warifteine, a
bisbenzylisoquilonine alkaloid, is allegedly the main bioactive substance in this
species, its low solubility in polar solvents suggested the presence of another
bioactive compound in the aqueous extracts. To investigate this alternative, PLS
modeling was used to study the correlation of the metabolome obtained by NMR of
diverse polar extracts with spasmolytic activity. The PLS model showed that the
signals from flavonoids were positively correlated with the activity. Therefore, they
concluded that identified flavonoids such as kaempferol and quercetin might be the
important contributors to the activity.
Many studies have applied successfully the combination of metabolomics and
bioassays with chemometrics in the identification of biologically active compounds
in complex mixtures. Researchers can identify active metabolites from the crude
extracts without isolation of single compound with this approach. However, from a
practical point of view, the metabolomics approach itself is not enough to identify
active molecules. Ultimately, the bioactivity of a compound must be confirmed with
tests on the pure compound. In any case, and particularly in case of novel
compounds, proper identification requires the determination of the full set of physi-
cochemical properties and spectral data. Nonetheless, the advantages of the
metabolomics approach are clear. To begin with, it allows the fast dereplication in
case of already-known active compounds. Secondly, there is no need for large-scale
isolation of the active compounds, nor bioassays and, if necessary, fractionation can
be guided by the NMR or (LC- or GC-) MS spectra of the fractions. Particularly in
the case of in-vivo bioassays (including clinical trials) only one first round of
bioassays is needed, after which chemistry-guided fractionation is sufficient to
isolate the active compound. In many cases this may have the additional advantage
of reducing the number of animal experiments.
130 H. K. Kim et al.

4 Application of Metabolomics-Based Bioactivity to Various

Disease Models

4.1 Antimicrobial Activity

Most infectious diseases are easy to diagnose and humans have been tirelessly
searching for medicines in nature to treat infections since ancient times. With the
discovery of penicillin in 1928 by Fleming, microorganisms have been targeted for
the discovery of other novel antibiotics. After a period of reduced interest, in recent
years there is a revival in antibiotic bioprospecting, mainly because of the rapid
spread of antibiotic-resistant pathogenic microorganisms and the failure to develop
active derivatives of the existing ones. The conventional lead-finding of antimicro-
bial drugs is bioactivity-guided fractionation much like for other disease models.
Aided by chemometrics, diverse in-silico methods have been recently introduced
into this field, most of which are applied in the post lead-finding steps such as hit
characterization and hit optimization (Xu et al. 2009). The application of
metabolomics in the process of lead-finding involves:

1. metabolic profiling of crude extracts of selected organisms with antimicrobial

activity, identification of active compounds
2. metabolomics-guided fractionation to isolate an active compound(s)
3. metabolic profiling of pathogenic microorganisms after treatment with potentially
antimicrobial extracts or leads to identify their potential mode of action.

Tang et al. (2015) identified 10 antimicrobial compounds from burdock (Arctium

lappa L.) leaves. The chemical composition of the leaf extracts was analyzed by
UPLC-MS. For the activity, the inhibition of biofilm formation was measured.
Biofilms are communities of microorganisms that are attached to a surface and are
considered to be a survival mechanism of bacteria (Donlan and Costerton 2002).
Anti-biofilm compounds can be used as antimicrobials. The obtained chemical and
biological data were correlated by PLS-DA and resulted in the identification of
10 activity-correlated metabolites: chlorogenic acid, caffeic acid, p-coumaric acid,
quercetin, ursolic acid, rutin, luteolin, crocin, benzoic acid, and tenacissoside. After
evaluating the activity of pure samples of these compounds, chlorogenic acid and
quercetin were determined to be the main anti-biofilm active compounds isolated
from burdock leaf.
Dos Santos et al. (2018) used a similar approach to examine the antimicrobial
activities of the volatile oils of several plant species. Eight volatile oils representing
different levels of antimicrobial activity (from inactive to very active) were selected
and profiled using GC-MS for their metabolomic analysis. The subsequent OPLS-
DA revealed a high correlation of several metabolites 7,8-epoxy-1-octene,
cis-α-bergamotene, methyl linolelaidate, alloaromadendrene, and veridiflorol with
a significant antimicrobial activity. Interestingly, evidence of a specific chemical
interaction between bornyl acetate and 4-terpineol was also observed, since they
Natural Products Drug Discovery: On Silica or In-Silico? 131

were negatively correlated with each other in terms of activity, suggesting antago-
nism of those compounds.
Streptomyces species are the most studied Actinomycetes for the production of
antibiotics. Wu et al. (2015a) applied NMR-based metabolomics to Actinomycetes
and rapidly discovered novel antibiotics. The metabolomic comparison of wild-type
and streptomycin-resistant actinomycetes strains revealed a stronger antimicrobial
activity of the latter. Using PLS analysis, 7-prenylisatin was identified as the
bioactive compound. The antimicrobial activity was confirmed by studies with the
pure isolated compound (Fig. 2).
Because of bacterial resistance to current antibiotics, it is extremely important to
find active compounds with a different mode of action to that of the current
antibiotics. Thus, the mode of action should be identified in an early stage of lead-
finding. A metabolomics approach in which the effect on the pathogenic
microorganisms of known antibiotics and novel candidates is compared could
provide an insight into the mode of action (Hoerr et al. 2016). Clearly, metabolomics
is an important tool that should speed up the process of finding novel antibiotics.
While not directly connected to the correlation between chemical ingredients and
antibiotic activity, the possibility of the induction of new antibiotics in a co-culture
system (Streptomyces and Aspergillus) was probed using an analogous approach.
Though well-known as a source of antibiotics, the Streptomyces metabolome is very
sensitive to a number of external conditions. By co-culturing two different species,
the interactions between the organisms may induce the production of novel
compounds in either species. Moreover, catabolism of the compounds formed may
add even further chemodiversity. Wu et al. (2015b) showed that the co-culture of
Streptomyces with Aspergillus produced significant modifications in their individual
biosynthetic processes yielding many new metabolites which could not be detected
in the individual organisms.

4.2 Anticancer Activity

Natural products and their derivatives have a tremendous potential for the develop-
ment of anticancer drugs. More than 50% of all anticancer drugs approved for
therapeutic use in the past 70 years are either natural product or semi-synthetic
analogs (Butler et al. 2014). Metabolomics has been used not only for drug devel-
opment, but also in many fields related to cancer research including cancer progno-
sis, diagnosis, and treatment efficacy. The in-silico approach described above has
been applied to the search of new lead compounds in many natural resources. An
interesting example of its application in cancer research is reported by Graziani et al.
(2018). Extracts of 14 legumes were screened against an array of human colorectal
cancer cell lines in the search for bioactive natural products. Two plant extracts were
selected for their strong antiproliferative activity using PCA and Hierarchical Cluster
Analysis (HCA) (Fig. 3). The 2D NMR analysis of these plant extracts allowed the
identification of two putative active compounds: a cycloartane glycoside and a
protodioscin derivative. To confirm their bioactivity, these two compounds were
132 H. K. Kim et al.

Fig. 2 Example of multivariate data analysis. PLS score plot (a) shows a good separation between
wild-type Streptomyces sp. (blue) and its two mutant derivatives (MBT28-30: light blue and
MBT28-91: red). The corresponding loading plot (b) presents the NMR signals (dashed circles)
which contribute mostly to the separation of the samples and their bioactivity. The arrow refers to
the characteristic proton signal at δ 7.03 (t, J = 7.2 Hz). X, primary variable of the chemical shift; Y,
bioactivity. (Adopted from Wu et al. 2015a)
Natural Products Drug Discovery: On Silica or In-Silico? 133

Fig. 3 Data of cell growth percentage were analyzed by HCA dendrogram (a) and PCA (b).
Control and colon cancer cell lines which were treated with the plant extracts were distinguished.
These analyses enabled classification of the species into 3 subsets; Groups I – the active species
134 H. K. Kim et al.

isolated and tested positively for antiproliferative activity against colon cancer cells.
These results are clear evidence of the efficiency of metabolomics as a tool to speed
up the lead-finding process, since the active compounds are already identified in the
crude extracts avoiding the time-consuming on-silica isolation procedure.
Using a similar approach, Bao et al. (2018) applied UPLC-MS-based
metabolomics to identify anticancer compounds in Forsythiae fructus (from For-
sythia suspensa Vahl). Crude extracts obtained with diverse solvents were profiled
and the results were correlated with their anticancer activity against murine mela-
noma B16-F10 cell lines. OPLS-DA was applied to maximize the discrimination of
extracts with different degrees of anticancer activity resulting in the selection of
betulinic acid as a potential anticancer compound in Forsythiae fructus.
Gao et al. (2010) performed antiproliferative activity tests against human lung
cell line SK-MES-1 on Scutellaria baicalensis Georgi root extracts. The data
provided by the HPLC-UV and 1H NMR analysis of the extracts were combined
with PCA and PLS analysis for more complete metabolomics data, allowing the
identification of baicalin, baicalein, and wogonin as the compounds responsible for
the cell growth inhibition activity of the extracts.
Tawfike et al. (2019) adopted a different approach to study anticancer activity of
the endophytic fungus Aspergillus flocculus, isolated from the stem of the medicinal
plant Markhamia lutea (Benth.) K.Schum. (Markhamia platycalyx is a synonym). Its
in-vitro culture was found to yield high levels of anticancer compounds active
against the chronic myelogenous leukemia cell line K562. Several fractions of the
fungal culture were analyzed by LC-MS. A combination of a molecular interaction
network and OPLS-DA of the chemical data resulted in the identification of five
active metabolites, namely, cis-4-hydroxymellein, 5-hydroxymellein, diorcinol,
botryoisocoumarin A, and mullein. This study clearly shows the advantage of
combining different forms of in-silico analysis prior to any purification attempts.
Another successful study using this approach was recently published by Ory et al.
(2019) involving an anti-breast cancer compound (tested on MCF-7 cells) from a
marine-derived Penicillium chrysogenum extract. Using different correlation and
discriminant analysis including PLS-DA and PLS modeling, ergosterol was found to
have anticancer activity with an antiproliferative activity on MCF-7 cells with an
IC50 at 0.10 mM concentration. An interesting aspect of this report is the description
of a workflow that allows the combination of both the chemical and biological data, a
key step in this sort of application. Sometimes, a metabolomics approach alone is not
sufficient to identify active metabolites in the crude extracts due not only to the
complexity of the mixture but also to the low concentration levels of potentially
interesting metabolites. In that case, conventional bio-guided fractionation can be
useful as a supplementary tool. The work of Graziani et al. (2021) is a good example
of this. They evaluated the anticancer activity of two plant species, Ononis diffusa

Fig. 3 (continued) (black), group II – active only at the highest tested doses (gray), and group III –
no significant effect (white). Adapted from Graziani et al. (2018)
Natural Products Drug Discovery: On Silica or In-Silico? 135

Ten. and O. variegata L. using colorectal cancer cell lines. When the NMR-based
metabolomics data were correlated with the anticancer activity, most of the poten-
tially interesting activity related signals were concentrated in the aliphatic region of
the NMR spectra, a region which is difficult to annotate. The partial purification of
the extracts using column chromatography and the application of different 2D NMR
techniques led to the identification of a mixture of oxylipins as putative bioactive
compounds. However, the report provides no information on activity confirmation of
pure isolated compounds.
Gao et al. (2010) reported the application of metabolomics to study the mode of
action of (-)-5-hydroxyequol, an isoflavone metabolite obtained by microbial
biotransformation known to affect lung cancer cells through an unknown mechanism
of action. Applying 1H NMR-based metabolomics it was possible to observe rapid
changes in the metabolism in human lung cancer cells, especially in the
glycometabolism.

4.3 Anti-inflammatory Activity

Inflammation is involved in complex diseases such as autoimmune diseases, meta-

bolic syndrome, neurodegenerative diseases, cancers, and cardiovascular diseases
(Chen et al. 2017). The development of new inflammatory modulators to treat such
diseases is therefore of great interest in the pharmaceutical industry. However, their
development has been hampered by several issues regarding their potency, efficacy,
and adverse effects. Since the introduction of acetylsalicylic acid in the nineteenth
century, based on the use of salicylic acid-containing medicinal plants, nature has
been regarded as an important resource to screen in the search for anti-inflammatory
drugs resulting in multiple reports of natural products with alleged anti-inflammatory
effects both in vitro and in vivo. Unsurprisingly, metabolomics has also been applied
frequently in recent times to detect novel lead compounds from nature. A good
example was recently reported for the methanolic extracts of Cyrtanthus contractus
N.E.Br. (Amaryllidaceae) bulbs by Raŕova et al. (2019). These bulb extracts showed
a significant anti-inflammatory activity decreasing the level of E-selectin, a key
player in the initiation of inflammation in a dose-dependent manner. Fractions
from the extracts were profiled using LC-MS and evaluated for their biological
activity. Subsequently, the correlation between biological activity and metabolite
levels was calculated, resulting in the identification of narciclasine as a putative
active compound. The bioactivity of pure narciclasine confirmed the findings.
Alcântara et al. (2021) reported a similar approach to study the leaves of Ocotea
odorifera (Vell.) Rohwer. This plant has been used traditionally for the treatment of
rheumatism. Chemical profiles of leaf extracts and their fractions were obtained
using LC-HRMS and their in vivo anti-inflammatory activity was tested. PLS-DA of
the chemical data suggested that the activity is correlated with S-(+)-reticuline, a
known alkaloid from this plant. Yet another unknown compound, which was not
identified in this study was also found to be active. The researchers proved their aim,
136 H. K. Kim et al.

which was to show the competitiveness of metabolomics as a rapid tool for

dereplication.
Domingos et al. (2019) demonstrated the successful identification of a new anti-
inflammatory compound using this approach. The extracts from Poincianella
pluviosa var. peltophoroides (Benth.) L.P.Queiroz (unresolved name) showed sig-
nificant in vivo anti-inflammatory activity. The UHPLC-HRMS profiling of the
extracts allowed the isolation of four compounds using a metabolomics-guided
chromatographic process. Of these, two were previously unknown compounds and
were identified as 4′′′-methoxycaesalpinioflavone and 7-methoxycaesalpinioflavone
while the other two were the known compounds, rhuschalcone VI and
caesalpinioflavone. The activity of all four was confirmed using pure compounds.
The combination of metabolomics and bio-guided fractionation has thus proved
to be very efficient for the identification of active metabolites. A further example of
its successful application is the LC-MS-based metabolomics of Actinidia arguta
(Siebold & Zucc.) Planch. ex Miq. leaves which combined with bio-guided fraction-
ation resulted in the identification of the active compounds caffeoylthreonic acid and
danshensu (Kim et al. 2019). The identification of known compounds is thus
possible by using spectral data obtained from NMR and/or MS spectroscopy.
Thanks to the development of analytical techniques and technological improvements
in existing instrumentation, it is now much easier to analyze mixtures and identify
the biologically active components within them.

5 Conclusions and Perspective

We have selected representative examples among the hundreds of published reports

that describe the role of metabolomics in natural products lead-finding. Nature has
been and will most likely continue to be a relevant, if not the most relevant, source
for novel drugs as reflected in, among others, the review by Newman and Cragg
(2007). Biodiversity is essential for chemodiversity. This was naturally appreciated
by our ancestors, who understood its power and were able to take advantage of
it. Medicinal plants such as Atropa belladonna L. and Papaver somniferum L. are
just two among the many examples of plants that have provided humanity with a
number of important therapeutic solutions that are part of the core of western
medicine.
Analyzing natural products research for novel drugs, it is clear that there have
been two ways to go about it. One consists in the at-random screening of plants and
other organisms for a certain activity. The other is to use the accrued traditional
knowledge on medicinal plants as a starting point. Whichever the approach chosen,
researchers will have to deal with complex mixtures of compounds among which a
few active ones will need to be identified. Active compounds might be present in
very low concentrations, might be very labile, or present in an inactive form that may
require some kind of transformation to acquire activity. Furthermore, experience has
shown that in many cases, the alleged activity of a traditional medicine is due to a
particular combination of compounds that act through synergistic and/or
Natural Products Drug Discovery: On Silica or In-Silico? 137

antagonistic interactions. Bioprospecting throughout the past decades used the

classical approach of at-random screening for a certain activity followed by
bioassay-guided fractionation. The present industrial approach is at-random high-
throughput screening using fast molecular level bioassays. However, this approach
is an elaborate, time- and energy-consuming process with results that do not often
enough match with the invested resources. This experimental design is based on the
single target-single compound paradigm and having been undoubtedly successful in
some cases was used for many years and continues to be used. An example of this
approach was the NCI program for antitumor drugs which yielded some very active
molecules and leads. However, the present day perspective of big pharmaceutical
companies regarding novel drugs development is one of caution, considering the
level of risk of the investment required, leaving the lead-finding process in hands of
small start-up companies, universities, or public research institutions. The biodiver-
sity treaty (Nagoya protocol on Access and Benefit-sharing, https://www.cbd.int/
abs/about/) did not really encourage the pharmaceutical industry to invest in
bioprospecting, whereas on national level investment in such research is limited.
The studies on local traditional medicines have often higher priority. It is clear that to
increase the chances of success, the whole process has to be speeded up, and
particularly in connection with traditional medicine sources, more in-vivo assays
should be performed to validate their use, using a systems biology-based approach.
Whichever the chosen approach, metabolomics can be used to shorten the time to
identify the active compounds. Metabolomics of different extracts from a plant in
combination with the bioassay data can in both designs be used to identify the
signals from the metabolomics data that correlate with activity. Based on the
information obtained from the set of analytical data (e.g., NMR signals, Mass
spectra, retention time), a preliminary identification is possible by comparing the
data with already available data from a metabolomics database. This information
allows an informed decision on whether it is worthwhile to isolate the presumably
active compound(s) for further studies. If the compound is deemed of interest, the
isolation procedure must be scaled up to obtain sufficient amounts of the pure
compound(s) to submit it to full spectral data analysis and the relevant bioassay(s).
For this metabolomics-guided fractionation should be preferred over the bioassay-
guided fractionation process since it is faster and may reduce the use of animal
experiments.
In conclusion, the “on-silica” approach remains an excellent tool to fractionate
extracts on a small scale for metabolomics analysis and on a larger scale for the
isolation of sufficient amounts of pure active compounds for full spectral analysis for
identification or for structure elucidation of novel compounds.
The “in-silico” approach, based on the systems biology paradigm of holistic
observations, accelerates the dereplication of known active compounds, among
others, by reducing the required number of bioassays and animal experiments.
Moreover the “in-silico” approach will allow deeper insight into the mode of action
by measuring the metabolic changes observed in living cells and organisms when
treated with drugs. A further not inconsiderate advantage of the application of “in-
silico” tools is the reduction of the time needed to identify an active compound, but
138 H. K. Kim et al.

most importantly in reduction of the amount of toxic, contaminating solvents and

consumables required by early stage dereplication of known active compounds. In
other words, the “in-silico” approach is not only faster but also a greener procedure
for finding novel leads.

References
Akhtar MHT, Mushtaq MY, Verpoorte R et al (2016) Zebrafish as a model for systems medicine
R&D: rethinking the metabolic effects of carrier solvents and culture buffers determined by 1H
NMR metabolomics. OMICS 20:42–52. https://doi.org/10.1089/omi.2015.0119
Alcântara BGV, Oliveira FP, Katchborian-Neto A et al (2021) Confirmation of
ethnopharmacological anti-inflammatory properties of Ocotea odorifera and determination of
its main active compounds. J Ethnopharmacol 264:113378. https://doi.org/10.1016/j.jep.2020.
113378
Ali K, Iqbal M, Yuliana ND et al (2013) Identification of bioactive metabolites against adenosine
A1 receptor using NMR-based metabolomics. Metabolomics 9:778–785. https://doi.org/10.
1007/s11306-013-0498-9
Allard PM, Peŕesse T, Bisson J et al (2016) Integration of molecular networking and in-silico
MS/MS fragmentation for natural products dereplication. Anal Chem 88:3317–3323. https://
doi.org/10.1021/acs.analchem.5b04804
Baerheim Svendsen A, Karlsen J (1967) Gaschromatographie von Monoterpenkohlenwasserstoffen
aus Aetherischen Oelen an gepackten Trennsaulen mit niedrigem Gehalt an fluessiger
stationaere Phase. Planta Med 15:1–5. https://doi.org/10.1055/s-0028-1099949
Banerjee P, Erehman J, Gohlke BO et al (2015) Super natural II – a database of natural products.
Nucleic Acids Res 43:D935–D939. https://doi.org/10.1093/nar/gku886
Bao JL, Ding RB, Jia XJ et al (2018) Fast identification of anticancer constituents in Forsythiae
Fructus based on metabolomics approaches. J Pharm Biomed Anal 154:312–320. https://doi.
org/10.1016/j.jpba.2018.03.020
Boezio B, Audouze K, Ducrot P et al (2017) Network-based approaches in pharmacology. Mol
Informatics 36:1700048. https://doi.org/10.1002/minf.201700048
Butler MS, Robertson AAB, Cooper MA (2014) Natural product and natural product derived drugs
in clinical trials. Nat Prod Rep 31:1612–1661. https://doi.org/10.1039/c4np00064a
Cardoso-Taketa AT, Pereda-Miranda R, Choi YH et al (2008) Metabolic profiling of the Mexican
anxiolytic and sedative plant Galphimia glauca using nuclear magnetic resonance spectroscopy
and multivariate data analysis. Planta Med 74:1295–1301. https://doi.org/10.1055/s-
2008-1074583
Chen L, Deng H, Cui H et al (2017) Inflammatory responses and inflammation-associated diseases
in organs. Oncotarget 9:7204–7218. https://doi.org/10.18632/oncotarget.23208
Cloarec O, Dumas ME, Craig A et al (2005) Statistical total correlation spectroscopy: an explor-
atory approach for latent biomarker identification from metabolic 1H NMR data sets. Anal Chem
77:1282–1289. https://doi.org/10.1021/ac048630x
Cragg GM, Newman DJ (2005) Plants as a source of anti-cancer agents. J Ethnopharmacol 100:72–
79. https://doi.org/10.1016/j.jep.2005.05.011
Cuperlovic-Culf M, Culf AS (2016) Applied metabolomics in drug discovery. Expert Opin Drug
Discovery 11(8):759–770. https://doi.org/10.1080/17460441.2016.1195365
De Donno A, Grassi T, Idolo A et al (2012) First-time comparison of the in vitro antimalarial
activity of Artemisia annua herbal tea and artemisinin. Trans R Soc Trop Med Hyg 106:696–
700. https://doi.org/10.1016/j.trstmh.2012.07.008
de Melo ICAR, de Souza ILL, Vasconcelos LHC et al (2020) Metabolomic fingerprinting of
Cissampelos sympodialis Eichler leaf extract and correlation with its spasmolytic activity. J
Ethnopharmacol 253:112678. https://doi.org/10.1016/j.jep.2020.112678
Natural Products Drug Discovery: On Silica or In-Silico? 139

Domingos OD, Alcântara BGV, Santos MFC et al (2019) Anti-inflammatory derivatives with dual
mechanism of action from the metabolomic screening of Poincianella pluviosa. Molecules
24:4375. https://doi.org/10.3390/molecules24234375
Donlan RM, Costerton JW (2002) Biofilms: survival mechanisms of clinically relevant
microorganisms. Clin Microb Rev 15:167–193. https://doi.org/10.1128/CMR.15.2.167-193.
2002
dos Santos FA, Sousa IP, Furtado NAJC et al (2018) Combined OPLS-DA and decision tree as a
strategy to identify antimicrobial biomarkers of volatile oils analyzed by gas chromatography–
mass spectrometry. Rev Bras Farm 28:647–653. https://doi.org/10.1016/j.bjp.2018.08.006
Emwas AH, Roy R, McKay RT, Tenori L, Saccenti E, Gowda GAN, Raftery D, Alahmari F,
Jaremko L, Jaremko M, Wishart DS (2019) NMR spectroscopy for metabolomics research.
Metabolites 9(7):123. https://doi.org/10.3390/metabo9070123
Gao JY, Zhao HY, Hylands PJ et al (2010) Secondary metabolite mapping identifies Scutellaria
inhibitors of human lung cancer cells. J Pharm Biomed Anal 53:723–728. https://doi.org/10.
1016/j.jpba.2010.04.019
Graziani V, Scognamiglio M, Belli V et al (2018) Metabolomic approach for a rapid identification
of natural products with cytotoxic activity against human colorectal cancer cells. Sci Rep 8:
5309. https://doi.org/10.1038/s41598-018-23704-9
Graziani V, Potenza N, D’Abrosca B et al (2021) NMR profiling of Ononis diffusa identifies
cytotoxic compounds against cetuximab-resistant colon cancer cell lines. Molecules 26:3266.
https://doi.org/10.3390/molecules26113266
Gromski PS, Muhamadali H, Di E et al (2015) A tutorial review: metabolomics and partial least
squares-discriminant analysis – a marriage of convenience or a shotgun wedding. Anal Chim
Acta 879:10–23. https://doi.org/10.1016/j.aca.2015.02.012
Hoerr V, Duggan GE, Zbytnuik L et al (2016) Characterization and prediction of the mechanism of
action of antibiotics through NMR metabolomics. BMC Microbiol 16:82. https://doi.org/10.
1186/s12866-016-0696-5
Hopkins AL (2007) Network pharmacology: network biology illuminates our understanding of
drug action. Nat Biotechnol 25:1110–1111. https://doi.org/10.1038/nbt1007-1110
Hopkins AL (2008) Network pharmacology: the next paradigm in drug discovery. Nat Chem Biol
4:682–690. https://doi.org/10.1038/nchembio.118
Ingkaninan K, de Best CM, van der Heijden R et al (2000) HPLC with on-line coupled UV, mass
spectrometric and biochemical detection for identification of acetylcholinesterase inhibitors
from natural products. J Chromatogr A 872:61–73. https://doi.org/10.1016/s0021-9673(99)
01292-3
Jeffryes JG, Colastani RL, Elbadawi-Sidhu M et al (2015) MINEs: open access databases of
computationally predicted enzyme promiscuity products for untargeted metabolomics. J
Chem Inform 7:44. https://doi.org/10.1186/s13321-015-0087-1
Jiang H, Hu C, Chen M (2021) The advantages of connectivity map applied in traditional Chinese
medicine. Front Pharmacol 12:474267. https://doi.org/10.3389/fphar.2021.474267
Kim GD, Lee JY, Auh JH (2019) Metabolomic screening of anti-inflammatory compounds from the
leaves of Actinidia arguta (Siebold & Zucc.) Planch. ex Miq. (Hardy Kiwi). Foods 8:47. https://
doi.org/10.3390/foods8020047
Klingelhofer I, Ngoc LP, van der Burg B et al (2021) A bioimaging system combining human
cultured reporter cells and planar chromatography to identify novel bioactive molecules. Anal
Chim Acta 1183:338956. https://doi.org/10.1016/j.aca.2021.338956
Leonti M, Verpoorte R (2017) Traditional mediterranean and European herbal medicines. J
Ethnopharmacol 199:161–167. https://doi.org/10.1016/j.jep.2017.01.052
Li S, Zhang B (2013) Traditional Chinese medicine network pharmacology: theory, methodology
and application. Chin J Nat Med 11:0110–0120. https://doi.org/10.1016/S1875-5364(13)
60037-0
140 H. K. Kim et al.

Lia ZY, Zhi HJ, Zhang FS et al (2013) Metabolomic profiling of the antitussive and expectorant
plant Tussilago farfara L. by nuclear magnetic resonance spectroscopy and multivariate data
analysis. J Pharm Biomed Anal 75:158–164. https://doi.org/10.1016/j.jpba.2012.11.023
Malone MH, Robichaud RC (1962) A Hippocratic screen for pure or crude drug materials. Lloydia
25:320–332
Maree J, Kamatou G, Gibbons S et al (2014) The application of GC–MS combined with
chemometrics for the identification of antimicrobial compounds from selected commercial
essential oils. Chemom Intel Lab Syst 130:172–181. https://doi.org/10.1016/j.chemolab.2013.
11.004
Morlock GE (2021) High-performance thin-layer chromatography combined with effect directed
assays and high-resolution mass spectrometry as an emerging hyphenated technology: a tutorial
review. Anal Chim Acta 1180:338644. https://doi.org/10.1016/j.aca.2021.338644
Morlock GE, Drotleff L, Brinkmann S (2021) Miniaturized all-in-one nanoGIT(+active) system for
on-surface metabolization, separation and effect imaging. Anal Chim Acta 1154:33830. https://
doi.org/10.1016/j.aca.2021.338307
Mushtaq MY, Verpoorte R, Kim HK (2013) Zebrafish as a model for systems biology. Biotechnol
Gen Engin Rev 29:187–205. https://doi.org/10.1080/02648725.2013.801238
Newman DJ, Cragg GM (2007) Natural products as sources of new drugs over the last 25 years. J
Nat Prod 70:461–477. https://doi.org/10.1021/np068054v
Newman DJ, Cragg GM (2020) Natural products as sources of new drugs over the nearly four
decades from 01/1981 to 09/2019. J Nat Prod 83:770–803. https://doi.org/10.1021/acs.jnatprod.
9b01285
Ntungwe NE, Dominguez-Martin EM, Roberto A et al (2020) Artemia species: an important tool to
screen general toxicity samples. Curr Pharm Design 26:2892–2908. https://doi.org/10.2174/
1381612826666200406083035
Ory L, Nazih EH, Daoud S et al (2019) Targeting bioactive compounds in natural extracts –
development of a comprehensive workflow combining chemical and biological data. Anal Chim
Acta 1070:29–42. https://doi.org/10.1016/j.aca.2019.04.038
Parng C, Seng WL, Semino C et al (2002) Zebrafish: a preclinical model for drug screening. Assay
Drug Dev Technol 1:41–48. https://doi.org/10.1089/154065802761001293
Paul A, de Boves Harrington P (2021) Chemometric applications in metabolomic studies using
chromatography-mass spectrometry. Trends Anal Chem 135:116165. https://doi.org/10.1016/j.
trac.2020.116165
Pimm SL, Russell GJ, Gittleman JL, Brooks TM (1995) The future of biodiversity. Science
269:347–350. https://doi.org/10.1126/science.269.5222.347
Raŕova L, Ncube B, Van Staden J et al (2019) Identification of narciclasine as an in vitro anti-
inflammatory component of Cyrtanthus contractus by correlation-based metabolomics. J Nat
Prod 82:1372–1376. https://doi.org/10.1021/acs.jnatprod.8b00973
Rhee IK, van Rijn RM, Verpoorte R (2003) Qualitative determination of false-positive effects in the
acetylcholinesterase assays using thin layer chromatography. Phytochem Anal 14:127–131.
https://doi.org/10.1002/pca.675
Salzer L, Witting M (2021) Quo Vadis Caenorhabditis elegans metabolomics-a review of current
methods and applications to explore metabolism in the nematode. Metabolites 11:284. https://
doi.org/10.3390/metabo11050284
Samuelsson G, Bohlin L (eds) (2009) Drugs of natural origin: a treatise of pharmacognosy. Swedish
Academy of Pharmaceutical Sciences, Stockholm
Schreiner T, Morlock GE (2021) Non-target bioanalytical eight-dimensional hyphenation including
bioassay, heart-cut trapping, online desalting, orthogonal separations and mass spectrometry. J
Chromatogr 1647:462154. https://doi.org/10.1016/j.chroma.2021.462154
Szymańska E, Saccenti E, Smilde AK et al (2012) Double-check: validation of diagnostic statistics
for PLS-DA models in metabolomics studies. Metabolomics 8:3–16. https://doi.org/10.1007/
s11306-011-0330-3
Natural Products Drug Discovery: On Silica or In-Silico? 141

Tang Y, Lou Z, Yang L et al (2015) Screening of antimicrobial compounds against Salmonellaty

phimurium from burdock (Arctium lappa) leaf based on metabolomics. Eur Food Res Technol
240:1203–1209. https://doi.org/10.1007/s00217-015-2423-0
Tawfike TAF, Romli M, Clements C et al (2019) Isolation of anticancer and anti-trypanosome
secondary metabolites from the endophytic fungus Aspergillus flocculus via bioactivity guided
isolation and MS based metabolomics. J Chromatogr B 1106–1107:71–83. https://doi.org/10.
1016/j.jchromb.2018.12.032
Verpoorte R, Kim HK, Choi YH (2018) Synergy: easier to say than to prove. Synergy 7:34–35.
https://doi.org/10.1016/j.synres.2018.10.004
Wang M, Lamers RJAN, Korthout HAAJ et al (2005) Metabolomics in the context of systems
biology: bridging traditional Chinese medicine and molecular pharmacology. Phytother Res 19:
173–182. https://doi.org/10.1002/ptr.1624
Westerhuis JA, Hoefsloot HCJ, Smit S et al (2008) Assessment of PLSDA cross validation.
Metabolomics 4:81–89. https://doi.org/10.1007/s11306-007-0099-6
Wishart DS (2008) Applications of metabolomics in drug discovery and development. Drugs R D
9:307–322. https://doi.org/10.2165/00126839-200809050-00002
Wu C, Du C, Gubbens J et al (2015a) Metabolomics-driven discovery of a prenylated isatin
antibiotic produced by Streptomyces species MBT28. J Nat Prod 78:2355–2363. https://doi.
org/10.1021/acs.jnatprod.5b00276
Wu C, Zacchetti B, Ram AFJ et al (2015b) Expanding the chemical space for natural products by
Aspergillus niger and Streptomyces coelicolor co-cultivation and biotransformation. Nat Sci
Rep 4:10868. https://doi.org/10.1038/srep/10868
Xu EY, Schaefer WH, Xu Q (2009) Metabolomics in pharmaceutical research and development:
metabolites, mechanisms and pathways. Curr Opin Drug Disc 12:40–52. https://doi.org/10.
1016/j.copbio.2015.04.004
Yang JY, Sanchez LM, Rath CM et al (2013) Molecular networking as a dereplication strategy. J
Nat Prod 76:1686–1699. https://doi.org/10.1021/np400413s
Ye H, Wei J, Tang KL et al (2016) Drug repositioning Ttrough network pharmacology. Curr Top
Med Chem 16:3646–3656. https://doi.org/10.2174/1568026616666160530181328
Yuliana ND, Khatib A, Choi YH et al (2011) Comprehensive extraction integrated with NMR
metabolomics: a new way of bioactivity screening methods for plants, adenosine A1 receptor
binding compounds in Orthosiphon stamineus Benth. Anal Chem 83:6902–6906. https://doi.
org/10.1021/ac201458n
Quantitative NMR Methods
in Metabolomics

G. A. Nagana Gowda and Daniel Raftery

Contents
1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 144
2 Quantitation Approaches Using NMR . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 146
2.1 Internal Reference Standards for Absolute Quantitation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 146
2.2 Alternative Reference Standards for Absolute Quantitation . . . . . . . . . . . . . . . . . . . . . . . . . 149
2.3 Quantitation of Metabolites Using Intact Samples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 149
2.3.1 Intact Serum and Plasma Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 150
2.3.2 Intact Urine Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 150
2.3.3 Intact Tissue Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 151
2.4 Metabolite Quantitation Using Processed Samples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 152
2.4.1 Analysis of Aqueous Metabolites . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 152
2.4.2 Analysis of Coenzymes and Antioxidants . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 154
2.4.3 Analysis of Lipids . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 156
2.5 Quantitation Methods Using Stable Isotope Labeling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 157
2.5.1 Isotope Labeling Focused on Metabolic Fluxes and Pathways . . . . . . . . . . . . . . 157
2.5.2 Isotope Labeling Focused on Metabolite Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . 157
3 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 158
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 159

G. A. Nagana Gowda (✉)

Northwest Metabolomics Research Center, University of Washington, Seattle, WA, USA
Mitochondria and Metabolism Center, Anesthesiology and Pain Medicine, University of
Washington, Seattle, WA, USA
e-mail: [email protected]

D. Raftery (✉)
Northwest Metabolomics Research Center, University of Washington, Seattle, WA, USA
Mitochondria and Metabolism Center, Anesthesiology and Pain Medicine, University of
Washington, Seattle, WA, USA
Fred Hutchinson Cancer Research Center, Seattle, WA, USA
e-mail: [email protected]

# The Author(s), under exclusive license to Springer Nature Switzerland AG 2022 143
V. Ghini et al. (eds.), Metabolomics and Its Impact on Health and Diseases,
Handbook of Experimental Pharmacology 277, https://doi.org/10.1007/164_2022_612
144 G. A. Nagana Gowda and D. Raftery

Abstract

Nuclear Magnetic Resonance (NMR) spectroscopy is one of the two major

analytical platforms in the field of metabolomics, the other being mass spectrom-
etry (MS). NMR is less sensitive than MS and hence it detects a relatively small
number of metabolites. However, NMR exhibits numerous unique characteristics
including its high reproducibility and non-destructive nature, its ability to identify
unknown metabolites definitively, and its capabilities to obtain absolute
concentrations of all detected metabolites, sometimes even without an internal
standard. These characteristics outweigh the relatively low sensitivity and resolu-
tion of NMR in metabolomics applications. Since biological mixtures are highly
complex, increased demand for new methods to improve detection, better identify
unknown metabolites, and provide more accurate quantitation continues
unabated. Technological and methodological advances to date have helped to
improve the resolution and sensitivity and detection of a larger number of
metabolite signals. Efforts focused on measuring unknown metabolite signals
have resulted in the identification and quantitation of an expanded pool of
metabolites including labile metabolites such as cellular redox coenzymes,
energy coenzymes, and antioxidants. This chapter describes quantitative NMR
methods in metabolomics with an emphasis on recent methodological
developments, while highlighting the benefits and challenges of NMR-based
metabolomics.

Keywords
Fast NMR methods · Isotope tagging · Metabolomics · Nuclear magnetic
resonance (NMR) · Quantitation

1 Introduction

The field of metabolomics represents the parallel analysis of large numbers of

metabolites in biological systems. Metabolites provide information on the instanta-
neous biological state of an organism or system along with the functions of upstream
cellular molecular species such as genes, transcripts, and proteins in health and
pathological conditions. Using a variety of advanced methodologies, comprehensive
analysis of metabolite data enables understanding biological phenotypes,
deciphering mechanisms, and identifying disease biomarkers or drug targets
(Raftery 2014; Nagana Gowda and Raftery 2019). Metabolomics applications span
a wide range of disciplines including human health and diseases, pharmacology,
drug development, toxicology, environment, plants, food, and nutrition. However, a
majority of the studies to date are focused on improving the mechanistic understand-
ing, prevention, early diagnosis, and management of human diseases (Kodama et al.
2020; Goldman et al. 2019; Johnson et al. 2016; Wishart 2016).
Quantitative NMR Methods in Metabolomics 145

Nuclear magnetic resonance (NMR) spectroscopy and mass spectrometry

(MS) are the two most widely used methods in the metabolomics field. MS typically
provides two to three orders of magnitude higher sensitivity than NMR and thereby
enables analysis of several hundreds to thousands of metabolites from a single
measurement. Generally, in MS analysis, metabolites from biological mixtures are
subjected to separation using methods such as liquid chromatography, gas chroma-
tography, or electrophoresis prior to detection. Separation using liquid chromatog-
raphy, however, is the most popular and nearly 80% of the metabolomics methods
use liquid chromatography resolved MS method (Edison et al. 2021). Absolute
quantitation of metabolites in MS involves using internal or external standards,
ideally, for each metabolite. However, finding isotopically labeled internal standards
for each metabolite is challenging and hence, one standard that represents a class of
metabolites is often used (Djukovic et al. 2020). This approach, however, can result
in a loss of accuracy. In contrast, and as will be described below, NMR provides
several approaches for accurate quantitation.
Although NMR spectroscopy is less sensitive than MS, it exhibits numerous
unique and favorable characteristics that are beneficial to the field of metabolomics
(Edison et al. 2021; Wishart 2019; Nagana Gowda and Raftery 2014a, 2015, 2017a,
2019). Notably: (1) NMR is highly reproducible and has excellent linearity (Mo and
Raftery 2008); (2) NMR provides absolute quantitation of all metabolites in the
spectrum using a single internal standard or even without the need for an internal
standard; (3) it provides the gold standard approach in establishing the identity of
unknown metabolites; (4) it enables the analysis of intact biofluid and tissue samples
with little to no need for sample preprocessing; (5) it is non-destructive, which means
the sample remains intact after the analysis and can be reused for analysis using
NMR or using other methods such as MS; (6) it enables tracing of metabolic
pathways and measuring metabolic fluxes utilizing stable isotope-labeled precursors;
(7) using NMR, the same metabolites can be detected through one or more types of
atomic nuclei such as 1H, 13C, 31P, or 15N, which provides flexibility to measure
metabolite levels; (8) NMR’s ability to detect essentially all molecular species with a
given nucleus makes it extremely useful for following methods development; and
(9) NMR offers new avenues to measure unstable metabolites that are fundamental
to cellular functions. Such characteristics far outweigh the poor sensitivity and
resolution of NMR and have been exploited extensively in the metabolomics field.
Human blood serum/plasma, urine, and tissue continue to be the most widely
used biological specimens in the metabolomics field. However, other biological
specimens including saliva (Lohavanichbutr et al. 2018), cerebrospinal fluid
(Albrecht et al. 2020), gut aspirate (Bala et al. 2006), bile (Nagana Gowda 2011),
amniotic fluid (Orczyk-Pawilowicz et al. 2016), synovial fluid (Anderson et al.
2020), fecal samples (Zierer et al. 2018), exhaled breath condensate (Maniscalco
et al. 2020), tear (Yazdani et al. 2019), and sperm-seminal fluid (Engel et al. 2019)
have also been analyzed. In addition, specimens from animal models, cell lines,
yeast (Airoldi et al. 2015), bacteria (Lussu et al. 2017), tumor cells (Lane et al. 2017),
tumor spheroids (Kalfe et al. 2015), exosomes (Zebrowska et al. 2019), and isolated
mitochondria (Xu et al. 2018) have been used.
146 G. A. Nagana Gowda and D. Raftery

The key steps involved in almost all metabolomics investigations include metab-
olite detection, unknown peak identification, and quantification. Relative or absolute
concentrations of metabolites thus obtained are then subjected to statistical and/or
metabolic pathway analysis focused on a wide variety of applications in the areas of
basic and medical sciences. Typically, metabolite data are analyzed using univariate
and multivariate statistical analysis focused on the discovery and validation of
putative metabolite biomarkers. Alternatively, metabolite levels or isotope labeled
metabolites are used for identifying the perturbed metabolic pathways, which pro-
vide mechanistic understanding of cellular functions including information on drug
targets for therapeutic development.

2 Quantitation Approaches Using NMR

The quantitative ability of NMR makes it an important platform complementary to

MS in metabolomics. NMR can be used for quantitative analysis of metabolites in
intact samples, extracted samples, live organisms, cells, or subcellular organelles
such as mitochondria. In NMR, generally, metabolite peaks are identified prior to
their relative or absolute quantitation. The identities of metabolites are established
using databases of standard compounds, the comprehensive analysis of 1D and 2D
NMR spectra, and/or spiking with authentic compounds.
Quantitation generally involves either (a) relative quantitation, in which metabo-
lite levels are measured relative to one another; or (b) absolute quantitation, in which
molar concentrations of metabolites are determined using an internal or external
standard. Currently, relative quantitation is the most widely used approach owing to
its ease of use combined with challenges associated with absolute quantitation,
especially for some sample types, such as cells, tissue, fecal samples, etc. However,
absolute quantitation promises a number of benefits. Importantly, it provides a basic
platform of metabolite levels for a specific type of biological specimen. This is
important for assessment of data quality, such as to compare samples across different
geographical regions, different batches or analysis times, or perhaps most impor-
tantly to compare to known values, such as clinical ranges for blood or urine
metabolites. Considering the increased interest for absolute quantitation, there
have been numerous efforts in recent years focused on establishing reference
standards for absolute quantitation using NMR as described in the following section.

2.1 Internal Reference Standards for Absolute Quantitation

Many compounds (>25) have been evaluated as potential internal standards for
applications in numerous areas including organic chemistry, natural product chem-
istry, agriculture, drug discovery, and pharmaceuticals (Maniara et al. 1998;
Holzgrabe 2010; Pauli et al. 2012; Rundlöf et al. 2010; Salem and Mossa 2012).
These compounds exhibit favorable physical characteristics, such as unique chemi-
cal shift, purity, stability, solubility, and suitability for accurate gravimetry.
Quantitative NMR Methods in Metabolomics 147

However, most of these are not suitable for metabolomics due to aqueous solubility
concerns or chemical shift overlap. Chemical shift reference compounds such as TSP
(trimethylsilylpropionic acid) and DSS (trimethylsilylpropanesulfonic acid) have
been used as internal standards for absolute quantitation of metabolites. It was
realized some years ago, however, that these compounds are unsuitable for quantita-
tion owing to their peak suppression arising from the interaction with proteins. One
alternative, formic acid, was evaluated as an alternative to TSP for quantitation of
metabolites in intact serum many years ago (Kriat et al. 1992). However, formic acid
is unsuitable as a reference since it is an endogenous metabolite; the endogenous
concentration in serum varies significantly from person to person (~40 to 350 μM)
and hence, it interferes with externally added formic acid (Kubáň and Boček 2013;
d'Alessandro et al. 1994; Kapur et al. 2007). In another study, DSA (4,4-dimethyl-4-
silapentane-1-ammonium trifluoroacetate), which is a derivative of DSS, was
evaluated as a potential internal standard using intact rat plasma (Alum et al.
2008). However, it is also unsuitable as a reliable internal standard since several
factors including the increased line broadening by a factor of >2 at pH 7.4 relative to
pH 3.0 indicate that DSA interacts with sample matrix. One remedy for analysis of
metabolites in samples such as blood serum/plasma that contain copious
macromolecules is to remove the macromolecules effectively by ultrafiltration; in
such a case, TSP or DSS can still be used as standards for absolute quantitation
(Psychogios et al. 2011; Barding et al. 2012; Simón-Manso et al. 2013). The
challenge with ultrafiltration, however, is that it attenuates many metabolite peaks
(Nagana Gowda and Raftery 2014b), requires larger sample volumes, and is cum-
bersome for large-scale studies. In addition, ultrafiltration cannot be used for analysis
of samples such as tissue and whole blood. Ultrafiltration is also incompatible with
MS analysis, the other major analytical platform used in metabolomics, since MS
analysis invariably employs protein precipitation to remove macromolecules, prior
to analysis (Nagana Gowda et al. 2018).
Protein precipitation that removes macromolecules from samples provides an
alternative approach to quantitate metabolites and is well suited for large-scale
studies. However, even in such samples, peaks from the traditional internal
standards, TSP and DSS, are attenuated by up to 35% and hence they are unsuitable
as internal standards. More recently, two compounds, maleic acid and fumaric acid,
were evaluated for their utility as potential internal standards for quantitation of
metabolites since both provide a single peak in NMR spectrum and their peaks do
not overlap with peaks from bio-specimen spectra (Nagana Gowda et al. 2021)
(Fig. 1). It was shown that fumaric acid is a robust standard for protein precipitated
serum, plasma, and whole blood; and maleic acid is suitable for plasma and serum,
but it overlaps with coenzyme peaks in whole blood samples. These findings provide
new opportunities for improved and accurate quantitation of metabolites in human
plasma, serum, and whole blood using NMR spectroscopy. The potential utility of
maleic acid and fumaric acid as internal standards may be extended to other
biological specimens, as long as they do not overlap with bio-specimen peaks.
148 G. A. Nagana Gowda and D. Raftery

TSP

Maleic acid

Fumaric acid

Serum

Plasma

Whole Blood

Fig. 1 Typical 800 MHz 1H NMR spectra of a protein precipitated sample of human (a) whole
blood, (b) plasma, and (c) serum, solubilized in D2O buffer containing a mixture of three internal
standards (TSP, 238 μM; maleic acid, 350 μM; fumaric acid, 293 μM). Each spectrum is overlaid
with a spectrum from the blank buffer consisting of the same three standards (spectrum shown in
red) to enable the visualization of peak heights for the three internal standards; the spectra of the
bio-specimens are slightly right shifted relative to the blank spectrum for clarity. Peaks from the
blank are marked with asterisks. Heights for the fumaric acid peaks from the bio-specimen
and blank are approximately matched; however, a significant attenuation of the TSP peak in all
three bio-specimens spectra is noticeable. TSP: Trimethylsilylpropionic acid-d4. (reproduced with
permission from Nagana Gowda et al. 2021)
Quantitative NMR Methods in Metabolomics 149

2.2 Alternative Reference Standards for Absolute Quantitation

An altogether different approach is to determine metabolite concentrations without

the need for an internal standard. One such method is ERETIC (electronic reference
to access in vivo concentrations) (Akoka et al. 1999). In the ERETIC method, a
synthetic signal is generated in NMR spectra with the desired peak intensity, line
width, and chemical shift. This peak is then calibrated and used as a reference for
quantitation. A drawback of this method is that the quantitation error can be large
when NMR peaks are attenuated due to altered pulse widths arising, for example,
from lossy biological solutions. More recently, a method known as PULCON (pulse
length based concentration determination) alleviates the limitation of ERETIC and
provides a more robust approach to quantitation without the need for an internal
standard (Wider and Dreier 2006). PULCON, also known as ERETIC 2, works
based on the principle of reciprocity (Hoult and Richards 1976; Hoult 2000; Van der
Klink 2001) and allows the correlation of signal strength from a reference spectrum
with the spectrum of interest. This method shows immense promise for
metabolomics applications (Jiménez et al. 2018; Goldoni et al. 2016). The method,
however, requires that reference and test spectra are obtained using the same probe,
probe tuning and matching for the test samples should be identical to the reference
sample, and the same RF power needs to be delivered to the coil for each NMR
spectrum. Poor probe matching for salty test samples, for example, will lead to
inaccurate quantitation of metabolites. Despite this limitation, however, by using
standardized operating procedures quantitative data can be obtained that can poten-
tially enable sharing of inter-laboratory results.
Another approach for quantitation is to use the solvent signal as a concentration
reference (Mo and Raftery 2008). Most solvents can be observed by NMR and
solvent concentrations can be readily determined independently. In particular, a
widely used solvent such as water can serve as a primary concentration standard
for metabolite quantitation. The potential problems of radiation damping associated
with a strong NMR signal can be alleviated by small pulse angle excitation. The fact
that the solvent signal can be detected by the NMR receiver with the same efficiency
as analytes enables their accurate quantitation. It is shown by this approach that
analyte concentration can be accurately determined from 4 μM to more than 100 M.

2.3 Quantitation of Metabolites Using Intact Samples

The ability to analyze intact samples with no need for sample preparation or
separation using NMR is an important characteristic that continues to drive
NMR-based metabolomics. Initially, widely used bio-specimens including blood
serum and plasma were used only in their intact forms and this approach continues to
be widely used. In the following sections, analyses of intact bio-specimens that are
most widely used such as serum/plasma, urine, and tissue are described. The
methods presented here are also applicable for other specimen types.
150 G. A. Nagana Gowda and D. Raftery

2.3.1 Intact Serum and Plasma Analysis

Analysis of intact serum or plasma enables quantitation of aqueous metabolites as
well as lipids and various classes of lipoproteins in serum and plasma (Würtz et al.
2017). Two widely used, one-dimensional (1D) NMR pulse techniques are 1D
NOESY (nuclear Overhauser enhancement spectroscopy) and CPMG (Carr-
Purcell-Meiboom-Gill) with water signal suppression (often using presaturation)
(Nicholson et al. 1995). The 1D NOESY detects both small molecules such as
metabolites and macromolecules such as lipids and lipoproteins. On the other
hand, the CPMG experiment detects only small molecules; the macromolecule
signals from proteins and lipoproteins are suppressed based on a T2 (transverse
relaxation) filter (Beckonert et al. 2007); metabolites exhibit longer T2 relaxation
times compared to macromolecules and hence they are selectively retained in the
CPMG spectra. Numerous large-scale epidemiological studies have demonstrated
quantitation of 50–70 metabolite peaks and over 200 metabolic measures (which
include ratios of metabolite peaks) on a routine basis (Soininen et al. 2015; Würtz
et al. 2017). As described above, recent advances in NMR enable absolute quantita-
tion using an external reference, with no need for an internal standard (Wider and
Dreier 2006). This is remarkable considering that internal standards largely cannot
be used for absolute quantitation since they interact with copious proteins present in
the samples. However, a challenge for reliable analysis of metabolites in intact
samples is that metabolite binding to proteins causes signal attenuation (Nicholson
and Gartland 1989; Chatham and Forder 1999; Bell et al. 1988; Nagana Gowda and
Raftery 2014b). Moreover, exchange of metabolites between free and protein bound
forms results in broader NMR peaks. Further, residual macromolecule signals cause
distorted spectral baseline in CPMG spectra, which together adversely affect metab-
olite quantitation.

2.3.2 Intact Urine Analysis

Urine provides a rich source of information as it contains a significantly higher
number of detectable metabolites, compared to serum/plasma, and with a vast
concentration range (~106). In addition, urine has a relatively low concentration of
proteins and hence macromolecular interference is minimal for metabolite analysis.
A step-by-step procedure for NMR analysis of urine is provided as a guide for
routine applications (Beckonert et al. 2007; Emwas et al. 2016). The pH of normal
human urine varies widely, from approximately 5 to 8 (Hernandez et al. 2001;
Rylander et al. 2006; Welch et al. 2008) and the salt concentration also varies
significantly from sample to sample. Such pH and salt concentration variations
alter chemical shifts of many peaks in the urine NMR spectra. Such peak shifts are
significant for metabolites with functional groups with pKa’s near the physiological
pH. This causes a challenge for peak identification, comparison of different spectra,
and quantitation of metabolites. Therefore, urine samples are generally mixed with
buffer solution typically in a 1:1 (v/v) ratio (at pH = 7.4). Using Chenomx software
and authentic compound spiking, >200 metabolites in urine have been identified
(Bouatra et al. 2013). However, considering the high complexity of the urine NMR
spectrum and the sensitivity of chemical shifts to factors such as pH and salt
Quantitative NMR Methods in Metabolomics 151

concentration, the number of metabolites that can be analyzed on a routine basis is

restricted to ~60 to 70. Factors such as diet, medications, physical activity, smoking,
gender, age, gut microbe diversity greatly affect the metabolome and they should be
carefully accounted for disease biomarker identification (Emwas et al. 2015, 2016).
Large-scale (>1,000 samples) high-throughput studies now enable quantitative
analysis of urinary metabolites using automated or semi-automated regression-
based spectral analysis (Tynkkynen et al. 2019). In a large and impressive study, it
was shown that prediction of metabolite concentrations, including many invisible
inorganic ions, could be made based on the interrelationships between chemical
shifts and concentrations, for automated urine analysis (Takis et al. 2017). Such
advances promise new applications to areas including clinical, epidemiological, and
pharmaceutical research.

2.3.3 Intact Tissue Analysis

NMR spectra of intact tissue are obtained using high-resolution magic angle
spinning (HR-MAS) techniques (Tilgner et al. 2019). HR-MAS provides highly
resolved spectra, which are comparable to those of bio-fluids. Tissue specimens
typically collected from a surgical procedure or biopsy are often snap-frozen and
stored for later analyses. The use of fresh samples for direct analysis, however, is
advantageous for sensitive and structurally delicate biopsy samples. Resected or
biopsied tissue is washed by quickly rinsing, typically with D2O, to remove any
blood contamination prior to freezing or direct analysis. The use of fresh samples
avoids any deleterious effects caused by the freeze/thaw process and protects tissue
integrity. Care should, however, be exercised to ensure fresh samples specifically
from biopsy are kept under cold and humid conditions until the analyses are
performed to retain the integrity of the metabolite profiles and reduce the possibility
of metabolic changes. The ability to recover tissue after NMR analysis provides an
opportunity to use the same specimens for other studies such as proteomic and
genomic analysis or even histology. Advances in probe technologies with a 2H field-
frequency lock channel and a magnetic field gradient coil offer spectral stability and
resolution sufficient for routine metabolomics studies of tissue samples as small as a
few ng (Wong et al. 2012). Such capabilities, combined with minimal sample
preparation and fast data acquisition, promise to extend the application of metabolic
profiling of biopsied tissue to clinical applications. As examples, studies have shown
that HR-MAS NMR of core needle biopsy tissue can predict breast tumor
aggressiveness prior to surgery (Choi et al. 2012). Tissue metabolite profiles offer
numerous benefits owing to the close association of tissue with disease pathologies.
For example, alteration in tissue metabolite profiles has been shown to differentiate
breast cancer tumors from normal tissue (Paul et al. 2018; Sitter et al. 2010).
Importantly, HR-MAS NMR potentially enables diagnosis, prognosis, and staging
of cancers (Dinges et al. 2019; Chen et al. 2017).
152 G. A. Nagana Gowda and D. Raftery

2.4 Metabolite Quantitation Using Processed Samples

Sample processing involves separation of metabolites from the macromolecular

matrix. Such an approach enables detection of a significantly expanded pool of
metabolites. A number of sample processing methods exist, to date, focused on
analysis of aqueous metabolites or lipids or both. Further, new methods are being
continuously developed that focus on improving the extraction efficiency, preserv-
ing metabolite integrity and simplifying the extraction process.

2.4.1 Analysis of Aqueous Metabolites

Methods such as ultrafiltration, solid phase extraction, and protein precipitation
using organic solvents such as methanol, acetonitrile, acetone, perchloric acid, or
trichloroacetic acid have been explored for many years to extract metabolites
(Wevers et al. 1994; Daykin et al. 2002; Tiziani et al. 2008; Fan 2012). Among
them, ultrafiltration using low molecular weight (~3 kDa) cut-off filter removes
proteins most effectively. Using this method, nearly 50 aqueous metabolites could
be identified and quantified (Psychogios et al. 2011). However, ultrafiltration
attenuates many metabolite peaks (Nagana Gowda and Raftery 2014b), requires
larger sample volumes, and is particularly cumbersome for large-scale studies.
Nearly half of the detected metabolites in ultrafiltered serum exhibited lower
concentrations ranging from nearly 10 to 75% (Nagana Gowda and Raftery
2014b). Further, ultrafiltration is incompatible with analysis of samples such as
whole blood and tissue as well as with analysis using MS, which generally employs
protein precipitation using organic solvents (Nagana Gowda et al. 2018).
Detailed studies have focused on increasing the number of detected metabolites,
identifying unknown metabolites and optimizing their quantitation in blood serum
and plasma (Nagana Gowda and Raftery 2014b; Nagana Gowda et al. 2015). These
studies have shown that protein precipitation using methanol in a 2:1 ratio (v/v) with
the sample offers an optimal approach for analysis of aqueous metabolites in blood
serum/plasma. The use of acetonitrile for protein precipitation, on the other hand,
revealed a surprisingly poor performance; one-third of the detected metabolites were
attenuated by up to 70% compared to methanol precipitation at the same solvent to
serum ratio of 2:1 (v/v) (Fig. 2). A further attenuation of nearly two-third of the
metabolites was observed for an acetonitrile to serum ratio of 4:1 (v/v). As the
analysis of metabolites using MS invariably employs protein precipitation prior to
analysis, methods developed for NMR analysis also help analysis using MS. The
performance of sample processing for MS analysis is typically evaluated using the
total number of ions detected, which is problematic (Ivanisevic et al. 2013) and is an
inaccurate approach as far as quantitation is concerned.
Protein precipitation, however, does not remove macromolecules completely and
the residual macromolecules (~2%) are water-soluble, which cause broad baselines
in NMR spectra when obtained using the one-pulse or 1D NOESY pulse sequence
(Nagana Gowda et al. 2015, 2021). The use of the CPMG sequence helps to suppress
signals from these residual proteins (~2%) and provides a flat baseline. The CPMG
sequence, however, causes a small attenuation for many signals due to differential T2
Quantitative NMR Methods in Metabolomics 153

22 22
2:1 MeOH:Serum 2:1 ACN:Serum
20 20
18
3:1 MeOH:Serum 18 3:1 ACN:Serum
16 4:1 MeOH:Serum 16 4:1 ACN:Serum
14 14
(a) (e)
12 12
10 10
8 8
6 6
4 4
2 2
0 0

100 100
90 90
80 (b) 80 (f
70 70
60 60
50 50
40 40
30 30
20 20
10 10
0 0

450 450
400 400
350 (c) 350 (g)
300 300
250 250
200 200
150 150
100 100
50 50
0 0

4500 (d) (h)

4500
4000 4000
3500 3500
3000 3000
2500 2500
2000 2000
1500 1500
1000 1000
500 500
0 0

Fig. 2 Comparison of absolute concentrations (in μM) of metabolites detected in pooled human
blood serum and quantitated using 800 MHz NMR spectroscopy after protein precipitation using
methanol (MeOH) (a, b, c and d) or acetonitrile (ACN) (e, f, g and h) at a solvent to serum ratios of
2:1, 3:1, and 4:1. Methanol performs most optimally over a wide range and a methanol to serum
ratio of 2:1 provides the best performance (reproduced with permission from Nagana Gowda et al.
2015)
154 G. A. Nagana Gowda and D. Raftery

relaxation rates; for example, an evaluation of 20 metabolite peaks revealed an

average of ~6% attenuation for plasma and serum when a 32 ms CPMG 180o echo
pulse train was used. The peak attenuation increased with increasing duration of the
echo pulse train and it exceeded 10% for a 256 ms echo pulse train. Hence, for
accurate quantitation, signal attenuation due to T2 relaxation in the CPMG spectra
should be carefully accounted. Potential alternatives to the CPMG sequence, includ-
ing the use of stimulated echo (STE) pulse sequence (Lucas et al. 2005) have proved
unsuitable for metabolomics applications.

2.4.2 Analysis of Coenzymes and Antioxidants

Coenzymes, including coenzyme A (CoA), acetyl coenzyme A (acetyl-CoA),
coenzymes of redox reactions and energy, and antioxidants mediate biochemical
reactions fundamental to the functioning of all living cells. The most common redox
coenzymes include NAD+ (oxidized nicotinamide adenine dinucleotide), NADH
(reduced nicotinamide adenine dinucleotide), NADP+ (oxidized nicotinamide ade-
nine dinucleotide phosphate), and NADPH (reduced nicotinamide adenine dinucle-
otide phosphate). The coenzymes of energy include ATP (adenosine triphosphate),
ADP (adenosine diphosphate), and AMP (adenosine monophosphate). Major
antioxidants include GSSG (oxidized glutathione) and GSH (reduced glutathione).
Conventional enzymatic assays are suboptimal, as separate protocols are needed for
analysis of each coenzyme or their ratios. The interference from sample matrix and
the finite linear range of these assays further add to the challenges. Although MS is
extensively used, ion suppression, interference due to the unit mass difference in
targeted analysis, and in-source fragmentation pose challenges for reliable coenzyme
analysis (Evans et al. 2010; Trammell and Brennera 2013). Hence, the ability to
analyze these coenzymes in one-step using NMR represents an important advance-
ment in the metabolomics field. A major challenge unconnected with any analytical
method, however, is the notoriously unstable nature of these compounds. Enzyme
activity and oxidation affect their levels, deleteriously. Somewhat recently, sample
harvesting, processing, and analysis conditions were optimized for heart tissue from
mouse models and showed that a simple NMR experiment can simultaneously
measure NAD+, NADH, NADP+, NADPH, ATP, ADP, and AMP in one step
apart from other metabolites (Nagana Gowda et al. 2016, 2018). Later, the scope
of NMR was extended to the analysis of CoA, acetyl-CoA, and antioxidants (GSH,
GSSG) along with a large pool of other metabolites and coenzymes, in one step
(Nagana Gowda et al. 2019) (Fig. 3). Further, as an important alternative to serum/
plasma metabolomics, it was shown that using whole blood, the coenzymes and
antioxidants can be measured simultaneously in addition to the nearly 70 metabolites
that can be quantitated in serum/plasma with essentially no additional effort (Nagana
Gowda and Raftery 2017b). The analysis protocols and the annotated characteristic
fingerprints for these newly identified coenzymes and other metabolites are provided
for easy identification and absolute quantification using a single internal reference.
The ability to measure the unstable but ubiquitous coenzymes fundamental to
cellular functions, simultaneously and reliably, offers a new avenue to investigate
the mechanistic details of cellular function in health and diseases.
Quantitative NMR Methods in Metabolomics 155

(c)
NAD+ Acetyl CoA (b)
ATP CoA
NADH CoA-S-S-G

AMP

Acetyl-CoA NADPH
NADP+
CoA ADP

Gluta-
mate
Glutamine

Ace-tate
TSP
α-Glucose
Aromatic amino acids

β-Glucose

Lactate

BCCA
Fumarate

A
(a)

(e)
GSH (d)
Cr

GSSG
PCr

Fig. 3 (a) Typical 800 MHz 1H NMR spectrum of a mouse heart tissue extract with labeling of
some of the metabolites: BCCA: branched chain amino acids; TSP: reference peak; (b–e) expanded
spectral regions highlighting characteristic peaks for (b) coenzyme A (CoA), acetyl coenzyme A
(acetyl-CoA), and coenzyme A glutathione disulfide (CoA-S-S-G); (c) CoA, acetyl-CoA, oxidized
nicotinamide adenine dinucleotide (NAD+), oxidized nicotinamide adenine dinucleotide phosphate
(NADP+), reduced nicotinamide adenine dinucleotide (NADH), reduced nicotinamide adenine
dinucleotide phosphate (NADPH), adenosine triphosphate (ATP), adenosine diphosphate (ADP),
and adenosine monophosphate (AMP); (d) reduced glutathione (GSH) and oxidized glutathione
(GSSG); and (e) creatine (Cr) and phosphocreatine (PCr) (reproduced with permission from Nagana
Gowda et al. 2019)
156 G. A. Nagana Gowda and D. Raftery

2.4.3 Analysis of Lipids

NMR spectroscopy is widely used for analysis of lipids and lipoprotein particles in
serum and plasma (Mallol et al. 2013). Identification and quantitation of lipoprotein
particles by NMR exploits the characteristic chemical shifts of the methyl resonances
of fatty acid chains of lipids from different particle sizes, with peaks from smaller
particles appearing at lower frequencies. Methodologies used to characterize lipo-
protein particles based on methyl resonances utilize either deconvolution (Jeyarajah
et al. 2006, Kaess et al. 2008) or statistical (Soininen et al. 2009) methods. These
methods have enabled determination of particle size and number for lipoprotein
classes such as VLDL (very low-density lipoprotein), LDL (low-density lipopro-
tein), and HDL (high-density lipoprotein) and up to 14 (or more) lipoprotein
subclasses. The ability to quantitate a variety of lipoprotein particles using NMR
has opened avenues for clinical assessment and management of cardiovascular
disease risk. In view of the fact that such lipoprotein classification and
sub-classification using NMR is superior to the conventional methods, the method
has been commercialized to manage the risk of heart diseases. Somewhat recently, a
diffusion-based method was proposed to characterizing lipoprotein particles (Mallol
et al. 2015). Here, two-dimensional diffusion-ordered 1H NMR spectroscopy
(DOSY) was used to measure diffusion coefficients, which provide information on
the particle sizes of lipoproteins (Johnson 1999). The lipoprotein particle numbers
are then calculated by dividing the peak volume by the size of lipoprotein particles.
The ability to directly calculate lipoprotein sizes using the DOSY method was
purported (Mallol et al. 2015) to provide a more accurate results for the particle
numbers than the commercialized methods, which are based on 1D NMR.
After extraction, typically using a mixture of organic solvents, the analysis of
tissue or blood samples provides quantitative information on individual lipids or
lipid classes. The Folch extraction, consisting of chloroform/methanol/water in a
volumetric ratio of 8:4:3 (v/v/v) is one of the earliest and most popular methods
(Folch et al. 1957). Since then, numerous different lipid extraction protocols with
modification to Folch et al. (1957) or Bligh and Dyer method (Bligh and Dyer 1959)
have been proposed for biological specimens such as blood, tissue, and cells. A more
recent method, involving butanol-methanol (BUME), eliminates the need for chlo-
roform, which is an hazardous solvent (Löfgren et al. 2012; Cruz et al. 2016). More
recently, the BUME method was modified to suit the analysis of lipids using NMR
spectroscopy (Barrilero et al. 2018). This method replaces heptane with diisopropyl
ether as the organic solvent, since peaks from the residual heptane overlap with lipid
signals. Notably, this method has enabled identification and quantitation of 15 differ-
ent lipid classes including fatty acids, triglycerides, phospholipids, and cholesterols
in serum. A semiautomatic software, LipSpin, converts raw NMR data based on
mathematical and reference spectral models and provides quantitative information
on lipids (Barrilero et al. 2018). Detailed protocols for extraction and quantitative
analysis of lipids in biological specimens such as serum, tissue, and cells are
provided, which serve as a practical guide for beginners in the field (Gil et al. 2019).
Quantitative NMR Methods in Metabolomics 157

2.5 Quantitation Methods Using Stable Isotope Labeling

Stable isotope incorporation in vivo or ex vivo offers opportunities to quantitate

metabolites using NMR with improved resolution and sensitivity. In vivo analysis of
metabolites in live systems enables monitoring of dynamic changes, measuring
fluxes and monitoring metabolism in real time. The use of heteronuclear 2D
(two-dimensional) NMR pulse techniques involving stable isotopes offers a combi-
nation of selectivity, sensitivity, and resolution and alleviates major challenges in
NMR experiments involving nuclei with low natural abundance. To date, stable
isotopes including 13C, 15N, 2H, and/or 31P have been employed for analysis of
metabolites in biological mixtures and investigation of metabolic pathways.

2.5.1 Isotope Labeling Focused on Metabolic Fluxes and Pathways

Isotope labeling in vivo enables measurement of fluxes and tracing of metabolic
pathways. Using this approach, the same metabolite that flows through multiple
pathways can be distinguished. A growing number of pathways, including glycoly-
sis, pentose phosphate pathway, glutaminolysis, fatty acid oxidation, and TCA cycle
can be investigated using the combination of NMR and selective or uniformly
isotope labeled substrates such as 13C-glucose and 13C/15N-glutamine (Lin et al.
2019). Quantitative analysis of in vitro or in vivo isotope labeled metabolites can be
measured either ex vivo, after extraction of metabolites, or in live systems in vitro or
in vivo. While analysis after extraction provides a snapshot of metabolite levels at a
particular time point, in situ analysis using live systems enables the measurement of
the dynamic changes in metabolite levels and monitoring of metabolism in real time.
Analysis after extraction of metabolites has been widely used in the metabolomics
field. However, the growing technological and methodological advances in NMR are
witnessing an increasing number of in vitro or in vivo investigations using live
systems such as C. elegans, cells, and isolated mitochondria (Nguyen et al. 2020;
Wen et al. 2015; Xu et al. 2018). Isotope labeled studies using cells and subcellular
organelles enable understanding of metabolic pathways under controlled conditions.
And the use of organisms, animal models, or humans can translate the findings from
studies of cells and subcellular organelles to investigate the pathogenesis of human
diseases (Fan et al. 2009; Locasale et al. 2011; Lane et al. 2011).

2.5.2 Isotope Labeling Focused on Metabolite Analysis

Isotope labeling in vivo in plants and organisms such as bacteria and yeast offers
significant enhancement to spectral resolution and the detection sensitivity (Zhang
et al. 2012; Chikayama et al. 2008; Bingol et al. 2012, 2013). In particular, it
alleviates the challenges invariably met with the analysis involving low natural
abundance heteronuclei and enables analysis of a large number of metabolites
using conventional high-resolution 2D NMR experiments such as HSQC and
HMBC. The uniform labeling using nuclei such as 13C also enables characterization
of metabolites based on homonuclear 2D 13C NMR experiments. Carbon-bond
topology networks obtainable from such homonuclear 2D 13C experiments provide
158 G. A. Nagana Gowda and D. Raftery

additional avenues for metabolite identification (Chikayama et al. 2008; Bingol et al.
2012).
An altogether different approach is to label different classes of metabolites based
on the specific functional group (Shanaiah et al. 2007; Desilva et al. 2009; Ye et al.
2009). Chemical derivatization of metabolites using a substrate that contains isotope
such as 13C, 15N, or 31P offers both sensitivity and resolution enhancement, owing to
the high isotopic abundance and wide chemical shift dispersion imparted by the
incorporated isotope. The 1H decoupled 1D or 2D NMR spectrum involving the
isotope labeled heteronuclei provides a single peak for each metabolite, which
further adds to the sensitivity and resolution. Metabolite classes including amines,
carboxylic acids, and hydroxyls have thus been tagged with isotopes and analyzed
using 1D or 2D NMR (Shanaiah et al. 2007; Desilva et al. 2009; Ye et al. 2009;
Vicente-Muñoz et al. 2021). Owing to its high natural abundance, 31P signals from
metabolites, however, can show up as strong background peaks in the 31P enriched
experiments, unlike the other nuclei. Incorporation of a “smart isotope tag” such as
15
N-cholamine enables analysis of the carboxylic acid class of metabolites using
both NMR and MS methods (Tayyari et al. 2013). The smart isotope tag possesses
an NMR sensitive isotope (15N) that offers good chemical shift dispersion and a
permanent positive charge that improves MS sensitivity and enables quantitation of
metabolites more accurately by both NMR and MS. Such analysis allows direct
comparison of NMR and MS data, which is an important characteristic for biomarker
discovery and biological interpretation in the metabolomics field.

3 Conclusion

The ability to identify unknown metabolites, absolute quantitation and analysis of

intact bio-specimens including live cells and subcellular organelles, is expanding the
application of NMR to new and exciting areas in metabolomics. Technological
advances have provided significant improvements to sensitivity and resolution,
which have led to the identification and quantitation of an expanded pool of
metabolites. NMR spectroscopy offers opportunities to gain mechanistic insights
into biochemical pathways in health and diseases, to discover biomarkers and
potential therapy targets, and to translate laboratory findings to clinical applications.
Continuing, multifaceted efforts to boost sensitivity, resolution, and the speed of
data acquisition and to improve quantitative accuracy promise to alleviate the
increasingly realized complexity of biological mixtures and large-scale
metabolomics studies. Moreover, ongoing technical and methodological advances
contribute to further expanding the routinely quantifiable metabolites in biological
specimens and hence NMR-based metabolomics is anticipated to greatly improve
and impact our understanding of systems biology and to help make progress in the
treatment and management of human diseases.

Acknowledgments The authors gratefully acknowledge the financial support from the NIH grants
R01GM138465, R01GM131491, P30CA015704, and P30DK035816.
Quantitative NMR Methods in Metabolomics 159

References
Airoldi C, Tripodi F, Guzzi C, Nicastro R, Coccetti P (2015) NMR analysis of budding yeast
metabolomics: a rapid method for sample preparation. Mol Biosyst 11(2):379–383
Akoka S, Barantin L, Trierweiler M (1999) Concentration measurement by proton NMR using the
ERETIC method. Anal Chem 71(13):2554–2557
Albrecht B, Voronina E, Schipke C, Peters O, Parr MK, Díaz-Hernández MD, Schlörer NE (2020)
Pursuing experimental reproducibility: an efficient protocol for the preparation of cerebrospinal
fluid samples for NMR-based metabolomics and analysis of sample degradation. Metabolites
10(6):251
Alum MF, Shaw PA, Sweatman BC, Ubhi BK, Haselden JN, Connor SC (2008) 4,4-Dimethyl-4-
silapentane-1-ammonium trifluoroacetate (DSA), a promising universal internal standard for
NMR-based metabolic profiling studies of biofluids, including blood plasma and serum.
Metabolomics 4:122–127
Anderson JR, Phelan MM, Rubio-Martinez LM, Fitzgerald MM, Jones SW, Clegg PD, Peffers MJ
(2020) Optimization of synovial fluid collection and processing for NMR metabolomics and
LC-MS/MS proteomics. J Proteome Res 19(7):2585–2597
Bala L, Ghoshal UC, Ghoshal U, Tripathi P, Misra A, Nagana Gowda GA, Khetrapal CL (2006)
Malabsorption syndrome with and without small intestinal bacterial overgrowth: a study on
upper-gut aspirate using 1H NMR spectroscopy. Magn Reson Med 56(4):738–744
Barding GA Jr, Salditos R, Larive CK (2012) Quantitative NMR for bioanalysis and metabolomics.
Anal Bioanal Chem 404(4):1165–1179
Barrilero R, Gil M, Amigó N, Dias CB, Wood LG, Garg ML et al (2018) LipSpin: a new
bioinformatics tool for quantitative 1 H NMR lipid profiling. Anal Chem 90:2031–2040
Beckonert O, Keun HC, Ebbels TM, Bundy J, Holmes E, Lindon JC, Nicholson JK (2007)
Metabolic profiling, metabolomic and metabonomic procedures for NMR spectroscopy of
urine, plasma, serum and tissue extracts. Nat Protoc 2(11):2692–2703
Bell JD, Brown JC, Kubal G, Sadler PJ (1988) NMR-invisible lactate in blood plasma. FEBS Lett
235:81–86
Bingol K, Zhang F, Bruschweiler-Li L, Brüschweiler R (2012) Carbon backbone topology of the
metabolome of a cell. J Am Chem Soc 134:9006–9011
Bingol K, Zhang F, Bruschweiler-Li L, Brüschweiler R (2013) Quantitative analysis of metabolic
mixtures by two-dimensional 13C constant-time TOCSY NMR spectroscopy. Anal Chem 85:
6414–6420
Bligh EG, Dyer WJ (1959) A rapid method of total lipid extraction and purification. Can J Biochem
Physiol 37(8):911–917
Bouatra S, Aziat F, Mandal R, Guo AC, Wilson MR, Knox C, Bjorndahl TC et al (2013) The human
urine metabolome. Plos One 8(9):e73076
Chatham JC, Forder JR (1999) Lactic acid and protein interactions: implications for the NMR
visibility of lactate in biological systems. Biochim Biophys Acta 1426(1):177–184
Chen W, Lu S, Wang G, Chen F, Bai C (2017) Staging research of human lung cancer tissues by
high-resolution magic angle spinning proton nuclear magnetic resonance spectroscopy
(HRMAS 1 H NMR) and multivariate data analysis. Asia Pac J Clin Oncol 13(5):e232–e238
Chikayama E, Suto M, Nishihara T, Shinozaki K, Kikuchi J (2008) Systematic NMR analysis of
stable isotope labeled metabolite mixtures in plant and animal systems: coarse grained views of
metabolic pathways. PLoS One 3:e3805
Choi JS, Baek HM, Kim S, Kim MJ, Youk JH, Moon HJ, Kim EK et al (2012) HR-MAS MR
spectroscopy of breast cancer tissue obtained with core needle biopsy: correlation with prog-
nostic factors. PLoS One 7:e51712
Cruz M, Wang M, Frisch-Daiello J, Han X (2016) Improved butanol-methanol (BUME) method by
replacing acetic acid for lipid extraction of biological samples. Lipids 51(7):887–896
160 G. A. Nagana Gowda and D. Raftery

d'Alessandro A, Osterloh JD, Chuwers P, Quinlan PJ, Kelly TJ, Becker CE (1994) Formate in
serum and urine after controlled methanol exposure at the threshold limit value. Environ Health
Perspect 102(2):178–181
Daykin CA, Foxall PJ, Connor SC et al (2002) The comparison of plasma deproteinization methods
for the detection of low-molecular-weight metabolites by (1)H nuclear magnetic resonance
spectroscopy. Anal Biochem 304(2):220–230
DeSilva MA, Shanaiah N, Nagana Gowda GA, Raftery MA, Hainline BE, Raftery D (2009)
Application of 31P NMR spectroscopy and chemical derivatization for metabolite profiling of
lipophilic compounds in human serum. Magn Reson Chem 47Suppl 1:S74–S80
Dinges SS, Vandergrift LA, Wu S, Berker Y, Habbel P, Taupitz M, Wu C-L et al (2019)
Metabolomic prostate cancer fields in HRMAS MRS-profiled histologically benign tissue
vary with cancer status and distance from cancer. NMR Biomed 32(10):e4038
Djukovic D, Raftery D, Nagana Gowda GA (2020) Chapter 16 – Mass spectrometry and NMR
spectroscopy based quantitative metabolomics. In: Issaq HJ, Veenstra TD (eds) Proteomic and
metabolomic approaches to biomarker discovery, 2nd edn. Academic Press, pp 289–311
Edison S, Colonna M, Gouveia GJ, Holderman NR, Judge MT, Shen X, Zhang S (2021) NMR:
unique strengths that enhance modern metabolomics research. Anal Chem 93:478
Emwas AH, Luchinat C, Turano P et al (2015) Standardizing the experimental conditions for using
urine in NMR-based metabolomic studies with a particular focus on diagnostic studies: a
review. Metabolomics 11(4):872–894
Emwas AH, Roy R, McKay RT, Ryan D, Brennan L, Tenori L, Luchinat C, Gao X, Zeri AC,
Nagana Gowda GA, Raftery D, Steinbeck C, Salek RM, Wishart DS (2016) Recommendations
and standardization of biomarker quantification using NMR-based metabolomics with particular
focus on urinary analysis. J Proteome Res 15(2):360–373
Engel KM, Baumann S, Rolle-Kampczyk U, Schiller J, von Bergen M, Grunewald S (2019)
Metabolomic profiling reveals correlations between spermiogram parameters and the
metabolites present in human spermatozoa and seminal plasma. PLoS One 14(2):e0211679
Evans C, Bogan KL, Song P, Burant CF, Kennedy RT, Brenner C (2010) NAD+ metabolite levels
as a function of vitamins and calorie restriction: evidence for different mechanisms of longevity.
BMC Chem Biol 10:2
Fan TW (2012) In: Fan TW, Higashi RM, Lane AN (eds) The handbook of metabolomics, methods
in pharmacology and toxicology, vol 2012. Springer, New York, pp 7–27
Fan TW, Lane AN, Higashi RM, Farag MA, Gao H, Bousamra M, Miller DH (2009) Altered
regulation of metabolic pathways in human lung cancer discerned by (13)C stable isotope-
resolved metabolomics (SIRM). Mol Cancer 8:41
Folch J, Lees M, Stanley G (1957) A simple method for the isolation and purification of total lipides
from animal tissues. J Biol Chem 226:497–509
Gil M, Samino S, Barrilero R, Correig X (2019) Lipid profiling using 1 H NMR spectroscopy.
Methods Mol Biol 2037:35–47
Goldman A, Khiste S, Freinkman E, Dhawan A, Majumder B, Mondal J, Pinkerton AB, Eton E,
Medhi R, Chandrasekar V, Rahman MM, Ichimura T, Gopinath KS, Majumder P, Kohandel M,
Sengupta S (2019) Targeting tumor phenotypic plasticity and metabolic remodeling in adaptive
crossdrug tolerance. Sci Signal 12(595):eaas8779. https://doi.org/10.1126/scisignal.aas8779
Goldoni L, Beringhelli T, Rocchia W, Realini N, Piomelli D (2016) A simple and accurate protocol
for absolute polar metabolite quantification in cell cultures using quantitative nuclear magnetic
resonance. Anal Biochem 501:26–34
Hernandez ME, Lopez AC, Calatayud AG et al (2001) Vesical uric acid lithiasis in a child with
renal hypouricemia. An Esp Pediatr 55(3):273–276
Holzgrabe U (2010) Quantitative NMR spectroscopy in pharmaceutical applications. Prog Nucl
Magn Reson Spectrosc 57(2):229–240
Hoult DI (2000) The principle of reciprocity in signal strength calculations – a mathematical guide.
Concepts Magn Reson 12:173–187
Quantitative NMR Methods in Metabolomics 161

Hoult DI, Richards RE (1976) The signal-to-noise ratio of the nuclear magnetic resonance experi-
ment. J Magn Reson 24:71–85
Ivanisevic J, Zhu ZJ, Plate L, Tautenhahn R, Chen S, O'Brien PJ, Johnson CH, Marletta MA, Patti
GJ, Siuzdak G (2013) Toward 'omic scale metabolite profiling: a dual separation-mass spec-
trometry approach for coverage of lipid and central carbon metabolism. Anal Chem 85(14):
6876–6884
Jeyarajah EJ, Cromwell WC, Otvos JD (2006) Lipoprotein particle analysis by nuclear magnetic
resonance spectroscopy. Clin Lab Med 26:847–870
Jiménez B, Holmes E, Heude C, Tolson RF, Harvey N, Lodge SL, Chetwynd AJ et al (2018)
Quantitative lipoprotein subclass and low molecular weight metabolite analysis in human serum
and plasma by 1H NMR spectroscopy in a multilaboratory trial. Anal Chem 90(20):
11962–11971
Johnson CS (1999) Diffusion ordered nuclear magnetic resonance spectroscopy: principles and
applications. Prog Nucl Magn Reson Spectrosc 34:203–256
Johnson CH, Ivanisevic J, Siuzdak G (2016) Metabolomics: beyond biomarkers and towards
mechanisms. Nat Rev Mol Cell Biol 17:451–459
Kaess B, Fischer M, Baessler A, Stark K, Huber F, Kremer W, Kalbitzer HR et al (2008) The
lipoprotein subfraction profile: heritability and identification of quantitative trait loci. J Lipid
Res 49:715–723
Kalfe A, Telfah A, Lambert J, Hergenröder R (2015) Looking into living cell systems: planar
waveguide microfluidic NMR detector for in vitro metabolomics of tumor spheroids. Anal
Chem 87(14):7402–7410
Kapur BM, Vandenbroucke AC, Adamchik Y, Lehotay DC, Carlen PL (2007) Formic acid, a novel
metabolite of chronic ethanol abuse, causes neurotoxicity, which is prevented by folic acid.
Alcohol Clin Exp Res 31(12):2114–2120
Kodama M, Oshikawa K, Shimizu H, Yoshioka S, Takahashi M, Izumi Y, Bamba T, Tateishi C,
Tomonaga T, Matsumoto M, Nakayama KI (2020) A shift in glutamine nitrogen metabolism
contributes to the malignant progression of cancer. Nat Commun 11:1320
Kriat M, Confort-Gouny S, Vion-Dury J, Sciaky M, Viout P, Cozzone PJ (1992) Quantitation of
metabolites in human blood serum by proton magnetic resonance spectroscopy. A comparative
study of the use of formate and TSP as concentration standards. NMR Biomed 5(4):179–184
Kubáň P, Boček P (2013) Direct analysis of formate in human plasma, serum and whole blood by
in-line coupling of microdialysis to capillary electrophoresis for rapid diagnosis of methanol
poisoning. Anal Chim Acta 768:82–89
Lane AN, Fan TW, Bousamra M II, Higashi RM, Yan J, Milleret DM (2011) Stable isotope-
resolved metabolomics (SIRM) in cancer research with clinical application to NonSmall cell
lung cancer OMICS. J Integr Biol 15:173–182
Lane AN, Tan J, Wang Y et al (2017) Probing the metabolic phenotype of breast cancer cells by
multiple tracer stable isotope resolved metabolomics. Metab Eng 43(Pt B):125–136
Lin P, Lane AN, Fan TW (2019) Stable isotope-resolved metabolomics by NMR. Methods Mol
Biol 2037:151–168
Locasale JW, Grassian AR, Melman T, Lyssiotis CA, Mattaini KR, Bass AJ, Heffron G et al (2011)
Phosphoglycerate dehydrogenase diverts glycolytic flux and contributes to oncogenesis. Nat
Genet 43:869–874
Löfgren L, Ståhlman M, Forsberg GB, Saarinen S, Nilsson R, Hansson GI (2012) The BUME
method: a novel automated chloroform-free 96-well total lipid extraction method for blood
plasma. J Lipid Res 53:1690–1700
Lohavanichbutr P, Zhang Y, Wang P, Gu H, Nagana Gowda GA, Djukovic D, Buas MF, Raftery D,
Chen C (2018) Salivary metabolite profiling distinguishes patients with oral cavity squamous
cell carcinoma from normal controls. PLoS One 13(9):e0204249
Lucas LH, Larive CK, Wilkinson PS, Huhn S (2005) Progress toward automated metabolic
profiling of human serum: comparison of CPMG and gradient-filtered NMR analytical methods.
J Pharm Biomed Anal 39(1–2):156–163
162 G. A. Nagana Gowda and D. Raftery

Lussu M, Camboni T, Piras C, Serra C, Del Carratore F, Griffin J, Atzori L, Manzin A (2017) 1H
NMR spectroscopy-based metabolomics analysis for the diagnosis of symptomatic E. coli-
associated urinary tract infection (UTI). BMC Microbiol 17(1):201
Mallol R, Rodriguez MA, Brezmes J, Masana L, Correig X (2013) Human serum/plasma lipopro-
tein analysis by NMR: application to the study of diabetic dyslipidemia. Prog Nucl Magn Reson
Spectrosc 70:1–24
Mallol R, Amigó N, Rodríguez MA, Heras M, Vinaixa M, Plana N, Rock E et al (2015) Liposcale: a
novel advanced lipoprotein test based on 2D diffusion-ordered 1H NMR spectroscopy. J Lipid
Res 56(3):737–746
Maniara G, Rajamoorthi K, Rajan S, Stockton GW (1998) Method performance and validation for
quantitative analysis by 1H and 31P NMR spectroscopy. Applications to analytical standards and
agricultural chemicals. Anal Chem 70(23):4921–4928
Maniscalco M, Cutignano A, Paris D, Melck DJ, Molino A, Fuschillo S, Motta A (2020)
Metabolomics of exhaled breath condensate by nuclear magnetic resonance spectroscopy and
mass spectrometry: a methodological approach. Curr Med Chem 27(14):2381–2399
Mo H, Raftery D (2008) Solvent signal as an NMR concentration reference. Anal Chem 80(24):
9835–9839
Nagana Gowda GA (2011) NMR spectroscopy for discovery and quantitation of biomarkers of
disease in human bile. Bioanalysis 3(16):1877–1890
Nagana Gowda GA (2018) Profiling redox and energy coenzymes in whole blood, tissue and cells
using NMR spectroscopy. Metabolites 8(2):32
Nagana Gowda GA, Raftery D (2014a) Advances in NMR based metabolomics. In: Simó C,
Cifuentes A, García-Cañas V (eds) Fundamentals of advanced Omics technologies: from
genes to metabolites, comprehensive analytical chemistry, vol 63. Elsevier, New York, pp
187–211
Nagana Gowda GA, Raftery D (2014b) Quantitating metabolites in protein precipitated serum using
NMR spectroscopy. Anal Chem 86(11):5433–5440
Nagana Gowda GA, Raftery D (2015) Can NMR solve some significant challenges in
metabolomics? J Magn Reson 260:144–160
Nagana Gowda GA, Raftery D (2017a) Recent advances in NMR-based metabolomics. Anal Chem
89(1):490–510
Nagana Gowda GA, Raftery D (2017b) Whole blood metabolomics by 1H NMR spectroscopy
provides a new opportunity to evaluate coenzymes and antioxidants. Anal Chem 89(8):
4620–4627
Nagana Gowda GA, Raftery D (eds) (2019) NMR based metabolomics: methods and protocols,
Methods in molecular biology, vol 2037. Humana Press/Springer Science, New York
Nagana Gowda GA, Gowda YN, Raftery D (2015) Expanding the limits of human blood metabolite
quantitation using NMR spectroscopy. Anal Chem 87(1):706–715
Nagana Gowda GA, Abell L, Lee CF, Tian R, Raftery D (2016) Simultaneous analysis of major
coenzymes of cellular redox reactions and energy using ex vivo (1)H NMR spectroscopy. Anal
Chem 88(9):4817–4824
Nagana Gowda GA, Djukovic D, Bettcher LF, Gu H, Raftery D (2018) NMR-guided mass
spectrometry for absolute quantitation of human blood metabolites. Anal Chem 90(3):
2001–2009
Nagana Gowda GA, Abell L, Tian R (2019) Extending the scope of 1H NMR spectroscopy for the
analysis of cellular coenzyme a and acetyl coenzyme A. Anal Chem 91(3):2464–2471
Nagana Gowda GA, Hong NN, Raftery D (2021) Evaluation of Fumaric acid and maleic acid as
internal standards for NMR analysis of protein precipitated plasma, serum, and whole blood.
Anal Chem 93(6):3233–3240
Nguyen TTM, An YJ, Cha JW, Ko YJ, Lee H, Chung CH, Jeon SM et al (2020) Real-time
in-organism NMR metabolomics reveals different roles of AMP-activated protein kinase cata-
lytic subunits. Anal Chem 92(11):7382–7387
Quantitative NMR Methods in Metabolomics 163

Nicholson JK, Gartland KP (1989) 1H NMR studies on protein binding of histidine, tyrosine and
phenylalanine in blood plasma. NMR Biomed 2(2):77–82
Nicholson JK, Foxall PJD, Spraul M, Farrant RD, Lindon JC (1995) 750 MHz 1H and 1H-13C
NMR spectroscopy of human blood plasma. Anal Chem 67:793–811
Orczyk-Pawilowicz M, Jawien E, Deja S, Hirnle L, Zabek A, Mlynarz P (2016) Metabolomics of
human amniotic fluid and maternal plasma during normal pregnancy. PLoS One 11(4):
e0152740
Paul A, Kumar S, Raj A, Sonkar AA, Jain S, Singhai A, Roy R (2018) Alteration in lipid
composition differentiates breast cancer tissues: a 1 H HRMAS NMR metabolomic study.
Metabolomics 14(9):119
Pauli GF, Gödecke T, Jaki BU, Lankin DC (2012) Quantitative 1H NMR. Development and
potential of an analytical method: an update. J Nat Prod 75(4):834–851
Psychogios N, Hau DD, Peng J, Guo AC, Mandal R, Bouatra S, Sinelnikov I et al (2011) The
human serum metabolome. PLoS One 6(2):e16957
Raftery D (ed) (2014) Mass spectrometry in metabolomics: methods and protocols, Methods in
molecular biology, vol 1198. Humana Press/Springer Science, New York
Rundlöf T, Mathiasson M, Bekiroglu S, Hakkarainen B, Bowden T, Arvidsson T (2010) Survey and
qualification of internal standards for quantification by 1H NMR spectroscopy. J Pharm Biomed
Anal 52(5):645–651
Rylander R, Remer T, Berkemeyer S, Vormann J (2006) Acid–base status affects renal magnesium
losses in healthy, elderly persons. J Nutr 136(9):2374–2377
Salem AA, Mossa HA (2012) Method validation and determinations of levofloxacin, metronidazole
and sulfamethoxazole in an aqueous pharmaceutical, urine and blood plasma samples using
quantitative nuclear magnetic resonance spectrometry. Talanta 88:104–114
Shanaiah N, Desilva MA, Nagana Gowda GA et al (2007) Class selection of amino acid metabolites
in body fluids using chemical derivatization and their enhanced 13C NMR. Proc Natl Acad Sci
U S A 104(28):11540–11544
Simón-Manso Y, Lowenthal MS, Kilpatrick LE, Sampson ML, Telu KH, Rudnick PA, Mallard WG
et al (2013) Metabolite profiling of a NIST standard reference material for human plasma (SRM
1950): GC-MS, LC-MS, NMR, and clinical laboratory analyses, libraries, and web-based
resources. Anal Chem 85(24):11725–11731
Sitter B, Bathen TF, Singstad TE, Fjøsne HE, Lundgren S, Halgunset J, Gribbestad IS (2010)
Quantification of metabolites in breast cancer patients with different clinical prognosis using HR
MAS MR spectroscopy. NMR Biomed 23(4):424–431
Soininen P, Kangas AJ, Wurtz P, Tukiainen T, Tynkkynen T, Laatikainen R, Jarvelin MR et al
(2009) High-throughput serum NMR metabonomics for cost-effective holistic studies on
systemic metabolism. Analyst 134:1781–1785
Soininen P, Kangas AJ, Würtz P, Suna T, Ala-Korpela M (2015) Quantitative serum nuclear
magnetic resonance metabolomics in cardiovascular epidemiology and genetics. Circ
Cardiovasc Genet 8(1):192–206
Takis PG, Schäfer H, Spraul M, Luchinat C (2017) Deconvoluting interrelationships between
concentrations and chemical shifts in urine provides a powerful analysis tool. Nat Commun
8(1):1662
Tayyari F, Nagana Gowda GA, Gu H, Raftery D (2013) 15N-cholamine – a smart isotope tag for
combining NMR- and MS-based metabolite profiling. Anal Chem 85(18):8715–8721
Tilgner M, Vater TS, Habbel P, Cheng LL (2019) High-resolution magic angle spinning (HRMAS)
NMR methods in metabolomics. Methods Mol Biol 2037:49–67
Tiziani S, Emwas AH, Lodi A, Ludwig C, Bunce CM, Viant MR, Güntheret UL (2008) Optimized
metabolite extraction from blood serum for 1H nuclear magnetic resonance spectroscopy. Anal
Biochem 377(1):16–23
Trammell SA, Brennera C (2013) Targeted, LCMS-based metabolomics for quantitative measure-
ment of NAD(+) metabolites. Comput Struct Biotechnol J 4:e201301012
164 G. A. Nagana Gowda and D. Raftery

Tynkkynen T, Wang Q, Ekholm J, Anufrieva O, Ohukainen P, Vepsäläinen J, Männikkö M et al

(2019) Proof of concept for quantitative urine NMR metabolomics pipeline for large-scale
epidemiology and genetics. Int J Epidemiol 48(3):978–993
Van der Klink JJ (2001) The NMR reciprocity theorem for arbitrary probe geometry. J Magn Reson
148:147–154
Vicente-Muñoz S, Lin P, Fan TW, Lane AN (2021) NMR analysis of carboxylate isotopomers of
13C-metabolites by chemoselective derivatization with 15N-cholamine. Anal Chem 93
(17):6629–6637
Welch AA, Mulligan A, Bingham SA, Khaw K-T (2008) Urine pH is an indicator of dietary acid-
base load, fruit and vegetables and meat intakes: results from the European prospective investi-
gation into cancer and nutrition (EPIC)- Norfolk population study. Br J Nutr 99(6):1335–1343
Wen H, An YJ, Xu WJ, Kang KW, Park S (2015) Real-time molnitoring of cancer cell metabolism
and effects of an anticancer agent using 2D in-cell NMR spectroscopy. Angew Chem Int Ed
Engl 54(18):5374–5377
Wevers RA, Engelke U, Heerschap A (1994) High-resolution 1H-NMR spectroscopy of blood
plasma for metabolic studies. Clin Chem 40(7 Pt 1):1245–1250
Wider G, Dreier L (2006) Measuring protein concentrations by NMR spectroscopy. J Am Chem
Soc 128(8):2571–2576
Wishart DS (2016) Emerging applications of metabolomics in drug discovery and precision
medicine. Nat Rev Drug Discov 15:473–484
Wishart DS (2019) NMR metabolomics: a look ahead. J Magn Reson 306:155–161
Wong A, Jiménez B, Li X, Holmes E, Nicholson JK, Lindon JC, Sakellariou D (2012) Evaluation of
high resolution magic-angle coil spinning NMR spectroscopy for metabolic profiling of
nanoliter tissue biopsies. Anal Chem 84(8):3843–3848
Würtz P, Kangas AJ, Soininen P, Lawlor DA, Smith GD, Ala-Korpela M (2017) Quantitative serum
nuclear magnetic resonance metabolomics in large-scale epidemiology: a primer on -Omic
technologies. Am J Epidemiol 186(9):1084–1096
Xu WJ, Wen H, Kim HS, Ko YJ, Dong SM, Park IS, Yook JI, Park S (2018) Observation of acetyl
phosphate formation in mammalian mitochondria using real-time in-organelle NMR
metabolomics. Proc Natl Acad Sci U S A 115(16):4152–4157
Yazdani M, Elgstøen KBP, Rootwelt H, Shahdadfar A, Utheim ØA, Utheim TP (2019) Tear
metabolomics in dry eye disease: a review. Int J Mol Sci 20(15):3755. https://doi.org/10.
3390/ijms20153755
Ye T, Mo H, Shanaiah N, Nagana Gowda GA, Zhang S, Raftery D (2009) Chemoselective 15N tag
for sensitive and high-resolution nuclear magnetic resonance profiling of the carboxyl-
containing metabolome. Anal Chem 81(12):4882–4888
Zebrowska A, Skowronek A, Wojakowska A, Widlak P, Pietrowska M (2019) Metabolome of
exosomes: focus on vesicles released by cancer cells and present in human body fluids. Int J Mol
Sci 20(14):3461
Zhang F, Bruschweiler-Li L, Brüschweiler R (2012) High-resolution homonuclear 2D NMR of
carbon-13 enriched metabolites and their mixtures. J Magn Reson 225:10–13
Zierer J, Jackson MA, Kastenmüller G, Mangino M, Long T, Telenti A, Mohney RP, Small KS,
Bell JT, Steves CJ, Valdes AM, Spector TD, Menni C (2018) The fecal metabolome as a
functional readout of the gut microbiome. Nat Genet 50:790–795
Advancements in Pulsed Stable
Isotope-Resolved Metabolomics

Martin Forbes, Sabrina Geisberger, Matthias Pietzke,

Guido Mastrobuoni, and Stefan Kempa

Contents
1 Introduction: A Brief History of Isotopic Labelling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 166
2 Time- and Isotope-Resolved Metabolomics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 168
2.1 Stable Isotope-Resolved Metabolomics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 168
2.2 Time Resolved Isotope Labelling Studies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 169
2.3 Isotope Tracing at the Crossroad of Metabolic Pathways . . . . . . . . . . . . . . . . . . . . . . . . . . . . 172
2.4 The Application of Ultra-High Resolution Mass Spectrometry Allows the Tracing
of Different Isotopic Species . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 174
3 Applications and Future Perspectives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 176
3.1 Applications of pSIRM . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 176
3.2 Perspective Towards Single Cells . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 177
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 178

Abstract

The understanding of biochemical processes of metabolism is gained through the

measurement of the concentration of intermediates and the rate of metabolite
conversion. However, the measurement of metabolite concentrations does not
give a full representation of this dynamic system. To understand the kinetics of
metabolism, the system must be described and quantified in terms of metabolite
flow as a function of time. In order to measure the metabolite flow, or more
precisely the metabolic flux through a biological system, substrates of the cell are
labelled with stable isotopes. The usage of these substrates by the cell leads to the
incorporation of the isotopes into downstream intermediates.

M. Forbes · S. Geisberger · M. Pietzke · G. Mastrobuoni · S. Kempa (✉)

Max Delbrück Center for Molecular Medicine in the Helmholtz Association, Berlin Institute for
Medical Systems Biology, Berlin, Germany
e-mail: [email protected]; [email protected];
[email protected]; [email protected]; [email protected]

# The Author(s), under exclusive license to Springer Nature Switzerland AG 2022 165
V. Ghini et al. (eds.), Metabolomics and Its Impact on Health and Diseases,
Handbook of Experimental Pharmacology 277, https://doi.org/10.1007/164_2022_621
166 M. Forbes et al.

The most important metabolic pathways are encompassed in the central carbon
metabolism (CCM). According to the Kyoto Encyclopedia of Genes and
Genomes (KEGG), the central carbon metabolism “is the most basic aspect of
life”. It includes all metabolites and enzymatic reactions within: glycolysis and
gluconeogenesis, pentose phosphate pathway (PPP), tricarboxylic acid (TCA)
cycle, oxidative phosphorylation (OXPHOS), amino acids and nucleotide meta-
bolic pathways. Some molecules are at the crossroad of metabolic pathways,
interconnecting diverse metabolic and therefore functional outcomes. Labelling
these nodal metabolites and analysing their isotopic composition allows the
precise determination of the metabolic flow within the biochemical networks
that they are in.
Application of stable isotope labelled substrates allows the measurement of
metabolic flux through a biochemical pathway. The rapid turnover of metabolites
in pathways requires pulse-feeding cells with a labelled substrate. This method
allows for the determination of different cell states. For example, the action of a
drug from immediate impact until the compensatory response of the metabolic
system (cell, organs, organisms). Pulsed labelling is an elegant way to analyse the
action of small molecules and drugs and enables the analysis of regulatory
metabolic processes in short time scales.

Keywords

Cancer metabolism · Isotope-resolved metabolomics · Mass spectrometry

methods · Metabolic flux analysis

1 Introduction: A Brief History of Isotopic Labelling

The use of isotope labelled molecules in metabolic research began within the first
decades of the last century. In the early days of metabolism research, radioactive
isotopes were used to investigate the metabolic flow in bacteria, plants and animals.
In order to investigate the structure of biochemical pathways and specifically
metabolic cycles, e.g. tricarboxylic acid cycle, reverse tricarboxylic acid cycle or
Calvin-Benson cycle, isotopes were applied in a time resolved manner. Time
resolution allowed for the elucidation of the consecutive order of chemical reactions
within the investigated pathways. The use of radioactive isotopes was of paramount
importance to decipher the flow of carbon within cells and organisms.
In 1910, English chemist Frederick Soddy observed that “elements of different
atomic weights may possess identical (chemical) properties” and therefore belong to
the same position in the periodic table. This included not only radioactive elements
but also stable isotopes of atoms, i.e. atoms with the same number of protons but
different numbers of neutrons in their nucleus. We can refer to these elements as
“hot” or “cold” isotopes of an atom in dependency of their radio- or
non-radioactivity. The presence of radioactive isotopes was discovered by black
spots occurring on photosensitive emulsions, as the decaying radioactive element
Advancements in Pulsed Stable Isotope-Resolved Metabolomics 167

produced traceable radiation. This phenomenon gave rise to years of research into
the isotopes of the periodic table and their many uses in a broad range of analytical
fields, including: chemistry, geology, biology, physics and medicine (Wilikinson
2018).
Georg Charles de Hevesy was a Hungarian radio-chemist and Nobel Prize in
Chemistry laureate, recognised in 1943 for his achievements in the development of
radioactive tracers in the study of metabolism in animals. Hevesy is considered the
first to use radioisotopes to measure metabolic flux in biological systems. One such
(sadly unpublished) experiment of Hevesy and Rutherford describes the practices of
a shrewd landlady in 1911, Manchester. Hevesy was convinced that his landlady was
recycling food, much to the denial of such practice. In order to thwart the thrifty
landlady, Hevesy “spiked” a portion of leftover meat with a tiny amount of a
radioactive material. A few days later he brought an electroscope to the table to
demonstrate to the indignant landlady that the food served that day was radioactive.
Hevesy is thought to have enjoyed fresh hot meals as a result of his radioactive
tracing experiment (Myers 1996). Hevesy continued his endeavours into the usage of
tracers in biological systems. In 1934, Hevesy and Hofer first used tracers in
medicine by using an enriched stable isotope to determine the rate of elimination
of water from the human body. Hevesy and Hofer drank dilute deuterated water and
assayed the isotopic dilution of the deuterium in their urine. From their results they
concluded the average time a water molecule spent in their bodies was 13 ± 1.5 days
(Hevesy and Hofer 1934).
The accounts of Hevesy are far more than comical anecdotes. They provide an
understandable example for the use of radioisotopes in analytical chemistry.
Through the development of mass spectrometry, stable isotopes replaced
radioisotopes in tracing experiments. In mass spectrometry, a metabolite can be
identified by the mass spectrum of its fragments, i.e. the intensity of the fragment’s
peaks at a defined mass-to-charge (m/z) ratio. The additional neutron in the atomic
nucleus of isotopes makes the atom 1 Da heavier and increases the m/z ratio. The
incorporation of isotopes into a metabolite changes the atomic composition of this
intermediate and induces a shift in the respective fragment’s mass spectrum. Specifi-
cally, stable isotopes of carbon (13C), nitrogen (15N), oxygen (18O) and hydrogen
(2H) can be introduced into organic compounds. By tracing the incorporation of the
isotope, the metabolic fate of these compounds within biological systems can be
characterised (Wilikinson 2018). The incorporation of these isotopes can be resolved
in molecules such as sugars, amino acids or nucleotides. This method of isotope
labelling was used to describe the effect of oncogenes on cancer cell metabolism. For
example, Le et al. applied 13C-Glucose and 13C,15N-Glutamine to trace glucose and
glutamine metabolism when the MYC oncogene was induced in P493 cancer cells.
The authors showed that glutamine plays an essential role in the cells’ proliferation
and survival, highlighting targets of glutamine metabolism for cancer therapy
(Le et al. 2012). The usage of isotope labelling provides insights into the dynamics
and kinetics of metabolism, as a function of time and cell state.
168 M. Forbes et al.

2 Time- and Isotope-Resolved Metabolomics

2.1 Stable Isotope-Resolved Metabolomics

In order to measure the metabolite flow, or more precisely the metabolic flux through
a biological system, substrates of the system are labelled with stable isotopes. These
substrates may be, glucose or glutamine which provide carbon and nitrogen to the
central metabolic pathways (Fig. 1). The usage of these substrates by the cell leads to
the incorporation of the isotopes into downstream intermediates. After isotopically
labelling a biological system and extracting the cellular metabolites, mass spectrom-
etry is then employed to analyse the number of heavy atoms incorporated
(isotopologues) and their positions (isotopomers) in detected metabolites (Bruntz
et al. 2017). Isotopologues can be identified by increasing mass shifts, as every
heavy atom incorporated rises the m/z ratio by one. Isotopomers, on the other hand,
can be distinguished by the heavy atom incorporation visible in different fragments

Fig. 1 Scheme of central metabolic pathways indicating carbon and nitrogen flow. The scheme
displays main central metabolic pathways: glycolysis, pentose phosphate pathway, nucleotide
metabolism, tricarboxylic acid cycle and glutaminolysis. The different colours depict the contribu-
tion of the distinct carbon and nitrogen atoms stemming from glucose or glutamine, respectively
Advancements in Pulsed Stable Isotope-Resolved Metabolomics 169

of the same metabolite. In summary, the usage of isotope labelled substrates allows
for the follow-up of interconnected anabolic and catabolic pathways. The tracing of
isotope labels in metabolites, and the atomic position in which the label resides, is
referred to as stable isotope-resolved metabolomics (SIRM).
Cellular metabolism is a rapid, dynamic process, in which both anabolism and
catabolism of metabolites produce energy, build macromolecules for biomass and
generate intermediates involved in intra- and inter-cellular signalling. The under-
standing of biochemical processes of metabolism is inferred through the concentra-
tion of intermediates and the rate of metabolite conversion. However, the
measurement of metabolite concentrations does not give a full representation of
this dynamic system. In order to understand the kinetics of metabolism, it must be
described and quantified in terms of metabolite flow as a function of time (Sauer
2006; Pietzke et al. 2014; Buescher et al. 2015; Jang et al. 2018).

2.2 Time Resolved Isotope Labelling Studies

Metabolism is a highly dynamic system in which biochemical reactions occur

rapidly. This allows the cell to constantly meet its needs and adapt to changing
stimuli and micro-environments. The time for which cells are exposed to a labelled
substrate determines the amount of isotopic label incorporated into its metabolites.
Supplying a cell with an isotope labelled substrate for extended periods of time will
lead to complete usage and saturation of downstream pathways. This condition is
referred to as isotopic steady state and solely reflects the usage of metabolic
pathways in relation to each other, i.e. in which pathways the given substrate is
involved in. In other words, if one applies the labelled substrate for such a period of
time that the label incorporation has reached saturation, then the model of flux would
be stationary (Fig. 2, left (II)).
To understand the flux of labelled metabolites, we aim to assess metabolism in a
non-stationary state. In order to measure the metabolic activity, the time a labelled
substrate is offered to a cell is limited. A labelling time is chosen where the label
incorporation into intermediates of interest is in linear relationship to time (Fig. 2,
left (I)). At this chosen labelling time, one can analyse the speed of a specific
pathway. It is also possible to assess how different conditions the cell is exposed
to affect the velocity of the pathway in use.
By avoiding saturating the system, information about pathway preference and
directionality is provided, as discussed in Sect. 2.3. The reduction in labelling time
can be implemented by a pulsed labelling-quenching strategy, referred to as pulsed
stable isotope-resolved metabolomics (pSIRM).
We may take glycolysis as a metabolic pathway that can be analysed by a tracing
experiment. 13C6-Glc is used as a substrate for the glycolytic pathway. This labelled
tracer contains 13C at every carbon position on the glucose molecule. As glycolysis
proceeds 13C is incorporated into downstream metabolites (Fig. 2, right). The rate of
label incorporation can be described in both functions of time and quantity of
labelled metabolite. The rate of label incorporation, or indeed the metabolic rate,
170 M. Forbes et al.

Fig. 2 Modelling the kinetics of metabolite labelling. As cells are incubated with a labelled
substrate for a period of time, the quantity of label incorporated into metabolites increases. The
rate of reaction, or the percentage of label in the metabolites relative to the unlabelled fraction, is
dependent on the position of the metabolite in the pathway and the efficiency of associated
enzymes. In order to understand the rate and kinetics of metabolic processes, we aim to observe
labelled metabolites in the instationary phase (Fig. 2, left (I)). On the contrary, if all metabolites
are saturated with label after a given period of time (Fig. 2, left (II)), then no kinetic information
can be derived

of glycolysis is rapid. For instance, glucose-6-phosphate (G6P) – the primary step in

glycolysis – is labelled to saturation in approximately 2–5 min (Pietzke et al.
2014). Longer labelling times are required for the label to reach saturation in
metabolites downstream of G6P. Through a pulse-quench-harvest labelling strategy
we can determine the rate of label incorporation into metabolites of the glycolytic
pathway, as a function of the time cells are exposed to the labelled substrate. With
this information we can determine how “glycolytic” a certain cell type is.
Furthermore, we can study the effects on inhibitors on pathways by monitoring
the relative changes in label incorporation upon inhibitor treatment.
3-Bromopyruvate (BrPyr), for example, is a strong glycolytic inhibitor. More spe-
cifically, BrPyr inhibits glycolysis at the glyceraldehyde-3-phosphate dehydroge-
nase (GAPDH) reaction (Pietzke et al. 2014). In order to localise BrPyr action on
GAPDH, cells were pre-incubated with BrPyr for 12 min followed for a tracing with
13
C6-Glc for 3 min to label glycolytic metabolites. They showed that the carbon flow
through metabolites downstream of GAPDH was almost completely inhibited in
BrPyr treated cells (Fig. 3). The concentration of 3PGA drops below the detection
limit and the dramatic decrease in the label incorporation into pyruvate, lactate and
citrate proved bromopyruvate’s inhibitory activity in the lower glycolysis. In con-
trast fructose-1,6-bisphosphate (above the blockage) accumulates roughly fourfold,
while at the same time the label incorporation decreases by a factor 4, so the total
flow into this compound remains constant. This illustrates the connection of metab-
olite pool size and label incorporation and highlights the importance of considering
both readouts simultaneously during a pSIRM experiment.
Advancements in Pulsed Stable Isotope-Resolved Metabolomics 171

Fig. 3 pSIRM data in 2D. Each point represents one single metabolite of the CCM. On the x-axis
fold-changes of metabolite concentrations and on the y-axis changes of isotope incorporation are
shown. For this experiment cancer cells were incubated for a total time of 15 min with BrPyr and
metabolites were measured using the pSIRM technology (see Pietzke et al. (2014)). The graph
shows that in some metabolites a quantitative change in metabolite concentration can be observed.
However, the effect on glycolytic inhibition can only be observed in the label incorporation of
metabolites downstream of the GAPDH reaction

We have described the use of isotope labelled substrates for investigating the rate
of reactions in metabolic pathways. One may also investigate the direction of
metabolic pathways or deduce the relative dominance of a pathway by using
substrates labelled at specific carbon positions. For example, the pentose phosphate
pathway (PPP) – which links glycolysis to nucleotide synthesis – generates ribose-5-
phosphate (R5P) from two reactions. The oxidative PPP utilises glucose-6-phos-
phate to produce 6-phosphogluconate which is de-carboxylated to ribulose-6-phos-
phate (Ru5P). In turn Ru5P is isomerised to R5P. The non-oxidative PPP cycles
carbons from fructose-6-phosphate, glyceraldehyde-3-phosphate and erythrose-4-
phosphate to produce xyulose-5-phosphate and R5P. This branch does not contain a
de-carboxylation event and carbon number is maintained in the process. Therefore,
we may employ a labelling strategy to decipher the ratio of the oxidative and
non-oxidative branches by assessing the state of the de-carboxlyation event in the
oxidative branch. By using 1,2-[13C]-Glc as the labelling substrate we can measure
the ratio of R5P present in the sample which contains a single (m/z + 1) or a double
(m/z + 2) label. R5P which is derived from the oxidative PPP will only contain a
single label as the labelled carbon is lost during the de-carboxylation event (Fig. 4).
Oxythiamine is an inhibitor of the non-oxidative PPP enzyme, transketolase
(TKT). Boros et al. (1997) showed that oxythiamine inhibits Mia pancreatic adeno-
carcinoma cell growth by 39%. By using 1,2-[13C]-Glc labelling the authors
172 M. Forbes et al.

Fig. 4 Utilising 1,2-[13C]-Glc in the analysis of the pentose phosphate pathway. The pentose
phosphate pathway is comprised of the oxidative and non-oxidative branches. The oxidative branch
observes a de-carboxylation event which removes a labelled carbon from glucose-6-phosphate. In
the non-oxidative branch, labelled carbons are maintained as de-carboxylation does not occur. The
ratio of single to double labelled R5P shows the relative dominance of the branches. One may
observe the single or double labelled R5P in a mass spectrum generated by MS, by the relative
abundance of R5P (m/z + 1) and R5P (m/z + 2). This is depicted in the simulated MS spectrum on
the right

discovered that this tumour cell line generated R5P predominantly through
transketolase and the non-oxidative PPP (85%). It was deduced that oxythiamine
was an effective anti-tumour inhibitor due to the dependency of R5P synthesis, and
downstream nucleotide synthesis, on the non-oxidative PPP (Boros et al. 1997). In
summary, differential labelling of substrates permits not only the analysis of the
kinetics of metabolism but also gives information on the relative dependencies of
parallel reactions in different cell lines for targeted therapy development.

2.3 Isotope Tracing at the Crossroad of Metabolic Pathways

The most important metabolic pathways are encompassed in the central carbon
metabolism (CCM). According to the Kyoto Encyclopedia of Genes and Genomes
(KEGG), the central carbon metabolism “is the most basic aspect of life”. It includes
all enzymatic reactions within: glycolysis and gluconeogenesis, pentose phosphate
pathway (PPP), tricarboxylic acid (TCA) cycle and oxidative phosphorylation
(OXPHOS), amino acids and nucleotide metabolism pathways. Additionally, the
CCM includes six known carbon fixation pathways (reductive pentose phosphate
cycle (Calvin cycle), reductive citrate cycle, 3-hydroxypropionate bi-cycle, two
variants of 4-hydroxybutyrate pathway and reductive acetyl-coenzyme A (CoA)
pathway) as well as some pathways of methane metabolism, all not relevant in
animal cells (Qiu 2013). Some molecules are at the crossroad of metabolic pathways,
Advancements in Pulsed Stable Isotope-Resolved Metabolomics 173

Fig. 5 13C and 15N label integration into pathways of the central carbon metabolism. Pictogram of
glycolysis, TCA cycle, pentose phosphate pathway, as well as de novo purine and pyrimidine
biosynthesis. Red circles represent glucose-derived 13C atoms. Blue represent the 13C (dark blue) or
15
N (light blue) atoms derived from glutamine. Unlabelled C and N atoms are displayed in white

interconnecting diverse metabolic and therefore functional outcomes. Labelling

these nodal metabolites and analysing their isotopic composition allows the precise
determination of the metabolic flow within the biochemical networks that they are in
(Fig. 5).
One example of such intersections is citrate. Glucose-derived pyruvate can be
metabolised by pyruvate dehydrogenase to citrate or by pyruvate decarboxylase to
oxaloacetate, which itself feeds into citrate during TCA cycling. On the other hand,
glutamine can enter the TCA cycle via glutamic acid and 2-oxo-glutaric acid and
feed into citrate via both, oxidative (classic) and reductive (reverse) TCA cycling.
Via ATP-citrate lyase, citrate is further interconnected with fatty acid oxidation and
fatty acid biosynthesis. When labelling with 13C-glucose, 13C-glutamine or 13C-fatty
acids one will be able to distinguish between different mass shifts in citrate,
depending on the source of the 13C-label as well as the directionality of TCA cycling.
13
C atoms derived from pyruvate dehydrogenase activity or fatty acid oxidation-
derived acetyl-CoA will lead to an m/z + 2 mass shift in citrate, as two heavy carbon
atoms are incorporated in the latter. On the contrary, an m/z + 3 mass shift is
observed when pyruvate integrates into oxaloacetate and subsequently into citrate.
Similarly, when providing cells with 13C-glutamine, mass shifts of m/z + 4 and
174 M. Forbes et al.

m/z + 5 in citrate are the result of anaplerotic fuelling from glutamine into the TCA
cycle and oxidative or reductive cycling, respectively.
Other examples of pathway interconnection are pyruvate and glutamine, as both
can be metabolised into very different intermediates and play a role in distinct
cellular functions. Pyruvate-derived 13C-atoms can integrate, as described above,
into citrate or oxaloacetate, but also into alanine or lactate. Thereby, metabolic flow
through pyruvate merges amino acid metabolism, anaerobic glycolysis and TCA
cycle. Glutamine, on the other hand, is an important nitrogen donor and is involved
in amino acids metabolism and de novo biosynthesis of purine and pyrimidine
nucleotides (Bayram et al. 2020). Additionally, via glutamic acid, the carbon
backbone of glutamine feeds into TCA cycle, as well as polyamine and glutathione
synthesis. The example of glutamine highlights the necessity of dual carbon and
nitrogen labelling in order to greatly increase our understanding of pathway connec-
tivity and metabolic fate. Combining 13C- and 15N-labelling allows the follow-up of
glutamine usage, while distinguishing between the amino group, amido group and
carbon backbone utilisation.

2.4 The Application of Ultra-High Resolution Mass Spectrometry

Allows the Tracing of Different Isotopic Species

Using stable isotopes of different elements allows to determine the utilisation of two
metabolic precursors (e.g. glucose and glutamine) in a single experiment simulta-
neously, or to follow the fate of different atoms from the same precursors (e.g. C and
N from glutamine) into the downstream metabolic network.
For a long time such experiments have not been possible, since mass analysers
with nominal mass resolution are not able to distinguish the mass increase due to the
presence of different elemental isotopes. For example, the difference in the mass
increase given by one 13C or one 15N is only 0.00632 Da (Fig. 6). According to the
IUPAC definition (McNaught and Wilkinson 2008), the resolution needed to sepa-
rate such mass difference for two molecules with nominal mass of 400 Da is around
60,000, far beyond the resolution of conventional quadrupole or ion trap analysers
and barely reached by Time-of-Flight instruments. Only with the advent of (ultra)
high resolution Fourier transform mass spectrometry mass analysers it became
possible to analyse the isotopic fine structure of labelled compounds (Werner et al.
2008; Marshall and Hendrickson 2008).
One implementation of (ultra) high resolution MS for isotope tracing is the direct
infusion of the metabolites in an Fourier-transform ion cyclotron resonance mass
spectrometer (Le et al. 2012; Yang et al. 2017). This approach offers excellent results
in terms of achievable resolution and accuracy in the determination of the ratios
between different isotopes, because the ion signal in the ion cyclotron resonance cell
can be averaged for long times. However, in direct infusion highly abundant analytes
can suppress the ionisation of low abundant ones (Han et al. 2008). In addition, in
absence of another mean of separation, isomers cannot be distinguished because they
have the same accurate m/z ratio, like G6P and F6P. MS/MS experiments can help
 Advancements in Pulsed Stable Isotope-Resolved Metabolomics

Fig. 6 MS spectrum of 15N labelled Glutamine. The ion at m/z 271.16635 corresponds to the unlabelled fragment of TBDMS-derivatised glutamine with
molecular formula C12H27ON2Si2. It contains 4 carbon and all nitrogen atoms of glutamine. Insert shows the magnification of the m/z + 1 peaks, revealing the
separation of peaks due to the incorporation of 15N or 13C. The peak of 29Si isotope, present in the derivatisation agent, is also resolved. The masses reported in
the insert are the Δm relative to the monoisotopic peak
175
176 M. Forbes et al.

confirming or excluding the presence/absence of one or more isomers. Nevertheless,

isomers often require different collision energies, making an accurate quantification
difficult, if not impossible. Orbitrap mass analysers also can achieve high resolution
with fast scanning rate (Makarov et al. 2006). With a scan rate up to 40 Hz (Kelstrup
et al. 2018), this analyser is fast enough to be seamlessly hyphenated with GC or LC
separation.

3 Applications and Future Perspectives

In line with the general “OMICS” concept, metabolomics aims to measure all
metabolic components of a biological system at once in a quantitative manner.
Unfortunately that’s not possible so far. The chemical space of biomolecules regard-
ing polarity and size as well as their quantities – ranging from a few molecules up to
millimolar concentrations – is hindering its simultaneous detection. One step further
is the measurement of the dynamics of metabolism. The combination of stable
isotope labelling and mass spectrometric detection made a big contribution in this
regard. More and more methods allow the simultaneous measurement of metabolite
concentrations and isotope incorporation (Pietzke et al. 2014). Such integrated
methods are a prerequisite when metabolic dynamics in vivo or in tissues will be
analysed in a clinical context. Often a consecutive sampling and parallel quantifica-
tion in additional samples is not possible. The application of pSIRM together with
the present knowledge of the biochemical network allows to define certain metabolic
nodes that contain superior information of the usage of metabolic pathways,
e.g. citrate. Future method development may focus on such nodes, allowing for
more directed pSIRM applications with tailored isotopically labelled substrates.

3.1 Applications of pSIRM

In the last years, pSIRM has been applied to successfully analyse metabolic changes
within short and defined windows, e.g. during differentiation of cells. Delp et al.
found that immature and mature neurons rely on different fuels. Precursor cells were
found to be mainly glycolytic and strongly dependent on glutamine. During differ-
entiation, however, they lost their glutamine dependency while gaining flexibility in
energy production (Delp et al. 2018). By analysing the glutamine-derived carbon
utilisation in high salt-treated differentiating macrophages a specific down-
regulation of the succinate to fumarate conversion could be revealed. This observa-
tion led to the identification of sodium as strong regulator of complex II activity
(Geisberger et al. 2021).
pSIRM can also be applied in vivo as well as ex vivo in tissue slices or organs. By
administering 13C6-glucose into the peritoneum of a hepatocellular carcinoma mouse
model, Berndt et al. described enhanced glycolytic rates in tumours compared to
normal liver. These data contributed to the creation of individualised metabolic
profiles of tumours and modelling predictions of the efficacy of drug therapies
Advancements in Pulsed Stable Isotope-Resolved Metabolomics 177

(Berndt et al. 2020). Similarly, application of stable isotopes in situ to cancerous and
non-cancerous lung tissue revealed higher activity of glycolysis and the Krebs cycle
in the tumour tissue (Fan et al. 2009). In brain tissue slices from naked mole rats,
e.g. labelling with 13C6-fructose uncovered the ability of these animals to metabolise
fructose as fuel for neuronal tissues under hypoxic conditions (Park et al. 2017).
Beside the application of stable isotope labelled organic compounds, inorganic
compounds such as CO2 can be used for labelling experiments. For example,
Rohwer et al. used 13CO2 to demonstrate the reductive carboxylation in gastric
cancer cells (Rohwer et al. 2016). These are only few examples of the wide range
of applications pSIRM; a very versatile technique to describe metabolism in function
of quantity and time.

3.2 Perspective Towards Single Cells

The advent of single cell “OMICS” has allowed for the analysis of the architecture of
heterogenous cell samples at single cell resolution. Currently, single cell proteomics,
genomics and transcriptomics dominate the single cell technology landscape, while
single cell metabolomics is still in its infancy. The discrepancy of single cell
metabolomics is mainly due to the hardware of mass spectrometry being incompati-
ble with single cell resolution. A mammalian cell contains roughly 1 pL of analytical
volume. Due to this extremely small sample volume GC-MS methods often use
around two million cells per analysis. To further add complexity, metabolism is
highly dynamic – temporally and spatially – in tissue. Therefore, acquisition of
single cells and their processing presents many challenges in the context of meta-
bolic profiling.
Firstly, we will address the process of sampling single cells. The sorting of cells
from a sample is essential to single cell analysis. Fluorescence-activated cell sorting
(FACS) employs fluorescent labels to sort cells of specified origin. FACS may be
coupled to mass spectrometry methods to assess the profiles of sorted cells. How-
ever, the FACS process may interfere with metabolic profiles and therefore FACS
coupled with MS is more suited to proteomic analysis (Bandura et al. 2009). To
minimise the sampling time methods such as matrix-assisted laser desorption/
ionisation (MALDI) employ a laser to ablate cells and their metabolites in situ.
Following laser ablation the ionised metabolites are analysed via MS. Advances
towards single cell resolution have been reported. By using a combination of
computational imaging techniques and nuclei staining, the points of laser ablation
during MALDI-MS analysis can be inferred as metabolite acquisition from a single
cell (Rappez et al. 2021). However, this method is not quantitative and is biased to
high abundant metabolites and proteins.
Current technologies are aiming to sample single cells through microfluidic
trapping coupled with acquisition of intracellular metabolites through micro
capillaries. However, efforts to move to a microfluidic environment present their
own complications, mainly due to maintaining homeostatic environment in a
miniaturised platform (Ali et al. 2019).
178 M. Forbes et al.

In conclusion, as mass spectrometers become more sensitive coupled with inven-

tive single cell trapping techniques, metabolomics will be a valid and useful addition
to the single cell “OMICS” universe.

References
Ali A, Abouleila Y, Shimizu Y, Hiyama E, Emara S, Mashaghi A, Hankemeier T (2019) Single-cell
metabolomics by mass spectrometry: advances, challenges, and future applications. TrAC
Trends Anal Chem 120:115436., ISSN 0165-9936. https://doi.org/10.1016/J.TRAC.2019.
02.033
Bandura DR, Baranov VI, Ornatsky OI, Antonov A, Kinach R, Lou X, Pavlov S, Vorobiev S, Dick
JE, Tanner SD (2009) Mass cytometry: technique for real time single cell multitarget immuno-
assay based on inductively coupled plasma time-of-flight mass spectrometry. Anal Chem
81(16):6813–6822., ISSN 00032700. https://pubs.acs.org/doi/full/10.1021/ac901049w. https://
doi.org/10.1021/AC901049W
Bayram S, Fürst S, Forbes M, Kempa S (2020) Analysing central metabolism in ultra-high
resolution: at the crossroads of carbon and nitrogen. Mol Metab 33:38–47., ISSN 2212-8778.
https://doi.org/10.1016/J.MOLMET.2019.12.002
Berndt N, Egners A, Mastrobuoni G, Vvedenskaya O, Fragoulis A, Dugourd A, Bulik S, Pietzke M,
Bielow C, van Gassel R, Damink SW, Erdem M, Saez-Rodriguez J, Holzhütter HG, Kempa S,
Cramer T (2020. ISSN 15321827. https://www.nature.com/articles/s41416-019-0659-3)
Kinetic modelling of quantitative proteome data predicts metabolic reprogramming of liver
cancer. Br J Cancer 122(2):233–244. https://doi.org/10.1038/s41416-019-0659-3
Boros LG, Puigjaner J, Cascante M, Lee WNP, Brandes JL, Bassilian S, Yusuf FI, Williams RD,
Muscarella P, Melvin WS, Schirmer WJ (1997) Oxythiamine and dehydroepiandrosterone
inhibit the nonoxidative synthesis of ribose and tumor cell proliferation. Cancer Res. ISSN
00085472
Bruntz RC, Lane AN, Higashi RM, Fan TW-M (2017) Exploring cancer metabolism using stable
isotope-resolved metabolomics (SIRM). J Biol Chem 292(28):11601–11609., ISSN 0021-9258.
http://www.jbc.org/article/S0021925820370289/fulltext. http://www.jbc.org/article/S002192
5820370289/abstract. https://doi.org/10.1074/JBC.R117.776054
Buescher JM, Antoniewicz MR, Boros LG, Burgess SC, Brunengraber H, Clish CB, DeBerardinis
RJ, Feron O, Frezza C, Ghesquiere B, Gottlieb E, Hiller K, Jones RG, Kamphorst JJ, Kibbey
RG, Kimmelman AC, Locasale JW, Lunt SY, Maddocks OD, Malloy C, Metallo CM, Meuillet
EJ, Munger J, Nöh K, Rabinowitz JD, Ralser M, Sauer U, Stephanopoulos G, St-Pierre J,
Tennant DA, Wittmann C, Heiden MGV, Vazquez A, Vousden K, Young JD, Zamboni N,
Fendt SM (2015) A roadmap for interpreting 13C metabolite labeling patterns from cells. Curr
Opin Biotechnol 34:189–201. ISSN 0958-1669. https://doi.org/10.1016/J.COPBIO.2015.
02.003
Delp J, Gutbier S, Cerff M, Zasada C, Niedenführ S, Zhao L, Smirnova L, Hartung T,
Borlinghaus H, Schreiber F, Bergemann J, Gätgens J, Beyss M, Azzouzi S, Waldmann T,
Kempa S, Nöh K, Leist M (2018) Stage-specific metabolic features of differentiating neurons:
implications for toxicant sensitivity. Toxicol Appl Pharmacol 354:64–80. ISSN 1096-0333.
https://pubmed.ncbi.nlm.nih.gov/29278688/. https://doi.org/10.1016/J.TAAP.2017.12.013
Fan TW, Lane AN, Higashi RM, Farag MA, Gao H, Bousamra M, Miller DM (2009) Altered
regulation of metabolic pathways in human lung cancer discerned by 13C stable isotope-
resolved metabolomics (SIRM). Mol Cancer 8(1):1–19. ISSN 14764598. https://molecular-
cancer.biomedcentral.com/articles/10.1186/1476-4598-8-41. https://doi.org/10.1186/1476-
4598-8-41/TABLES/6
Geisberger S, Bartolomaeus H, Neubert P, Willebrand R, Zasada C, Bartolomaeus T, McParland V,
Swinnen D, Geuzens A, Maifeld A, Krampert L, Vogl M, Mähler A, Wilck N, Markó L, Tilic E,
Advancements in Pulsed Stable Isotope-Resolved Metabolomics 179

Forslund SK, Binger KJ, Stegbauer J, Dechend R, Kleinewietfeld M, Jantsch J, Kempa S,

Müller DN (2021. ISSN 1524-4539. https://pubmed.ncbi.nlm.nih.gov/33906377/) Salt tran-
siently inhibits mitochondrial energetics in mononuclear phagocytes. Circulation 144(2):
144–158. https://doi.org/10.1161/CIRCULATIONAHA.120.052788
Han J, Danell R, Patel J, Gumerov D, Scarlett C, Speir J, Parker C, Rusyn I, Zeisel S, Borchers C
(2008) Towards high-throughput metabolomics using ultrahigh-field Fourier transform ion
cyclotron resonance mass spectrometry. Metabolomics 4(2):128–140. ISSN 1573-3882.
https://pubmed.ncbi.nlm.nih.gov/19081807/. https://doi.org/10.1007/S11306-008-0104-8
Hevesy G, Hofer E (1934) Elimination of water from the human body. Nature 134(3397):879–879.,
ISSN 1476-4687. https://www.nature.com/articles/134879a0. https://doi.org/10.1038/
134879a0
Jang C, Chen L, Rabinowitz JD (2018) Leading edge primer metabolomics and isotope tracing. Cell
173:822–837. https://doi.org/10.1016/j.cell.2018.03.055
Kelstrup C, Bekker-Jensen D, Arrey T, Hogrebe A, Harder A, Olsen J (2018) Performance
evaluation of the Q exactive HF-X for shotgun proteomics. J Proteome Res 17(1):727–738.
ISSN 1535-3907. https://pubmed.ncbi.nlm.nih.gov/29183128/. https://doi.org/10.1021/ACS.
JPROTEOME.7B00602
Le A, Lane AN, Hamaker M, Bose S, Gouw A, Barbi J, Tsukamoto T, Rojas CJ, Slusher BS,
Zhang H, Zimmerman LJ, Liebler DC, Slebos RJ, Lorkiewicz PK, Higashi RM, Fan TW, Dang
CV (2012) Glucose-independent glutamine metabolism via TCA cycling for proliferation and
survival in b cells. Cell Metab 15(1):110–121. ISSN 15504131. https://pubmed.ncbi.nlm.nih.
gov/22225880/. https://doi.org/10.1016/j.cmet.2011.12.009
Makarov A, Denisov E, Kholomeev A, Balschun W, Lange O, Strupat K, Horning S (2006)
Performance evaluation of a hybrid linear ion trap/orbitrap mass spectrometer. Anal Chem
78(7):2113–2120. https://doi.org/10.1021/AC0518811. https://pubs.acs.org
Marshall A, Hendrickson C (2008) High-resolution mass spectrometers. Annu Rev Anal Chem
1(1):579–599. https://doi.org/10.1146/annurev.anchem.1.031207.112945. https://www.
annualreviews.org/
McNaught D, Wilkinson A (2008) Mass resolving power in mass spectrometry. IUPAC Compend
Chem Terminol. https://doi.org/10.1351/GOLDBOOK.M03730
Myers W (1996) Georg charles de hevesy: the father of nuclear medicine. J Nucl Med Technol 24:
291–294
Park TJ, Reznick J, Peterson BL, Blass G, Omerbašić D, Bennett NC, Kuich PHJ, Zasada C, Browe
BM, Hamann W, Applegate DT, Radke MH, Kosten T, Lutermann H, Gavaghan V,
Eigenbrod O, Bégay V, Amoroso VG, Govind V, Minshall RD, Smith ESJ, Larson J,
Gotthardt M, Kempa S, Lewin GR (2017) Fructose-driven glycolysis supports anoxia resistance
in the naked mole-rat. Science 356(6335):307–311. ISSN 1095-9203. https://pubmed.ncbi.nlm.
nih.gov/28428423/. https://doi.org/10.1126/SCIENCE.AAB3896
Pietzke M, Zasada C, Mudrich S, Kempa S (2014) Decoding the dynamics of cellular metabolism
and the action of 3-bromopyruvate and 2-deoxyglucose using pulsed stable isotope-resolved
metabolomics. Cancer Metab 2(1):9. https://doi.org/10.1186/2049-3002-2-9
Qiu Y-Q (2013) KEGG pathway database. In: Encyclopedia of systems biology. Springer,
New York, pp 1068–1069. https://doi.org/10.1007/978-1-4419-9863-7_472
Rappez L, Stadler M, Triana S, Gathungu RM, Ovchinnikova K, Phapale P, Heikenwalder M,
Alexandrov T (2021) SpaceM reveals metabolic states of single cells. Nat Methods 18(7):
799–805. ISSN 1548-7105. https://www.nature.com/articles/s41592-021-01198-0. https://doi.
org/10.1038/s41592-021-01198-0
Rohwer N, Bindel F, Grimm C, Lin SJ, Wappler J, Klinger B, Blüthgen N, Du Bois I, Schmeck B,
Lehrach H, De Graauw M, Goncalves E, Saez-Rodriguez J, Tan P, Grabsch HI, Prigione A,
Kempa S, Cramer T (2016. ISSN 19492553. https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4
872743/) Annexin A1 sustains tumor metabolism and cellular proliferation upon stable loss of
HIF1A. Oncotarget 7(6):6693. https://doi.org/10.18632/ONCOTARGET.6793
180 M. Forbes et al.

Sauer U (2006) Metabolic networks in motion: 13C-based flux analysis. Mol Syst Biol 2(1):62.,
ISSN 1744-4292. https://onlinelibrary.wiley.com. https://doi.org/10.1038/MSB4100109
Werner E, Heilier J, Ducruix C, Ezan E, Junot C, Tabet J (2008. ISSN 1570-0232. https://pubmed.
ncbi.nlm.nih.gov/18672410/) Mass spectrometry for the identification of the discriminating
signals from metabolomics: current status and future trends. J Chromatogr B Analyt Technol
Biomed Life Sci 871(2):143–163. https://doi.org/10.1016/J.JCHROMB.2008.07.004
Wilikinson D (2018) Historical and contemporary stable isotope tracer approaches to studying
mammalian protein metabolism. Mass Spectrom Rev 37(1):57–80. ISSN 1098-2787. https://
pubmed.ncbi.nlm.nih.gov/27182900/
Yang Y, Fan T, Lane A, Higashi RM (2017) Chloroformate derivatization for tracing the fate of
amino acids in cells and tissues by multiple stable isotope resolved metabolomics (mSIRM).
Anal Chim Acta 976:63–73. ISSN 1873-4324. https://pubmed.ncbi.nlm.nih.gov/28576319/.
https://doi.org/10.1016/J.ACA.2017.04.014
Metabolomics in Cell Biology

Zuhal Eraslan, Marta Cascante, and Ulrich L. Günther

Contents
1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 182
1.1 Omics Context . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 182
1.2 History of Metabolomics Using Cellular Samples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 183
2 Metabolomics Approaches: Targeted vs. Untargeted . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 184
2.1 Sample Types . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 186
2.2 Preparation of Cell Extracts for Metabolomics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 186
2.3 Metabolomics Technologies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 188
2.3.1 NMR-Based Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 189
2.3.2 MS-Based Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 191
3 Case Studies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 193
3.1 Examples of Cancer Metabolomics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 193
3.2 Real-Time Measurements of Cells . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 195
3.3 Metabolic Studies in Mycobacteria . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 197
3.4 Metabolic Fluxes by NMR and GC-MS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 198
3.5 Examples for Fluxes in Cancer Cells by GC-MS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 199
3.6 Fluxes and Metabolic Modelling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 201
3.7 Metabolic Fluxes in Liver Disease . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 201

Z. Eraslan
Department of Dermatology, Weill Cornell Medicine, New York, NY, USA
M. Cascante
Department of Biochemistry and Molecular Biomedicine, University of Barcelona, Barcelona,
Spain
Institute of Biomedicine of University of Barcelona (IBUB), University of Barcelona, Barcelona,
Spain
CIBER of Hepatic and Digestive Diseases (CIBEREHD), Institute of Health Carlos III (ISCIII),
Madrid, Spain

U. L. Günther (✉)
Institute of Chemistry and Metabolomics, University of Lübeck, Lübeck, Germany
e-mail: [email protected]

# The Author(s), under exclusive license to Springer Nature Switzerland AG 2022 181
V. Ghini et al. (eds.), Metabolomics and Its Impact on Health and Diseases,
Handbook of Experimental Pharmacology 277, https://doi.org/10.1007/164_2022_619
182 Z. Eraslan et al.

4 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 202
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 203

Abstract

Metabolomics has long been used in a biomedical context. The most typical
samples are body fluids in which small molecules can be detected and quantified
using technologies such as Nuclear Magnetic Resonance (NMR) and Mass
Spectrometry (MS). Many studies, in particular in the wider field of cancer
research, are based on cellular models. Different cancer cells can have vastly
different ways of regulating metabolism and responses to drug treatments depend
on specific metabolic mechanisms which are often cell type specific. This has led
to a series of publications using metabolomics to study metabolic mechanisms.
Cell-based metabolomics has specific requirements and allows for interesting
approaches where metabolism is followed in real-time. Here applications of
metabolomics in cell biology have been reviewed, providing insight into specific
technologies used and showing exemplary case studies with an emphasis towards
applications which help to understand drug mechanisms.

Keywords
Cell biology · Metabolomics

1 Introduction

1.1 Omics Context

Metabolomics is the omics-science that analyses the small molecule compartment of

biological samples. Originally, Nicholson coined the term metabonomics as the
science that studies responses of living systems to metabolic changes (Nicholson
et al. 1999). What distinguishes metabolomics from conventional analytical
approaches is the way data are analysed, looking at many components of a sample
at the same time in an untargeted manner, often in a high-through-put manner with
subsequent statistical data analysis. Metabolomics is now often found in the context
of other omics technologies, as part of multi-omics workflows, usually involving
computational systems biology to integrate the different types of data. On the other
hand, targeted analyses have been used to study mechanisms based on biological
hypotheses in a targeted manner, often using isotopically labelled metabolic
precursors as tracers.
Overall, metabolomics has become a broadly used technology in many fields
of biomedical and nutritional sciences (Wishart 2016). Biomedical analyses
are often based on body fluids with the goal to derive biomarkers. This often
involves large sample numbers and high-throughput approaches and is increasingly
linked to genetic and epigenetic features. Wishart, who has established the most
Metabolomics in Cell Biology 183

comprehensive repositories and online tools for metabolomics, has recently

discussed the relevance of metabolomics to study disease mechanisms, to customise
treatments and monitor treatment response and to identify new drugs (Wishart
2016). The expression phenotyping is often used, considering that metabolomics
links genomics, epigenomics and transcriptomics with biological phenotypes.

1.2 History of Metabolomics Using Cellular Samples

Besides the use of metabolomics in a medical context, often with the intent to find
biomarkers, there are an increasing number of applications in biological studies with
the goal to identify specific mechanisms in cells. When looking at such applications
of metabolomics in cellular systems the boundaries between targeted and
non-targeted analysis are often more fluid. As a matter of fact, some of the earliest
applications of what would now be called metabolomics were conducted in cellular
systems. As early as 1978, Ugurbil and Shulman used NMR to study metabolic
turnover in Escherichia coli cells (E. coli), first by 31P-NMR and soon afterwards by
13
C-NMR. The first report used 31P-NMR to measure glycolysis rates in E. coli, and
also looked at the effect of ATPase inhibition (Ugurbil et al. 1978a). Even at a
spectrometer with 360 MHz proton frequency, a 2 min time resolution was achieved.
In a second publication, E. coli cells were incubated with [1-13C]-labelled glucose in
an NMR tube and followed the turnover of glucose via fructose bisphosphate into
amino acids (alanine and valine) along with lactate, succinate and acetate using
13
C-NMR spectra (Ugurbil et al. 1978b). Upon oxygenation, the authors also
observed glutamate, labelled at the C-4, C-3 and C-2 positions. Time-courses were
reported over up to 3 h with a time resolution of 1 min, which is remarkable
considering that the NMR instrumentation in 1978 was at 90 MHz proton frequency.
This early work demonstrates the advantage of NMR to identify site-specific label
incorporation. In a subsequent publication from Shulman’s group, rat hepatocytes
were used to quantitatively assess the metabolic flux of glycerol in gluconeogenesis,
showing label incorporation in several glycolytic intermediates and products (Cohen
et al. 1979) and to understand whether or to what extent hyperthyroid cells (after
pre-treatment of rats with triiodothyronine hormone (T3)) increases the rate of
glycerol consumption and glucose formation. In this work not only signal intensities
were used but also scalar couplings between 13C nuclei to quantitatively assess label
incorporation. In hyperthyroid cells labelling was also reported in lactate, alanine,
aspartate, glutamate and ketone bodies. By analysing time-courses of label
incorporation in different positions of glucose from differently labelled glycerol
precursors pentose phosphate pathway activity could be determined, along with
activities of transaldolase and transketolase. These early findings which provide a
historical background for cell-based metabolic studies are confirmed by more recent
work taking similar approaches to measure fluxes by NMR (Jin et al. 2013).
Soon afterwards, Jardetzky and co-workers described a system for continuous-
flow monitoring of metabolism in mammalian cells (Chinese hamster ovarian cells)
184 Z. Eraslan et al.

(Gonzalez-Mendez et al. 1982). These seminal experiments laid the ground for using
NMR to look at metabolism at a time when nobody thought about metabolomics.
They exemplify the virtues of NMR for examining metabolism in cellular systems,
which include real-time analysis of intact cells, using isotopic labels as tracers and
the ability to observe site-specific label incorporation in molecules. Although any
analysis that uses labelled precursors is by definition not untargeted, NMR was
always used to simultaneously detect many metabolites in one NMR spectrum.
Another milestone in this development has come from Szyperski et al. (Szyperski
1995; Szyperski et al. 1996) in 1995, using 1H-13C_HSQC spectra to analyse cellular
label incorporation. Today HSQC spectra are often preferred as they offer higher
sensitivity than13C-observed spectra. Szyperski was able to make sense of highly
complex JCC couplings patterns which arise from mixtures of isotopomers. Today’s
most advanced NMR systems work at 1.2 GHz, commonly used metabolomics
systems work at 600 MHz. Cryoprobes and other improvements of NMR technology
have boosted the sensitivity of NMR experiments by at least an order of magnitude.
There are now cryoprobes available that have been optimised for mass limited
samples which provide immense sensitivity advantages for cell-based studies
(Saborano et al. 2019). Microprobes hold great promise to work with very few
cells (Finch et al. 2016).
Here we review approaches of metabolomics in cells and present the methods
used, with a specific focus on advanced NMR methods. Exemplary case studies are
presented which demonstrate how metabolomics can be used in cell-based
investigations in the wider context of drug discovery.

2 Metabolomics Approaches: Targeted vs. Untargeted

Metabolomics can be carried out at different levels and using very different
approaches (Fig. 1). Methods are typically classified as targeted or non-targeted. In
early metabolomics non-targeted fingerprinting was commonly used where
“features” or raw spectra were employed without specific assignments, often using
multivariate statistical approaches to identify the most relevant features for given
classes (Nicholson et al. 1999). The general advantage of fingerprinting is that data
can be left unassigned. For NMR most initial work has used raw spectra for
subsequent statistical, often multivariate analyses (Nicholson et al. 1999), although
this approach is becoming increasingly uncommon as metabolites are now easily
assigned in many types of samples, including cell extracts where it is extremely well-
known which signals are unique identifiers for individual metabolites. In the case of
GC-MS, the assignments of the signals are usually also well-known thus favouring
targeted approaches. This has included tracer-based approaches for which GC-MS is
equally well suited (Bruntz et al. 2017). Employing typical LC-MS methods,
thousands of features can be detected which can either be used in univariate or
multivariate statistical analyses in an untargeted manner. Alternatively in targeted
approaches features are assigned and only fully assigned signals are used for further
Metabolomics in Cell Biology 185

Fig. 1 Approaches towards metabolomics

analysis. In the special case of direct-injection MS, only a small percentage of

signals can be uniquely assigned which makes the method better suitable for
fingerprinting.
In today’s applications of metabolomics in cell-based biological systems, targeted
approaches have become much more common, although we also find high-
throughput studies without assigning all signals. Whenever hypothesis-driven
projects aim to decipher metabolic mechanisms in cells, targeted approaches are
needed. This has often also included tracer-based metabolomics using 13C- or
15
N-labelled metabolic precursors to study the fate of metabolites along often
complex metabolic pathways. In the specific case of flux analysis, time-courses are
used to derive kinetic data, or at least to determine the direction of metabolic fluxes
(Kempa, “Advancements in pulsed stable isotope re-solved Metabolomics”). Com-
monly used isotopically labelled metabolic precursors have been glucose and gluta-
mine, but there have also been reports using other metabolites (reviewed in
(Saborano et al. 2019; Cascante and Marin 2008; Jang et al. 2018)). To a certain
degree, it is also possible to obtain fluxes by looking at rates of consumption and
production of metabolites in media. The expression “metabolic flux” is also often
also used for modelling metabolic pathways using tracer-based data (Selivanov et al.
2006, 2020).
186 Z. Eraslan et al.

2.1 Sample Types

When working with cellular systems or organelles, there is a wide array of possible
samples. Primary cultures are directly taken from organisms and can be subject to
investigation for at least a short time. Cell lines arise from primary cells which have
been immortalized. Cell cultures are available from public resources, such as ATCC
(http://www.atcc.org), Coriell (http://ccr.coriell.org), DSMZ (http://www.dsmz.de),
ECACC (https://www.culturecollections.org.uk), JCRB (https://cellbank.nibiohn.
go.jp) or RIKEN Bioresource centre (https://web.brc.riken.jp/en), for which robust
quality control protocols have been established (Yu et al. 2015). Most studies use
immortalised cell lines which can be readily grown to obtain sufficient amounts of
metabolites. Cell extracts represent the most common type of sample, but it is also
possible and often reasonable to measure metabolite concentrations in the medium in
which cells were grown, which can be used to calculate turnover for individual
nutrients. Media-sampling can also be carried out in bioreactors where metabolic
turnover can be directly detected. Moreover, NMR is also suitable for non-invasive
analysis of metabolism in living cells and organelles which has led to several
interesting studies focussed on specific metabolic mechanisms.

2.2 Preparation of Cell Extracts for Metabolomics

Several recent publications have reported protocols for the preparation of

metabolomic samples from cells (Bhinderwala and Powers 2019; Halama 2014;
Dietmair et al. 2012; Rais et al. 1999). A consensus protocol has been summarised in
Fig. 2. Whether primary cells, or certain strains of cells or cell lines are used is less
relevant for metabolomics sample preparation than the way the cells are grown. The
protocol is different for cells grown in suspension or adherently on an inert surface
(typically plastic or glass). Among mammalian cells, it is mainly haematological
cancer cells which are grown in suspension. In the case of adherent cells it is useful
to wash cells with ice cold buffer (usually PBS) as the first step of the preparation for
metabolomics extraction. For suspension cells we wash with prewarmed PBS and
then transfer cells into Eppendorf tubes. Suspension cells can be harvested by
centrifugation or by filtration, in our lab we prefer centrifugation at 4°C as filters
need to be washed to avoid contamination with small molecules from the filter. Cells
grown adherently need to be removed from the surface on which they were grown.
This can be achieved by scraping them off the surface or by trypsinization, where the
latter is generally more common, as it is easier to be carried out reproducibly.
Metabolism needs to be quenched by cooling cells rapidly, which is often carried
out with precooled methanol and/or acetonitrile. Other authors prefer sonication
and/or submerging cells in liquid nitrogen. Subsequently, chloroform can be added
to obtain two phases. Often only the aqueous phase is used but it is also possible to
analyse the lipids from the chloroform layer. Precipitated proteins are found at the
interface between the two phases and the polar and non-polar extracts can readily be
obtained with a syringe. We found it to be important to use glass vials from the first
Metabolomics in Cell Biology 187

Adherent cell lines Suspension cell lines

Grow cells to 80% confluency Cells in exponential growth until 106

Preparation

Wash cells 2-3 times with ice cold PBS to Centrifuge at 1500rpm, for 5 minutes, at
remove residual media 21° C and remove supernatant

Quickly wash cells just once with PBS at

ca.37° C to remove residual media

Centrifuge at 1400 RPM, for 10 seconds at

21° C and remove supernatant

Add 400 μL of methanol to stop metabolism

Scrape cells off the flask and transfer to a Vortex cells in methanol and transfer to a
glass vial glass vial

Add 200 μL of chloroform and vortex for Add 325 μL of water and 400 μL of
30 seconds chloroform
Extraction

Add a further 200 μL of chloroform and

300 μL of water.

Vortex for 30 seconds and put sample on ice for 10 minutes

Centrifuge for 10 minutes at 4000 RPM

Keep on the bench for 5 minutes (optional)

Transfer polar layer to an Eppendorf and non-polar layer to a glass vial

Fig. 2 Protocol for the preparation of metabolomics samples from cells

addition of methanol as tests with plastic vials showed compounds leaching from the
plastic with methanol and chloroform. More recently plastic tubes have become
available that are supposed to be chloroform proof.
Cell extracts recapitulate metabolic levels inside cells at one particular point of
time. Such cell extracts can be examined by NMR or by MS. If NMR is used, a
one-dimensional spectrum allows to reliably identify and quantify 30–40
188 Z. Eraslan et al.

metabolites using a 600 MHz spectrometer with a cryoprobe. The number of cells
required for such measurements depends on the cell size. For reference,
haematological cancer cells such as HL60 or K562, approximately 1–10 million
cells are required for one-dimensional spectra, depending on the specific type of
NMR experiments. For tracer-based flux experiments the number of cells required is
larger and should be at least 10 million for 1H-13C-HSQC spectra or isotope-filtered
spectra. For directly observed 13C-spectra even higher concentrations of metabolites
are required, although recently developed 13C-optimised micro-cryoprobes offer
huge sensitivity advantages with significant potential for metabolomics in cell-
based systems (Ramaswamy et al. 2013; Clendinen et al. 2014; Thomas et al.
2021). Most MS studies of metabolic mechanisms used GC-MS which is somewhat
more sensitive than NMR, >1 million cells are typically needed. LC-MS is much
more sensitive and can work with very small numbers of cells, very much depending
on the specific version of mass spectrometer that is used.

2.3 Metabolomics Technologies

As already introduced, the main technologies behind metabolomics are NMR and
MS, each with specific advantages (Verpoorte, “Natural products drug discovery: on
silica or in-silico?”; Evans, “Compound Identification Strategies in Mass
Spectrometry-Based Metabolomics and Pharmacometabolomics”; Raftery, “Quanti-
tative NMR Methods in Metabolomics; Wishart, “Practical Aspects of NMR-Based
Metabolomics”; McKay, “Metabolomics using NMR- avoiding the “Black-Box””.).
NMR methods are well established in drug discovery, mainly for studying the
interaction of proteins with small molecules (Meyer and Peters 2003; Becker et al.
2018; Ludwig and Guenther 2009), and there is a range of methods to study cells.
Table 1 lists the overall advantages of NMR and MS approaches along with specific
sensitivities. For biologically driven users, it is particularly relevant to understand
the level of application in a cellular context, translated into accessible metabolite
concentrations, cell numbers and types of samples that can be studied. NMR requires
millions of cells (assuming haematological cancer cells such as HL60 and K562 as a
reference) whereas MS based methods are substantially more sensitive. NMR has a
significant advantage for quantification and to detect small but relevant changes.
Other advantages of NMR are the possibility to look a living cells non-invasively
enabling the measurement of time-courses, and to identify site-specific label
incorporation in tracer-based experiments. On the other hand, MS offers several
orders higher sensitivity, enabling even single-cell metabolomics and providing
access to a much wider range of metabolites.
MS can now be applied on a single-cell level, even combined with imaging
(Rappez et al. 2021). Table 1 provides an overview over the different analytical
methods that can be used providing information, their applicability, advantages and
limitations. GC-MS has commonly preferred for tracer-based approaches and there
are well-established protocols for this (Cascante and Marin 2008).
Metabolomics in Cell Biology 189

Table 1 Metabolomics methods used for cellular systems

1
Technology H-NMR GC-MS LC-MS
Metabolites Water-soluble Water-soluble Water-soluble
metabolites; lipid fraction metabolites which metabolites; lipid
must be derivatized fraction for lipidomics
Sample types Whole cells, cell extracts, Cell extracts Cell extracts
media, organoids
Number of cellsa Min 10 million 5–10 million <1 million
Single cell
Run-time 1D spectrum: 10 min 10 min 10–15 min
Detection limit 5–10 μM <1 μM pM–nM
Number of 30–40 150 Thousands of features,
metabolites in ca. 250 assigned
cell-based metabolites
samples
Excellent reproducibility; Good No simple
concentrations can be reproducibility; quantification, except
quantified semi-quantitative for BIOCRATES
approach
Simple sample Sample extraction Sample extraction
preparation; HR-MAS can and derivatization required
be used without required
preparation
a
Cell numbers for typical haematological cancer reference cells (HL60 or K562)

2.3.1 NMR-Based Methods

NMR in the context of metabolomics has been the subject of previous reviews which
cover NMR methods and protocols used (Wishart 2016; Bhinderwala and Powers
2019; Halama 2014; Duarte et al. 2009; Powers 2009; ÄŒuperloviÄ 2010; Palmnas
and Vogel 2013; Zhang et al. 2013; Markley et al. 2017; Vignoli et al. 2019).
Usually, one-dimensional (1D) NMR spectra are used for metabolomics. For this,
a pulse sequence called 1D-NOESY is most commonly employed, mainly because it
allows for quantitative measurements with excellent suppression of the water signal.
The 30–40 metabolites which can be quantified in 1D NMR spectra from cell
extracts include amino acids, glutathione, taurine, several energy metabolites, such
as UDP, ATP, UDP-glucose, lactate, pyruvate, succinate, fumarate and
ɑ-ketoglutarate, a range of sugars, myo-inositol, cholesterol, fatty acids (typically
the CH, CH2 and CH3 groups), choline and phosphatidylcholine, phosphatidyletha-
nolamine and triglycerides. While this list is short compared to what MS can
observe, concentrations can be reproduced within 1–2% between repeats of cell
cultures as illustrated by Tiziani et al. (2009). Raftery and co-workers have used a
combination of 1D and 2D NMR methods to distinguish and quantify common
phosphorylated coenzymes such as AMP, ADP and ATP, NADH, NADPH, NAD+
and NADP+ (Gowda et al. 2016).
190 Z. Eraslan et al.

1D-spectra can also be used for absolute quantification of metabolites and this in a
highly reproducible manner. This can either be achieved by using an internal
standard such as Trimethylsilylpropanoic acid (TMSP) or by using an electronic or
software-generated reference signals using technologies which are now incorporated
in most spectrometers such as ERETIC and PULCON (Akoka et al. 1999; Watanabe
et al. 2016). These methods are now commonly used in NMR protocols for the
analysis of blood samples (Dona et al. 2014).
NMR also offers a large number of alternative pulse sequences which each yield
specific spectral features (reviewed in (Vignoli et al. 2019)). J-resolved (J-RES)
spectra are frequently used, in particular in samples with significant overlap between
signals. This is more common in plant-based samples than in samples arising from
mammalian cell lines. For the assignment of J-RES spectra, the Birmingham Metab-
olite Library (Ludwig et al. 2011) provides a unique resource. J-RES spectra are
usually processed in a manner where scalar couplings are removed thus providing
considerably simplified spectra. J-RES is also a quantitative method suitable to
calculate metabolite concentrations. Recently, a sophisticated 31P-selective J-RES
experiment has been used to identify several phosphometabolites from glycolysis
(Cox et al. 2021). While this method has limited sensitivity, it can distinguish
G6P/F6P or 2PG/3PG which is not possible using MS methods.
Many other NMR pulse sequences have been used to study metabolomics
samples. The most important ones are the TOCSY experiment which links chemi-
cally connected protons and HSQC spectra which correlate 1H and 13C resonances.
The two can also be combined in 2D-HSQC-TOCSY spectra which represent
invaluable tools for the assignment of metabolites in metabolomics. Importantly,
HSQC and HSQC-TOCSY spectra have been compiled in the COLMAR database
which also offers online tools for metabolite identification (Robinette et al. 2008;
Bingol et al. 2014, 2015a; Wang et al. 2020). HSQC-TOCSY spectra have also been
used in conjunction with MS for the assignment of unknown metabolites in an
approach that involves the simulation of NMR spectra using preliminary assignment
from MS (Bingol et al. 2015b). Although generally important for metabolomics, this
approach is less relevant for mammalian cells where the same 30–40 metabolites are
observed in almost all cell extracts. Although TOCSY spectra cannot be used to
derive absolute concentrations of metabolites, they have been used extensively for
the analysis of tracer-based metabolism, mainly by Fan and Lane (2008). HSQC
spectra offer considerable deconvolution of NMR spectra although at a much lower
sensitivity than 1D spectra and with a loss of quantification. The reason for this lies
in the nature of HSQCs which depend on scalar 1H-13C couplings (1JCH) which vary
considerably between metabolites. As a consequence, the signal intensity of a HSQC
spectrum depends not only on the concentration of a particular molecule but also on
the size of the coupling constant. Wan et al. suggested to determine these coupling
constants for a large number of metabolites and calculate the effect of1JCH on the
overall HSQC signal intensities (Wan et al. 2017), but this would require a signifi-
cant effort of collecting such a spectral database.
Nevertheless, HSQC spectra have been used in metabolomics, predominantly in
tracer-based metabolism (Saborano et al. 2019). For labelled samples, the effect of
Metabolomics in Cell Biology 191

the coupling constant on the HSQC intensity can be ignored when signals are
compared for the same metabolite for samples grown under different conditions.
In such cases, ratios between signals provide the relevant information on relative
label incorporations. Alternatively specific filtering methods have been employed to
quantify label incorporation. Such filters can be designed to become largely inde-
pendent of the JCH coupling constant (Reed et al. 2019).
A key advantage of NMR is that analyses are non-invasive, i.e. can be carried out
using living cells. This has been exploited in various real-time approaches where
metabolites have been measured over extended periods of time. Two main types of
real-time experiments have been reported. High-Resolution Magic Angle Spinning
(HR-MAS) stands for a technology which eliminates line broadening due to aniso-
tropic interactions in samples (Raftery, “Quantitative NMR methods in
metabolomics”). HR-MAS technologies for metabolomics have recently been
reviewed in (Edison et al. 2021). HR-MAS requires almost no sample preparation
and is non-destructive as long as cells can tolerate the high spinning forces.
HR-MAS enables the measurement of metabolites in intact tissue samples, small
animals (C. elegans (Blaise et al. 2007; Mobarhan et al. 2017) or drosophila (Sarou-
Kanian et al. 2015)) or cells. The need for MAS arises from the limited mobility of
molecules in semi-solid samples leading to broad lines in NMR spectra, either
caused by anisotropic dipole–dipole interactions or by local variations in magnetic
susceptibility. Both effects can be eliminated by rotating samples at the “magic
angle” of 54.7° to the magnetic field. This requires placing the sample in a small
rotor spinning at circa 6,000 Hz. For a 1.4 mm rotor at this spinning rate the sample
experiences 200,000 g acceleration which induces serious stress to the sample
(Edison et al. 2021). Recent developments may help to overcome this limitation,
with rotor synchronised pulse sequences it has been possible to obtain good spectra
of earth worms with as low as 100 Hz (Mobarhan et al. 2017). Micro-sized coils for
1 mm rotors were shown to further improve mass sensitivity (Lucas-Torres and
Wong 2019). HR-MAS has been used, for example, to study bacterial cells (Righi
et al. 2013), mycobacteria (Hanoulle et al. 2005a, 2006a; Lee et al. 2005) and also for
examination of mammalian cells (Nyblom et al. 2008; Garcia-Alvarez et al. 2011;
Gogiashvili et al. 2019; Judge et al. 2019; García-Álvarez et al. 2009; Vermathen
et al. 2021) and recently a cellular fungus (Edison et al. 2021). Although such
g-forces are seen as being too large for haematological cancer cells, Edison et al.
demonstrated applicability to chronic lymphoid leukaemia cells (Edison et al. 2021).
Recent works by Edison et al. were focussed towards real-time monitoring using
HR-MAS (Edison et al. 2021). As an alternative approach, cells have been used in
flow bioreactors (Hall et al. 2016) or have been embedded in matrices such as
agarose for real-time measurements (Koczula et al. 2016; Alshamleh et al. 2020).
Examples will be shown below.

2.3.2 MS-Based Methods

GC-MS has been used in the context of metabolomics for more than 50 years and
protocols have been established which can identify and quantify more than
100 metabolites form cell cultures (reviewed in (Fiehn 2016)). First applications of
192 Z. Eraslan et al.

GC-MS were reported in the 1960s covering sugars, amino acids, sterols, hormones,
hydroxyl acids, fatty acids, aromatics and many other small molecules (<650 Da), as
reviewed in (Fiehn 2016), before the term metabolomics had been coined. GC-MS
requires derivatization which has been established for a large range of small
molecules and is also well suited for metabolic flux analysis. The method benefits
greatly from large data bases (e.g. the NIST Mass Spectral Library (Babushok et al.
2007) and others reviewed in (Fiehn 2016)) and has played an important role in the
study of cellular metabolism and metabolic fluxes (Fischer et al. 2004). Unlike
NMR, GC-MS is not inherently quantitative but quantification is feasible using
external or internal standards (Fiehn 2016).
Most modern applications rely on LC-MS which does not require any derivatiza-
tion and is potentially truly global, covering thousands of features. LC-MS is a
widely used technology which is now available in many bioanalytical laboratories.
Unfortunately, there is much heterogeneity between LC-MS methods, different
upfront chromatographic approaches, different ionisation sources and different MS
hardware. The advantage of LC-MS for studying metabolomics in the context of
drug discovery is well illustrated by a recent workflow which combined
metabolomics, proteomics and transcriptomics datasets of 54 cancer cells to derive
a map of metabolite–transcriptional regulator interactions which was also used to
look at drug sensitivities (Ortmayr et al. 2019). Such studies typically involve a
systems biology layer that is used to combine large amounts of data. It should be
mentioned that direct flow-injection analysis time-of-flight mass spectrometry
(FIA-TOFMS) shows a lot of potential, in particular because measurements are
very fast, allowing for >5,000 samples to be run per day (Fuhrer et al. 2011) as
illustrated by a recent study by Ortmayr et al. (2019) who screened 54 cancer cell
lines linked with transcriptomic and proteomic data. This study reports 689 putative
assignments based on HMDB (Ortmayr et al. 2019). Typically, 250–350 metabolites
can be uniquely assigned using LC-MS, although not easily quantified.
For cell-based applications, the Biocrates kits approach has become popular as it
offers excellent quantification by using isotopically labelled reference compounds
(see (Illig et al. 2010) as an early example where this technology was used).
BIOCRATES kits now cover a wide range of hydrophilic and lipophilic metabolites
(Thompson et al. 2019) and offer a simple protocol for sample preparation.
Highly specialised MS metabolomics techniques have now reached a sensitivity
where single cells become amenable to investigation, even combined with imaging
(Rappez et al. 2021). This is particularly interesting to assess cancer cell heteroge-
neity and plasticity, manifested as metabolic heterogeneity. Several studies have
attempted to characterise metabolic heterogeneity in cancers (Rappez et al. 2021;
DeVilbiss et al. 2021; Lau et al. 2020), although this remains a challenging area of
MS research.
Metabolomics in Cell Biology 193

3 Case Studies

3.1 Examples of Cancer Metabolomics

Tiziani and Lodi started to extract cells in 2009 to profile drug responses in Acute
Myeloid Leukaemia (AML) cell lines through NMR spectroscopy (Tiziani et al.
2009). For this work, they established protocols for sample preparation and charac-
terization which yield highly reproducible NMR spectra. PCA shows narrow clusters
for HL-60, K562 and KG1a cell lines, perfectly separated from each other. Using
these AML cell lines a combination of two repurposed drugs, bezafibrate (BEZ) and
medroxyprogesterone acetate (MPA) was tested. This work has established the
mechanism of action of these drugs to be in part related to the generation of reactive
oxygen species (ROS) with a profound effect on cell viability. The mechanism
behind this is that high ROS causes chemically driven decarboxylation of ɑ-keto-
acids, converting ɑ-ketoglutarate into succinate, pyruvate to acetate and oxaloacetate
into malonate, a mechanism that has recently also been observed in acute myeloid
leukaemia cells interacting with stromal cells (Vilaplana-Lopera et al. 2021).
Tiziani et al. later developed this approach into a high-content drug screening tool
(Tiziani et al. 2011). For this, he developed a new protocol which combined growing
cells in 96-well plates with in situ lysing and quenching of metabolism via SDS
addition and sonication. The cell extracts were subsequently analysed by J-RES-
NMR spectra. Using this protocol, several inhibitors could be detected by monitor-
ing changes of the lactate/pyruvate ratio. Validation in cell lines and in primary
cancer cells was demonstrated. The key advantage of this approach is that it is not
based on a single read-out but rather a high-content metabolic measurement
reflecting increasingly well-understood metabolic changes in cells. Similarly, Lodi
showed in a cellular model how a combinatorial treatment using phytochemical
combinations affects prostate cancer cells (Lodi et al. 2017). In this work, flux
measurements were used to monitor changes in glutamine metabolism in response
to treatment.
Eraslan et al. have conducted metabolomic studies to reveal therapeutic metabolic
targets for germinal centre-derived Burkitt lymphoma (BL) and Diffuse large B-cell
lymphoma (DLBCL) by applying various 1D and 2D NMR techniques (Eraslan
et al. 2021). A principal component analysis (PCA) that was performed on the 1D
NMR spectra of the media showed a clear separation between the BL cells and
DLBCL (Fig. 3a). The loading plot of the corresponding PCA analysis depicted that
the separation of BL from DLBCL mostly derived from the difference in extracellu-
lar asparagine level of BL and DLBL (Fig. 3b, c). 13C-tracer-based NMR metabolic
analysis that was carried out with cell extracts derived from BL and DLBCL cells
cultured in a medium with or no asparagine containing 13C-stable-isotope labelled
tracers depicted that asparagine regulates the synthesis of serine from glucose
(Fig. 3d) and serine uptake (Fig. 3e) in both BL and DLBCL cells. By combining
a metabolomic platform with a transcriptomic platform, they uncovered that BL cells
express the genes involved in serine biosynthesis at a higher level than DLBCL cells
do (Fig. 3f) and defined a new treatment model which solely works for BL. They
194 Z. Eraslan et al.

Fig. 3 The role of extracellular asparagine in regulation of serine metabolism in BL and DLBCL.
(a) Principal Component Analysis (PCA) for 1D 1H-NMR for media samples of Burkitt lymphoma
cell lines (red) and diffuse large B-cell lymphoma cell lines (blue) (different symbols for different
types of cells). (b) Representation of corresponding loadings plot showing metabolites which
mostly contributed to separation of BL cells from DLBCL cells. (c) Relative 1D 1H-NMR peak
intensities for asparagine in the growth media of DLBCL and BL cells. (d) 2D 13C-NMR analysis of
the cell extraction of Glor and Farage cells cultured in a medium with or without asparagine cultured
with [U-13C]-glucose. (e) 2D 13C-NMR analysis of the cell extraction of Glor and Farage cells
cultured in a medium with or without asparagine cultured with [U-13C]-serine. (f) Heatmap of
statistically significant altered genes (FDR <0.01) associated with serine metabolism from differ-
ential expression analysis. (g) Analysis of 2D 13C-NMR spectra of serine extracted from Glor and
Farage cells cultured in a medium cultured [U-13C]-glucose treated with the PHGDH inhibitor
NCT-503 at 10 μM. (h) The viability of BL31 (BL cell line) and SUDHL6 (DLBCL cell line) cells
after treatment with ASNase at 0.1 U/ml and ASNase at 0.1 U/ml plus NCT-503 for 24, 48 and 72 h

firstly performed a metabolomic study to assess the metabolic effect of an inhibitor

(NCT503) that targets a rate-limiting enzyme PHGDH, in the serine biosynthesis
pathway utilising a 13C-tracer-based NMR metabolic approach (Fig. 3g). Then, they
combined asparaginase (ASNase), which is used to treat acute lymphoblastic leu-
kaemia (ALL) patients, at a very low dose with the PHGDH inhibitor. The combi-
nation of ASNase with the PHGDH inhibitor had a synergistic effect on cell viability
Metabolomics in Cell Biology 195

in BL cells while no synergistic effect on the viability of DLBCL cells (Fig. 3h).
Thus, the integrated multi-omic approach with metabolomics and transcriptomic
data has suggested an attractive new treatment model for BL.
To the best of our knowledge, the first application of HR-MAS NMR spectros-
copy to study cells is a publication by Nyblom et al. studying fatty acid formation
from [1-13C]glucose in insulinoma cell lines (Nyblom et al. 2008)the effect of
an antimitotic glycoside in glioma cells (García-Álvarez et al. 2009). Righi has
demonstrated the applicability of HR-MAS for Pseudomonas aeruginosa, a patho-
genic gram-negative bacterium (Righi et al. 2013). 25 metabolites could be assigned
in 1D and 2D spectra. In an elegant experiment, Hanoulle et al. showed the activa-
tion of the anti-tuberculosis prodrug ethionamide in mycobacteria (Hanoulle et al.
2005b, 2006b). Judge has demonstrated the use of HR-MAS for human multiple
myeloma cells showing that spectra can be obtained in 4 min thus allowing to
measure time-courses for continuous in vivo monitoring of metabolism. In multiple
myeloma cells, employing 13C-labelled ɑ-keto-isovalerate, conversion into valine
could be monitored over a time-course of 60 min (Fig. 4) (Judge et al. 2018). The
production of branched chain amino acids is an essential process for multiple
myeloma cells which have a high demand for amino acids needed for protein
production. For multicellular Neurospora crassa fungal cells, time-courses for
central carbon metabolism, amino acids, TCA cycle intermediates, energy storage
molecules and lipid and cell wall precursors were monitored for up to 10 h (Judge
et al. 2018).
Vermathen et al. have used non-small cell lung cancer (NSCLC) cells to study
Cisplatin (cisPt)-resistance metabolic adaptations employing HR-MAS (Vermathen
et al. 2021). High-quality spectra allowed the detection of 53 metabolites. PCA
showed close clustering of replicates and a clear separation with increasing resis-
tance against cisPt. De-induced cells showed similar behaviour as cisPt-resistant
cells, indicating a long-term memory after cisPt treatment. Metabolites predomi-
nantly changed in cisPt-resistant cells and their de-induced counterparts include
glutathione and taurine (Vermathen et al. 2021). This study clearly demonstrates
how HR-MAS NMR can be used to explain metabolic adaptations during drug
resistance.

3.2 Real-Time Measurements of Cells

Koczula et al. described experiments where chronic lymphoid leukaemia (CLL) cells
were embedded in a low-density agarose matrix to monitor metabolism over several
hours yielding time-courses as shown in Fig. 5 (Koczula et al. 2016). The agarose
matrix first helped to maintain cells afloat in a homogenous matrix and enable
measurements with small numbers of cells (0.5–1.0 million) which would otherwise
only have covered 1–2 mm at the bottom of the NMR tube. The concentration of
cells in this experiment determines the overall turnover rate. Limited diffusion of the
matrix caused oxygen depletion after circa 2 h which causes an abrupt increase of
lactate, alanine and glutamate production along with changes in the rate of glutamine
196 Z. Eraslan et al.

Fig. 4 Targeted isotopic CIVM-NMR measurements of metabolic flux in human leukaemia cells.
(a) 13C-labelled keto-isovalerate (KIV) was converted to valine. (b) 13C-labelled valine was not
converted to KIV, confirming unidirectional flux in ML cells. (c, d) Relative concentrations over
time of 13C-labelled KIV (orange) and 13 C-labelled valine (purple) compared to baseline noise
(grey), from spectral intensities within each region of the representative experiments in (a, b),
respectively, for 3 independent replicates. Reproduced from (Judge et al. 2018)

consumption (Koczula et al. 2016). By placing cells back in normoxic conditions

before repeating the real-time NMR experiment, reversible metabolic adaptation of
quiescent CLL cells to hypoxic conditions was demonstrated. By using chetomin as
a HIF-1ɑ inhibitor, it could be shown that this process of metabolic adaptation to
hypoxia is HIF-1ɑ dependent.
One advantage of this approach is that small numbers of cells are sufficient, the
number of cells merely determines the overall rate of metabolic turnover. A major
disadvantage of this approach is however that only the extracellular metabolome is
observed because signals arising from intra-cellular metabolites are too weak and
T2-broadened to be observed. Alshamleh et al. designed a similar experiment to
follow time-courses of metabolism using primary acute myeloid leukaemia (AML)
cells and used methylcellulose as a more cell-friendly matrix, not affecting ATP
levels as reported for agarose (Alshamleh et al. 2020).
Metabolomics in Cell Biology 197

Glucose
21
15 60 Lactate 25 110

Oxygen [%]
14
50 100
10 20
7 40 90

Oxygen [%]
5
15
0 0 30 80
0 2 4 6 10
20 70
5
10 60

0 0 50
0 5 10 15 20 0 5 10 15 20
Intensity [AU]

Glutamate 1,1 Glutamine 0,05 Alanine

0,2

0,15 0,9
0,04

0,1 0,7

0,03
0,05 0,5

0 0,3 0,02
0 5 10 15 20 0 5 10 15 20 0 5 10 15 20

Time [hours]

Control 20nM CTM 100nM CTM Oxygen %

Fig. 5 Metabolic adaptation of CLL cells to hypoxia involves HIF-1α. Representative NMR time-
course data for a CLL pre-treated for 3 h with either 0, 20 or 100 nM chetomin, before transferring
into NMR for a further 24 h. Dashed lines on the lactate graph show oxygen levels inside the NMR
tube. The top left panel shows an expanded view of lactate kinetics during the first 6 h with a visible
shift after oxygen depletion which is inhibited by CTM. Data shown are representative of a
minimum of N = 3 CLL samples. Reproduced from (Koczula et al. 2016)

3.3 Metabolic Studies in Mycobacteria

The applicability of HR-MAS for bacterial cells has been demonstrated by Li

using M. smegmatis whole cells (Li 2006). Two early studies were focussed on
mycobacteria where drug metabolites were monitored using HR-MAS. Ethion-
amide, the drug used in these studies needs to be activated by the monooxygenase
EthA. HR-MAS spectra show nicely how this activation occurs when EthA is
present (Hanoulle et al. 2005b, 2006b).
3D-cell cultures add a level of sophistication by allowing inter-cellular
interactions which are known to be relevant in particular in cancer cells (Knitsch
et al. 2021). Recently, Knitsch et al. (2021) designed a particularly interesting NMR
experiment. Their experimental design allowed to acquire spatially resolved spectra
from a 700 μm spheroid over a time-course of 112 h (Fig. 6). Profiles show spatially
resolved depletion of, for example, glucose with concomitant build-up of lactate.
Such 3D-culture systems are highly promising as they overcome main limitations of
198 Z. Eraslan et al.

Fig. 6 (a) Light microscopy images of three spheroids aligned to the spatially resolved intensity
profiles of NMR detected metabolites shown in (b) and enlarged (y-axis) in (c). Reproduced with
permission from (Knitsch et al. 2021)

cell-based models. NMR is very well suited to study such systems in real time as
illustrated and spacially resolved across the speroid.

3.4 Metabolic Fluxes by NMR and GC-MS

Both NMR and GC-MS have been used to study metabolic fluxes. Cascante has
established GC-MS based fluxomics using mainly GC-MS, first by establishing the
theoretically possible distribution of isotopes arising from selectively labelled glu-
cose, comparing [1-13C]glucose or [1,2-13C]glucose (Cascante and Marin 2008).
One of the key discoveries arising from this work has been the possibility to assess
pentose phosphate pathway activity from the label distribution in pyruvate and
Metabolomics in Cell Biology 199

lactate (Boros et al. 1998). Chong et al. have shown how the complementary nature
of GC-MS and NMR can be utilised to experimentally elucidate metabolic fluxes in
mammalian cells (Chong et al. 2017). While GC-MS yields mass increments for
each 13C-atom that is incorporated into a metabolite, without positional information,
NMR offers exactly the opposite, site-specific information of label incorporation.
For this, high-resolution HSQC NMR spectra were used, which resolve adjacent
13 13
C- C couplings. These couplings can become rather complex in mixtures of
metabolites with different label incorporations. An algorithm termed Combined
Analysis of NMR and MS Spectra (CANMS) has been described to simulate such
line shapes and use complementary GC-MS data to determine the exact label
incorporation for different atoms in glycolytic or TCA cycle metabolites. Examples
are shown for lactate and glutamate, for [1,2-13C]glucose as a metabolic precursor.
This approach is somewhat cumbersome but yields highly accurate site-specific label
incorporation.

3.5 Examples for Fluxes in Cancer Cells by GC-MS

MYC addition has long been understood to be major driver for many cancers,
associated with a strong glycolytic metabolism. Tarrado-Castellarnau et al. have
used a tracer-based GC-MS and transcriptomics analysis to study metabolic
consequences arising from the inhibition of CDK4/6 (Tarrado-Castellarnau et al.
2017). Cyclin-dependent kinases (CDK) are rational cancer therapeutic targets
owing to their essential role in the regulation of cell cycle progression at the G1
restriction point. Upon inhibition of CDK4/6, an accumulation of MYC was
observed. As a consequence, glutamine metabolism is upregulated, the mTOR
pathway gets activated while HIF1α-mediated responses to hypoxia get disturbed.
These MYC-driven adaptations to CDK4/6 inhibition render cancer cells highly
sensitive to inhibitors of MYC, glutaminase or mTOR and to hypoxia. This targeted
analysis was used to explore one avenue of a synergistic strategy by also inhibiting
GLS1 demonstrating strong antiproliferative effects. This work exemplifies how
metabolomics can be efficiently used in drug discovery to identify vulnerabilities to
develop new synergistic treatment regimes.
Moonira et al. used GC-MS metabolomics to study the mechanism of the anti-
diabetic drug metformin in hepatocytes (Moonira et al. 2020). They next used
[1,2-13C2]glucose to measure partitioning of flux between glycolysis (via PFK1)
and the pentose pathway by looking at mass isotopologues of lactate. Metabolism of
[1,2-13C2]glucose by the pentose pathway generates m1 and m0 lactate, whereas
glycolysis generates m2 and m0 lactate. Metformin increased the fraction of lactate
derived from glucose and increased the m2/m1 ratio. An AMPK activator decreased
the m2/m1 ratio. Cumulatively, metformin stimulates glycolysis via PFK1 in abso-
lute terms and relative to flux via the pentose pathway, whereas the AMPK activator
had the converse effect on the m2/m1 ratio. Metformin lowered cell G6P but not
ATP and decreased G6pc mRNA at high glucose. In contrast, direct allosteric
activators of AMPK were shown to have opposite effects compared to metformin
200 Z. Eraslan et al.

on glycolysis, gluconeogenesis and cell G6P. Overall this study showed how
metformin acts in hepatocytes by lowering G6P when challenged with glucose by
stimulation of glycolysis by an AMP-activated protein kinase-independent mecha-
nism through changes in allosteric effectors of phosphofructokinase-1 and fructose
bisphosphatase-1.
Selivanov has recently developed computational workflows based on their previ-
ously published software ISODYN (Selivanov et al. 2006) for the estimation of
quantitative dynamic flux maps using stable-isotope resolved metabolomics (SIRM)
(Selivanov et al. 2020) involving the extraction of mass spectra for metabolites of
interest, correction of the spectra for natural isotope abundance and the evaluation of
fluxes by simulation of mass spectra using a kinetic model based on ordinary
differential equations (ODEs) for isotopomers of metabolites. The biochemical
network is considered to derive a dynamic flux map. Foguet et al. has also extended
a 13C MFA-approach to seamlessly incorporate gene expression with metabolic flux
data (Foguet et al. 2019).
Metabolic flux maps used for 13C-MFA typically include less than 10% of the
reactions contained in human cells Genome-Scale Metabolic Models (GSMMs) used
to predict cell-level metabolic flux maps using human metabolic network
reconstructions and constraint-based methods (Nilsson and Nielsen 2017). In recent
years, it has been shown that GSMMs provide an advantageous platform for the
integration of transcriptomics and metabolomics data to study the mechanisms
underlying cell metabolic reprogramming and to identify new drug targets for cancer
therapy (Karakitsou et al. 2021). Recently, Cascante’s team developed genome-scale
metabolic flux maps for colon cancer cell lines integrating multi-omics data on
RECON2 human metabolic network reconstruction (Tarragó-Celada et al. 2021).
In brief, transcriptomic data, growth rates, respiration parameters, rates of metabolite
uptake and secretion, 13C resolved metabolomics and targeted metabolomics have
been used to parametrize the GSMMs. From the inferred metabolic flux maps,
cystine transport xCT system and methylenetetrahydrofolate dehydrogenase
1 (MTHFD1) have been identified and validated as new potential targets for meta-
static colorectal cancer.
Two studies from recent years exemplify how large-scale metabolomics screen-
ing can be used to map the larger landscape of cancer cell line metabolism. Li et al.
have profiled 225 metabolites in 928 cell lines from more than 20 cancer types from
the Cancer Cell Line Encyclopedia (CCLE) using LC-MS. Interrogation of metabo-
lite/gene associations reiterated the well-known link between IDH1 mutations and
the onco-metabolite 2-hydrogxyglutarate, which was also found in renal cell carci-
noma cell lines that did not have a mutated IDH1. Interestingly, by combining a
CRISPR–Cas9 knockout dataset with the Dependency Map (DepMap) exhibited that
the reduced asparagine levels in cells are associated with hypermethylation of the
promoter region of asparagine synthetase (ASNS). Further experiments showed that
asparagine starvation depletes cells with lowered ASNS expression. This is in
agreement with the targeted specific results of Eraslan et al. (2021) who showed
that some Burkitt’s lymphoma cell lines depend on asparagine for serine synthesis
(needed for one-carbon metabolism). Both studies allow the conclusion that
Metabolomics in Cell Biology 201

asparaginase should be pursued as a cancer therapeutic beyond its current use in

treating acute lymphoblastic leukaemia (ALL).

3.6 Fluxes and Metabolic Modelling

A second excellent example for high-throughput screening of cell lines used time-
of-flight mass spectrometry (FIA-TOFMS) to obtain metabolic profiles of 54 cancer
cell lines which were correlated with transcriptomic and proteomics data
highlighting the role of transcription regulators (Ortmayr et al. 2019).
FIA-TOFMS is by far the fastest metabolomics method which requires only 1 min
per sample (Fuhrer et al. 2011). The disadvantage is that it is somewhat limited to
fingerprinting, as it can yield multiple putative assignments for each feature. This
particular study has used HMDB (Wishart et al. 2009) and RECON2 (Thiele et al.
2013) for the annotation of metabolites. The authors constructed a global network
model incorporating the transcriptome, the proteome and the metabolome to predict
regulatory associations with central metabolic pathways. One key finding was a
global coordination between glucose and one-carbon metabolism, predicting a
selective sensitivity to antifolate drugs in cell lines with low glucose uptake as a
potential diagnostic marker for cancer cells that are more likely to respond to folate
synthesis inhibitors.
Such high-throughput metabolomics approaches generally require a high level
of computational support considering that the identification of metabolites is
difficult.

3.7 Metabolic Fluxes in Liver Disease

Most cell-based metabolomics is currently focussed on cancer metabolism. Key

mechanisms of cancer metabolomics along with a detailed list of drugs or drug
candidates that target metabolism were recently published by Schmidt et al. (2021)
One reason why cancer has been a major focus of metabolomics lies in the mere
nature of cancer in which numerous tumour-associated metabolic alterations accu-
mulate (Hanahan and Weinberg 2011). Most other disease models cannot be studied
on a cellular basis but require at least tissue of organoids as models. For example, to
study the action of ketohexokinase (KHK) inhibitors to improve steatosis, fibrosis
and inflammation in the context of non-alcoholic fatty liver disease (NAFLD),
human co-cultures are required (Shepherd 2021). While we struggled to use those
for metabolomics studies, liver perfusion with [U-13C]fructose worked well and
allowed to obtain high-quality HSQC spectra (Fig. 7) which helped to decipher
the mode of action of a KHK inhibitor showing reduction in fructose derived
glycerol. These spectra illustrate the level of spectral quality that can be obtained
using a micro-cryoprobe and HSQC spectra measured using non-uniform sampling
with a 25% schedule. A time sequence of HSQC spectra of control and KHK
inhibited samples confirmed uptake of labelled fructose for the control and
202 Z. Eraslan et al.

Fig. 7 NMR analysis of perfused human livers exposed to stable isotope-labelled fructose confirms
ketohexokinase (KHK) inhibition reduces lipogenesis and glycolysis. 1H-13C-HSQC spectra from
matched controls (CT: blue) and KHK inhibited (KHKi red) donor liver samples exposed to labelled
fructose for indicated periods (0.5–3 h). Reproduced with permission from (Shepherd 2021)

inhibitor-treated livers, at a similar rate and accumulated over time. Gradual conver-
sion of fructose to glycerol and glycerate was observed in control livers, as was
conversion to labelled sorbitol. Although labelled sorbitol appeared in the KHK
inhibited samples, there was a distinct lack of labelled glycerol and glycerate (Fig. 7)
(Shepherd 2021).

4 Conclusions

Metabolomics has become an essential element in multi-omics studies, mainly

because it builds a bridge between genetics and the phenotypical end point of
biological systems. Cell lines remain a gold standard for many preclinical
investigations despite obvious limitations. In many ways, cell lines are essential to
investigate specific genetic features, for example in cancer. Metabolomics
incorporates a wide range of methods, predominantly using NMR and mass spec-
trometry. We have shown several examples of high-throughput investigations that
are probably only feasible using fast LC-MS methods. However, the lions’ share of
studies at this point focus on specific targets of mechanisms where targeted
approaches are needed. To understand metabolic alterations in cancer cells, tracer-
based analyses employing GC-MS and NMR metabolic data have played an impor-
tant role. Moreover, NMR offers a unique angle to study metabolism in living cells
using different types of real-time experiments. Such methods will likely be further
developed to work under conditions better suitable to mimic physiological
conditions. Overall, the methodologies reviewed here open new avenues for drug
discovery by providing multi-parametric read-outs for cell-based experiments. The
results of many targeted metabolic studies lend themselves to be integrated into now
well-established Genome-Scale Metabolic Models to further predict drug targets.
Metabolomics in Cell Biology 203

Acknowledgments U. Günther and M. Cascante acknowledge the funding from the European
Commission (HaemMetabolome [EC-675790]). M. Cascante acknowledges grant PID2020-
115051RB-I00 funded by MCIN/AEI/10.13039/501100011033, AGAUR-Generalitat de
Catalunya (2017SGR1033) and ICREA foundation (ICREA-Academia-2021).

References
Akoka S, Barantin L, Trierweiler M (1999) Concentration measurement by proton NMR using the
ERETIC method. Anal Chem 71:2554–2557
Alshamleh I et al (2020) Real-time NMR spectroscopy for studying metabolism. Angew Chem Int
Ed 59:2304–2308
ÄŒuperloviÄ M (2010) Cell culture metabolomics: applications and future directions. Drug Discov
Today 15:12
Babushok VI et al (2007) Development of a database of gas chromatographic retention properties of
organic compounds. J Chromatogr A 1157:414–421
Becker W, Bhattiprolu KC, Gubensäk N, Zangger K (2018) Investigating protein-ligand
interactions by solution nuclear magnetic resonance spectroscopy. ChemPhysChem 19:895–
906
Bhinderwala F, Powers R (2019) NMR metabolomics protocols for drug discovery. In: Gowda
GAN, Raftery D (eds) NMR-based metabolomics, vol 2037. Springer, New York, pp 265–311
Bingol K, Bruschweiler-Li L, Li D-W, Brüschweiler R (2014) Customized metabolomics database
for the analysis of NMR 1H–1H-TOCSY and 13C–1H-HSQC-TOCSY spectra of complex
mixtures. Anal Chem 86:5494–5501
Bingol K et al (2015a) Unified and isomer-specific NMR metabolomics database for the accurate
analysis of 13 C– 1 H HSQC spectra. ACS Chem Biol 10:452–459
Bingol K et al (2015b) Metabolomics beyond spectroscopic databases: a combined MS/NMR
strategy for the rapid identification of new metabolites in complex mixtures. Anal Chem 87:
3864–3870
Blaise BJ et al (2007) Metabotyping of Caenorhabditis elegans reveals latent phenotypes. Proc Natl
Acad Sci U S A 104:19808–19812
Boros LG et al (1998) Nonoxidative pentose phosphate pathways and their direct role in ribose
synthesis in tumors: is cancer a disease of cellular glucose metabolism? Med Hypotheses 50:55–
59
Bruntz RC, Lane AN, Higashi RM, Fan TW-M (2017) Exploring cancer metabolism using stable
isotope-resolved metabolomics (SIRM). J Biol Chem 292:11601–11609
Cascante M, Marin S (2008) Metabolomics and fluxomics approaches. Essays Biochem 45:67–82
Chong M et al (2017) Combined analysis of NMR and MS spectra (CANMS). Angew Chem Int Ed
56:4140–4144
Clendinen CS et al (2014) 13C NMR metabolomics: applications at natural abundance. Anal Chem
86:9242–9250
Cohen SM, Ogawa S, Shulman RG (1979) 13C NMR studies of gluconeogenesis in rat liver cells:
utilization of labeled glycerol by cells from euthyroid and hyperthyroid rats. Proc Natl Acad Sci
76:1603–1607
Cox N, Millard P, Charlier C, Lippens G (2021) Improved NMR detection of phospho-metabolites
in a complex mixture. Anal Chem 93:4818–4824
DeVilbiss AW et al (2021) Metabolomic profiling of rare cell populations isolated by flow
cytometry from tissues. Elife 10:e61980
Dietmair S et al (2012) Metabolite profiling of CHO cells with different growth characteristics.
Biotechnol Bioeng 109:1404–1414
Dona AC et al (2014) Precision high-throughput proton NMR spectroscopy of human urine, serum,
and plasma for large-scale metabolic phenotyping. Anal Chem 86:9887–9894
204 Z. Eraslan et al.

Duarte IF, Lamego I, Rocha C, Gil AM (2009) NMR metabonomics for mammalian cell metabo-
lism studies. Bioanalysis 1:1597–1614
Edison AS et al (2021) NMR: unique strengths that enhance modern metabolomics research. Anal
Chem 93:478–499
Eraslan Z, Papatzikas G, Cazier J-B, Khanim FL, Günther UL (2021) Targeting asparagine and
serine metabolism in germinal centre-derived B cells non-Hodgkin lymphomas (B-NHL). Cell
10:2589
Fan TW-M, Lane AN (2008) Structure-based profiling of metabolites and isotopomers by NMR.
Prog Nucl Magn Reson Spectrosc 52:69–117
Fiehn O (2016) Metabolomics by gas chromatography–mass spectrometry: combined targeted and
untargeted profiling. Curr Protoc Mol Biol 114:30.4.1
Finch G, Yilmaz A, Utz M (2016) An optimised detector for in-situ high-resolution NMR in
microfluidic devices. J Magn Reson 262:73–80
Fischer E, Zamboni N, Sauer U (2004) High-throughput metabolic flux analysis based on gas
chromatography–mass spectrometry derived 13C constraints. Anal Biochem 325:308–316
Foguet C, Jayaraman A, Marin S, Selivanov VA, Moreno P, Messeguer R, de Atauri P, Cascante M
(2019) p13CMFA: Parsimonious 13C metabolic flux analysis. PLoS Comput Biol 15(9):
e1007310
Fuhrer T, Heer D, Begemann B, Zamboni N (2011) High-throughput, accurate mass metabolome
profiling of cellular extracts by flow injection–time-of-flight mass spectrometry. Anal Chem 83:
7074–7080
García-Álvarez I, Garrido L, Doncel-Pérez E, Nieto-Sampedro M, Fernández-Mayoralas A (2009)
Detection of metabolite changes in C6 glioma cells cultured with antimitotic Oleyl glycoside by
1
H MAS NMR. J Med Chem 52:1263–1267
Garcia-Alvarez I, Fernandez-Mayoralas A, Garrido L (2011) Effect of drugs in cells and tissues by
NMR spectroscopy. Curr Top Med Chem 11:27–42
Gogiashvili M et al (2019) HR-MAS NMR based quantitative metabolomics in breast cancer.
Metabolites 9:19
Gonzalez-Mendez R, Wemmer D, Hahn G, Wade-Jardetzky N, Jardetzky O (1982) Continuous-
flow NMR culture system for mammalian cells. Biochim Biophys Acta 720:274–280
Gowda GAN, Abell L, Lee CF, Tian R, Raftery D (2016) Simultaneous analysis of major
coenzymes of cellular redox reactions and energy using ex vivo 1H NMR spectroscopy. Anal
Chem 8
Halama A (2014) Metabolomics in cell culture – a strategy to study crucial metabolic pathways in
cancer development and the response to treatment. Arch Biochem Biophys 564:100–109
Hall AMR et al (2016) Practical aspects of real-time reaction monitoring using multi-nuclear high
resolution flow NMR spectroscopy. Cat Sci Technol 6:8406–8417
Hanahan D, Weinberg RA (2011) Hallmarks of cancer: the next generation. Cell 144:646–674
Hanoulle et al (2005a) Monitoring of the ethionamide pro-drug activation .pdf
Hanoulle X et al (2005b) Monitoring of the ethionamide pro-drug activation in mycobacteria by 1H
high resolution magic angle spinning NMR. Biochem Biophys Res Commun 331:452–458
Hanoulle et al (2006a) Selective intracellular accumulation of the major .pdf
Hanoulle X et al (2006b) Selective intracellular accumulation of the major metabolite issued from
the activation of the prodrug ethionamide in mycobacteria. J Antimicrob Chemother 58:768–
772
Illig T et al (2010) A genome-wide perspective of genetic variation in human metabolism. Nat
Genet 42:137–141
Jang C, Chen L, Rabinowitz JD (2018) Metabolomics and isotope tracing. Cell 173:822–837
Jin ES, Sherry AD, Malloy CR (2013) Evidence for transaldolase activity in the isolated heart
supplied with [U-13C3]glycerol. J Biol Chem 288:2914–2922
Judge MT et al (2018) Continuous in vivo metabolism by NMR. bioRxiv. https://doi.org/10.1101/
501577
Judge MT et al (2019) Continuous in vivo metabolism by NMR. Front Mol Biosci 6:26
Metabolomics in Cell Biology 205

Karakitsou E et al (2021) Genome-scale integration of transcriptome and metabolome unveils

squalene synthase and dihydrofolate reductase as targets against AML cells resistant to chemo-
therapy. Comput Struct Biotechnol J 19:4059–4066
Knitsch R, AlWahsh M, Raschke H, Lambert J, Hergenröder R (2021) In vitro spatio-temporal
NMR metabolomics of living 3D cell models. Anal Chem 93:13485–13494
Koczula KM et al (2016) Metabolic plasticity in CLL: adaptation to the hypoxic niche. Leukemia
30:65–73
Lau AN et al (2020) Dissecting cell-type-specific metabolism in pancreatic ductal adenocarcinoma.
Elife 9:e56782
Lee REB, Li W, Chatterjee D, Lee RE (2005) Rapid structural characterization of the
arabinogalactan and lipoarabinomannan in live mycobacterial cells using 2D and 3D
HR-MAS NMR: structural changes in the arabinan due to ethambutol treatment and gene
mutation are observed. Glycobiology 15:139–151
Li W (2006) Multidimensional HRMAS NMR: a platform for in vivo studies using intact bacterial
cells. Analyst 131:777
Lodi A et al (2017) Combinatorial treatment with natural compounds in prostate cancer inhibits
prostate tumor growth and leads to key modulations of cancer cell metabolism. NPJ Precision
Oncol 1:18
Lucas-Torres C, Wong A (2019) Current developments in μMAS NMR analysis for metabolomics.
Metabolites 9:29
Ludwig C, Guenther UL (2009) Ligand based NMR methods for drug discovery. Front Biosci 14:
24
Ludwig C et al (2011) Birmingham metabolite library: a publicly accessible database of 1-D 1H and
2-D 1H J-resolved NMR spectra of authentic metabolite standards (BML-NMR). Metabolomics
8:8–18
Markley JL et al (2017) The future of NMR-based metabolomics. Curr Opin Biotechnol 43:34–40
Meyer B, Peters T (2003) NMR spectroscopy techniques for screening and identifying ligand
binding to protein receptors. Angew Chem Int Ed 42:864–890
Mobarhan YL, Struppe J, Fortier-McGill B, Simpson AJ (2017) Effective combined water and
sideband suppression for low-speed tissue and in vivo MAS NMR. Anal Bioanal Chem 409:
5043–5055
Moonira T et al (2020) Metformin lowers glucose 6-phosphate in hepatocytes by activation of
glycolysis downstream of glucose phosphorylation. J Biol Chem 295:3330–3346
Nicholson JK, Lindon JC, Holmes E (1999) ‘Metabonomics’: understanding the metabolic
responses of living systems to pathophysiological stimuli via multivariate statistical analysis
of biological NMR spectroscopic data. Xenobiotica 29:1181–1189
Nilsson A, Nielsen J (2017) Genome scale metabolic modeling of cancer. Metab Eng 43:103–112
Nyblom HK, Nord LI, Andersson R, Kenne L, Bergsten P (2008) Glucose-induced de novo
synthesis of fatty acyls causes proportional increases in INS-1E cellular lipids. NMR Biomed
21:357–365
Ortmayr K, Dubuis S, Zampieri M (2019) Metabolic profiling of cancer cells reveals genome-wide
crosstalk between transcriptional regulators and metabolism. Nat Commun 10:1841
Palmnas M, Vogel H (2013) The future of NMR metabolomics in cancer therapy: towards
personalizing treatment and developing targeted drugs? Metabolites 3:373–396
Powers R (2009) NMR metabolomics and drug discovery: NMR metabolomics and drug discovery.
Magn Reson Chem 47:S2–S11
Rais B et al (1999) Oxythiamine and dehydroepiandrosterone induce a G(1) phase cycle arrest in
Ehrlich’s tumor cells through inhibition of the pentose cycle. FEBS Lett 456:113–118
Ramaswamy V et al (2013) Development of a 13C-optimized 1.5-mm high temperature
superconducting NMR probe. J Magn Reson 235:58–65
Rappez L et al (2021) SpaceM reveals metabolic states of single cells. Nat Methods 18:799–805
206 Z. Eraslan et al.

Reed MAC, Roberts J, Gierth P, Kupče Ē, Günther UL (2019) Quantitative isotopomer rates in real-
time metabolism of cells determined by NMR methods. Chembiochem 20(17):2207–2211.
https://doi.org/10.1002/cbic.201900084
Righi V, Constantinou C, Kesarwani M, Rahme LG, Tzika AA (2013) Live-cell high resolution
magic angle spinning magnetic resonance spectroscopy for in vivo analysis of Pseudomonas
aeruginosa metabolomics. Biomed Rep 1:707–712
Robinette SL, Zhang F, Brüschweiler-Li L, Brüschweiler R (2008) Web server based complex
mixture analysis by NMR. Anal Chem 80:3606–3611
Saborano R et al (2019) A framework for tracer-based metabolism in mammalian cells by NMR. Sci
Rep 9(1):2520
Sarou-Kanian V et al (2015) Metabolite localization in living drosophila using high resolution
magic angle spinning NMR. Sci Rep 5:9872
Schmidt DR et al (2021) Metabolomics in cancer research and emerging applications in clinical
oncology. CA Cancer J Clin 71:333–358
Selivanov VA, Marin S, Lee PWN, Cascante M (2006) Software for dynamic analysis of tracer-
based metabolomic data: estimation of metabolic fluxes and their statistical analysis. Bioinfor-
matics 22:2806–2812
Selivanov VA et al (2020) Software supporting a workflow of quantitative dynamic flux maps
estimation in central metabolism from SIRM experimental data. In: Nagrath D (ed) Metabolic
flux analysis in eukaryotic cells, vol 2088. Springer, pp 271–298
Shepherd EL (2021) Ketohexokinase inhibition improves NASH by reducing fructose-induced
steatosis and fibrogenesis. JHEP Rep 3:12
Szyperski T (1995) Biosynthetically directed fractional 13C-labeling of Proteinogenic amino acids.
Eur J Biochem 232:433–448
Szyperski T, Bailey JE, Wüthrich K (1996) Detecting and dissecting metabolic fluxes using
biosynthetic fractional 13C labeling and two-dimensional NMR spectroscopy. Trends
Biotechnol 14:453–459
Tarrado-Castellarnau M, de Atauri P, Tarragó-Celada J, Perarnau J, Yuneva M, Thomson TM,
Cascante M (2017) De novo MYC addiction as an adaptive response of cancer cells to CDK4/6
inhibition. Mol Syst Biol 13(10):940
Tarragó-Celada J et al (2021) Cysteine and folate metabolism are targetable vulnerabilities of
metastatic colorectal cancer. Cancers 13:425
Thiele I et al (2013) A community-driven global reconstruction of human metabolism. Nat
Biotechnol 31:419–425
Thomas JN et al (2021) Progress towards a higher sensitivity 13C-optimized 1.5 mm HTS NMR
probe. IEEE Trans Appl Supercond 31:1–4
Thompson JW et al (2019) International ring trial of a high resolution targeted metabolomics and
lipidomics platform for serum and plasma analysis. Anal Chem 91:14407–14416
Tiziani S et al (2009) Metabolomic profiling of drug responses in acute myeloid Leukaemia cell
lines. PLoS One 4:e4251
Tiziani S, Kang Y, Choi JS, Roberts W, Paternostro G (2011) Metabolomic high-content nuclear
magnetic resonance-based drug screening of a kinase inhibitor library. Nat Commun 2:545
Ugurbil K, Rottenberg H, Glynn P, Shulman RG (1978a) 31P nuclear magnetic resonance studies
of bioenergetics and glycolysis in anaerobic Escherichia coli cells. Proc Natl Acad Sci 75:
2244–2248
Ugurbil K, Brown TR, Den Hollander JA, Glynn P, Shulman RG (1978b) High-resolution 13C
nuclear magnetic resonance studies of glucose metabolism in Escherichia coli. Proc Natl Acad
Sci 75:3742–3746
Vermathen M, von Tengg-Kobligk H, Hungerbühler MN, Vermathen P, Ruprecht N (2021) 1H
HR-MAS NMR based metabolic profiling of lung cancer cells with induced and De-induced
cisplatin resistance to reveal metabolic resistance adaptations. Molecules 26:6766
Vignoli A et al (2019) High-throughput metabolomics by 1D NMR. Angew Chem Int Ed 58:
968–994
Metabolomics in Cell Biology 207

Vilaplana-Lopera N et al (2021) Crosstalk between AML and stromal cells triggers acetate secretion
through the metabolic rewiring of stromal cells. 2021.01.21.427406 Preprint. https://doi.org/10.
1101/2021.01.21.427406
Wan Q, Wang Y, Tang H (2017) Quantitative 13 C traces of glucose fate in hepatitis B virus-
infected hepatocytes. Anal Chem 89:3293–3299
Wang C et al (2020) COLMAR lipids web server and ultrahigh-resolution methods for
two-dimensional nuclear magnetic resonance- and mass spectrometry-based lipidomics. J
Proteome Res 19:1674–1683
Watanabe R et al (2016) Quantitative nuclear magnetic resonance spectroscopy based on PULCON
methodology: application to quantification of invaluable marine toxin, okadaic acid. Toxins
(Basel) 8:294
Wishart DS (2016) Emerging applications of metabolomics in drug discovery and precision
medicine. Nat Rev Drug Discov 15:473–484
Wishart DS et al (2009) HMDB: a knowledgebase for the human metabolome. Nucleic Acids Res
37:D603–D610
Yu M et al (2015) A resource for cell line authentication, annotation and quality control. Nature
520:307–311
Zhang A, Sun H, Xu H, Qiu S, Wang X (2013) Cell metabolomics. OMICS 17:495–501
NMR-Based Metabolomics to Evaluate
Individual Response to Treatments

Alessia Vignoli, Gaia Meoni, Veronica Ghini, Francesca Di Cesare,

Leonardo Tenori, Claudio Luchinat, and Paola Turano

Contents
1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 210
1.1 The Individual Metabolic Phenotype . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 211
1.2 Modulation of the Individual Metabolic Phenotype . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 212
2 Disease Fingerprinting and Individual Response to Pharmacological Therapy . . . . . . . . . . . . 214
2.1 Celiac Disease . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 214
2.2 Cancer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 215
2.2.1 Breast Cancer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 216
2.2.2 Lung Cancer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 219
2.2.3 Colon Cancer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 223
2.3 Viral Infections . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 226
2.3.1 Hepatitis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 226
2.3.2 COVID-19 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 229
2.4 Chronic Obstructive Pulmonary Disease . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 232
2.5 Drug Toxicity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 233
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 234

Alessia Vignoli and Gaia Meoni contributed equally to this work.

A. Vignoli · G. Meoni · V. Ghini · F. Di Cesare

Magnetic Resonance Center (CERM), University of Florence, Sesto Fiorentino, Italy
Department of Chemistry “Ugo Schiff”, University of Florence, Sesto Fiorentino, Italy

L. Tenori · C. Luchinat · P. Turano (✉)

Magnetic Resonance Center (CERM), University of Florence, Sesto Fiorentino, Italy
Department of Chemistry “Ugo Schiff”, University of Florence, Sesto Fiorentino, Italy
Consorzio Interuniversitario Risonanze Magnetiche MetalloProteine (CIRMMP), Sesto Fiorentino,
Italy
e-mail: [email protected]fi.it

# The Author(s), under exclusive license to Springer Nature Switzerland AG 2022 209
V. Ghini et al. (eds.), Metabolomics and Its Impact on Health and Diseases,
Handbook of Experimental Pharmacology 277, https://doi.org/10.1007/164_2022_618
210 A. Vignoli et al.

Abstract

The aim of this chapter is to highlight the various aspects of metabolomics in

relation to health and diseases, starting from the definition of metabolic space and
of how individuals tend to maintain their own position in this space. Physio-
pathological stimuli may cause individuals to lose their position and then regain
it, or move irreversibly to other positions. By way of examples, mostly selected
from our own work using 1H NMR on biological fluids, we describe the effects on
the individual metabolomic fingerprint of mild external interventions, such as diet
or probiotic administration. Then we move to pathologies (such as celiac disease,
various types of cancer, viral infections, and other diseases), each characterized
by a well-defined metabolomic fingerprint. We describe the effects of drugs on
the disease fingerprint and on its reversal to a healthy metabolomic status. Drug
toxicity can be also monitored by metabolomics. We also show how the individ-
ual metabolomic fingerprint at the onset of a disease may discriminate responders
from non-responders to a given drug, or how it may be prognostic of e.g., cancer
recurrence after many years. In parallel with fingerprinting, profiling (i.e., the
identification and quantification of many metabolites and, in the case of selected
biofluids, of the lipoprotein components that contribute to the 1H NMR spectral
features) can provide hints on the metabolic pathways that are altered by a disease
and assess their restoration after treatment.

Keywords
1
H-NMR · Cancer · Celiac disease · COPD · Drug toxicity · Fingerprinting ·
Profiling · Response to treatment · Viral infections

1 Introduction

The use of metabolomics in personalized medicine originates from two basic facts:

1. The existence of an individual metabolic phenotype characteristic for each

individual, i.e. of an invariant part of the metabolome that allows each subject
to be discriminated from all others.
2. The existence of a signature of the disease, which may range from very weak to
very strong depending on the pathology or its severity. The disease transiently
alters the individual metabolic phenotype, but this alteration disappears when the
individual reverts to the “healthy” status following medical or pharmaceutical
interventions.

Despite 1H NMR features a lower sensitivity (detection limit in the order of μM)
with respect to MS analysis, which downsizes the total number of measurable
NMR-Based Metabolomics to Evaluate Individual Response to Treatments 211

metabolites, it is exquisitely amenable to the untargeted fingerprinting of the sample

metabolome due to its very high reproducibility and capability for high throughput
analysis (Griffiths 2008; Takis et al. 2019; Vignoli et al. 2019).
With the exception of inorganic ions, almost all small molecules contain hydro-
gen atoms that can be measured simultaneously, providing rapid and distinct global
spectral patterns (or fingerprints) of the samples under investigation (Klupczyńska
et al. 2015; Takis et al. 2019; Vignoli et al. 2019). Thus, the fingerprint of a sample is
a recapitulation of its current metabolome, independently of the identification of the
metabolites. Conversely, metabolomic profiling is a global evaluation of the metab-
olite content of all the samples in a comparative fashion. The final aim is to identify
and accurately quantify as many compounds as possible. This approach enables the
detection of changes in the concentration of the measurable metabolites related to
specific physiological conditions.
Here, we first provide a series of examples demonstrating how untargeted 1H
NMR metabolomic fingerprinting can be used to identify the individual metabolic
phenotype and assess its stability over a long timescale in the absence of important
physio-pathological alterations. The existence of this individual fingerprint has been
observed in several biosamples (such as urine, saliva, blood, breath condensate) that
are commonly used in metabolomics. Then we provide a series of examples span-
ning from celiac disease to cancer and viral infections. These pathologies are very
different in their etiologies and clinical manifestations and involve different organs.
Likewise, in our lab we have observed metabolomic alterations that are clearly
characteristic of each of them. Additionally, we have used NMR-based
metabolomics to monitor the response to pharmaceutical treatments, and to identify,
before treatment, features of the individual metabolome that are prognostics to
discriminate responders from non-responders.
To help readers assess the potentiality of the NMR-based approach, we describe a
series of case studies drawn on the many years of experience of our group.

1.1 The Individual Metabolic Phenotype

In 2008, our research group (Assfalg et al. 2008) demonstrated, for the first time, that
the NMR detectable part of the metabolome of urine contains an invariant part,
which can be considered as the chemical signature of each individual. Multivariate
statistical analysis of multiple urine samples of different individuals enables the
definition of their “metabolic space”, where the metabolomic fingerprint of each
subject can be visualized and discriminated from that of the other subjects with an
accuracy close to 100%. Over a timescale of 10 years (Ghini et al. 2015a), in the
absence of important stressful perturbations, each individual still occupies its spe-
cific region of the metabolic space; daily intra-individual variability leads to small
fluctuations inside the individual metabolic space; conversely, shifts to other distinct
regions are associated to significant changes of the individual metabolomic pheno-
type as a consequence of the occurrence of important physio-pathological conditions
(Fig. 1).
212 A. Vignoli et al.

Fig. 1 A schematic representation of the individual metabolic phenotype. Each individual

(represented by a different color code in the picture) has a stable phenotype, which distinguishes
him/her from the other subjects (inter-individual variability). This phenotype is flexible enough to
compensate for day-by-day changes and external stimuli (intra-individual variability). This capacity
is lost at the onset of a disease but can be restored after a complete recovery

A strong individual phenotype also exists in saliva (Wallner-Liebmann et al.

2016) and blood (Holmes et al. 2008). For the latter two biofluids, their intrinsic
nature causes daily intra-individual fluctuations in metabolome composition that are
smaller than what occurs in urine; this is particularly true for blood, the composition
of which is tightly regulated by homeostasis. Additionally, at variance with urine and
saliva that have individual metabolomic fingerprints dominated by low-molecular
weight molecules, in both plasma and serum the presence of lipids and lipoproteins
significantly contributes to the individual fingerprint.
For serum/plasma, the stability of the individual phenotype has been reported
over a period of 7 years (but is likely to be much longer), while no information about
the long-term stability of the metabolic phenotype in saliva is yet available.

1.2 Modulation of the Individual Metabolic Phenotype

Beside the genetic components, all endogenous and exogenous metabolites derived
from extrinsic factors such as diet, drugs, gut microflora, stressors, and pollutants
contribute to the definition of the human metabolic phenotype. Thus, several studies
have been performed to understand the effects of mild dietary interventions on the
NMR-Based Metabolomics to Evaluate Individual Response to Treatments 213

metabolomic profile (Andersen et al. 2014; O’Sullivan et al. 2011). Importantly,

despite the levels of a large set of metabolites have been reported to be modulated by
the different dietary habits and to become more homogeneous in individuals
undergoing diet standardization for a short period of time, diet regimes do not
significantly change the individual signatures, which remain equally evident in all
types of biofluids (Lenz et al. 2003; Marianne and Walsh 2006; Wallner-Liebmann
et al. 2015; Winnike et al. 2009). The same effects are reported for probiotics
assumption in 22 healthy volunteers for 8 weeks (Ghini et al. 2020b). Because of
the probiotic treatment, significant modulations in the levels of a few urine and
serum metabolites are observed, but the observed changes are not so strong as to
hamper the individual recognition in the metabolic space. Another example is
represented by the evaluation of the metabolomic effects induced by the administra-
tion of bioactive molecules in the serum profiles of volunteers at risk of metabolic
syndrome. This study was performed in the context of the EC funded project
PATHWAY-27 (FP-KBBE # 311876). In this framework, the beneficial effects of
DHA (docosahexaenoic acid) administration on serum metabolomic profiles could
be assessed. However, this was only possible by suppressing the intra-individual
variations using paired approaches that compare the serum samples collected at the
beginning and at the end of the trial for each subject (Ghini et al. 2017).
The above examples demonstrate that little or no alteration of the individual
phenotype is brought about by mild treatments, whereas it can be profoundly
affected by the presence of pathologies and major surgical interventions, such as
bariatric surgery. Currently, bariatric surgery is the only available treatment to
provide sustained weight loss (Buchwald et al. 2009), to improve glucose regulation,
and to even promote complete remission of type 2 diabetes in severely obese patients
(Meijer et al. 2011). All three different bariatric procedures, i.e. sleeve gastrectomy,
proximal Roux-en-Y gastric bypass, and distal bypass, are associated with a strong
alteration of the serum metabolomic fingerprints of the patients (Gralka et al. 2015).
Within the common strong alterations, distal bypass patients could be discriminated
from the other two groups of patients, suggesting a stronger impact of this procedure
on the metabolic fingerprint. Different short-term and long-term alterations are also
observed. The serum metabolomic profiles of severe obese patients are characterized
by high levels of aromatic and branched-chain amino acids (AAA and BCAA,
respectively), and of metabolites related to energy metabolism (pyruvate and citrate).
Elevated levels of some metabolites related to the gut microbiota such as formate,
methanol, and isopropanol are also associated with obesity. Interestingly, after
bariatric surgery, independently of the type of procedure used, a significant reduction
of AAA, BCAA, pyruvate, methanol, and isopropanol concentrations, along with an
increase in arginine and glutamine levels are observed, indicating that surgically
induced weight loss can, at least in part, normalize the alterations in amino acid
metabolism associated with obesity (Gralka et al. 2015).
214 A. Vignoli et al.

2 Disease Fingerprinting and Individual Response

to Pharmacological Therapy

As anticipated, several pathologies have a strong metabolomic fingerprint at the

systemic level. Below we provide some examples from our lab of the metabolomic
characterization of different types of diseases. In these examples, NMR-based
metabolomics successfully provided a medium-to-strong fingerprint of the disease
and a definition of the molecular profile of the pathology in terms of small molecules
and lipoprotein parameters. Additionally, NMR was used to monitor the change of
the metabolomic fingerprint and profile along the disease evolution and in particular
following pharmacological treatments. Indeed, pharmacometabolomics is defined as
the determination of the individual metabolic phenotype to characterize signatures,
both before and after drug administration, that might inform treatment outcomes,
map the effects of drugs on metabolism and identify molecular pathways
contributing to drug-response and drug-toxicity phenotypes (Kaddurah-Daouk
et al. 2014).

2.1 Celiac Disease

The celiac disease (CD) originates from an aberrant adaptive immune response
against gluten. The presence of a characteristic metabolomic fingerprint of CD has
been demonstrated by multivariate analysis of the NMR spectra of serum and urine
samples from affected subjects and healthy controls (Bertini et al. 2009, 2011a;
Vignoli et al. 2019). For example, when comparing adult patients with sex- and age-
matched controls, a discrimination accuracy in the range of 75–83% was obtained in
urine, which rises to 81–94% in serum, depending on the study cohorts. The
differences originate from variations in the levels of metabolites related to three
main mechanisms: malabsorption, altered energy metabolism, and altered gut micro-
flora (Bertini et al. 2009).
In serum samples, the main reported differences between CD patients and
controls consist in lower levels of several amino acids (asparagine, isoleucine,
methionine, proline, and valine) and of pyruvate, lactate, and lipids, and in higher
levels of glucose and 3-hydroxybutyric acid. The decreased levels of pyruvate and
lactate along with higher levels of glucose are indicative of an impaired glycolysis
process. Enhanced beta-oxidation and malabsorption can instead explain the lower
levels of amino acids and lipids observed. A possible increase in the use of ketonic
bodies as a source of energy in celiac patients is consistent with the high levels of
3-hydroxybutyric acid. Energy conversion from lipids and catabolism of ketonic
bodies is far less efficient than that from glucids; consistently, celiac patients often
report symptoms of fatigue. Further, urine samples of CD patients are characterized
by high levels of some metabolites related to the gut microbiota, i.e. indoxyl sulfate,
meta-[hydroxyphenyl] propionic acid, phenylacetylglycine, and p-cresolsulfate. All
these findings are consistent with the hypothesis that in CD patients the gut
NMR-Based Metabolomics to Evaluate Individual Response to Treatments 215

microflora of the small bowel is altered or presents peculiar microbial species with
their own metabolome.
Notably, the so-called potential celiac patients, i.e., those individuals who tested
positive for the antibodies but have no evidence of intestinal damage, have an NMR
fingerprint that is similar to that of the overt patients, suggesting that the
dysmetabolism precedes the intestinal damage (Bernini et al. 2011a, b).
Celiac disease represents one example where dietary intervention induces rele-
vant changes in the metabolomic fingerprint (Bernini et al. 2011a, b). The only
available treatment for this disorder is to follow a gluten-free diet. One of the most
interesting findings was that the metabolic fingerprint of CD patients reverts to
normality after 12 months of a strict gluten-free diet. After the gluten-free diet,
normal levels of glucose and 3-hydroxybutyric acids are restored, and the sense of
fatigue tends to be reduced.

2.2 Cancer

Metabolomics can be used to derive information on cancer at various levels. Several

studies concern the characterization of the metabolome of cancer cell lines as a
complementary approach with respect to classical biochemical analyses and to other
omics, with the aim to derive information on altered metabolic pathways
(Cuperlovic-Culf et al. 2012; Li et al. 2017) and to conduct preclinical tests of
new anticancer agents (D’Alessandro et al. 2019; Ghini et al. 2021; Li et al. 2019;
Resendiz-Acevedo et al. 2021). NMR is particularly suitable for this purpose, as it
allows for a fast untargeted characterization of the endo and exometabolome (i.e., the
intracellular and extracellular composition in terms of metabolites), also at multiple
post-drug treatment times. Additionally, isotopically labeled substrates (e.g., 13C
labeled glucose) can be used to define metabolic fluxes by following the metabolism
of labeled substrates into their pathway products at specific time points (Antoniewicz
2018; Saborano et al. 2019).
Tumors can also be characterized at the level of the tissue metabolome; this
approach provides a specific view of cancer cells and their cross-talk with the tumor
microenvironment (Márquez and Matés 2021). Cancer tissues represent the localized
site of the disease. 1H NMR, in its HR-MAS version, is applicable to acquire spectra
of small quantities of cryo-preserved tissues (of the order of 10 mg), provided they
have not undergone any further sample manipulation (Vignoli et al. 2019). Indeed,
the use of tissues for ex vivo metabolomics raises a number of criticalities, ranging
from ethical to technical issues. Surgical specimens collected at different times of
intraoperative (warm) or post-operative (cold) ischemia undergo significant molecu-
lar degradation, so that the measured “metabolome” no longer reflects the original
physiological state of the tissues before intervention (Cacciatore et al. 2013). The
need of suitable frozen material collected and handled under strict preanalytical
conditions conflicts with ethics that assigns precedence, during intraoperative
procedures, to the best surgical performances and afterward to the diagnostic
needs of the pathologist.
216 A. Vignoli et al.

Fig. 2 Types of samples that can be analyzed via NMR-based metabolomics: cells, tissues, and
biofluids. In particular, in Sect. 2.2, we will focus on the analysis of biofluids (i.e. serum, plasma,
and urine) to investigate pathophysiological conditions associated with breast cancer, lung cancer,
and colon cancer

Instead, the metabolomic analysis of biofluids, such as blood serum or plasma, or

urine, presents several advantages. Sample collection is only minimally invasive,
and multiple collections at different time points can be easily obtained to establish a
metabolomic signature both before and after a given drug therapy. The preanalytical
SOPs are simple and easily satisfied (Ghini et al. 2019; ISO 23118:2021; Vignoli
et al. 2022). The resulting analyses might inform on the presence of specific features
of prognostic value and on treatment outcomes, respectively.
Below we provide examples that have seen a significant contribution from our
group and that regard the characterization of three of the most common types of
cancer, namely breast cancer, colorectal cancer, and lung cancer; all of them are
based on patients’ serum/plasma (Fig. 2).

2.2.1 Breast Cancer

Female breast cancer (BC) is the most commonly diagnosed cancer, with an
estimated 2.3 million new cases per year (11.7%), and is the leading cause of cancer
death in women (Sung et al. 2021). For many years BC has been considered a unique
clinical entity and treated with only one general approach. However, it has become
extremely clear that there is a high degree of diversity between and within the BC
subclasses (Polyak 2011). Based on their molecular characteristics, BC is classified
into three major subtypes: luminal (positive for estrogen and progesterone
receptors), human epidermal growth factor receptor 2+ (HER2+) enriched, and
basal (the majority of the latter tumors are also called triple-negative BC). Currently,
these different tumor subtypes are treated with specific therapies improving patient
survival: hormone receptor-positive disease is generally treated with endocrine
therapy (Early Breast Cancer Trialists’ Collaborative Group (EBCTCG) 2005;
Lerner et al. 1976; Wiggans et al. 1979); HER2+ disease is now cured with targeted
agents, which have significantly improved outcomes in both the (neo)adjuvant
NMR-Based Metabolomics to Evaluate Individual Response to Treatments 217

(Gianni et al. 2012; Martin et al. 2017; Slamon et al. 2011; von Minckwitz et al.
2017, 2019) and the metastatic setting (Swain et al. 2013; Verma et al. 2012).
Metabolomics can distinguish patients with BC with respect to healthy controls
with high predictive accuracies: PLS models with Q2 > 0.6 for studies on serum/
plasma (Cala et al. 2018; Lécuyer et al. 2018; Silva et al. 2019; Singh et al. 2017;
Sitter et al. 2006; Slupsky et al. 2010; Suman et al. 2018; Tayyari et al. 2018;
Wojtowicz et al. 2020; Zhou et al. 2017), and models with Q2 > 0.5 for urinary
studies (Silva et al. 2019; Slupsky et al. 2010; Zhou et al. 2017) were obtained.
Metabolomics proved useful in discriminating the plasma profiles of patients with
different BC molecular subtypes in comparison to controls (Díaz-Beltrán et al.
2021): as compared to HER2-patients, the HER2+ group showed elevated aerobic
glycolysis, gluconeogenesis, and increased fatty acid biosynthesis with reduced
Krebs cycle. ER+ patients, as compared to ER- ones, showed elevated alanine,
aspartate, and glutamate metabolism, decreased glycerol-lipid catabolism, and
enhanced purine metabolism (Fan et al. 2016).
Given these premises, in the framework of BC precision oncology, it is crucial to
identify, in each BC subtype, patients at higher risk of cancer recurrence and drug-
response profiles able to guide patients’ management (Bendinelli et al. 2021; Vignoli
et al. 2021b). The NMR-based metabolomics has shown to be a valuable prognostic
instrument and examples of its application with the abovementioned objectives are
provided below.

Response to Chemotherapy
The possible association between different metabolic profiles and response to
chemotherapy has been extensively investigated via NMR metabolomics of blood
derivatives.
Pharmacometabolomics on blood plasma/serum with the aim of predicting which
patients will benefit most from a specific treatment has provided significant results.
First in 2012, our group demonstrated that NMR-based metabolomics may play a
role in identifying the best responding patients with HER2+ metastatic breast cancer
(MBC) treated with paclitaxel plus lapatinib (Tenori et al. 2012). With a similar
NMR approach, Jiang and colleagues predicted the response to gemcitabine/
carboplatin chemotherapy (Jiang et al. 2018).
More recently, the general research attention has focused on neo-adjuvant che-
motherapy (NAC), the preferred treatment strategy for patients with large primary
tumors and/or locally advanced disease (Thompson and Moulder-Thompson 2012).
NAC offers the unquestionable benefit of downstaging disease and reducing the
tumor size prior to surgery, thus making patients with inoperable tumors candidates
for surgical resection (Debik et al. 2019; Thompson and Moulder-Thompson 2012).
Nonetheless, only less than 30% of patients overall exhibit pathological complete
response (disappearance of all invasive cancer in the breast) to NAC (Wei et al.
2013), with lower rates of response observed in endocrine receptor-positive, HER2-
disease.
In 2020, our research group, for the first time, investigated the capability of
predicting pathological complete response, using the baseline host immune
218 A. Vignoli et al.

cytokines and 1H NMR metabolomic fingerprints in a highly homogeneous popula-

tion of HER2+ BC patients enrolled for NAC treatment (Vignoli et al. 2020a). For
this study, 43 HER2+ BC patients, stratified according to their ER status in 22 ER+
and 21 ER-, were enrolled at baseline prior to any interventions. The pathological
complete response was obtained in 11 out of 22 ER+ patients and in 13 out of 21 ER-
patients. The plasma metabolomic fingerprint can distinguish ER+ and ER- patients
with 74.4% discrimination accuracy, suggesting a differential host–cancer interac-
tion in these two subtypes of BC. Moreover, in the ER+ BC patients the baseline
metabolomic fingerprint can be predictive of pathological complete response (72.7%
accuracy). The good responders, as compared to poor responders, are characterized
by lower concentrations of branched-chain amino acids isoleucine, and valine, as
well as ethanol, several phospholipids and cholesterol associated to almost all classes
of lipoproteins assigned by NMR. In the ER+ subgroup the combination of a
cytokine (TNF-α) and a metabolite (valine) was found to significantly enhance
discrimination between complete and partial response to NAC, yielding an area
under the receiver operating characteristic (ROC) curve of 0.92, and an accuracy of
90.9%. Conversely, no predictivity was observed in ER- BC patients. In their pilot
study, Wei et al. 2013 had already shown that metabolic profiling, performed by
combining NMR and LC–MS method, can distinguish groups of BC patients with
no, partial, or complete response; consistently, our study confirms this evidence and
enriches this scenario by coupling NMR-based metabolomics with the analysis of a
panel of 10 different cytokines.
Conversely, from blood serum analyses, NMR studies performed on breast
biopsies did not reveal significant metabolic differences between complete patho-
logical response and pathological non-responders before treatment (Choi et al. 2013;
Euceda et al. 2017), implying that the host–cancer interaction at a systemic level
plays a crucial role in the response to treatment. As a confirmation, a study
performed via MS on both blood serum and tissue biopsies (collected before, during,
and after NAC) showed that only marginally correlations are present between the
two biospecimens, and that only the serum profile is predictive of NAC response
(Debik et al. 2019).
Other relevant serum/plasma metabolomic studies have been targeted at
characterizing the impact of NAC on the patient metabolism. The NAC induces
relevant alterations in patient metabolism during and after treatment (Corona et al.
2021; Debik et al. 2019; Jobard et al. 2017). In particular, Jobard et al. (2017) have
shown that the administration of trastuzumab and everolimus in combination
induced systemic effects by altering lipid, glucose, and ketone bodies metabolisms.
These alterations are observable on the metabolic profiles of patients even several
weeks after the end of the drug intervention.

Disease Recurrence
Despite all efforts aimed at better stratifying BC patients, there is still a significant
proportion of early breast cancer patients who are overtreated. Clinicians are looking
for prognostic tools able to distinguish early BC patients at high risk of disease
recurrence, who need to be treated with aggressive adjuvant therapies, with respect
NMR-Based Metabolomics to Evaluate Individual Response to Treatments 219

to low-risk patients, who may be cured by locoregional therapy alone (McCartney

et al. 2017, 2018). NMR-based metabolomics of blood derivatives has shown to play
a role in this scenario.
The first evidence supporting the application of metabolomics as a potential
prognostic tool for recurrence prediction was published by Asiago et al. in 2010
(Asiago et al. 2010). They utilized metabolic profiling approach, obtained by
combining metabolites detected by both NMR and MS, to identify breast cancer
relapse before it occurs. Over the past decade, our research group has pursued this
research line using a fingerprinting approach. We have established a reproducible
method, based on serum NMR metabolomic analysis, able to distinguish early and
metastatic breast cancer patients with high discrimination accuracy. Furthermore, we
demonstrated that this model can be used to predict cancer relapse: early BC patients
“misclassified” as metastatic on the basis of their metabolomic fingerprint showed
indeed high risk of recurrence, whereas early BC patients correctly classified as early
BC can be considered at low risk. The recurrence prediction with this approach has
been validated and reproduced in monocentric and multicentric cohorts of patients
(Fig. 3) providing successful results (Tenori et al. 2015; Hart et al. 2017; McCartney
et al. 2019). In our 2015 and 2019 studies, serum samples of early BC patients were
collected before surgery, thus when the tumor was still in place, whereas in the
multicentric study of 2017, samples were collected after surgery but before starting
(when indicated) adjuvant chemotherapy or radiotherapy. Moreover, the early BC
patients enrolled for the studies of 2017 and 2019 had ER+ breast cancer, whereas
those of the study of 2015 were diagnosed with ER- breast cancer. The reproducibil-
ity of our approach, despite these differences in the study design, reinforces the
evidence that NMR-based metabolomics is really a promising instrument for the
stratification of patients with early breast cancer.
From our multicentric study (Hart et al. 2017) it emerged that, as compared with
metastatic BC patients, patients with early ER+ BC are characterized by lower serum
levels of citrate, choline, acetate, formate, lactate, glutamate, 3-hydroxybutyrate,
phenylalanine, glycine, leucine, alanine, proline, tyrosine, isoleucine, creatine, cre-
atinine, and methionine, and higher serum levels of glucose and glutamine. Interest-
ingly, in the subgroup of early BC patients with either relapse or no-relapse (with a
follow-up of at least 6 years), the patients who relapsed showed significantly higher
serum levels of choline, phenylalanine, leucine, histidine, glutamate, glycine, tyro-
sine, valine, lactate, and isoleucine, thus resembling a “micro-metastatic” profile
already at diagnosis. The decrease of tyrosine and lactate in early BC patients
confirmed what already seen in the ER- negative cohort examined in 2015, whereas
glucose and histidine showed opposite trends in ER- BC patients.

2.2.2 Lung Cancer

Lung cancer (LC) is the second most common cancer in both men and women and is
by far the leading cause of cancer death worldwide, making up almost 25% of all
cancer deaths. Most of the patients (>84%) with LC are affected by non-small cell
lung cancer (NSCLC), with the majority of patients presenting with advanced,
unresectable disease at the time of diagnosis (Siegel et al. 2021). Currently used
220 A. Vignoli et al.

Fig. 3 (a) The study design of the three metabolomic studies performed in our laboratory (Tenori
et al. 2015; Hart et al. 2017; McCartney et al. 2019) on serum samples of BC patients. Early BC
patients were randomly divided into a training and a validation set. Using the early BC training set
and the metastatic BC group, a random forest model was calculated on the serum NMR data. Then
the early BC validation set was used to test the hypothesis that early BC patients at high risk of
cancer recurrence could be classified as metastatic by the RF model. (b) The results of the three
studies are summarized by reporting for each of them the RF proximity plot (early BC green dots,
metastatic BC red dots) along with the RF model accuracy, and the area under the ROC curve of the
validation set used for the recurrence prediction

treatments for advanced NSCLC include chemotherapy, targeted drug therapy,

immunotherapy, or chemo-immunotherapy (Hirsch et al. 2017). Treatment options
are based mainly on the tumor histology but other factors, such as certain cancer
traits like PDL-1 TPS (tumor proportion score) and the presence of specific genomic
mutations, are also important.

Response to Immunotherapy
Among all the therapeutic strategies, immunotherapy has become an attractive
treatment modality for chemo-refractory solid tumors (Postow et al. 2015).
NSCLC cells have the ability to evade the immune system by expressing on their
surfaces certain “immune checkpoint” molecules that normally protect against
autoimmunity and inflammation, such as cytotoxic T-lymphocyte antigen-4
(CTLA-4), programmed cell death protein 1 and its ligand (PD-1 and PDL-1,
NMR-Based Metabolomics to Evaluate Individual Response to Treatments 221

respectively). Immune Checkpoint Inhibitor agents (ICIs), such as the monoclonal

antibodies nivolumab and pembrolizumab, reactivate T lymphocyte-mediated
immune response against the cancer cells by blocking the immune checkpoint
molecules (Brahmer 2014; Hamada et al. 2018). Several ICIs have shown outstand-
ing early success in many tumor types and have established an important role in the
first line of treatment of advanced lung cancer as a monotherapy or in combination
with chemotherapy, as well as in the second line after standard treatment (Borghaei
et al. 2015; Gandhi et al. 2018; Gettinger et al. 2015; Herbst et al. 2016; Rittmeyer
et al. 2017).
Unfortunately, not all patients respond to ICIs; the response rates are modest
(approximately 30% or less in LC), the associated costs are high, and true predictive
markers of treatment efficacy do not exist. Thus, the identification of biomarkers able
to identify the patients that are most likely to respond to, and benefit from, ICIs
treatment is of pivotal importance (Brahmer 2014). In this framework, metabolomic
fingerprinting of biofluids may represent a timely tool to define metabolomic
signatures that might inform on treatment outcomes.
In 2020, our research group conducted a pilot study based on 1H NMR
metabolomic investigation of sera samples from NSCLC patients treated with
immune checkpoint inhibitors (Ghini et al. 2020a). The experimental scheme of
the study is reported in Fig. 4. A total of 53 patients with advanced NSCLC were
enrolled; 34 patients were treated with nivolumab (monoclonal antibody directed
against PD-1) and 19 patients were treated with pembrolizumab (monoclonal anti-
body against PDL-1). All the analyzed samples were collected before the beginning
of the treatment (T0) with the aim to a priori identify responder and non-responder
subjects. Significantly, we could show that the metabolomic fingerprint of T0 serum
acts as a predictive biomarker of immune checkpoint inhibitors response, being able
to predict individual therapy outcome with >80% accuracy (Ghini et al. 2020a). In
the serum samples of non-responder subjects, we detected significantly higher levels
of pyruvate and alanine along with, even if not statistically significant, higher lactate
and glycine levels and lower citrate levels. All these changes are evocative of
increased glycolysis and decreased TCA pathway in non-responders. It is important
to underline that the significance of the univariate analysis performed on single
metabolite levels strongly relies on the number of subjects included in the study.
Thus, further investigations enrolling much higher numbers of subjects are necessary
to confirm these findings.
To the authors’ knowledge, this represents the first study using NMR-based
metabolomic fingerprinting of serum/plasma samples to predict the individual
response to anti PD-1 therapy in NSCLC. Instead, a few other examples based on
MS metabolomic profiling can be found in the literature. In 2020, Hatae et al.
analyzed by GC-MS the plasma samples of 55 NSCLC patients treated with
nivolumab and found that a combination of four plasma metabolites and several T
cell markers could be used as a good biomarker for responder identification (area
under the ROC curve = 0.96). The four selected metabolites include molecules
related to (1) gut microbiota (hippuric acid), (2) fatty acid oxidation
(butyryl-carnitine), and (3) redox-related metabolites (cystine and glutathione
222 A. Vignoli et al.

Fig. 4 (a) Experimental scheme to evaluate individual response to immunotherapy. (b) O-PLS
discrimination between NSCLC responders and non-responders to immune check points inhibitors,
adapted from Ghini et al. (2020a). Score plot, PC1 vs. PC2. Each symbol in the O-PLS score plot
represents the fingerprint of the NMR spectrum of each patient recruited for the study. Red dots:
nivolumab responder subjects; blue dots: nivolumab non-responder subjects. Red crosses:
pembrolizumab subjects predicted as responders; blue crosses: pembrolizumab subjects predicted
as non-responders

disulfide) (Hatae et al. 2020). One year later, in (2021), Nie X. et al. used LC-MS
profiling of early on-treatment serum to explore predictors of clinical outcomes of
anti-PD-1 treatment in 74 Chinese NSCLC patients. Serum samples were collected
2–3 weeks after the first infusion of PD-1 inhibitor. A small metabolite panel
consisting of hypoxanthine and histidine was identified and validated as a predictor
of treatment response, and high levels of both hypoxanthine and histidine were
associated with improved progression-free survival and overall survival (Nie et al.
2021). The completely different metabolites observed in the reported investigations
and proposed as treatment-efficacy biomarkers derive from different analytical
platforms, which allow for the observation of different panels of metabolites.
NMR-Based Metabolomics to Evaluate Individual Response to Treatments 223

The strength of our study relies on the uniqueness of the 1H NMR fingerprinting,
which takes advantage of its intrinsically untargeted nature and high reproducibility.
This approach allowed us to identify a metabolomic signature associated with ICIs
response that is independent of metabolite assignment and acts as a stronger “col-
lective” biomarker with respect to a single molecule or to a panel of a few molecules.

2.2.3 Colon Cancer

Colorectal cancer (CRC) is the third most prevalent malignancy after breast and lung
cancer, and the second most lethal disease in the world, with an anticipated 1.9
million new cases and 0.9 million deaths in 2020 (Xi and Xu 2021). The stage of
CRC at diagnosis is the most important predictor of survival: a relative 5-year
survival rate of around 90% has been demonstrated for patients diagnosed with
localized-stage disease, declining to around 71% and 14% for those diagnosed with
regional and distant metastasis respectively (Salmerón et al. 2022). Despite that, over
the last 20 years, breakthroughs in CRC treatment have resulted in a steady increase
in median overall survival (Dekker et al. 2019), CRC is still one of the most lethal
diseases. CRC is known for its significant variety in clinical presentation and
underlying tumor biology, as well as its relationship with numerous types of
etiological variables (Cunningham et al. 2010). Metabolomics of several biofluids
is increasingly used for successful patient classification in CRC, determining a
strong signature of the disease. This allows for the discrimination of CRC patients
from healthy subjects (Nannini et al. 2020; Turano 2014) and permits the prediction
of the overall survival (OS) within a set of metastatic patients using serum samples
(Bertini et al. 2012).
NMR-based metabolomics of minimally invasive biospecimens such as feces
can, not only, correctly classify CRC patients from healthy subjects, as it was already
shown by several studies (Le Gall et al. 2018; Lin et al. 2016, 2019; Monleón et al.
2009), but also discriminate CRC from patients with adenomatous polyps
(AP) (Nannini et al. 2021). Some of the identified metabolites suggest that metabolic
changes in CRC and adenoma are associated with different pathways, mainly
involving amino acid metabolism. It is well known that 95% of CRCs begin as
colonic AP or adenomas, and the possibility to correctly differentiate these two
forms is of primary importance for the early detection of the tumor.

Disease Recurrence
Using metabolomics to determine the distinct profiles of CRCs might allow for more
personalized or informed cancer therapy adjustments, contributing to precision
medicine (Wishart 2015). Indeed, CRC shows different characteristics of clinical
onset and individual response, even at the same pathological stage. Despite 80% of
CRCs are diagnosed at early stage and immediately treated with surgery, 35% of
these treated patients develop cancer relapse within 2–3 years after surgery (Guraya
2019). The assessment of recurrence risk in colon cancer primarily relies on the
pathological stage defined by the TNM system, based on the depth of tumor invasion
(T), lymph nodes involvement (N), and distant metastases (M). Risk stratification is
of fundamental importance for the choice of adjuvant treatment; however, only a
224 A. Vignoli et al.

small portion of patients benefits from it, with the majority being already cured by
primary surgery, and others experiencing disease relapse despite having received
adjuvant therapy. Improved identification of individuals who would benefit most
from adjuvant chemotherapy is a critical aim, particularly in older patients who are
more susceptible to treatment-related toxicity.
In 2021, our group demonstrated that NMR-based metabolomics on serum
samples can improve risk stratification in elderly patients with early CRC
(Di Donato et al. 2021). In this study, 169 serum samples, taken from three distinct
clinical trials, were collected before treatment from elderly CRC patients. Of these,
94 were collected from patients with early CRC (65 relapse free and 29 relapsed) and
75 from patients with metastatic CRC (Fig. 5). The model was built using a
supervised algorithm to discriminate the serum fingerprint of the relapse-free
patients in the early CRC cohort and the patients in the metastatic CRC cohort,
yielded 70% accuracy, 71% sensitivity, and 69% specificity. Then, with the hypoth-
esis that the metabolomic fingerprint of relapsed early CRC patients would be more
similar to that of patients with metastatic CRC, the model was used to predict the
remaining early CRC who had disease recurrence. Among the early CRC patients,
69% were correctly predicted as metastatic (and therefore considered at high relapse
risk). This suggests that even in the absence of clinically visible metastatic spread,
the metabolomic fingerprint of individuals with early CRC, who may have cancer
recurrence, has a potential “metastatic signature.”
Furthermore, when the metabolomic classification of all patients with early CRC
was analyzed using Kaplan–Meier curves, a strong prognostic effect was observed,
with patients classified at high risk by metabolomics (thus with a metabolomic
profile more similar to that of metastatic CRC patients) having a significantly higher
probability of disease relapse than those with a low risk metabolomic score
(High vs. Low risk: Hazard Ratio = 3.68, p-value = 0.001). Histidine and glutamine
were shown to be significantly decreased in the serum of metastatic CRC patients.
Previous evidences already suggested an association of glutamine with cancer
progression and poor cancer-specific survival (Bertini et al. 2012; Sirniö et al.
2019). Also, the downregulation of histidine was already observed in other studies
(Tan et al. 2013; Zhu et al. 2014) and its alteration was correlated with an increased
activity of histidine decarboxylase.
Our research group, in another recent paper (Vignoli et al. 2021a), investigated
the CRC relapse in the serum samples of a group of pre-chemotherapy CRC patients
undergoing surgery. We demonstrated that several differences between post- and
pre-operative serum samples are associated with cancer recurrence, in particular an
increment of HDL-Chol subfractions coupled with a decrement of VLDL-Chol
subfractions. These results may corroborate the hypothesis that the development of
CRC disrupts the physiological equilibrium of lipids and lipoproteins, leading to
lipid metabolic disorders (Zhang et al. 2014).
Urine samples have been also used for an NMR-based metabolomic approach to
predict relapse in a group of 62 CRC patients, yielding an area under the ROC curve
of 0.650 for cancer recurrence (Dykstra et al. 2017). Interestingly, here the authors
NMR-Based Metabolomics to Evaluate Individual Response to Treatments 225

Fig. 5 Experimental scheme (a) and results (b) extracted from Di Donato et al. (2021). Dots in the
score plots of the multivariate supervised PCA-CA kNN model represent the fingerprint of the
NMR spectra of each patient recruited for the study. Red dots code for metastatic CRC (mCRC)
patients, blue dots for early CRC (eCRC) relapse-free patients, green triangles for eCRC relapsed
patients predicted as “metastatic,” while green squares represent the CRC relapsed predicted as
early CRC free from relapse. The Kaplan–Meier plot on the bottom right was used by the authors to
estimate the outcome distribution
226 A. Vignoli et al.

used the NMR data to predict treatment delay which could depend on reaction to
chemotherapy, reaching an area under the ROC curve values of 0.750.

Pre-Surgical Effects of Anesthesia

The biofluids used by the various studies have been shown to contain important
information detectable with metabolomics for the characterization of CRC patients.
However, it is important to use the right preanalytical procedures on the samples to
obtain reliable results. We have highlighted (Ghini et al. 2015b) that the moment of
blood sample collection can strongly influence the plasma metabolomic profile.
Blood samples were taken from 70 CRC patients (40 non-metastatic and 30 with
liver metastasis) preoperatively, both prior and after anesthesia administration.
Anesthesia depresses the metabolisms in uneven ways, thus reducing the informa-
tion content of the metabolic profile and hence reducing the discrimination capability
of the method. Consequently, post-anesthesia samples are not very suitable for
standard metabolomics studies.

2.3 Viral Infections

Viruses utilize and/or rewire the host metabolism. Therefore, metabolomics is an

excellent tool to study the effect of viral infection, either in vitro using infected cell
cultures, or ex vivo in biosamples from infected animals or humans. In terms of
practical applicability, obvious limitations arise from biosecurity, which requires the
evaluation of the viral load and infectious risk of the different biological matrices as
a function of their nature and of the nature of the infective agent, as well as of the
biocontainment level of the laboratory.
On the other hand, identifying the metabolic pathways utilized by a virus has the
potential to help revealing drug targets, to monitor the response to antiviral agents
but also to evaluate the effect of vaccine administration, although very little is
available on this aspect.
Here, we report examples from our laboratory on the metabolomic characteriza-
tion of viral hepatitis and of SARS-CoV-2 infection.

2.3.1 Hepatitis
Viral hepatitis is a global health issue that affects millions of individuals and is
associated with a high fatality rate. Except for the hepatitis A virus (HAV), all
hepatotropic viruses, including hepatitis B, C, D, and E viruses (HBV, HCV, HDV,
and HEV), can induce chronic infections. HBV and HCV are the most common viral
causes of liver disease. According to the updated estimate of the WHO, there are
about 300 million people suffering from chronic HBV and six million people
suffering from chronic HCV (World Health Organization 2021).
Viral hepatitis chronic infection can cause progressive liver damage leading to
fibrosis and cirrhosis. Cirrhosis is the end-stage of every chronic liver disease and is
the major risk factor for hepatocellular carcinoma (HCC). HBV and HCV are the
NMR-Based Metabolomics to Evaluate Individual Response to Treatments 227

leading cause of HCC worldwide, accounting for a significant mortality of more than
1.3 million death per year (Ringehan et al. 2017).
Prevention campaigns are one of the main weapons to limit the incidence of these
viruses, especially for HCV for which a vaccine is not yet available. Direct-acting
antivirals (DAAs) were approved in 2014, revolutionizing HCV therapy and
allowing almost all patients to be cured. DAAs are very effective and well-tolerated
and they constitute the gold standard for the treatment of HCV chronic infection in
patients at all stages of liver disease.
Metabolomics analysis is being utilized to better understand host–pathogen
interactions and screen host biospecimen for biomarkers that are characteristic of
the viral infection. In 2019, an interesting picture emerged of the metabolomic
fingerprint of HCV infection compared to both healthy subjects (HS) and
HBV-infected patients (Meoni et al. 2019), suggesting that the two viruses exert a
different impact on human metabolism. Indeed, by the comparison of the 1H-NMR
serum profiles of HCV- and HBV-infected patients, we identified characteristic
metabolomic fingerprints of the two viral infections, obtaining an overall discrimi-
nation accuracy of 86% (OPLS-DA algorithm). As expected, the serum fingerprint
of HCV- and HBV-infected patients resulted to be extremely different also from the
serum fingerprint of HS, with a classification accuracy of 98.7% in the model built to
discriminate HCV vs. HS, and a classification accuracy of 80% in the model built of
HBV vs. HS (Fig. 6). Similarly, Godoy et al. (2010), using the 1H-NMR fingerprint-
ing approach, were able to accurately discriminate (95% predictive accuracy) the
urine samples of HCV-infected patients from those of HCV-negative subjects,
corroborating the potential of 1H-NMR fingerprinting for the fast, non-invasive
diagnosis of HCV infection using a urine sample.
The common changes we detected in the serum metabolomic profile of HBV- and
HCV-infected patients when compared to HS (e.g., increased levels of
3-hydroxybutyrate, acetate, lactate, and pyruvate) support the hypothesis that these
viruses preferentially stimulate glycolysis over oxidative phosphorylation, analo-
gously to the Warburg effect in cancer (Okuda et al. 2002). Instead, in the compari-
son of HCV- and HBV-infected patients we noted a different behavior of several
metabolites, suggesting that the perturbation could be attributable to a direct action
of the two types of viruses rather than to the host response. Interestingly, the higher
levels of 2-oxoglutarate and 3-hydroxybutyrate in HCV patients compared to HBV,
also identified as biomarkers of cardiovascular disease and ketoacidosis, could
explain why some extrahepatic manifestations, such as cardiovascular diseases and
diabetes, are common in patients with chronic HCV and not in patients with HBV
(Bernini et al. 2011a, b; Chen et al. 2014; Du et al. 2014).
Metabolomics proved useful also for disease staging and for characterizing the
response to treatment. Anti-HCV treatment has advanced significantly in recent
years, with direct-acting antivirals (DAAs) replacing pegylated interferon and riba-
virin and providing effective treatment and less adverse effects. In our work, we
characterized also the metabolomic fingerprint and the profile of HCV patients
before and after effective DAA treatment. In this case, we identified a major
contribution of the low-molecular weight molecules in characterizing the changes
228 A. Vignoli et al.

Fig. 6 Fingerprinting of the Hepatitis C and B viruses in serum samples resulting from OPLS-DA
models. (a) HCV- (red dots) vs HBV- (purple triangles) infected patients; (b) HCV-infected
patients (red dots) vs HS (sky-blue triangles); (c) HBV-infected patients (purple triangles) vs
HS (sky-blue triangles). Adapted from Meoni et al. (2019)

introduced in the individual metabolomic profile by the therapy, suggesting also that
the perturbation in lipid metabolism induced by the infection persists after viral
eradication.
According to other studies (Cano et al. 2017; Nguyen et al. 2021; Sarfaraz et al.
2016) tyrosine and formate levels increase on passing from no/mild fibrosis to severe
fibrosis. Furthermore, differences in metabolite levels between patients with higher
and lower fibrosis scores were reduced after DAAs therapy, confirming that altered
metabolites are restored, most likely due to liver damage regression after viral
eradication.
Regarding the effect of treatment on sera of HBV-infected patients, Nguyen et al.
2021 recently published a study confirming that NMR-based metabolomics is
capable of revealing in serum samples a gradual metabolic transition from
NMR-Based Metabolomics to Evaluate Individual Response to Treatments 229

pretreatment to early treatment and then to a longer treatment period, as well as of

accurately distinguishing the serum of patients who needed medical treatment
(patients who would commence treatment within 6 months from sampling) from
those who did not.

2.3.2 COVID-19
The COVID-19 pandemic has heavily reshaped research activities worldwide.
Along with an incredibly fast development of vaccines that could bring the pandemic
under control, efforts were also directed toward the development of antiviral drugs
and monoclonal antibodies. Additionally, the pandemic has led to a rush to repur-
pose existing drugs. The pandemic has also stimulated researchers operating in fields
complementary to pharmacology and vaccinology to contribute to the understanding
of the physiopathology of the disease and to the characterization of risk factors and
response to treatments. Coordinated efforts have been searched via the creation of
networks that facilitate discussion among participants and communication of the key
results prior to publication, together with the establishment of strategic transnational
collaborations. Worth mentioning in this context are: (1) the Covid19-NMR project
(https://covid19-nmr.de/), dealing with the determination of the structures of RNA
and proteins of SARS-CoV-2 to investigate their drugability by small molecules;
(2) the NMR international COVID-19 Research Network (CV19 Research Net-
work), a metabolomics-based initiative consisting of several institutions that collab-
orate, using standardized NMR procedures, to detect the infection, predict outcomes
during hospitalization, and direct efforts toward Long COVID.
Suitable biofluids from COVID-19 patients for metabolomics are serum, plasma,
and urine, thanks to their low viral load, albeit collection of research samples during
the worse phases of the COVID-19 pandemic was a further burden for the clinicians.
In this frame, we present here some activities from our lab directed toward the use of
NMR-based metabolomics and lipoproteomics to characterize the COVID-19
metabolomic fingerprints, to monitor the effect of repurposed drugs and vaccination
follow-up (Ghini et al. 2022a; Meoni et al. 2021).
We and others (Ghini et al. 2022b; Ballout et al. 2021; Baranovicova et al. 2021;
Bizkarguenaga et al. 2021; Bruzzone et al. 2020; Julkunen et al. 2021; Kimhofer
et al. 2020; Lodge et al. 2021; Masuda et al. 2021) have shown that SARS-CoV-
2 infection induces profound changes in the metabolic phenotype of the patients
(Millet, “Prospective metabolomic studies in precision medicine. The AKRIBEA
project.”; Rogers, “The Metabolomics of Critical Illness”). Accordingly, 1H NMR
spectra of plasma samples of COVID-19 patients could be strongly discriminated
from the spectra of both healthy subjects and COVID-19-recovered subjects, with a
discrimination accuracy higher than 90% in both cases. The differences originate
from significant alterations in the concentrations of several metabolites and of a
panel of lipoprotein components. The metabolites and lipoprotein parameters that
are significantly dysregulated in COVID-19 acute subjects are listed in Fig. 7.
Characteristic trends in metabolite and lipoprotein levels are also observed as a
function of the disease severity (Ghini et al. 2022b). The analysis of the specific
changes and correlations with clinical data enabled the identification of potential
230 A. Vignoli et al.

Fig. 7 Metabolites and lipoprotein parameters that are found significantly up- or down-regulated in
COVID-19 patients with respect to healthy controls according to different Metabolomics papers
(Meoni et al. 2021; Bruzzone et al. 2020; Kimhofer et al. 2020; Lodge et al. 2021; Ballout et al. 2021;
Masuda et al. 2021; Julkunen et al. 2021; Baranovicova et al. 2021; Bruzzone et al. 2020) and COMETA
project (Ghini et al. 2022b). Up-/down-regulated features are indicated by red/blue cells. (a) Metabolites;
(b) Lipoprotein Main Parameters; (c) Lipoprotein Subfractions. Adapted from (Ghini et al. 2022b)
NMR-Based Metabolomics to Evaluate Individual Response to Treatments 231

biochemical determinants of the disease fingerprint, which found confirmation in

other studies performed at different centers worldwide and in some cases dealing
with much larger cohorts. The parameters that are found altered in COVID-19
patients with respect to recovered individuals overlap with the acute infection
biomarkers identified in the comparison with healthy subjects, indicating the sub-
stantial metabolic healing of COVID-19-recovered subjects. During the healing
process, the metabolome and lipoproteome revert back to the “healthy” state with
different rates; during either spontaneous healing or pharmacological treatments the
metabolites are reverted faster than lipoproteins. Notably, several other
metabolomics papers have been published, which identify common molecular
features as characteristic of the COVID-19 profile. High convergence on common
biomarkers from different metabolomics studies is not so common and denotes the
presence of a strong profile, independent of confounding factors like place of origin,
sex, age, and comorbidities.
Regarding the pharmacological treatments, we had the chance to analyze the
effect of tocilizumab on a very small cohort (8 patients) treated at the Florence
University hospital during the first wave of COVID-19 in spring 2020. Tocilizumab
is a monoclonal antibody that attaches to the receptor of the cytokine interleukin-6,
whose levels are elevated in response to systemic inflammation and plays an
important role in severe COVID-19 disease and associated respiratory failure. On
December 6th 2021, the European Medicines Agency (EMA) recommended
extending the indication of RoActemra (tocilizumab) to include the treatment of
adult COVID-19 patients who are receiving systemic treatment with corticosteroids
and require supplemental oxygen or mechanical ventilation.
When measuring the post-treatment levels of metabolites and lipoproteins that are
significantly altered by the infection, we found that eight metabolites (namely,
acetone, citrate, glutamine, glycine, lactate, mannose, phenylalanine, and pyruvate)
partially or completely revert toward the levels of CTR subjects. Some lipoprotein
main- and sub-fractions were also significantly affected by the tocilizumab treat-
ment, but, they did not revert back to “healthy” values, in line with the above-
reported observation that the recovery of the lipoproteome is slower than that of the
metabolome along the healing process (Meoni et al. 2021).
Finally, 1H NMR spectra of sera have also been used to define the changes
induced by vaccination with Pfizer-BioNTech vaccine in a cohort of 20 healthcare
workers, 10 COVID-19 naïve, and 10 with a previous history of COVID-19 infec-
tion (infected in the period March–April 2020 with the Wuhan strain and recovered
from the disease 208–280 days before vaccination). All of them received two doses
of vaccine, 21 days apart, and their serum samples were collected at 6 different time
points to monitor time-dependent changes induced by the vaccination. Importantly,
the vaccination does not induce a major modification of the metabolic phenotype of
the subjects; the intra-individual differences remain smaller than the inter-individual
ones during all the course of the study, with an individual discrimination accuracy
>85% (considering the six samples collected for each subject). Nevertheless, in
response to vaccination we could observe some common changes that are consis-
tently occurring in all subjects within each group. While vaccination does not induce
any significant variation in the metabolome, it causes changes at the level of
232 A. Vignoli et al.

lipoproteins that are smaller for COVID-19-recovered subjects with respect to naïve
subjects, suggesting that a previous infection reduces the vaccine modulation of the
lipoproteome composition. The differences between the two groups involve the
nature and number of affected lipoprotein parameters. Additionally, the effect of
the second dose is essentially negligible for the COVID-19-recovered subjects
(Ghini et al. 2022a).

2.4 Chronic Obstructive Pulmonary Disease

As highlighted in chapter (Lacy, “Metabolomics of Respiratory Diseases”), chronic

obstructive pulmonary disease (COPD) is a pathological condition characterized by
the chronic, poorly reversible, and progressive development of airflow limitation
often associated with parenchymal destruction and emphysema (Barnes 2000).
Although abnormal respiratory inflammation is crucial for COPD development,
the complex COPD pathophysiology is not yet fully understood. Currently, no
validated biomarker is accepted for disease prognosis or COPD therapy monitoring
(Ghosh et al. 2016).
Several studies have identified serum, exhaled breath condensate (EBC), and
urine 1H NMR-based metabolomic fingerprints of COPD patients, also showing the
ability of EBC metabolomics to assess airway inflammation (Airoldi et al. 2016; de
Laurentiis et al. 2008; Motta et al. 2012; Wang et al. 2013; Ząbek et al. 2015). In this
frame, a study from our group (Bertini et al. 2014) showed that NMR metabolomics
of EBC could discriminate COPD patients from controls with an overall accuracy of
86%. As compared to controls, EBC from COPD patients featured significantly
lower levels of acetone, valine, and lysine, and significantly higher levels of lactate,
acetate, propionate, serine, proline, and tyrosine. Lower levels of valine and lysine
(two essential amino acids) appear consistent with muscle wasting and weight loss
that are known to occur in advanced COPD (Agusti et al. 2002). The hypothesis of a
possible presence of subclinical malnutrition in COPD is also discussed in chapter
(Lacy, “Metabolomics of Respiratory Diseases”), where the reader can find a more
comprehensive overview of the metabolic alterations observed in sera of COPD
patients.
The application of pharmacometabolomics in the context of COPD is aimed at
identifying the appropriate treatment for each COPD patient, predicting his/her
response to therapy. Our research group applied an original and holistic approach,
dubbed “breathomics,” to monitor the effects of treatment with and withdrawal from
inhaled beclometasone/formoterol in patients with COPD (Montuschi et al. 2018). In
our application, breathomics combined two electronic noses (carbon polymer sensor
e-nose, quartz crystal sensor e-nose), EBC NMR-based metabolomics, sputum cell
counts, sputum supernatant and EBC prostaglandin E2 (PGE2) and
15-F2t-isoprostane, fraction of exhaled nitric oxide, and spirometry data.
Breathomics improves the identification of pharmacological treatment-induced
effects as compared with standard spirometry. Furthermore, this approach provides
insights into the anti-inflammatory effects of inhaled corticosteroids in COPD
patients as reflected by reduced levels of sputum PGE2 and EBC acetate during
NMR-Based Metabolomics to Evaluate Individual Response to Treatments 233

Fig. 8 Heatmap showing correlations among metabolites quantified in the urine, serum, exhaled
breath condensate (EBC), and sputum supernatant samples of 14 patients with COPD. Correlation
values (R) are reported as different degrees of color intensity (blue, negative correlation; red,
positive correlations). The figure is adapted and reprinted with permission from Vignoli et al.
(2020b). Copyright 2020 American Chemical Society

treatment with formoterol alone. This research line was further explored with the
analysis of urine, serum and sputum supernatant, demonstrating that different
biological matrices provide complementary information on the effects of
beclometasone/formoterol administration in COPD patients, and thus their integra-
tion could be useful for elucidating the metabolic mechanism of action of inhaled
corticosteroids (Vignoli et al. 2020b). An overview of the metabolite correlation
patterns among the different biofluids is presented in Fig. 8.

2.5 Drug Toxicity

Pharmacometabolomics includes drug safety evaluation that is an important step in

the drug pipeline and a main concern for regulatory agencies. Safety evaluation is
234 A. Vignoli et al.

required across the whole process of drug development, from preclinical studies to
clinical trials, as well as in post-approval safety surveillance. Biochemical and
histological analyses are the major approaches used for drug safety evaluation.
These approaches are effective in most cases to determine the safety profile of
drug candidates (Wang et al. 2017). However, these methods can neither provide
detailed information nor explore the mechanisms of drug toxicity (Nicholson et al.
2002). Metabolomics is ideally positioned to address the challenges of drug toxicity
(Hertz, “Chemotherapy-induced Peripheral Neuropathy”). It represents a powerful
tool for collecting mechanistic information, indicating not only the extent of a toxic
insult but also its underlying mechanisms (Ramirez et al. 2013). Several examples
exist in the literature of NMR-based metabolomic applications in toxicology and
drug safety evaluation. Just to cite few of them, drugs and toxicants like flutamide
(Choucha Snouber et al. 2013), hydrazine (Garrod et al. 2005; Lindon et al. 2003),
and gentamicin (Lenz et al. 2005) have been assessed.
A typical example of a clinically useful class of molecules that presents high rates
of therapy discontinuation due to acute side effects is represented by
V-phosphodiesterase (PDE5) inhibitors (PDE5i) (Corona et al. 2016). These
molecules are the first-line therapy for erectile dysfunction (ED), a widespread health
problem in the general population of middle-aged men (Ayta et al. 1999; Yafi et al.
2016). Administration of PDE5i proved beneficial in 60–70% of patients with
varying etiologies of sexual dysfunction (Yafi et al. 2016). However, adverse effects
are a common drawback. In a study from our research group, we retrospectively
evaluated serum and urine NMR-based metabolomic profiles to identify prognostic
biomarkers of unfavorable efficacy/safety profile of PDE5i before drug administra-
tion. To the best of the authors’ knowledge, this is the first and only NMR-based
metabolomic study focused on PDE5i toxicity. Patients who are likely to experience
adverse effects can be identified with an accuracy of 77% using pretreatment serum
samples. Adverse drug reactions showed to be associated with high levels of
LDL-lipoprotein subfractions at baseline (Rocca et al. 2020). The results of this
pilot study underline how metabolomics may help in identifying the metabolic bases
underlying efficacy/safety profile of the PDE5i therapy.

References
Agusti AG, Sauleda J, Miralles C, Gomez C, Togores B, Sala E, Batle S, Busquets X (2002)
Skeletal muscle apoptosis and weight loss in chronic obstructive pulmonary disease. Am J
Respir Crit Care Med 166:485–489. https://doi.org/10.1164/rccm.2108013
Airoldi C, Ciaramelli C, Fumagalli M, Bussei R, Mazzoni V, Viglio S, Iadarola P, Stolk J (2016) 1H
NMR To explore the metabolome of exhaled breath condensate in α1-antitrypsin deficient
patients: a pilot study. J Proteome Res 15:4569–4578. https://doi.org/10.1021/acs.jproteome.
6b00648
Andersen M-BS, Rinnan Å, Manach C, Poulsen SK, Pujos-Guillot E, Larsen TM, Astrup A,
Dragsted LO (2014) Untargeted metabolomics as a screening tool for estimating compliance
to a dietary pattern. J Proteome Res 13:1405–1418. https://doi.org/10.1021/pr400964s
Antoniewicz MR (2018) A guide to 13C metabolic flux analysis for the cancer biologist. Exp Mol
Med 50:1–13. https://doi.org/10.1038/s12276-018-0060-y
NMR-Based Metabolomics to Evaluate Individual Response to Treatments 235

Asiago VM, Alvarado LZ, Shanaiah N, Gowda GAN, Owusu-Sarfo K, Ballas RA, Raftery D
(2010) Early detection of recurrent breast cancer using metabolite profiling. Cancer Res 70:
8309–8318. https://doi.org/10.1158/0008-5472.CAN-10-1319
Assfalg M, Bertini I, Colangiuli D, Luchinat C, Schäfer H, Schütz B, Spraul M (2008) Evidence of
different metabolic phenotypes in humans. Proc Natl Acad Sci U S A 105:1420–1424. https://
doi.org/10.1073/pnas.0705685105
Ayta IA, McKinlay JB, Krane RJ (1999) The likely worldwide increase in erectile dysfunction
between 1995 and 2025 and some possible policy consequences. BJU Int 84:50–56. https://doi.
org/10.1046/j.1464-410x.1999.00142.x
Ballout RA, Kong H, Sampson M, Otvos JD, Cox AL, Agbor-Enoh S, Remaley AT (2021) The
NIH Lipo-COVID study: a pilot NMR investigation of lipoprotein subfractions and other
metabolites in patients with severe COVID-19. Biomedicine 9:1090. https://doi.org/10.3390/
biomedicines9091090
Baranovicova E, Bobcakova A, Vysehradsky R, Dankova Z, Halasova E, Nosal V, Lehotsky J
(2021) The ability to normalise energy metabolism in advanced COVID-19 disease seems to be
one of the key factors determining the disease progression – a metabolomic NMR study on
blood plasma. Appl Sci 11:4231. https://doi.org/10.3390/app11094231
Barnes PJ (2000) Chronic obstructive pulmonary disease. N Engl J Med 343:269–280. https://doi.
org/10.1056/NEJM200007273430407
Bendinelli B, Vignoli A, Palli D, Assedi M, Ambrogetti D, Luchinat C, Caini S, Saieva C, Turano
P, Masala G (2021) Prediagnostic circulating metabolites in female breast cancer cases with low
and high mammographic breast density. Sci Rep 11:13025. https://doi.org/10.1038/s41598-
021-92508-1
Bernini P, Bertini I, Calabrò A, la Marca G, Lami G, Luchinat C, Renzi D, Tenori L (2011a) Are
patients with potential celiac disease really potential? The answer of metabonomics. J Proteome
Res 10:714–721. https://doi.org/10.1021/pr100896s
Bernini P, Bertini I, Luchinat C, Tenori L, Tognaccini A (2011b) The cardiovascular risk of healthy
individuals studied by NMR metabonomics of plasma samples. J Proteome Res 10:4983–4992.
https://doi.org/10.1021/pr200452j
Bertini I, Cacciatore S, Jensen BV, Schou JV, Johansen JS, Kruhøffer M, Luchinat C, Nielsen DL,
Turano P (2012) Metabolomic NMR fingerprinting to identify and predict survival of patients
with metastatic colorectal cancer. Cancer Res 72:356–364. https://doi.org/10.1158/0008-5472-
11-1543
Bertini I, Calabrò A, De Carli V, Luchinat C, Nepi S, Porfirio B, Renzi D, Saccenti E, Tenori L
(2009) The metabonomic signature of celiac disease. J Proteome Res 8:170–177. https://doi.org/
10.1021/pr800548z
Bertini I, Luchinat C, Miniati M, Monti S, Tenori L (2014) Phenotyping COPD by H-1 NMR
metabolomics of exhaled breath condensate. Metabolomics 10:302–311. https://doi.org/10.
1007/s11306-013-0572-3
Bizkarguenaga M, Bruzzone C, Gil-Redondo R, SanJuan I, Martin-Ruiz I, Barriales D, Palacios A,
Pasco ST, González-Valle B, Laín A, Herrera L, Azkarate A, Vesga MA, Eguizabal C, Anguita
J, Embade N, Mato JM, Millet O (2021) Uneven metabolic and lipidomic profiles in recovered
COVID-19 patients as investigated by plasma NMR metabolomics. NMR Biomed 35:e4637.
https://doi.org/10.1002/nbm.4637
Borghaei H, Paz-Ares L, Horn L, Spigel DR, Steins M, Ready NE, Chow LQ, Vokes EE, Felip E,
Holgado E, Barlesi F, Kohlhäufl M, Arrieta O, Burgio MA, Fayette J, Lena H, Poddubskaya E,
Gerber DE, Gettinger SN, Rudin CM, Rizvi N, Crinò L, Blumenschein GR, Antonia SJ,
Dorange C, Harbison CT, Graf Finckenstein F, Brahmer JR (2015) Nivolumab versus docetaxel
in advanced nonsquamous non-small-cell lung cancer. N Engl J Med 373:1627–1639. https://
doi.org/10.1056/NEJMoa1507643
Brahmer JR (2014) Immune checkpoint blockade: the hope for immunotherapy as a treatment of
lung cancer? Semin Oncol 41:126–132. https://doi.org/10.1053/j.seminoncol.2013.12.014
236 A. Vignoli et al.

Bruzzone C, Bizkarguenaga M, Gil-Redondo R, Diercks T, Arana E, García de Vicuña A, Seco M,

Bosch A, Palazón A, San Juan I, Laín A, Gil-Martínez J, Bernardo-Seisdedos G, Fernández-
Ramos D, Lopitz-Otsoa F, Embade N, Lu S, Mato JM, Millet O (2020) SARS-CoV-2 infection
dysregulates the metabolomic and lipidomic profiles of serum. iScience 23:101645. https://doi.
org/10.1016/j.isci.2020.101645
Buchwald H, Estok R, Fahrbach K, Banel D, Jensen MD, Pories WJ, Bantle JP, Sledge I (2009)
Weight and type 2 diabetes after bariatric surgery: systematic review and meta-analysis. Am J
Med 122:248–256.e5. https://doi.org/10.1016/j.amjmed.2008.09.041
Cacciatore S, Hu X, Viertler C, Kap M, Bernhardt GA, Mischinger H-J, Riegman P, Zatloukal K,
Luchinat C, Turano P (2013) Effects of intra- and post-operative ischemia on the metabolic
profile of clinical liver tissue specimens monitored by NMR. J Proteome Res 12:5723–5729.
https://doi.org/10.1021/pr400702d
Cala MP, Aldana J, Medina J, Sánchez J, Guio J, Wist J, Meesters RJW (2018) Multiplatform
plasma metabolic and lipid fingerprinting of breast cancer: a pilot control-case study in
Colombian Hispanic women. PLoS One 13:e0190958. https://doi.org/10.1371/journal.pone.
0190958
Cano A, Mariño Z, Millet O, Martínez-Arranz I, Navasa M, Falcón-Pérez JM, Pérez-Cormenzana
M, Caballería J, Embade N, Forns X, Bosch J, Castro A, Mato JM (2017) A metabolomics
signature linked to liver fibrosis in the serum of transplanted hepatitis C patients. Sci Rep
7:10497. https://doi.org/10.1038/s41598-017-10807-y
Chen P-A, Xu Z-H, Huang Y-L, Luo Y, Zhu D-J, Wang P, Du Z-Y, Yang Y, Wu D-H, Lai W-Y,
Ren H, Xu D-L (2014) Increased serum 2-oxoglutarate associated with high myocardial energy
expenditure and poor prognosis in chronic heart failure patients. Biochim Biophys Acta 1842:
2120–2125. https://doi.org/10.1016/j.bbadis.2014.07.018
Choi JS, Baek H-M, Kim S, Kim MJ, Youk JH, Moon HJ, Kim E-K, Nam YK (2013) Magnetic
resonance metabolic profiling of breast cancer tissue obtained with core needle biopsy for
predicting pathologic response to neoadjuvant chemotherapy. PLoS One 8:e83866. https://
doi.org/10.1371/journal.pone.0083866
Choucha Snouber L, Bunescu A, Naudot M, Legallais C, Brochot C, Dumas ME, Elena-Herrmann-
B, Leclerc E (2013) Metabolomics-on-a-chip of hepatotoxicity induced by anticancer drug
flutamide and its active metabolite hydroxyflutamide using HepG2/C3a microfluidic biochips.
Toxicol Sci 132:8–20. https://doi.org/10.1093/toxsci/kfs230
Corona G, Di Gregorio E, Vignoli A, Muraro E, Steffan A, Miolo G (2021) 1H-NMR plasma
lipoproteins profile analysis reveals lipid metabolism alterations in HER2-positive breast cancer
patients. Cancers 13:5845. https://doi.org/10.3390/cancers13225845
Corona G, Rastrelli G, Burri A, Serra E, Gianfrilli D, Mannucci E, Jannini EA, Maggi M (2016)
First-generation phosphodiesterase type 5 inhibitors dropout: a comprehensive review and meta-
analysis. Andrology 4:1002–1009. https://doi.org/10.1111/andr.12255
Cunningham D, Atkin W, Lenz H-J, Lynch HT, Minsky B, Nordlinger B, Starling N (2010)
Colorectal cancer. Lancet 375:1030–1047. https://doi.org/10.1016/S0140-6736(10)60353-4
Cuperlovic-Culf M, Ferguson D, Culf A, Morin P, Touaibia M (2012) 1H NMR metabolomics
analysis of glioblastoma subtypes correlation between metabolomics and gene expression
characteristics. J Biol Chem 287:20164–20175. https://doi.org/10.1074/jbc.M111.337196
D’Alessandro G, Quaglio D, Monaco L, Lauro C, Ghirga F, Ingallina C, De Martino M, Fucile S,
Porzia A, Di Castro MA, Bellato F, Mastrotto F, Mori M, Infante P, Turano P, Salmaso S,
Caliceti P, Di Marcotullio L, Botta B, Ghini V, Limatola C (2019) 1H-NMR metabolomics
reveals the Glabrescione B exacerbation of glycolytic metabolism beside the cell growth
inhibitory effect in glioma. Cell Commun Signal 17:108. https://doi.org/10.1186/s12964-019-
0421-8
Debik J, Euceda LR, Lundgren S, Gythfeldt H v d L, Garred Ø, Borgen E, Engebraaten O, Bathen
TF, Giskeødegård GF (2019) Assessing treatment response and prognosis by serum and tissue
metabolomics in breast cancer patients. J Proteome Res 18:3649–3660. https://doi.org/10.1021/
acs.jproteome.9b00316
NMR-Based Metabolomics to Evaluate Individual Response to Treatments 237

Dekker E, Tanis PJ, Vleugels JLA, Kasi PM, Wallace MB (2019) Colorectal cancer. Lancet 394:
1467–1480. https://doi.org/10.1016/S0140-6736(19)32319-0
Di Donato S, Vignoli A, Biagioni C, Malorni L, Mori E, Tenori L, Calamai V, Parnofiello A, Di
Pierro G, Migliaccio I, Cantafio S, Baraghini M, Mottino G, Becheri D, Del Monte F, Miceli E,
McCartney A, Di Leo A, Luchinat C, Biganzoli L (2021) A serum metabolomics classifier
derived from elderly patients with metastatic colorectal cancer predicts relapse in the adjuvant
setting. Cancers 13:2762. https://doi.org/10.3390/cancers13112762
Díaz-Beltrán L, González-Olmedo C, Luque-Caro N, Díaz C, Martín-Blázquez A, Fernández-
Navarro M, Ortega-Granados AL, Gálvez-Montosa F, Vicente F, Pérez del Palacio J,
Sánchez-Rovira P (2021) Human plasma metabolomics for biomarker discovery: targeting the
molecular subtypes in breast cancer. Cancers 13:147. https://doi.org/10.3390/cancers13010147
Du Z, Shen A, Huang Y, Su L, Lai W, Wang P, Xie Z, Xie Z, Zeng Q, Ren H, Xu D (2014)
1H-NMR-based metabolic analysis of human serum reveals novel markers of myocardial
energy expenditure in heart failure patients. PLoS One 9:e88102. https://doi.org/10.1371/
journal.pone.0088102
Dykstra MA, Switzer N, Eisner R, Tso V, Foshaug R, Ismond K, Fedorak R, Wang H (2017) Urine
metabolomics as a predictor of patient tolerance and response to adjuvant chemotherapy in
colorectal cancer. Mol Clin Oncol 7:767–770. https://doi.org/10.3892/mco.2017.1407
Early Breast Cancer Trialists’ Collaborative Group (EBCTCG) (2005) Effects of chemotherapy and
hormonal therapy for early breast cancer on recurrence and 15-year survival: an overview of the
randomised trials. Lancet 365:1687–1717. https://doi.org/10.1016/S0140-6736(05)66544-0
Euceda LR, Haukaas TH, Giskeødegård GF, Vettukattil MR, Engel J, Silwal-Pandit L, Lundgren S,
Borgen E, Garred Ø, Postma G, Buydens LMC, Børresen-Dale A-L, Engebråten O, Bathen TF
(2017) Evaluation of metabolomic changes during neoadjuvant chemotherapy combined with
bevacizumab in breast cancer using MR spectroscopy. Metabolomics 13:37. https://doi.org/10.
1007/s11306-017-1168-0
Fan Y, Zhou X, Xia T-S, Chen Z, Li J, Liu Q, Alolga RN, Chen Y, Lai M-D, Li P, Zhu W, Qi L-W
(2016) Human plasma metabolomics for identifying differential metabolites and predicting
molecular subtypes of breast cancer. Oncotarget 7:9925–9938. https://doi.org/10.18632/
oncotarget.7155
Gandhi L, Rodríguez-Abreu D, Gadgeel S, Esteban E, Felip E, De Angelis F, Domine M,
Clingan P, Hochmair MJ, Powell SF, Cheng SY-S, Bischoff HG, Peled N, Grossi F, Jennens
RR, Reck M, Hui R, Garon EB, Boyer M, Rubio-Viqueira B, Novello S, Kurata T, Gray JE,
Vida J, Wei Z, Yang J, Raftopoulos H, Pietanza MC, Garassino MC, KEYNOTE-189
Investigators (2018) Pembrolizumab plus chemotherapy in metastatic non-small-cell lung
cancer. N Engl J Med 378:2078–2092. https://doi.org/10.1056/NEJMoa1801005
Garrod S, Bollard ME, Nicholls AW, Connor SC, Connelly J, Nicholson JK, Holmes E (2005)
Integrated metabonomic analysis of the multiorgan effects of hydrazine toxicity in the rat. Chem
Res Toxicol 18:115–122. https://doi.org/10.1021/tx0498915
Gettinger SN, Horn L, Gandhi L, Spigel DR, Antonia SJ, Rizvi NA, Powderly JD, Heist RS,
Carvajal RD, Jackman DM, Sequist LV, Smith DC, Leming P, Carbone DP, Pinder-Schenck
MC, Topalian SL, Hodi FS, Sosman JA, Sznol M, McDermott DF, Pardoll DM, Sankar V,
Ahlers CM, Salvati M, Wigginton JM, Hellmann MD, Kollia GD, Gupta AK, Brahmer JR
(2015) Overall survival and long-term safety of nivolumab (anti-programmed death 1 antibody,
BMS-936558, ONO-4538) in patients with previously treated advanced non-small-cell lung
cancer. J Clin Oncol 33:2004–2012. https://doi.org/10.1200/JCO.2014.58.3708
Ghini V, Di Nunzio M, Tenori L, Valli V, Danesi F, Capozzi F, Luchinat C, Bordoni A (2017)
Evidence of a DHA signature in the lipidome and metabolome of human hepatocytes. Int J Mol
Sci 18:359. https://doi.org/10.3390/ijms18020359
Ghini V, Laera L, Fantechi B, del Monte F, Benelli M, McCartney A, Tenori L, Luchinat C,
Pozzessere D (2020a) Metabolomics to assess response to immune checkpoint inhibitors in
patients with non-small-cell lung cancer. Cancers 12:3574. https://doi.org/10.3390/
cancers12123574
238 A. Vignoli et al.

Ghini V, Maggi L, Mazzoni A, Spinicci M, Zammarchi L, Bartoloni A, Annunziato F, Turano P

(2022a) Serum NMR profiling reveals differential alterations in the lipoproteome induced by
Pfizer-BioNTech vaccine in COVID-19 recovered subjects and Naïve subjects. Front Mol
Biosci 9:839809. https://doi.org/10.3389/fmolb.2022.839809
Ghini V, Meoni G, Pelagatti L, Celli T, Veneziani F, Petrucci F, Vannucchi V, Bertini L,
Luchinat C, Landini G, Turano P (2022b) Profiling metabolites and lipoproteins in COMETA,
an Italian cohort of COVID-19 patients. PLoS Pathog 18:e1010443. https://doi.org/10.1371/
journal.ppat.1010443
Ghini V, Quaglio D, Luchinat C, Turano P (2019) NMR for sample quality assessment in
metabolomics. N Biotechnol 52:25–34. https://doi.org/10.1016/j.nbt.2019.04.004
Ghini V, Saccenti E, Tenori L, Assfalg M, Luchinat C (2015a) Allostasis and resilience of the
human individual metabolic phenotype. J Proteome Res 14:2951–2962. https://doi.org/10.1021/
acs.jproteome.5b00275
Ghini V, Senzacqua T, Massai L, Gamberi T, Messori L, Turano P (2021) NMR reveals the
metabolic changes induced by auranofin in A2780 cancer cells: evidence for glutathione
dysregulation. Dalton Trans 50:6349–6355. https://doi.org/10.1039/D1DT00750E
Ghini V, Tenori L, Pane M, Amoruso A, Marroncini G, Squarzanti DF, Azzimonti B, Rolla R,
Savoia P, Tarocchi M, Galli A, Luchinat C (2020b) Effects of probiotics administration on
human metabolic phenotype. Metabolites 10:396. https://doi.org/10.3390/metabo10100396
Ghini V, Unger FT, Tenori L, Turano P, Juhl H, David KA (2015b) Metabolomics profiling of
pre-and post-anesthesia plasma samples of colorectal patients obtained via Ficoll separation.
Metabolomics 11:1769–1778. https://doi.org/10.1007/s11306-015-0832-5
Ghosh N, Dutta M, Singh B, Banerjee R, Bhattacharyya P, Chaudhury K (2016) Transcriptomics,
proteomics and metabolomics driven biomarker discovery in COPD: an update. Expert Rev Mol
Diagn 16:897–913. https://doi.org/10.1080/14737159.2016.1198258
Gianni L, Pienkowski T, Im Y-H, Roman L, Tseng L-M, Liu M-C, Lluch A, Staroslawska E, de la
Haba-Rodriguez J, Im S-A, Pedrini JL, Poirier B, Morandi P, Semiglazov V, Srimuninnimit V,
Bianchi G, Szado T, Ratnayake J, Ross G, Valagussa P (2012) Efficacy and safety of
neoadjuvant pertuzumab and trastuzumab in women with locally advanced, inflammatory, or
early HER2-positive breast cancer (NeoSphere): a randomised multicentre, open-label, phase
2 trial. Lancet Oncol 13:25–32. https://doi.org/10.1016/S1470-2045(11)70336-9
Godoy MGG, Lopes EPA, Silva RO, Hallwass F, Koury LCA, Moura IM, Gonçalves SMC, Simas
AM (2010) Hepatitis C virus infection diagnosis using metabonomics. J Viral Hepat 17:854–
858. https://doi.org/10.1111/j.1365-2893.2009.01252.x
Gralka E, Luchinat C, Tenori L, Ernst B, Thurnheer M, Schultes B (2015) Metabolomic fingerprint
of severe obesity is dynamically affected by bariatric surgery in a procedure-dependent manner.
Am J Clin Nutr 102:1313–1322. https://doi.org/10.3945/ajcn.115.110536
Griffiths WJ (2008) Metabolomics, metabonomics and metabolite profiling. RSC Publishing
Guraya SY (2019) Pattern, stage, and time of recurrent colorectal cancer after curative surgery. Clin
Colorectal Cancer 18:e223–e228. https://doi.org/10.1016/j.clcc.2019.01.003
Hamada T, Kosumi K, Nakai Y, Koike K (2018) Surrogate study endpoints in the era of cancer
immunotherapy. Ann Transl Med 6:S27. https://doi.org/10.21037/atm.2018.09.31
Hart CD, Vignoli A, Tenori L, Uy GL, To TV, Adebamowo C, Hossain SM, Biganzoli L, Risi E,
Love RR, Luchinat C, Di Leo A (2017) Serum metabolomic profiles identify ER-positive early
breast cancer patients at increased risk of disease recurrence in a multicenter population. Clin
Cancer Res 23:1422–1431. https://doi.org/10.1158/1078-0432.CCR-16-1153
Hatae R, Chamoto K, Kim YH, Sonomura K, Taneishi K, Kawaguchi S, Yoshida H, Ozasa H,
Sakamori Y, Akrami M, Fagarasan S, Masuda I, Okuno Y, Matsuda F, Hirai T, Honjo T (2020)
Combination of host immune metabolic biomarkers for the PD-1 blockade cancer immunother-
apy. JCI Insight 5:133501. https://doi.org/10.1172/jci.insight.133501
Herbst RS, Baas P, Kim D-W, Felip E, Pérez-Gracia JL, Han J-Y, Molina J, Kim J-H, Arvis CD,
Ahn M-J, Majem M, Fidler MJ, de Castro G, Garrido M, Lubiniecki GM, Shentu Y, Im E,
Dolled-Filhart M, Garon EB (2016) Pembrolizumab versus docetaxel for previously treated,
NMR-Based Metabolomics to Evaluate Individual Response to Treatments 239

PD-L1-positive, advanced non-small-cell lung cancer (KEYNOTE-010): a randomised con-

trolled trial. Lancet 387:1540–1550. https://doi.org/10.1016/S0140-6736(15)01281-7
Hirsch FR, Scagliotti GV, Mulshine JL, Kwon R, Curran WJ, Wu Y-L, Paz-Ares L (2017) Lung
cancer: current therapies and new targeted treatments. Lancet 389:299–311. https://doi.org/10.
1016/S0140-6736(16)30958-8
Holmes E, Loo RL, Stamler J, Bictash M, Yap IKS, Chan Q, Ebbels T, De Iorio M, Brown IJ,
Veselkov KA, Daviglus ML, Kesteloot H, Ueshima H, Zhao L, Nicholson JK, Elliott P (2008)
Human metabolic phenotype diversity and its association with diet and blood pressure. Nature
453:396–400. https://doi.org/10.1038/nature06882
ISO 23118:2021. Molecular in vitro diagnostic examinations – specifications for pre-examination
processes in metabolomics in urine, venous blood serum and plasma. https://www.iso.org/
standard/74605.html
Jiang L, Lee SC, Ng TC (2018) Pharmacometabonomics analysis reveals serum formate and acetate
potentially associated with varying response to gemcitabine-carboplatin chemotherapy in meta-
static breast cancer patients. J Proteome Res 17:1248–1257. https://doi.org/10.1021/acs.
jproteome.7b00859
Jobard E, Trédan O, Bachelot T, Vigneron AM, Aït-Oukhatar CM, Arnedos M, Rios M,
Bonneterre J, Diéras V, Jimenez M, Merlin J-L, Campone M, Elena-Herrmann B (2017)
Longitudinal serum metabolomics evaluation of trastuzumab and everolimus combination as
pre-operative treatment for HER-2 positive breast cancer patients. Oncotarget 8:83570–83584.
https://doi.org/10.18632/oncotarget.18784
Julkunen H, Cichońska A, Slagboom PE, Würtz P (2021) Metabolic biomarker profiling for
identification of susceptibility to severe pneumonia and COVID-19 in the general population.
Elife 10:e63033. https://doi.org/10.7554/eLife.63033
Kaddurah-Daouk R, Weinshilboum RM, Pharmacometabolomics Research Network (2014)
Pharmacometabolomics: implications for clinical pharmacology and systems pharmacology.
Clin Pharmacol Ther 95:154–167. https://doi.org/10.1038/clpt.2013.217
Kimhofer T, Lodge S, Whiley L, Gray N, Loo RL, Lawler NG, Nitschke P, Bong S-H, Morrison
DL, Begum S, Richards T, Yeap BB, Smith C, Smith KCG, Holmes E, Nicholson JK (2020)
Integrative modelling of quantitative plasma lipoprotein, metabolic and amino acid data reveals
a multi-organ pathological signature of SARS-CoV-2 infection. J Proteome Res 19:4442–4454.
https://doi.org/10.1021/acs.jproteome.0c00519
Klupczyńska A, Dereziński P, Kokot ZJ (2015) Metabolomics in medical sciences – trends,
challenges and perspectives. Acta Pol Pharm 72:629–641
de Laurentiis G, Paris D, Melck D, Maniscalco M, Marsico S, Corso G, Motta A, Sofia M (2008)
Metabonomic analysis of exhaled breath condensate in adults by nuclear magnetic resonance
spectroscopy. Eur Respir J 32:1175–1183. https://doi.org/10.1183/09031936.00072408
Le Gall G, Guttula K, Kellingray L, Tett AJ, ten Hoopen R, Kemsley KE, Savva GM, Ibrahim A,
Narbad A (2018) Metabolite quantification of faecal extracts from colorectal cancer patients and
healthy controls. Oncotarget 9:33278–33289. https://doi.org/10.18632/oncotarget.26022
Lécuyer L, Victor Bala A, Deschasaux M, Bouchemal N, Nawfal Triba M, Vasson M-P, Rossary A,
Demidem A, Galan P, Hercberg S, Partula V, Le Moyec L, Srour B, Fiolet T, Latino-Martel P,
Kesse-Guyot E, Savarin P, Touvier M (2018) NMR metabolomic signatures reveal predictive
plasma metabolites associated with long-term risk of developing breast cancer. Int J Epidemiol
47:484–494. https://doi.org/10.1093/ije/dyx271
Lenz EM, Bright J, Knight R, Westwood FR, Davies D, Major H, Wilson ID (2005) Metabonomics
with 1H-NMR spectroscopy and liquid chromatography-mass spectrometry applied to the
investigation of metabolic changes caused by gentamicin-induced nephrotoxicity in the rat.
Biomarkers 10:173–187. https://doi.org/10.1080/13547500500094034
Lenz EM, Bright J, Wilson ID, Morgan SR, Nash AFP (2003) A 1H NMR-based metabonomic
study of urine and plasma samples obtained from healthy human subjects. J Pharm Biomed Anal
33:1103–1115. https://doi.org/10.1016/S0731-7085(03)00410-2
240 A. Vignoli et al.

Lerner HJ, Band PR, Israel L, Leung BS (1976) Phase II study of tamoxifen: report of 74 patients
with stage IV breast cancer. Cancer Treat Rep 60:1431–1435
Li C, Li Z, Zhang T, Wei P, Li N, Zhang W, Ding X, Li J (2019) 1H NMR-based metabolomics
reveals the antitumor mechanisms of triptolide in BALB/c mice bearing CT26 Tumors. Front
Pharmacol 10:1175. https://doi.org/10.3389/fphar.2019.01175
Li Y, Wang C, Li D, Deng P, Shao X, Hu J, Liu C, Jie H, Lin Y, Li Z, Qian X, Zhang H, Zhao Y
(2017) 1H-NMR-based metabolic profiling of a colorectal cancer CT-26 lung metastasis model
in mice. Oncol Rep 38:3044–3054. https://doi.org/10.3892/or.2017.5954
Lin Y, Ma C, Bezabeh T, Wang Z, Liang J, Huang Y, Zhao J, Liu X, Ye W, Tang W, Ouyang T, Wu
R (2019) 1 H NMR-based metabolomics reveal overlapping discriminatory metabolites and
metabolic pathway disturbances between colorectal tumor tissues and fecal samples. Int J
Cancer 145:1679–1689. https://doi.org/10.1002/ijc.32190
Lin Y, Ma C, Liu C, Wang Z, Yang J, Liu X, Shen Z, Wu R (2016) NMR-based fecal metabolomics
fingerprinting as predictors of earlier diagnosis in patients with colorectal cancer. Oncotarget 7:
29454–29464. https://doi.org/10.18632/oncotarget.8762
Lindon JC, Nicholson JK, Holmes E, Antti H, Bollard ME, Keun H, Beckonert O, Ebbels TM,
Reily MD, Robertson D, Stevens GJ, Luke P, Breau AP, Cantor GH, Bible RH, Niederhauser U,
Senn H, Schlotterbeck G, Sidelmann UG, Laursen SM, Tymiak A, Car BD, Lehman-
McKeeman L, Colet J-M, Loukaci A, Thomas C (2003) Contemporary issues in toxicology
the role of metabonomics in toxicology and its evaluation by the COMET project. Toxicol Appl
Pharmacol 187:137–146. https://doi.org/10.1016/S0041-008X(02)00079-0
Lodge S, Nitschke P, Kimhofer T, Coudert JD, Begum S, Bong S-H, Richards T, Edgar D, Raby E,
Spraul M, Schaefer H, Lindon JC, Loo RL, Holmes E, Nicholson JK (2021) NMR spectroscopic
windows on the systemic effects of SARS-CoV-2 infection on plasma lipoproteins and
metabolites in relation to circulating cytokines. J Proteome Res 20:1382–1396. https://doi.org/
10.1021/acs.jproteome.0c00876
Marianne C, Walsh LB (2006) Effect of acute dietary standardization on the urinary, plasma, and
salivary metabolomic profiles of healthy humans. Am J Clin Nutr 84:531–539. https://doi.org/
10.1093/ajcn/84.3.531
Márquez J, Matés JM (2021) Tumor metabolome: therapeutic opportunities targeting cancer
metabolic reprogramming. Cancers (Basel) 13:314. https://doi.org/10.3390/cancers13020314
Martin M, Holmes FA, Ejlertsen B, Delaloge S, Moy B, Iwata H, von Minckwitz G, Chia SKL,
Mansi J, Barrios CH, Gnant M, Tomašević Z, Denduluri N, Šeparović R, Gokmen E,
Bashford A, Ruiz Borrego M, Kim S-B, Jakobsen EH, Ciceniene A, Inoue K, Overkamp F,
Heijns JB, Armstrong AC, Link JS, Joy AA, Bryce R, Wong A, Moran S, Yao B, Xu F,
Auerbach A, Buyse M, Chan A, ExteNET Study Group (2017) Neratinib after trastuzumab-
based adjuvant therapy in HER2-positive breast cancer (ExteNET): 5-year analysis of a
randomised, double-blind, placebo-controlled, phase 3 trial. Lancet Oncol 18:1688–1700.
https://doi.org/10.1016/S1470-2045(17)30717-9
Masuda R, Lodge S, Nitschke P, Spraul M, Schaefer H, Bong S-H, Kimhofer T, Hall D, Loo RL,
Bizkarguenaga M, Bruzzone C, Gil-Redondo R, Embade N, Mato JM, Holmes E, Wist J,
Millet O, Nicholson JK (2021) Integrative modeling of plasma metabolic and lipoprotein
biomarkers of SARS-CoV-2 infection in Spanish and Australian COVID-19 patient cohorts. J
Proteome Res 20:4139–4152. https://doi.org/10.1021/acs.jproteome.1c00458
McCartney A, Vignoli A, Biganzoli L, Love R, Tenori L, Luchinat C, Di Leo A (2018)
Metabolomics in breast cancer: a decade in review. Cancer Treat Rev 67:88–96. https://doi.
org/10.1016/j.ctrv.2018.04.012
McCartney A, Vignoli A, Hart C, Tenori L, Luchinat C, Biganzoli L, Di Leo A (2017)
De-escalating and escalating treatment beyond endocrine therapy in patients with luminal breast
cancer. Breast 34:S13–S18. https://doi.org/10.1016/j.breast.2017.06.021
McCartney A, Vignoli A, Tenori L, Fornier M, Rossi L, Risi E, Luchinat C, Biganzoli L, Di Leo A
(2019) Metabolomic analysis of serum may refine 21-gene expression assay risk recurrence
stratification. NPJ Br Cancer 5:26. https://doi.org/10.1038/s41523-019-0123-9
NMR-Based Metabolomics to Evaluate Individual Response to Treatments 241

Meijer RI, van Wagensveld BA, Siegert CE, Eringa EC, Smulders YM (2011) Bariatric surgery as a
novel treatment for type 2 diabetes mellitus: a systematic review. Arch Surg 146:744–750.
https://doi.org/10.1001/archsurg.2011.134
Meoni G, Ghini V, Maggi L, Vignoli A, Mazzoni A, Salvati L, Capone M, Vanni A, Tenori L,
Fontanari P, Lavorini F, Peris A, Bartoloni A, Liotta F, Cosmi L, Luchinat C, Annunziato F,
Turano P (2021) Metabolomic/lipidomic profiling of COVID-19 and individual response to
tocilizumab. PLoS Pathog 17:e1009243. https://doi.org/10.1371/journal.ppat.1009243
Meoni G, Lorini S, Monti M, Madia F, Corti G, Luchinat C, Zignego AL, Tenori L, Gragnani L
(2019) The metabolic fingerprints of HCV and HBV infections studied by nuclear magnetic
resonance spectroscopy. Sci Rep 9:4128. https://doi.org/10.1038/s41598-019-40028-4
Monleón D, Morales JM, Barrasa A, López JA, Vázquez C, Celda B (2009) Metabolite profiling of
fecal water extracts from human colorectal cancer. NMR Biomed 22:342–348. https://doi.org/
10.1002/nbm.1345
Montuschi P, Santini G, Mores N, Vignoli A, Macagno F, Shohreh R, Tenori L, Zini G, Fuso L,
Mondino C, Di Natale C, D’Amico A, Barnes PJ, Higenbottam T (2018) BREATHOMICS for
ASSESSING the effects of treatment and withdrawal with inhaled beclomethasone/formoterol
in patients with COPD. Front Pharmacol 9:258. https://doi.org/10.3389/fphar.2018.00258
Motta A, Paris D, Melck D, de Laurentiis G, Maniscalco M, Sofia M, Montuschi P (2012) Nuclear
magnetic resonance-based metabolomics of exhaled breath condensate: methodological aspects.
Eur Respir J 39:498–500. https://doi.org/10.1183/09031936.00036411
Nannini G, Meoni G, Amedei A, Tenori L (2020) Metabolomics profile in gastrointestinal cancers:
update and future perspectives. World J Gastroenterol 26:2514–2532. https://doi.org/10.3748/
wjg.v26.i20.2514
Nannini G, Meoni G, Tenori L, Ringressi MN, Taddei A, Niccolai E, Baldi S, Russo E, Luchinat C,
Amedei A (2021) Fecal metabolomic profiles: a comparative study of patients with colorectal
cancer vs adenomatous polyps. World J Gastroenterol 27:6430–6441. https://doi.org/10.3748/
wjg.v27.i38.6430
Nguyen HTT, Wimmer R, Le VQ, Krarup HB (2021) Metabolic fingerprint of progression of
chronic hepatitis B: changes in the metabolome and novel diagnostic possibilities.
Metabolomics 17:16. https://doi.org/10.1007/s11306-020-01767-y
Nicholson JK, Connelly J, Lindon JC, Holmes E (2002) Metabonomics: a platform for studying
drug toxicity and gene function. Nat Rev Drug Discov 1:153–161. https://doi.org/10.1038/
nrd728
Nie X, Xia L, Gao F, Liu L, Yang Y, Chen Y, Duan H, Yao Y, Chen Z, Lu S, Wang Y, Yang C
(2021) Serum metabolite biomarkers predictive of response to PD-1 blockade therapy in
non-small cell lung cancer. Front Mol Biosci 8:472. https://doi.org/10.3389/fmolb.2021.678753
Okuda M, Li K, Beard MR, Showalter LA, Scholle F, Lemon SM, Weinman SA (2002) Mitochon-
drial injury, oxidative stress, and antioxidant gene expression are induced by hepatitis C virus
core protein. Gastroenterology 122:366–375. https://doi.org/10.1053/gast.2002.30983
O’Sullivan A, Gibney MJ, Brennan L (2011) Dietary intake patterns are reflected in metabolomic
profiles: potential role in dietary assessment studies. Am J Clin Nutr 93:314–321. https://doi.
org/10.3945/ajcn.110.000950
Polyak K (2011) Heterogeneity in breast cancer. J Clin Invest 121:3786–3788. https://doi.org/10.
1172/JCI60534
Postow MA, Callahan MK, Wolchok JD (2015) Immune checkpoint blockade in cancer therapy. J
Clin Oncol 33:1974–1982. https://doi.org/10.1200/JCO.2014.59.4358
Ramirez T, Daneshian M, Kamp H, Bois FY, Clench MR, Coen M, Donley B, Fischer SM, Ekman
DR, Fabian E, Guillou C, Heuer J, Hogberg HT, Jungnickel H, Keun HC, Krennrich G, Krupp
E, Luch A, Noor F, Peter E, Riefke B, Seymour M, Skinner N, Smirnova L, Verheij E, Wagner
S, Hartung T, van Ravenzwaay B, Leist M (2013) Metabolomics in toxicology and preclinical
research. ALTEX 30:209–225. https://doi.org/10.14573/altex.2013.2.209
242 A. Vignoli et al.

Resendiz-Acevedo K, García-Aguilera ME, Esturau-Escofet N, Ruiz-Azuara L (2021) 1H -NMR

metabolomics study of the effect of cisplatin and CasiopeinaIIgly on MDA-MB-231 breast
tumor cells. Front Mol Biosci 8:1135. https://doi.org/10.3389/fmolb.2021.742859
Ringehan M, McKeating JA, Protzer U (2017) Viral hepatitis and liver cancer. Philos Trans R Soc
Lond B Biol Sci 372:20160274. https://doi.org/10.1098/rstb.2016.0274
Rittmeyer A, Barlesi F, Waterkamp D, Park K, Ciardiello F, von Pawel J, Gadgeel SM, Hida T,
Kowalski DM, Dols MC, Cortinovis DL, Leach J, Polikoff J, Barrios C, Kabbinavar F, Frontera
OA, De Marinis F, Turna H, Lee J-S, Ballinger M, Kowanetz M, He P, Chen DS, Sandler A,
Gandara DR, OAK Study Group (2017) Atezolizumab versus docetaxel in patients with
previously treated non-small-cell lung cancer (OAK): a phase 3, open-label, multicentre
randomised controlled trial. Lancet 389:255–265. https://doi.org/10.1016/S0140-6736(16)
32517-X
Rocca MS, Vignoli A, Tenori L, Ghezzi M, De Rocco Ponce M, Vatsellas G, Thanos D, Padrini R,
Foresta C, De Toni L (2020) Evaluation of serum/urine genomic and metabolomic profiles to
improve the adherence to sildenafil therapy in patients with erectile dysfunction. Front
Pharmacol 11:602369. https://doi.org/10.3389/fphar.2020.602369
Saborano R, Eraslan Z, Roberts J, Khanim FL, Lalor PF, Reed MAC, Günther UL (2019) A
framework for tracer-based metabolism in mammalian cells by NMR. Sci Rep 9:2520. https://
doi.org/10.1038/s41598-018-37525-3
Salmerón AM, Tristán AI, Abreu AC, Fernández I (2022) Serum colorectal cancer biomarkers
unraveled by NMR metabolomics: past, present, and future. Anal Chem 94:417–430. https://
doi.org/10.1021/acs.analchem.1c04360
Sarfaraz MO, Myers RP, Coffin CS, Gao Z-H, Shaheen AAM, Crotty PM, Zhang P, Vogel HJ,
Weljie AM (2016) A quantitative metabolomics profiling approach for the noninvasive assess-
ment of liver histology in patients with chronic hepatitis C. Clin Transl Med 5:33. https://doi.
org/10.1186/s40169-016-0109-2
Siegel RL, Miller KD, Fuchs HE, Jemal A (2021) Cancer statistics, 2021. CA Cancer J Clin 71:7–
33. https://doi.org/10.3322/caac.21654
Silva CL, Olival A, Perestrelo R, Silva P, Tomás H, Câmara JS (2019) Untargeted urinary 1H
NMR-based metabolomic pattern as a potential platform in breast cancer detection. Metabolites
9:269. https://doi.org/10.3390/metabo9110269
Singh A, Sharma RK, Chagtoo M, Agarwal G, George N, Sinha N, Godbole MM (2017) 1H NMR
metabolomics reveals association of high expression of inositol 1, 4, 5 trisphosphate receptor
and metabolites in breast cancer patients. PLoS One 12:e0169330. https://doi.org/10.1371/
journal.pone.0169330
Sirniö P, Väyrynen JP, Klintrup K, Mäkelä J, Karhu T, Herzig K-H, Minkkinen I, Mäkinen MJ,
Karttunen TJ, Tuomisto A (2019) Alterations in serum amino-acid profile in the progression of
colorectal cancer: associations with systemic inflammation, tumour stage and patient survival.
Br J Cancer 120:238–246. https://doi.org/10.1038/s41416-018-0357-6
Sitter B, Lundgren S, Bathen TF, Halgunset J, Fjosne HE, Gribbestad IS (2006) Comparison of HR
MAS MR spectroscopic profiles of breast cancer tissue with clinical parameters. NMR Biomed
19:30–40. https://doi.org/10.1002/nbm.992
Slamon D, Eiermann W, Robert N, Pienkowski T, Martin M, Press M, Mackey J, Glaspy J, Chan A,
Pawlicki M, Pinter T, Valero V, Liu M-C, Sauter G, von Minckwitz G, Visco F, Bee V,
Buyse M, Bendahmane B, Tabah-Fisch I, Lindsay M-A, Riva A, Crown J, Breast Cancer
International Research Group (2011) Adjuvant trastuzumab in HER2-positive breast cancer.
N Engl J Med 365:1273–1283. https://doi.org/10.1056/NEJMoa0910383
Slupsky CM, Steed H, Wells T, Dabbs K, Schepansky A, Capstick V, Faught W, Sawyer MB
(2010) Urine metabolite analysis offers potential early diagnosis of ovarian and breast cancers.
Clin Cancer Res 16:5835–5841. https://doi.org/10.1158/1078-0432.CCR-10-1434
Suman S, Sharma RK, Kumar V, Sinha N, Shukla Y (2018) Metabolic fingerprinting in breast
cancer stages through 1H NMR spectroscopy-based metabolomic analysis of plasma. J Pharm
Biomed Anal 160:38–45. https://doi.org/10.1016/j.jpba.2018.07.024
NMR-Based Metabolomics to Evaluate Individual Response to Treatments 243

Sung H, Ferlay J, Siegel RL, Laversanne M, Soerjomataram I, Jemal A, Bray F (2021) Global
cancer statistics 2020: GLOBOCAN estimates of incidence and mortality worldwide for
36 cancers in 185 countries. CA Cancer J Clin 71:209–249. https://doi.org/10.3322/caac.21660
Swain SM, Kim S-B, Cortés J, Ro J, Semiglazov V, Campone M, Ciruelos E, Ferrero J-M,
Schneeweiss A, Knott A, Clark E, Ross G, Benyunes MC, Baselga J (2013) Pertuzumab,
trastuzumab, and docetaxel for HER2-positive metastatic breast cancer (CLEOPATRA
study): overall survival results from a randomised, double-blind, placebo-controlled, phase
3 study. Lancet Oncol 14:461–471. https://doi.org/10.1016/S1470-2045(13)70130-X
Takis PG, Ghini V, Tenori L, Turano P, Luchinat C (2019) Uniqueness of the NMR approach to
metabolomics. TRAC-Trends Anal Chem 120:115300. https://doi.org/10.1016/j.trac.2018.10.
036
Tan B, Qiu Y, Zou X, Chen T, Xie G, Cheng Y, Dong T, Zhao L, Feng B, Hu X, Xu LX, Zhao A,
Zhang M, Cai G, Cai S, Zhou Z, Zheng M, Zhang Y, Jia W (2013) Metabonomics identifies
serum metabolite markers of colorectal cancer. J Proteome Res 12:3000–3009. https://doi.org/
10.1021/pr400337b
Tayyari F, Gowda GAN, Olopade OF, Berg R, Yang HH, Lee MP, Ngwa WF, Mittal SK,
Raftery D, Mohammed SI (2018) Metabolic profiles of triple-negative and luminal A breast
cancer subtypes in African-American identify key metabolic differences. Oncotarget 9:11677–
11690. https://doi.org/10.18632/oncotarget.24433
Tenori L, Oakman C, Claudino WM, Bernini P, Cappadona S, Nepi S, Biganzoli L, Arbushites MC,
Luchinat C, Bertini I, Di Leo A (2012) Exploration of serum metabolomic profiles and outcomes
in women with metastatic breast cancer: a pilot study. Mol Oncol 6:437–444. https://doi.org/10.
1016/j.molonc.2012.05.003
Tenori L, Oakman C, Morris PG, Gralka E, Turner N, Cappadona S, Fornier M, Hudis C, Norton L,
Luchinat C, Di Leo A (2015) Serum metabolomic profiles evaluated after surgery may identify
patients with oestrogen receptor negative early breast cancer at increased risk of disease
recurrence. Results from a retrospective study. Mol Oncol 9:128–139. https://doi.org/10.1016/
j.molonc.2014.07.012
Thompson AM, Moulder-Thompson SL (2012) Neoadjuvant treatment of breast cancer. Ann Oncol
23:x231–x236. https://doi.org/10.1093/annonc/mds324
Turano P (2014) Colorectal cancer: the potential of metabolic fingerprinting. Expert Rev
Gastroenterol Hepatol 8:847–849. https://doi.org/10.1586/17474124.2014.945912
Verma S, Miles D, Gianni L, Krop IE, Welslau M, Baselga J, Pegram M, Oh D-Y, Diéras V,
Guardino E, Fang L, Lu MW, Olsen S, Blackwell K, EMILIA Study Group (2012) Trastuzumab
emtansine for HER2-positive advanced breast cancer. N Engl J Med 367:1783–1791. https://
doi.org/10.1056/NEJMoa1209124
Vignoli A, Ghini V, Meoni G, Licari C, Takis PG, Tenori L, Turano P, Luchinat C (2019) High-
throughput metabolomics by 1D NMR. Angew Chem Int Ed 58:968–994. https://doi.org/10.
1002/anie.201804736
Vignoli A, Mori E, Di Donato S, Malorni L, Biagioni C, Benelli M, Calamai V, Cantafio S,
Parnofiello A, Baraghini M, Garzi A, Monte FD, Romagnoli D, Migliaccio I, Luchinat C,
Tenori L, Biganzoli L (2021a) Exploring serum NMR-based metabolomic fingerprint of
colorectal cancer patients: effects of surgery and possible associations with cancer relapse.
Appl Sci 11:11120. https://doi.org/10.3390/app112311120
Vignoli A, Muraro E, Miolo G, Tenori L, Turano P, Di Gregorio E, Steffan A, Luchinat C, Corona
G (2020a) Effect of Estrogen receptor status on circulatory immune and metabolomics profiles
of HER2-positive breast cancer patients enrolled for neoadjuvant targeted chemotherapy.
Cancers 12:314. https://doi.org/10.3390/cancers12020314
Vignoli A, Orlandini B, Tenori L, Biagini MR, Milani S, Renzi D, Luchinat C, Calabrò AS (2019)
Metabolic signature of primary biliary cholangitis and its comparison with celiac disease. J
Proteome Res 18:1228–1236. https://doi.org/10.1021/acs.jproteome.8b00849
244 A. Vignoli et al.

Vignoli A, Risi E, McCartney A, Migliaccio I, Moretti E, Malorni L, Luchinat C, Biganzoli L,

Tenori L (2021b) Precision oncology via NMR-based metabolomics: a review on breast cancer.
Int J Mol Sci 22:4687. https://doi.org/10.3390/ijms22094687
Vignoli A, Santini G, Tenori L, Macis G, Mores N, Macagno F, Pagano F, Higenbottam T,
Luchinat C, Montuschi P (2020b) NMR-based metabolomics for the assessment of inhaled
pharmacotherapy in chronic obstructive pulmonary disease patients. J Proteome Res 19:64–74.
https://doi.org/10.1021/acs.jproteome.9b00345
Vignoli A, Tenori L, Morsiani C, Turano P, Capri M, Luchinat C (2022) Serum or plasma (and
which plasma), that is the question. J Proteome Res 21:1061–1072. https://doi.org/10.1021/acs.
jproteome.1c00935
von Minckwitz G, Huang C-S, Mano MS, Loibl S, Mamounas EP, Untch M, Wolmark N,
Rastogi P, Schneeweiss A, Redondo A, Fischer HH, Jacot W, Conlin AK, Arce-Salinas C,
Wapnir IL, Jackisch C, DiGiovanna MP, Fasching PA, Crown JP, Wülfing P, Shao Z, Rota
Caremoli E, Wu H, Lam LH, Tesarowski D, Smitt M, Douthwaite H, Singel SM, Geyer CE,
Katherine Investigators (2019) Trastuzumab emtansine for residual invasive HER2-positive
breast cancer. N Engl J Med 380:617–628. https://doi.org/10.1056/NEJMoa1814017
von Minckwitz G, Procter M, de Azambuja E, Zardavas D, Benyunes M, Viale G, Suter T,
Arahmani A, Rouchet N, Clark E, Knott A, Lang I, Levy C, Yardley DA, Bines J, Gelber
RD, Piccart M, Baselga J, Aphinity Steering Committee and Investigators (2017) Adjuvant
pertuzumab and trastuzumab in early HER2-positive breast cancer. N Engl J Med 377:122–131.
https://doi.org/10.1056/NEJMoa1703643
Wallner-Liebmann S, Gralka E, Tenori L, Konrad M, Hofmann P, Dieber-Rotheneder M, Turano P,
Luchinat C, Zatloukal K (2015) The impact of free or standardized lifestyle and urine sampling
protocol on metabolome recognition accuracy. Genes Nutr 10:441. https://doi.org/10.1007/
s12263-014-0441-3
Wallner-Liebmann S, Tenori L, Mazzoleni A, Dieber-Rotheneder M, Konrad M, Hofmann P,
Luchinat C, Turano P, Zatloukal K (2016) Individual human metabolic phenotype analyzed
by (1)H NMR of saliva samples. J Proteome Res 15:1787–1793. https://doi.org/10.1021/acs.
jproteome.5b01060
Wang L, Tang Y, Liu S, Mao S, Ling Y, Liu D, He X, Wang X (2013) Metabonomic profiling of
serum and urine by 1H NMR-based spectroscopy discriminates patients with chronic obstruc-
tive pulmonary disease and healthy individuals. PLoS One 8:e65675. https://doi.org/10.1371/
journal.pone.0065675
Wang P, Shehu AI, Ma X (2017) The opportunities of metabolomics in drug safety evaluation. Curr
Pharmacol Rep 3:10–15. https://doi.org/10.1007/s40495-016-0079-5
Wei S, Liu L, Zhang J, Bowers J, Gowda GAN, Seeger H, Fehm T, Neubauer HJ, Vogel U, Clare
SE, Raftery D (2013) Metabolomics approach for predicting response to neoadjuvant chemo-
therapy for breast cancer. Mol Oncol 7:297–307. https://doi.org/10.1016/j.molonc.2012.10.003
Wiggans RG, Woolley PV, Smythe T, Hoth D, Macdonald JS, Green L, Schein PS (1979) Phase-II
trial of tamoxifen in advanced breast cancer. Cancer Chemother Pharmacol 3:45–48. https://doi.
org/10.1007/BF00254419
Winnike JH, Busby MG, Watkins PB, O’Connell TM (2009) Effects of a prolonged standardized
diet on normalizing the human metabolome. Am J Clin Nutr 90:1496–1501. https://doi.org/10.
3945/ajcn.2009.28234
Wishart DS (2015) Is cancer a genetic disease or a metabolic disease? EBioMedicine 2:478–479.
https://doi.org/10.1016/j.ebiom.2015.05.022
Wojtowicz W, Wróbel A, Pyziak K, Tarkowski R, Balcerzak A, Bębenek M, Młynarz P (2020)
Evaluation of MDA-MB-468 cell culture media analysis in predicting triple-negative breast
cancer patient sera metabolic profiles. Metabolites 10:173. https://doi.org/10.3390/
metabo10050173
World Health Organization (2021) Global progress report on HIV, viral hepatitis and sexually
transmitted infections, 2021: accountability for the global health sector strategies 2016–2021:
actions for impact: web annex 1: key data at a glance. World Health Organization, Geneva
NMR-Based Metabolomics to Evaluate Individual Response to Treatments 245

Xi Y, Xu P (2021) Global colorectal cancer burden in 2020 and projections to 2040. Transl Oncol
14:101174. https://doi.org/10.1016/j.tranon.2021.101174
Yafi FA, Jenkins L, Albersen M, Corona G, Isidori AM, Goldfarb S, Maggi M, Nelson CJ, Parish S,
Salonia A, Tan R, Mulhall JP, Hellstrom WJG (2016) Erectile dysfunction. Nat Rev Dis Primers
2:16003. https://doi.org/10.1038/nrdp.2016.3
Ząbek A, Stanimirova I, Deja S, Barg W, Kowal A, Korzeniewska A, Orczyk-Pawiłowicz M,
Baranowski D, Gdaniec Z, Jankowska R, Młynarz P (2015) Fusion of the 1H NMR data of
serum, urine and exhaled breath condensate in order to discriminate chronic obstructive
pulmonary disease and obstructive sleep apnea syndrome. Metabolomics 11:1563–1574.
https://doi.org/10.1007/s11306-015-0808-5
Zhang X, Zhao X-W, Liu D-B, Han C-Z, Du L-L, Jing J-X, Wang Y (2014) Lipid levels in serum
and cancerous tissues of colorectal cancer patients. World J Gastroenterol 20:8646–8652.
https://doi.org/10.3748/wjg.v20.i26.8646
Zhou J, Wang Y, Zhang X (2017) Metabonomics studies on serum and urine of patients with breast
cancer using 1 H-NMR spectroscopy. Oncotarget 5. https://doi.org/10.18632/oncotarget.16210
Zhu J, Djukovic D, Deng L, Gu H, Himmati F, Chiorean EG, Raftery D (2014) Colorectal cancer
detection using targeted serum metabolic profiling. J Proteome Res 13:4120–4130. https://doi.
org/10.1021/pr500494u
Pharmacometabolomics of Asthma
as a Road Map to Precision Medicine

Rachel S. Kelly, Margaret F. Cote, Sofina Begum,

and Jessica Lasky-Su

Contents
1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 248
2 Pharmacometabolomics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 248
3 Applications of Pharmacometabolomics to Asthma . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 249
4 Metabolomic Responses to Asthma Therapeutics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 250
4.1 Steroids (Inhaled and Oral) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 252
4.2 Bronchodilators (Beta-Agonists) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 254
4.3 Leukotriene Inhibitors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 256
4.4 Combined Therapeutics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 257
5 Therapy or Severity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 257
6 Clinical Subgroups and Endotypes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 258
6.1 Clinically Derived Subgroups (The “Bottom-Up” Approach) . . . . . . . . . . . . . . . . . . . . . . . 259
6.2 Omic-Driven Endotypes (The “Top-Down” Approach) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 261
6.3 Aspirin-Exacerbated Asthma . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 263
7 Prevention . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 264
8 Challenges of Pharmacometabolomics in Asthma . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 265
9 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 267
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 268

Abstract
Pharmacometabolomics applies the principles of metabolomics to therapeutics in
order to elucidate the biological mechanisms underlying the variation in
responses to drugs between groups and individuals. Asthma is associated with
broad systemic effects and heterogeneity in treatment response and as such is

R. S. Kelly (✉) · M. F. Cote · S. Begum · J. Lasky-Su

Channing Division of Network Medicine, Brigham and Women’s Hospital and Harvard Medical
School, Boston, MA, USA
e-mail: [email protected]; [email protected]; [email protected];
[email protected]

# The Author(s), under exclusive license to Springer Nature Switzerland AG 2022 247
V. Ghini et al. (eds.), Metabolomics and Its Impact on Health and Diseases,
Handbook of Experimental Pharmacology 277, https://doi.org/10.1007/164_2022_615
248 R. S. Kelly et al.

ideally suited to pharmacometabolomics. In this chapter, we discuss the state of

the emerging field of asthma pharmacometabolomics, with a particular focus on
studies of steroids, bronchodilators, and leukotriene inhibitors. We also consider
those studies concerned with subtyping cases to better understand the pharmacol-
ogy of those groups and those looking to leverage pharmacometabolomics for
asthma prevention. We finish with a discussion of the challenges and
opportunities of asthma pharmacometabolomics and reflect upon where this
field must go next in order to realize its precision medicine potential.

Keywords
Asthma · Endotyping · Pharmacometabolomics · Steroids · Treatment response

1 Introduction

Approximately 334 million people are affected by asthma worldwide, attributing to

nearly 400,000 deaths annually (GBD 2015 Chronic Respiratory Disease
Collaborators 2017). It is predicted that in the next decade, over 100 million new
cases will arise (Enilari and Sinha 2019). In addition to the public health burden
associated with asthma, there is also a severe economic burden (Yaghoubi et al.
2019). It is estimated that by 2040 the direct costs of uncontrolled asthma will rise to
300.6 billion dollars in the USA alone (Yaghoubi et al. 2019). Consequently, the
need for increased asthma control via improved therapeutics is paramount.
The definition of asthma has not changed in over 50 years (Hargreave and Nair
2009). However, studies have demonstrated its complex, heterogeneous nature
(Darveaux and Busse 2015; Moore et al. 2010; Sinha et al. 2017; Tyler and
Bunyavanich 2019). Asthma differs between individuals in its pathology, manifes-
tation, molecular biology, and clinical response (Eder et al. 2006; Tyler and
Bunyavanich 2019). Hence, there is a need for novel targeted therapeutics to replace
the current “one-size-fits-all” approach (Tyler and Bunyavanich 2019). Precision
medicine for asthma could improve management, reduce instances of serious
outcomes, and prevent adverse drug responses through disease stratification
(Moore et al. 2010).

2 Pharmacometabolomics

The metabolome reflects transcriptional and translational processes as well as

environmental interactions (Fiehn 2002; Tyler and Bunyavanich 2019).
Metabolomics utilizes spectroscopic techniques and methods for the comprehensive
profiling of the metabolome in a biological specimen (Clish 2015). This powerful
tool is central in precision medicine, facilitating deconvolution of the complex
metabolic landscapes which underpin disease. Pharmacometabolomics utilizes the
principles of metabolomics and extends these principles to therapeutics in order to
Pharmacometabolomics of Asthma as a Road Map to Precision Medicine 249

elucidate the underlying biological mechanisms behind complex diseases and pro-
vide insights into how individuals respond to the drugs designed for these conditions
(Kaddurah-Daouk and Weinshilboum 2014). Through the implementation of
pharmacometabolomic techniques and methods, we can improve understanding of
why different individuals may respond differently to the same treatment (Kaddurah-
Daouk and Weinshilboum 2014).
Asthma is associated with broad systemic effects including inflammation, oxida-
tive stress, and tissue remodeling (Sahiner et al. 2011), which arise as a consequence
of the complex interplay between genetic and environmental factors. Thus, it is ideal
for the application of a systems approach, such as metabolomics and subsequently
pharmacometabolomics (Zhu et al. 2019). Pharmacometabolomics offers the poten-
tial to improve our understanding of the actions of drugs on individuals across the
spectrum of asthma and facilitate the development of novel targeted therapies with
improved efficacy for all. Pharmacometabolomics offers a compelling and novel
route to precision medicine for individuals with asthma.

3 Applications of Pharmacometabolomics to Asthma

Over the last two decades, there has been growing interest in the role of the
metabolome in asthma. A large number of studies have demonstrated that the
metabolome, as measured in a variety of biosamples including blood, urine, and
exhaled breath consendate (EBC) can be leveraged to predict, diagnose, and assess
the severity of asthma (Wang et al. 2021; Kelly et al. 2017; Papamichael et al. 2021).
Importantly, several of the metabolomic perturbations reported as being associated
with asthma and asthma phenotypes including pathways relating to hypoxia
response, oxidative stress, immunity, inflammation, lipid metabolism, and the tricar-
boxylic acid cycle have been independently validated (Chung et al. 2014). More
recently, a body of asthma metabolomics researchers have shifted their focus to
study how an individual’s metabolome can influence their response to therapeutic
strategies. There is a growing interest in the utility of metabolomics for the develop-
ment of more efficacious asthma treatments (McGeachie et al. 2015; Svenningsen
and Nair 2017; Kelly et al. 2017, 2019b; Kuruvilla et al. 2019). Although only a
small number of studies of asthma have specifically labelled themselves as
“pharmacometabolomic” studies, many more have focused on the integration of
asthma phenotypes, metabolomics, and therapeutics.
Pharmacometabolomic asthma studies tend to take one of the two broad
approaches (Fig. 1): (1) the assessment of the metabolomic responses of individuals
with asthma to various therapeutics; or (2) the subgrouping of individuals with
asthma. These subgrouping or disease stratification studies can be further subdivided
into those that take a “bottom-up” and those that take a “top-down” approach. The
“bottom-up” approach clusters asthmatics based on their clinical characteristics as
they relate to therapy and/or management and then investigates the metabolomic
profiles of these subgroups (Fig. 1 [2a]). The “top-down” studies stratify disease
based on their metabolome, clustering individuals with asthma into subgroups based
250 R. S. Kelly et al.

Fig. 1 Asthma pharmacometabolomics study design approaches

on their metabolic profile and investigating the therapeutic needs of these subgroups
(Fig. 1 [2b]). In both instances, the end goal is to understand the mechanisms and
biology underlying disease in the different subgroups in order to identify novel
treatments targeted toward those underlying mechanisms.
In this chapter, we explore current perspectives and uses of
pharmacometabolomics in the study of asthma, describe the findings of the studies
to date, and reflect upon their potential roles in optimizing the treatment of
individuals with asthma as a route toward precision medicine.

4 Metabolomic Responses to Asthma Therapeutics

Multiple therapeutics are currently utilized in the management of asthma and its
symptoms. These include inhaled corticosteroids (ICSs), oral corticosteroids, long-
acting β2-agonists (LABAs), short-acting β2-agonists (SABAs), leukotriene
inhibitors (blockers of the CysLT receptor, specifically), long-acting muscarinic
antagonists (LAMAs), short-acting muscarinic antagonists (SAMAs), and anti-IL
and anti-IgE antibody therapies, theophylline and cromolyn (Wendell et al. 2020)
(Table 1). These treatments, alone or in combination, comprise both long-term and
quick-relief medications that aid in reducing symptoms among asthmatic patients by
targeting different biological pathways and processes. However, their efficacy varies
Pharmacometabolomics of Asthma as a Road Map to Precision Medicine 251

Table 1 Common asthma therapeutics

Example
Type Function medications
Steroids Prohibits gene transcription of inflammatory genes Beclomethasone
(inhaled) Dipropionate
Budesonide
Ciclesonide
Flunisolide
Fluticasone
Furoate
Mometasone
furoate
Fluticasone
Propionate
Steroids (Oral) Prohibits gene transcription of inflammatory genes Prednisolone
Prednisone
Methylprednisolone
Leukotriene Prevents the formation of leukotrienes and blocks Zileuton
inhibitors leukotriene receptors Montelukast
Zafirlukast
Pranlukast
Anti-IL Binds to the IL-5 receptor to reduce eosinophil Anti-IL5:
treatment: IL5 survival Mepolizumab
Reslizumab
Benralizumab
Anti-IL Binds to the IL-4 receptors to reduce airway Anti-IL4:
treatment: IL-4 hyperresponsiveness Dupilumab
Bronchodilators: Activates the beta2-adrenoreceptors present on the SABAS:
SABAS surfaces of airway smooth muscle cells for muscle Epinephrine
relaxation Isoproterenol
Salbutamol
Pirbuterol
Terbutaline
Bronchodilators: Activates the beta2-adrenoreceptors present on the LABAS:
LABAS surfaces of airway smooth muscle cells for muscle Salmeterol
relaxation Formoterol
BI-167107
Vilanterol
Indacaterol
Olodaterol

across populations (Larsson et al. 2020), and many asthmatics do not respond to the
“one-size-fits-all” management approach. Patients who do not respond to standard
care are responsible for most of the asthma-related economic burden (Chung et al.
2014). Researchers are now beginning to leverage pharmacometabolomics to try and
better understand the mechanisms of action of these therapeutics and the subgroups
that may benefit from these treatments.
252 R. S. Kelly et al.

4.1 Steroids (Inhaled and Oral)

Steroids reduce the airway inflammation associated with asthma and are the most
commonly prescribed drugs for this condition (Wendell et al. 2020). However, their
efficacy varies between individuals; approximately 25–35% of asthma patients
either do not respond or respond poorly to inhaled corticosteroids (ICS) (Kachroo
et al. 2021). As such, they are often prescribed in combination with other therapeu-
tics in order to optimize treatment. Identifying non- or poor-responders could help to
improve treatment efficacy and to reduce the potential side effects associated with
overtreatment (Kachroo et al. 2021), which can include weight gain, high blood
pressure, muscle weakness, osteoporosis, and an increased risk of infection, in
addition to the increasingly understood consequence of adrenal suppression
(Grennan and Wang 2019).
There are two different categories of steroids that work to reduce asthma
symptoms: long-term control and quick-relief medications (Wendell et al. 2020).
Long-term control medications are typically ICS such as beclomethasone
dipropionate, budesonide, ciclesonide, flunisolide, fluticasone furoate, mometasone
furoate, and fluticasone propionate. Quick-relief medications include both oral and
intravenous corticosteroids such as prednisolone, prednisone, and methylpredniso-
lone (Wendell et al. 2020). The main action of corticosteroids is to prohibit tran-
scription of inflammatory genes (Barnes 2006). They also induce transcription of
anti-inflammatory and immune regulation genes including beta2-adrenergic
receptors, secretory leukocyte inhibitory protein, and mitogen-activated protein
kinase phosphatase-1 through the glucocorticoid receptor-mediated transactivation,
further contributing to their anti-inflammatory properties (Newton and Giembycz
2016; Matera et al. 2020; Wendell et al. 2020).
Among the treatment options for asthma, steroids are by far the most commonly
studied in a pharmacometabolomic framework. Within this category, studies of ICS
dominate the research. Yet despite the relatively large body of literature, there is
some disagreement regarding the influences of steroids on the metabolome.
Ferraro et al. performed breathomics (metabolomic profiling of the EBC) in
26 children with asthma before and after a 10-week course of inhaled
beclomethasone dipropionate, a common corticosteroid, and in 16 children without
asthma (Ferraro et al. 2020). Although the children with asthma demonstrated
improvements in lung function and asthma control following treatment, there was
no change in the EBC metabolome, and no differences in the urinary endogenous
steroid profile. They also demonstrated that the EBC metabolome could distinguish
between children with and without asthma and that steroid treatment did not affect
this ability, with notable differences in prostaglandin, fatty acid, and
glycerophospholipid metabolites between children with and without asthma. Inter-
estingly, they identified more metabolomic differences between the asthmatic and
non-asthmatic children before the steroid treatment. This could be interpreted to
suggest that rather than causing further dysregulation to the metabolome the treat-
ment actually returned it to a state more akin to a non-asthmatic control. (Ferraro
et al. 2020). Another breathomics study investigated ethane, a product of lipid
Pharmacometabolomics of Asthma as a Road Map to Precision Medicine 253

peroxidation that occurs in response to oxidative stress, and it was further

demonstrated that ethane levels in the non-treated individuals were correlated with
exhaled nitric oxide, a marker of oxidative stress and inflammation (Paredi et al.
2000). In contrast to Ferraro et al.’s findings, which found no differences in the
metabolome associated with treatment, this study of ethane in exhaled air (Paredi
et al. 2000) determined that levels of ethane were higher in 12 patients not receiving
steroids as compared to 14 steroid-treated patients and 12 non-smoking controls.
This again points to the effectiveness of steroids in stabilizing treated subjects.
Oxidative stress was also a mechanism of interest in one (Loureiro et al. 2014) of
two studies that considered the relationship between metabolites and exacerbations
among individuals, with the hypothesis that exacerbation-associated metabolites
may act as markers of treatment response(Kachroo et al. 2021; Loureiro et al.
2014). Loureiro et al. conducted a small study of 10 individuals with asthma on
inhaled corticosteroids and long-acting b-agonists who had urinary metabolomics
performed during exacerbations(Loureiro et al. 2014). They determined levels of
threonine (and/or lactate), alanine, carnitine, acetylcarnitine, and trimethylamine-N-
oxide, aldehydes and alkanes were increased during the exacerbated state compared
to the stable condition, while acetate, citrate, malonate, hippurate, dimethylglycine,
and phenylacetylglutamine were decreased (Loureiro et al. 2014). As such, these
metabolites, several of which are involved in pathways of oxidative stress, can be
considered as markers of steroid non-responsiveness. In a larger study of the plasma
metabolome including 170 adults with asthma, Kachroo et al. identified a set of
steroid-response associated metabolites. They reported eight plasma metabolites that
were associated with episodes of exacerbation while on ICS after correction for
multiple testing (Kachroo et al. 2021). Two of the top metabolites, hexadecanedioate
and tetradecanedioate are involved in omega fatty acid oxidation, which is known to
play a key role in immune function, airway remodeling, but which has not previously
been linked to ICS response. They further identified associations with amino acids
and urea cycle metabolites and reported inverse associations between cortisol and
cortisone and exacerbations. The authors postulate that, as these metabolites are
known markers of treatment adherence, this suggests individuals with asthma who
consistently used their ICS medication also suffered less frequent exacerbations.
Finally, the authors reported significant sex-metabolite and age-metabolite
interactions for exacerbations. These findings raise a crucial point; the relationship
between treatment and metabolite may not be consistent between the sexes or over in
individuals of differing age. This underlies the importance of considering such
variables when exploring the influence of any treatment on the metabolome.
One of the most intriguing aspects of the asthma pharmacometabolomic literature
is the evidence regarding the long-term effects of ICS on the metabolome. In fact, the
enduring effects of steroid treatment on the metabolome and therefore on overall
health are a concern that goes beyond asthma to include any condition treated with
steroids (Pandya et al. 2014). Exogenous corticosteroids have been shown to
precipitate negative feedback via their action on the hypothalamus and anterior
pituitary gland to reduce the release of cortisol from the adrenal cortex leading to
shrinkage of the adrenals. The adrenals can therefore no longer produce the
254 R. S. Kelly et al.

necessary endogenous corticosteroids which can result in adrenal insufficiency and

adrenal suppression (Pandya et al. 2014), conditions associated with multiple
morbidities and increased mortality (Gurnell et al. 2021).
Within the asthma pharmacometabolomics literature, a recent analysis of over
14,000 individuals from four independent cohorts identified and validated 35 plasma
metabolites significantly associated with asthma status. Of these, 34 metabolites
were annotated to canonical curated pathways for corticosteroids, pregnenolone, and
androgenic steroids and all 34 were detected at lower levels among asthmatics
(Kachroo et al. 2022). Strikingly, these reductions were more pronounced for
asthmatics on ICS as compared to asthmatics not taking ICS. Further exploration
of these cohorts determined that, in agreement with the literature, long-term ICS use
among asthmatics was associated with significant long-term reduction in cortisone
and cortisol. This study is among the first to report that these reductions are evident
at even the lowest dosages of ICS. Importantly, when tracking cortisol levels over
the course of a 24-h period it was found that the decrease in cortisol levels was
greatest in the early morning, when subjects were most susceptible to an asthma
attack. Furthermore, 31% of asthmatics using ICS had cortisol levels low enough to
meet the clinical threshold for adrenal suppression (Kachroo et al. 2022). These
results confirm the long-term clinical implications of steroid treatment in asthmatics
using a pharmacometabolomic framework, while highlighting the potential of
pharmacometabolomics to generate and test novel hypotheses.
Evidence that treatment with steroids acts to alter the metabolome in the long term
is further supported by studies in both children (Kannisto et al. 2001) and adults
(Kannisto et al. 2004) with asthma who were found to have significantly reduced
levels of serum dehydroepiandrosterone sulfate following treatment with inhaled
glucocorticosteroids. As dehydroepiandrosterone sulfate is the most abundant andro-
gen secreted by the adrenals, this lends further credence to the hypothesis that long-
term steroid use may lead to adrenal suppression.

4.2 Bronchodilators (Beta-Agonists)

Bronchodilators, also known as beta-agonists, work to activate the β2-adrenoceptors

(β2-ARs) that are present on the surfaces of airway smooth muscle cells, epithelium,
and vascular cells, as well as submucosal glands (Ruffin et al. 1982; Carstairs et al.
1985). β2-ARs are G-protein coupled receptors that are activated by catecholamines,
which transmit downstream signaling (typically resulting in calcium channel inter-
action) to mediate the relaxation of smooth muscle. β2-ARs also exist on the surface
of inflammatory and immune cells including mast cells, macrophages, neutrophils,
lymphocytes, eosinophils, epithelial and endothelial cells, and type I and type II
alveolar cells (Cazzola et al. 2012). The purpose of bronchodilators is to utilize the
β2-adrenoceptors’ pathway to eventually open K+ channels that lead to the relaxa-
tion of airway smooth muscle (Jones et al. 1999), although the effects on inflamma-
tion and the inflammatory pathway have not yet been fully understood (Matera et al.
2020). For this reason, bronchodilators and ICS are commonly prescribed together to
Pharmacometabolomics of Asthma as a Road Map to Precision Medicine 255

ensure that both inflammation and smooth muscle relaxation are achieved (Wendell
et al. 2020). The most common types of bronchodilators include short-acting-beta-
agonists (SABAs) which are used when an individual needs an immediate muscle
relaxation effect during an asthma exacerbation, while long-acting-beta-agonists are
taken daily to maintain muscle relaxation over long periods of time (Wendell et al.
2020). Commonly prescribed SABAs include albuterol, epinephrine, isoproterenol,
metaproterenol, salbutamol, pirbuterol, and terbutaline. The most commonly utilized
LABAs are salmeterol, formoterol, BI-167107, vilanterol, indacaterol, and
olodaterol (Wendell et al. 2020).
To date, only a small number of studies have considered beta-agonists and asthma
in a metabolomics framework, with most investigating treatment related
metabolomic changes. Urinary levels of bromotyrosine, a marker reactive
brominating oxidants formed by eosinophil-catalyzed oxidation, have been shown
to be associated with the use of beta-agonists (Wedes et al. 2011). Similarly
increased serum levels of lactate have been significantly associated with albuterol
use in adult asthmatics into two independent studies (Lewis et al. 2014; Rodrigo and
Rodrigo 2005), which is thought to be a direct result of therapy (Rodrigo 2014).
McGeachie et al. also focused on SABAs, investigating differences in the levels
of serum metabolites between eight individuals with asthma who reported SABA
use in the last 7 days compared to 12 individuals with asthma who did not use
SABAs in the same time period (McGeachie et al. 2015). While no metabolites
survived stringent correction for multiple testing, they did observe a nominally
significant decrease in a monohydroxy derivative of arachidonic acid in those
individuals not using SABAs. Together with linoleic acid metabolism and
sphingolipid metabolism, arachidonic acid metabolism ranked as the top
dysregulated pathway with respect to SABA use as a proxy for asthma control
(McGeachie et al. 2015). They then integrated the metabolomic data with genome-
wide genotype, gene expression data and methylation data using a Conditional
Gaussian Bayesian Network (CGBN) to identify the strongest predictors of asthma
control across these omic types. Integrative over representation analyses of the top
multi-omic predictors supported the importance of these pathways in SABA
use/asthma control (McGeachie et al. 2015).
Taking a slightly different approach two studies considered the influence of
age-related changes on the beta-agonist – metabolite relationship. Kelly et al.
focused on the metric of bronchodilator response (BDR). BDR measures spirometry
before and after beta-agonists/bronchodilator treatment to establish reversibility of
airflow. As BDR is known to decline with age, this study aimed to determine
whether the BDR-age relationship was mediated by underlying metabolite levels.
They were able to identify and validate in two independent populations blood
metabolites that interacted with age in BDR. Increased levels of
2-hydroxyglutarate in the blood of an individual with asthma may exacerbate the
association between age and BDR decline, while increased levels of cholesterol
esters, GABA, and ribothymidine may attenuate the age-associated BDR decline
(Kelly et al. 2019b). These metabolites have been shown to play various roles in lung
function, inflammation, and immune function, but have not previously been directly
256 R. S. Kelly et al.

related to treatment (Kelly et al. 2019b). A second study (Sordillo et al. 2020) also
investigated whether circulating metabolites mediate age-related changes in bron-
chodilator response (BDR) for individuals with asthma. Sordillo et al. determined
that a portion of the effect of age on BDR acted indirectly through phosphatidylcho-
line plasmalogens. It is hypothesized that plasmalogens which protect against
oxidative stress may mediate lung function responses through their ability to alter
the structural properties of lung surfactants, and they are known to be enriched in the
smooth muscle that bronchodilator acts upon (Bozelli et al. 2021). As such, the
authors concluded that plasmalogens may serve as potential pharmacologic targets
for enhancement of lung function in individuals with asthma (Sordillo et al. 2020).

4.3 Leukotriene Inhibitors

Cysteinyl leukotrienes (CysLTs) are a class of bioactive fatty acids associated with
airway inflammation (Wendell et al. 2020). These fatty acids are released by
eosinophils during an asthma exacerbation as proinflammatory mediators, which
contribute to the bronchoconstriction seen in asthma patients (Wendell et al. 2020).
The receptors for these leukotrienes exist on the cell surface of monocytes,
eosinophils, lung macrophages, and resident mast cells, and the interaction of the
CysLTs with their receptors has been found to induce anaphylaxis and cause the
prolonged contraction of bronchial smooth muscle, that is characteristic of asthma
(Augstein et al. 1973; Yokomizo et al. 2018). Leukotrienes further contribute to the
pathogenesis of asthma and airway obstruction by recruiting eosinophils and
neutrophils in the mucosa of the airway (Bisgaard 2001; Foster and Chan 1991;
Smith et al. 1993; Henderson et al. 1996).
There are two critical stages along the pathway at which the formation of
leukotrienes and the subsequent inflammatory response in the airways can be
prevented. The first is at the 5-LOX enzymatic step where arachidonic acid is
converted into LTA4 by 5-lipoxygenase, where Zileuton blocks this pathway
(Walter et al. 1987; Peters-Golden and Henderson 2007). The second is at the
G-protein coupled receptor (GPCR) CysLTR1 location, where LTC4 and LTD4
are the substrates for downstream signaling events (Wendell et al. 2020). Drugs
including Montelukast, Zafirlukast, and Pranlukast act as leukotriene receptor
antagonists (LTRAs) by targeting the CysLTR1 to obstruct the signaling pathway
(Peters-Golden and Henderson 2007; Wendell et al. 2020). Treatment with these
drugs therefore results in a decrease in the level of leukotrienes present in the body,
leading to fewer and less severe asthma exacerbations. Montelukast specifically was
found to significantly reduce eosinophil infiltration in asthma patients, while
Zileuton reduces the production of leukotrienes (Yokomizo et al. 2018; Bruno
et al. 2018).
The actions of these therapeutics are supported by cell-based studies (Werner
et al. 2019). Profiling of the lipid metabolome of healthy human macrophages
demonstrated that Zileuton reduces the intensity of the immune response by
blocking the pathway for the creation of the enzyme 5-LOX. It also showed a
Pharmacometabolomics of Asthma as a Road Map to Precision Medicine 257

decrease in leukotriene levels as a reaction to FLAP and 5-LOX inhibitors (Werner

et al. 2019). However, to date only a single human study has investigated the effect
of leukotriene inhibitors on the metabolome of asthmatics (Quan-Jun et al. 2017).
This study focused on the serum and urine NMR metabolomic profile of children
with asthma treated with Montelukast and concluded that it imparted no
metabolomic effect (Quan-Jun et al. 2017).

4.4 Combined Therapeutics

In individuals with severe asthma, beta-agonists and ICS are commonly prescribed
in combination. To date very few studies have explored the combined metabolomic
effects of these therapeutics. In addition to the Loureiro et al. article (Loureiro et al.
2014), which was primarily concerned with steroid non-responsiveness in
individuals with asthma on inhaled corticosteroids and long-acting b-agonists,
Quan-Jun et al. investigated the impact on the metabolome of childhood asthmatics
receiving both the glucocorticosteroid budesonide and the beta-2-agonist salbutamol
(Quan-Jun et al. 2017). There was a clear difference in the serum metabolomic
profile of children on the combined therapeutic when compared to children receiving
neither, and this difference was driven by increased levels of 4-hydroxybutyrate,
lactate, cis-aconitate, 5-HIAA, taurine, trans-4-hydroxy-l-proline, tiglylglycine,
3-hydroxybutyrate, 3-methylhistidine, and glucose and decreased levels of alanine,
glycerol, arginine, glycylproline, 2-hydroxy-3-methylvalerate, creatine, citrulline,
glutamate, asparagine, 2-hydroxyvalerate, and citrate. There was no difference in the
metabolome when considering either treatment alone, suggesting it is the
combination of therapies that drives the metabolomic shift (Quan-Jun et al. 2017).
Furthermore, the authors determined there were no metabolomic effects of a further
beta-2-agonist, procaterol. Similar results were observed when considering the urine
metabolome of these children. Taken together these results suggested possible
“metabolic reprogramming” as a result of inhaled budesonide and salbutamol in
asthmatic children, that is particularly pronounced in the arginine and proline
metabolism pathway.

5 Therapy or Severity

Several studies report metabolomic changes that associate with the use of commonly
administered asthma therapeutics. However, when considering these results, the
issue of cause and effect must be considered. Treatment is inherently linked to the
disease itself, whereby more severe cases or cases with a particular phenotype are
more likely to receive therapy (confounding by indication). Considering this, many
pharmacometabolomic studies of the effect of a particular therapeutic on the
metabolome may be confounded by the clinical indication for that therapy. This is
reflected in the fact that a number of metabolites and metabolomic pathways found to
be associated with treatment in the studies previously mentioned have also been
258 R. S. Kelly et al.

linked to both the presence of asthma and its severity. For example, linoleic acid
metabolism and arachidonic acid metabolism have been associated with asthma
through their roles as lipid mediators of inflammation (Wendell et al. 2014).
A number of the asthma pharmacometabolomics studies have addressed the
question of cause versus effect directly. In Paredi et al.’s study of exhaled ethane
in asthma, they determined that ethane was higher in the non-treated cases, but that
even within this group it was highest in those with the greatest degree of lung
function. This suggests oxidative stress is correlated with severity itself and that the
observed association between ethane as a marker of oxidative stress and with steroid
use is due to the fact that treatment is reducing oxidative stress in this group (Paredi
et al. 2000). The fact that Ferraro et al. found few differences between the steroid-
treated EBC metabolome and healthy controls, as compared to the untreated
metabolome versus controls could support this hypothesis, particularly given no
differences were seen within the same individuals before and after treatment (Ferraro
et al. 2020). In further agreement McGeachie et al. remarked that in the confines of
their study “whether these metabolomic changes reflect the drug or the phenotype
cannot be determined” (McGeachie et al. 2015). It is also of interest to note that
lactate was implicated in several different therapeutics, which may suggest that
dysregulation of this metabolite and its pathways relates more to the need for
treatment than to the treatment itself.
However, it should be noted that the long-term effects of ICS use on endogenous
cortisol and cortisone levels and on dehydroepiandrosterone sulfate does suggest
that treatments for asthma can cause metabolomic shifts independent of the disease.
Furthermore, changes in lactate levels have been directly attributed to treatment
(Kelly et al. 2019b). This suggests that whether or not therapy directly effects the
metabolome is metabolite specific and therapy specific, it may occur in some
instances but not others. Given findings from several studies that have considered
interactions, these relationships are also likely to be impacted by external factors
such as age and sex.

6 Clinical Subgroups and Endotypes

Given the potential challenges that accompany consideration of the influence of a

treatment on the metabolome of a potentially heterogeneous group, increasingly,
pharmacometabolomic studies of asthma have been leveraging the heterogeneous
nature of this condition to explore clinical subtypes or derive endotypes using
metabolomics. In this way, studies aim to explain why the metabolomic responses
to therapeutics may differ between individuals and ultimately to identify the optimal
therapeutics in each case. Although to date, there have been no instances where
pharmacometabolomic findings in asthma have been translated into the clinic,
important insights into the mechanisms and biology of treatment responses have
already been reported.
Pharmacometabolomics of Asthma as a Road Map to Precision Medicine 259

6.1 Clinically Derived Subgroups (The “Bottom-Up” Approach)

The “bottom-up” approach argues that therapeutic approaches should be targeted

toward groups of individuals with the same clinical characteristics and phenotypes
(Wenzel 2012). Various asthma subtypes have been proposed based on key clinical
features including symptoms, exacerbations, lung function, treatment response, and
disease severity. Others have attempted to derive subtypes using laboratory
parameters, such as peripheral blood or sputum immune cell count (most typically
eosinophil and neutrophil counts), or percentage and fractional exhaled nitric oxide
(FENO) (Wenzel 2012; Tyler and Bunyavanich 2019), characteristics which inform
on underlying inflammation.
It is well known that differences in treatment related features between individuals,
including therapeutic regimen, exacerbation frequency while on controller medica-
tion, and response to medication can correspond to differences in various omic
profiles (Hastie et al. 2010; Loza et al. 2016; Bigler et al. 2017; Svenningsen and
Nair 2017). Accordingly, several research efforts have been investigating the
metabolomic profiles of the clinically relevant subgroups derived in their
populations. Among these, only one reported no difference in the EBC metabolome
as measured using NMR spectroscopy between their subgroups. Carro et al. could
not distinguish children classified as non-severe asthma regularly treated with
controller medications from those non-severe cases who were well controlled and
steroid naïve based on their EBC NMR profile (Carraro et al. 2013).
In a study of 22 healthy subjects and 54 asthmatics between 18 and 70 years old,
Reinke et al. classified patients into several subgroups based on their clinical
characteristics and therapeutic regimens; healthy controls (n = 22), mild asthmatics
with intermittent symptoms and only treated with β2-agonists alone (n = 12);
moderate asthmatics with frequent symptoms and treated with ICS (n = 20); and
severe asthmatics with persistent symptoms treated with ICS therapy (n = 22) or
with ICS and oral corticosteroid (OCS) therapy (n = 5) (Reinke et al. 2017). Serum
metabolomic profiling using targeted mass spectrometry revealed metabolomic
differences between the asthma cases versus controls, but also differences between
the asthma subgroups. Six metabolites were significantly different in between the
healthy controls and the individuals in the three asthma subgroups with ICS or
OCS use: DHEA-S, cortisone, ProHyp, pipecolate, N-palmitoyltaurine, and cortisol.
Differences in these six metabolites were observed with even moderate ICS dosage,
but with the exception of pipecolate, were greatest between the healthy controls and
those who were on both ICS and OCS. This is in agreement with the findings from
the steroids literature which found that cortisol and cortisone levels are lowest in
asthmatics treated with high doses of exogenous steroids (Kachroo et al. 2022). To
address the issue of potential confounding in their results, Reinke et al. further
interrogated the data. While they did identify some metabolite shifts that were
intrinsic to the disease process itself, for example Oleoylethanolamide increased
with asthma severity independently of steroid treatment, they concluded that others
were specific to treatment. Intriguingly, they reported that overall the association
260 R. S. Kelly et al.

between metabolic profile and ICS treatment is greater than that between metabolic
profile and disease (Reinke et al. 2017).
In their study of the discriminatory ability of the urine metabolome for asthma
phenotypes, Mattarucchi et al. addressed the question of whether the observed
metabolomic changes were due to the treatment or due to the disease. They built
two models, one which sought to distinguish between children with asthma who
took controller medication and those who did not, and one which sought to distin-
guish cases from healthy controls. They determined that their controller medication
model was highly effective, with only a single misclassification in a population of
42 (Mattarucchi et al. 2012). In keeping with the findings of Reinke et al., they found
the discriminatory model was primarily driven by intermediates in the metabolism of
C21-steroid hormones urocortisone and urocortisol. Importantly, this model was
distinct from that that distinguished all 42 asthma cases from 12 age-matched
controls demonstrating that the metabolomic differences between asthma cases and
controls were not driven by consumption of controller medication (Mattarucchi et al.
2012).
Comhair et al. classified severe and non-severe asthmatics based on criteria
outlined in the proceedings of the American Thoracic Society Workshop on Refrac-
tory Asthma, which were largely based on treatment regimen (Comhair et al. 2015).
In their study all 10 severe adult asthmatics received a high dose of inhaled or oral
corticosteroids either singly or in combination with long-acting beta-agonists, while
the 10 non-severe asthmatics did not receive any corticosteroids or long-acting beta-
agonists and received inhaled beta-agonist (rescue medication) infrequently but less
than 2 times per week. A comparison of the plasma metabolomic profile of the severe
and non-severe asthma cases identified 18 significant metabolites, all but one of
which were lower in the severe asthmatics. Most of these metabolites were again
related to steroid metabolism including 1-steraroyylglycerol,
dehydroisoandrosterone sulfate, epi-androsterone sulfate, and androsterone sulfate.
These findings are consistent with the suppression of adrenal steroids as a conse-
quence of therapeutic use of corticosteroids as mentioned previously. They also
observed differences in several amino acids, including beta-alanine, between the two
groups of asthmatics, but further observed that the severe group shared many of the
same metabolomic signatures as non-severe asthmatics, who were not on
corticosteroids. The authors interpreted these findings as further evidence that
corticosteroid therapy is not the driver of differences between asthma cases and
healthy controls (Comhair et al. 2015). This hypothesis was further supported by
Dallinga et al. also who showed that the discrimination between asthma cases and
controls based on an analysis of volatile organic compound in breath samples was
not driven by medication usage (Dallinga et al. 2010). They predicated this conclu-
sion on the observation that the metabolomic products of known asthma medications
were not reflected in the principal components that distinguished between asthma
cases and controls.
Park et al. took a somewhat different approach, focusing on the metabolomic
differences between those who respond to corticosteroids as compared to those who
do not (Park et al. 2017). They performed high resolution mass spectrometry based
Pharmacometabolomics of Asthma as a Road Map to Precision Medicine 261

urinary metabolomic profiling of children with severe asthma; 15 children who did
not responded to corticosteroids, defined by persistently poor asthma control with
symptoms more than twice weekly and a less than 15% improvement in FEV1
2 weeks after systemic triamcinolone administration, and 15 children who did
respond. They identified 30 metabolomic features that differed significantly between
the two groups, representing differences in the metabolism of tyrosine, glutathione,
and the degradation of aromatic compounds. Of these five, 3,6-dihydronicotinic acid
3-methoxy-4-hydroxyphenyl(ethylene)glycol, 3,4-dihydroxy-phenylalanine,
γ-glutamylcysteine, Cys-Gly, and Flavin mononucleotide were determined to con-
tribute most to the distinction of corticosteroid responders and non-responders (Park
et al. 2017). These results point to the reduced synthesis and increased degradation
of the antioxidant glutathione, which in its thiol-reduced form (GSH) is the most
abundant antioxidant in the airway epithelial lining fluid and plays an important role
in the pathogenesis of asthma (Papamichael et al. 2021). The significant metabolites
also included constituents of cigarette smoke, which were higher in the urine of
corticosteroid-resistant children. This is consistent with evidence that smoking can
impair the efficacy of corticosteroid treatment in asthma (Chaudhuri et al. 2003).
Fitzpatrick et al. compared the plasma metabolome of mild to moderate childhood
asthmatics treated with ICS or ICS/ long-acting beta-agonists (LABA) combination
therapy to that of severe asthmatics treated with high-dose ICS and LABA
(Fitzpatrick et al. 2014). Interestingly, among the severe asthmatics they determined
no children displayed complete corticosteroid responsiveness (defined as symptoms,
lung function, bronchodilator reversibility, and exhaled nitric oxide values
normalized 2 weeks after treatment with systemic triamcinolone acetonide).
Among the severe asthmatics 89% demonstrated partial responsiveness, while the
remainder had no discernable response, taken together these children were consid-
ered a corticosteroid refractory population. Their results indicated that the cortico-
steroid refractory children demonstrated a vast array of metabolic derangements
relative to mild/moderate asthmatics. These differences again pointed to increased
lipid peroxidation and dysregulation in thiol redox balance and oxidative stress
related pathways, with dysregulation of the glycine, serine, and threonine metabo-
lism pathway and the N-acylethanolamine and N-acyltransferase pathways
observed. The results may support the hypothesis that oxidative stress is a contribu-
tory factor to corticosteroid refractory severe asthma in children. However, like
many of the previous studies the authors determined that confounding by indication
could not be ruled out, as individuals taking asthma medication for severe diseases
are also more likely to have co-morbid conditions including obesity sinus disease
and obstructive sleep apnea (Fitzpatrick et al. 2014).

6.2 Omic-Driven Endotypes (The “Top-Down” Approach)

The alternative to the bottom-up clinical subgrouping approach instead classifies

patients based on the underlying mechanisms of their disease. This approach is
predicated on the hypothesis that treating a disease based on its underlying biological
262 R. S. Kelly et al.

mechanisms may be more effective than treating it on its clinical manifestations

(Svenningsen and Nair 2017; Tyler and Bunyavanich 2019). The inherently mecha-
nistic nature of the metabolome makes it particularly compelling for the derivation of
such endotypes (Tyler and Bunyavanich 2019).
Within U-BIOPRED (Unbiased Biomarkers for the PREDiction of respiratory
disease outcomes), a pan-European cohort of severe asthma in adults and children,
exhaled metabolomic fingerprints measured via electronic noses (eNoses) were
available on 78 adults with severe asthma. Severe asthma was classified as having
a prescription for high-dose ICS plus one other asthma control medication, a daily
OCS prescription, two or more severe exacerbations, or a combination of these
(Brinkman et al. 2019). Unsupervised clustering of these metabolite-derived profiles
identified three clusters of asthmatics (n = 26/33/19) that differed in the clinically
important metrics of circulating eosinophil and neutrophil percentages. The clusters
also differed in the ratios of patients using oral corticosteroids; with the neutrophilic
inflammation predominant cluster 2 containing the higher percentage of patients
using OCS maintenance therapy. In contrast cluster 3 had the highest percentages of
peripheral blood eosinophilia and the lowest percentage of subjects with chronic
OCS use (Brinkman et al. 2019). These findings suggest that hypothesis-free
clustering of individuals with asthma based on the metabolome can generate clini-
cally meaningful subgroups that differ in their medication usage.
This hypothesis was further expanded upon by Kelly et al. in a larger study
including over 2,000 children with asthma from two independent and well-
characterized cohorts (Kelly et al. 2021). This allowed validation of their findings,
a critical step when using metabolomic datasets and unsupervised approaches. In this
study five metabolomic driven endotypes or “metabo-endotypes” were derived from
unsupervised clustering of the plasma metabolome in a population of children with
mild to moderate asthma. These metabo-endotypes differed in metrics of lung
function and more specifically lung obstruction. There were also differences in
medication uses across the five metabo-endotypes that were independent of lung
function. Exploration of the metabolites contributing most to the formation of the
metabo-endotypes identified multiple metabolites involved in the formation and
homeostatic regulation of pulmonary surfactant, thereby pinpointing an actionable
mechanism that precision therapeutics could be directed toward. Importantly, these
metabo-endotypes were validated in an independent population of children with
mild-moderate asthma (Kelly et al. 2021).
In contrast; Sinha et al. performed clustering on 61 asthmatics based on their EBC
profile. Although they did identify phenotypic differences across these clusters,
including significant differences in exacerbations and eosinophil count, they did
not determine any differences in the number of patients using corticosteroids
between the three clusters (Sinha et al. 2017).
Overall, the “top-down” metabolomics-driven approach seems promising, how-
ever it should be noted that given the cross-sectional nature of the sample collection
for profiling and measurement of relevant clinical phenotypes, it is still challenging
to determine causation; whether the profiles that are driving the clusters are driven by
their need for a given therapy, or by the therapy itself. Kelly et al. reported that the
Pharmacometabolomics of Asthma as a Road Map to Precision Medicine 263

lung function related differences between their endotypes were robust to additional
adjustment for medication use, suggesting medication is not driving the clusters.
However, further work incorporating longitudinal data is necessary to explore this
further.

6.3 Aspirin-Exacerbated Asthma

An estimated 5–15% of patients with asthma are considered to have aspirin-

exacerbated respiratory disease (AERD), which is defined as the formation of
nasal polyposis, the development of asthma and a respiratory reaction on ingestion
of aspirin or another nonsteroidal anti-inflammatory drug (NSAID) (Haque et al.
2021). To date, only a small number of studies have utilized metabolomics to
explore this subset of asthma cases characterized by their response to a common
pharmacologic agent, with a focus on measured of leukotrienes. Ban et al. utilized
untargeted metabolomic profiling of urine and serum from 45 AERD asthmatics and
44 aspirin-tolerant asthmatics (ATA) on urine and serum (Ban et al. 2017). Six
metabolites were shown to differ between the AERD and the ATA subjects, includ-
ing several members of the arachidonic acid pathway. This reflects the importance of
the imbalance of the eicosanoid cascade in the pathogenesis of AERD. They further
explored these results by targeted measurement of the arachidonic acid pathway and
validation of their findings in an independent population. Their results suggested that
serum baseline levels of Leukotriene E4 (LTE4) and LTE4/Prostaglandin F2alpha
ratios can be useful diagnostic biomarkers for AERD (Ban et al. 2017).
The utility of LTE4, which provides a measure of the cysteinyl leukotrienes
production associated with asthma attacks, has been further demonstrated in other
studies. In a study of urinary levels in 240 cases of AERD and 226 cases of ATA
significantly higher levels of urinary LTE4 were observed in the AERD population.
Asthmatics with high levels of Urinary 3-bromotyrosine (uBrTyr) (> 0.101 ng/mg
Cr), urinary LTE4 levels (> 800 pg/mg Cr), and blood eosinophils (> 300 cells/μl)
were 7 times more likely to have AERD, however they did not observe a statistically
significant difference between AERD and ATA for uBrTyr alone (Comhair et al.
2018).
Similarly, comparing the concentration of uLTE4 in 247 AERD patients and
239 ATA patients, levels were significantly higher in the AERD cases, and the
authors determined that at a concentration of 8,000 pg/mg creatinine,TE4 could
discriminate between the two groups with an area under the curve of 0.70
(95% CI: 0.66–0.74), a sensitivity of 49%, and a specificity of 81%. However, its
discriminatory power was outperformed by a set of typical clinical parameters,
calling into question its potential clinical translatability (Bochenek et al. 2018).
Another study reported that at a uLTE threshold of 241 pg/mg Cr, specificity was
even better at 92% (Divekar et al. 2016), but both these studies lacked validation of
the predictive ability of their models in an independent population.
264 R. S. Kelly et al.

7 Prevention

A growing body of literature is focused on the prevention of asthma. Given the early
life origins of this condition in most asthmatics much of this work is concentrated on
maternal or prenatal interventions (Gern et al. 1999). These have largely been in the
form of nutritional supplementations, such as vitamins D, C, and E (Litonjua et al.
2020), fish oil (Bisgaard et al. 2016), folic acid (Veeranki et al. 2015), pre and
probiotics (Cuello-Garcia et al. 2016), and antioxidants (Gref et al. 2017). However,
the results have been somewhat contradictory in terms of the efficacy of these
interventions in preventing asthma (Gur et al. 2017). Here again,
pharmacometabolomics, which in this instance can be expanded to include the use
of metabolomics for understanding differing responses to the same nutritional
intervention, can be leveraged to try and understand the mechanisms of success, or
otherwise.
To date, there has been little work on preventative pharmacometabolomics of
asthma. However, the existing literature does show that preventative measures
administered to mothers can influence their offspring’s metabolome (Blighe et al.
2017; Rago et al. 2019). For example, increased exposure to vitamin D in utero via
supplementation has been found to be associated with lower concentrations of fatty
acids, in particular linoleate, linolenate (18:3n–3/3n–6), myristate, oleate, palmitate,
palmitoleate (16:1n–7), and stearate (18:0) in infants (Blighe et al. 2017). Similarly,
prenatal supplementation with fish oil has been shown to result in lower levels of the
n-6 LCPUFA pathway-related metabolites and saturated and monounsaturated long-
chain fatty acids-containing compounds, lower levels of metabolites of the trypto-
phan pathway, and higher levels of metabolites in the tyrosine and glutamic acid
pathway in early life (Rago et al. 2019). It has also been demonstrated that vitamin D
and fish oil interventions can reduce the risk of asthma among offspring (Bisgaard
et al. 2016; Wolsk et al. 2017). Given it is known that many of these altered
metabolites play a direct role in the biological pathways and processes that underlie
the development and progression of asthma, such as oxidative stress and inflamma-
tion, this leads to the hypothesis that an altered offspring metabolomic milieu
resulting from a prenatal intervention may be mediating the association between
that intervention and asthma risk. Indeed, a prenatal fish oil supplementation related
metabolic profile at age 6 months was significantly associated with a reduced risk of
asthma by age 5 and this profile could explain 24% of the observed asthma-
protective effect of this supplementation (Rago et al. 2019). Similarly, the
metabolomic changes in children accompanying vitamin D supplementation of
their mothers have been linked to the reduced risk of asthma (Blighe et al. 2017).
However, these relationships can be complicated by the genetic profile of both the
mother and child. Sphingolipids, which among their multitude of roles act as
signaling molecules involved in immune response, inflammation, have been
suggested to mediate the relationship between maternal vitamin D supplementation
and child asthma risk (Kelly et al. 2019a). Vitamin D metabolites have been shown
to be capable of activating the sphingolipid pathway and decreased synthesis of
sphingolipids has been associated with increased airway hyperresponsiveness and
Pharmacometabolomics of Asthma as a Road Map to Precision Medicine 265

inflammation (Kelly et al. 2019a). Work within the VDAART (Vitamin D Antenatal
Asthma Reduction Trial) prenatal supplementation trial seemed to confirm this
relationship with one important caveat. There was no increase in child sphingolipid
levels and therefore no protective effect for asthma among children with a key
variant in a functional SNP in the region of ORMLD3, which regulates a
rate-limiting step in the de novo synthesis of sphingolipids (Kelly et al. 2019a).
Correspondingly, vitamin D supplementation conferred no protective effect for the
children of mothers with this same variant (Knihtilä et al. 2021).
Further studies are required to explore the role that pharmacometabolomics may
be able to play in the prevention of asthma.

8 Challenges of Pharmacometabolomics in Asthma

The utility of metabolomics to obtain insights into pathophysiology is clear (Fiehn

and Kim 2014; Wishart 2019), and this utility has been demonstrated repeatedly for
asthma (Kelly et al. 2017). The evidence that pharmacometabolomics can help to
identify dysregulated metabolomic markers of disease, leading to more personalized
and efficacious treatment options is growing. However, several important challenges
remain. Many of these challenges pertain broadly to the field of metabolomics.
While a battery of spectroscopic technologies is used in metabolomics, these are
continually undergoing advancements for more accurate and improved resolution.
Though NMR spectroscopy provides a reduced resolution in comparison with mass
spectroscopy, the technique is highly reproducible and robust, not requiring separa-
tion or derivatization of molecules, factors that can sometimes hinder true resolution
in mass spectroscopy with poor sample preparation or quality (Dona et al. 2014).
Regardless of the technology used, the fact that the complexity and magnitude of the
metabolomic map has yet to be completely characterized means that pertinent
findings or critical information may still be missed (Franklin and Vondriska 2011;
Mulvihill and Nomura 2014). While metabolite databases are continually being
updated, curation and accuracy of identified metabolites remains an issue. A major-
ity of studies use untargeted profiling which by definition does not include chemical
standards and so only measures relative abundances of even those metabolites that
have been characterized (Bowler et al. 2017), making it difficult to compare and
combine studies.
Further issues relate to the fact that, although efforts are ongoing, there remains an
overall lack of standardization in approaches, measurements, and analysis for
metabolomic data (Bowler et al. 2017). One particular point of contention is the
issue of imputing missing data, different studies take different approaches rendering
comparison of the results challenging (Wei et al. 2018). Even more problematic is
the fact that, as with several the studies included here, some do not even report upon
the method used. However, it should be noted that such issues are not unique to
metabolomics studies and are common to most epidemiological and particularly
omic studies. The issue of multiple testing correction is similarly of concern to most
omic studies, but is arguably a particular challenge for metabolomics given the
266 R. S. Kelly et al.

collinear nature of metabolites within coregulated biological pathways (Wei et al.

2018; Peluso et al. 2021). This is again evident among the pharmacometabolomics
of asthma studies; while some applied the most stringent approaches for multiple
testing, namely Bonferroni correction, others did not account for multiple testing at
all, leading to a disconnect in the results presented and a complexity in comparing
them. The highly correlated nature of metabolites within coregulated biological
pathways has the additional effect of complicating biological interpretation and
understanding (Bowler et al. 2017). Similarly, the dynamic nature of the
metabolome and its sensitivity to environmental influences, which are in many
ways one of the greatest strengths of metabolomics, can be considered a weakness
when trying to deconvolute causation versus correlation and deal with confounding.
In addition to these broad metabolomic and epidemiological issues, there are also
challenges that are applicable to asthma pharmacometabolomics, more specifically.
Chiefly, due to the relatively recent development of the field, the body of literature is
somewhat sparse, and several studies are based on very small sample sizes. The
optimal biosample on which to conduct metabolomic profiling for
pharmacometabolomic studies in individuals with asthma remains to be determined
and various biosamples have been employed to date including sputum, plasma,
serum, and exhaled breath condensate. The lung as the organ most proximal to the
disease is of particular interest and can be considered to be uniquely informative in a
metabolomic framework, but it is also uniquely challenging, with particular issues
relating to the invasiveness of sample collection and the likelihood of sample
contamination (Bowler et al. 2017). Similarly EBC, which has been widely used
in the asthma pharmacometabolomics literature to date is limited by the lack of
standardization in collection and the issue of appropriate dilution (Bowler et al.
2017). Consequently, much of the literature is focused on blood and urine; findings
have been promising, but their relevance to lung disease has been at times called into
question (Bowler et al. 2017). This issue is magnified when the tissue type of greatest
relevance to the pharmacological agents and their biological processes may not be
the same as that most relevant for asthma. Thus, the use of different biosamples
across the asthma pharmacometabolomics literature further complicates its overall
assessment.
Of the current pharmacometabolomics literature most studies are cross-sectional
in nature, rendering the determination of a causal pathway between medications and
metabolomic effects challenging. While several studies reported that the observed
relationships between medication and therapy were independent of asthma, asthma
phenotypes and severity, others cautioned that they could not rule out confounding
by indication. Regardless, the majority of reported associations remain to be
validated in independent populations or utilizing a randomized control design.
Additional well-designed, large-scale longitudinal studies are required to address
this further. Crucially, these studies need to recognize and consider the heteroge-
neous nature of asthma.
Pharmacometabolomics of Asthma as a Road Map to Precision Medicine 267

9 Conclusions

In conclusion, the field of pharmacometabolomics is an emerging area of research,

with evidence that offers genuine clinical translational potential. Yet, much work
remains to be done. The literature to date is somewhat disparate in terms of study
design, biosample, characteristics of included asthma cases, therapeutics of interest,
statistical approach, and metabolomic profiling technique.
Nevertheless, some common themes have emerged (Fig. 2). The largest number
of studies focus on ICS use, and in several studies spanning different biosamples and
metabolomic profiling techniques, it has been reported that adrenal suppression is a
potential side effect of long-term ICS use. This is particularly important given recent
changes to the GINA guidelines, which propose a shift from SABAs to ICS, given
the risks of SABA-only treatment and SABA overuse (Reddel et al. 2021). Interest-
ingly, the potential harms of SABA do not seem to be a focus of the current asthma
pharmacometabolomics literature. However, as these new guidelines were only
published at the end of 2021, we may see an increase in SABA related
pharmacometabolomics literature moving forward.
There was also a consistent theme of oxidative stress, although there was some
disagreement as to whether this was a consequence of the treatment or was instead
related to phenotype. Further work is needed to clarify. There was similarly a lack of
consensus regarding lactate. The discriminatory ability of leukotrienes for aspirin-
exacerbated asthma was more definitive, although it should be noted that these were
not metabolome-wide studies, rather they focused only on one or a small number of
compounds.
The literature does suggest a growing interest in endotyping in asthma, which
leverages the heterogeneity of this condition to derive subgroups based on mecha-
nism. The small number of studies to date provide some compelling results and
support the notion that endotyping may represent a particularly rich resource for the
identification of therapeutic markers targeted to the mechanisms of a specific
subgroup of disease (Hunt et al. 2002; Sinha et al. 2012, 2017; Brinkman et al.
2019). Endotypes derived via the metabolome are uniquely positioned for clinical

Fig. 2 Summary of findings

268 R. S. Kelly et al.

translational potential through both the assignment of individuals to subgroups that

can receive treatment targeted to the mechanisms of their disease and through the
identification of the therapeutic targets within those subgroups underlying that
treatment (Kelly et al. 2021). However, it should be cautioned that the endotypes
reported here are a long way from being used in clinical practice, and to date there
are no examples of endotypes derived via any omic technology being used to
determine treatment. Further investigation, including diverse validation populations,
assessment of stability over time, and targeted quantification of metabolites that may
have biomarker or therapeutic target potential, is vital (Kelly et al. 2021).
Pharmacometabolomics can be considered the natural successor of
pharmacogenetics, which has several demonstrated successes in asthma, although
validation and translation into clinical practice still remains somewhat of a bottle-
neck (Kersten and Koppelman 2017). It could therefore be hypothesized that
combining these two technologies and other omics, such as proteomics, epigenetics,
in a pharmaco-multi-omics approach could offer improved understanding of
mechanisms in a systems biology framework. However, to date no such studies
exist. This represents an area of untapped potential that is ripe for exploration.
A further development that could help propel the field of asthma
pharmacometabolomics forward is advancements in the handling and analysis of
electronic medical record data which can provide detailed information on medical
and treatment history on huge numbers of well-phenotyped individuals. Similarly,
exploiting the potential of pre-existing biobanks could support studies of individuals
with asthma in far greater numbers than seen in the literature to date. This could also
help to provide more diverse populations allowing further exploration of the
influences of factors such as age sex and smoking status, that have been flagged in
the literature as influencing the relationship between therapy and the metabolome.
Pharmacometabolomics is a nascent field, requiring further development and
exploration. Although much can be taken from the parallel ongoing advancements
in the field of metabolomics, there are many other issues that pertain specifically to
the investigation of medications and therapies, which need to be addressed by the
pharmacometabolomics community. Only then can we consider precision medicine
initiatives underlined by pharmacometabolomics, as a realistic approach to combat
the substantial public health burden of asthma.

References
Augstein J, Farmer JB, Lee TB et al (1973) Selective inhibitor of slow reacting substance of
anaphylaxis. Nat New Biol 245:215–217. https://doi.org/10.1038/newbio245215a0
Ban G-Y, Cho K, Kim S-H et al (2017) Metabolomic analysis identifies potential diagnostic
biomarkers for aspirin-exacerbated respiratory disease. Clin Exp Allergy 47:37–47. https://
doi.org/10.1111/cea.12797
Barnes PJ (2006) Corticosteroid effects on cell signalling. Eur Respir J 27:413–426. https://doi.org/
10.1183/09031936.06.00125404
Pharmacometabolomics of Asthma as a Road Map to Precision Medicine 269

Bigler J, Boedigheimer M, Schofield JPR et al (2017) A severe asthma disease signature from gene
expression profiling of peripheral blood from U-BIOPRED cohorts. Am J Respir Crit Care Med
195:1311–1320. https://doi.org/10.1164/rccm.201604-0866OC
Bisgaard H (2001) Pathophysiology of the cysteinyl leukotrienes and effects of leukotriene receptor
antagonists in asthma. Allergy 56(Suppl 66):7–11. https://doi.org/10.1034/j.1398-9995.56.s66.
2.x
Bisgaard H, Stokholm J, Chawes BL et al (2016) Fish oil-derived fatty acids in pregnancy and
wheeze and asthma in offspring. N Engl J Med 375:2530–2539. https://doi.org/10.1056/
NEJMoa1503734
Blighe K, Chawes BL, Kelly RS et al (2017) Vitamin D prenatal programming of childhood
metabolomics profiles at age 3 y. Am J Clin Nutr 106:1092–1099. https://doi.org/10.3945/
ajcn.117.158220
Bochenek G, Stachura T, Szafraniec K et al (2018) Diagnostic accuracy of urinary LTE4 measure-
ment to predict aspirin-exacerbated respiratory disease in patients with asthma. J Allergy Clin
Immunol Pract 6:528–535. https://doi.org/10.1016/j.jaip.2017.07.001
Bowler RP, Wendt CH, Fessler MB et al (2017) New strategies and challenges in lung proteomics
and metabolomics. An official American Thoracic Society workshop report. Ann Am Thorac
Soc 14:1721–1743. https://doi.org/10.1513/AnnalsATS.201710-770WS
Bozelli JC, Azher S, Epand RM (2021) Plasmalogens and chronic inflammatory diseases. Front
Physiol 12. https://doi.org/10.3389/fphys.2021.730829
Brinkman P, Wagener AH, Hekking P-P et al (2019) Identification and prospective stability of
electronic nose (eNose)-derived inflammatory phenotypes in patients with severe asthma. J
Allergy Clin Immunol 143:1811–1820.e7. https://doi.org/10.1016/j.jaci.2018.10.058
Bruno F, Spaziano G, Liparulo A et al (2018) Recent advances in the search for novel
5-lipoxygenase inhibitors for the treatment of asthma. Eur J Med Chem 153:65–72. https://
doi.org/10.1016/j.ejmech.2017.10.020
Carraro S, Giordano G, Reniero F et al (2013) Asthma severity in childhood and metabolomic
profiling of breath condensate. Allergy 68:110–117. https://doi.org/10.1111/all.12063
Carstairs JR, Nimmo AJ, Barnes PJ (1985) Autoradiographic visualization of beta-adrenoceptor
subtypes in human lung. Am Rev Respir Dis 132:541–547. https://doi.org/10.1164/arrd.1985.
132.3.541
Cazzola M, Page CP, Calzetta L, Matera MG (2012) Pharmacology and therapeutics of
bronchodilators. Pharmacol Rev 64:450–504. https://doi.org/10.1124/pr.111.004580
Chaudhuri R, Livingston E, McMahon AD et al (2003) Cigarette smoking impairs the therapeutic
response to oral corticosteroids in chronic asthma. Am J Respir Crit Care Med 168:1308–1311.
https://doi.org/10.1164/rccm.200304-503OC
Chung KF, Wenzel SE, Brozek JL et al (2014) International ERS/ATS guidelines on definition,
evaluation and treatment of severe asthma. Eur Respir J 43:343–373. https://doi.org/10.1183/
09031936.00202013
Clish CB (2015) Metabolomics: an emerging but powerful tool for precision medicine. Cold Spring
Harb Mol Case Stud 1:a000588. https://doi.org/10.1101/mcs.a000588
Comhair SAA, McDunn J, Bennett C et al (2015) Metabolomic endotype of asthma. J Immunol
195:643–650. https://doi.org/10.4049/jimmunol.1500736
Comhair SAA, Bochenek G, Baicker-McKee S et al (2018) The utility of biomarkers in diagnosis of
aspirin exacerbated respiratory disease. Respir Res 19:210. https://doi.org/10.1186/s12931-018-
0909-6
Cuello-Garcia CA, Fiocchi A, Pawankar R et al (2016) World allergy organization-McMaster
University guidelines for allergic disease prevention (GLAD-P): prebiotics. World Allergy
Organ J 9:10. https://doi.org/10.1186/s40413-016-0102-7
Dallinga JW, Robroeks CMHHT, van Berkel JJBN et al (2010) Volatile organic compounds in
exhaled breath as a diagnostic tool for asthma in children. Clin Exp Allergy 40:68–76. https://
doi.org/10.1111/j.1365-2222.2009.03343.x
270 R. S. Kelly et al.

Darveaux J, Busse WW (2015) Biologics in asthma – the next step toward personalized treatment. J
Allergy Clin Immunol Pract 3:152–160.; quiz 161. https://doi.org/10.1016/j.jaip.2014.09.014
Divekar R, Hagan J, Rank M et al (2016) Diagnostic utility of urinary LTE4 in asthma, allergic
rhinitis, chronic rhinosinusitis, nasal polyps, and aspirin sensitivity. J Allergy Clin Immunol
Pract 4:665–670. https://doi.org/10.1016/j.jaip.2016.03.004
Dona AC, Jiménez B, Schäfer H et al (2014) Precision high-throughput proton NMR spectroscopy
of human urine, serum, and plasma for large-scale metabolic phenotyping. Anal Chem 86:9887–
9894. https://doi.org/10.1021/ac5025039
Eder W, Ege MJ, von Mutius E (2006) The asthma epidemic. N Engl J Med 355:2226–2235.
https://doi.org/10.1056/NEJMra054308
Enilari O, Sinha S (2019) The global impact of asthma in adult populations. Ann Glob Health 85.
https://doi.org/10.5334/aogh.2412
Ferraro VA, Carraro S, Pirillo P et al (2020) Breathomics in asthmatic children treated with inhaled
corticosteroids. Metabolites 10. https://doi.org/10.3390/metabo10100390
Fiehn O (2002) Metabolomics – the link between genotypes and phenotypes. Plant Mol Biol 48:
155–171
Fiehn O, Kim J (2014) Metabolomics insights into pathophysiological mechanisms of interstitial
cystitis. Int Neurourol J 18:106–114. https://doi.org/10.5213/inj.2014.18.3.106
Fitzpatrick AM, Park Y, Brown LAS, Jones DP (2014) Children with severe asthma have unique
oxidative stress-associated metabolomic profiles. J Allergy Clin Immunol 133(258–61):e1–e8.
https://doi.org/10.1016/j.jaci.2013.10.012
Foster A, Chan CC (1991) Peptide leukotriene involvement in pulmonary eosinophil migration
upon antigen challenge in the actively sensitized guinea pig. Int Arch Allergy Appl Immunol 96:
279–284. https://doi.org/10.1159/000235508
Franklin S, Vondriska TM (2011) Genomes, proteomes, and the central dogma. Circ Cardiovasc
Genet 4:576. https://doi.org/10.1161/CIRCGENETICS.110.957795
GBD 2015 Chronic Respiratory Disease Collaborators (2017) Global, regional, and national deaths,
prevalence, disability-adjusted life years, and years lived with disability for chronic obstructive
pulmonary disease and asthma, 1990-2015: a systematic analysis for the Global Burden of
Disease Study 2015. Lancet Respir Med 5:691–706. https://doi.org/10.1016/S2213-2600(17)
30293-X
Gern JE, Lemanske RF, Busse WW (1999) Early life origins of asthma. J Clin Invest 104:837–843.
https://doi.org/10.1172/JCI8272
Gref A, Rautiainen S, Gruzieva O et al (2017) Dietary total antioxidant capacity in early school age
and subsequent allergic disease. Clin Exp Allergy 47:751–759. https://doi.org/10.1111/cea.
12911
Grennan D, Wang S (2019) Steroid side effects. JAMA 322:282–282. https://doi.org/10.1001/
JAMA.2019.8506
Gur M, Hakim F, Bentur L (2017) Better understanding of childhood asthma, towards primary
prevention – are we there yet? Consideration of pertinent literature. F1000Res 6(2152):
10.12688/f1000research.11601.1
Gurnell M, Heaney LG, Price D, Menzies-Gow A (2021) Long-term corticosteroid use, adrenal
insufficiency and the need for steroid-sparing treatment in adult severe asthma. J Intern Med
290:240–256. https://doi.org/10.1111/joim.13273
Haque R, White AA, Jackson DJ, Hopkins C (2021) Clinical evaluation and diagnosis of aspirin-
exacerbated respiratory disease. J Allergy Clin Immunol 148:283–291. https://doi.org/10.1016/
j.jaci.2021.06.018
Hargreave FE, Nair P (2009) The definition and diagnosis of asthma. Clin Exp Allergy 39:1652–
1658. https://doi.org/10.1111/j.1365-2222.2009.03321.x
Hastie AT, Moore WC, Meyers DA et al (2010) Analyses of asthma severity phenotypes and
inflammatory proteins in subjects stratified by sputum granulocytes. J Allergy Clin Immunol
125:1028–1036.e13. https://doi.org/10.1016/j.jaci.2010.02.008
Pharmacometabolomics of Asthma as a Road Map to Precision Medicine 271

Henderson WR, Lewis DB, Albert RK et al (1996) The importance of leukotrienes in airway
inflammation in a mouse model of asthma. J Exp Med 184:1483–1494. https://doi.org/10.1084/
jem.184.4.1483
Hunt JF, Erwin E, Palmer L et al (2002) Expression and activity of pH-regulatory glutaminase in the
human airway epithelium. Am J Respir Crit Care Med 165:101–107. https://doi.org/10.1164/
ajrccm.165.1.2104131
Jones KA, Perkins WJ, Lorenz RR et al (1999) F-actin stabilization increases tension cost during
contraction of permeabilized airway smooth muscle in dogs. J Physiol 519(Pt 2):527–538.
https://doi.org/10.1111/j.1469-7793.1999.0527m.x
Kachroo P, Sordillo JE, Lutz SM et al (2021) Pharmaco-metabolomics of inhaled corticosteroid
response in individuals with asthma. J Pers Med 11:1148. https://doi.org/10.3390/jpm11111148
Kachroo P, Stewart I, Kelly R et al (2022) The systematic use of metabolomic epidemiology,
biobanks, and electronic medical records for precision medicine initiatives in asthma: findings
suggest new guidelines to optimize treatment. https://doi.org/10.21203/rs.3.rs-268507/v1
Kaddurah-Daouk R, Weinshilboum RM (2014) Pharmacometabolomics: implications for clinical
pharmacology and systems pharmacology. Clin Pharmacol Therap 95:154–167. https://doi.org/
10.1038/clpt.2013.217
Kannisto S, Korppi M, Remes K, Voutilainen R (2001) Serum dehydroepiandrosterone sulfate
concentration as an indicator of adrenocortical suppression in asthmatic children treated with
inhaled steroids. J Clin Endocrinol Metab 86:4908–4912. https://doi.org/10.1210/jcem.86.10.
7975
Kannisto S, Laatikainen A, Taivainen A et al (2004) Serum dehydroepiandrosterone sulfate
concentration as an indicator of adrenocortical suppression during inhaled steroid therapy in
adult asthmatic patients. Eur J Endocrinol 150:687–690. https://doi.org/10.1530/eje.0.1500687
Kelly RS, Dahlin A, McGeachie MJ et al (2017) Asthma metabolomics and the potential for
integrative omics in research and the clinic. Chest 151:262–277. https://doi.org/10.1016/j.
chest.2016.10.008
Kelly RS, Chawes BL, Guo F et al (2019a) The role of the 17q21 genotype in the prevention of
early childhood asthma and recurrent wheeze by vitamin D. Eur Respir J 54. https://doi.org/10.
1183/13993003.00761-2019
Kelly RS, Sordillo JE, Lutz SM et al (2019b) Pharmacometabolomics of bronchodilator response in
asthma and the role of age-metabolite interactions. Metabolites 9. https://doi.org/10.3390/
metabo9090179
Kelly RS, Mendez KM, Huang M et al (2021) Metabo-endotypes of asthma reveal differences in
lung function: discovery and validation in two TOPMed cohorts. Am J Respir Crit Care Med.
https://doi.org/10.1164/rccm.202105-1268OC
Kersten ETG, Koppelman GH (2017) Pharmacogenetics of asthma: toward precision medicine.
Curr Opin Pulm Med 23:12–20. https://doi.org/10.1097/MCP.0000000000000335
Knihtilä HM, Kelly RS, Brustad N et al (2021) Maternal 17q21 genotype influences prenatal
vitamin D effects on offspring asthma/recurrent wheeze. Eur Respir J 58. https://doi.org/10.
1183/13993003.02012-2020
Kuruvilla ME, Lee FE-H, Lee GB (2019) Understanding asthma phenotypes, endotypes, and
mechanisms of disease. Clin Rev Allergy Immunol 56:219–233. https://doi.org/10.1007/
s12016-018-8712-1
Larsson K, Kankaanranta H, Janson C et al (2020) Bringing asthma care into the twenty-first
century. NPJ Prim Care Respir Med 30:25. https://doi.org/10.1038/s41533-020-0182-2
Lewis LM, Ferguson I, House SL et al (2014) Albuterol administration is commonly associated
with increases in serum lactate in patients with asthma treated for acute exacerbation of asthma.
Chest 145:53–59. https://doi.org/10.1378/chest.13-0930
Litonjua AA, Carey VJ, Laranjo N et al (2020) Six-year follow-up of a trial of antenatal vitamin D
for asthma reduction. N Engl J Med 382:525–533. https://doi.org/10.1056/NEJMoa1906137
Loureiro CC, Duarte IF, Gomes J et al (2014) Urinary metabolomic changes as a predictive
biomarker of asthma exacerbation. J Allergy Clin Immunol 133:261-3.e1–261-3.e5. https://
doi.org/10.1016/j.jaci.2013.11.004
272 R. S. Kelly et al.

Loza MJ, Djukanovic R, Chung KF et al (2016) Validated and longitudinally stable asthma
phenotypes based on cluster analysis of the ADEPT study. Respir Res 17:165. https://doi.org/
10.1186/s12931-016-0482-9
Matera MG, Page CP, Calzetta L et al (2020) Pharmacology and therapeutics of bronchodilators
revisited. Pharmacol Rev 72:218–252. https://doi.org/10.1124/pr.119.018150
Mattarucchi E, Baraldi E, Guillou C (2012) Metabolomics applied to urine samples in childhood
asthma; differentiation between asthma phenotypes and identification of relevant metabolites.
Biomed Chromatogr 26:89–94. https://doi.org/10.1002/bmc.1631
McGeachie MJ, Dahlin A, Qiu W et al (2015) The metabolomics of asthma control: a promising
link between genetics and disease. Immun Inflamm Dis 3:224–238. https://doi.org/10.1002/
iid3.61
Moore WC, Meyers DA, Wenzel SE et al (2010) Identification of asthma phenotypes using cluster
analysis in the severe asthma research program. Am J Respir Crit Care Med 181:315–323.
https://doi.org/10.1164/rccm.200906-0896OC
Mulvihill MM, Nomura DK (2014) Metabolomic strategies to map functions of metabolic
pathways. Am J Physiol Endocrinol Metab 307:E237–E244. https://doi.org/10.1152/ajpendo.
00228.2014
Newton R, Giembycz MA (2016) Understanding how long-acting β2 -adrenoceptor agonists
enhance the clinical efficacy of inhaled corticosteroids in asthma – an update. Br J Pharmacol
173:3405–3430. https://doi.org/10.1111/bph.13628
Pandya D, Puttanna A, Balagopal V (2014) Systemic effects of inhaled corticosteroids: an over-
view. Open Respir Med J 8:59–65. https://doi.org/10.2174/1874306401408010059
Papamichael MM, Katsardis C, Sarandi E et al (2021) Application of metabolomics in pediatric
asthma: prediction, diagnosis and personalized treatment. Metabolites 11. https://doi.org/10.
3390/metabo11040251
Paredi P, Kharitonov SA, Barnes PJ (2000) Elevation of exhaled ethane concentration in asthma.
Am J Respir Crit Care Med 162:1450–1454. https://doi.org/10.1164/ajrccm.162.4.2003064
Park YH, Fitzpatrick AM, Medriano CA, Jones DP (2017) High-resolution metabolomics to
identify urine biomarkers in corticosteroid-resistant asthmatic children. J Allergy Clin Immunol
139:1518–1524.e4. https://doi.org/10.1016/j.jaci.2016.08.018
Peluso A, Glen R, Ebbels TMD (2021) Multiple-testing correction in metabolome-wide association
studies. BMC Bioinform 22:67. https://doi.org/10.1186/s12859-021-03975-2
Peters-Golden M, Henderson WR (2007) Leukotrienes. N Engl J Med 357:1841–1854. https://doi.
org/10.1056/NEJMra071371
Quan-Jun Y, Jian-Ping Z, Jian-Hua Z et al (2017) Distinct metabolic profile of inhaled budesonide
and salbutamol in asthmatic children during acute exacerbation. Basic Clin Pharmacol Toxicol
120:303–311. https://doi.org/10.1111/bcpt.12686
Rago D, Rasmussen MA, Lee-Sarwar KA et al (2019) Fish-oil supplementation in pregnancy, child
metabolomics and asthma risk. EBioMedicine 46:399–410. https://doi.org/10.1016/j.ebiom.
2019.07.057
Reddel HK, Bacharier LB, Bateman ED et al (2021) Global initiative for asthma (GINA) strategy
2021 – executive summary and rationale for key changes. J Allergy Clin Immunol Pract. https://
doi.org/10.1016/j.jaip.2021.10.001
Reinke SN, Gallart-Ayala H, Gómez C et al (2017) Metabolomics analysis identifies different
metabotypes of asthma severity. Eur Respir J 49. https://doi.org/10.1183/13993003.01740-2016
Rodrigo GJ (2014) Serum lactate increase during acute asthma treatment: a new piece of the puzzle.
Chest 145:6–7. https://doi.org/10.1378/chest.13-2042
Rodrigo GJ, Rodrigo C (2005) Elevated plasma lactate level associated with high dose inhaled
albuterol therapy in acute severe asthma. Emerg Med J 22:404–408. https://doi.org/10.1136/
emj.2003.012039
Ruffin RE, McIntyre E, Crockett AJ et al (1982) Combination bronchodilator therapy in asthma. J
Allergy Clin Immunol 69:60–65. https://doi.org/10.1016/0091-6749(82)90089-6
Pharmacometabolomics of Asthma as a Road Map to Precision Medicine 273

Sahiner UM, Birben E, Erzurum S et al (2011) Oxidative stress in asthma. World Allergy Organ J 4:
151–158. https://doi.org/10.1097/WOX.0b013e318232389e
Sinha A, Krishnan V, Sethi T et al (2012) Metabolomic signatures in nuclear magnetic resonance
spectra of exhaled breath condensate identify asthma. Eur Respir J 39:500–502. https://doi.org/
10.1183/09031936.00047711
Sinha A, Desiraju K, Aggarwal K et al (2017) Exhaled breath condensate metabolome clusters for
endotype discovery in asthma. J Transl Med 15:262. https://doi.org/10.1186/s12967-017-
1365-7
Smith LJ, Shamsuddin M, Houston M (1993) Effect of leukotriene D4 and platelet-activating factor
on human alveolar macrophage eicosanoid and PAF synthesis. Am Rev Respir Dis 148:682–
688. https://doi.org/10.1164/ajrccm/148.3.682
Sordillo JE, Lutz SM, Kelly RS et al (2020) Plasmalogens mediate the effect of age on bronchodi-
lator response in individuals with asthma. Front Med (Lausanne) 7:38. https://doi.org/10.3389/
fmed.2020.00038
Svenningsen S, Nair P (2017) Asthma Endotypes and an overview of targeted therapy for asthma.
Front Med (Lausanne) 4:158. https://doi.org/10.3389/fmed.2017.00158
Tyler SR, Bunyavanich S (2019) Leveraging -omics for asthma endotyping. J Allergy Clin
Immunol 144:13–23. https://doi.org/10.1016/j.jaci.2019.05.015
Veeranki SP, Gebretsadik T, Mitchel EF et al (2015) Maternal folic acid supplementation during
pregnancy and early childhood asthma. Epidemiology 26:934–941. https://doi.org/10.1097/
EDE.0000000000000380
Walter MA, Linsley PS, Cox DW (1987) Apa I polymorphism of a human immunoglobin VH3
subclass locus. Nucleic Acids Res 15:4697. https://doi.org/10.1093/nar/15.11.4697
Wang et al (2021) Research progress of metabolomics in asthma. Metabolites. PMID: 34564383
Wedes SH, Wu W, Comhair SAA et al (2011) Urinary bromotyrosine measures asthma control and
predicts asthma exacerbations in children. J Pediatr 159:248–55.e1. https://doi.org/10.1016/j.
jpeds.2011.01.029
Wei R, Wang J, Su M et al (2018) Missing value imputation approach for mass spectrometry-based
metabolomics data. Sci Rep 8:663. https://doi.org/10.1038/s41598-017-19120-0
Wendell SG, Baffi C, Holguin F (2014) Fatty acids, inflammation, and asthma. J Allergy Clin
Immunol 133:1255–1264. https://doi.org/10.1016/j.jaci.2013.12.1087
Wendell SG, Fan H, Zhang C (2020) G protein-coupled receptors in asthma therapy: pharmacology
and drug action. Pharmacol Rev 72:1–49. https://doi.org/10.1124/pr.118.016899
Wenzel SE (2012) Asthma phenotypes: the evolution from clinical to molecular approaches. Nat
Med 18:716–725. https://doi.org/10.1038/nm.2678
Werner M, Jordan PM, Romp E et al (2019) Targeting biosynthetic networks of the
proinflammatory and proresolving lipid metabolome. FASEB J 33:6140–6153. https://doi.org/
10.1096/fj.201802509R
Wishart DS (2019) Metabolomics for investigating physiological and pathophysiological processes.
Physiol Rev 99:1819–1875. https://doi.org/10.1152/physrev.00035.2018
Wolsk HM, Chawes BL, Litonjua AA et al (2017) Prenatal vitamin D supplementation reduces risk
of asthma/recurrent wheeze in early childhood: a combined analysis of two randomized
controlled trials. PLoS One 12:e0186657. https://doi.org/10.1371/journal.pone.0186657
Yaghoubi M, Adibi A, Safari A et al (2019) The projected economic and health burden of
uncontrolled asthma in the United States. Am J Respir Crit Care Med 200:1102–1112. https://
doi.org/10.1164/rccm.201901-0016OC
Yokomizo T, Nakamura M, Shimizu T (2018) Leukotriene receptors as potential therapeutic
targets. J Clin Invest 128:2691–2701. https://doi.org/10.1172/JCI97946
Zhu Z, Camargo CA, Hasegawa K (2019) Metabolomics in the prevention and management of
asthma. Expert Rev Respir Med 13:1135–1138. https://doi.org/10.1080/17476348.2019.
1674650
Prospective Metabolomic Studies
in Precision Medicine: The AKRIBEA Project

Maider Bizkarguenaga, Rubén Gil-Redondo, Chiara Bruzzone,

Ganeko Bernardo-Seisdedos, Ana Laín, Beatriz González-Valle,
Nieves Embade, José M. Mato, and Oscar Millet

Contents
1 Metabolomics as a Tool for Precision Medicine . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 276
1.1 Precision Medicine and Molecular Medicine . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 276
1.2 Biomarkers and Metabolomics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 277
1.3 Description of the Accessible NMR Metabolic Landscape . . . . . . . . . . . . . . . . . . . . . . . . . . 278
1.4 On the Cohort Size for Applications in Precision Medicine . . . . . . . . . . . . . . . . . . . . . . . . . 282
2 Large-Cohort Sample Collection and Biobanking for Metabolomic Studies . . . . . . . . . . . . . . 283
3 Standard Operating Procedures for Metabolomics in Precision Medicine . . . . . . . . . . . . . . . . . 285
4 Artificial Intelligence and Data Mining: Data Is the Way . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 286
4.1 Data Management . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 286
4.2 Data Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 288
4.3 Biological Contextualization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 289
4.4 From the Laboratory to the Clinic . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 289
5 Applications of Precision Medicine: Non-oriented and Oriented Metabolomics
Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 290
5.1 Non-oriented Metabolomics Study: Metabolic Syndrome . . . . . . . . . . . . . . . . . . . . . . . . . . . 290
5.2 Oriented Metabolomics Study: COVID-19 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 293
6 Concluding Remarks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 295
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 295

Abstract
For a long time, conventional medicine has analysed biomolecules to diagnose
diseases. Yet, this approach has proven valid only for a limited number of
metabolites and often through a bijective relationship with the disease

M. Bizkarguenaga · R. Gil-Redondo · C. Bruzzone · G. Bernardo-Seisdedos · A. Laín ·

B. González-Valle · N. Embade · J. M. Mato · O. Millet (✉)
Precision Medicine and Metabolism Laboratory, CIC bioGUNE, Basque Research and Technology
Alliance, Bizkaia Science and Technology Park, Derio, Bizkaia, Spain
e-mail: [email protected]; [email protected]; [email protected];
[email protected]; [email protected]; [email protected];
[email protected]; [email protected]; [email protected]

# The Author(s), under exclusive license to Springer Nature Switzerland AG 2022 275
V. Ghini et al. (eds.), Metabolomics and Its Impact on Health and Diseases,
Handbook of Experimental Pharmacology 277, https://doi.org/10.1007/164_2022_610
276 M. Bizkarguenaga et al.

(i.e. glucose relationship with diabetes), ultimately offering incomplete diagnos-

tic value. Nowadays, precision medicine emerges as an option to improve the
prevention and/or treatment of numerous pathologies, focusing on the molecular
mechanisms, acting in a patient-specific dimension, and leveraging multiple
contributing factors such as genetic, environmental, or lifestyle. Metabolomics
grasps the required subcellular complexity while being sensitive to all these
factors, which results in a most suitable technique for precision medicine. The
aim of this chapter is to describe how NMR-based metabolomics can be
integrated in the design of a precision medicine strategy, using the Precision
Medicine Initiative of the Basque Country (the AKRIBEA project) as a case
study. To that end, we will illustrate the procedures to be followed when
conducting an NMR-based metabolomics study with a large cohort of
individuals, emphasizing the critical points. The chapter will conclude with the
discussion of some relevant biomedical applications.

Keywords
Biomarker · NMR metabolomics · Precision medicine

1 Metabolomics as a Tool for Precision Medicine

1.1 Precision Medicine and Molecular Medicine

A silent revolution is steadily transforming traditional medicine into the novel field
of personalized medicine, which tailors medical treatment to the individual
characteristics of each patient (Kohler et al. 2017). Currently, medical practice is
mostly based in evidence, organized at the organ level and administering a rather
uniform treatment for a given disease. In contrast, precision medicine concentrates
on the mechanistic aspects of the disease, is organized at the cellular and molecular
levels, and aims to provide the most suitable personalized treatment for each patient.
As a consequence of this innovative strategy, personalized medicine focuses on early
detection and prospective evaluation, while canonical medical praxis is based on
statistically based retrospective analyses to provide curative and palliative care.
To achieve this goal, the concept of precision medicine, first used in 2009 in the
book The Innovator’s Prescription by Clayton Christensen (Christensen et al. 2009),
exploits the analysis of clinical and molecular data of patient subpopulations, which
tries to classify a subject into a healthy individual or a patient by contrasting
biological information with datasets from large population cohorts, enriched in
biological diversity. This strategy allows to endotype a particular disease, ultimately
leading to personalized care (Śliwczyński and Orlewska 2016). Furthermore, the
generation of prediction models for prevention and diagnosis paves the way for
application of precision medicine in a rationally designed drug treatment for geneti-
cally or metabolically stratified groups of patients (Authors 2021). In fact, this
strategy not only tailors the appropriate drug treatment for each patient (Tranvåg
Prospective Metabolomic Studies in Precision Medicine: The AKRIBEA Project 277

et al. 2021) but also provides an explanation for the non-responding population in
clinical phase trials of drugs, where patient sub-classification also actively
contributes to the better understanding of the mode-of-action of drugs (John et al.
2020; Puchades-Carrasco and Pineda-Lucena 2017).
As already mentioned, traditional medicine is rooted on medical evidence (mea-
surable symptomatology), from which only a limited fraction of it arises from
biochemical analyses (i.e. molecular medicine): glucose, cholesterol, high-density
lipoprotein/low-density lipoprotein, transaminases, or creatinine, among others, to
monitor diabetes, cardiovascular, liver, and renal disorders, respectively. Perhaps,
the most powerful diagnostic use of metabolites currently occurs in the heel test to
screen for neonatal congenital methabolopathies (Beger et al. 2016). In contrast,
precision medicine heavily relies on genomics, proteomics, and metabolomics for an
appropriate biochemical characterization of the individual. To that end, it integrates
molecular information originated by all the abovementioned “omic” techniques with
the medical records, environmental surroundings, diet, and lifestyle of the popula-
tion under consideration. Altogether, this approach can provide predictive, diagnos-
tic, and prognostic markers for a plethora of diseases, informing on their underlying
molecular mechanisms and enabling their sub-classification and the stratification of
patients based on the metabolic pathways involved (Śliwczyński and Orlewska
2016; Clish 2015).

1.2 Biomarkers and Metabolomics

Precision medicine is intimately coupled to biomarker validation. Indeed,

biomarkers have a great potential as a diagnostic tool for the clinicians and to help
deciding optimal treatments, dietary restrictions, or other major therapeutic lifestyle
changes (Kohler et al. 2017; Beger et al. 2016). Biomarkers can be classified into
two categories: dynamic biomarkers, that can monitor disease progression and
treatment responses, and static biomarkers, used as a prognostic tool for the
abovementioned conditions (Kohler et al. 2017; Hartl et al. 2021). In this context,
metabolomics is based on the measurement of the low-molecular-weight molecules
and metabolites of different cellular processes of a biological sample and has a
crucial role in defining the molecular data that can have an impact in precision
medicine (Kohler et al. 2017). Metabolomics quantitatively characterizes the inter-
mediate- and end-products of the biochemical processes, which not only are
associated to genetic diversity, but also to environmental events, and disease
conditions or treatment responses (Kohler et al. 2017; Puchades-Carrasco and
Pineda-Lucena 2017; Clish 2015). Furthermore, it provides a time-specific definition
of the metabolic state since metabolome is in constant adaptation to gene expression,
nutrient and drug intake, other external influences, variations in the gut microbiota
and, importantly, to the natural history of the diseases.
A proper implementation of metabolomics into precision medicine and clinical
practice optimally implies that the biomarkers monitorization must be conducted in a
non-invasive and accessible matrix (i.e. biofluids). In theory, a biofluid should be
278 M. Bizkarguenaga et al.

proximal to the organic focus of the disease but, in reality, most diseases also have
significant systemic manifestations (Hartl et al. 2021). In this context, serum,
plasma, and urine are the most suitable biofluids, characterized by optimal sensitivity
to the time-dependent metabolome, disponible nature, and easy storage. Plasma and
serum metabolomes undergo limited daily variation, also providing information of
the extracellular metabolome, whereas urine is the most on-demand available
biofluid (Kohler et al. 2017).
Nowadays, metabolomic research technology enables the identification and
quantification of hundreds of distinct metabolites from biofluids, using technologies
that can be easily extrapolated to large-scale cohorts of donors and thousands of
samples. For the sake of reproducibility, Standardized Operating Procedures (SOPs)
must be implemented for sample collection, storage, handling, analytical analyses,
and data processing (Beger et al. 2016). Under these premises, research laboratories
with standardized infrastructures will generate dataset that become additive,
enabling large-scale and multicentred studies in metabolomics. Techniques that
nowadays can fulfil these requirements include Nuclear Magnetic Resonance spec-
troscopy (NMR), Liquid Chromatography coupled to Mass Spectrometry (LC-MS),
and Gas Chromatography with Mass Spectrometry (GC-MS). In here, we will focus
on NMR spectroscopy, which is able to simultaneously quantify hundreds of
metabolites in a biological sample with an acquisition time of less than 20 min.

1.3 Description of the Accessible NMR Metabolic Landscape

The human metabolome comprehends up to 100,000 metabolites that participate in

more than 18,000 enzymatic reactions, which have to be in constant healthy equilib-
rium (Wishart et al. 2018). Metabolic phenotyping of biofluids and polar/apolar
tissue extracts are powerful ways to monitor metabolite concentrations (Bernardo-
Seisdedos et al. 2021; Dona et al. 2014). In 1974, 13C-NMR spectroscopy was first
used in metabolic studies to investigate metabolic processes (Wilson et al. 1974).
Since then, NMR metabolomics experienced a huge development thanks to its
ability to characterize the chemical composition of complex biological mixtures
(Wishart 2019). Actually, in a few minutes, a one-dimensional proton NMR spec-
trum can detect tenths of molecules, which are in the micromolar concentration
range, typically including amino acids, carbohydrates, alcohols, organic acids, the
intermediates or end-products of the main metabolic pathways, and also the overall
lipid composition of a biosystem (reviewed in (Vignoli et al. 2019)).
The metabolic landscape that is accessible to NMR depends on the investigated
matrix and the experimental choice (i.e. the active nucleus to be monitored) (Fig. 1).
Urine is an especially well suited biofluid for metabolomics since it perfectly reflects
the body's metabolism. In addition, as a biological waste material, it typically
contains catabolic products from a wide range of foods, drinks, drugs, environmental
contaminants, endogenous waste metabolites, and bacterial by-products. Urine
NMR analysis can routinely quantify over 100 metabolites including glucose and
other carbohydrates, lipids, aromatic amino acids, and p-cresol sulphate, maltitol,
 Prospective Metabolomic Studies in Precision Medicine: The AKRIBEA Project

Fig. 1 Central metabolism pathways. NMR analysis allows the characterization of certain metabolites that are spread all over the pathways. The colour code
represents the biofluid or methodology sensitive to the detected metabolite (see legend). The scheme is based on KEGG hsa01100 map. Abbreviations: (R)-3-
Hydroxybutyric acid, 3-HB; (S)-3-Hydroxy-3-methylglutaryl-CoA, HMG-CoA; (S)-3-Methyl-2-oxopentanoic acid, KMV; 1,2-Diacylglycerol, LPC;
1-Methylnicotinamide, MNA; 2,5-Dihydroxyphenylacetic acid, 4-HPP; 2-Hydroxyphenylacetic acid, HPAA; 2-Hydroxytricarballylic acid, CA;
279

2-Oxobutanoic acid; FA; 2-Phospho-D-glycerate, 2PG; 3,4-Dihydroxy-L-phenylalanine, L-DOPA; 3',5'-Cyclic AMP, NMPc; 3',5'-Cyclic GMP, NMPc;
3-Methyl-2-oxobutanoic acid, KIV; 3-Phospho-D-glycerate, 3PG; 4-Fumarylacetoacetic acid, 4-FAcAc; 4-Maleylacetoacetate, 4-MAcAc; 4-Methyl-2-
280 M. Bizkarguenaga et al.

among others (Fig. 1). The abnormal quantity of some of these metabolites can be
associated with clinical pathologies and prevalent syndromes (Bruzzone et al. 2021).
Additionally, the urine metabolome can also be scrutinized for the absence of
metabolites, to confirm that the individual does not suffer a congenital
metabolopathy, as an alternative to the newborn screening heel test (Embade et al.
2019).
Serum is the blood moiety devoid of cell and clotting factors, rich in proteins, and
has been used in numerous diagnostic tests. Serum is metabolically very homeostatic
and over 50 components can be easily identified and quantified by NMR spectros-
copy (Fig. 1). Yet, the added value of NMR-based analysis of serum is the lipopro-
tein profiling (Raftery, “Quantitative NMR methods in metablomics”). Lipoproteins
are a group of micelles consisting of lipids and proteins whose main function is to

Fig. 1 (continued) oxopentanoate, KIC; 4-Pyridoxate, 4-PA; 5,6-Dihydrouracil, DHU;

5-Aminolevulinate, ALA; 6-Phospho-D-gluconate, 6PG; Acetic acid, AcOH; Acetoacetic acid,
AcAc; Adenine, A; Adenosine, Ado; Adenosine diphosphate, ADP; ADPribose, ADPR; Alanine,
Ala; alpha-D-Galactose, Gal; alpha-Ketoglutaric acid, KGA; Adenosine monophosphate, AMP;
Apolipoproteins, Apo; Arginine, Arg; Asparagine, Asp; Adenosine triphosphate, ATP; beta-D-
Glucose, G; beta-D-Glucose 6-phosphate, G6P; Betaine, BT; Bilirubin, BR; Biliverdin, BV;
Butyric acid, BA; C00570, CDPE; C04475, GPE; C05977, LPA; Caffeine, CAF; Carnosine, Car;
Cytidine diphosphate, CDP; Choline, Ch; Choline phosphate, PCH; Cytidine monophosphate,
CMP; Coproporphyrin III, COPROIII; Coproporphyrinogen I, COPROgenI; Coproporphyrinogen
III, COPROgenIII; Creatine, Cr; Cytidine triphosphate, CTP; Cytidine 5'-diphosphocholine,
CDPCh; D-Erythrose 4-phosphate, E4P; D-Fructose 1,6-bisphosphate, FBP; D-Fructose 6-phos-
phate, F6P; D-Gluconic acid, GlA; D-Glucose 1-phosphate, G1P; Diphosphatidylglycerol, CL;
D-Sedoheptulose 7-phosphate, S7P; D-Sedoheptulose 7-phosphate, S7P; Deoxythymidine diphos-
phate dTDP; D-Xylulose 5-phosphate, X5P; Ethanol, EtOH; Ethanolamine phosphate, PEA;
Ethylenesuccinic acid, Suc; Formic acid, FA; Fumaric acid, FuA; Galactitol, GAT; gamma-
Aminobutyric acid, GABA; Guanosine diphosphate, GDP; GDP-L-fucose, GDPF; GDPmannose,
GDPMan; Glutamic acid, Glu; Glutamine, Gln; Glycerol, Gly; Glycine, Gly; Guanosine
monophosphate, GMP; Guanosine triphosphate, GTP, Guanidinoacetate, GAA; Heme, Heme;
Hippurate, HIP; Histidine, His; Hydracrylic acid, 3-HPA; Hydroxymethylbilane, HMB; Inosine,
I; Isoleucine, Ile; L-2-Hydroxybutanedioic acid, MA; L-Argininosuccinate, ArgSuc; L-Citrulline,
Citru; L-Cystine, Cys; Leucine, Leu; L-Fucose, Fuc; L-Kynurenine, Kyn; L-Lactic acid, Lac;
L-Ornithine, L-NIO; Lysine, Lys; Methionine, Met; Methylmalonic acid, MMA; myo-Inositol,
MI; N(pai)-Methyl-L-histidine, 1-MH; N,N-Dimethylglycine, DMG; N-Acetyl-D-glucosamine
6-phosphate, GlcNAc-6P; N-Acetyl-L-aspartate, NAA; N-Acetyl-L-glutamic acid, NAcGlu; Nico-
tinamide adenine dinucleotide, NAD+; Nicotinamide adenine dinucleotide phosphate, NADP+;
Oxosuccinic acid, OAA; Pantothenic acid, PA; Phenylacetic acid, PAA; Phenylalanine, Phe;
Phenylpyruvate, PPY; Phosphatidylcholine, PC; Phosphatidylethanolamine, PE;
Phosphatidylglycerol, PG; Phosphatidylinositol, PI; Phosphatidylserine, PS; Phosphocreatine,
PCr; Phosphoenolpyruvate, PEP; p-Hydroxyphenylpyruvic acid, HPPA; Pidolic acid, PCA;
Porphobilinogen, PBG; Proline, Pro; Propionic acid, ProA; Protoporphyrin, PROTOIX;
Protoporphyrinogen IX, PROTOgenIX; Pyroracemic acid, Pyr; Quinolinic acid, QA; Riboflavin,
RF; Sarcosine, Sarc; Sphingomyelin, SPM; Taurine, Tau; Thiamine monophosphate, TMP; Thia-
mine triphosphate, TTP; Threonine, Thr; Thymine, T; Tryptophan, Try; Tyramine, Tyra; Tyrosine,
Tyr; Uridine diphosphate, UDP; UDP-D-galactose, UDPgal; UDPglucose, UDPG;
UDPglucuronate, UDPGlu; UDP-N-acetyl-D-glucosamine, UDPNAcGA; Uridine monophosphate,
UMP; Uracil, U; Uracil-6-carboxylic acid, OA, Uridine, Uri; Uroporphyrinogen I, UROgenI;
Uroporphyrinogen III, UROgenIII; Uridine triphosphate, UTP; Valine, Val
Prospective Metabolomic Studies in Precision Medicine: The AKRIBEA Project 281

transport insoluble plasma lipids such as cholesterol or triglycerides from absorption

and/or synthesis organs such as the intestine or liver to the places where they need to
be processed or used. These are chylomicrons, very low-density lipoproteins
(VLDL), intermediate density lipoproteins (IDL), low-density lipoproteins (LDL),
and high-density lipoproteins (HDL) (Martin et al. 2015).
Proteins found in these compounds or apolipoproteins are necessary for the
assembly, structure, and function of lipoproteins. They activate key enzymes for
the metabolism of lipoproteins and are also capable of acting as ligands for cell
surface receptors. ApoA-I, which is synthesized in the liver and intestine, is found in
almost all HDL particles while ApoA-II is the second most abundant HDL apolipo-
protein. ApoB is the main structural protein of chylomicrons, VLDL, IDL, and LDL
and only one ApoB molecule is present in each lipoprotein particle. ApoB is the only
apolipoprotein that does not transfer between different lipoprotein particles. ApoE is
present in chylomicrons, VLDL, and IDL, and plays a critical role in the metabolism
of triglyceride-rich particles. Three apolipoproteins of the C series (ApoC-I, ApoC-
II, and ApoC-III) are also involved in the metabolism of triglyceride-rich
lipoproteins.
NMR has been extensively used to identify the different classes of lipoproteins
(Otvos et al. 1992) and to elucidate and quantify subclasses of the known
lipoproteins (Jiménez et al. 2018). This strategy has proven especially useful to
unravel the molecular mechanisms associated to cardiovascular risk, metabolic
syndrome (Bruzzone et al. 2021), or infections such as SARS-CoV-2 virus
(Bruzzone et al. 2020a).
One caveat for the metabolic characterization is its intrinsic overwhelming
complexity. In this regard, our group hypothesized that the analysis of human
metabolome could be simplified by just analysing the phosphorylated chemical
compounds (Bernardo-Seisdedos et al. 2021). Due to the strategic deployment of
phosphorylated metabolites (Fig. 1), the analysis of the phosphorylated moiety of the
metabolism reduces complexity at an acceptable functional information loss. Spe-
cifically, phosphorylated metabolites occupy a prominent position in all metabolic
and catabolic pathways that could be used as readouts or reporters of the balance
between glycolysis, the tricarboxylic acid cycle (TCA), gluconeogenesis, the pen-
tose phosphate pathway (PPP), the oxidative phosphorylation, and phospholipid
metabolism. To this aim, we developed a strict SOP for the extraction of mouse
liver and hepatocellular carcinoma tissues although it could be applied to different
tissues, fluids, or biological systems. The detection and quantification of multiple
hydrophilic and hydrophilic phosphorylated metabolites relies on 31P-NMR–based
methodology. This procedure worked both with murine and human liver samples
and provides the absolute concentration of up to 54 phosphorylated metabolites
(Fig. 1). Apart from the already mentioned pathways, obtained from the hydrophilic
phase, the hydrophobic (lipophilic) phase provides information on the phospholipid
metabolism such as phosphatidylethanolamine (PE), phosphatidylcholine (PC),
phosphatidylinositol (PI), phosphatidylserine (PS), sphingomyelin (SM), and
cardiolipin (CL) (Bernardo-Seisdedos et al. 2021).
282 M. Bizkarguenaga et al.

1.4 On the Cohort Size for Applications in Precision Medicine

Metabolomic studies investigating human populations can be carried out as a single

interventional study or by longitudinal studies (Nabi et al. 2021; Mignot et al. 2021;
Carrat et al. 2021; Ruiz et al. 2016), where different samples are recollected over
time from the same individual. Ideally, the cohort must be large enough to ensure
statistical power and to embrace the natural metabolic variability. Actually, a proper
planning, optimal quality controls as well as unified data collection are essential to
minimize biases. That said, working with large cohorts is challenging due to a
number of reasons that include complex logistics to ensure the proper sample
recollection and handling (i.e. abiding the SOPs), volunteer’ adherence or change
in their lifestyle during the longitudinal studies, and the associated economic burden,
which may span over several years of sample collection. All these complications
may explain why most of the studies in the literature (for metabolomics and other
omics as well) scrutinized cohorts with a limited number of participants. Of course,
exceptional studies with a large number of samples can also be found, including a
very complete study (genomics, epigenomics, transcriptomics, genomics, and
metabolomics) of a Dutch cohort of more than 1,000 participants (Tigchelaar et al.
2015). Another example targeted the Spanish Mediterranean population (1,011
individuals) to investigate genome-wide association studies (GWAS) (Ortega-
Azorín et al. 2019). Also, the LIFE-Adult-Study recently completed the baseline
examination of 10,000 randomly selected participants in Germany with the aim of
investigating prevalence, markers, genetic predisposition, and lifestyle associated to
the most common diseases (Loeffler et al. 2015). For this purpose, they used a
mixture of the usual clinical trials for the diagnosis of numerous diseases combined
with genomics and metabolomics. Finally, one of the large-scale projects developed
in recent years is the Tohoku Medical Megabank (TMM) with the aim of conducting
prospective cohort studies in several of the regions that were affected by the 2011
Japan earthquake (Koshiba et al. 2018). During the first stage of these studies (from
2013 to 2017), more than 150,000 participants have been successfully recruited.
At CIC bioGUNE we are developing a precision medicine initiative, vehiculated
through a longitudinal study (AKRIBEA) that targets a cohort of 10,000 individuals
from the Basque Country. The study is non-oriented and the only inclusion criteria is
to belong to the working population of the Mondragón Cooperative (with about
40,000 employees), since the volunteer recruitment and the sample collection are
carried out during their annual medical check-up. Available data includes general
characteristics, biochemical and lifestyle data, the NMR-based identification and
quantification of metabolites in urine and serum, the lipoprotein analysis of serum
also obtained by NMR spectroscopy and, to a lesser extent, the genotyping and
proteomic analysis of the population. This sample dataset can be interpreted as a
reference cohort to investigate a plethora of diseases and syndromes, including
metabolic syndrome, prostate cancer, Chron´s disease, Non-Alcoholic
Steatohepatitis (NASH), and COVID-19, among others.
Due to the abovementioned recruitment characteristics, the cohort is geographi-
cally and culturally constrained, but it has been cross validated by comparing it to
Prospective Metabolomic Studies in Precision Medicine: The AKRIBEA Project 283

Fig. 2 Flowchart of the AKRIBEA initiative in the context of precision medicine

smaller dataset of samples from other unrelated cohorts measured under the same
SOPs. There are two main applications of this large dataset (Fig. 2): (1) integral
analysis of the dataset to characterize the cohort (i.e. environmental factors and
habits, disease biomarkers for prevalent disease such as metabolic syndrome, density
of rare diseases, big data and artificial intelligence analyses oriented to medicine and
drug discovery), and (2) to compare the reference dataset to samples from smaller
cohorts of patients from the pathology under study. Such cohorts must have been
previously validated using an independent gold-standard methodology. In this
chapter, we will use the AKRIBEA project as a case study to do an in-depth
description of the required steps needed in precision medicine studies with large
cohorts. Moreover, we will present illustrative examples for the two abovementioned
exploration strategies: metabolic syndrome and COVID-19.

2 Large-Cohort Sample Collection and Biobanking

for Metabolomic Studies

When considering large cohorts of subjects, it is essential to ensure collaboration

between the different involved entities, the biobank coordination, and the implemen-
tation of a comprehensive, yet rigorous, methodology. In this context, the
AKRIBEA project is a long-standing collaborative effort between CIC bioGUNE,
the Basque Biobank, and Osarten Kooperatiba Elkartea, the prevention service of the
284 M. Bizkarguenaga et al.

Mondragón Cooperative, whose mission is to promote the safety and health of the
workers of the associated companies.
In projects involving human samples, ethical and government permits are
required before starting the recollection of biofluids. It is important to plan this
step quite in advance due to the considerable associated bureaucracy. For the
AKRIBEA project, the approval was granted by the Euskadi Drug Research Ethics
Committee (CEIm-E), an independent and multidisciplinary agency whose main
purpose is to ensure the protection of the rights, safety, and well-being of people who
participate in a biomedical research project, guaranteeing the correct application of
ethical, methodological, and legal principles. At this stage, it is important to properly
specify the intended use of the data in the informed consent and whether the data will
be used in a coded or anonymized manner. It is convenient to circulate quite in
advance a summary of the project among the potential donors. This document should
describe the goals of the project and be redacted in an understandable way (i.e. an
easy-to-read leaflet, distributed with the internal correspondence of the Corporation).
Finally, a database must be created to correctly link the samples with the metadata
(De Souza and Greenspan 2013). This is of particular importance in longitudinal
studies.
Along AKRIBEA project, samples are collected using homogeneous disposable
material and under the same conditions for all the involved patients: fasting
conditions for serum and first urine in the morning. Extracted blood is conserved
in the form of serum or plasma. Serum is richer in metabolites concentration, but at
the expense of higher variability associated to the clotting procedure. Thus,
pre-analytical SOPs (Vignoli et al. 2019) become critical as they properly define
the clotting time and temperature, among other variables. In the case of plasma
extraction, the use of anticoagulant is needed but this can lead also to the presence of
contaminants in the samples (González-Domínguez et al. 2020) and, for example,
EDTA becomes largely observable in the NMR spectra. In turn, urine samples are
centrifuged to eliminate cell residuals and maintained at low temperature (4°C) for
the shortest time possible until freezing to avoid bacterial contamination and to
guarantee the stability of the metabolites present in this biofluid.
Automatization is key to minimize variability. In this line, the German National
Cohort health study (https://www.dkfz.de/en/epidemiologie-krebserkrankungen/
units/NAKO_Studienzentrum_eng/NAKO_engl.html), that will follow the long-
term medical histories of 200,000 participants over 25–30 years, has been equipped
with pipetting robots and all the aliquots are carried out automatically. In the
AKRIBEA project, the efficiency of the personnel associated to the biobank has
been instrumental to fulfil this goal.
In principle, biobanks can collect and store all kinds of human samples for
analysis on a large scale. Samples must be collected and stored in a way that
makes them suitable for metabolomic analysis, typically frozen at -80°C. Previous
studies have demonstrated that negligible changes can be observed after long-term
stored samples (Yin et al. 2015; Loo et al. 2020). Yet, a crucial point is to ensure
dividing specimens into aliquots adjusting the final volumes and number of aliquots
from every sample to the requirements of the project. This precaution will minimize
Prospective Metabolomic Studies in Precision Medicine: The AKRIBEA Project 285

the number of freezing/thawing cycles experienced by the sample, which has drastic
adverse effects in the metabolic integrity of the biofluids (Lodge et al. 2021a).

3 Standard Operating Procedures for Metabolomics

in Precision Medicine

Standardized experiments use extensively validated NMR pulse programs and

acquisition parameters to enable data exchange worldwide in order to undertake
large-scale studies. Many SOPs have been specially designed to guarantee time and
interlaboratory reproducibility (Wishart 2019; Vignoli et al. 2019; Emwas et al.
2019; Ghini et al. 2019; Monsonis Centelles et al. 2017), critically compared in an
excellent recent review (Vignoli et al. 2019).
One valid strategy to ensure reproducibility and data transferability is to adopt
analytical in vitro diagnostic research (IVDr) SOPs which, in NMR metabolomics
studies imply the daily calibration of the spectrometer for temperature (MeOD
sample), stability, water suppression and magnetic field homogeneity (2mM sucrose
sample), and quantification performance (QuantRef routine) (Bruzzone et al. 2020a),
prior to the sample measurement.
The SOPs also normally adapt to the given biological matrix. Urine samples are
defrosted at room temperature for 30 min followed by centrifugation (at 4°C, for
5 min at 6,000 rpm). The supernatant is mixed with a specific buffer (1:10 buffer-to-
sample ratio) to ensure pH stability and transferred into a 5 mm NMR tube for the
measurement (Bruzzone et al. 2020b). For this biofluid, two experiments are
acquired at 300 K: a one-dimensional 1H spectrum with water presaturation for
metabolite quantification (noesypr1d) and a two-dimensional (2D) J-resolved 1H
spectrum ( jresgpprqf). In turn, serum samples are defrosted, mixed with phosphate
buffer (1:1 ratio) and transferred into a 5 mm NMR tube for the measurement, where
three different experiments are acquired at 310K: a Carr–Purcell–Meiboom–Gill
(CPMG) spin-echo experiment (cpmgpr) and the noesypr1d and jresgpprqf
experiments (Dona et al. 2014).
Chosen routine experiments are always time-optimized to enable high-throughput
acquisition. However, it is very convenient to sparsely interleave 2D 1H,1H- TOCSY
(Total Correlation SpectroscopY) for a subset of samples, for metabolite identifica-
tion purposes.
Standard processing of spectra is followed by a quality control analysis that
includes the monitorization of several parameters (i.e. linewidth, water suppression,
proper referencing) and a comparison to a reference sample (obtained from a pool),
which is measured by duplicate daily. In general, it may be necessary to perform a
normalization to correct potential bias from several sources: batch effects, different
machines, experimental conditions, etc. Typically, this is done either by division by
total spectrum intensity or by a quantitative normalization with respect to a reference
compound/signal (i.e. creatinine or DSS).
286 M. Bizkarguenaga et al.

4 Artificial Intelligence and Data Mining: Data Is the Way

4.1 Data Management

In order to be applicable to metabolomic studies, the NMR spectra must be unam-

biguously linked to the relevant metainformation. This is typically done using
unique identifiers (ID), which can be a single attribute or a combination of them.
Yet, it is necessary to design a relational system with enough flexibility to deal with
the multiplicity problem.
Since the focus is on precision medicine, the data focuses on people (donors or
patients). An internal ID is normally associated to each donor, which also contains
the invariant information (i.e. gender, ethnic, etc.). On the other hand, samples of
different type (i.e.: urine, serum) and/or collected at different moments are associated
to their own IDs, which is associated to the specific person ID via a timestamp
(collection time, Fig. 3). This time stamp also includes anthropometric information
(weight, height, etc.), questionaries, medical records, and all the time-dependent
metadata. Finally, data must be curated for outliers (i.e. incompatible types, mixed
units, typos, etc.).
In a similar way, NMR data also requires hierarchical organization and integra-
tion, as shown in Fig. 4. NMR data (the fid) can be processed in different ways
(Processing types) that shall be stored independently, when necessary. In the
AKRIBEA project, processed data goes through several quantification algorithms
(i.e. one for small metabolites and another for blood lipoproteins) to yield a list of
quantified metabolites with their values and associated units.

Fig. 3 Overview of main data entities and their relations

Prospective Metabolomic Studies in Precision Medicine: The AKRIBEA Project 287

Fig. 4 Flowchart for NMR data acquisition and processing

288 M. Bizkarguenaga et al.

4.2 Data Analysis

Data analysis may include unsupervised analysis (i.e. segmentation of the database
into two or more non-previously defined subgroups which can be mutually com-
pared) or supervised analysis (the segmentation groups have been previously defined
according to metadata). The type of analysis may, respectively, help to respond to
the questions whether there are metabolic differences in given classes or whether it is
possible to identify subgroups of different metabolic profiles. Such analyses can be
performed on each variable independently (univariate) or using two or more
variables together (multivariate).
Univariate analysis involves the unidimensional analysis of a single variable,
which usually results in straightforward biological interpretation. The analysis
includes effect size (i.e. directional variation of the metabolite concentration) and
the statistical significance of the variation, normally expressed as p-value. For the
sake of comparison, effect size is normally expressed in fold-change as binary
logarithm units. In univariate analysis, it is important to take precautions to control
Type I error (i.e. the false positive rate) in multiple testing and to properly account
for the confounding factors in the analysis, when necessary.
Multivariate analysis involves the concurrent consideration of several variables.
In here, the distinction between unsupervised and supervised analyses becomes even
stronger. The unsupervised multivariate analysis is often carried out by principal
component analysis (PCA), a method for dimensionality reduction that comprises
the maximum data variability in their first components. PCA also allows the visual
representation of multidimensional data, helps to identify patterns and that can be
combined with clustering algorithms (hierarchical or not) to discover groups of
individuals with similar characteristics. In turn, multivariate supervised analyses
identify the group or a specific value for each individual/sample (dependent variable)
and fit the input (independent) variables to a model that explains/predicts the
dependent variable. In metabolomics, it is common to use linear regression models
or PLS (Partial Least Squares) because of their explanatory capacity through their
coefficients. Instead, for classification problems, PLS Discriminant Analysis
(PLS-DA) or orthogonal PLS-DA (OPLS-DA) is widely used (Verpoorte, “Natural
products drug discovery: on silica or in-silico?”; Wishart, “Practical Aspects of
NMR-Based Metabolomics”). Other machine learning methods such as support
vector machines (SVM), random forest (RF), or neural networks are also
commonly used.
In supervised techniques it is important to pay attention to method performance
evaluation and tuning. Building a model is an iterative process where parameters
shall be adjusted, and results must be evaluated. Sensitivity, specificity, accuracy,
Area Under the Receiver Operating Characteristic curve (AUROC), coefficient of
determination (R2), and Root Mean Square Error (RMSE) are among the perfor-
mance metrics that can be used alone or in combination.
Prospective Metabolomic Studies in Precision Medicine: The AKRIBEA Project 289

4.3 Biological Contextualization

The final goal for prospective any metabolomic study with human cohorts is to try to
adequately contextualize the observed metabolic changes within the underlying
biochemical pathways. Biological interpretation of metabolic changes is challenging
and, consequently, largely depends on knowledge and experience. Yet, nowadays,
there are databases and tools that help to support this task. They are mostly based on
network analysis, where the main representative is metabolic pathway analysis.
These computer programs use the identified metabolites (from metabolomics or
indirectly from other omics) to pinpoint and analyse the involved pathways at
both, the topological and the enrichment level. These analyses pave the way to
formulate new hypothesis based on the experimental data interpretation, which can
be further contrasted by searching for additional information in available online
databases or by designing validation experiments.

4.4 From the Laboratory to the Clinic

Of course, the final goal of any putative biomarker is to end up validated and being
used by clinicians for the diagnostic and/or prognostic of diseases. In general, for
application in the clinic, the results should be validated at multiple levels, including
analytical validation and using samples from many different sources. The validation
method may involve conducting large corroboratory studies by directly testing each
candidate biomarker with an independently collected new set of samples from the
target population that will comprise a minimum of three groups (control, targeted
disease group, and targeted disease group with comorbidities). Since this study
includes significantly larger set of samples, the applied analytical techniques must
have high-throughput capacity as, for instance, NMR can provide. Finally, it is worth
noting that validated biomarkers are not directly used as direct diagnostic methods,
but they can act as diagnostic tools to round-up the diagnostic.
An important technical aspect is that for a validated biomarker the methodology
used for its determination has to fulfil the In Vitro Diagnostic (IVD) and/or the In
Vitro Medical-devices Regulation (IVDR), a new harmonized regulatory framework
to ensure the safety and performance of in vitro diagnostic medical devices on the
European market. This implementation is technically complex and often requires to
obtain several international and national certifications. Such regulatory standards
will undoubtedly modify the laboratory operativity, accounting for many aspects
such as the identification of the persons responsible for regulatory compliance, the
implementation of unique device identification, and post-market surveillance, if that
is finally the case. The implementation of this procedures often requires a risk
analysis (typically a SWOT analysis, Strengths, Weaknesses, Threats, and
Opportunities) for all processes that allows anticipating and minimizing the proba-
bility of failures and/or errors in the procedures. Moreover, the laboratory has to
comply with some regulations such as ISO 31000 Risk Management standard and
ISO 14971: Risk Management System for Medical Devices.
290 M. Bizkarguenaga et al.

Unfortunately, the elaborated procedures and the difficult access to large cohort
studies have prevented an avenue of validated biomarkers. In fact, in the past years,
150,000 studies have been conducted on biomarker, but only 100 biomarkers have
been validated and implemented for clinical use.

5 Applications of Precision Medicine: Non-oriented

and Oriented Metabolomics Analysis

5.1 Non-oriented Metabolomics Study: Metabolic Syndrome

The project “Metabotype of the active labour population of the Basque Country
through the metabolomic analysis of serum and urine by NMR” recruited serum and
urine samples from 10,000 subjects belonging to the working population of the
Basque Country. This sample repository was scrutinized to generate a metabolomic
(and to a less extent also a genomic) profiling for this community, that could set up
the basis for studying different pathologies impacting in the metabolism. In here, we
describe the application of the metabolomic analysis to investigate the Metabolic
Syndrome (MetS).
MetS is a group of metabolic abnormalities whose existence is considered a
predominant risk factor for the development of cardiovascular diseases (Day
2007). Currently, it is of great importance due to its high prevalence rate, since
35% of the adult population and 50% of the elderly population have MetS. Although
the pathogenesis of MetS is not fully understood, it is generally diagnosed after the
existence of 3 or more risk factors in an individual, altered glucose metabolism,
obesity, dyslipidaemia, and/or hypertension (Neuhauser 2005). Yet, this definition is
not adequate and is not supported at the molecular level by validated biomarkers.
Many studies have metabolically investigated MetS or its participating risk factors
(Reddy et al. 2018; Blouin et al. 2005), including studies based on NMR
metabolomics (Monnerie et al. 2020). For instance, Würtz et al. (2012) analysed
more than 7,000 serum samples from apparently healthy young adults using high-
throughput NMR to investigate diabetes prevalence. The metabolic signatures of
insulin resistance were modulated by obesity (Würtz et al. 2012) and found different
for men and women.
In a recent study (Bruzzone et al. 2021), we have investigated MetS using
NMR-based metabolomics of urine samples. The goal of the project was to investi-
gate the natural history of the syndrome, evaluating the metabolic fingerprint at each
step of the progression between apparently healthy individuals (at least devoid of
any risk factor related to MetS) up to subjects that suffer all the MetS-associated risk
factors. Indeed, the study was designed not only to investigate the contribution of
each of the previously risk factors that can be involved in the development of MetS
independently, but also to evaluate all their possible combinations between the
presence or absence of them. Four different risk factors and their possible interme-
diate conditions were considered: obesity, dyslipidaemia, hypertension, and diabetes
and all the possible intermediate conditions (Table 1). The reference AKRIBEA
Prospective Metabolomic Studies in Precision Medicine: The AKRIBEA Project 291

Table 1 MetS conditions, definitions, and number of individuals allocated to each condition in the
metabolic syndrome study (Bruzzone et al. 2021)
Total Female Male
MetS condition N = 10,792 N = 4,351 N = 6,441
0000 Apparently healthy 6,925 2,935 3,990
(64.17%) (67.46%) (61.95%)
0001 Hypertension 692 (6.41%) 276 416
(6.34%) (6.46%)
0010 Dyslipidaemia 733 (6.79%) 120 613
(2.76%) (9.52%)
0011 Obesity 170 (1.58%) 53 (1.22%) 117
(1.82%)
0100 Diabetes 504 (4.67%) 232 272
(5.33%) (4.22%)
0101 Dyslipidaemia + hypertension 310 (2.87%) 169 141
(3.88%) (2.19%)
0110 Obesity + hypertension 170 (1.58%) 37 (0.85%) 133
(2.06%)
0111 Obesity + dyslipidaemia 148 (1.37%) 62 (1.42%) 86 (1.34%)
1000 Diabetes + hypertension 282 (2.61%) 89 (2.05%) 193
(3.00%)
1001 Diabetes + dyslipidaemia 188 (1.74%) 83 (1.91%) 105
(1.63%)
1010 Diabetes + obesity 84 (0.78%) 18 (0.41%) 66 (1.02%)
1011 Obesity + dyslipidaemia + hypertension 90 (0.83%) 32 (0.74%) 58 (0.90%)
1100 Diabetes + dyslipidaemia + 92 (0.85%) 44 (1.01%) 48 (0.75%)
hypertension
1101 Diabetes + obesity + hypertension 202 (1.87%) 111 91 (1.41%)
(2.55%)
1110 Diabetes + obesity + dyslipidaemia 62 (0.57%) 17 (0.39%) 45 (0.70%)
1111 Diabetes + obesity + dyslipidaemia + 140 (1.30%) 73 (1.68%) 67 (1.04%)
hypertension
Legend: Diabetes as fasting plasma glucose >100 mg/dL, previously diagnosed type 2, impaired
fasting glucose, impaired glucose tolerance, or insulin resistance taking medication for
hyperglycaemia; obesity as BMI > 30 kg/m2; dyslipidaemia as triglycerides >150 mg/dL HDL
Cholesterol <34.75 mg/dL in men or <38.61 in women, previously diagnosed
hypercholesterolaemia, hyperlipidaemia or hypertriglyceridaemia taking medication for
dyslipidaemia; hypertension as blood pressure ≥140/90 mmHg, previously diagnosed hypertension
taking medication for hypertension

cohort was complemented with three additional cohorts of individuals specially

enriched in risk factors associated with MetS to ensure high statistical power for
each intermediate condition. Thus, the less populated condition still remained with
up to 62 associated samples (Table 1).
According to PCA analysis, NMR spectroscopy of urine can efficiently discrimi-
nate all the intermediate conditions, with a stronger power for diabetes and hyper-
tension as compared to obesity and dyslipidaemia. Notably, the responsible
metabolites for this statistical separation gradually adapt their concentrations when
292 M. Bizkarguenaga et al.

Fig. 5 Spearman correlation distances to the healthy condition for all 4 conditions. Colours
represent the distance to the apparently healthy (0000) condition and the lines connect adjacent
conditions. MetS definition according to WHO, European Group for the Study of Insulin (EGIR),
and American Association of Clinical Endocrinology (AACE) is represented by squares and
triangles; definition from National Cholesterol Education Program-Third Adult Treatment Panel
(NCEP:ATPIII) is represented by squares, triangles, and rhombus; definition by International
Diabetes Federation (IDF) is represented by squares and rhombus

moving from an apparently healthy condition (0000 according to our notation)

towards MetS patients (i.e. 1111, 1011, etc.) (Bruzzone et al. 2021). This result
strongly suggests that MetS is a continuous metabolic dysregulation, and a putative
definition deducted from the metabolic analysis would clearly identify the
conditions: 1001, 1011, 1101, and 1111 as MetS (Fig. 5). This definition is similar,
but not identical, to other existing definitions such as the WHO or the harmonized
definition, which are based on compatible symptomatology only.
The discriminant urine metabolic fingerprint contains metabolites associated to
all the risk factors under consideration, but it is largely dominated by the high
glucose levels, always associated to conditions proximal to MetS. Other significant
metabolites were p-cresol sulphate, and 4-hydroxyphenylpyruvic acid, also involved
in insulin resistance (Koppe et al. 2013).
Finally, the metabolic fingerprint was used to generate a model (AUROC of
0.87), able to discriminate the pool of samples (more than 13,000) according to MetS
probability.
Prospective Metabolomic Studies in Precision Medicine: The AKRIBEA Project 293

5.2 Oriented Metabolomics Study: COVID-19

COVID-19 is an infectious disease caused by severe acute respiratory syndrome

coronavirus type 2 (SARS-CoV-2) (WHO, 2022). The spreading of the viral
outbreaks around the world has been considered as a health problem associated to
the current globalization, with frequent social contacts between people from differ-
ent countries and continents, causing the virus to spread worldwide. As of August,
2021, more than 200 million cases have been reported in worldwide (WHO, 2022).
There is a plethora of symptoms associated with COVID-19 and patients can go
through different stages of the disease. The initial infectious stage often involves an
incubation period associated with mild and nonspecific symptoms, such as malaise,
fever, and a dry cough (Ayres 2020). Symptomatic patients may undergo an acute
phase, characterized by mild symptomatology: a pulmonary phase that may be
accompanied by shortness of breath or, eventually, a severe hyperinflammatory
phase that can lead to acute respiratory distress syndrome and/or heart failure
(Ayres 2020). In many cases the prognosis is favourable, but approximately 20%
of patients with COVID-19 require admission to the intensive care unit due to a life-
threatening acute respiratory syndrome (ARDS) and an extrapulmonary systemic
hyperinflammation syndrome, which could be accompanied by multiple organ
failure (Niazkar et al. 2020). A growing number of studies in this area suggest that
SARS-CoV-2 infection is not just an infection that affects the lungs but a systemic
syndrome with numerous metabolic manifestations (Bruzzone et al. 2020a; Ayres
2020; Kimhofer et al. 2020; San Juan et al. 2020; Thomas et al. 2020).
In principle, a metabolomic approach, based on the study of serum from patients
infected with SARS-CoV-2, should effectively address the problem of the great
disparity in the evolution of COVID-19 between individuals, from the most general
and complementary angle possible, similar to previous studies with other viruses.
For instance, in dengue virus, a specific metabolomic fingerprint included altered
metabolites associated with lipid metabolism, the regulation of inflammatory pro-
cesses by the signal of fatty acids and phospholipids, as well as endothelial cell
homeostasis and the vascular barrier (Voge et al. 2016). In this line, several studies
have used NMR metabolomics to investigate COVID-19 patients (Bruzzone et al.
2020a; Holmes et al. 2021; Lodge et al. 2021b; Meoni et al. 2021) (Turano, “NMR-
based metabolomics to evaluate individual response to treatment”; Rogers, “The
metabolomics of critical illness”). In our laboratory, we hypothesized that the
analysis of metabolic dysregulations measurable by NMR spectroscopy can, when
combined with other specific inflammation markers, predict the different stages of
SARS-CoV-2 infection associated to several biomarkers of the disease. To investi-
gate that hypothesis, in collaboration with the Cruces University Hospital and
Basurto University Hospital (Basque Country), we analysed 260 sera from patients
in a mostly severe stage of the development of the disease and compared them to
280 pre-COVID sera (healthy population samples from the AKRIBEA cohort, that
were collected in the two-year period prior to the COVID-19 outbreak), used as
healthy reference cohort.
294 M. Bizkarguenaga et al.

This study revealed that COVID-19 patients dysregulate ApoA-I and ApoA-II
lipoproteins, which points to an increased risk of atherosclerosis. This is in addition
to a singular profile of low-molecular-weight metabolites, which include high levels
of ketone bodies (acetoacetic acid, 3-hydroxybutyric acid, and acetone) and
2-hydroxybutyric acid, related to liver damage, dyslipidaemia, and oxidative stress
(Bruzzone et al. 2020a).
A follow-up study, in collaboration with the group of Jeremy Nicholson (ANPC,
Murdoch University, Australia), reinforces the idea that COVID-19 disease
dysregulates amino acids, biogenic amines, and tryptophan pathway, a bona fide
biomolecular marker of the disease (Kimhofer et al. 2020). Also, in the acute phase
of the disease the systemic inflammation markers appear to be elevated. This is in
line with other COVID-19 studies (Kimhofer et al. 2020; Thomas et al. 2020), which
show decreased counts of T-helper cells (regulators of T lymphocytes) and elevated
inflammatory cytokines and biomarkers such as IL-2, IL-6, IL-7, macrophage
inflammatory protein 1-a, tumour necrosis factor a, C-reactive protein, ferritin, and
D-dimer.
The natural history of COVID-19 disease reveals that some people infected by
SARS-CoV-2 do not return to a normal/healthy state, establishing a new deteriorated
health-baseline (Ayres 2020). Long-term COVID-19 or post-acute COVID-19 syn-
drome patients may retain some symptoms and/or have developed unprecedented
adverse effects (Holmes et al. 2021; Nalbandian et al. 2021; Yong 2021). Known
manifestations include a wide variety of physical symptoms, such as severe fatigue,
malaise, headaches, and other more worrisome phenomena such as increased risk of
damage to the heart, lungs, and brain (López-León et al. 2021; Davis et al. 2021). It
is estimated that around 1 out of 10 COVID-19 patients will undergo symptoms up
to 12 weeks after having suffered the disease, according to a new report from the
European Observatory of Health Systems and Policies promoted by the World
Health Organization (WHO).
An important open question is whether post-acute sequela of COVID-19 are
accompanied by incomplete metabolic phenoreversion. Metabolic alterations on a
cohort of 27 recovered, non-hospitalized patients have been recently analysed using
a combination of LC-MS spectrometry and NMR spectroscopy to show that a subset
of these patients still display a biochemical pathophysiology (Holmes et al. 2021).
Consistently, our lab studied a cohort of 97 post-COVID-19 recovered patients,
6 months after the infection on average, and compared them to a matched reference
set (n = 87), created ad hoc from the AKRIBEA cohort. Our results show a bimodal
distribution, with half the patient's population still experiencing a significant meta-
bolic and lipidomic dysregulation after 6 months of the infection. Finally, a recent
study analysed blood samples of PCR-/controls, PCR+/not-hospitalized, PCR+/
hospitalized, and PCR+/intubated patients by mass spectrometry to show that
kynurenine/tryptophan ratio, C26:0-lysoPC, and pyruvic acid can discriminate
non-COVID controls from PCR+/not-hospitalized patients (López-Hernández
et al. 2021). Moreover, C10:2-lysoPC, butyric acid, and pyruvic acid could distin-
guish between PCR+/ not-hospitalized and PCR+/hospitalized and PCR+/intubated.
Altogether, these studies highlight the potential power of metabolomics for the
Prospective Metabolomic Studies in Precision Medicine: The AKRIBEA Project 295

characterization of the pathophysiology of complex diseases such as COVID-19 and

pave the way for their integration in the clinical practice, after exploring larger
validation cohorts of patients.

6 Concluding Remarks

The optimization of the analytical techniques combined with the availability of

large-scale logistics under strict SOPs enables the implementation of precision
medicine into the clinical practice. In this not-so-futuristic scenario, NMR
metabolomics may play a key role thanks to its analytical power, very large
reproducibility, and the realistic possibility of deploying magnets in hospitals and
surgery rooms. To that end, large spectral databases combined with AI techniques
will become instrumental to validate the diagnostic value that metabolomics can
potentially offer. Moreover, precision medicine could be applied in the clinic with
this metabolism analysis, which is constantly changing, allowing the clinician to
prescribe the right treatment and the specific dose of it at the right time of the disease
state.

References
Authors (2021) Ultra-precision medicine. Nat Biotechnol 39:645. https://doi.org/10.1038/s41587-
021-00967-8
Ayres JS (2020) A metabolic handbook for the COVID-19 pandemic. Nat Metab 2:572–585
Beger RD et al (2016) Metabolomics enables precision medicine: “A White Paper, Community
Perspective”. Metabolomics 12
Bernardo-Seisdedos G et al (2021) Metabolic landscape of the mouse liver by quantitative 31P
nuclear magnetic resonance analysis of the phosphorome. Hepatology 74:148–163
Blouin K et al (2005) Contribution of age and declining androgen levels to features of the metabolic
syndrome in men. Metabolism 54:1034–1040
Bruzzone C et al (2020a) SARS-CoV-2 infection dysregulates the metabolomic and lipidomic
profiles of serum. iScience 23
Bruzzone C et al (2020b) 1H NMR-based urine metabolomics reveals signs of enhanced carbon and
nitrogen recycling in prostate cancer. J Proteome Res 19:2419–2428
Bruzzone C et al (2021) A molecular signature for the metabolic syndrome by urine metabolomics.
Cardiovasc Diabetol 20:155
Carrat F et al (2021) Evidence of early circulation of SARS-CoV-2 in France: findings from the
population-based “CONSTANCES” cohort. Eur J Epidemiol 36:219–222
Christensen CM, Grossman JH, Hwang J (2009) The innovator’s prescription: a disruptive solution
for health care. McGraw-Hill
Clish CB (2015) Metabolomics: an emerging but powerful tool for precision medicine. Mol Case
Stud 1:a000588
Davis HE et al (2021) Characterizing Long COVID in an international cohort: 7 months of
symptoms and their impact. SSRN Electron J. https://doi.org/10.2139/ssrn.3820561
Day C (2007) Metabolic syndrome, or what you will: definitions and epidemiology. Diab Vasc Dis
Res 4:32–38
De Souza YG, Greenspan JS (2013) Biobanking past, present and future: responsibilities and
benefits. AIDS 27:303–312
296 M. Bizkarguenaga et al.

Dona AC et al (2014) Precision high-throughput proton NMR spectroscopy of human urine, serum,
and plasma for large-scale metabolic phenotyping. Anal Chem 86:9887–9894
Embade N et al (2019) NMR-based newborn urine screening for optimized detection of inherited
errors of metabolism. Sci Rep 9:1–9
Emwas AH et al (2019) NMR spectroscopy for metabolomics research. Metabolites 9
Ghini V, Quaglio D, Luchinat C, Turano P (2019) NMR for sample quality assessment in
metabolomics. N Biotechnol 52:25–34
González-Domínguez R, González-Domínguez Á, Sayago A, Fernández-Recamales Á (2020)
Recommendations and best practices for standardizing the pre-analytical processing of blood
and urine samples in metabolomics. Metabolites 10:1–18
Hartl D et al (2021) Translational precision medicine: an industry perspective. J Transl Med 19:1–
14
Holmes E, Wist J, Masuda R, Lodge S, Nitschke P, Kimhofer T, Loo RL, Begum S, Boughton B,
Yang R, Morillon AC, Chin ST, Hall D, Ryan M, Bong SH, Gay M, Hartmut Lawler NG,
Gray N, Whiley L, Nicholson J (2021) Incomplete systemic recovery and metabolic
phenoreversion in post-acute-phase nonhospitalized COVID-19 patients: implications for
assessment of post-acute COVID-19 syndrome. J Proteome Res
Jiménez B et al (2018) Quantitative lipoprotein subclass and low molecular weight metabolite
analysis in human serum and plasma by (1)H NMR spectroscopy in a multilaboratory trial. Anal
Chem 90:11962–11971
John A, Qin B, Kalari KR, Wang L, Yu J (2020) Patient-specific multi-omics models and the
application in personalized combination therapy. Future Oncol 16:1737–1750
Kimhofer T et al (2020) Integrative modeling of quantitative plasma lipoprotein, metabolic, and
amino acid data reveals a multiorgan pathological signature of SARS-CoV-2. J Proteome Res
Kohler I, Hankemeier T, van der Graaf PH, Knibbe CAJ, van Hasselt JGC (2017) Integrating
clinical metabolomics-based biomarker discovery and clinical pharmacology to enable precision
medicine. Eur J Pharm Sci 109:S15–S21
Koppe L et al (2013) p-Cresyl sulfate promotes insulin resistance associated with CKD. J Am Soc
Nephrol 24:88–99
Koshiba S et al (2018) Omics research project on prospective cohort studies from the Tohoku
Medical Megabank Project. Genes Cells 23:406–417
Lodge S et al (2021a) Low volume in vitro diagnostic proton NMR spectroscopy of human blood
plasma for lipoprotein and metabolite analysis: application to SARS-CoV-2 biomarkers. J
Proteome Res 20:1415–1423
Lodge S et al (2021b) NMR spectroscopic windows on the systemic effects of SARS-CoV-
2 infection on plasma lipoproteins and metabolites in relation to circulating cytokines. J
Proteome Res 20:1382–1396
Loeffler M et al (2015) The LIFE-adult-study: objectives and design of a population-based cohort
study with 10,000 deeply phenotyped adults in Germany. BMC Public Health 15:1–14
Loo RL et al (2020) Quantitative in-vitro diagnostic NMR spectroscopy for lipoprotein and
metabolite measurements in plasma and serum: recommendations for analytical artifact minimi-
zation with special reference to COVID-19/SARS-CoV-2 samples. J Proteome Res 19:4428–
4441
López-Hernández Y et al (2021) Targeted metabolomics identifies high performing diagnostic and
prognostic biomarkers for COVID-19. Sci Rep 11:14732
López-León S et al (2021) More than 50 long-term effects of COVID-19: a systematic review and
meta-analysis. SSRN Electron J 1–22. https://doi.org/10.2139/ssrn.3769978
Martin SS et al (2015) HDL cholesterol subclasses, myocardial infarction, and mortality in
secondary prevention: the Lipoprotein Investigators Collaborative. Eur Heart J 36:22–30
Meoni G et al (2021) Metabolomic/lipidomic profiling of COVID-19 and individual response to
tocilizumab. PLoS Pathog 17
Mignot S et al (2021) Correlates of premature pap test screening, under 25 years old: analysis of
data from the CONSTANCES cohort study. BMC Public Health 21:1–9
Prospective Metabolomic Studies in Precision Medicine: The AKRIBEA Project 297

Monnerie S et al (2020) Metabolomic and lipidomic signatures of metabolic syndrome and its
physiological components in adults: a systematic review. Sci Rep 10:1–13
Monsonis Centelles S et al (2017) Toward reliable lipoprotein particle predictions from NMR
spectra of human blood: an interlaboratory ring test. Anal Chem 89:8004–8012
Nabi O et al (2021) Comorbidities are associated with fibrosis in NAFLD subjects: a nationwide
study (NASH-CO study). Dig Dis Sci. https://doi.org/10.1007/s10620-021-07032-z
Nalbandian A et al (2021) Post-acute COVID-19 syndrome. Nat Med 27:601–615
Neuhauser HK (2005) The metabolic syndrome. 366
Niazkar HR, Zibaee B, Nasimi A, Bahri N (2020) The neurological manifestations of COVID-19.
Prat Neurol FMC 11:145–146
Ortega-Azorín C et al (2019) Candidate gene and genome-wide association studies for circulating
leptin levels reveal population and sex-specific associations in high cardiovascular risk Medi-
terranean subjects. Nutrients 11:2751
Otvos JD, Jeyarajah EJ, Bennett DW, Krauss RM (1992) Development of a proton nuclear
magnetic resonance spectroscopic method for determining plasma lipoprotein concentrations
and subspecies distributions from a single, rapid measurement. Clin Chem 38:1632–1638
Puchades-Carrasco L, Pineda-Lucena A (2017) Metabolomics applications in precision medicine:
an oncological perspective. Curr Top Med Chem 17:2740–2751
Reddy P, Leong J, Jialal I (2018) Amino acid levels in nascent metabolic syndrome: a contributor to
the pro-inflammatory burden. J Diabetes Complications 32:465–469
Ruiz F et al (2016) High quality standards for a large-scale prospective population-based observa-
tional cohort: constances. BMC Public Health 16:1–10
San Juan I et al (2020) Abnormal concentration of porphyrins in serum from COVID-19 patients. Br
J Haematol 190:e265–e267
Śliwczyński A, Orlewska E (2016) Precision medicine for managing chronic diseases. Pol Arch
Med Wewn 126:681–687
Thomas T et al (2020) COVID-19 infection alters kynurenine and fatty acid metabolism, correlating
with IL-6 levels and renal status. JCI Insight 5
Tigchelaar EF et al (2015) Cohort profile: LifeLines DEEP, a prospective, general population cohort
study in the northern Netherlands: Study design and baseline characteristics. BMJ Open 5:1–10
Tranvåg EJ, Strand R, Ottersen T, Norheim OF (2021) Precision medicine and the principle of equal
treatment: a conjoint analysis. BMC Med Ethics 22:1–9
Vignoli A et al (2019) High-throughput metabolomics by 1D NMR. Angew Chem Int Ed 58:968–
994
Voge NV et al (2016) Metabolomics-based discovery of small molecule biomarkers in serum
associated with dengue virus infections and disease outcomes. PLoS Negl Trop Dis 10:1–27
WHO (2022). https://www.who.int/health-topics/coronavirus
WHO (2022). https://covid19.who.int/
Wilson D, Burlingame A, Cronholm T, Sjövall J (1974) Deuterium and carbon-13 tracer studies of
ethanol metabolism in the rat by 2H, 1H-decoupled 13C nuclear magnetic resonance. Biochem
Biophys Res Commun 56:828–835
Wishart DS (2019) NMR metabolomics: a look ahead. J Magn Reson 306:155–161
Wishart DS et al (2018) HMDB 4.0: the human metabolome database for 2018. Nucleic Acids Res
46:D608–D617
Würtz P et al (2012) Metabolic signatures of insulin resistance in 7,098 young adults. Diabetes 61:
1372–1380
Yin P, Lehmann R, Xu G (2015) Effects of pre-analytical processes on blood samples used in
metabolomics studies. Anal Bioanal Chem 407:4879–4892
Yong SJ (2021) Long COVID or post-COVID-19 syndrome: putative pathophysiology, risk
factors, and treatments. Infect Dis 0:1–18
Chemotherapy-Induced Peripheral
Neuropathy

Ciao-Sin Chen and Daniel L. Hertz

Contents
1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 300
2 Review of CIPN Metabolomics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 302
2.1 Metabolomics Analyses Design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 302
2.2 CIPN Study Design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 303
2.3 Clinical CIPN Metabolomics Studies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 303
2.4 Animal and Cellular CIPN Metabolomics Studies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 306
2.5 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 308
3 Investigation of Peripheral Neuropathy Metabolomic Biomarker Pathways . . . . . . . . . . . . . . 308
3.1 Histidine . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 313
3.1.1 1-Methylhistidine . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 315
3.1.2 Carnosine . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 315
3.1.3 Ergothioneine . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 316
3.1.4 Histamine . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 316
3.1.5 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 316
3.2 Phenylalanine . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 317
3.2.1 Phenylalanine and Tyrosine . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 317
3.2.2 Phenethylamine . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 319
3.2.3 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 319
3.3 N-Acetylornithine . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 319
3.3.1 Ornithine and Arginine . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 320
3.3.2 Ornithine and Polyamines: Putrescine, Spermidine, and Spermine . . . . . . . . . . 321
3.3.3 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 321
3.4 Threonine . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 321
3.5 Glycogen . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 322
3.6 Adenosine Phosphates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 323
4 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 324
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 324

C.-S. Chen · D. L. Hertz (✉)

Department of Clinical Pharmacy, University of Michigan College of Pharmacy, Ann Arbor, MI,
USA
e-mail: [email protected]; [email protected]

# The Author(s), under exclusive license to Springer Nature Switzerland AG 2022 299
V. Ghini et al. (eds.), Metabolomics and Its Impact on Health and Diseases,
Handbook of Experimental Pharmacology 277, https://doi.org/10.1007/164_2022_609
300 C.-S. Chen and D. L. Hertz

Abstract

Chemotherapy-induced peripheral neuropathy (CIPN) is a debilitating side effect

of many common anti-cancer agents that can lead to dose reduction or treatment
discontinuation, which decrease chemotherapy efficacy. Long-term CIPN can
interfere with activities of daily living and diminish the quality of life. The
mechanism of CIPN is not yet fully understood, and biomarkers are needed to
identify patients at high risk and potential treatment targets. Metabolomics can
capture the complex behavioral and pathophysiological processes involved in
CIPN. This chapter is to review the CIPN metabolomics studies to find metabolic
pathways potentially involved in CIPN. These potential CIPN metabolites are
then investigated to determine whether there is evidence from studies of other
neuropathy etiologies such as diabetic neuropathy and Leber hereditary optic
neuropathy to support the importance of these pathways in peripheral neuropathy.
Six potential biomarkers and their putative mechanisms in peripheral neuropathy
were reviewed. Among these biomarkers, histidine and phenylalanine have clear
roles in neurotransmission or neuroinflammation in peripheral neuropathy. Fur-
ther research is needed to discover and validate CIPN metabolomics biomarkers
in large clinical studies.

Keywords

Chemotherapy · Histidine · Metabolomics · Peripheral neuropathy ·

Phenylalanine

1 Introduction

Chemotherapy-induced peripheral neuropathy (CIPN) is a debilitating side effect of

many common anti-cancer agents, including taxane, vinca alkaloids, platinum,
bortezomib, and thalidomide-related agents (Miltenburg and Boogerd 2014; Seretny
et al. 2014). The prevalence and clinical characteristics of CIPN vary by regimen
(Miltenburg and Boogerd 2014; Seretny et al. 2014) with prevalence estimates as
high as 70% (Seretny et al. 2014). CIPN is a chronic distal symmetric
polyneuropathy that predominantly presents as sensory symptoms, including numb-
ness (stocking-glove), tingling, and shooting or burning pain (Miltenburg and
Boogerd 2014; Quasthoff and Hartung 2002). Motor symptoms such as loss of
balance or dexterity caused by muscle weakness can be found in severe cases,
with prevalence up to 40% depending on the regimen (Argyriou et al. 2008; Haim
et al. 1994; Lee and Swain 2006; Miltenburg and Boogerd 2014; Plasmati et al.
2007). Autonomic symptoms are usually rare, but constipation is common (up to
90%) in patients receiving certain neurotoxic agents including vincristine,
bortezomib, and thalidomide-related agents (Miltenburg and Boogerd 2014).
Chemotherapy-Induced Peripheral Neuropathy 301

CIPN symptoms can limit instrumental activities of daily living such as preparing
meals, shopping for groceries or clothes, or using the phone. CIPN symptoms persist
in about 40% of CIPN patients for at least 2 years and are irreversible in some
patients (Bao et al. 2016; Hershman et al. 2018; Mustafa Ali et al. 2017). Irreversible
CIPN diminishes long-term quality of life (Bandos et al. 2018; Mols et al. 2014),
increases risk for falls, and causes disability in severe cases (Kolb et al. 2016;
Winters-Stone et al. 2017). These life-long consequences are particularly concerning
in patients with favorable long-term prognosis, such as those with early stage breast
cancer (McGale et al. 2014).
The mechanism of CIPN is not yet fully understood and is likely to be multifac-
torial. The neurotoxicity caused by chemotherapy may involve DNA damage,
microtubule disruption, mitochondrial damage, pain detection receptor remodeling,
oxidative stress, and neuroinflammation in the neurons of the peripheral nervous
system (Chan et al. 2019; Kerckhove et al. 2017; Zajączkowska et al. 2019). The
lack of mechanistic understanding is one reason for the lack of effective strategies for
CIPN prevention or treatment; the only guideline-recommended strategy for treat-
ment of painful CIPN is duloxetine, which has limited efficacy (Loprinzi et al. 2020;
Smith et al. 2013). Otherwise, guidelines recommend delaying, decreasing, or
discontinuing neurotoxic chemotherapy treatment in patients experiencing CIPN,
which reduces treatment efficacy (Hertz et al. 2021).
Biomarkers that could be used prior to or early in treatment to predict which
patients are most at risk for severe, and perhaps irreversible, CIPN would be
extremely clinically useful. The most established CIPN risk factors are cumulative
treatment, prior neuropathy or neurotoxic treatment, age, race, diabetes, and obesity
(Gu et al. 2021; Hershman et al. 2016; Kandula et al. 2016; Schneider et al. 2017;
Seretny et al. 2014). There may be some confounding between these factors, as age,
race, and obesity are all risk factors for diabetes, which is also a cause of peripheral
neuropathy (Bao et al. 2016; Cox-Martin et al. 2017; Gu et al. 2021; Hershman et al.
2016; Petrovchich et al. 2019; Smith and Singleton 2013; Timmins et al. 2021;
Wiggin et al. 2009). Modifiable nutritional factors such as diet and nutrient supple-
mentation have also been reported to influence CIPN risk (Mongiovi et al. 2018;
Stankovic et al. 2020; Zirpoli et al. 2017). More severe CIPN has been reported in
patients with lower grain consumption, not taking vitamins or taking antioxidants
(Mongiovi et al. 2018; Stankovic et al. 2020; Zirpoli et al. 2017), deficiency of
vitamin D or saturated fatty acids (Grim et al. 2017; Jennaro et al. 2020), or a
sedentary lifestyle (Bulls et al. 2020; Duregon et al. 2018; Greenlee et al. 2017;
Kanzawa-Lee et al. 2020; Mols et al. 2015). Again, these risk factors may be due to
shared mechanistic pathways with neuropathy caused by poor nutrition, obesity,
and/or diabetes. Finally, substantial research has been undertaken to identify
inherited genetic markers, including those that affect neurotoxic drug elimination
or pharmacology, diabetes, and neuronal cell function or survival (Chan et al. 2019;
Cliff et al. 2017; Ng et al. 2014; Sucheston-Campbell et al. 2018; Terrazzino et al.
2015), however, no genetic markers have been sufficiently validated for use in
clinical practice (Chan et al. 2019).
302 C.-S. Chen and D. L. Hertz

Biomarkers that capture the complex behavioral and pathophysiological pro-

cesses involved in CIPN are needed to identify patients with high CIPN risk or to
gain further mechanistic understanding of CIPN to identify potential treatment
targets. Metabolomics is a biomarker discovery strategy that analyzes the
metabolome, which is the comprehensive endogenous metabolite profile of
biospecimens. Metabolites are at the end of the DNA-RNA-protein central dogma
axis. By providing a direct snapshot of the current metabolic status, metabolomics
offers an opportunity to explain the inter-individual variability that cannot be
detected by DNA sequence, such as environmental and nutritional factors (Clayton
et al. 2009; O’Gorman and Brennan 2017; Shin et al. 2014). These factors are
sometimes too complicated to be extracted from medical records and are often not
fully examined or stratified (Everett 2019; Kantae et al. 2017; Rattray et al. 2018;
Wishart 2016). Metabolomics can capture the end effect of these factors in quantifi-
able metabolic signatures, which have been found to be useful in predicting toxicity
from cancer drugs (Backshall et al. 2011). In addition, compared to other omics,
such as transcriptomics and proteomics, metabolomics is closer to the clinical
phenotype and easier to quantify (Karczewski and Snyder 2018). In this chapter,
we will review CIPN metabolomics studies to find metabolic pathways potentially
involved in CIPN and then investigate whether there is evidence from studies of
other neuropathy etiologies to support these pathways. CIPN biomarker studies that
only investigated lipidomics are outside of the scope and will not be included
(Kramer et al. 2015; Maekawa et al. 2019).

2 Review of CIPN Metabolomics

2.1 Metabolomics Analyses Design

The two most common techniques in metabolomics analyses are nuclear magnetic
resonance (NMR) and mass spectrometry (MS). Each has its own advantages and
disadvantages in profile coverage, sensitivity, stability, sample volume, sample
preparation, quantification, high-throughput automation, time, and cost (Everett
2017; Wishart 2016). MS-based methods are 10–100 times more sensitive than
NMR. Whereas a typical LC-MS study can detect more than 1,000 metabolites
with concentration of 10–100 nM in human plasma, a typical NMR study can only
detect 50–200 metabolites with concentration at the microM scale. On the other
hand, only a few hundred of the metabolites detected by MS-based methods can be
identified (Emwas et al. 2019). NMR sensitivity can be improved with digital
spectrometers, cryoprobes, and low volume probes (Everett 2017). There is a trend
toward increasing use of NMR due to its greater reproducibility and the ease of full
automation (Everett 2017; Wishart 2016). Although NMR requires larger sample
volume, sample preparation is minimal (Kirwan et al. 2018; Wishart 2016), and
NMR can provide absolute quantification of metabolite concentrations in only a few
minutes without reference standards (Wishart 2016). The start-up cost of NMR
instrumentation is expensive, but the cost per sample is low (Everett 2017; Wishart
Chemotherapy-Induced Peripheral Neuropathy 303

2016). There are two approaches to metabolite identification, targeted and untargeted
metabolomics. Targeted approaches focus on a set of defined metabolites, whereas
global untargeted approaches capture all metabolites that are present in a sample but
may not provide the exact identification of some of the metabolites due to the
limitations of databases or libraries. Untargeted metabolomics is usually qualitative,
or can provide relative quantitation, and is used for hypothesis-agnostic discovery,
while targeted metabolomics can provide absolute quantitation and is often used to
confirm results from untargeted studies or confirm hypotheses (Roberts et al. 2012).
Both NMR and MS-based methods can be used in untargeted and targeted analyses,
but MS-based methods are generally superior for untargeted metabolomics discov-
ery due to high sensitivity and for targeted metabolomics validation due to its
selectivity, while NMR is used primarily in untargeted analyses (Emwas et al. 2019).

2.2 CIPN Study Design

Metabolomics studies of CIPN enroll patients with cancer who are going to receive
neurotoxic chemotherapy. CIPN data are collected from clinician or patient assess-
ment, and patients can be defined as whether or not they experienced CIPN, the
severity of their CIPN, or the cumulative dose at the time that CIPN occurred (Hertz
2019). Biosamples are collected pretreatment (baseline) and on-treatment to acquire
metabolite information. The pretreatment metabolite levels can be used to identify
metabolic signatures that predict peripheral neuropathy, whereas the on-treatment
changes from baseline can be used to find metabolic signatures that are associated
with the neurotoxic effect of the chemotherapy, which can still be predictive if they
are sufficiently early in treatment that CIPN has not yet become clinically observ-
able. Metabolic changes at the time of clinically evident CIPN are not useful for
prediction but may be useful to understand mechanisms underlying CIPN.

2.3 Clinical CIPN Metabolomics Studies

Two CIPN metabolomics studies have been conducted in patients receiving chemo-
therapy (Table 1). One untargeted whole blood NMR study by Sun et al. (2018)
enrolled 48 patients with stage I-III or oligometastatic breast cancer planning to
receive weekly paclitaxel and obtained metabolomics profiles prior to treatment
(baseline), at the end of the first paclitaxel infusion, and 24 h after the first infusion.
Paclitaxel-induced peripheral neuropathy was assessed by a patient-reported ques-
tionnaire at baseline and before each infusion, and CIPN severity was defined by the
maximum change from baseline of the sensory symptom subscale. Without strict
correction for multiple comparisons, lower pretreatment levels of three amino acids,
histidine, phenylalanine, and threonine, were moderately correlated with more
severe sensory neuropathy symptoms. Statistically uncorrected secondary
analyses identified many metabolites that changed from pretreatment to
post-infusion or 24-h later, however, none of these were associated with CIPN
Table 1 Human metabolomics studies of CIPN
304

Neurotoxic CIPN Metabolomics Sample

Subjects chemotherapy data techniques type Statistical methods Findings Reference
48 patients Paclitaxel 80 mg/m2 Max NMR, Whole Pearson correlation Lower pretreatment (Sun et al.
with breast 1-h infusion weekly for change untargeted blood, coefficient with false levels of histidine (r = - 2018)
cancer 12 weeks from 34 metabolites non-fasting discovery 0.38, p = 0.01),
baseline rate < 25% phenylalanine (r = -
of 0.34, p = 0.02), and
CIPN8 threonine (r = -0.36,
p = 0.01) were
moderately correlated
with the severity of
sensory peripheral
neuropathy during
treatment
32 pediatric Vincristine 1.5 mg/m2 Max of LC-MS/MS, Plasma, Linear support vector N-acetylornithine, (Verma
patients with (max 2 mg) during the TNS-PV untargeted, non-fasting machine with glycogen, adenosine et al.
acute induction (day 8 and >8 total number of recursive feature monophosphate, and 2020)
lymphoblastic 29) and consolidation metabolites not elimination and five- adenosine diphosphate
leukemia phases (day 180) reported fold cross validation on day 29 were identified
to be predictive
(AUC = 0.75). Arginine,
1, 7-dimethylguanosine,
gamma-glutamyl-
isoleucine, glutathione
disulfide, phenyl-alanyl-
proline, pipecolate,
5-hydroxy-tryptophol,
sphingomyelin SM (d18:
0/16: 1(9Z)), and
glycerophosphocholine
PC (16: 1(9Z)/0:0) on
C.-S. Chen and D. L. Hertz
 day 180 were identified
indicative of CIPN
development
(AUC = 0.92). No
metabolite
concentrations were
reported
CIPN8, the subscale of the first eight sensory items of European Organization for Research and Treatment of Cancer Quality of Life Questionnaire for
Chemotherapy-induced Peripheral Neuropathy 20-item scale
TNS-PV, total neuropathy score-pediatric vincristine
Chemotherapy-Induced Peripheral Neuropathy
305
306 C.-S. Chen and D. L. Hertz

severity. A subsequent analysis identified metabolites that were moderately

correlated with paclitaxel pharmacokinetics, but these statistically uncorrected
discovery-phase associations did not mechanistically explain the CIPN
metabolomics findings in the initial study (Chen et al. 2021).
The other clinical CIPN metabolomics study was an untargeted plasma LC-MS/
MS study by Verma et al. (2020) which enrolled 32 pediatric patients receiving
vincristine for precursor B-cell acute lymphoblastic leukemia. Patients with
preexisting peripheral neuropathy or taking erythropoietin, itraconazole, or vitamin
supplementation were excluded due to the possible interference with neurological
assessment. Vincristine-induced peripheral neuropathy was assessed via physician-
evaluation regularly throughout treatment, and patients were classified as peripheral
neuropathy if they experienced a high symptom score (total neuropathy score-
pediatric vincristine >8) at any time during treatment. The investigators analyzed
the metabolome at days 8, 29, and 180 and used a linear support vector machine to
build metabolomic models that predict (days 8 or 29) or indicate (day 180) the
development of CIPN. The final models included 2, 14, and 21 metabolomics
features, respectively, and all three models had excellent performance with recursive
feature elimination and five-fold cross validation. However, metabolite
concentrations were not reported, and some metabolites were not identified, for
which only mass and retention time were reported in the supplementary information.
The identified metabolites in the day 29 and 180 models can be found in Table 1.
Pathway analyses did not identify any relevant metabolic pathways after multiple
comparison correction.

2.4 Animal and Cellular CIPN Metabolomics Studies

Although clinical CIPN metabolomics studies are very scarce, animal and cellular
studies may assist with identifying metabolic pathways of interest (Table 2). Two
untargeted NMR studies by Ferrier et al. (2013, 2015) analyzed the metabolomics of
central nervous system tissue in rat models of oxaliplatin-induced neurotoxicity. A
single injection of oxaliplatin increased glutamine and decreased creatinine and
adenosine phosphate in spinal dorsal horn tissues in rats that developed acute pain
hypersensitivity compared to vehicle-treated rats, and these metabolic changes and
pain hypersensitivity were abrogated by a polyamine-deficient diet (Ferrier et al.
2013). In a similar study of multiple oxaliplatin injections, higher choline and
glycerophosphocholine and lower aspartate, glutamine, creatinine, and GABA
were found in various brain tissues in rats that developed painful neuropathy
compared to vehicle-treated rats (Ferrier et al. 2015). An untargeted LC-MS/MS
study by Wu et al. (2018) analyzed the plasma metabolomics of a rat model of
paclitaxel-induced peripheral neuropathy and found 19 lipids, including fatty acids,
ketones, and glycerophosphocholine, which differentiated paclitaxel-treated from
vehicle-treated rats. An untargeted study by Qin et al. (2012) used a different
metabolomics technique, capillary electrophoresis (CE-MS), which is increasingly
used in metabolomics due to its low sample volume requirement and high efficiency
Table 2 Animal and cellular metabolomics studies of CIPN
Neurotoxic Metabolomics Statistical
Subjects chemotherapy CIPN data techniques Sample type methods Findings Reference
Adult male rat Single Behavioral tests of NMR, Spinal dorsal Mann– Increased glutamine and (Ferrier
model of intraperitoneal mechanical allodynia, untargeted horn tissue, Whitney decreased creatinine and et al.
oxaliplatin- oxaliplatin pain hypersensitivity, 18 metabolites synthetic diet U tests adenosine phosphate 2013)
induced acute injection and cold available ad were found in
pain hypersensitivity libitum oxaliplatin-treated rats,
hypersensitivity and polyamine-deficient
diet prevented such
metabolic changes and
oxaliplatin-induced
acute pain
hypersensitivity
Chemotherapy-Induced Peripheral Neuropathy

Male rat model Intravenous Behavioral tests of NMR, Various brain Mann– Higher choline and (Ferrier
of oxaliplatin- oxaliplatin mechanical allodynia, untargeted tissues, food Whitney glycerophosphocholine et al.
induced painful injection twice a cold hypersensitivity 17 metabolites available ad U tests and lower aspartate, 2015)
neuropathy week for four libitum creatinine, glutamine,
and half weeks and GABA were found
in oxaliplatin-treated
rats
Female rat model Intraperitoneal Behavioral tests of LC-MS/MS, Plasma, food PLSDA Fatty acids, ketones, and (Wu et al.
of paclitaxel- paclitaxel mechanical allodynia untargeted, available ad and t tests glycerophosphocholine 2018)
induced injection on total number of libitum differentiated between
peripheral three alternate metabolites not paclitaxel-treated rats
neuropathy days reported and controls
Human neural Low-dose Immunofluorescence CE-MS, Cell culture Mann– Lower glutathione was (Qin et al.
progenitor cells thalidomide for imaging of neuronal untargeted Whitney found in thalidomide- 2012)
3 days differentiation 74 metabolites U tests treated cells during
neural differentiation
307

Mechanical allodynia is a painful sensation caused by innocuous stimuli such as light touch of clothes
308 C.-S. Chen and D. L. Hertz

for separating polar metabolites (Harada et al. 2018; Kristoff et al. 2020; Zhang and
Ramautar 2021). This study analyzed human neural progenitor cell cultures treated
with low-dose thalidomide and found, compared to vehicle-treated controls, lower
glutathione in thalidomide-treated cells during neural differentiation. Of the
metabolites found in these animal and cellular studies, only adenosine phosphate
was consistent with the metabolites identified in the CIPN clinical studies. This may
be due to the use of vehicle-treated controls, as opposed to treated rats that did not
develop CIPN, suggesting these results may reflect the direct effect of chemotherapy
and be unrelated to CIPN.

2.5 Summary

Two pioneering metabolomics studies have been conducted in patients with cancer
receiving neurotoxic chemotherapy to identify metabolic predictors of CIPN. These
studies suggest six metabolites with putative pathways that may be pretreatment or
early-in-treatment predictors of CIPN: histidine, phenylalanine, threonine (Sun et al.
2018), N-acetylornithine, glycogen, and adenosine mono/diphosphate (Verma et al.
2020). Except for adenosine phosphate, studies in animal models and neuronal
progenitor cells found different metabolites that are affected by chemotherapy
treatment, but their relationship to CIPN is unknown (Ferrier et al. 2013, 2015;
Qin et al. 2012; Wu et al. 2018).

3 Investigation of Peripheral Neuropathy Metabolomic

Biomarker Pathways

The rest of this chapter further investigates the putative associations between these
six potential metabolomic biomarkers of CIPN, which will be listed in descending
order of the strength of evidence (Table 3). These pathways are examined for any
evidence of their involvement in human metabolomics studies of other types of
peripheral neuropathy, including diabetic neuropathy, hereditary neuropathy, and
peripheral neuropathy caused by autoimmune disorders. It can be difficult to collect
metabolic or peripheral neuropathy information before the onset of these other
etiologies of chronic peripheral neuropathy. Therefore, metabolomics studies usu-
ally compare metabolic profiles between patients with peripheral neuropathy and
controls, who are either healthy volunteers or patients who share the underlying
disease state but are not experiencing peripheral neuropathy. Some studies also
analyzed the association between metabolite levels and the severity of peripheral
neuropathy. Due to the large number of metabolites and the complex and unclear
mechanisms underlying peripheral neuropathy, many studies also adjust for clinical
covariates to try to identify the most representative metabolites.
In addition to reviewing metabolomics studies of other etiologies of peripheral
neuropathy, for each of these six potential metabolomic biomarkers of CIPN, the
putative mechanistic relationship with neuropathy is reviewed and any observational
Table 3 Evidence of association for each of the putative CIPN metabolomics biomarkers
Other
CIPN Other PN Other observational interventional Putative
Metabolites Description metabolomics metabolomics studies studies mechanisms
Histidine Essential from food Lower pretreatment Higher plasma Nominally higher Carnosine Histidine,
Precursor of plasma histidine histidine associated blood supplementation carnosine, and
histamine, which is associated with with diabetic ergothioneine in prevents oxaliplatin ergothioneine have
a neurotransmitter more severe neuropathy (Lin patients with less CIPN (Yehia et al. anti-oxidant and
Metabolites such as sensory paclitaxel- et al. 2019) sensory CIPN 2019) anti-inflammatory
carnosine and induced peripheral Lower plasma (Winkels et al. Ergothioneine properties
ergothioneine have neuropathy (Sun histidine correlated 2020) prevents platinum Ergothioneine
also been found to et al. 2018) with worse (Jong et al. 2011; inhibits oxaliplatin
be associated with cardiovascular Nishida et al. 2018; uptake in neurons
PN autonomic Song et al. 2010) Histamine
neuropathy CIPN in animals inhibitors reduce
(Mathew et al. Histamine pain detection in
Chemotherapy-Induced Peripheral Neuropathy

2019) inhibitors prevent sensory neurons

Lower skin taxane (Gao et al. 1-Methylhistidine
fibroblast histidine 2016; Nagano et al. via unknown
levels associated 2012), oxaliplatin mechanism
with Leber (Azevedo et al.
hereditary optic 2013), or
neuropathy (Chao bortezomib
de la Barca et al. (Tsukaguchi et al.
2016) 2013) CIPN pain in
Higher plasma patients or animals
1-methylhistidine
levels associated
with Leber
hereditary optic
neuropathy (Bocca
et al. 2021)
309

(continued)
Table 3 (continued)
310

Other
CIPN Other PN Other observational interventional Putative
Metabolites Description metabolomics metabolomics studies studies mechanisms
Phenylalanine Essential from food Lower pretreatment Lower skin Dietary Levodopa relieves Deficiencies of
Precursor of plasma fibroblast phenylalanine diabetic phenylalanine and
tyrosine and phenylalanine phenylalanine and restriction neuropathic pain tyrosine cause
catecholamine associated with tyrosine levels associated with (Ertas et al. 1998) sensory
neurotransmitters more severe associated with sensory neuropathy A phenethylamine neuropathy
such as levodopa sensory paclitaxel- Leber hereditary in patients derivative prevents
Precursor of induced peripheral optic neuropathy (Ludolph et al. and relieves acute
phenethylamine, neuropathy (Sun (Chao de la Barca 1992) and cats oxaliplatin-induced
which is a et al. 2018) et al. 2016; Morvan (Dickinson et al. neurotoxicity
neurotransmitter and Demidem 2004) (Durand et al.
2018) 2012) and diabetic
neuropathic pain
(Kadiroglu et al.
2008; Razazian
et al. 2014;
Rowbotham et al.
2004)
N-acetylornithine Non-essential N-acetylornithine Lower plasma Lower serum Arginine Arginine is anti-
Metabolized to levels on day ornithine correlated arginine associated supplementation oxidant
ornithine and then 29 predictive of with worse with diabetic prevents diabetic Polyamine-
polyamines vincristine-induced cardiovascular neuropathy (Ganz neuropathy pain in deficient diet
Ornithine peripheral autonomic et al. 2017) rats (Rondón et al. reduces neuronal
predominantly neuropathy (Verma neuropathy in 2018), but not excitotoxicity
from arginine et al. 2020) diabetic patients significant in
(Mathew et al. patients (Jude et al.
2019) 2010)
Higher plasma A polyamine-
N-acetylputrescine deficient diet
C.-S. Chen and D. L. Hertz
 levels associated prevents pain
with peripheral hypersensitivity
neuropathy in (Rivat et al. 2008)
obese patients (Guo including
et al. 2021) oxaliplatin-induced
Lower skin acute pain
fibroblast arginine hypersensitivity in
and polyamines rats (Ferrier et al.
levels associated 2013)
with Leber
hereditary optic
neuropathy (Chao
de la Barca et al.
2016)
Threonine Essential from food Lower pretreatment Lower skin Higher glycine Glycine antagonist No mechanism
Chemotherapy-Induced Peripheral Neuropathy

Small amount plasma threonine fibroblast threonine levels in spinal cord did not improve determined
converts to glycine associated with and glycine levels tissues in mice with neuropathic pain
more severe associated with Charcot-Marie- (Wallace et al.
sensory paclitaxel- Leber hereditary Tooth type 2D 2002)
induced peripheral optic neuropathy neuropathy (Bais
neuropathy (Sun (Chao de la Barca et al. 2016)
et al. 2018) et al. 2016; Morvan
and Demidem
2018)
Higher
cerebrospinal fluid
threonine levels
found in patients
with endemic optic
neuropathy
(Gonzalez-
Quevedo et al.
311

2001)
(continued)
Table 3 (continued)
312

Other
CIPN Other PN Other observational interventional Putative
Metabolites Description metabolomics metabolomics studies studies mechanisms
Glycogen Glucose storage Glycogen levels on N/A Glycogen Anti-glycemic Glycogen
Glucose is its day 29 predictive of accumulation in agents prevent accumulation
precursor and vincristine-induced neurons or Schwan diabetic neuropathy causes neuropathy
metabolite peripheral cells of patients (El Mouhayyar Hyperglycemia
neuropathy (Verma with late-onset et al. 2020) and causes neuronal
et al. 2020) Pompe disease possibly treat CIPN injury and
(Finsterer et al. (Erdoğan et al. ischemia
2017) and diabetic 2020; Fujita et al.
neuropathy 2015; Shigematsu
(Bischoff 1980; et al. 2020)
Powell et al. 1985;
Yagihashi and
Matsunaga 1979)
Adenosine DNA and RNA Adenosine mono/ N/A Impaired adenosine An adenosine Adenosine
phosphates purine precursors diphosphate levels phosphate phosphate phosphate
Energy storage on day signaling pathway signaling pathway signaling
29 predictive of in mice with activator prevents pathways
vincristine-induced diabetic neuropathy oxaliplatin CIPN in promoting
peripheral (Roy Chowdhury patients (El-Fatatry mitochondrial
neuropathy (Verma et al. 2012) et al. 2018) and biogenesis and
et al. 2020) Patients taking an mice (Mao-Ying inhibiting
Decreased adenosine et al. 2014) autophagy and
adenosine phosphate neuroinflammation
phosphate signaling pathway
indicative of activator had lower
oxaliplatin-induced risk of diabetic
acute pain neuropathy
hypersensitivity (Pop-Busui et al.
C.-S. Chen and D. L. Hertz

(Ferrier et al. 2013) 2009)

Chemotherapy-Induced Peripheral Neuropathy 313

and interventional studies that support the association are described. Four of these
metabolites are amino acids, which are the building blocks for protein synthesis. In
humans, there are 20 proteinogenic amino acids, of which nine essential amino acids
are not produced endogenously and must be supplied from the diet. Many amino
acids participate in neurotransmitter biosynthesis, providing an apparent mechanism
by which they may be involved in peripheral neuropathy (Dalangin et al. 2020;
Kölker 2018). The other two metabolites are glycogen and adenosine phosphate.
Glycogen plays a critical role in carbohydrate metabolism (Nordlie et al. 1999), and
adenosine phosphate is the precursor of RNA and provides energy for cellular
metabolic process (Hardie et al. 2012).

3.1 Histidine

Histidine is an essential amino acid supplied from dietary sources, such as meat.
Histidine has an imidazole function group that can scavenge reactive oxygen species
to achieve anti-oxidant and anti-inflammatory effects (Peterson et al. 1998). This
section will also discuss four of its potentially peripheral neuropathy-related
metabolites, 1-methylhistidine, carnosine, ergothioneine, and histamine (Fig. 1).
Similar to the paclitaxel-induced peripheral neuropathy metabolomics study (Sun
et al. 2018), two metabolomics studies identified that histidine was associated with
diabetic neuropathy, but the directions of associations were inconclusive, and the
associations were not significant after adjusting for clinical covariates (Lin et al.

Fig. 1 Histidine and its peripheral neuropathy-related metabolites have several putative
mechanisms (Bocca et al. 2021; Chao de la Barca et al. 2016; Jong et al. 2011; Lin et al. 2019;
Mathew et al. 2019; Nishida et al. 2018; Song et al. 2010; Sun et al. 2018; Winkels et al. 2020;
Yehia et al. 2019). Quercetin and lafutidine are not histidine metabolites, but they are histamine
inhibitors (Azevedo et al. 2013; Gao et al. 2016; Nagano et al. 2012; Tsukaguchi et al. 2013). Blue
boxes are neurotransmitters, such as histamine. Orange boxes are metabolites or compounds that
have been found to prevent or treat CIPN. Solid arrows are metabolic reactions that can happen in
humans, and dashed arrows are the ones that happen exclusively in animals. X-shaped arrows
indicate inhibition
314 C.-S. Chen and D. L. Hertz

2019; Mathew et al. 2019). A longitudinal MS-based study by Mathew et al. (2019)
measured plasma amino acids and tricarboxylic acid metabolites in patients with
type I diabetes but without any signs of microvascular complications and age- and
sex-matched healthy controls at baseline and for 3 years of observation. Correlation
coefficients between baseline metabolite levels and end-of-study electrocardiogram
parameters were calculated to identify metabolites that predict cardiovascular auto-
nomic neuropathy. Lower histidine levels were moderately correlated with worse
neuropathy but did not remain significant after adjusting for baseline clinical
covariates, including blood glucose, HbA1c, years of diabetes, body mass index,
cholesterol, estimated glomerular filtration rate, and urine microalbumin-to-creati-
nine ratio. An untargeted NMR study by Lin et al. (2019) enrolled patients with type
II diabetes with and without microangiopathy and healthy controls. Orthogonal
partial least squares discriminant analysis (OPLSDA) was used to identify predictive
metabolites and build predictive models of diabetes and diabetic microangiopathy
complications including sensory diabetic neuropathy. Higher plasma histidine was
identified in multiple models that predicted the occurrence of diabetic neuropathy,
but the association was not significant after adjusting for clinical covariates, includ-
ing age, sex, body mass index, and the use of medication for hypertension and
hyperlipidemia. The lack of association after covariate adjustment and seemingly
conflicting findings in these studies does not eliminate the possibility of histidine
being involved in the development of peripheral neuropathy, but indicates that the
causal relationship between histidine, diabetes, and peripheral neuropathy may be
complicated. For example, the two studies of diabetic neuropathy required fasting
before sample collection (Lin et al. 2019; Mathew et al. 2019), but the paclitaxel-
induced peripheral neuropathy study did not (Sun et al. 2018). Since diet can affect
CIPN (Mongiovi et al. 2018), diabetic neuropathy (Bunner et al. 2015; Oza et al.
2021), and diabetes (Neuenschwander et al. 2019; Schwingshackl et al. 2017), and
most of the histidine in the body is supplied by dietary sources, it is difficult to
determine whether lower or higher plasma histidine is causally related to peripheral
neuropathy or this relationship is due to statistical confounding. More studies are
needed to elucidate the effect of histidine on, and its potential for use as a predictive
biomarker of, peripheral neuropathy.
Several peripheral neuropathy metabolomics studies were focused on Leber
hereditary optic neuropathy (LHON), which is an inherited disease that results in
vision loss that can include other elements of peripheral neuropathy such as tremors
and movement disorders. LHON is caused by mutations of mitochondrial DNA
including MT-ND1, MT-ND4, MT-ND4L, or MT-ND6, which are critical for proper
mitochondrial function. Studies have analyzed skin fibroblast metabolomics in
patients with LHON to understand the metabolic changes caused by mitochondrial
dysfunction without the confounding of the short-term effects of diet, medication,
and circadian cycles, which complicate research of systemic metabolomics (Wilkins
et al. 2019). A targeted LC-MS/MS metabolomics study by Chao de la Barca et al.
(2016) used OPLSDA to compare skin fibroblast metabolomics between patients
with LHON and healthy controls. Levels of almost all proteinogenic amino acids,
including histidine, were significantly lower in patients with LHON (Chao de la
Chemotherapy-Induced Peripheral Neuropathy 315

Barca et al. 2016), which is consistent with the known function of mitochondria
synthesizing amino acids to release stored energy.
Histidine has not been tested for prevention or treatment of any types of periph-
eral neuropathy. However, histidine supplementation in obese women with meta-
bolic syndrome has been demonstrated in a randomized controlled trial to suppress
inflammation and oxidative stress (Feng et al. 2013), which may also protect against
CIPN. Interventional studies of histidine supplementation for peripheral neuropathy
prevention and treatment may be warranted.

3.1.1 1-Methylhistidine
1-Methylhistidine is a histidine metabolite that is found exclusively in red meat or
poultry, making it a biomarker for meat consumption (Altorf-van der Kuil et al.
2013; Dragsted 2010; Fraser et al. 2016; Khodorova et al. 2019; Kochlik et al. 2018;
Lloyd et al. 2011; Mitry et al. 2019; Sjölin et al. 1987; Yin et al. 2017), and has a
half-life about half day (Sjölin et al. 1987). Another untargeted LC-MS
metabolomics study of LHON by Bocca et al. (2021) found higher plasma
1-methylhistidine in patients than healthy controls in an OPLSDA multivariable
model. This study required fasting before sample collection (Bocca et al. 2021), so
this result is likely attributed to habitual meat consumption. However, prior studies
have not found that higher meat consumption increases CIPN (Kenkhuis et al. 2021;
Mongiovi et al. 2018). Alternatively, two small interventional studies demonstrated
that patients with diabetic neuropathy had less pain after receiving a low-fat and
plant-based diet (Bunner et al. 2015; Smith et al. 2006). Again, the causal
relationships between diet, metabolites, and neuropathy are likely complex and not
fully understood.

3.1.2 Carnosine
Carnosine is a dipeptide of histidine and alanine that is abundant in human and
animal muscle and brain tissue and is a biomarker for red meat consumption (Altorf-
van der Kuil et al. 2013; Cuparencu et al. 2019; Mitry et al. 2019). Observational
metabolomics studies of meat intake often find increased carnosine (Cheung et al.
2017; Cuparencu et al. 2019; Vázquez-Fresno et al. 2015). Similar to histidine,
carnosine has anti-oxidant and anti-inflammatory properties. In a randomized con-
trolled trial, carnosine supplementation was protective against oxaliplatin-induced
peripheral neuropathy, possibly by reducing nuclear factor erythroid-2 related
factor-2 (NRF2) and nuclear factor kappa light chain enhancer of activated B cells
(NFκB) (Yehia et al. 2019). Nrf2 is involved in the redox pathway and has been
proposed as a potential treatment target for CIPN (Miao et al. 2019; Yang et al. 2018)
and diabetic neuropathy (Ganesh Yerra et al. 2013; Gupta et al. 2021; Kumar and
Mittal 2017; Xu et al. 2013). Similarly, NFκB contributes to CIPN (Janes et al. 2014;
Li et al. 2015; Wang et al. 2017) and diabetic neuropathy (Cameron and Cotter 2008;
Dewanjee et al. 2018; Ganesh Yerra et al. 2013) through its involvement in inflam-
matory pathways.
316 C.-S. Chen and D. L. Hertz

3.1.3 Ergothioneine
The other histidine downstream metabolite of interest, ergothioneine, is synthesized
exclusively by fungi and bacteria, which makes it a biomarker for mushroom
consumption (Pallister et al. 2016). An observational study of patients with colorec-
tal cancer receiving chemotherapy found that patients with higher blood
ergothioneine had nominally, but not significantly, less sensory CIPN symptoms
(Winkels et al. 2020). Similar to histidine and carnosine, ergothioneine is a potent
anti-oxidant that may protect injured tissue (Cheah and Halliwell 2012; Halliwell
et al. 2016). In addition to its anti-oxidant properties, ergothioneine has also been
found in animal studies to reduce oxaliplatin accumulation in dorsal root ganglion
neurons via inhibition of organic cation/carnitine transporter (OCTN1)-mediated
oxaliplatin transport (Fujita et al. 2019; Gründemann 2012; Jong et al. 2011;
Nishida et al. 2018). Ergothioneine has been found to protect against cisplatin
(Song et al. 2010) and oxaliplatin-induced peripheral neuropathy in animal studies
(Jong et al. 2011; Nishida et al. 2018), but we are not aware of any interventional
studies in humans.

3.1.4 Histamine
Histidine is a precursor of histamine, which is a neurotransmitter that participates in
regulation of pain and neurogenic inflammation (Rosa and Fantozzi 2013; Yu et al.
2016). Histamine may have a role in diabetes and diabetic neuropathy (Pini et al.
2016; Wei et al. 2016), and there is a potential that histidine levels affect the nervous
system through histamine pathways, but we are not aware of any observational
peripheral neuropathy metabolomics studies that have directly assessed histamine. A
flavonoid from plants, quercetin, has been found to inhibit mast cells from releasing
antigen-induced histamine (Kimata et al. 2000; Pearce et al. 1984) and subsequently
inhibit transient receptor potential cation channel subfamily V member 1 (TRPV1)
(Gao et al. 2016), which has functions of detection and regulation of heat and pain
(Benítez-Angeles et al. 2020; Cui et al. 2016; Frias and Merighi 2016; Gouin et al.
2017; Romanovsky et al. 2009). Quercetin has been found to have anti-oxidant
properties via the nitric oxide pathways (Di Carlo et al. 1999; Gao et al. 2016) and
anti-inflammatory property via the arachidonic acid pathways (Di Carlo et al. 1999;
Ferrándiz and Alcaraz 1991) and has been found in animal studies to prevent pain
from oxaliplatin-(Azevedo et al. 2013) and paclitaxel-induced peripheral neuropathy
(Gao et al. 2016). A histamine H2 receptor antagonist, lafutidine, has been found in
small clinical trials to potentially be effective for treating taxane-induced peripheral
neuropathy (Nagano et al. 2012) and prevent bortezomib-induced peripheral neu-
ropathy (Tsukaguchi et al. 2013). The pain relieving effect of lafutidine has been
hypothesized to be similar to capsaicin, which activates and then reduces the
expression of TRPV1 (TRPV1 defunctionalization) in sensory neurons (Onodera
et al. 1995, 1999; Umeda et al. 1999; Yamamoto et al. 2001).

3.1.5 Summary
Lower histidine levels have been found to be associated with paclitaxel-induced
peripheral neuropathy (Sun et al. 2018) and LHON (Chao de la Barca et al. 2016),
Chemotherapy-Induced Peripheral Neuropathy 317

but the directions of associations in diabetic neuropathy were inconclusive (Lin et al.
2019; Mathew et al. 2019). The evidence suggests that histidine plays some role in
peripheral neuropathy. However, the causal mechanistic relationship is unclear,
which likely explains the lack of significance after adjusting for clinical covariates
(Lin et al. 2019; Mathew et al. 2019).
Metabolomics studies of LHON found higher levels of 1-methylhistidine (Bocca
et al. 2021), which has not been identified in observational dietary analyses
(Kenkhuis et al. 2021; Mongiovi et al. 2018), but suggested that reducing meat
consumption may be a possible dietary intervention of peripheral neuropathy
(Bunner et al. 2015; Smith et al. 2006). If meat consumption is demonstrated to
increase CIPN risk, 1-methylhistidine and carnosine are indicators of meat consump-
tion that could be used as CIPN biomarkers (Altorf-van der Kuil et al. 2013;
Cuparencu et al. 2019; Mitry et al. 2019).
Histidine, carnosine, and ergothioneine all have anti-oxidant and anti-
inflammatory properties that can protect injured tissues and possibly protect against
peripheral neuropathy. Carnosine supplementation may prevent oxaliplatin-induced
peripheral neuropathy through anti-oxidant and anti-inflammatory pathways (Yehia
et al. 2019), whereas ergothioneine may have an additional mechanism of reducing
oxaliplatin accumulation in neurons (Fujita et al. 2019; Gründemann 2012; Jong
et al. 2011; Nishida et al. 2018; Winkels et al. 2020).
Histidine may affect the nervous system through histamine, a neurotransmitter
that regulates pain and neurogenic inflammation (Rosa and Fantozzi 2013; Yu et al.
2016). Histamine inhibitors such as quercetin (Azevedo et al. 2013; Gao et al. 2016)
and lafutidine (Nagano et al. 2012; Tsukaguchi et al. 2013) may prevent or relieve
CIPN pain. Further experimental work is needed to confirm these mechanisms and
randomized interventional trials are needed to confirm the effectiveness for preven-
tion or treatment of CIPN.

3.2 Phenylalanine

Phenylalanine is an essential aromatic amino acid found in high-protein food

sources, such as meat and cheese. Phenylalanine is abundant in human brain tissue
and plasma. Phenylalanine regulates insulin secretion (Floyd et al. 1966; Guasch-
Ferré et al. 2016; Urpi-Sarda et al. 2019; Wang et al. 2011; Wishart 2019) and is
associated with higher risk of diabetes (Guasch-Ferré et al. 2016; Park et al. 2018;
Sun et al. 2020). This section will discuss phenylalanine and two of its potentially
peripheral neuropathy-related metabolites, tyrosine and phenethylamine (Fig. 2).

3.2.1 Phenylalanine and Tyrosine

Phenylalanine is the precursor of tyrosine, which is also an essential aromatic amino
acid. Tyrosine is rapidly metabolized to 3,4-dihydroxyphenylalanine (levodopa),
which is the precursor of many catecholamine neurotransmitters, including dopa-
mine, norepinephrine, and epinephrine. Tyrosine is associated with higher risk of
diabetes (Guasch-Ferré et al. 2016; Park et al. 2018; Sun et al. 2020) and many
318 C.-S. Chen and D. L. Hertz

Fig. 2 Phenylalanine and its peripheral neuropathy-related metabolites (Chao de la Barca et al.
2016; Dickinson et al. 2004; Ertas et al. 1998; Ludolph et al. 1992; Morvan and Demidem 2018;
Sun et al. 2018). Blue boxes are neurotransmitters. Orange boxes are metabolites or compounds that
have been found to relieve neuropathic pain. Solid arrows are metabolic reactions that can happen in
humans, and circle-shaped arrows indicate induction. Venlafaxine is a phenethylamine derivative
that inhibits the reuptake of norepinephrine, which leads to increased norepinephrine levels (Durand
et al. 2012; Kadiroglu et al. 2008; Razazian et al. 2014; Rowbotham et al. 2004)

disorders of the central nervous system, including headache (D’Andrea et al. 2019),
depression (Bot et al. 2020), and Alzheimer’s disease (Albrahim 2020), but that is
beyond the scope of this chapter.
Even though phenylalanine and its metabolites seem to play a role in neurotrans-
mission, only two peripheral neuropathy metabolomics studies identified differences
in phenylalanine levels. These two LHON skin fibroblast metabolomics studies
found most of the proteinogenic amino acids levels, including phenylalanine and
tyrosine, were significantly lower in patients with LHON (Chao de la Barca et al.
2016; Morvan and Demidem 2018). In a previously mentioned diabetic neuropathy
study, higher plasma tyrosine, but not phenylalanine, was identified in some multi-
variable models, but the performance of these models was not ideal, and tyrosine was
not significant after adjusting for clinical covariates (Lin et al. 2019).
There are no interventional trials of phenylalanine supplementation to prevent or
treat peripheral neuropathy, to the best of our knowledge. However, an observational
study suggests that dietary phenylalanine restriction in patients with hereditary
hyperphenylalaninemia causes sensory neuropathy (Ludolph et al. 1992). Another
study in cats also showed that restricting dietary phenylalanine and tyrosine can
cause sensory neuropathy (Dickinson et al. 2004). This strongly implicates
Chemotherapy-Induced Peripheral Neuropathy 319

phenylalanine deficiency as a causal mechanism for sensory neuropathy. A meta-

analysis found that tyrosine supplementation does not improve the neuropsychologi-
cal performance in patients with phenylketonuria who receive a phenylalanine-
restricted diet (Remmington and Smith 2021). Levodopa, which is within the
previously mentioned catecholamine pathway, has been shown to relieve diabetic
neuropathic pain in a small clinical trial (Ertas et al. 1998), which provides further
support for this pathway and suggests the potential for repurposing levodopa for
neuropathic pain.

3.2.2 Phenethylamine
The other phenylalanine metabolite, phenethylamine, is a brain neurotransmitter.
Human metabolomics studies of the use of methamphetamine (Boxler et al. 2017;
Kim et al. 2020), a phenethylamine derivative, have analyzed phenethylamine, but
we are not aware of any observational peripheral neuropathy metabolomics studies
that have assessed phenethylamine.
Another phenethylamine derivative, venlafaxine, is a serotonin-norepinephrine
reuptake inhibitor antidepressant that has been found to prevent and relieve acute
oxaliplatin-induced neurotoxicity (Durand et al. 2012), but not oxaliplatin-induced
peripheral neuropathy (Zimmerman et al. 2016). Venlafaxine has also been found to
reduce pain from diabetic neuropathy (Kadiroglu et al. 2008; Razazian et al. 2014;
Rowbotham et al. 2004), but is less effective than duloxetine and pregabalin for
treating CIPN (Farshchian et al. 2018) or diabetic neuropathy (Raskin et al. 2006;
Razazian et al. 2014).

3.2.3 Summary
Similar to the findings in CIPN (Sun et al. 2018), metabolomics studies in LHON
(Chao de la Barca et al. 2016; Morvan and Demidem 2018) indicate that phenylala-
nine or its metabolites may be involved in peripheral neuropathy. Two studies of
phenylalanine or tyrosine dietary restriction confirm this causes sensory peripheral
neuropathy (Dickinson et al. 2004; Ludolph et al. 1992), likely through the cate-
cholamine neurotransmitter pathway. Catecholamines have demonstrated effective-
ness in neuropathy-relevant conditions including the effectiveness of levodopa and
venlafaxine for relieving neuropathic pain caused by chemotherapy (Durand et al.
2012) or diabetes (Ertas et al. 1998; Kadiroglu et al. 2008; Razazian et al. 2014;
Rowbotham et al. 2004).

3.3 N-Acetylornithine

N-acetylornithine is an intermediate metabolic precursor of ornithine metabolism.

There is minimal information of the source or physiological role of N-acetylornithine
in humans. Other than the vincristine-induced peripheral neuropathy study (Verma
et al. 2020), only one study found higher N-acetylornithine levels to be associated
with the treatment effect of a tremor and involuntary movement disorder (Napoli
et al. 2019). No other observational or interventional peripheral neuropathy studies
320 C.-S. Chen and D. L. Hertz

Fig. 3 N-acetylornithine and

its peripheral neuropathy-
related metabolites (Chao de
la Barca et al. 2016; Ferrier
et al. 2013; Ganz et al. 2017;
Jude et al. 2010; Mathew et al.
2019; Rivat et al. 2008);
Rondón et al. (2018); (Verma
et al. 2020). Orange boxes are
metabolites that have been
found with effects in
peripheral neuropathy. Solid
arrows are metabolic reactions
that can happen in humans

have assessed N-acetylornithine. The major metabolite of N-acetylornithine is orni-

thine. In this section, we will discuss ornithine, together with ornithine’s major
precursor, arginine, and a group of ornithine downstream metabolites, polyamines
(Fig. 3).

3.3.1 Ornithine and Arginine

While ornithine is abundant in many plant food sources, ornithine is mostly pro-
duced from arginine in the urea cycle to dispose excess nitrogen. Only one
metabolomics study has identified an association between ornithine and neuropathy.
The longitudinal MS-based study by Mathew et al. (2019) found lower plasma
ornithine levels were moderately correlated with worse cardiovascular autonomic
neuropathy in patients with diabetes, and this association remained significant after
adjusting for clinical covariates at baseline. Another study found lower serum
arginine, but not ornithine, levels were associated with higher risk of peripheral
neuropathy in patients with diabetes (Ganz et al. 2017). Two previously mentioned
metabolomics studies found lower skin fibroblast levels of arginine, but not orni-
thine, in patients with LHON (Chao de la Barca et al. 2016; Morvan and Demidem
2018).
Ornithine has not been examined in interventional studies of peripheral neuropa-
thy. Arginine supplementation has been found to prevent painful diabetic neuropa-
thy in a rat diabetes model by reducing nitric oxide production (Rondón et al. 2018).
However, in a randomized controlled trial arginine supplementation was not effec-
tive in reducing neuropathy in patients with diabetic neuropathy, though this could
be due to the study being underpowered or confounding caused by glucose-lowering
medications (Jude et al. 2010).
Chemotherapy-Induced Peripheral Neuropathy 321

3.3.2 Ornithine and Polyamines: Putrescine, Spermidine,

and Spermine
Polyamines, including putrescine, spermidine, and spermine, are synthesized from
ornithine and are mainly from dietary intake and gut flora metabolism. Spermidine
and spermine positively modulate N-methyl-D-aspartate (NMDA) receptors by
shielding an inhibition site (Subramaniam et al. 1994; Traynelis et al. 1995), and a
prolonged activation of NMDA receptors can lead to neuronal damage (Paschen
1992). An untargeted LC-MS/MS metabolomics study by Guo et al. compared
glycemic status-matched obese patients with and without peripheral neuropathy
via PLSDA and group least absolute shrinkage and selection operator (LASSO)
regression after adjusting for age and sex. They found obese patients with peripheral
neuropathy had higher plasma N-acetylputrescine, but the mechanisms of
polyamines in PN have not been well studied (Guo et al. 2021). However, a
previously mentioned skin fibroblast metabolomics study found lower levels of
putrescine and spermidine in patients with LHON, but the mechanism for this
seemingly inverse relationship was unclear (Chao de la Barca et al. 2016).
A polyamine-deficient diet has been found to protect against pain hypersensitivity
(Rivat et al. 2008) including oxaliplatin-induced acute pain hypersensitivity in rats
(Ferrier et al. 2013). An interventional randomized controlled trial of the efficacy of
the polyamine-deficient diet in oxaliplatin-induced peripheral neuropathy is ongoing
(Balayssac et al. 2015). We are not aware of any interventional studies testing the
effect of restricting an individual polyamine.

3.3.3 Summary
The direct association between N-acetylornithine and peripheral neuropathy was
only found in the vincristine-induced peripheral neuropathy metabolomics study
(Verma et al. 2020), but lower levels of ornithine, arginine, and polyamines have
been found in patients with other etiologies of peripheral neuropathy. The minimal
data from interventional studies of arginine supplementation trials indicate limited
efficacy. Despite the seemingly conflicting findings in observational studies of
polyamines, polyamine-deficient diet may protect against peripheral neuropathic
pain through NMDA pathways (Ferrier et al. 2013; Rivat et al. 2008). More studies
with a larger sample size are needed to explore the role of ornithine, arginine, or
polyamines in peripheral neuropathy.

3.4 Threonine

The association between lower threonine and peripheral neuropathy was found in the
paclitaxel-induced peripheral neuropathy whole blood metabolomics study (Sun
et al. 2018) Threonine is an essential amino acid found in high-protein food sources,
such as meat and cheese. Threonine is abundant in human plasma and does not seem
to be directly associated with neuropathy, but a small amount of threonine is
converted to glycine (Darling et al. 2000; Edgar 2002; Zhao et al. 1986). This
section will discuss threonine and glycine together.
322 C.-S. Chen and D. L. Hertz

Three metabolomics studies have reported an association between threonine and

optic neuropathy (Chao de la Barca et al. 2016; Gonzalez-Quevedo et al. 2001;
Morvan and Demidem 2018). The first two are previously mentioned skin fibroblast
metabolomics studies, which found patients with LHON had lower levels of most of
the proteinogenic amino acids, including threonine and glycine (Chao de la Barca
et al. 2016; Morvan and Demidem 2018). The other study by Gonzalez-Quevedo
et al. (2001) was an LC metabolomics study of endemic optic neuropathy, which is
suspected to be due to nutrient insufficiency. Threonine levels were significantly
higher in cerebrospinal fluid, but not serum, in patients with optic neuropathy. This
study did not require fasting before sample collection, but the samples were collected
in the early morning to minimize the influence from circadian rhythms and food. No
other plasma metabolomics studies have found an association between threonine and
peripheral neuropathy to the best of our knowledge.
In a mouse study of Charcot-Marie-Tooth type 2D neuropathy, which is an
inherited neuropathy caused by mutations of glycyl-tRNA synthetase 1 (GARS1),
mice with neuropathy had higher glycine levels in spinal cord tissues. However,
glycine supplementation did not worsen or affect the neuropathy assessment (Bais
et al. 2016), which implied that high glycine was not the cause of neuropathy.
Threonine has not been tested in interventional studies of peripheral neuropathy,
and a glycine antagonist did not improve pain or neuropathy symptoms in a clinical
trial (Wallace et al. 2002). Overall, the evidence supporting an association of
threonine or its metabolites is weak and no interventional studies demonstrate
effectiveness of neuropathy prevention or treatment.

3.5 Glycogen

Glycogen is a polysaccharide that is the primary storage form of glucose. Other than
the vincristine-induced peripheral neuropathy study (Verma et al. 2020), no other
peripheral neuropathy metabolomics studies have assessed glycogen. An inherited
disorder, Pompe disease, is caused by mutations of alpha glucosidase (GAA), which
leads to glycogen accumulation in lysosomes, especially in heart and skeletal
muscles. Neuropathy has been reported in patients with late-onset Pompe disease
due to glycogen accumulation in neurons or Schwan cells (Finsterer et al. 2017).
Aggregates of glycogen particles have been found in neurons and Schwann cells in
patients with diabetic neuropathy as well (Bischoff 1980; Powell et al. 1985;
Yagihashi and Matsunaga 1979). These studies strongly suggest that glycogen
accumulation is associated with peripheral neuropathy, but the role of glycogen in
CIPN is unclear.
Chronic elevation of glucose, the precursor and metabolite of glycogen, is the
main cause of diabetes, and high glucose levels in peripheral nerves is the pathogen-
esis of diabetic neuropathy (Freeman et al. 2016). Hyperglycemia causes injury to
neuronal cells and excessive glucose metabolism promotes thickening of capillary
basement membrane, which leads to neuronal ischemia (Filla and Edwards 2016;
Forbes and Cooper 2013). Numerous anti-glycemic agents have been approved for
Chemotherapy-Induced Peripheral Neuropathy 323

reducing blood glucose in patients with diabetes and is expected to reduce the risk of
diabetic microvascular complication including peripheral neuropathy (El Mouhayyar
et al. 2020), and some of these drugs also showed efficacy in treating CIPN through
glucagon-like peptide-1 signaling pathways (Erdoğan et al. 2020; Fujita et al. 2015;
Shigematsu et al. 2020). However, blood glucose varies by dietary intake, and
glycosylated hemoglobin is considered to be a more accurate long-term biomarker
of glycemic control than short-term blood glucose (Koenig et al. 1976).

3.6 Adenosine Phosphates

Adenosine monophosphate and adenosine diphosphate were found in a plasma

metabolomics study of vincristine-induced peripheral neuropathy, but the
concentrations were not reported (Verma et al. 2020). Adenosine monophosphate
and adenosine diphosphate are purine-based nucleotides with one and two phosphate
groups, respectively. They are the precursors of RNA and can convert to each other
or adenosine triphosphate, commonly referred to as ATP, which provides energy for
a variety of signaling process in cells. A possible mechanistic pathway connecting
these metabolites to neuropathy is that both adenosine monophosphate and adeno-
sine diphosphate activate adenosine monophosphate activated protein kinase
(AMPK), which has been found to play an important role in diabetic neuropathy
(Madhavi et al. 2019; Shrikanth and Nandini 2020). Impaired AMPK signaling is a
component of mitochondrial dysfunction and diabetic neuropathy in a diabetic
mouse model (Roy Chowdhury et al. 2012). AMPK activation can prevent and
reverse the neuronal injury by promoting mitochondrial biogenesis and inhibiting
autophagy and neuroinflammation (Roy Chowdhury et al. 2012; Yerra et al. 2017),
which suggests that activation of AMPK can be a prevention or treatment target for
diabetic neuropathy and possibly other peripheral neuropathy (Madhavi et al. 2019;
Price et al. 2016; Shaw et al. 2020; Shrikanth and Nandini 2020). This may explain
the finding that patients receiving the AMPK activator metformin, the first-line
treatment for diabetes, had lower risk of developing neuropathy than patients
receiving insulin, and the association was independent of the duration of diabetes
and glycemic control (Pop-Busui et al. 2009). Clinical and mouse studies also
suggests metformin may protect against oxaliplatin-induced peripheral neuropathy
(El-Fatatry et al. 2018; Mao-Ying et al. 2014). Reduced AMPK activity can result in
upregulation of transient receptor potential cation channel subfamily A member
1 (TRPA1). TRPA1 is related to, but in a different receptor subfamily as, the
TRVP1 receptor that is highly associated with neuropathic pain and was previously
mentioned in the histamine subsection (Wang et al. 2018). In summary, adenosine
phosphates were identified in both clinical (Verma et al. 2020) and animal (Ferrier
et al. 2013) CIPN metabolomics studies, and may work via AMPK signaling
pathway activation. AMPK activators, such as metformin, have shown potentials
to prevent and treat diabetic neuropathy and CIPN (El-Fatatry et al. 2018; Mao-Ying
et al. 2014; Pop-Busui et al. 2009), further supporting the possibility that this
pathway has a role in CIPN.
324 C.-S. Chen and D. L. Hertz

4 Conclusion

Although scarce, there have been a few CIPN metabolomics biomarker studies in
clinical, animal, and cellular models. Our review of the clinical studies identified six
metabolites that may predict CIPN risk. We were able to find supportive evidence for
these associations by reviewing metabolomics, observational, and interventional
studies of CIPN and other etiologies of peripheral neuropathy. Histidine and phe-
nylalanine have clear roles in neurotransmission or neuroinflammation in peripheral
neuropathy, including CIPN. Furthermore, interventional studies of their
metabolites, including carnosine and levodopa, have shown promise in preventing
or treating CIPIN and other types of peripheral neuropathy, though definitive
interventional trials are needed. Further research is needed to discover and validate
CIPN metabolomics biomarkers in large clinical studies and to elucidate the
mechanisms and causal relationships through animal or cell culture experiments.
After retrospective validation, prospective biomarker-based interventional studies
are essential to translate the biomarker findings into clinical practice. For example,
metabolomics biomarkers could be used to identify patients at high CIPN risk before
or early in treatment, in whom CIPN monitoring could be enhanced or investiga-
tional preventive interventions could be tested. Metabolomic biomarkers can also be
indicative of effective interventions, such as nutrient supplementation, to prevent or
reduce CIPN. The ultimate goal is to improve treatment outcomes and quality of life
of patients with cancer.

References
Albrahim T (2020) The potential role of nutritional components in improving brain function among
patients with Alzheimers disease: a meta-analysis of RCT studies. Neurosciences 25:4–17.
https://doi.org/10.17712/nsj.2020.1.20190037
Altorf-van der Kuil W, Brink EJ, Boetje M, Siebelink E, Bijlsma S, Engberink MF, van’t Veer P,
Tomé D, Bakker SJ, van Baak MA, Geleijnse JM (2013) Identification of biomarkers for intake
of protein from meat, dairy products and grains: a controlled dietary intervention study. Br J
Nutr 110:810–822. https://doi.org/10.1017/s0007114512005788
Argyriou AA, Iconomou G, Kalofonos HP (2008) Bortezomib-induced peripheral neuropathy in
multiple myeloma: a comprehensive review of the literature. Blood 112:1593–1599. https://doi.
org/10.1182/blood-2008-04-149385
Azevedo MI, Pereira AF, Nogueira RB, Rolim FE, Brito GA, Wong DV, Lima-Júnior RC, de
Albuquerque RR, Vale ML (2013) The antioxidant effects of the flavonoids rutin and quercetin
inhibit oxaliplatin-induced chronic painful peripheral neuropathy. Mol Pain 9:53. https://doi.
org/10.1186/1744-8069-9-53
Backshall A, Sharma R, Clarke SJ, Keun HC (2011) Pharmacometabonomic profiling as a predictor
of toxicity in patients with inoperable colorectal cancer treated with capecitabine. Clin Cancer
Res 17:3019–3028. https://doi.org/10.1158/1078-0432.Ccr-10-2474
Bais P, Beebe K, Morelli KH, Currie ME, Norberg SN, Evsikov AV, Miers KE, Seburn KL,
Guergueltcheva V, Kremensky I, Jordanova A, Bult CJ, Burgess RW (2016) Metabolite profile
of a mouse model of Charcot-Marie-Tooth type 2D neuropathy: implications for disease
mechanisms and interventions. Biol Open 5:908–920. https://doi.org/10.1242/bio.019273
Chemotherapy-Induced Peripheral Neuropathy 325

Balayssac D, Ferrier J, Pereira B, Gillet B, Pétorin C, Vein J, Libert F, Eschalier A, Pezet D (2015)
Prevention of oxaliplatin-induced peripheral neuropathy by a polyamine-reduced diet-
NEUROXAPOL: protocol of a prospective, randomised, controlled, single-blind and
monocentric trial. BMJ Open 5:e007479. https://doi.org/10.1136/bmjopen-2014-007479
Bandos H, Melnikow J, Rivera DR, Swain SM, Sturtz K, Fehrenbacher L, Wade JL 3rd, Brufsky
AM, Julian TB, Margolese RG, McCarron EC, Ganz PA (2018) Long-term peripheral neuropa-
thy in breast cancer patients treated with adjuvant chemotherapy: NRG oncology/NSABP B-30.
J Natl Cancer Inst 110. https://doi.org/10.1093/jnci/djx162
Bao T, Basal C, Seluzicki C, Li SQ, Seidman AD, Mao JJ (2016) Long-term chemotherapy-induced
peripheral neuropathy among breast cancer survivors: prevalence, risk factors, and fall risk.
Breast Cancer Res Treat 159:327–333. https://doi.org/10.1007/s10549-016-3939-0
Benítez-Angeles M, Morales-Lázaro SL, Juárez-González E, Rosenbaum T (2020) TRPV1: struc-
ture, endogenous agonists, and mechanisms. Int J Mol Sci 21. https://doi.org/10.3390/
ijms21103421
Bischoff A (1980) Morphology of diabetic neuropathy. Horm Metab Res Suppl 9:18–28
Bocca C, Le Paih V, Barca JMC, Nzoughet JK, Amati-Bonneau P, Blanchet O, Védié B,
Géromin D, Simard G, Procaccio V, Bonneau D, Leaners G, Orssaud C, Reynier P (2021) A
plasma metabolomic signature of Leber hereditary optic neuropathy showing taurine and
nicotinamide deficiencies. Hum Mol Genet. https://doi.org/10.1093/hmg/ddab013
Bot M, Milaneschi Y, Al-Shehri T, Amin N, Garmaeva S, Onderwater GLJ, Pool R, Thesing CS,
Vijfhuizen LS, Vogelzangs N, Arts ICW, Demirkan A, van Duijn C, van Greevenbroek M, van
der Kallen CJH, Köhler S, Ligthart L, van den Maagdenberg A, Mook-Kanamori DO, de
Mutsert R, Tiemeier H, Schram MT, Stehouwer CDA, Terwindt GM, Willems van Dijk K,
Fu J, Zhernakova A, Beekman M, Slagboom PE, Boomsma DI, Penninx B (2020)
Metabolomics profile in depression: a pooled analysis of 230 metabolic markers in 5283
cases with depression and 10,145 controls. Biol Psychiatry 87:409–418. https://doi.org/10.
1016/j.biopsych.2019.08.016
Boxler MI, Liechti ME, Schmid Y, Kraemer T, Steuer AE (2017) First time view on human
metabolome changes after a single intake of 3,4-methylenedioxymethamphetamine in healthy
placebo-controlled subjects. J Proteome Res 16:3310–3320. https://doi.org/10.1021/acs.
jproteome.7b00294
Bulls HW, Hoogland AI, Small BJ, Kennedy B, James BW, Arboleda BL, Shahzad MMK,
Gonzalez BD, Jim HSL (2020) Lagged relationships among chemotherapy-induced peripheral
neuropathy, sleep quality, and physical activity during and after chemotherapy. Ann Behav
Med. https://doi.org/10.1093/abm/kaaa101
Bunner AE, Wells CL, Gonzales J, Agarwal U, Bayat E, Barnard ND (2015) A dietary intervention
for chronic diabetic neuropathy pain: a randomized controlled pilot study. Nutr Diabetes 5:e158.
https://doi.org/10.1038/nutd.2015.8
Cameron NE, Cotter MA (2008) Pro-inflammatory mechanisms in diabetic neuropathy: focus on
the nuclear factor kappa B pathway. Curr Drug Targets 9:60–67. https://doi.org/10.2174/
138945008783431718
Chan A, Hertz DL, Morales M, Adams EJ, Gordon S, Tan CJ, Staff NP, Kamath J, Oh J, Shinde S,
Pon D, Dixit N, D'Olimpio J, Dumitrescu C, Gobbo M, Kober K, Mayo S, Pang L, Subbiah I,
Beutler AS, Peters KB, Loprinzi C, Lustberg MB (2019) Biological predictors of chemotherapy-
induced peripheral neuropathy (CIPN): MASCC neurological complications working group
overview. Support Care Cancer 27:3729–3737. https://doi.org/10.1007/s00520-019-04987-8
Chao de la Barca JM, Simard G, Amati-Bonneau P, Safiedeen Z, Prunier-Mirebeau D, Chupin S,
Gadras C, Tessier L, Gueguen N, Chevrollier A, Desquiret-Dumas V, Ferré M, Bris C, Kouassi
Nzoughet J, Bocca C, Leruez S, Verny C, Miléa D, Bonneau D, Lenaers G, Martinez MC,
Procaccio V, Reynier P (2016) The metabolomic signature of Leber's hereditary optic neuropa-
thy reveals endoplasmic reticulum stress. Brain 139:2864–2876. https://doi.org/10.1093/brain/
aww222
326 C.-S. Chen and D. L. Hertz

Cheah IK, Halliwell B (2012) Ergothioneine; antioxidant potential, physiological function and role
in disease. Biochim Biophys Acta 1822:784–793. https://doi.org/10.1016/j.bbadis.2011.09.017
Chen L, Chen CS, Sun Y, Henry NL, Stringer KA, Hertz DL (2021) Feasibility of
pharmacometabolomics to identify potential predictors of paclitaxel pharmacokinetic
variability. Cancer Chemother Pharmacol. https://doi.org/10.1007/s00280-021-04300-7
Cheung W, Keski-Rahkonen P, Assi N, Ferrari P, Freisling H, Rinaldi S, Slimani N, Zamora-Ros R,
Rundle M, Frost G, Gibbons H, Carr E, Brennan L, Cross AJ, Pala V, Panico S, Sacerdote C,
Palli D, Tumino R, Kühn T, Kaaks R, Boeing H, Floegel A, Mancini F, Boutron-Ruault MC,
Baglietto L, Trichopoulou A, Naska A, Orfanos P, Scalbert A (2017) A metabolomic study of
biomarkers of meat and fish intake. Am J Clin Nutr 105:600–608. https://doi.org/10.3945/ajcn.
116.146639
Clayton TA, Baker D, Lindon JC, Everett JR, Nicholson JK (2009) Pharmacometabonomic
identification of a significant host-microbiome metabolic interaction affecting human drug
metabolism. Proc Natl Acad Sci U S A 106:14728–14733. https://doi.org/10.1073/pnas.
0904489106
Cliff J, Jorgensen AL, Lord R, Azam F, Cossar L, Carr DF, Pirmohamed M (2017) The molecular
genetics of chemotherapy-induced peripheral neuropathy: a systematic review and meta-
analysis. Crit Rev Oncol Hematol 120:127–140. https://doi.org/10.1016/j.critrevonc.2017.
09.009
Cox-Martin E, Trahan LH, Cox MG, Dougherty PM, Lai EA, Novy DM (2017) Disease burden and
pain in obese cancer patients with chemotherapy-induced peripheral neuropathy. Support Care
Cancer 25:1873–1879. https://doi.org/10.1007/s00520-017-3571-5
Cui M, Gosu V, Basith S, Hong S, Choi S (2016) Polymodal transient receptor potential vanilloid
type 1 Nocisensor: structure, modulators, and therapeutic applications. Adv Protein Chem Struct
Biol 104:81–125. https://doi.org/10.1016/bs.apcsb.2015.11.005
Cuparencu C, Rinnan Å, Dragsted LO (2019) Combined markers to assess meat intake-human
metabolomic studies of discovery and validation. Mol Nutr Food Res 63:e1900106. https://doi.
org/10.1002/mnfr.201900106
D'Andrea G, Gucciardi A, Perini F, Leon A (2019) Pathogenesis of cluster headache: from episodic
to chronic form, the role of neurotransmitters and neuromodulators. Headache 59:1665–1670.
https://doi.org/10.1111/head.13673
Dalangin R, Kim A, Campbell RE (2020) The role of amino acids in neurotransmission and
fluorescent tools for their detection. Int J Mol Sci 21. https://doi.org/10.3390/ijms21176197
Darling PB, Grunow J, Rafii M, Brookes S, Ball RO, Pencharz PB (2000) Threonine dehydroge-
nase is a minor degradative pathway of threonine catabolism in adult humans. Am J Physiol
Endocrinol Metab 278:E877–E884. https://doi.org/10.1152/ajpendo.2000.278.5.E877
Dewanjee S, Das S, Das AK, Bhattacharjee N, Dihingia A, Dua TK, Kalita J, Manna P (2018)
Molecular mechanism of diabetic neuropathy and its pharmacotherapeutic targets. Eur J
Pharmacol 833:472–523. https://doi.org/10.1016/j.ejphar.2018.06.034
Di Carlo G, Mascolo N, Izzo AA, Capasso F (1999) Flavonoids: old and new aspects of a class of
natural therapeutic drugs. Life Sci 65:337–353. https://doi.org/10.1016/s0024-3205(99)
00120-4
Dickinson PJ, Anderson PJ, Williams DC, Powell HC, Shelton GD, Morris JG, LeCouteur RA
(2004) Assessment of the neurologic effects of dietary deficiencies of phenylalanine and
tyrosine in cats. Am J Vet Res 65:671–680. https://doi.org/10.2460/ajvr.2004.65.671
Dragsted LO (2010) Biomarkers of meat intake and the application of nutrigenomics. Meat Sci 84:
301–307. https://doi.org/10.1016/j.meatsci.2009.08.028
Durand JP, Deplanque G, Montheil V, Gornet JM, Scotte F, Mir O, Cessot A, Coriat R,
Raymond E, Mitry E, Herait P, Yataghene Y, Goldwasser F (2012) Efficacy of venlafaxine
for the prevention and relief of oxaliplatin-induced acute neurotoxicity: results of EFFOX, a
randomized, double-blind, placebo-controlled phase III trial. Ann Oncol 23:200–205. https://
doi.org/10.1093/annonc/mdr045
Chemotherapy-Induced Peripheral Neuropathy 327

Duregon F, Vendramin B, Bullo V, Gobbo S, Cugusi L, Di Blasio A, Neunhaeuserer D, Zaccaria M,

Bergamin M, Ermolao A (2018) Effects of exercise on cancer patients suffering chemotherapy-
induced peripheral neuropathy undergoing treatment: a systematic review. Crit Rev Oncol
Hematol 121:90–100. https://doi.org/10.1016/j.critrevonc.2017.11.002
Edgar AJ (2002) The human L-threonine 3-dehydrogenase gene is an expressed pseudogene. BMC
Genet 3:18. https://doi.org/10.1186/1471-2156-3-18
El-Fatatry BM, Ibrahim OM, Hussien FZ, Mostafa TM (2018) Role of metformin in oxaliplatin-
induced peripheral neuropathy in patients with stage III colorectal cancer: randomized, con-
trolled study. Int J Colorectal Dis 33:1675–1683. https://doi.org/10.1007/s00384-018-3104-9
El Mouhayyar C, Riachy R, Khalil AB, Eid A, Azar S (2020) SGLT2 inhibitors, GLP-1 agonists,
and DPP-4 inhibitors in diabetes and microvascular complications: a review. Int J Endocrinol
2020:1762164. https://doi.org/10.1155/2020/1762164
Emwas A-H, Roy R, McKay RT, Tenori L, Saccenti E, Gowda GAN, Raftery D, Alahmari F,
Jaremko L, Jaremko M, Wishart DS (2019) NMR spectroscopy for metabolomics research.
Metabolites 9:123. https://doi.org/10.3390/metabo9070123
Erdoğan MA, Taşkıran E, Yiğittürk G, Erbaş O, Taşkıran D (2020) The investigation of therapeutic
potential of oxytocin and liraglutide on vincristine-induced neuropathy in rats. J Biochem Mol
Toxicol 34:e22415. https://doi.org/10.1002/jbt.22415
Ertas M, Sagduyu A, Arac N, Uludag B, Ertekin C (1998) Use of levodopa to relieve pain from
painful symmetrical diabetic polyneuropathy. Pain 75:257–259. https://doi.org/10.1016/s0304-
3959(98)00003-7
Everett JR (2017) NMR-based pharmacometabonomics: a new paradigm for personalised or
precision medicine. Prog Nucl Magn Reson Spectrosc 102-103:1–14. https://doi.org/10.1016/
j.pnmrs.2017.04.003
Everett JR (2019) Pharmacometabonomics: the prediction of drug effects using metabolic profiling.
Handb Exp Pharmacol 260:263–299. https://doi.org/10.1007/164_2019_316
Farshchian N, Alavi A, Heydarheydari S, Moradian N (2018) Comparative study of the effects of
venlafaxine and duloxetine on chemotherapy-induced peripheral neuropathy. Cancer
Chemother Pharmacol 82:787–793. https://doi.org/10.1007/s00280-018-3664-y
Feng RN, Niu YC, Sun XW, Li Q, Zhao C, Wang C, Guo FC, Sun CH, Li Y (2013) Histidine
supplementation improves insulin resistance through suppressed inflammation in obese women
with the metabolic syndrome: a randomised controlled trial. Diabetologia 56:985–994. https://
doi.org/10.1007/s00125-013-2839-7
Ferrándiz ML, Alcaraz MJ (1991) Anti-inflammatory activity and inhibition of arachidonic acid
metabolism by flavonoids. Agents Actions 32:283–288. https://doi.org/10.1007/bf01980887
Ferrier J, Bayet-Robert M, Dalmann R, El Guerrab A, Aissouni Y, Graveron-Demilly D, Chalus M,
Pinguet J, Eschalier A, Richard D, Daulhac L, Marchand F, Balayssac D (2015) Cholinergic
neurotransmission in the posterior insular cortex is altered in preclinical models of
neuropathic pain: key role of muscarinic M2 receptors in donepezil-induced antinociception. J
Neurosci 35:16418–16430. https://doi.org/10.1523/jneurosci.1537-15.2015
Ferrier J, Bayet-Robert M, Pereira B, Daulhac L, Eschalier A, Pezet D, Moulinoux JP, Balayssac D
(2013) A polyamine-deficient diet prevents oxaliplatin-induced acute cold and mechanical
hypersensitivity in rats. PLoS One 8:e77828. https://doi.org/10.1371/journal.pone.0077828
Filla LA, Edwards JL (2016) Metabolomics in diabetic complications. Mol Biosyst 12:1090–1105.
https://doi.org/10.1039/c6mb00014b
Finsterer J, Wanschitz J, Quasthoff S, Iglseder S, Löscher W, Grisold W (2017) Causally treatable,
hereditary neuropathies in Fabry's disease, transthyretin-related familial amyloidosis, and
Pompe's disease. Acta Neurol Scand 136:558–569. https://doi.org/10.1111/ane.12758
Floyd JC Jr, Fajans SS, Conn JW, Knopf RF, Rull J (1966) Stimulation of insulin secretion by
amino acids. J Clin Invest 45:1487–1502. https://doi.org/10.1172/JCI105456
Forbes JM, Cooper ME (2013) Mechanisms of diabetic complications. Physiol Rev 93:137–188.
https://doi.org/10.1152/physrev.00045.2011
328 C.-S. Chen and D. L. Hertz

Fraser GE, Jaceldo-Siegl K, Henning SM, Fan J, Knutsen SF, Haddad EH, Sabaté J, Beeson WL,
Bennett H (2016) Biomarkers of dietary intake are correlated with corresponding measures from
repeated dietary recalls and food-frequency questionnaires in the adventist health study-2. J Nutr
146:586–594. https://doi.org/10.3945/jn.115.225508
Freeman OJ, Unwin RD, Dowsey AW, Begley P, Ali S, Hollywood KA, Rustogi N, Petersen RS,
Dunn WB, Cooper GJ, Gardiner NJ (2016) Metabolic dysfunction is restricted to the sciatic
nerve in experimental diabetic neuropathy. Diabetes 65:228–238. https://doi.org/10.2337/
db15-0835
Frias B, Merighi A (2016) Capsaicin, nociception and pain. Molecules 21. https://doi.org/10.3390/
molecules21060797
Fujita S, Hirota T, Sakiyama R, Baba M, Ieiri I (2019) Identification of drug transporters
contributing to oxaliplatin-induced peripheral neuropathy. J Neurochem 148:373–385. https://
doi.org/10.1111/jnc.14607
Fujita S, Ushio S, Ozawa N, Masuguchi K, Kawashiri T, Oishi R, Egashira N (2015) Exenatide
facilitates recovery from oxaliplatin-induced peripheral neuropathy in rats. PLoS One 10:
e0141921. https://doi.org/10.1371/journal.pone.0141921
Ganesh Yerra V, Negi G, Sharma SS, Kumar A (2013) Potential therapeutic effects of the
simultaneous targeting of the Nrf2 and NF-κB pathways in diabetic neuropathy. Redox Biol
1:394–397. https://doi.org/10.1016/j.redox.2013.07.005
Ganz T, Wainstein J, Gilad S, Limor R, Boaz M, Stern N (2017) Serum asymmetric
dimethylarginine and arginine levels predict microvascular and macrovascular complications
in type 2 diabetes mellitus. Diabetes Metab Res Rev 33. https://doi.org/10.1002/dmrr.2836
Gao W, Zan Y, Wang ZJ, Hu XY, Huang F (2016) Quercetin ameliorates paclitaxel-induced
neuropathic pain by stabilizing mast cells, and subsequently blocking PKCε-dependent activa-
tion of TRPV1. Acta Pharmacol Sin 37:1166–1177. https://doi.org/10.1038/aps.2016.58
Gonzalez-Quevedo A, Obregon F, Fernandez R, Santiesteban R, Serrano C, Lima L (2001) Amino
acid levels and ratios in serum and cerebrospinal fluid of patients with optic neuropathy in Cuba.
Nutr Neurosci 4:51–62. https://doi.org/10.1080/1028415x.2001.11747350
Gouin O, L'Herondelle K, Lebonvallet N, Le Gall-Ianotto C, Sakka M, Buhé V, Plée-Gautier E,
Carré JL, Lefeuvre L, Misery L, Le Garrec R (2017) TRPV1 and TRPA1 in cutaneous
neurogenic and chronic inflammation: pro-inflammatory response induced by their activation
and their sensitization. Protein Cell 8:644–661. https://doi.org/10.1007/s13238-017-0395-5
Greenlee H, Hershman DL, Shi Z, Kwan ML, Ergas IJ, Roh JM, Kushi LH (2017) BMI, lifestyle
factors and Taxane-induced neuropathy in breast cancer patients: the pathways study. J Natl
Cancer Inst 109. https://doi.org/10.1093/jnci/djw206
Grim J, Ticha A, Hyspler R, Valis M, Zadak Z (2017) Selected risk nutritional factors for
chemotherapy-induced polyneuropathy. Nutrients 9:535. https://doi.org/10.3390/nu9060535
Gründemann D (2012) The ergothioneine transporter controls and indicates ergothioneine activity--
a review. Prev Med 54 Suppl: S71–S74. doi: https://doi.org/10.1016/j.ypmed.2011.12.001
Gu J, Lu H, Chen C, Gu Z, Hu M, Liu L, Yu J, Wei G, Huo J (2021) Diabetes mellitus as a risk
factor for chemotherapy-induced peripheral neuropathy: a meta-analysis. Support Care Cancer.
https://doi.org/10.1007/s00520-021-06321-7
Guasch-Ferré M, Hruby A, Toledo E, Clish CB, Martínez-González MA, Salas-Salvadó J, Hu FB
(2016) Metabolomics in prediabetes and diabetes: a systematic review and meta-analysis.
Diabetes Care 39:833–846. https://doi.org/10.2337/dc15-2251
Guo K, Savelieff MG, Rumora AE, Alakwaa FM, Callaghan BC, Hur J, Feldman EL (2021) Plasma
metabolomics and lipidomics differentiate obese individuals by peripheral neuropathy status. J
Clin Endocrinol Metab. https://doi.org/10.1210/clinem/dgab844
Gupta A, Behl T, Sehgal A, Bhatia S, Jaglan D, Bungau S (2021) Therapeutic potential of Nrf-2
pathway in the treatment of diabetic neuropathy and nephropathy. Mol Biol Rep 48:2761–2774.
https://doi.org/10.1007/s11033-021-06257-5
Chemotherapy-Induced Peripheral Neuropathy 329

Haim N, Epelbaum R, Ben-Shahar M, Yarnitsky D, Simri W, Robinson E (1994) Full dose

vincristine (without 2-mg dose limit) in the treatment of lymphomas. Cancer 73:2515–2519.
https://doi.org/10.1002/1097-0142(19940515)73:10<2515::aid-cncr2820731011>3.0.co;2-g
Halliwell B, Cheah IK, Drum CL (2016) Ergothioneine, an adaptive antioxidant for the protection
of injured tissues? A hypothesis. Biochem Biophys Res Commun 470:245–250. https://doi.org/
10.1016/j.bbrc.2015.12.124
Harada S, Hirayama A, Chan Q, Kurihara A, Fukai K, Iida M, Kato S, Sugiyama D, Kuwabara K,
Takeuchi A, Akiyama M, Okamura T, Ebbels TMD, Elliott P, Tomita M, Sato A, Suzuki C,
Sugimoto M, Soga T, Takebayashi T (2018) Reliability of plasma polar metabolite
concentrations in a large-scale cohort study using capillary electrophoresis-mass spectrometry.
PLoS One 13:e0191230. https://doi.org/10.1371/journal.pone.0191230
Hardie DG, Ross FA, Hawley SA (2012) AMPK: a nutrient and energy sensor that maintains
energy homeostasis. Nat Rev Mol Cell Biol 13:251–262. https://doi.org/10.1038/nrm3311
Hershman DL, Till C, Wright JD, Awad D, Ramsey SD, Barlow WE, Minasian LM, Unger J (2016)
Comorbidities and risk of chemotherapy-induced peripheral neuropathy among participants
65 years or older in southwest oncology group clinical trials. J Clin Oncol 34:3014–3022.
https://doi.org/10.1200/jco.2015.66.2346
Hershman DL, Unger JM, Crew KD, Till C, Greenlee H, Minasian LM, Moinpour CM, Lew DL,
Fehrenbacher L, Wade JL 3rd, Wong SF, Fisch MJ, Lynn Henry N, Albain KS (2018) Two-year
trends of taxane-induced neuropathy in women enrolled in a randomized trial of acetyl-L-
carnitine (SWOG S0715). J Natl Cancer Inst 110:669–676. https://doi.org/10.1093/jnci/djx259
Hertz DL (2019) Concerns regarding use of patient-reported outcomes in biomarker studies of
chemotherapy-induced peripheral neuropathy. Pharmacogenomics J 19:411–416. https://doi.
org/10.1038/s41397-019-0093-1
Hertz DL, Childs DS, Park SB, Faithfull S, Ke Y, Ali NT, McGlown SM, Chan A, Grech L,
Loprinzi CL (2021) Patient-centric decision framework for treatment alterations in patients with
chemotherapy-induced peripheral neuropathy (CIPN). Cancer Treat Rev
Janes K, Little JW, Li C, Bryant L, Chen C, Chen Z, Kamocki K, Doyle T, Snider A, Esposito E,
Cuzzocrea S, Bieberich E, Obeid L, Petrache I, Nicol G, Neumann WL, Salvemini D (2014) The
development and maintenance of paclitaxel-induced neuropathic pain require activation of the
sphingosine 1-phosphate receptor subtype 1. J Biol Chem 289:21082–21097. https://doi.org/10.
1074/jbc.M114.569574
Jennaro TS, Fang F, Kidwell KM, Smith EML, Vangipuram K, Burness ML, Griggs JJ, Van
Poznak C, Hayes DF, Henry NL, Hertz DL (2020) Vitamin D deficiency increases severity of
paclitaxel-induced peripheral neuropathy. Breast Cancer Res Treat 180:707–714. https://doi.
org/10.1007/s10549-020-05584-8
Jong NN, Nakanishi T, Liu JJ, Tamai I, McKeage MJ (2011) Oxaliplatin transport mediated by
organic cation/carnitine transporters OCTN1 and OCTN2 in overexpressing human embryonic
kidney 293 cells and rat dorsal root ganglion neurons. J Pharmacol Exp Ther 338:537–547.
https://doi.org/10.1124/jpet.111.181297
Jude EB, Dang C, Boulton AJ (2010) Effect of L-arginine on the microcirculation in the neuropathic
diabetic foot in type 2 diabetes mellitus: a double-blind, placebo-controlled study. Diabet Med
27:113–116. https://doi.org/10.1111/j.1464-5491.2009.02876.x
Kadiroglu AK, Sit D, Kayabasi H, Tuzcu AK, Tasdemir N, Yilmaz ME (2008) The effect of
venlafaxine HCl on painful peripheral diabetic neuropathy in patients with type 2 diabetes
mellitus. J Diabetes Complications 22:241–245. https://doi.org/10.1016/j.jdiacomp.2007.
03.010
Kandula T, Park SB, Cohn RJ, Krishnan AV, Farrar MA (2016) Pediatric chemotherapy induced
peripheral neuropathy: a systematic review of current knowledge. Cancer Treat Rev 50:118–
128. https://doi.org/10.1016/j.ctrv.2016.09.005
Kantae V, Krekels EHJ, Esdonk MJV, Lindenburg P, Harms AC, Knibbe CAJ, Van der Graaf PH,
Hankemeier T (2017) Integration of pharmacometabolomics with pharmacokinetics and
330 C.-S. Chen and D. L. Hertz

pharmacodynamics: towards personalized drug therapy. Metabolomics 13:9. https://doi.org/10.

1007/s11306-016-1143-1
Kanzawa-Lee GA, Larson JL, Resnicow K, Smith EML (2020) Exercise effects on chemotherapy-
induced peripheral neuropathy: a comprehensive integrative review. Cancer Nurs 43:E172–
e185. https://doi.org/10.1097/ncc.0000000000000801
Karczewski KJ, Snyder MP (2018) Integrative omics for health and disease. Nat Rev Genet 19:299–
310. https://doi.org/10.1038/nrg.2018.4
Kenkhuis MF, van der Linden BWA, Breedveld-Peters JJL, Koole JL, van Roekel EH, Breukink
SO, Mols F, Weijenberg MP, Bours MJL (2021) Associations of the dietary World Cancer
Research Fund/American Institute for Cancer Research (WCRF/AICR) recommendations with
patient-reported outcomes in colorectal cancer survivors 2-10 years post-diagnosis: a cross-
sectional analysis. Br J Nutr 125:1188–1200. https://doi.org/10.1017/s0007114520003487
Kerckhove N, Collin A, Condé S, Chaleteix C, Pezet D, Balayssac D (2017) Long-term effects,
pathophysiological mechanisms, and risk factors of chemotherapy-induced peripheral
neuropathies: a comprehensive literature review. Front Pharmacol 8:86. https://doi.org/10.
3389/fphar.2017.00086
Khodorova NV, Rutledge DN, Oberli M, Mathiron D, Marcelo P, Benamouzig R, Tomé D,
Gaudichon C, Pilard S (2019) Urinary metabolomics profiles associated to bovine meat inges-
tion in humans. Mol Nutr Food Res 63:e1700834. https://doi.org/10.1002/mnfr.201700834
Kim S, Jang WJ, Yu H, Kim J, Lee SK, Jeong CH, Lee S (2020) Revealing metabolic perturbation
following heavy methamphetamine abuse by human hair metabolomics and network analysis.
Int J Mol Sci 21. https://doi.org/10.3390/ijms21176041
Kimata M, Shichijo M, Miura T, Serizawa I, Inagaki N, Nagai H (2000) Effects of luteolin,
quercetin and baicalein on immunoglobulin E-mediated mediator release from human cultured
mast cells. Clin Exp Allergy 30:501–508. https://doi.org/10.1046/j.1365-2222.2000.00768.x
Kirwan JA, Brennan L, Broadhurst D, Fiehn O, Cascante M, Dunn WB, Schmidt MA, Velagapudi
V (2018) Preanalytical processing and biobanking procedures of biological samples for
metabolomics research: a white paper, community perspective (for "Precision Medicine and
Pharmacometabolomics Task Group" – the metabolomics society initiative). Clin Chem 64:
1158–1182. https://doi.org/10.1373/clinchem.2018.287045
Kochlik B, Gerbracht C, Grune T, Weber D (2018) The influence of dietary habits and meat
consumption on plasma 3-methylhistidine – a potential marker for muscle protein turnover. Mol
Nutr Food Res 62:e1701062. https://doi.org/10.1002/mnfr.201701062
Koenig RJ, Peterson CM, Jones RL, Saudek C, Lehrman M, Cerami A (1976) Correlation of
glucose regulation and hemoglobin AIc in diabetes mellitus. N Engl J Med 295:417–420.
https://doi.org/10.1056/nejm197608192950804
Kolb NA, Smith AG, Singleton JR, Beck SL, Stoddard GJ, Brown S, Mooney K (2016) The
association of chemotherapy-induced peripheral neuropathy symptoms and the risk of falling.
JAMA Neurol 73:860–866. https://doi.org/10.1001/jamaneurol.2016.0383
Kölker S (2018) Metabolism of amino acid neurotransmitters: the synaptic disorder underlying
inherited metabolic diseases. J Inherit Metab Dis 41:1055–1063. https://doi.org/10.1007/
s10545-018-0201-4
Kramer R, Bielawski J, Kistner-Griffin E, Othman A, Alecu I, Ernst D, Kornhauser D,
Hornemann T, Spassieva S (2015) Neurotoxic 1-deoxysphingolipids and paclitaxel-induced
peripheral neuropathy. FASEB J 29:4461–4472. https://doi.org/10.1096/fj.15-272567
Kristoff CJ, Bwanali L, Veltri LM, Gautam GP, Rutto PK, Newton EO, Holland LA (2020)
Challenging bioanalyses with capillary electrophoresis. Anal Chem 92:49–66. https://doi.org/
10.1021/acs.analchem.9b04718
Kumar A, Mittal R (2017) Nrf2: a potential therapeutic target for diabetic neuropathy.
Inflammopharmacology 25:393–402. https://doi.org/10.1007/s10787-017-0339-y
Lee JJ, Swain SM (2006) Peripheral neuropathy induced by microtubule-stabilizing agents. J Clin
Oncol 24:1633–1642. https://doi.org/10.1200/jco.2005.04.0543
Chemotherapy-Induced Peripheral Neuropathy 331

Li Y, Zhang H, Kosturakis AK, Cassidy RM, Zhang H, Kennamer-Chapman RM, Jawad AB,
Colomand CM, Harrison DS, Dougherty PM (2015) MAPK signaling downstream to TLR4
contributes to paclitaxel-induced peripheral neuropathy. Brain Behav Immun 49:255–266.
https://doi.org/10.1016/j.bbi.2015.06.003
Lin HT, Cheng ML, Lo CJ, Lin G, Lin SF, Yeh JT, Ho HY, Lin JR, Liu FC (2019) (1)H nuclear
magnetic resonance (NMR)-based cerebrospinal fluid and plasma metabolomic analysis in type
2 diabetic patients and risk prediction for diabetic microangiopathy. J Clin Med 8. https://doi.
org/10.3390/jcm8060874
Lloyd AJ, Favé G, Beckmann M, Lin W, Tailliart K, Xie L, Mathers JC, Draper J (2011) Use of
mass spectrometry fingerprinting to identify urinary metabolites after consumption of specific
foods. Am J Clin Nutr 94:981–991. https://doi.org/10.3945/ajcn.111.017921
Loprinzi CL, Lacchetti C, Bleeker J, Cavaletti G, Chauhan C, Hertz DL, Kelley MR, Lavino A,
Lustberg MB, Paice JA, Schneider BP, Lavoie Smith EM, Smith ML, Smith TJ, Wagner-
Johnston N, Hershman DL (2020) Prevention and management of chemotherapy-induced
peripheral neuropathy in survivors of adult cancers: ASCO guideline update. J Clin Oncol 38:
3325–3348. https://doi.org/10.1200/jco.20.01399
Ludolph AC, Ullrich K, Nedjat S, Masur H, Bick U (1992) Neurological outcome in 22 treated
adolescents with hyperphenylalaninemia. A clinical and electrophysiological study. Acta
Neurol Scand 85:243–248. https://doi.org/10.1111/j.1600-0404.1992.tb04039.x
Madhavi YV, Gaikwad N, Yerra VG, Kalvala AK, Nanduri S, Kumar A (2019) Targeting AMPK in
diabetes and diabetic complications: energy homeostasis, autophagy and mitochondrial health.
Curr Med Chem 26:5207–5229. https://doi.org/10.2174/0929867325666180406120051
Maekawa K, Ri M, Nakajima M, Sekine A, Ueda R, Tohkin M, Miyata N, Saito Y, Iida S (2019)
Serum lipidomics for exploring biomarkers of bortezomib therapy in patients with multiple
myeloma. Cancer Sci 110:3267–3274. https://doi.org/10.1111/cas.14178
Mao-Ying QL, Kavelaars A, Krukowski K, Huo XJ, Zhou W, Price TJ, Cleeland C, Heijnen CJ
(2014) The anti-diabetic drug metformin protects against chemotherapy-induced peripheral
neuropathy in a mouse model. PLoS One 9:e100701. https://doi.org/10.1371/journal.pone.
0100701
Mathew AV, Jaiswal M, Ang L, Michailidis G, Pennathur S, Pop-Busui R (2019) Impaired amino
acid and TCA metabolism and cardiovascular autonomic neuropathy progression in type
1 diabetes. Diabetes 68: 2035–2044. doi: https://doi.org/10.2337/db19-0145
McGale P, Taylor C, Correa C, Cutter D, Duane F, Ewertz M, Gray R, Mannu G, Peto R, Whelan T,
Wang Y, Wang Z, Darby S (2014) Effect of radiotherapy after mastectomy and axillary surgery
on 10-year recurrence and 20-year breast cancer mortality: meta-analysis of individual patient
data for 8135 women in 22 randomised trials. Lancet 383:2127–2135. https://doi.org/10.1016/
s0140-6736(14)60488-8
Miao H, Xu J, Xu D, Ma X, Zhao X, Liu L (2019) Nociceptive behavior induced by chemothera-
peutic paclitaxel and beneficial role of antioxidative pathways. Physiol Res 68:491–500. https://
doi.org/10.33549/physiolres.933939
Miltenburg NC, Boogerd W (2014) Chemotherapy-induced neuropathy: a comprehensive survey.
Cancer Treat Rev 40:872–882. https://doi.org/10.1016/j.ctrv.2014.04.004
Mitry P, Wawro N, Rohrmann S, Giesbertz P, Daniel H, Linseisen J (2019) Plasma concentrations
of anserine, carnosine and pi-methylhistidine as biomarkers of habitual meat consumption. Eur J
Clin Nutr 73:692–702. https://doi.org/10.1038/s41430-018-0248-1
Mols F, Beijers AJ, Vreugdenhil G, Verhulst A, Schep G, Husson O (2015) Chemotherapy-induced
peripheral neuropathy, physical activity and health-related quality of life among colorectal
cancer survivors from the PROFILES registry. J Cancer Surviv 9:512–522. https://doi.org/10.
1007/s11764-015-0427-1
Mols F, Beijers T, Vreugdenhil G, van de Poll-Franse L (2014) Chemotherapy-induced peripheral
neuropathy and its association with quality of life: a systematic review. Support Care Cancer 22:
2261–2269. https://doi.org/10.1007/s00520-014-2255-7
332 C.-S. Chen and D. L. Hertz

Mongiovi JM, Zirpoli GR, Cannioto R, Sucheston-Campbell LE, Hershman DL, Unger JM, Moore
HCF, Stewart JA, Isaacs C, Hobday TJ, Salim M, Hortobagyi GN, Gralow JR, Thomas Budd G,
Albain KS, Ambrosone CB, McCann SE (2018) Associations between self-reported diet during
treatment and chemotherapy-induced peripheral neuropathy in a cooperative group trial
(S0221). Breast Cancer Res 20:146. https://doi.org/10.1186/s13058-018-1077-9
Morvan D, Demidem A (2018) NMR metabolomics of fibroblasts with inherited mitochondrial
complex I mutation reveals treatment-reversible lipid and amino acid metabolism alterations.
Metabolomics 14:55. https://doi.org/10.1007/s11306-018-1345-9
Mustafa Ali M, Moeller M, Rybicki L, Moore HCF (2017) Long-term peripheral neuropathy
symptoms in breast cancer survivors. Breast Cancer Res Treat 166:519–526. https://doi.org/
10.1007/s10549-017-4437-8
Nagano H, Sanai H, Muraoka M, Takagi K (2012) Efficacy of lafutidine, a histamine H2-receptor
antagonist, for taxane-induced peripheral neuropathy in patients with gynecological
malignancies. Gynecol Oncol 127:172–174. https://doi.org/10.1016/j.ygyno.2012.06.029
Napoli E, Schneider A, Wang JY, Trivedi A, Carrillo NR, Tassone F, Rogawski M, Hagerman RJ,
Giulivi C (2019) Allopregnanolone treatment improves plasma metabolomic profile associated
with GABA metabolism in fragile X-associated tremor/ataxia syndrome: a pilot study. Mol
Neurobiol 56:3702–3713. https://doi.org/10.1007/s12035-018-1330-3
Neuenschwander M, Ballon A, Weber KS, Norat T, Aune D, Schwingshackl L, Schlesinger S
(2019) Role of diet in type 2 diabetes incidence: umbrella review of meta-analyses of prospec-
tive observational studies. BMJ 366:l2368. https://doi.org/10.1136/bmj.l2368
Ng T, Chan M, Khor CC, Ho HK, Chan A (2014) The genetic variants underlying breast cancer
treatment-induced chronic and late toxicities: a systematic review. Cancer Treat Rev 40:1199–
1214. https://doi.org/10.1016/j.ctrv.2014.10.001
Nishida K, Takeuchi K, Hosoda A, Sugano S, Morisaki E, Ohishi A, Nagasawa K (2018)
Ergothioneine ameliorates oxaliplatin-induced peripheral neuropathy in rats. Life Sci 207:
516–524. https://doi.org/10.1016/j.lfs.2018.07.006
Nordlie RC, Foster JD, Lange AJ (1999) Regulation of glucose production by the liver. Annu Rev
Nutr 19:379–406. https://doi.org/10.1146/annurev.nutr.19.1.379
O'Gorman A, Brennan L (2017) The role of metabolomics in determination of new dietary
biomarkers. Proc Nutr Soc 76:295–302. https://doi.org/10.1017/s0029665116002974
Onodera S, Shibata M, Tanaka M, Inaba N, Arai Y, Aoyama M, Lee B, Yamaura T (1999)
Gastroprotective mechanism of lafutidine, a novel anti-ulcer drug with histamine H2-receptor
antagonistic activity. Arzneimittelforschung 49:519–526. https://doi.org/10.1055/s-
0031-1300454
Onodera S, Shibata M, Tanaka M, Inaba N, Yamaura T, Ohnishi H (1995) Gastroprotective activity
of FRG-8813, a novel histamine H2-receptor antagonist, in rats. Jpn J Pharmacol 68:161–173.
https://doi.org/10.1254/jjp.68.161
Oza MJ, Laddha AP, Gaikwad AB, Mulay SR, Kulkarni YA (2021) Role of dietary modifications in
the management of type 2 diabetic complications. Pharmacol Res 168:105602. https://doi.org/
10.1016/j.phrs.2021.105602
Pallister T, Jennings A, Mohney RP, Yarand D, Mangino M, Cassidy A, MacGregor A, Spector
TD, Menni C (2016) Characterizing blood metabolomics profiles associated with self-reported
food intakes in female twins. PLoS One 11:e0158568. https://doi.org/10.1371/journal.pone.
0158568
Park JE, Lim HR, Kim JW, Shin KH (2018) Metabolite changes in risk of type 2 diabetes mellitus in
cohort studies: a systematic review and meta-analysis. Diabetes Res Clin Pract 140:216–227.
https://doi.org/10.1016/j.diabres.2018.03.045
Paschen W (1992) Polyamine metabolism in different pathological states of the brain. Mol Chem
Neuropathol 16:241–271. https://doi.org/10.1007/bf03159973
Pearce FL, Befus AD, Bienenstock J (1984) Mucosal mast cells. III. Effect of quercetin and other
flavonoids on antigen-induced histamine secretion from rat intestinal mast cells. J Allergy Clin
Immunol 73:819–823. https://doi.org/10.1016/0091-6749(84)90453-6
Chemotherapy-Induced Peripheral Neuropathy 333

Peterson JW, Boldogh I, Popov VL, Saini SS, Chopra AK (1998) Anti-inflammatory and
antisecretory potential of histidine in salmonella-challenged mouse small intestine. Lab Invest
78:523–534
Petrovchich I, Kober KM, Wagner L, Paul SM, Abrams G, Chesney MA, Topp K, Smoot B,
Schumacher M, Conley YP, Hammer M, Levine JD, Miaskowski C (2019) Deleterious effects
of higher body mass index on subjective and objective measures of chemotherapy-induced
peripheral neuropathy in cancer survivors. J Pain Symptom Manage 58:252–263. https://doi.
org/10.1016/j.jpainsymman.2019.04.029
Pini A, Obara I, Battell E, Chazot PL, Rosa AC (2016) Histamine in diabetes: is it time to
reconsider? Pharmacol Res 111:316–324. https://doi.org/10.1016/j.phrs.2016.06.021
Plasmati R, Pastorelli F, Cavo M, Petracci E, Zamagni E, Tosi P, Cangini D, Tacchetti P, Salvi F,
Bartolomei I, Michelucci R, Tassinari CA (2007) Neuropathy in multiple myeloma treated with
thalidomide: a prospective study. Neurology 69:573–581. https://doi.org/10.1212/01.wnl.
0000267271.18475.fe
Pop-Busui R, Lu J, Lopes N, Jones TL (2009) Prevalence of diabetic peripheral neuropathy and
relation to glycemic control therapies at baseline in the BARI 2D cohort. J Peripher Nerv Syst
14:1–13. https://doi.org/10.1111/j.1529-8027.2009.00200.x
Powell HC, Rosoff J, Myers RR (1985) Microangiopathy in human diabetic neuropathy. Acta
Neuropathol 68:295–305. https://doi.org/10.1007/bf00690832
Price TJ, Das V, Dussor G (2016) Adenosine monophosphate-activated protein kinase (AMPK)
activators for the prevention, treatment and potential reversal of pathological pain. Curr Drug
Targets 17:908–920. https://doi.org/10.2174/1389450116666151102095046
Qin XY, Akanuma H, Wei F, Nagano R, Zeng Q, Imanishi S, Ohsako S, Yoshinaga J, Yonemoto J,
Tanokura M, Sone H (2012) Effect of low-dose thalidomide on dopaminergic neuronal differ-
entiation of human neural progenitor cells: a combined study of metabolomics and morphologi-
cal analysis. Neurotoxicology 33:1375–1380. https://doi.org/10.1016/j.neuro.2012.08.016
Quasthoff S, Hartung HP (2002) Chemotherapy-induced peripheral neuropathy. J Neurol 249:9–17.
https://doi.org/10.1007/pl00007853
Raskin J, Smith TR, Wong K, Pritchett YL, D'Souza DN, Iyengar S, Wernicke JF (2006)
Duloxetine versus routine care in the long-term management of diabetic peripheral neuropathic
pain. J Palliat Med 9:29–40. https://doi.org/10.1089/jpm.2006.9.29
Rattray NJW, Deziel NC, Wallach JD, Khan SA, Vasiliou V, Ioannidis JPA, Johnson CH (2018)
Beyond genomics: understanding exposotypes through metabolomics. Hum Genomics 12:4.
https://doi.org/10.1186/s40246-018-0134-x
Razazian N, Baziyar M, Moradian N, Afshari D, Bostani A, Mahmoodi M (2014) Evaluation of the
efficacy and safety of pregabalin, venlafaxine, and carbamazepine in patients with painful
diabetic peripheral neuropathy. A randomized, double-blind trial. Neurosciences (Riyadh) 19:
192–198
Remmington T, Smith S (2021) Tyrosine supplementation for phenylketonuria. Cochrane Database
Syst Rev 1:Cd001507. https://doi.org/10.1002/14651858.CD001507.pub4
Rivat C, Richebé P, Laboureyras E, Laulin JP, Havouis R, Noble F, Moulinoux JP, Simonnet G
(2008) Polyamine deficient diet to relieve pain hypersensitivity. Pain 137:125–137. https://doi.
org/10.1016/j.pain.2007.08.021
Roberts LD, Souza AL, Gerszten RE, Clish CB (2012) Targeted metabolomics. Curr Protoc Mol
Biol. Chapter 30: Unit 30.2.1-24. https://doi.org/10.1002/0471142727.mb3002s98
Romanovsky AA, Almeida MC, Garami A, Steiner AA, Norman MH, Morrison SF, Nakamura K,
Burmeister JJ, Nucci TB (2009) The transient receptor potential vanilloid-1 channel in thermo-
regulation: a thermosensor it is not. Pharmacol Rev 61:228–261. https://doi.org/10.1124/pr.109.
001263
Rondón LJ, Farges MC, Davin N, Sion B, Privat AM, Vasson MP, Eschalier A, Courteix C (2018)
L-arginine supplementation prevents allodynia and hyperalgesia in painful diabetic neuropathic
rats by normalizing plasma nitric oxide concentration and increasing plasma agmatine concen-
tration. Eur J Nutr 57:2353–2363. https://doi.org/10.1007/s00394-017-1508-x
334 C.-S. Chen and D. L. Hertz

Rosa AC, Fantozzi R (2013) The role of histamine in neurogenic inflammation. Br J Pharmacol
170:38–45. https://doi.org/10.1111/bph.12266
Rowbotham MC, Goli V, Kunz NR, Lei D (2004) Venlafaxine extended release in the treatment of
painful diabetic neuropathy: a double-blind, placebo-controlled study. Pain 110:697–706.
https://doi.org/10.1016/j.pain.2004.05.010
Roy Chowdhury SK, Smith DR, Saleh A, Schapansky J, Marquez A, Gomes S, Akude E,
Morrow D, Calcutt NA, Fernyhough P (2012) Impaired adenosine monophosphate-activated
protein kinase signalling in dorsal root ganglia neurons is linked to mitochondrial dysfunction
and peripheral neuropathy in diabetes. Brain 135:1751–1766. https://doi.org/10.1093/brain/
aws097
Schneider BP, Shen F, Jiang G, O'Neill A, Radovich M, Li L, Gardner L, Lai D, Foroud T, Sparano
JA, Sledge GW Jr, Miller KD (2017) Impact of genetic ancestry on outcomes in ECOG-ACRIN-
E5103. JCO Precis Oncol 2017. https://doi.org/10.1200/po.17.00059
Schwingshackl L, Hoffmann G, Lampousi AM, Knüppel S, Iqbal K, Schwedhelm C, Bechthold A,
Schlesinger S, Boeing H (2017) Food groups and risk of type 2 diabetes mellitus: a systematic
review and meta-analysis of prospective studies. Eur J Epidemiol 32:363–375. https://doi.org/
10.1007/s10654-017-0246-y
Seretny M, Currie GL, Sena ES, Ramnarine S, Grant R, MacLeod MR, Colvin LA, Fallon M (2014)
Incidence, prevalence, and predictors of chemotherapy-induced peripheral neuropathy: a sys-
tematic review and meta-analysis. Pain 155:2461–2470. https://doi.org/10.1016/j.pain.2014.
09.020
Shaw S, Uniyal A, Gadepalli A, Tiwari V, Belinskaia DA, Shestakova NN, Venugopala KN, Deb
PK, Tiwari V (2020) Adenosine receptor signalling: probing the potential pathways for the
ministration of neuropathic pain. Eur J Pharmacol 889:173619. https://doi.org/10.1016/j.ejphar.
2020.173619
Shigematsu N, Kawashiri T, Kobayashi D, Shimizu S, Mine K, Hiromoto S, Uchida M, Egashira N,
Shimazoe T (2020) Neuroprotective effect of alogliptin on oxaliplatin-induced peripheral
neuropathy in vivo and in vitro. Sci Rep 10:6734. https://doi.org/10.1038/s41598-020-62738-w
Shin SY, Fauman EB, Petersen AK, Krumsiek J, Santos R, Huang J, Arnold M, Erte I, Forgetta V,
Yang TP, Walter K, Menni C, Chen L, Vasquez L, Valdes AM, Hyde CL, Wang V, Ziemek D,
Roberts P, Xi L, Grundberg E, Waldenberger M, Richards JB, Mohney RP, Milburn MV, John
SL, Trimmer J, Theis FJ, Overington JP, Suhre K, Brosnan MJ, Gieger C, Kastenmüller G,
Spector TD, Soranzo N (2014) An atlas of genetic influences on human blood metabolites. Nat
Genet 46:543–550. https://doi.org/10.1038/ng.2982
Shrikanth CB, Nandini CD (2020) AMPK in microvascular complications of diabetes and the
beneficial effects of AMPK activators from plants. Phytomedicine 73:152808. https://doi.org/
10.1016/j.phymed.2018.12.031
Sjölin J, Hjort G, Friman G, Hambraeus L (1987) Urinary excretion of 1-methylhistidine: a
qualitative indicator of exogenous 3-methylhistidine and intake of meats from various sources.
Metabolism 36:1175–1184. https://doi.org/10.1016/0026-0495(87)90245-9
Smith AG, Russell J, Feldman EL, Goldstein J, Peltier A, Smith S, Hamwi J, Pollari D, Bixby B,
Howard J, Singleton JR (2006) Lifestyle intervention for pre-diabetic neuropathy. Diabetes Care
29:1294–1299. https://doi.org/10.2337/dc06-0224
Smith AG, Singleton JR (2013) Obesity and hyperlipidemia are risk factors for early diabetic
neuropathy. J Diabetes Complications 27:436–442. https://doi.org/10.1016/j.jdiacomp.2013.
04.003
Smith EM, Pang H, Cirrincione C, Fleishman S, Paskett ED, Ahles T, Bressler LR, Fadul CE,
Knox C, Le-Lindqwister N, Gilman PB, Shapiro CL (2013) Effect of duloxetine on pain,
function, and quality of life among patients with chemotherapy-induced painful peripheral
neuropathy: a randomized clinical trial. JAMA 309:1359–1367. https://doi.org/10.1001/jama.
2013.2813
Chemotherapy-Induced Peripheral Neuropathy 335

Song TY, Chen CL, Liao JW, Ou HC, Tsai MS (2010) Ergothioneine protects against neuronal
injury induced by cisplatin both in vitro and in vivo. Food Chem Toxicol 48:3492–3499. https://
doi.org/10.1016/j.fct.2010.09.030
Stankovic JSK, Selakovic D, Mihailovic V, Rosic G (2020) Antioxidant supplementation in the
treatment of neurotoxicity induced by platinum-based chemotherapeutics – a review. Int J Mol
Sci 21. https://doi.org/10.3390/ijms21207753
Subramaniam S, O'Connor MJ, Masukawa LM, McGonigle P (1994) Polyamine effects on the
NMDA receptor in human brain. Exp Neurol 130:323–330. https://doi.org/10.1006/exnr.1994.
1210
Sucheston-Campbell LE, Clay-Gilmour AI, Barlow WE, Budd GT, Stram DO, Haiman CA,
Sheng X, Yan L, Zirpoli G, Yao S, Jiang C, Owzar K, Hershman D, Albain KS, Hayes DF,
Moore HC, Hobday TJ, Stewart JA, Rizvi A, Isaacs C, Salim M, Gralow JR, Hortobagyi GN,
Livingston RB, Kroetz DL, Ambrosone CB (2018) Genome-wide meta-analyses identifies
novel taxane-induced peripheral neuropathy-associated loci. Pharmacogenet Genomics 28:49–
55. https://doi.org/10.1097/fpc.0000000000000318
Sun Y, Gao HY, Fan ZY, He Y, Yan YX (2020) Metabolomics signatures in type 2 diabetes: a
systematic review and integrative analysis. J Clin Endocrinol Metab 105. https://doi.org/10.
1210/clinem/dgz240
Sun Y, Kim JH, Vangipuram K, Hayes DF, Smith EML, Yeomans L, Henry NL, Stringer KA,
Hertz DL (2018) Pharmacometabolomics reveals a role for histidine, phenylalanine, and
threonine in the development of paclitaxel-induced peripheral neuropathy. Breast Cancer Res
Treat 171:657–666. https://doi.org/10.1007/s10549-018-4862-3
Terrazzino S, Argyriou AA, Cargnin S, Antonacopoulou AG, Briani C, Bruna J, Velasco R,
Alberti P, Campagnolo M, Lonardi S, Cortinovis D, Cazzaniga M, Santos C, Kalofonos HP,
Canonico PL, Genazzani AA, Cavaletti G (2015) Genetic determinants of chronic oxaliplatin-
induced peripheral neurotoxicity: a genome-wide study replication and meta-analysis. J
Peripher Nerv Syst 20:15–23. https://doi.org/10.1111/jns.12110
Timmins HC, Mizrahi D, Li T, Kiernan MC, Goldstein D, Park SB (2021) Metabolic and lifestyle
risk factors for chemotherapy-induced peripheral neuropathy in taxane and platinum-treated
patients: a systematic review. J Cancer Surviv. https://doi.org/10.1007/s11764-021-00988-x
Traynelis SF, Hartley M, Heinemann SF (1995) Control of proton sensitivity of the NMDA receptor
by RNA splicing and polyamines. Science 268:873–876. https://doi.org/10.1126/science.
7754371
Tsukaguchi M, Shibano M, Matsuura A, Mukai S (2013) The protective effects of lafutidine for
bortezomib induced peripheral neuropathy. J Blood Med 4:81–85. https://doi.org/10.2147/jbm.
S44127
Umeda M, Fujita A, Nishiwaki H, Takeuchi K (1999) Effect of lafutidine, a novel histamine
H2-receptor antagonist, on monochloramine-induced gastric lesions in rats: role of capsaicin-
sensitive sensory neurons. J Gastroenterol Hepatol 14:859–865. https://doi.org/10.1046/j.
1440-1746.1999.01976.x
Urpi-Sarda M, Almanza-Aguilera E, Llorach R, Vázquez-Fresno R, Estruch R, Corella D, Sorli JV,
Carmona F, Sanchez-Pla A, Salas-Salvadó J, Andres-Lacueva C (2019) Non-targeted
metabolomic biomarkers and metabotypes of type 2 diabetes: a cross-sectional study of
PREDIMED trial participants. Diabetes Metab 45:167–174. https://doi.org/10.1016/j.diabet.
2018.02.006
Vázquez-Fresno R, Llorach R, Urpi-Sarda M, Lupianez-Barbero A, Estruch R, Corella D, Fitó M,
Arós F, Ruiz-Canela M, Salas-Salvadó J, Andres-Lacueva C (2015) Metabolomic pattern
analysis after mediterranean diet intervention in a nondiabetic population: a 1- and 3-year
follow-up in the PREDIMED study. J Proteome Res 14:531–540. https://doi.org/10.1021/
pr5007894
Verma P, Devaraj J, Skiles JL, Sajdyk T, Ho RH, Hutchinson R, Wells E, Li L, Renbarger J,
Cooper B, Ramkrishna D (2020) A metabolomics approach for early prediction of vincristine-
induced peripheral neuropathy. Sci Rep 10:9659. https://doi.org/10.1038/s41598-020-66815-y
336 C.-S. Chen and D. L. Hertz

Wallace MS, Rowbotham MC, Katz NP, Dworkin RH, Dotson RM, Galer BS, Rauck RL, Backonja
MM, Quessy SN, Meisner PD (2002) A randomized, double-blind, placebo-controlled trial of a
glycine antagonist in neuropathic pain. Neurology 59:1694–1700. https://doi.org/10.1212/01.
wnl.0000036273.98213.34
Wang J, Zhang XS, Tao R, Zhang J, Liu L, Jiang YH, Ma SH, Song LX, Xia LJ (2017)
Upregulation of CX3CL1 mediated by NF-κB activation in dorsal root ganglion contributes
to peripheral sensitization and chronic pain induced by oxaliplatin administration. Mol Pain 13:
1744806917726256. https://doi.org/10.1177/1744806917726256
Wang S, Kobayashi K, Kogure Y, Yamanaka H, Yamamoto S, Yagi H, Noguchi K, Dai Y (2018)
Negative regulation of TRPA1 by AMPK in primary sensory neurons as a potential mechanism
of painful diabetic neuropathy. Diabetes 67:98–109. https://doi.org/10.2337/db17-0503
Wang TJ, Larson MG, Vasan RS, Cheng S, Rhee EP, McCabe E, Lewis GD, Fox CS, Jacques PF,
Fernandez C, O'Donnell CJ, Carr SA, Mootha VK, Florez JC, Souza A, Melander O, Clish CB,
Gerszten RE (2011) Metabolite profiles and the risk of developing diabetes. Nat Med 17:448–
453. https://doi.org/10.1038/nm.2307
Wei H, Viisanen H, You HJ, Pertovaara A (2016) Spinal histamine in attenuation of mechanical
hypersensitivity in the spinal nerve ligation-induced model of experimental neuropathy. Eur J
Pharmacol 772:1–10. https://doi.org/10.1016/j.ejphar.2015.12.039
Wiggin TD, Sullivan KA, Pop-Busui R, Amato A, Sima AA, Feldman EL (2009) Elevated
triglycerides correlate with progression of diabetic neuropathy. Diabetes 58:1634–1640.
https://doi.org/10.2337/db08-1771
Wilkins J, Sakrikar D, Petterson XM, Lanza IR, Trushina E (2019) A comprehensive protocol for
multiplatform metabolomics analysis in patient-derived skin fibroblasts. Metabolomics 15:83.
https://doi.org/10.1007/s11306-019-1544-z
Winkels RM, van Brakel L, van Baar H, Beelman RB, van Duijnhoven FJB, Geijsen A, van
Halteren HK, Hansson BME, Richie JP, Sun D, Wesselink E, van Zutphen M, Kampman E,
Kok DE (2020) Are ergothioneine levels in blood associated with chronic peripheral neuropathy
in colorectal cancer patients who underwent chemotherapy? Nutr Cancer 72:451–459. https://
doi.org/10.1080/01635581.2019.1637005
Winters-Stone KM, Horak F, Jacobs PG, Trubowitz P, Dieckmann NF, Stoyles S, Faithfull S (2017)
Falls, functioning, and disability among women with persistent symptoms of chemotherapy-
induced peripheral neuropathy. J Clin Oncol 35:2604–2612. https://doi.org/10.1200/jco.2016.
71.3552
Wishart DS (2016) Emerging applications of metabolomics in drug discovery and precision
medicine. Nat Rev Drug Discov 15:473–484. https://doi.org/10.1038/nrd.2016.32
Wishart DS (2019) Metabolomics for investigating physiological and pathophysiological processes.
Physiol Rev 99:1819–1875. https://doi.org/10.1152/physrev.00035.2018
Wu FZ, Xu WJ, Deng B, Liu SD, Deng C, Wu MY, Gao Y, Jia LQ (2018) Wen-Luo-Tong
decoction attenuates paclitaxel-induced peripheral neuropathy by regulating linoleic acid and
glycerophospholipid metabolism pathways. Front Pharmacol 9:956. https://doi.org/10.3389/
fphar.2018.00956
Xu X, Luo P, Wang Y, Cui Y, Miao L (2013) Nuclear factor (erythroid-derived 2)-like 2 (NFE2L2)
is a novel therapeutic target for diabetic complications. J Int Med Res 41:13–19. https://doi.org/
10.1177/0300060513477004
Yagihashi S, Matsunaga M (1979) Ultrastructural pathology of peripheral nerves in patients with
diabetic neuropathy. Tohoku J Exp Med 129:357–366. https://doi.org/10.1620/tjem.129.357
Yamamoto H, Horie S, Uchida M, Tsuchiya S, Murayama T, Watanabe K (2001) Effects of
vanilloid receptor agonists and antagonists on gastric antral ulcers in rats. Eur J Pharmacol
432:203–210. https://doi.org/10.1016/s0014-2999(01)01481-9
Yang Y, Luo L, Cai X, Fang Y, Wang J, Chen G, Yang J, Zhou Q, Sun X, Cheng X, Yan H, Lu W,
Hu C, Cao P (2018) Nrf2 inhibits oxaliplatin-induced peripheral neuropathy via protection of
mitochondrial function. Free Radic Biol Med 120:13–24. https://doi.org/10.1016/j.
freeradbiomed.2018.03.007
Chemotherapy-Induced Peripheral Neuropathy 337

Yehia R, Saleh S, El Abhar H, Saad AS, Schaalan M (2019) L-carnosine protects against
oxaliplatin-induced peripheral neuropathy in colorectal cancer patients: a perspective on
targeting Nrf-2 and NF-κB pathways. Toxicol Appl Pharmacol 365:41–50. https://doi.org/10.
1016/j.taap.2018.12.015
Yerra VG, Areti A, Kumar A (2017) Adenosine monophosphate-activated protein kinase abates
hyperglycaemia-induced neuronal injury in experimental models of diabetic neuropathy: effects
on mitochondrial biogenesis, autophagy and neuroinflammation. Mol Neurobiol 54:2301–2312.
https://doi.org/10.1007/s12035-016-9824-3
Yin X, Gibbons H, Rundle M, Frost G, McNulty BA, Nugent AP, Walton J, Flynn A, Gibney MJ,
Brennan L (2017) Estimation of chicken intake by adults using metabolomics-derived markers.
J Nutr 147:1850–1857. https://doi.org/10.3945/jn.117.252197
Yu J, Tang YY, Wang RR, Lou GD, Hu TT, Hou WW, Yue JX, Ohtsu H, Shi LY, Zhang SH, Chen
Z (2016) A critical time window for the analgesic effect of central histamine in the partial sciatic
ligation model of neuropathic pain. J Neuroinflammation 13:163. https://doi.org/10.1186/
s12974-016-0637-0
Zajączkowska R, Kocot-Kępska M, Leppert W, Wrzosek A, Mika J, Wordliczek J (2019)
Mechanisms of chemotherapy-induced peripheral neuropathy. Int J Mol Sci 20:1451. https://
doi.org/10.3390/ijms20061451
Zhang W, Ramautar R (2021) CE-MS for metabolomics: developments and applications in the
period 2018-2020. Electrophoresis 42:381–401. https://doi.org/10.1002/elps.202000203
Zhao XH, Wen ZM, Meredith CN, Matthews DE, Bier DM, Young VR (1986) Threonine kinetics
at graded threonine intakes in young men. Am J Clin Nutr 43:795–802. https://doi.org/10.1093/
ajcn/43.5.795
Zimmerman C, Atherton PJ, Pachman D, Seisler D, Wagner-Johnston N, Dakhil S, Lafky JM,
Qin R, Grothey A, Loprinzi CL (2016) MC11C4: a pilot randomized, placebo-controlled,
double-blind study of venlafaxine to prevent oxaliplatin-induced neuropathy. Support Care
Cancer 24:1071–1078. https://doi.org/10.1007/s00520-015-2876-5
Zirpoli GR, McCann SE, Sucheston-Campbell LE, Hershman DL, Ciupak G, Davis W, Unger JM,
Moore HCF, Stewart JA, Isaacs C, Hobday TJ, Salim M, Hortobagyi GN, Gralow JR, Budd GT,
Albain KS, Ambrosone CB (2017) Supplement use and chemotherapy-induced peripheral
neuropathy in a cooperative group trial (S0221): the DELCaP study. J Natl Cancer Inst 109.
https://doi.org/10.1093/jnci/djx098
Metabolomics of Respiratory Diseases

Subhabrata Moitra, Arghya Bandyopadhyay, and Paige Lacy

Contents
1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 340
2 Respiratory Diseases with Metabolomic Signatures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 343
2.1 Asthma . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 343
2.2 Chronic Obstructive Pulmonary Disease (COPD) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 349
2.3 Pneumonia . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 352
2.4 Acute Lung Injury/Acute Respiratory Distress Syndrome (ARDS) . . . . . . . . . . . . . . . . . 352
2.5 Occupational and Environmental Lung Diseases . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 355
3 Metabolomics of Lung Microbiome in Respiratory Diseases . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 357
3.1 Asthma . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 357
3.2 Chronic Obstructive Pulmonary Disease (COPD) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 358
4 Conclusive Remarks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 359
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 360

Abstract
Metabolomics is an expanding field of systems biology that is gaining significant
attention in respiratory research. As a unique approach to understanding and
diagnosing diseases, metabolomics provides a snapshot of all metabolites present
in biological samples such as exhaled breath condensate, bronchoalveolar lavage,
plasma, serum, urine, and other specimens that may be obtained from patients
with respiratory diseases. In this article, we review the rapidly expanding field of
metabolomics in its application to respiratory diseases, including asthma, chronic
obstructive pulmonary disease (COPD), pneumonia, and acute lung injury, along
with its more severe form, adult respiratory disease syndrome. We also discuss

S. Moitra · A. Bandyopadhyay · P. Lacy (✉)

Department of Medicine, Alberta Respiratory Centre (ARC), University of Alberta, Edmonton, AB,
Canada
e-mail: [email protected]; [email protected]; [email protected]

# The Author(s), under exclusive license to Springer Nature Switzerland AG 2022 339
V. Ghini et al. (eds.), Metabolomics and Its Impact on Health and Diseases,
Handbook of Experimental Pharmacology 277, https://doi.org/10.1007/164_2022_614
340 S. Moitra et al.

the potential applications of metabolomics for monitoring exposure to aerosolized

occupational and environmental materials. With the latest advances in our under-
standing of the microbiome, we discuss microbiome-derived metabolites that
arise from the gut and lung in asthma and COPD that have mechanistic
implications for these diseases. Recent literature has suggested that metabolomics
analysis using nuclear magnetic resonance (NMR) and mass spectrometry
(MS) approaches may provide clinicians with the opportunity to identify new
biomarkers that may predict progression to more severe diseases which may be
fatal for many patients each year.

Keywords
Acute lung injury · Acute respiratory distress syndrome (ARDS) · Asthma ·
Chronic obstructive pulmonary disease (COPD) · Pneumonia

1 Introduction

Metabolomics is an established field of systems biology that has generated substan-

tial new findings in respiratory research. The ability of metabolomics to produce a
“snapshot” of small molecules within a given sample from the body provides a
powerful tool for temporal analyses to follow the distribution and concentration of
these molecules (Patti et al. 2012). Small molecules of interest include chemicals
(such as drugs) and metabolites (including waste products of metabolism). These
small molecules are a distinct group of compounds from the larger proteins and
nucleic acids (RNA, DNA), and their measurement provides a valuable complement
to other fields of systems biology (transcriptomics, genomics, proteomics, and
others). Further, metabolomics informs other areas of systems biology as it lies
downstream of proteins, RNA, and genes. Because of its ability to detect small
molecules, metabolomics has the potential to discover novel biomarkers of disease
as well as environmental and occupational exposure (Madsen et al. 2010; Robertson
et al. 2011). To understand the relevance of metabolomics in respiratory diseases, it
is important to establish how the metabolome is defined and how this aligns with
other approaches in systems biology.
The metabolome of an organism reflects events that occur in the proteome,
transcriptome, and genome. Changes in proteins, RNA, and genes result in
alterations of metabolite concentrations in biological fluids and tissues. Perhaps
unsurprisingly, measurement of metabolites in human samples is not a new proce-
dure, since metabolites have been used for millennia to aid in the diagnosis of
disease. For example, diabetes mellitus has been diagnosed since ancient times
based on the taste of glucose, a small molecule metabolite, in urine from patients
with type I diabetes. While this may be an unappetizing and perhaps unsafe practice
these days, this was essentially the first diagnostic test for a metabolite in urine
samples. The recognition that urine contains important biomarkers of disease led to
the development of analytical tools to measure these in a variety of samples from the
Metabolomics of Respiratory Diseases 341

body today. Today, the measurement of small molecules in human samples forms
the basis of clinical chemistry, established to assist health professionals in diagnosis
of illnesses.
Most clinical chemistry tests rely on the measurement of handful of metabolites,
and often these are only qualitative (positive or negative) rather than quantitative
tests. Because most metabolites measured using clinical tests are abundant and not
specific to any one disease, their detection must always be taken into consideration
with other clinical descriptors. Thus, the focus of clinical chemistry on such a small
group of metabolites is a significant limitation that prevents the applicability of
metabolite detection in the specific diagnosis of many diseases.
The limitations of traditional clinical chemistry highlight the advantages of
metabolomics. Recent improvements in the sensitivity and specificity of metabolite
detection using metabolomics have allowed the characterization and quantification
of complex metabolic profiles resulting in concurrent analysis of hundreds of
metabolites in a single sample. Metabolomics seeks to quantitatively assess complex
metabolic patterns in patient samples and is coupled with computational
technologies to allow the interpretation of data in the context of known metabolic
pathways. The complexity of the metabolome in a patient sample is further aug-
mented by the presence of metabolites that derive from the microbiome, which is
present in almost all samples obtained from the human body. The microbiome
generates metabolites that are unique to prokaryotic organisms and may be distin-
guished from the host’s own metabolome, thus providing another possible approach
for enhancing the diagnosis and prognosis of disease.
Despite substantial investments in genome analysis in diseases, genetic mutations
that result in the manifestation of disease are rare. Only 1–2% of disease risk for a
spectrum of conditions including asthma, chronic obstructive pulmonary disease
(COPD), and acute respiratory distress syndrome (ARDS) can be explained by
genetic mutations. Transcriptomic and proteomic analysis has generated more
insight into their potential as biomarkers, but these too have not developed into
standard disease indicators. In contrast, metabolomics and clinical chemistry repro-
ducibly demonstrate that metabolites are highly predictive for a large proportion of
complex diseases (Xia et al. 2013). Samples may be used from a broad range of
sources including saliva, nasal lavage, exhaled breath condensate (EBC), bronchial
washings, sweat, blood (plasma and serum), urine, feces, among others. Examples of
established metabolic biomarkers include glucose for diabetes, as mentioned above,
creatinine to detect kidney disease, cholesterol and triglycerides to evaluate the risk
of developing cardiovascular disease, uric acid for gout detection, and thyroxine for
hypo/hyperthyroidism. There are undoubtedly other metabolites that may be used to
serve as biomarkers in a range of diseases.
These findings indicate that the metabolome is a much more dynamic group of
analytes than the proteome, transcriptome, or genome, as it can change immediately
in response to environmental or physiological changes (Fig. 1) (Wishart 2005). To
appreciate the contribution that metabolomics may make to diagnosis of disease, it is
useful to compare the impact of environmental and physiological impact on proteins,
RNA, and genes. Environmental and physiological changes have negligible impact
342 S. Moitra et al.

(A) The Systems Biology Pyramid

Metabolomics

Metabolome

Environmental Influence
Physiological Influence

Proteomics
Proteome

Genomics
Genome

(B) Systems Biology Time Scales

Response

Metabolomics
Response

Proteomics
Response

Genomics

Time

Fig. 1 The systems biology pyramid and time scales of responses to environmental influences. (a)
Over 25,000 genes have been identified in human genomics, compared with a smaller number of
enzymes and even smaller number of metabolites. The responsiveness to physiological and
environmental insults of each of these components increases as we go from genomics to
metabolomics. (b) While metabolomics shows rapid changes in multiple metabolites in a short
period of time, proteomics shows smaller changes in abundance while genomics shows negligible
changes over the same period

on somatic gene expression, while some transcriptomic and proteomic changes have
been detected. In contrast, metabolomic changes in response to environmental and
physiological factors closely correlate with these events and can be altered within
seconds of exposure (Fig. 1b). Therefore, significant changes in metabolites may be
measurable in samples over far shorter time scales than by other systems biology
Metabolomics of Respiratory Diseases 343

approaches. This allows for a powerful approach for detection of changes in

biomarkers in real time and provides an opportunity to use metabolomics as a
biomonitoring tool in health and disease. Historically we have adhered to the
concept of a single biomarker for each disease, but this limits the accuracy, precision,
and sensitivity of the assay. New and developing metabolomics approaches suggest
that we may use a pattern of metabolites to describe a given disease. However, by
using multiple biomarkers for each disease, techniques become more sophisticated,
and the computing power used to analyze the data becomes much more complicated.
In this article, we review the expanding field of metabolomics in its application to
respiratory diseases, including asthma, chronic obstructive pulmonary disease
(COPD), pneumonia, acute lung injury/acute respiratory distress syndrome, and
occupational and environmental lung diseases. We also discuss the metabolomics
associated with the lung microbiome in asthma and COPD. These findings show that
there remains a considerable amount of experimental work to be done to understand
the role of the metabolome in respiratory diseases, and how this may be applied to
the diagnosis and/or prognosis of illness. Recent findings have shone some light onto
the relationship between the gut and lung microbiome metabolites in generating
metabolic signatures that may provide mechanistic insights into various lung
diseases, as well as deliver potential biomarkers associated with specific lung
conditions.

2 Respiratory Diseases with Metabolomic Signatures

2.1 Asthma

Asthma is an inflammatory disease of the airways that is often triggered by exoge-

nous perturbations. The recent Global Burden of Disease report stated that an
estimated 262 million people were affected by asthma in 2019, and 461,000 people
died from this disease in that year (Vos et al. 2020). Asthma is a highly heteroge-
neous disease with different phenotypic variations, as well as multiple causative
agents, etiology, and complex inflammatory and pathophysiological features. Thus,
it is proposed that significant metabolic changes are associated with different
phenotypes of disease.
Metabolic profiling has demonstrated significant variations in serological
and urinary metabolomic pathways that are distinct in various phenotypes of
asthma and provides valuable information about the accuracy and precision of
asthma diagnosis, disease progression, and response to treatment (Fig. 2) (Kelly,
“Pharmacometabolomics of Asthma as a road map to Precision Medicine”). Using
technologies such as nuclear magnetic resonance (NMR) spectroscopy and mass
spectrometry (MS), several metabolomic studies have demonstrated comprehensive
evidence of metabolic alterations in asthma. In Table 1, we have summarized some
of the recent reports of metabolomic studies in asthma.
Studies conducted on asthma patients with varying degrees of disease severity,
ages, or obesity have been reported. Interestingly, all these studies showed a high
344 S. Moitra et al.

Fig. 2 Metabolites and metabolic pathways in childhood asthma. Heatmap of Spearman’s rank
correlation coefficients between metabolites associated with lowly and highly sensitized asthma (a)
and metabolic pathways of metabolites associated with atopic and non-atopic asthma (b). Red color
Metabolomics of Respiratory Diseases 345

correlation among specific metabolites with disease. Park et al. (2017) showed that
in severe asthma in children (≤15 years of age) exhibiting corticosteroid resistance,
tyrosine metabolism, degradation of aromatic compounds, and glutathione metabo-
lism were suggested to be significant pathways related to corticosteroid resistance
based on urine metabolites. A targeted LC-MS-based study for the presence of a
unique biomarker in asthmatic children showed a combination of 2-isopropylmalic
acid and betaine can classify children with asthma and controls. It was also shown in
the same study that asthmatics had lower relative concentrations of serum ascorbic
acid, 2-isopropylmalic acid, shikimate-3-phosphate, 6-phospho-D-gluconate, and
reduced glutathione. In the case of overweight children, niacin concentrations
were elevated in serum samples (Checkley et al. 2016). Loureiro et al. showed that
lipid peroxidation-related metabolites in urine samples are associated with asthma
severity and lung function, along with eosinophilic inflammation in nonobese
asthmatic patients (Loureiro et al. 2016). In other studies, it was shown that meta-
bolic pathways and pathway components like arginine, proline, taurine, hypotaurine,
glyoxylate, and dicarboxylate in serum and urine samples were closely related to
acute exacerbations of asthma as well as the choice of corticosteroid treatment
(Quan-Jun et al. 2017). It was also found that a set of 15 volatile carbon compounds
may discriminate between controlled and uncontrolled asthma and that 7 of these
compounds detected in exhaled breath samples could predict exacerbation within the
next 14 days with 88% sensitivity and 75% specificity (van Vliet et al. 2017).
Lipids have also been correlated with the diagnosis of asthma. Kang et al. showed
that certain metabolites, primarily lipid biomolecules in bronchoalveolar lavage
(BAL) fluid, could be markedly elevated in asthma compared to non-asthmatic
healthy individuals (Kang et al. 2014). This observation, supported by other studies
(Loureiro et al. 2016; Ghosh et al. 2020), indicates that lipid metabolism is altered in
asthma, potentially as a result of increased oxidative stress. Such altered lipid
metabolism was also associated with asthma severity, reduced lung function, and
higher eosinophilic inflammation in asthmatic individuals (Loureiro et al. 2016).
While asthma and obesity are known to share common systemic manifestations,
Maniscalco et al. showed that methane, pyruvate, and glyoxylate and dicarboxylate
metabolic pathways in EBC also greatly vary between obese and nonobese asthma
patients (Maniscalco et al. 2017), which indicates more complex crosstalk between
asthma and obesity than previously recognized.
Other recent studies have reported intriguing results of altered profiles of struc-
tural lipid molecules in asthma compared to healthy individuals (Kang et al. 2014;
Ghosh et al. 2020; Reinke et al. 2017; Pang et al. 2018; Jiang et al. 2021). Bian et al.
reported that some derivatives of serum arachidonic acid that serve as potential
mediators for allergic responses were significantly elevated in asthma (Bian et al.

⁄


Fig. 2 (continued) represents positive correlations; blue color represents negative correlations; red
arrow represents increase; blue arrow represents decrease. + symbol means a P-value < 0.05; +
+ symbol means a P-value < 0.01. [Reproduced from Chiu et al. (2021)]
346 S. Moitra et al.

Table 1 Metabolomics studies – asthma

Study population
Study (adult/children) Sample/method Summary of results
Kang et al. Adults Quadrupole time- • " Lysophosphatidylcholine
(2014) (38 asthmatics and of-flight (QTOF) (LPC), triglyceride (TG),
13 healthy) MS of phosphatidylcholine (PC),
bronchoalveolar phosphatidylglycerol (PG),
lavage fluid phosphatidylserine (PS), and
(BALF) sphingomyelin (SM) in
non-steroidal bronchial asthma
NSBA) compared to healthy
subjects
• No difference was observed
between steroid-treated bronchial
asthma (SBA) and healthy subjects
Loureiro Adults Targeted solid Metabolites related to lipid
et al. (2016) (57 asthmatics) phase peroxidation were associated with
microextraction " asthma severity, # lung function,
(SPME) with and " eosinophilic inflammation in
two-dimensional nonobese patients with asthma
gas
chromatography
and time-of-flight
MS (GCxGC-TOF/
MS) of urine
Ghosh et al. Adults GC-MS of serum " 2-Palmitoylglycerol, cholesterol,
(2020) (34 asthmatics, serine, threonine, ethanolamine,
30 COPD, 35 ACO, glucose, stearic acid, linoleic acid,
33 healthy) D-mannose, succinic acid in
asthma than healthy
# Lactic acid, 2-palmitoylglycerol
in asthma than healthy
Maniscalco Adults (25 obese Untargeted LC-MS Participants with asthma, obesity,
et al. (2017) patients with asthma, of EBC and obesity + asthma showed
30 obese patients distinct variations in respiratory
without asthma, metabolic fingerprint
30 lean patients with
asthma and 72 adults
in the external
validation set)
Reinke Adults Untargeted LC-MS " Ceramide (C16:0, C18:0, C20:0,
et al. (2017) (54 asthmatics, of serum C22:0, C24:0, C24:1),
22 healthy) sphingomyelin (C18:0, C18:1),
hexosylceramide (C18:0, C24:1),
and cysteinyl leukotriene E4
(LTE4) in asthma than healthy
# 14,15-
Dihydroxyeicosatetraenoic acid
(DiHETE), 19,20-
Dihydroxydocosapentaenoic acid
(DiHDPA) in asthmatics than
healthy
(continued)
Metabolomics of Respiratory Diseases 347

Table 1 (continued)
Study population
Study (adult/children) Sample/method Summary of results
Pang et al. Adults Ultra-performance " Monosaccharides, LysoPC(o-18:
(2018) (29 asthmatics, liquid 0, 18:1), Retinyl ester, PC(18:1/2:
15 healthy) chromatography- 0), PC(16:0/18:1), arachidonic
tandem MS acid, PE(18:3/14:0) in asthma
(UPLC-MS) of # Glycerophosphocholine, PS(18:
serum 0/22:5), cholesterol glucuronide,
Phytosphingosine, Sphinganine,
LysoPC(p-18:1), retinols, PC(20:
4/16:1)
Jiang et al. Adults LC-MS/MS of " Phosphatidylethanolamine
(2021) (33 asthmatics, plasma (PE) (18:1p/22:6), PE (20:0/18:1),
28 healthy) PE (38:1), sphingomyelin
(SM) (d18:1/18:1), triglyceride
(TG) (16:0/16:0/18:1) in
asthmatics than healthy
# Phosphatidylinositol (PI) (16:0/
20:4), TG (17:0/18:1/18:1),
phosphatidylglycerol (PG) (44:0),
ceramide (d16:0/27:2), lysoPC
(22:4) in asthma
Chiu et al. Adults NMR of urine " Guanidoacetate
(2020, (30 asthmatics, # 1-Methylnicotinamide, allantoin
2018) 30 healthy)
Chiu et al. Children NMR of plasma " Lysine, isovalerate, histidine,
(2021) (28 asthmatics, tyrosine, glycine, citric acid,
25 healthy) ethanol, acetic acid, pyruvic acid in
asthma
Chang- Adults NMR of EBC " Lactate, formate, butyric acid,
Chien et al. (92 asthmatics, isobutyrate in asthma
(2021) 73 healthy)
Bian et al. Adults Ultra-high " 5(S)-Hydroxyeicosatetraenoic
(2017) (15 asthmatics and performance liquid acid (HETE), 8(S)-HETE, 11(S)-
15 healthy) chromatography HETE, 12(S)-HETE, 15(S)-HETE,
quadrupole time- 15(S)-Hydroxyeicosapentaenoic
of-flight (UHPLC)- acid (HEPE), prostaglandin
Q-TOF- MS of (PG)A2, PGB2, PGF1a, PGF2a,
serum PGJ2, 15-keto-PGF2a in asthma
compared to healthy
# Palmitic acid, Lauric acid in
asthma than healthy
Checkley Children Targeted liquid # Relative concentrations of serum
et al. (2016) (50 asthmatics and chromatography- ascorbic acid, 2-isopropylmalic
49 healthy between MS (LC-MS) of acid, shikimate-3-phosphate,
9 and 19 years) serum 6-phospho-D-gluconate, and
reduced glutathione in asthmatics
than healthy
(continued)
348 S. Moitra et al.

Table 1 (continued)
Study population
Study (adult/children) Sample/method Summary of results
Kelly et al. Children Targeted LC-MS • Metabolites (primarily
(2017) (380 asthmatics) (complementary glycerophospholipid, linoleic acid,
methods) of plasma and pyrimidine) were associated
with airway hyperreactivity, and
pre- and postbronchodilator FEV1/
FVC
• Distinct metabolites showed
moderate but important signatures
between disease severity
Tao et al. Children GC-MS of urine " Aspartic acid, Xanthosine,
(2019) (80 asthmatics, hypoxanthine,
29 healthy) N-acetylgalactosamine
# Stearic acid, Heptadecanoic acid,
uric acid, D-threitol
Li et al. Children GC-MS of urine " Azelaic acid, citraconic acid
(2020) (30 asthmatics, 4, D-altrose 1, D-erythro-
30 healthy) sphingosine 1, gentiobiose
2, 2-hydroxybutanoic acid,
L-allothreonine 1, leucine, stearic
acid, succinic acid, tyramine in
asthmatics than healthy
# 3,4-dihydroxycinnamic acid,
methionine 1, purine riboside,
malonic acid 1, cysteine, erythrose
1, lactamide 1, uric acid, valine in
asthma
Matysiak Children LC-MS/MS of " L-arginine, Β-alanine, Ƴ-amino-
et al. (2020) (13 asthmatics, blood N-butyric acid, L-histidine,
17 healthy) Hydroxy-L-proline in asthma
# D,L-Β-Aminoisobutyric acid,
taurine, L-tryptophan, L-valine in
asthma
Ferraro Children UPLC-MS of EBC " 9-amino-nonanoic acid,
et al. (2020) (26 asthmatics, 12-amino-dodecanoic acid, lactone
16 healthy) of PGF-MUM, N-linoleoyl taurine,
17-phenoxy trinor PGF2α ethyl
amide, lysoPC (18:2(9Z,12Z)) in
asthma
Van Vliet Children Targeted GC-TOF/ • 7 VOCs (3 aldehydes,
et al. (2017, (96 asthmatics) MS for VOCs in 1 hydrocarbon, 1 ketone,
2016) EBC 1 aromatic compound, and
1 unidentified VOC) in exhaled
breath could predict asthma
exacerbations
Metabolomics of Respiratory Diseases 349

2017). They described that some saturated fatty acids such as palmitic acid and lauric
acid levels were decreased in asthma. In addition, metabolites derived from protein
or carbohydrate metabolisms were found altered in asthma compared to
non-asthmatic healthy individuals in EBC, plasma, and urine (Chang-Chien et al.
2021; Chiu et al. 2018, 2020). However, in adults, the asthmatic response can be
caused, triggered, or aggravated by different risk factors such as allergy, environ-
mental exposures, active or passive smoking, and workplace conditions. Therefore,
more studies on different adult asthma phenotypes are required to better understand
those metabolic alterations.
On the other hand, asthma in children is mostly caused by allergic conditions or
genetic predisposition (such as parental atopy or asthma) and, to some extent,
pregnancy-related issues such as gestational smoking. In a group of asthmatic and
non-asthmatic children, Checkley et al. showed lowered relative concentrations of
serum ascorbic acid, reduced glutathione (GSH), and some carbohydrate derivatives
in asthma (Checkley et al. 2016). Kelly et al. further showed association between
certain plasma metabolites (glycerophospholipid, linoleic acid, and pyrimidine) and
airway hyperreactivity in asthmatic children (Kelly et al. 2017). They were able to
demonstrate moderate but clinically important signatures of distinct metabolites in
accordance with the disease severity (Kelly et al. 2017). Several other reports have
demonstrated distinct metabolomic profiles in asthmatic children compared to
non-asthmatic healthy individuals (Tao et al. 2019; Li et al. 2020; Matysiak et al.
2020; Chiu et al. 2021).
Recently, breath analysis has suggested some intriguing metabolic alterations in
asthma, particularly related to volatile organic compounds (VOCs) in the EBC that
could predict asthma exacerbations in children (van Vliet et al. 2017; Ferraro et al.
2020; Van Vliet et al. 2016). However, most of those analyses did not consider
potential risk factors or confounding factors as mentioned earlier. Therefore, clinical
correlations between metabolites and symptoms/severity are important to consider
while inferring those results into clinical practice.

2.2 Chronic Obstructive Pulmonary Disease (COPD)

COPD is a major lung disease worldwide that causes significant morbidity and
mortality and is among the top causes of death in many countries (Keogh and
Mark 2021). COPD is a chronic inflammatory disease of the lungs that is progressive
and irreversible in nature (Devine 2008). Although cigarette smoking is the most
common major risk factor for COPD, occupational or environmental insults are also
known to be prominent triggers for the onset and progression of this debilitating lung
condition.
Recent studies have demonstrated that COPD is a variable condition with multi-
modal phenotypic variants, particularly because of the differences in causal agents,
course, and progression of the disease. Although many metabolic alterations of
COPD were unknown until the beginning of the twenty-first century, these have
350 S. Moitra et al.

Fig. 3 Distinct metabolites identified in COPD-associated metabolomics studies. Metabolic

pathways analysis based on distinct metabolites published in chronic obstructive pulmonary disease
(COPD)-associated metabolomics studies performed by applying the Metabo-Analyst 4.0 platform.
The names of 44 disturbed metabolic pathways were marked in the pathway figure, which mainly
involved dysfunctions of amino acid metabolism, lipid metabolism, energy production pathways,
and imbalance of oxidation and antioxidation. [Reproduced from Ran et al. (2019)]

so far exhibited an intriguing panorama based on what has been discovered to date
(Fig. 3). In this section, we briefly describe some of the important findings from
metabolomic research in COPD.
Several studies have identified metabolites that are distinctive in COPD (Turano,
“NMR-based metabolomics to evaluate individual response to treatments”).
Novotna et al. (2018) examined 10 COPD patients and 10 healthy individuals and
observed that two amino acids, alanine and phenylalanine, were significantly lower
in the peripheral blood of COPD patients than healthy individuals, while
pyroglutamate level was higher in COPD patients. They also observed that the
free carnitine to acylcarnitine ratio was significantly lower in COPD patients than
the healthy individuals. Another report by Diao et al. (2019) further demonstrated
that COPD patients had reduced serum levels of creatine, glycine, histidine, and
threonine compared to non-COPD smokers. Although these findings indicate a
possible subclinical malnutrition in the context of respiratory disease, results are
still inconclusive regarding the association of these specific metabolites with COPD.
Body composition is greatly affected in COPD as the disease progresses (Schols
et al. 2005). Chronic bronchitis and emphysema are the two distinct phenotype
Metabolomics of Respiratory Diseases 351

variations of COPD and patients with these diseases have different body silhouettes,
presumably due to difference in lipid metabolic pathways. Some reports have
suggested that perturbation of lipid metabolism occurs in COPD (Chen et al. 2019;
Rafie et al. 2018). In the Subpopulations and Intermediate Outcomes in COPD Study
(SPIROMICS) cohort, Halper-Stromberg et al. (2019) observed that phosphatidyl-
ethanolamine, phosphatidylcholine, phosphatidylinositol, leucine, and lysine from
BAL fluid in COPD patients were associated with higher odds of developing
emphysema. Liang et al. (2019) identified that serum metabolites such as glutamine,
glycine, histidine, hypoxanthine, α-N-phenylacetyl-L-glutamine, L-pipecolic acid,
P-chlorophenylalanine, pseudouridine, and L-citrulline levels were markedly differ-
ent between asthma and COPD.
There is an increasing body of evidence suggesting that sphingolipids, which that
play crucial roles in the structure and function of plasma membranes and signal
transduction, also have roles in the pathogenesis of COPD, asthma, and other lung
conditions (Vlahos 2020). Lipidomic studies have shown that COPD patients have
higher plasma concentrations of very low density lipoprotein (VLDL) compared to
healthy individuals, which strongly correlates with higher central and peripheral
airway resistance (Rafie et al. 2018). Nambiar et al. (2021) found that blood
palmitoleic acid, linoleic acid, and dihydrotestosterone were lower in COPD patients
than healthy controls. Similarly, another study showed that the levels of serum
lysophosphatidylcholine (LPC) 18:3, lysophosphatidylethanolamine (LPE) 16:1,
and phosphatidylinositol (PI) 32:1 were markedly reduced in acute exacerbations
in COPD, thus highlighting the role of glycerophospholipids in the pathophysiology
of COPD (Gai et al. 2021).
Another recent report reiterated these findings in the context of disease onset
and stages in COPD where the authors observed that phosphatidylcholine and LPC
were key indicators of COPD onset and that phosphatidylserine and diacylglycerol
could potentially indicate the various COPD stages (Zhou et al. 2020). In line
with these observations, polyunsaturated acid metabolites were found to be
associated with reduced lung function and disease severity in COPD (Ran et al.
2019; Yu et al. 2019; Xue et al. 2020). Pinto-Plata et al. identified plasma lipid
metabolites that may predict survival differences in COPD patients (Pinto-Plata et al.
2019). Using the Karolinska COSMIC cohort, Naz et al. (2017) found that the
autotaxin-lysophosphatidic acid axis may be dysregulated due to oxidative stress
in COPD and that sex-regulated phenotypes are influential in the pathophysiology of
disease. However, despite several reports demonstrating associations between
metabolites with disease progression and severity in COPD, it is still not clear
whether these metabolites may influence pathophysiological mechanisms. Further-
more, there may be several residual confounders that influence the dysregulation of
metabolic pathways in disease conditions. Therefore, any interpretation and
conclusions made from these metabolic outcomes should be made cautiously (Kilk
et al. 2018).
352 S. Moitra et al.

2.3 Pneumonia

Community-acquired pneumonia remains a major cause of morbidity and mortality

around the world, with over a million hospitalizations each year in the USA prior to
COVID-19 (Griffin et al. 2013). Among the most common bacterial strains involved
in community-acquired pneumonia are Streptococcus pneumoniae and Staphylococ-
cus aureus, which are also found as commensal bacteria in healthy humans. The
challenge in controlling the incidence of pneumonia is to determine the etiological
process by which it occurs in individual patients. Using systems biology approaches,
it is hoped that diagnosis and monitoring of disease may be enhanced to allow for
more accurate prescription of drugs in pneumonia and similar inflammatory lung
diseases (Wheelock et al. 2013).
Application of NMR analysis of pneumonia patient urine suggests that definitive
metabolic profiles could be applied to infection with S. pneumoniae (Fig. 4). The
pattern of urinary metabolites detected in pneumococcal pneumonia could be distin-
guished from pneumonia associated with viruses and other bacterial strains (Slupsky
et al. 2009a). An animal model of pneumonia also demonstrated that distinct
metabolic profiles could be detected in the urine of mice infected with
S. pneumoniae or methicillin-resistant S. aureus, a major cause of antibiotic-resistant
pneumonia (Fig. 5) (Slupsky et al. 2009b). These studies indicate that metabolomics
has potential for the diagnosis, monitoring, and clinical management of pneumococ-
cal diseases.

2.4 Acute Lung Injury/Acute Respiratory Distress Syndrome

(ARDS)

Acute lung injury and its more severe form, ARDS, is characterized by infiltration of
an inflammatory, fibrin-rich exudate into the pulmonary interstitium and alveolar
spaces (Gattinoni et al. 2014; Martin and Matute-Bello 2011; Ware and Matthay

Fig. 4 Differentiating between different types of pneumonia in human patients. Urinary

metabolites were found to be distinct in pneumonia caused by S. pneumoniae and other pathogens.
These graphs show OPLS-DA models based on 61 measured metabolites found in the urine from
S. pneumoniae patients compared with those found in viral pneumonia and other bacteria (including
Mycoplasma tuberculosis, Legionella pneumophila, S. aureus, and others). Reprinted with permis-
sion from Slupsky et al. (2009b) J. Proteome Res. 8:5550–5558. Copyright 2009 American
Chemical Society
Metabolomics of Respiratory Diseases 353

Fig. 5 Metabolic profiles in mice infected with S. pneumoniae. An inbred strain of mice (C57BL/
6), maintained in specific virus antigen-free housing with autoclaved bedding and identical dietary
supplies, was infected intratracheally with a clinical isolate of S. pneumoniae, serotype 14. After
24 h of infection, bronchoalveolar lavage (BAL) samples were analyzed for cell counts (a) and
histology was carried out on lung sections (b) to confirm inflammation arising from infection. At the
same time, urine samples were collected from animals that were subjected to NMR analysis and a
PCA model of urinary metabolite concentrations was generated (c). Reprinted with permission from
Slupsky et al. (2009a) J. Proteome Res. 8(6):3029–3036. Copyright 2009 American Chemical
Society

2000; Li et al. 2011). This influx leads to impaired lung function and diminished gas
exchange (Ware and Matthay 2000). First described in 1967 by Ashbaugh et al.
(1967), ARDS is precipitated by many different causes, with the most common
being sepsis, pneumonia, severe trauma, and more recently, severe COVID-19
(Huang et al. 2020). ARDS is accompanied by an extraordinarily high mortality
rate (approximately 30% of patients die upon diagnosis of ARDS), and to date there
have been few effective pharmacotherapies for its treatment that mainly serve to
shorten the duration of illness rather than reverse it entirely. In addition, no effective
predictive or prognostic biomarkers are available to indicate the likelihood of a
patient developing ARDS. This has prompted a search for biomarkers of ARDS,
which has been led by genomics and proteomics, although neither field has yielded
suitable markers, and no candidate has progressed beyond the initial discovery phase
(Serkova et al. 2011; Rogers and Matthay 2014; Meyer 2013, 2014). This is likely
354 S. Moitra et al.

due to the heterogeneity of disease, and much of the variation could lie beyond the
proteome or genome, possibly in the metabolome (Serkova et al. 2011; Rogers and
Matthay 2014). Thus, metabolomics presents itself as a potentially valuable tool for
analysis in ARDS.
A challenge with understanding mechanisms associated with ARDS is that there
are no translational animal models that accurately mimic human disease (Martin and
Matute-Bello 2011; Matute-Bello et al. 2011; Matute-Bello and Downey 2013).
Despite this limitation, there have been several metabolomic studies carried out in
rodent models that demonstrate changes in metabolites (Stringer et al. 2016). In early
experimental models, mechanical ventilation-induced ARDS in rodents generated
metabolic profiles in serum, lung tissue, and BAL samples (Izquierdo-Garcia et al.
2014). Putative metabolites of ARDS were reported to be increased lactate and
decreased glucose and glycine in lung tissues, together with increased glucose,
lactate, acetate, 3-hydroxybutyrate, and creatine in BAL samples. NMR-detected
metabolites in lung samples were associated with markers of ARDS phenotype (peak
inspiratory pressure, PaO2, and lung histology), but there was no association
between these ARDS indices and serum metabolites. In one of the first studies
examining the metabolomics of experimental ARDS, a cytokine-induced lung injury
model was tested to determine the temporal association between inflammation in the
lungs and changes in lung metabolome (Serkova et al. 2008). Cytokine-induced lung
injury resulted in decreased ATP, energy balance, and energy charge levels,
suggesting a decreased energy state. Together with this there was a significant
increase in glycolytic activity, measured as elevated lactate-to-glucose levels that
normalized 24 h after the induction of injury. Collectively these findings indicate that
a shift in cell energy metabolism occurs in lung tissues in ARDS. The benefit of this
study was that it demonstrated an association between phenotypic and metabolic
changes, an important first step in biomarker discovery. To date, biomarkers have
not been found that can differentiate between the two extremes of mild interstitial
edema and extensive cellular injury in the spectrum of acute lung injury. However,
continued analysis by magnetic resonance imaging and metabolic NMR spectros-
copy may enhance the development of more robust and predictive longitudinal
processes of experimental lung injury. Other animal models have shown significant
metabolic shifts in ARDS induced by a variety of stimuli, reviewed in detail in
Stringer et al. (2016).
Few clinical studies have reported metabolomics analysis of patients with ARDS.
Several studies suggest that the use of BAL samples could provide insight into the
metabolomic profile associated with ARDS. In one study, at least 26 and 18 endoge-
nous metabolites, respectively, could be used to differentiate ARDS from healthy
BAL samples using liquid chromatography-MS analysis (Evans et al. 2014). These
included lactate and other energy metabolism-associated metabolites such as citrate,
creatine, and creatinine which are increased in the plasma of patients with ARDS
(Stringer et al. 2011). These findings demonstrate the utility of BAL as a biofluid for
metabolomics analysis.
Metabolomics of Respiratory Diseases 355

In addition, some reports have demonstrated the utility of exhaled breath as a

vehicle for metabolomics analysis (Schubert et al. 1998; Bos 2018). For example,
Schubert et al. demonstrated the utility of exhaled breath as a sample for
metabolomics analysis (Schubert et al. 1998). This was furthered in a study by
Bos et al. (2014) which found that three metabolites, octane, acetaldehyde, and
3-methylheptane, were able to discriminate ARDS from non-ARDS patients. Octane
is an end-product of lipid peroxidation, one of the degenerative processes caused by
oxidative stress (Riely et al. 1974; Horvat et al. 1964).
Interestingly, a recent study examining EBC from patients on mechanical venti-
lation due to severe COVID-19 or non-COVID-19 ARDS showed a characteristic
“breathprint” for COVID-19 (Grassin-Delyle et al. 2021) that could be distinguished
from non-COVID-19 ARDS. In this study, the four most prominent volatile
compounds in COVID-19 patients were methylpent-2-enal, 2,4-octadiene,
1-chlorohelptane, and nonanal, suggesting that real-time metabolomics analysis of
exhaled breath may identify patients with COVID-19. Nonanal is a sub-product of
oxidative stress-mediated destruction of the cell membrane (Rahman 2003).
In summary, the metabolomics data generated from experimental and clinical
studies of ARDS demonstrate that a disturbance in oxidative stress metabolism and
energy levels occur in this disease, which is consistent with what has been described
for the pathology of ARDS. To date, there appears to be no multi-center prospective
studies done for metabolomics analysis of ARDS. Our understanding of ARDS
metabolomics has been based on small studies that demonstrate feasibility for
evaluation of ARDS phenotypes and for determining lung injury severity. Going
forward, we will need to establish clinical trials aimed at testing prevention and
treatment strategies in ARDS patients by applying metabolomics analysis to the
spectrum of disease that presents in this population.

2.5 Occupational and Environmental Lung Diseases

Occupational exposure is one of the major risk factors associated with respiratory
illnesses, and the incidence of occupational lung diseases is increasing due to
expanding populations and consumer needs (Moitra et al. 2015). According to the
report of the International Labour Organization, nearly two million people die each
year due to workplace accidents, of which over 30% die due to lung cancer or other
lung diseases as a result of workplace exposure (Cullinan et al. 2017). In many cases,
occupational lung diseases are improperly recorded or detected, often due to a lack
of causal evidence, all of which contributes to a significant underestimation of the
true burden of these diseases. Although several biomarker-related reports have been
published in the context of occupational exposure, metabolomic studies have been
very limited to date. We discuss some of the few studies below on occupational and
environmental lung diseases.
Among a group of workers employed in carbon-coating friction systems,
Maniscalco et al. (2018) found that the concentrations of VOCs, including
356 S. Moitra et al.

1,2-propanediol, phenylalanine, 3-hydroxybutyrate, and isopropanol, were signifi-

cantly elevated in the EBC of the workers who did not wear a mask at the work,
compared to those who routinely wore masks.
Other markers such as polycyclic aromatic hydrocarbons (PAH) have been found
to be associated with occupational exposure. Wei et al. studied the joint effects of
arsenic exposure, smoking, and physical exercise on lung function changes among a
group of coke-oven workers and found that urinary concentrations of PAH were
significantly higher in coke-oven workers than office workers in the same industry
who were not directly exposed to the ovens (Wei et al. 2021). Using a nationwide
biomonitoring survey of the Korean National Environmental Health Survey, Koh
et al. collected measurements of urinary 1-hydroxypyrene (1-OHP) as a metabolite
of interest for PAH exposure at workplace. They found that the level of urinary
1-OHP was highest among people engaged in construction and mine-related
occupations. Although that study did not explicitly study associations between
urinary metabolites and respiratory health, the effect of PAH on respiratory health
is already well known and therefore, urinary 1-OHP could potentially be a marker of
PAH-associated respiratory dysfunctions.
The collapse of the World Trade Center (WTC) on 9/11 introduced a novel and
unprecedented exposure scenario in which hundreds of thousands of New Yorkers
were affected in the ensuing years. Firefighters and all first responders were exposed
to huge amounts of dust containing numerous fibrous, chemical, and hazardous
substances. To date, several reports have been published on the respiratory health of
the workers who were exposed to WTC dust, resulting in a condition known as WTC
lung disease (also known as WTC sarcoid-like granulomatous pulmonary disease).
For example, a recent study provided novel insights into metabolic syndrome as a
risk factor for lung function decline in a cohort of firefighters exposed to materials
arising from the collapse of the WTC (Kwon et al. 2021). They also proposed that
regulating metabolic syndrome, particularly dyslipidemia, could also help to
decrease the risk of developing WTC lung disease. This group also showed previ-
ously that the serum metabolome, particularly the sphingolipid cluster containing
sphingosine-1-phosphate, a pleiotropic inflammatory mediator, was low in WTC
lung diseases, suggesting decreased bioavailability and increased risk of
compromised vascular integrity in WTC lung disease (Crowley et al. 2018). A
mouse model of WTC particulate matter exposure was also investigated and showed
that several prominent metabolic pathways were affected, including advanced
glycation end-products and lipids (including sphingolipids), that correlated with
inflammatory changes and attenuation of antioxidant potential (Veerappan et al.
2020). However, despite these interesting outcomes, correlations between
metabolomics and clinical evaluation in occupational lung diseases remain limited,
and therefore more studies are required to elucidate the crosstalk between these two
aspects.
Metabolomics of Respiratory Diseases 357

3 Metabolomics of Lung Microbiome in Respiratory Diseases

3.1 Asthma

The lung microbiota and metabolome are likely to play a pivotal role in the onset of
disease in the case of asthma (Barcik et al. 2020). It is now emerging that metaboli-
cally active microbiota that reside in the lung under normal conditions maintain a
complex network of crosstalk with the host in a symbiotic manner. In disease
conditions, however, this symbiosis is transformed into dysbiosis that can alter the
host immune response, which influences the overall lung health (Loverdos et al.
2019). The composition of normal lung microbiota consists of Bacteroidetes and
Firmicutes (the most abundant two genera), and apart from these two,
Proteobacteria, Actinobacteria, and Fusobacteria have also been found by 16S
rRNA sequencing in endobronchial brushing samples (Charlson et al. 2011; Bassis
et al. 2015). Although normal lung microbiota consists of a relatively small bacterial
population, estimated to be around 103 to 105/cm2, their intensely intricate crosstalk
is thought to be primarily responsible for the conduct of most of the host-microbiome
interplay (Charlson et al. 2011; Bassis et al. 2015; Hilty et al. 2010; Mathieu et al.
2018; Denner et al. 2016; Goleva et al. 2013).
In asthma, the bacterial pattern of the pulmonary microbiome has been
characterized in several studies. It is evident that some bacterial species become
elevated in nasopharyngeal swabs from asthmatics, such as Haemophilus influenzae,
Streptococcus pneumoniae, Staphylococcus aureus, and Moraxella catarrhalis,
compared to healthy controls. These bacteria are well-known pathogens that can
cause infectious exacerbations (Dickson et al. 2016). Interestingly, Huang et al.
showed in patients with severe asthma, Actinobacteria is present at high abundance
in correlation with elevated sputum leukocytes and eosinophils in bronchial biopsies
(Huang et al. 2015). It has been also shown that elevated eosinophil numbers in
lavage, along with reduced FEV1, correlate with bacterial α-diversity (based on
comparison of different species present in same sample) in endobronchial brushings
of asthmatic subjects. Bacterial species associated with lower airway obstruction
show distinctive features associated with FEV1 levels. For example, patients with
asthma exhibiting FEV1 < 60% had low α-diversity but high β-diversity compared
to asthma patients with FEV1 > 80% (Denner et al. 2016).
Interestingly, the gut microbiome is an important component of asthma patho-
physiology which has not been explored in detail. The human gut possesses a surface
area of 150–200 m2, which harbors 100,000 to 100 billion bacteria per mL of
sample, depending on the region of sample collection (Sender et al. 2016). A
relationship between the gut and lung was discovered upon the observation that
different lung diseases may be influenced by changes in the gut microflora and vice
versa. The microbiota in these two sites is therefore connected by a gut-lung axis that
is important in relation to asthma (Marsland et al. 2015). Among many different
metabolites produced by the gut microbiome, short-chain fatty acids (SCFAs), such
as acetate, propionate, and butyrate, have been found to regulate physiological and
immunological responses in humans. It is well known that not only do SCFAs
358 S. Moitra et al.

provide a source of energy, but they also function as signaling molecules. SCFAs
have been shown to have multiple signaling effects: they inhibit histone deacetylases
(HDACs) that increase cytokine gene expression by promoting an anti-inflammatory
cell phenotype to maintain homeostasis, suppress transcription factors (nuclear
factor κ-light-chain-enhancer of activated B cells (NF-κB)), and reduce tumor
necrosis factor-α (TNF-α) production (Durack et al. 2017; Chambers et al. 2018;
Tan et al. 2014). Depletion of SCFA-producing bacteria as a mechanistic link
between the microbiome and asthma susceptibility or severity has been suggested
by Cait et al. (2018). Hence, SCFAs derived from the metabolic activity of gut
microbiota inhibit proinflammatory responses in the lungs. Although the mechanism
underlying this pathway is unclear, the most likely mechanistic explanation is that
the hepatic system may weaken innate immune responses by SCFAs binding to G
protein-coupled receptors and/or inhibition of the mevalonate/isoprenoid pathway
through HMGCoA reductase (Young et al. 2016). The specific mediators that make
up the communication between gut and lung is still unclear, but it has been
speculated that gut epithelial cells and immune cells absorb signals from the
endothelium to form local cytokine microenvironments, and eventually this alters
the immune response in distal sites such as the lung (Budden et al. 2017).
Overall, these studies demonstrate that the gut and lung microbiome, and its
associated metabolome, have an enormous impact on patient outcomes in asthma.
Findings from these reports could contribute to the discovery of mechanisms and
novel biomarkers for asthma and its associated exacerbations.

3.2 Chronic Obstructive Pulmonary Disease (COPD)

Recent evidence suggests an association between the lung microbiome and COPD,
suggesting a contribution of the lung bacterial community to disease progression in
the form of dysbiosis (Hilty et al. 2010; Erb-Downward et al. 2011; Zakharkina et al.
2013; Pragman et al. 2012). Phylogenetic analysis of microbial populations in
samples collected from the oropharynx and bronchial brushings from COPD patients
and healthy controls showed increased populations of pathogenic Proteobacteria
(Haemophilus spp.) over Bacteroidetes (Prevotella spp.), with the latter being
especially reduced (Hilty et al. 2010). Other studies also demonstrated that healthy
individuals commonly exhibit higher populations of Firmicutes, Bacteroidetes,
Proteobacteria, Fusobacteria, and Actinobacteria, in contrast to pathogenic
Haemophilus, Streptococcus, Klebsiella, Pseudomonas, and Moraxella in COPD
patients (Wu et al. 2014; Murphy et al. 2005). In addition, several reports have
described that exposure to tobacco smoke can modify bacterial populations in the
mouth and lungs. Though studies are limited in the context of COPD, numerous
reports indicate an alteration of the oral and respiratory bacterial microbiome as an
effect of tobacco smoking (Morris et al. 2013; Zhang et al. 2018; Huang and Shi
2019). In COPD patients, commensal colonization of H. influenzae, S. pneumoniae,
Metabolomics of Respiratory Diseases 359

Pseudomonas, and Moraxella are frequently observed in the lungs (Simpson et al.
2016).
The gut-lung axis also features prominently in COPD (Young et al. 2016). Fecal
microbiota derived from COPD patients have been demonstrated to contribute to the
development of COPD in a mouse model (Li et al. 2021). The fecal microbiota of
COPD patients were found to contain lower levels of SCFAs, which could contribute
to the manifestation of COPD.
In another recent study comparing the metabolomic profiles of COPD patients
with healthy humans (Bowerman et al. 2020), it was found that COPD patients and
healthy individuals manifest significantly different sets of microbial and metabolic
signatures in fecal samples. As many as 146 bacterial species differ in between these
two groups, along with a group of the top 50 indicator metabolites that distinguished
between healthy and COPD individuals, consisting of mostly lipids (46%), amino
acids (20%), and xenobiotic compounds (20%). Hence, it can be deduced that the
intricate mechanisms associated with the gut-lung axis and the host’s microbial
community play a crucial role in the manifestation and progression of COPD.

4 Conclusive Remarks

Taken together, we have reviewed some of the literature associated with

metabolomics analysis of biological fluids obtained from patients and experimental
animal models with a range of respiratory diseases. Metabolomics is a fundamental
part of systems biology analysis that has enormous clinical potential in discovering
novel biomarkers as well as understanding disease pathophysiology. Because of its
rapidly changing properties in health as well as disease, metabolomics has the power
to generate snapshots of metabolites from a given sample that can be followed over
time with repeated sampling. Several high-throughput systems have the ability to
capture the identities and qualities of metabolites in a rapid manner using NMR or
MS-based techniques. Challenges remain with the application of NMR in complex
biological samples, which is less sensitive to small amounts of metabolites in many
cases than MS. An important distinction to make is that analysis of metabolites in
lung-specific samples is predicted to provide greater sensitivity to the tissue-specific
metabolome over that of blood-derived (plasma, serum, or whole blood) or urinary
metabolites. This is especially evident in the case of analysis of the metabolomics of
the lung microbiome. Variability of NMR-measured metabolites is also an issue,
with differing results found within a single facility as well as multiple locations
(Lacy et al. 2014). In addition, a substantial number of unknown metabolites have
been detected by MS that await more detailed identification in biological samples.
We look forward to a future where we can implement increasingly sophisticated
analyses of biological samples using systems biology approaches in respiratory
diseases.
360 S. Moitra et al.

References
Ashbaugh DG, Bigelow DB, Petty TL, Levine BE (1967) Acute respiratory distress in adults.
Lancet 2(7511):319–323
Barcik W, Boutin RCT, Sokolowska M, Finlay BB (2020) The role of lung and gut microbiota in
the pathology of asthma. Immunity 52(2):241–255
Bassis CM, Erb-Downward JR, Dickson RP, Freeman CM, Schmidt TM, Young VB, Beck JM,
Curtis JL, Huffnagle GB (2015) Analysis of the upper respiratory tract microbiotas as the source
of the lung and gastric microbiotas in healthy individuals. MBio 6(2):e00037
Bian X, Sun B, Zheng P, Li N, Wu JL (2017) Derivatization enhanced separation and sensitivity of
long chain-free fatty acids: application to asthma using targeted and non-targeted liquid
chromatography-mass spectrometry approach. Anal Chim Acta 989:59–70
Bos LDJ (2018) Diagnosis of acute respiratory distress syndrome by exhaled breath analysis. Ann
Transl Med 6(2):33
Bos LD, Weda H, Wang Y, Knobel HH, Nijsen TM, Vink TJ, Zwinderman AH, Sterk PJ, Schultz
MJ (2014) Exhaled breath metabolomics as a noninvasive diagnostic tool for acute respiratory
distress syndrome. Eur Respir J 44(1):188–197
Bowerman KL, Rehman SF, Vaughan A, Lachner N, Budden KF, Kim RY, Wood DLA, Gellatly
SL, Shukla SD, Wood LG, Yang IA et al (2020) Disease-associated gut microbiome and
metabolome changes in patients with chronic obstructive pulmonary disease. Nat Commun
11(1):5886
Budden KF, Gellatly SL, Wood DL, Cooper MA, Morrison M, Hugenholtz P, Hansbro PM (2017)
Emerging pathogenic links between microbiota and the gut-lung axis. Nat Rev Microbiol 15(1):
55–63
Cait A, Hughes MR, Antignano F, Cait J, Dimitriu PA, Maas KR, Reynolds LA, Hacker L, Mohr J,
Finlay BB, Zaph C et al (2018) Microbiome-driven allergic lung inflammation is ameliorated by
short-chain fatty acids. Mucosal Immunol 11(3):785–795
Chambers ES, Preston T, Frost G, Morrison DJ (2018) Role of gut microbiota-generated short-chain
fatty acids in metabolic and cardiovascular health. Curr Nutr Rep 7(4):198–206
Chang-Chien J, Huang HY, Tsai HJ, Lo CJ, Lin WC, Tseng YL, Wang SL, Ho HY, Cheng ML,
Yao TC (2021) Metabolomic differences of exhaled breath condensate among children with and
without asthma. Pediatr Allergy Immunol 32(2):264–272
Charlson ES, Bittinger K, Haas AR, Fitzgerald AS, Frank I, Yadav A, Bushman FD, Collman RG
(2011) Topographical continuity of bacterial populations in the healthy human respiratory tract.
Am J Respir Crit Care Med 184(8):957–963
Checkley W, Deza MP, Klawitter J, Romero KM, Klawitter J, Pollard SL, Wise RA, Christians U,
Hansel NN (2016) Identifying biomarkers for asthma diagnosis using targeted metabolomics
approaches. Respir Med 121(1532–3064 (Electronic)):59–66
Chen H, Li Z, Dong L, Wu Y, Shen H, Chen Z (2019) Lipid metabolism in chronic obstructive
pulmonary disease. Int J Chron Obstruct Pulmon Dis 14:1009–1018
Chiu CY, Lin G, Cheng ML, Chiang MH, Tsai MH, Su KW, Hua MC, Liao SL, Lai SH, Yao TC,
Yeh KW et al (2018) Longitudinal urinary metabolomic profiling reveals metabolites for asthma
development in early childhood. Pediatr Allergy Immunol 29(5):496–503
Chiu CY, Cheng ML, Chiang MH, Wang CJ, Tsai MH, Lin G (2020) Metabolomic analysis reveals
distinct profiles in the plasma and urine associated with IgE reactions in childhood asthma. J
Clin Med 9(3):887
Chiu CY, Cheng ML, Chiang MH, Wang CJ, Tsai MH, Lin G (2021) Integrated metabolic and
microbial analysis reveals host–microbial interactions in IgE-mediated childhood asthma. Sci
Rep 11(1):23407
Crowley G, Kwon S, Haider SH, Caraher EJ, Lam R, St-Jules DE, Liu M, Prezant DJ, Nolan A
(2018) Metabolomics of world trade center-lung injury: a machine learning approach. BMJ
Open Respir Res 5(1):e000274
Metabolomics of Respiratory Diseases 361

Cullinan P, Munoz X, Suojalehto H, Agius R, Jindal S, Sigsgaard T, Blomberg A, Charpin D,

Annesi-Maesano I, Gulati M, Kim Y et al (2017) Occupational lung diseases: from old and
novel exposures to effective preventive strategies. Lancet Respir Med 5(5):445–455
Denner DR, Sangwan N, Becker JB, Hogarth DK, Oldham J, Castillo J, Sperling AI, Solway J,
Naureckas ET, Gilbert JA, White SR (2016) Corticosteroid therapy and airflow obstruction
influence the bronchial microbiome, which is distinct from that of bronchoalveolar lavage in
asthmatic airways. J Allergy Clin Immunol 137(5):1398–1405
Devine JF (2008) Chronic obstructive pulmonary disease: an overview. Am Health Drug Benefits
1(7):34–42
Diao W, Labaki WW, Han MK, Yeomans L, Sun Y, Smiley Z, Kim JH, McHugh C, Xiang P,
Shen N, Sun X et al (2019) Disruption of histidine and energy homeostasis in chronic
obstructive pulmonary disease. Int J Chron Obstruct Pulmon Dis 14:2015–2025
Dickson RP, Erb-Downward JR, Martinez FJ, Huffnagle GB (2016) The microbiome and the
respiratory tract. Annu Rev Physiol 78(1):481–504
Durack J, Lynch SV, Nariya S, Bhakta NR, Beigelman A, Castro M, Dyer AM, Israel E, Kraft M,
Martin RJ, Mauger DT et al (2017) Features of the bronchial bacterial microbiome associated
with atopy, asthma, and responsiveness to inhaled corticosteroid treatment. J Allergy Clin
Immunol 140(1):63–75
Erb-Downward JR, Thompson DL, Han MK, Freeman CM, McCloskey L, Schmidt LA, Young
VB, Toews GB, Curtis JL, Sundaram B, Martinez FJ et al (2011) Analysis of the lung
microbiome in the "healthy" smoker and in COPD. PLoS One 6(2):e16384
Evans CR, Karnovsky A, Kovach MA, Standiford TJ, Burant CF, Stringer KA (2014) Untargeted
LC-MS metabolomics of bronchoalveolar lavage fluid differentiates acute respiratory distress
syndrome from health. J Proteome Res 13(2):640–649
Ferraro VA, Carraro S, Pirillo P, Gucciardi A, Poloniato G, Stocchero M, Giordano G, Zanconato S,
Baraldi E (2020) Breathomics in asthmatic children treated with inhaled corticosteroids. Meta
10(10):390
Gai X, Guo C, Zhang L, Zhang L, Abulikemu M, Wang J, Zhou Q, Chen Y, Sun Y, Chang C (2021)
Serum glycerophospholipid profile in acute exacerbation of chronic obstructive pulmonary
disease. Front Physiol 12:646010
Gattinoni L, Cressoni M, Brazzi L (2014) Fluids in ARDS: from onset through recovery. Curr Opin
Crit Care 20(4):373–377
Ghosh N, Choudhury P, Kaushik SR, Arya R, Nanda R, Bhattacharyya P, Roychowdhury S,
Banerjee R, Chaudhury K (2020) Metabolomic fingerprinting and systemic inflammatory
profiling of asthma COPD overlap (ACO). Respir Res 21(1):126
Goleva E, Jackson LP, Harris JK, Robertson CE, Sutherland ER, Hall CF, Good JT Jr, Gelfand EW,
Martin RJ, Leung DY (2013) The effects of airway microbiome on corticosteroid
responsiveness in asthma. Am J Respir Crit Care Med 188(10):1193–1201
Grassin-Delyle S, Roquencourt C, Moine P, Saffroy G, Carn S, Heming N, Fleuriet J, Salvator H,
Naline E, Couderc LJ, Devillier P et al (2021) Metabolomics of exhaled breath in critically ill
COVID-19 patients: a pilot study. EBioMedicine 63:103154
Griffin MR, Zhu Y, Moore MR, Whitney CG, Grijalva CG (2013) U.S. Hospitalizations for
pneumonia after a decade of pneumococcal vaccination. N Engl J Med 369(2):155–163
Halper-Stromberg E, Gillenwater L, Cruickshank-Quinn C, O'Neal WK, Reisdorph N, Petrache I,
Zhuang Y, Labaki WW, Curtis JL, Wells J, Rennard S et al (2019) Bronchoalveolar lavage fluid
from COPD patients reveals more compounds associated with disease than matched plasma.
Metabolites 9(8):157
Hilty M, Burke C, Pedro H, Cardenas P, Bush A, Bossley C, Davies J, Ervine A, Poulter L,
Pachter L, Moffatt MF et al (2010) Disordered microbial communities in asthmatic airways.
PLoS One 5(1):e8578
Horvat RJ, Lane WG, Ng H, Shepherd AD (1964) Saturated hydrocarbons from autoxidizing
methyl linoleate. Nature 203:523–524
362 S. Moitra et al.

Huang C, Shi G (2019) Smoking and microbiome in oral, airway, gut and some systemic diseases. J
Transl Med 17(1):225
Huang YJ, Nariya S, Harris JM, Lynch SV, Choy DF, Arron JR, Boushey H (2015) The airway
microbiome in patients with severe asthma: associations with disease features and severity. J
Allergy Clin Immunol 136(4):874–884
Huang C, Wang Y, Li X, Ren L, Zhao J, Hu Y, Zhang L, Fan G, Xu J, Gu X, Cheng Z et al (2020)
Clinical features of patients infected with 2019 novel coronavirus in Wuhan, China. Lancet
395(10223):497–506
Izquierdo-Garcia JL, Naz S, Nin N, Rojas Y, Erazo M, Martinez-Caro L, Garcia A, de Paula M,
Fernandez-Segoviano P, Casals C, Esteban A et al (2014) A Metabolomic approach to the
pathogenesis of ventilator-induced lung injury. Anesthesiology 120(3):694–702
Jiang T, Dai L, Li P, Zhao J, Wang X, An L, Liu M, Wu S, Wang Y, Peng Y, Sun D et al (2021)
Lipid metabolism and identification of biomarkers in asthma by lipidomic analysis. Biochim
Biophys Acta Mol Cell Biol Lipids 1866(2):158853
Kang YP, Lee WJ, Hong JY, Lee SB, Park JH, Kim D, Park S, Park CS, Park SW, Kwon SW
(2014) Novel approach for analysis of bronchoalveolar lavage fluid (BALF) using HPLC-
QTOF-MS-based lipidomics: lipid levels in asthmatics and corticosteroid-treated asthmatic
patients. J Proteome Res 13(9):3919–3929
Kelly RS, Virkud Y, Giorgio R, Celedon JC, Weiss ST, Lasky-Su J (2017) Metabolomic profiling
of lung function in Costa-Rican children with asthma. Biochim Biophys Acta Mol basis Dis
1863(6):1590–1595
Keogh E, Mark WE (2021) Managing malnutrition in COPD: a review. Respir Med 176:106248
Kilk K, Aug A, Ottas A, Soomets U, Altraja S, Altraja A (2018) Phenotyping of chronic obstructive
pulmonary disease based on the integration of metabolomes and clinical characteristics. Int J
Mol Sci 19(3):666
Kwon S, Lee M, Crowley G, Schwartz T, Zeig-Owens R, Prezant DJ, Liu M, Nolan A (2021)
Dynamic metabolic risk profiling of world trade center lung disease: a longitudinal cohort study.
Am J Respir Crit Care Med 204(9):1035–1047
Lacy P, McKay RT, Finkel M, Karnovsky A, Woehler S, Lewis MJ, Chang D, Stringer KA (2014)
Signal intensities derived from different NMR probes and parameters contribute to variations in
quantification of metabolites. PLoS One 9(1):e85732
Li G, Malinchoc M, Cartin-Ceba R, Venkata CV, Kor DJ, Peters SG, Hubmayr RD, Gajic O (2011)
Eight-year trend of acute respiratory distress syndrome: a population-based study in Olmsted
County, Minnesota. Am J Respir Crit Care Med 183(1):59–66
Li S, Liu J, Zhou J, Wang Y, Jin F, Chen X, Yang J, Chen Z (2020) Urinary Metabolomic profiling
reveals biological pathways and predictive signatures associated with childhood asthma. J
Asthma Allergy 13:713–724
Li N, Dai Z, Wang Z, Deng Z, Zhang J, Pu J, Cao W, Pan T, Zhou Y, Yang Z, Li J et al (2021) Gut
microbiota dysbiosis contributes to the development of chronic obstructive pulmonary disease.
Respir Res 22(1):274
Liang Y, Gai XY, Chang C, Zhang X, Wang J, Li TT (2019) Metabolomic profiling differences
among asthma, COPD, and healthy subjects: a LC-MS-based metabolomic analysis. Biomed
Environ Sci 32(9):659–672
Loureiro CC, Oliveira AS, Santos M, Rudnitskaya A, Todo-Bom A, Bousquet J, Rocha SM (2016)
Urinary metabolomic profiling of asthmatics can be related to clinical characteristics. Allergy
71(9):1362–1365
Loverdos K, Bellos G, Kokolatou L, Vasileiadis I, Giamarellos E, Pecchiari M, Koulouris N,
Koutsoukou A, Rovina N (2019) Lung microbiome in asthma: current perspectives. J Clin Med
8(11):1967
Madsen R, Lundstedt T, Trygg J (2010) Chemometrics in metabolomics – a review in human
disease diagnosis. Anal Chim Acta 659(1–2):23–33
Maniscalco M, Paris D, Melck DJ, D'Amato M, Zedda A, Sofia M, Stellato C, Motta A (2017)
Coexistence of obesity and asthma determines a distinct respiratory metabolic phenotype. J
Allergy Clin Immunol 139(5):1536–1547
Metabolomics of Respiratory Diseases 363

Maniscalco M, Paris D, Melck D, Chiariello N, Di Napoli F, Manno M, Iavicoli I, Motta A (2018)

Biomonitoring of workers using nuclear magnetic resonance-based metabolomics of exhaled
breath condensate: a pilot study. Toxicol Lett 298:4–12
Marsland BJ, Trompette A, Gollwitzer ES (2015) The gut-lung axis in respiratory disease. Ann Am
Thorac Soc 12(Suppl 2):S150–S156
Martin TR, Matute-Bello G (2011) Experimental models and emerging hypotheses for acute lung
injury. Crit Care Clin 27(3):735–752
Mathieu E, Escribano-Vazquez U, Descamps D, Cherbuy C, Langella P, Riffault S, Remot A,
Thomas M (2018) Paradigms of lung microbiota functions in health and disease, particularly, in
asthma. Front Physiol 9(1168):1168
Matute-Bello G, Downey GP (2013) Reply: defining lung injury in animals. Am J Respir Cell Mol
Biol 48(2):267–268
Matute-Bello G, Downey G, Moore BB, Groshong SD, Matthay MA, Slutsky AS, Kuebler WM
(2011) Acute Lung Injury in Animals Study G. An official American Thoracic Society work-
shop report: features and measurements of experimental acute lung injury in animals. Am J
Respir Cell Mol Biol 44(5):725–738
Matysiak J, Klupczynska A, Packi K, Mackowiak-Jakubowska A, Breborowicz A, Pawlicka O,
Olejniczak K, Kokot ZJ, Matysiak J (2020) Alterations in serum-free amino acid profiles in
childhood asthma. Int J Environ Res Public Health 17(13):4758
Meyer NJ (2013) Future clinical applications of genomics for acute respiratory distress syndrome.
Lancet Respir Med 1(10):793–803
Meyer NJ (2014) Beyond single-nucleotide polymorphisms: genetics, genomics, and other 'omic
approaches to acute respiratory distress syndrome. Clin Chest Med 35(4):673–684
Moitra S, Puri R, Paul D, Huang YC (2015) Global perspectives of emerging occupational and
environmental lung diseases. Curr Opin Pulm Med 21(2):114–120
Morris A, Beck JM, Schloss PD, Campbell TB, Crothers K, Curtis JL, Flores SC, Fontenot AP,
Ghedin E, Huang L, Jablonski K et al (2013) Comparison of the respiratory microbiome in
healthy nonsmokers and smokers. Am J Respir Crit Care Med 187(10):1067–1075
Murphy TF, Brauer AL, Grant BJ, Sethi S (2005) Moraxella catarrhalis in chronic obstructive
pulmonary disease: burden of disease and immune response. Am J Respir Crit Care Med 172(2):
195–199
Nambiar S, Tan DBA, Clynick B, Bong SH, Rawlinson C, Gummer J, Corte TJ, Glaspole I,
Moodley YP, Trengove R (2021) Untargeted metabolomics of human plasma reveal lipid
markers unique to chronic obstructive pulmonary disease and idiopathic pulmonary fibrosis.
Proteomics Clin Appl 15(2–3):e2000039
Naz S, Kolmert J, Yang M, Reinke SN, Kamleh MA, Snowden S, Heyder T, Levanen B, Erle DJ,
Skold CM, Wheelock AM et al (2017) Metabolomics analysis identifies sex-associated
metabotypes of oxidative stress and the autotaxin-lysoPA axis in COPD. Eur Respir J 49(6):
1602322
Novotna B, Abdel-Hamid M, Koblizek V, Svoboda M, Hejduk K, Rehacek V, Bis J, Salajka F
(2018) A pilot data analysis of a metabolomic HPLC-MS/MS study of patients with COPD. Adv
Clin Exp Med 27(4):531–539
Pang Z, Wang G, Wang C, Zhang W, Liu J, Wang F (2018) Serum metabolomics analysis of asthma
in different inflammatory phenotypes: a cross-sectional study in Northeast China. Biomed Res
Int 2018:2860521
Park YH, Fitzpatrick AM, Medriano CA, Jones DP (2017) High-resolution metabolomics to
identify urine biomarkers in corticosteroid-resistant asthmatic children. J Allergy Clin Immunol
139(5):1518–1524
Patti GJ, Yanes O, Siuzdak G (2012) Innovation: metabolomics: the apogee of the omics trilogy.
Nat Rev Mol Cell Biol 13(4):263–269
Pinto-Plata V, Casanova C, Divo M, Tesfaigzi Y, Calhoun V, Sui J, Polverino F, Priolo C,
Petersen H, de Torres JP, Marin JM et al (2019) Plasma metabolomics and clinical predictors
of survival differences in COPD patients. Respir Res 20(1):219
364 S. Moitra et al.

Pragman AA, Kim HB, Reilly CS, Wendt C, Isaacson RE (2012) The lung microbiome in moderate
and severe chronic obstructive pulmonary disease. PLoS One 7(10):e47305
Quan-Jun Y, Jian-Ping Z, Jian-Hua Z, Yong-Long H, Bo X, Jing-Xian Z, Bona D, Yuan Z, Cheng
G (2017) Distinct metabolic profile of inhaled budesonide and salbutamol in asthmatic children
during acute exacerbation. Basic Clin Pharmacol Toxicol 120(3):303–311
Rafie S, Moitra S, Brashier BB (2018) Association between the serum metabolic profile and lung
function in chronic obstructive pulmonary disease. Turk Thorac J 19(1):13–18
Rahman I (2003) Oxidative stress, chromatin remodeling and gene transcription in inflammation
and chronic lung diseases. J Biochem Mol Biol 36(1):95–109
Ran N, Pang Z, Gu Y, Pan H, Zuo X, Guan X, Yuan Y, Wang Z, Guo Y, Cui Z, Wang F (2019) An
updated overview of Metabolomic profile changes in chronic obstructive pulmonary disease.
Meta 9(6):111
Reinke SN, Gallart-Ayala H, Gomez C, Checa A, Fauland A, Naz S, Kamleh MA, Djukanovic R,
Hinks TS, Wheelock CE (2017) Metabolomics analysis identifies different metabotypes of
asthma severity. Eur Respir J 49(3):1601740
Riely CA, Cohen G, Lieberman M (1974) Ethane evolution: a new index of lipid peroxidation.
Science 183(4121):208–210
Robertson DG, Watkins PB, Reily MD (2011) Metabolomics in toxicology: preclinical and clinical
applications. Toxicol Sci 120:S146–S170
Rogers AJ, Matthay MA (2014) Applying metabolomics to uncover novel biology in ARDS. Am J
Physiol Lung Cell Mol Physiol 306(11):L957–L961
Schols AM, Broekhuizen R, Weling-Scheepers CA, Wouters EF (2005) Body composition and
mortality in chronic obstructive pulmonary disease. Am J Clin Nutr 82(1):53–59
Schubert JK, Muller WP, Benzing A, Geiger K (1998) Application of a new method for analysis of
exhaled gas in critically ill patients. Intensive Care Med 24(5):415–421
Sender R, Fuchs S, Milo R (2016) Are we really vastly outnumbered? Revisiting the ratio of
bacterial to host cells in humans. Cell 164(3):337–340
Serkova NJ, Van Rheen Z, Tobias M, Pitzer JE, Wilkinson JE, Stringer KA (2008) Utility of
magnetic resonance imaging and nuclear magnetic resonance-based metabolomics for quantifi-
cation of inflammatory lung injury. Am J Physiol Lung Cell Mol Physiol 295(1):L152–L161
Serkova NJ, Standiford TJ, Stringer KA (2011) The emerging field of quantitative blood
metabolomics for biomarker discovery in critical illnesses. Am J Respir Crit Care Med
184(6):647–655
Simpson JL, Baines KJ, Horvat JC, Essilfie AT, Brown AC, Tooze M, McDonald VM, Gibson PG,
Hansbro PM (2016) COPD is characterized by increased detection of Haemophilus influenzae,
Streptococcus pneumoniae and a deficiency of Bacillus species. Respirology 21(4):697–704
Slupsky CM, Cheypesh A, Chao DV, Fu H, Rankin KN, Marrie TJ, Lacy P (2009a) Streptococcus
pneumoniae and Staphylococcus aureus pneumonia induce distinct metabolic responses. J
Proteome Res 8(6):3029–3036
Slupsky CM, Rankin KN, Fu H, Chang D, Rowe BH, Charles PG, McGeer A, Low D, Long R,
Kunimoto D, Sawyer MB et al (2009b) Pneumococcal pneumonia: potential for diagnosis
through a urinary metabolic profile. J Proteome Res 8(12):5550–5558
Stringer KA, Serkova NJ, Karnovsky A, Guire K, Paine R 3rd, Standiford TJ (2011) Metabolic
consequences of sepsis-induced acute lung injury revealed by plasma (1)H-nuclear magnetic
resonance quantitative metabolomics and computational analysis. Am J Physiol Lung Cell Mol
Physiol 300(1):L4–L11
Stringer KA, McKay RT, Karnovsky A, Quemerais B, Lacy P (2016) Metabolomics and its
application to acute lung diseases. Front Immunol 7:44
Tan J, McKenzie C, Potamitis M, Thorburn AN, Mackay CR, Macia L (2014) Chapter three – the
role of short-chain fatty acids in health and disease. In: Alt FW (ed) Advances in immunology,
vol 121. Academic Press, pp 91–119
Tao JL, Chen YZ, Dai QG, Tian M, Wang SC, Shan JJ, Ji JJ, Lin LL, Li WW, Yuan B (2019) Urine
metabolic profiles in paediatric asthma. Respirology 24(6):572–581
Metabolomics of Respiratory Diseases 365

Van Vliet D, Smolinska A, Jobsis Q, Rosias PP, Muris JW, Dallinga JW, van Schooten FJ,
Dompeling E (2016) Association between exhaled inflammatory markers and asthma control
in children. J Breath Res 10(1):016014
van Vliet D, Smolinska A, Jobsis Q, Rosias P, Muris J, Dallinga J, Dompeling E, van Schooten FJ
(2017) Can exhaled volatile organic compounds predict asthma exacerbations in children? J
Breath Res 11(1):016016
Veerappan A, Oskuei A, Crowley G, Mikhail M, Ostrofsky D, Gironda Z, Vaidyanathan S,
Wadghiri YZ, Liu M, Kwon S, Nolan A (2020) World trade center-cardiorespiratory and
vascular dysfunction: assessing the phenotype and metabolome of a murine particulate matter
exposure model. Sci Rep 10(1):3130
Vlahos R (2020) Lipids in chronic obstructive pulmonary disease: a target for future therapy? Am J
Respir Cell Mol Biol 62(3):273–274
Vos T, Lim SS, Abbafati C, Abbas KM, Abbasi M, Abbasifard M, Abbasi-Kangevari M,
Abbastabar H, Abd-Allah F, Abdelalim A, Abdollahi M et al (2020) Global burden of
369 diseases and injuries in 204 countries and territories, 1990–2019: a systematic analysis
for the global burden of disease study 2019. Lancet 396(10258):1204–1222
Ware LB, Matthay MA (2000) The acute respiratory distress syndrome. N Engl J Med 342(18):
1334–1349
Wei W, Wu X, Bai Y, Li G, Meng H, Feng Y, Li H, Li M, Guan X, Fu M, Wang C et al (2021)
Arsenic exposure and its joint effects with cigarette smoking and physical exercise on lung
function impairment: evidence from an occupational cohort study. Environ Res 196:110419
Wheelock CE, Goss VM, Balgoma D, Nicholas B, Brandsma J, Skipp PJ, Snowden S, Burg D,
D'Amico A, Horvath I, Chaiboonchoe A et al (2013) Application of 'omics technologies to
biomarker discovery in inflammatory lung diseases. Eur Respir J 42(3):802–825
Wishart DS (2005) Metabolomics: the principles and potential applications to transplantation. Am J
Transplant 5(12):2814–2820
Wu D, Hou C, Li Y, Zhao Z, Liu J, Lu X, Shang X, Xin Y (2014) Analysis of the bacterial
community in chronic obstructive pulmonary disease sputum samples by denaturing gradient
gel electrophoresis and real-time PCR. BMC Pulm Med 14(1):179
Xia J, Broadhurst DI, Wilson M, Wishart DS (2013) Translational biomarker discovery in clinical
metabolomics: an introductory tutorial. Metabolomics 9(2):280–299
Xue M, Cai C, Guan L, Xu Y, Lin J, Zeng Y, Hu H, Chen R, Wang H, Zhou L, Sun B (2020)
Exploration of n-6 and n-3 polyunsaturated fatty acids metabolites associated with nutritional
levels in patients with severe stable chronic obstructive pulmonary disease. Int J Chron Obstruct
Pulmon Dis 15:1633–1642
Young RP, Hopkins RJ, Marsland B (2016) The gut-liver-lung axis. Modulation of the innate
immune response and its possible role in chronic obstructive pulmonary disease. Am J Respir
Cell Mol Biol 54(2):161–169
Yu B, Flexeder C, McGarrah RW 3rd, Wyss A, Morrison AC, North KE, Boerwinkle E,
Kastenmuller G, Gieger C, Suhre K, Karrasch S et al (2019) Metabolomics identifies novel
blood biomarkers of pulmonary function and COPD in the general population. Meta 9(4):61
Zakharkina T, Heinzel E, Koczulla RA, Greulich T, Rentz K, Pauling JK, Baumbach J,
Herrmann M, Grunewald C, Dienemann H, von Muller L et al (2013) Analysis of the airway
microbiota of healthy individuals and patients with chronic obstructive pulmonary disease by
T-RFLP and clone sequencing. PLoS One 8(7):e68302
Zhang R, Chen L, Cao L, Li KJ, Huang Y, Luan XQ, Li G (2018) Effects of smoking on the lower
respiratory tract microbiome in mice. Respir Res 19(1):253
Zhou J, Li Q, Liu C, Pang R, Yin Y (2020) Plasma metabolomics and Lipidomics reveal perturbed
metabolites in different disease stages of chronic obstructive pulmonary disease. Int J Chron
Obstruct Pulmon Dis 15:553–565
The Metabolomics of Critical Illness

Ana E. Pacheco-Navarro and Angela J. Rogers

Contents
1Introduction: A Potential Role for Metabolomics in Understanding Critical Illness . . . . . 368
2Metabolic Alterations in Critical Illness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 369
3Endocrine Drivers of Altered Metabolism . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 369
4Adrenergic Mediators and Their Changes in Critical Illness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 371
5Inflammatory Mediators (Cytokines) and Their Contribution to Metabolism . . . . . . . . . . . . 371
6Metabolic Phases During Critical Illness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 371
7Autophagy and Mitophagy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 372
8Nutrient Changes During Critical Illness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 373
8.1 Lactate: The Original Metabolomic Biomarker . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 374
9 Metabolomics in Sepsis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 375
9.1 Diagnosis of Sepsis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 375
9.2 Prognostication in Sepsis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 376
9.3 Future Directions: Using Metabolomics to Guide Treatment . . . . . . . . . . . . . . . . . . . . . . . 377
10 Metabolomics Studies in ARDS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 377
10.1 Blood Metabolomics of ARDS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 378
10.2 Using Blood Metabolomics to Identify High-Risk Subsets of ARDS Cohorts . . . 379
10.3 Pulmonary Specific Samples for Metabolomic Profiling . . . . . . . . . . . . . . . . . . . . . . . . . . 379
11 Metabolomics of Critical Illness: Future Directions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 380
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 381

Abstract
Critical illness is associated with dramatic changes in metabolism driven by
immune, endocrine, and adrenergic mediators. These changes involve early
activation of catabolic processes leading to increased energetic substrate avail-
ability; later on, they are followed by a hypometabolic phase characterized by

A. E. Pacheco-Navarro · A. J. Rogers (✉)

Division of Pulmonary, Allergy and Critical Care Medicine, Stanford University School of
Medicine, Stanford, CA, USA
e-mail: [email protected]

# The Author(s), under exclusive license to Springer Nature Switzerland AG 2022 367
V. Ghini et al. (eds.), Metabolomics and Its Impact on Health and Diseases,
Handbook of Experimental Pharmacology 277, https://doi.org/10.1007/164_2022_622
368 A. E. Pacheco-Navarro and A. J. Rogers

deranged mitochondrial function. In sepsis and ARDS, these rapid clinical

changes are reflected in metabolomic profiles of plasma and other fluids,
suggesting that metabolomics could one day be used to assist in the diagnosis
and prognostication of these syndromes. Some metabolites, such as lactate, are
already in clinical use and define patients with septic shock, a high-mortality
subtype of sepsis. Larger-scale metabolomic profiling may ultimately offer a tool
to identify subgroups of critically ill patients who may respond to therapy, but
further work is needed before this type of precision medicine is readily employed
in the clinical setting.

Keywords
Acute respiratory distress syndrome (ARDS) · Metabolism · Metabolomics ·
Prognostication · Sepsis

1 Introduction: A Potential Role for Metabolomics

in Understanding Critical Illness

Each year, over four million critically ill Americans are admitted into intensive care
units (ICUs) around the country, a number that has only increased since the start of
the COVID-19 pandemic (ICU Outcomes 2022). Despite advances in medicine, ICU
mortality remains high, ranging between 8 and 20%, with approximately 500,000
American deaths per year. In particular, sepsis – a systemic inflammatory illness
caused by infection with a dysregulated host immune response – and the acute
respiratory distress syndrome (ARDS) – a syndrome of respiratory failure with a
diverse range of etiologies – are particularly challenging illnesses in the ICU, with
mortality upward of 40% (Fleischmann et al. 2016; Gaieski et al. 2013; Rudd et al.
2020).
Part of the challenge in treating these syndromes lies in their significant biological
heterogeneity. The majority of clinical trials in sepsis and ARDS have been negative,
and available treatment remains largely supportive and directed at treating underly-
ing infections, while not specifically addressing the inflammatory state. Recent
re-analyses of a number of negative trials have revealed that subgroups of patients
appear to respond to the intended therapy (Calfee et al. 2014; Puskarich et al. 2021),
underscoring the heterogeneity of these syndromes and the need for better
diagnostics in identifying high-risk patients and matching them with potential
treatments.
Given the dramatic and rapid metabolic alterations that are observed during
critical illness, metabolomics offers a powerful potential tool for better identifying
subphenotypes of critically ill patients who may differentially respond to treatments
and also identifying novel treatment pathways.
The Metabolomics of Critical Illness 369

2 Metabolic Alterations in Critical Illness

In health, energetic needs are largely met by nutrient intake through food, rather than
catabolism of the body’s macromolecules. ATP, the main source of cellular energy,
is derived by aerobic metabolism, and anabolic and catabolic processes are in
balance.
In critical illness, however, dramatic changes in the endocrine and autonomic
nervous system alter these metabolic processes, shifting them toward a catabolic
state in which the body’s carbohydrates, lipids, and protein stores are used to meet
increased energetic needs. This imbalance is often further driven by reduced caloric
intake, as critically ill patients frequently suffer from symptoms like encephalopathy
and anorexia that curb eating. Clinically, such catabolism leads to skeletal muscle
wasting, weakness, and prolonged recovery (Casaer and Van den Berghe 2014).

3 Endocrine Drivers of Altered Metabolism

Initially, acute stress such as that caused by sepsis and other forms of critical illness
leads to an inflammatory cascade which includes the release of cytokines, including
IL-1, IL-2, and IL-6 that can activate the hypothalamic-pituitary-adrenal axis. This
can occur through direct cytokine production by glial cells in the brain (Khardori and
Castillo 2012; Englert and Rogers 2016), and through free diffusion into the pitui-
tary, which is outside the blood brain barrier. This cytokine-derived activation
results in the release of corticotropin (ACTH) from the pituitary, which stimulates
production and secretion of cortisol from the adrenal glands (Englert and Rogers
2016; Cooper 2003). In addition to its increased production by the adrenal gland,
decreases in corticosteroid binding globulin during critical illness and impaired
hepatic metabolism of the hormone contribute to increased levels of free, biologi-
cally active cortisol (Wasyluk et al. 2021; Marik et al. 2008).
Cortisol acts as the primary endocrine mediator in sepsis, leading to a variety of
changes in metabolism, immune function, and cardiovascular function critical to
maintaining homeostasis (Fig. 1). Metabolically, it leads to increased blood glucose
through a variety of mechanisms: by activation of enzymes involved in gluconeo-
genesis, increasing insulin resistance, and decreasing glucose uptake by skeletal
muscle (Marik et al. 2008). It activates lipolysis in adipose tissue, leading the release
of free fatty acids (Marik et al. 2008).
In addition to increasing energy availability by increasing glucose and fatty acid
concentrations, cortisol has a major role in increasing blood pressure. In vascular
smooth muscle, it increases transcription of surface receptors for catecholamines and
angiotensin II, resulting in increased sensitivity to these vasoactive hormones.
Cortisol also plays a complex role in regulating and abrogating the immune
response, acting in a negative feedback loop that both reduces the number of
circulating immune cells and decreases their production of cytokines.
370 A. E. Pacheco-Navarro and A. J. Rogers

Fig. 1 Direct metabolic effects of cortisol. Figure created with biorender.com

While increased cortisol synthesis and secretion is part of the body’s typical
response to stress, this response can be attenuated or halted in a variety of conditions
in the setting of critical illness. Head injuries, central nervous system depressants,
and pituitary infarction can inhibit the CRH and corticotropin driven portion of the
hypothalamic-pituitary-adrenal axis (Cooper 2003). At the level of the adrenal
glands, drugs such as the induction agent etomidate – frequently used during
intubation – can impair cortisol synthesis, and adrenal hemorrhage, though rare,
can occur in septic patients with coagulopathies (Cooper 2003). More commonly,
chronic intake of exogenous corticosteroids – e.g., in patients with rheumatologic
disease – can lead to adrenal atrophy and a sluggish or absent response to
corticotropin.
While absolute adrenal insufficiency is rare in critically ill patients, studies have
shown that relative adrenal insufficiency – as defined by inadequate increase in
cortisol level in response to a corticotropin challenge – is common in patients with
septic shock (Annane et al. 2002). In 2007, a consensus statement between the
Society of Critical Care Medicine and the European Society of Intensive Care
Medicine first coined the term “CIRCI” (critical illness-related corticosteroid insuf-
ficiency) to recognize that many critically ill patients have an impaired
hypothalamic-pituitary-adrenal response.
Despite this, it is unclear whether such patients benefit from hormone replace-
ment (Annane et al. 2002; Sprung et al. 2008), as a number of large randomized
controlled trials studying the use of steroids in sepsis have had conflicting results
(Annane et al. 2002, 2018; Sprung et al. 2008; Venkatesh et al. 2018). At the
The Metabolomics of Critical Illness 371

minimum, however, these studies suggest that steroid administration in critically ill
patients is safe and should be considered, particularly for patients with refractory
shock. As in the sepsis literature, the benefits of steroids are unclear in ARDS, but
recent high-quality studies (Villar et al. 2020) – as well as the evidence in severe
COVID-19 (RECOVERY Collaborative Group et al. 2021; Tomazini et al. 2020) –
suggest that they may be beneficial.

4 Adrenergic Mediators and Their Changes in Critical Illness

The sympathetic nervous system is rapidly activated in the setting of critical illness.
These stress signals can occur at the level of the peripheral nerves (e.g., in the case of
trauma), by chemoreceptors (in the case of hypoxemia or hypercapnia), or by
baroreceptors (in the setting of shock). These events all result in the release of
norepinephrine from post-ganglionic neurons, and secretion of epinephrine and
norepinephrine from chromaffin cells in the adrenal medulla (Preiser et al. 2014).
Along with exogenously delivered catecholamines that may be administered
therapeutically, endogenous dopamine, norepinephrine, and epinephrine play
major roles in activating catabolic pathways and increasing nutrient availability.
Like cortisol, these hormones work to increase plasma glucose concentrations by
upregulating hepatic gluconeogenesis and glycogenolysis and increasing insulin
resistance.

5 Inflammatory Mediators (Cytokines) and Their

Contribution to Metabolism

In addition to the endocrine and sympathetic nervous systems, cytokines produced

by immune cells play important roles in mediating metabolic changes during critical
illness. Most heavily implicated are tumor necrosis factor α – previously known in
the cancer literature as “cachectin” for its upregulation of lipolysis – interleukin-1β
(IL-1β), and interleukin-6 (Il-6). TNFα is a potent inducer of the other two, and
together, they work to increase insulin resistance, proteolysis, and lipolysis.

6 Metabolic Phases During Critical Illness

Sepsis and other forms of critical illness are characterized by an early stress response,
as described above, associated with increased catabolism and energy availability.
This early phase is often followed by a period of decreased metabolic activity and
organ dysfunction (Singer et al. 2004; Pool et al. 2018). Initially, the organ dysfunc-
tion and elevations in lactate that are frequently seen in sepsis were thought to result
from impaired tissue perfusion, but studies have since shown that although
malperfusion may play a role in the early phase of critical illness, tissue oxygen
372 A. E. Pacheco-Navarro and A. J. Rogers

delivery is adequate in the later phases of sepsis (Sair et al. 2001). Instead, cellular
changes in metabolism – decreased oxidative phosphorylation and greater reliance
on glycolytic pathways – termed cytopathic hypoxia – likely underly this organ
dysfunction. These changes are hypothesized to be adaptive mechanisms to preserve
energy for critical functions, prevent DNA damage, and limit additional injury.
In the heart, for example, sepsis and critical illness can result in reduced cardiac
myocyte contractility through a process called myocardial hibernation, which
preserves myocyte ATP levels at the cost of reduced function. In animal models,
this change in contractility has been linked to inhibition of the oxidative phosphory-
lation enzyme cytochrome oxygenase, and this phenotype can be rescued by deliv-
ery of exogenous cytochrome c (Piel et al. 2007). Similar mitochondrial changes
have also been implicated in animal models of sepsis-induced dysfunction of the
lung, kidney, and liver (Pool et al. 2018). In the kidney, the presence of
pro-inflammatory cytokines or lipopolysaccharide from bacteria leads to
downregulation of energy-intensive sodium and chloride channels (Schmidt et al.
2007). Despite these observations, the full spectrum of metabolic changes during the
phases of critical illness is not known. Metabolomics provides a potential avenue
through which changes in multiple metabolic pathways can be better described.

7 Autophagy and Mitophagy

The mitochondrial dysfunction and electron transport chain inhibition described

above likely result in increased autophagy, a cellular housekeeping mechanism
that results in the lysosomal degradation of damaged and dysfunctional organelles
and proteins. In brief, autophagy is upregulated by cellular/organelle damage – e.g.,
the mitochondrial membrane depolarization that might occur due to uncoupled
respiration from an impaired electron transport chain – and inhibited by increased
nutrients (Vanhorebeek et al. 2011).
Despite being a mechanism for intracellular protein catabolism, autophagy has
been shown to play an important role in maintaining the health of myocytes in
skeletal muscle, and inhibition of autophagy has been linked to accelerated muscle
loss during fasting and muscle denervation (Masiero et al. 2009). For critically ill
patients, who are in a persistent catabolic state and are at very high risk of muscle
wasting and prolonged weakness (termed critical illness myopathy), the relationship
between fasting and autophagy has been of increasing interest. In the 4,600+ patient
EPaNIC trial (Early Parenteral Nutrition Completing Enteral Nutrition in Adult
Critically Ill Patients), the effect of initiating parenteral nutrition (PN) on an early
time scale (within 48 h of ICU admission) versus a late one (within 8 days) was
compared. Late parenteral nutrition initiation was associated with faster recovery
(as measured by ICU and hospital discharge) and fewer complications (Casaer et al.
2011). Significantly, in a prospectively planned subanalysis of the EPaNIC trial, in
which a subset of patients underwent skeletal muscle biopsies 8 days after randomi-
zation and sequential strength testing through 100 days, patients randomized to the
late PN group were found to have faster recovery of weakness and greater evidence
The Metabolomics of Critical Illness 373

of autophagosome formation (Hermans et al. 2013). Taken together, this finding

suggests that a tempered approach to early nutrition in the critically ill patient may
allow for more autophagy and faster recovery of skeletal muscle strength.

8 Nutrient Changes During Critical Illness

Critical illness leads to significant changes in the metabolism of carbohydrates,

protein, and lipids. Most notably, increased glycogenolysis, gluconeogenesis, and
reduced sensitivity to insulin lead to increased plasma glucose concentrations
through the endocrine, adrenergic, and inflammatory mechanisms described above
(Table 1).
Catabolism of proteins and lipids is also upregulated, particularly in early sepsis.
Amino acids generated in proteolytic processes are shuttled to the liver for synthesis
of acute phase reactants. Protein breakdown and a net negative nitrogen balance lead
to the wasting of skeletal muscles, which will contribute to deconditioning and
prolong patients’ recoveries. In adipose tissue, increased lipolysis leads to increased
serum triglycerides and free fatty acids and decreased circulating lipoproteins.
Micronutrient levels also fluctuate during critical illness. Of these, changes in
selenium and zinc levels in sepsis have been particularly well-described. Selenium
and zinc are both trace elements with important anti-inflammatory and antioxidant
properties, and levels of both minerals have been found to be lower in septic patients
than in healthy controls (Allingstrup and Afshari 2015; Besecker et al. 2011).
Supplementation of these micronutrients in the critically ill has been investigated
with promising initial results (Manzanares et al. 2012), but due to small study sizes,
further work is needed before their use in the ICU becomes widespread.

Table 1 Summary of metabolic changes in sepsis

Physiologic change in sepsis Metabolic impact
" Gluconeogenesis, glycolysis, Hyperglycemia
glycogenolysis
" Protein catabolism Altered circulating amino acids
" Lipolysis " Triglycerides, # lipoproteins
# Micronutrients " Oxidative stress
" Neuroendocrine activation " Catecholamines, " counterregulatory
hormones
" Cortisol Hyperglycemia
" Catecholamine release " Gluconeogenesis, " glycolysis
" Cytokine release Hyperglycemia, insulin resistance
Impaired oxygen utilization " Reactive oxygen species
Adapted from Englert and Rogers (Englert and Rogers 2016) with their permission
374 A. E. Pacheco-Navarro and A. J. Rogers

8.1 Lactate: The Original Metabolomic Biomarker

Plasma lactate is one of the most validated and frequently used biomarkers in the
intensive care unit. Lactate is a by-product of anerobic metabolic pathways, in which
glucose is broken down into two pyruvate molecules and then fermented to 2 lactate
molecules in a net reaction that produces 2 molecules of adenosine triphosphate
(ATP) and regenerates the NAD+ cofactor used in glycolysis (Fig. 2). Accordingly,
rising blood lactate levels can indicate inadequate tissue perfusion and oxygenation
or a perfusion-independent shift toward anerobic pathways as can occur in malig-
nancy (the Warburg effect) or in the later phases of sepsis.
In the ICU, lactate can be used for prognostication and also to guide resuscitation
during critical illness. From a risk stratification perspective, hospitalized patients
with infection-related admissions had a significantly higher risk of death if their
serum lactate was >4 mmol/L (as compared to those with normal concentrations),
and even modestly elevated lactates of 2.5–4 mmol were associated with increased
mortality (Shapiro et al. 2005). Similarly, in a study of septic patients in the
emergency department, intermediate and high lactate levels (2.5–3.9 mmol/L and
>4 mmol/L) were associated with increased mortality even in patients with initially
normal blood pressures and without evidence of organ dysfunction (Mikkelsen et al.
2009).
Lactate levels have also been applied to the goal-directed paradigm of sepsis
management and have been used to guide resuscitation, though its role in this
context has been less well-established. In the EMShockNet trial which compared
the use of ScvO2 to serial lactates in the goal-directed resuscitation of patients

Fig. 2 Glycolysis and lactate

fermentation. Adapted from
“Warburg Effect,” from
Biorender.com (2022).
Retrieved from https://app.
biorender.com/biorender-
templates and created in
biorender.com
The Metabolomics of Critical Illness 375

(in which both groups also had MAP and CVP goals), there was a trend toward
reduced mortality in the lactate group, but this difference was not statistically
significant (Jones et al. 2010). Similarly, in another trial comparing serial lactate-
guided therapy to traditional early goal-directed therapy as described by Rivers et al.,
mortality trended lower in the lactate group, but was just outside of the predefined
alpha for statistical significance ( p = 0.067) (Jansen et al. 2010).

9 Metabolomics in Sepsis

Sepsis, or life-threatening organ dysfunction in response to infection (Singer et al.

2016), is one of the leading causes of death worldwide, accounting for an estimated
five million deaths each year and many more hospitalizations (Fleischmann et al.
2016). The current treatment strategy is based on the delivery of early, broad-
spectrum antibiotics and is otherwise supportive. Because of the disease’s heteroge-
neity and the limited treatments available, precision medicine approaches – in which
subgroups of patients with different mechanistic drivers are identified, leading to
different treatment approaches – are appealing; these strategies have been effectively
applied to a range of diseases, from cancer to asthma. While broader plasma
metabolomic profiling studies have yet to impact care of septic patients to date, we
will discuss overarching trends in the field with selected examples from larger
studies and their clinical implications below.

9.1 Diagnosis of Sepsis

Noninfectious inflammatory responses – as might occur in pancreatitis or trauma –

can resemble sepsis, delaying the diagnosis until culture data are finalized and
potentially contributing to the administration of unnecessary antibiotics and increas-
ing the development of drug-resistant bacteria. Metabolomics has been used to
differentiate sepsis from sterile inflammation (Mickiewicz et al. 2013, 2014; Langley
et al. 2014). In these studies, plasma samples from septic patients (or in one study,
infected primates) were compared to those of critically ill uninfected controls with
evidence of systemic inflammation, key metabolic differences were identified, then
validated in an independent cohort of critically ill patients.
In one such study, Mickiewicz et al. (2013) used supervised analysis to investi-
gate 58 metabolites in the serum of 60 children with early septic shock and compared
profiles to those of 40 noninfected pediatric ICU patients with systemic inflamma-
tory response syndrome (SIRS) and 40 healthy children. They used principal
component analysis and partial least squares-discriminant analysis to identify
18 metabolites that differed in those who survived from those who did not. Interest-
ingly, the responses of infants and toddlers differed from those of school-aged
children, suggesting that the metabolic changes associated with sepsis may differ
depending on life stage. Three metabolites (2-hydroxybutyrate,
2-hydroxyisovalerate, and lactate) involved in fat breakdown and anerobic
376 A. E. Pacheco-Navarro and A. J. Rogers

metabolism were elevated in septic patients as compared to controls regardless of

age. Other studies reinforced these differences in lipid metabolism (Langley et al.
2013; Rogers et al. 2021) and also found differences in amino acid metabolism,
mitochondrial metabolism, and the citric acid cycle and pentose phosphate pathway
in septic patients versus noninfected inflamed controls (Eckerle et al. 2017).
More recently, Rydzak et al. used a metabolomic approach for rapid identification
of blood stream infection pathogens and their antimicrobial susceptibility patterns
(Rydzak et al. 2022), a highly valuable tool in sepsis, where blood stream infections
contribute to significant morbidity and mortality. They used a metabolic preference
assay that used the patterns of consumed and excreted metabolites of ex vivo
microbial cultures to identify pathogens and their antibiotic susceptibilities, monitor-
ing for changes in metabolite production in the presence of toxic antibiotics. They
found that just seven metabolites were sufficient to differentiate between the seven
most prevalent pathogens responsible for blood stream infections and identify their
antimicrobial susceptibilities in half the time of traditional microbial identification
and antimicrobial susceptibility testing protocols.

9.2 Prognostication in Sepsis

Sepsis has a wide range of outcomes, from full recovery to long-term disability or
death (Singer et al. 2016). As such, it is not surprising that many studies in critical
care metabolomics are focused on prognostication. In these studies, septic patients
with worse outcomes (death or multiorgan failure) are compared to survivors or
individuals with less organ failure.
Using targeted metabolomic profiling of more than 300 metabolites, Langley
et al. evaluated the plasma metabolomes and proteomes in two independent cohorts
(Langley et al. 2014). The group used 150 adult patients from the Community
Acquired Pneumonia and Sepsis Outcome Diagnostics (CAPSOD) study, which
enrolled patients who presented to the emergency department of three US hospitals
with symptoms of sepsis (Langley et al. 2013) as a discovery cohort, and used
patients with sepsis from the Brigham and Women’s Hospital Registry of Critical
Illness (RoCI) cohort as a validation set. They found that the metabolic signatures of
survivors differed markedly from those that died. In particular, levels of proteins
involved in fatty acid transport and β-oxidation, gluconeogenesis, and the citric acid
cycle were significantly higher in those that died. The group used Support Vector
Machine analysis to create an algorithm using two clinical features and five metabo-
lite levels to predict survival in sepsis, using the RoCI data as a validation set.
Though the RoCI cohort had higher mortality and more cancer as compared to the
discovery cohort, the metabolic changes identified and the sepsis prognostication
model were validated in the independent cohort.
Rogers et al. also used the CAPSOD and RoCI cohorts to perform targeted
metabolic profiling on sepsis survivors and nonsurvivors (Rogers et al. 2014).
Although the Langley and Rogers groups identified many of the same metabolites
that differed in survivors and nonsurvivors, the Bayesian model developed by the
The Metabolomics of Critical Illness 377

Rogers group using RoCI as the training cohort was very different, using seven
completely different metabolites than those used in the Langley algorithm. This may
reflect a redundancy of metabolites found in the same metabolic pathways, but also
illustrates the evolving nature of analytic methodology in the field, in which the
optimal statistical methods for dealing with highly correlated data are not yet known.

9.3 Future Directions: Using Metabolomics to Guide Treatment

One of the most promising applications for the use of metabolomics in critical care is
in the identification of patients who might differentially respond to therapies. Sepsis
treatment is currently focused on early antimicrobial administration and is otherwise
supportive, as no mechanism-targeted therapies are approved for use in the syn-
drome. The failure of many studies to identify effective, mechanism-based
treatments has been attributed to sepsis’s significant heterogeneity, which may
mask individual or subgroup responses to therapies in clinical trials.
Along these lines, Puskarich et al. used metabolomics to further phenotype
patients in a reanalysis of a phase II study of L-carnitine infusion for the treatment
of septic shock (Puskarich et al. 2015). The group performed metabolomics on
existing serum samples from septic patients in the study and found that three
metabolites involved in the synthesis and depredation of ketones differed in the
pre-intervention samples of L-carnitine-treated survivors and nonsurvivors. In later
work, they went on to identify acetylcarnitine as a particularly powerful biomarker
for predicting L-carnitine mortality benefit (Puskarich et al. 2021). While the
findings of retrospective reanalysis of a randomized controlled trials are likely
insufficient to change clinical practice, the hope is that metabolomics can be used
in the future to identify potential drug responders prior to their inclusion into
prospective clinical trials.

10 Metabolomics Studies in ARDS

Like sepsis, ARDS is a highly heterogeneous and clinically important disease,

defined by acute respiratory failure in the setting of a risk factor (such as sepsis,
pneumonia, trauma, aspiration), bilateral infiltrates on imaging and low levels of
blood oxygen (Ferguson et al. 2012; Thompson et al. 2017). It is estimated to affect
>100,000 Americans per year with a mortality rate over 25% despite clinical
improvements in recent decades (Bellani et al. 2016; Rubenfeld et al. 2005); those
numbers predate the COVID epidemic, with ARDS and hypoxemic respiratory
failure leading to the deaths of millions worldwide.
The heterogeneity of ARDS has no doubt contributed to a lack of effective
therapies. ARDS mortality has fallen because of improvements in approaches to
mechanical ventilation which decrease the pressure and volumes delivered to the
lung, preventing further lung injury. But despite more than 50 years of research and
substantial improvements in our understanding of the pathogenic underpinnings of
378 A. E. Pacheco-Navarro and A. J. Rogers

lung injury and inflammation, not a single drug had been approved to treat ARDS
prior to COVID-19 (Matthay et al. 2019). In COVID-19, numerous drug trials have
proved effective for the syndrome, likely because of increased homogeneity in the
patient population: all cases share an underlying etiology of SARS-CoV2 infection,
a typical time course of disease, and a substantial inflammatory component. Multiple
anti-inflammatory drug therapies are now available for COVID ARDS patients
(RECOVERY Collaborative Group et al. 2021; Investigators et al. 2021; Kalil
et al. 2021).
Metabolomics studies in ARDS are compelling because of the dynamic
metabolomic changes in the blood and lungs, with the potential for providing a
real-time snapshot of disease and identifying high-risk subjects who could be
targeted for trials (Millet, “Prospective metabolomic studies in precision medicine.
The AKRIBEA project.”; Turano, “NMR-based metabolomics to evaluate individ-
ual response to treatments”). Multiple studies have examined metabolomics of
ARDS; major studies are summarized below and in a thorough recent review
(Metwaly and Winston 2020).

10.1 Blood Metabolomics of ARDS

Several groups have performed metabolomic profiling of blood in ARDS patients

and controls. As noted, there is substantial variability between studies, including
(1) size of cohort, (2) how the control cohort is defined (healthy vs critically ill but
without ARDS), and (3) the metabolomic profiling methodology used. Given these
disparate methodologies and relatively small cohort sample sizes, it is perhaps
unsurprising that the metabolites and potential pathways identified vary
substantially.
Metwaly et al. examined 108 patients with ARDS, matched by age and gender
with 27 mechanically ventilated controls (Metwaly and Winston 2020). They
identified numerous pathways that differed in ARDS vs controls that were significant
in both a training and testing cohort, with a particular focus on serine-glycine
metabolism. As in prior studies, the mechanically ventilated controls were systemi-
cally much less sick than subjects with ARDS (e.g., less frequently required
vasopressors, less kidney dysfunction, and markedly lower mortality).
Rogers et al. examined a large cohort of 78 ARDS cases and compared them to
75 critically ill controls with sepsis (Rogers et al. 2021). While there were numerous
individual metabolites that differed, partial least squares-discriminant analysis
(PLS-DA) could not separate ARDS from critically ill controls, who frequently
had concomitant organ dysfunction (e.g., renal failure, liver failure, or GI
malperfusion). In fact, in this cohort, many of the previously identified “ARDS
metabolites” were associated with the Simplified Acute Physiology (SAPS) score,
which is predictive of general ICU mortality, and were not ARDS-specific. This
work suggests that many of the metabolic changes previously identified in ARDS
may reflect systemic illness and inflammation rather than ARDS per se. As these
The Metabolomics of Critical Illness 379

studies demonstrate, how to adjust for concomitant systemic organ dysfunction is an

active area of investigation in the metabolomics of ARDS.

10.2 Using Blood Metabolomics to Identify High-Risk Subsets

of ARDS Cohorts

Some of the most high-profile advances in ARDS pathogenesis in the past decade
have involved identifying high-risk subsets of patients who are at increased risk of
bad outcomes. Most of these risk markers have involved cytokines and proteins, for
example angiopoietin 2 (Calfee et al. 2012; Zinter et al. 2016), IL-18 (Rogers et al.
2019; Dolinay et al. 2012), and a hypoinflammatory/hyperinflammatory latent class
analysis (LCA) designation which can be assigned using a parsimonious model of
3 blood biomarkers (IL-6, TNF receptor alpha, and bicarbonate) (Calfee et al. 2014;
Sinha et al. 2020). The latent class hyperinflammatory subset has been particularly
well-studied and shown to be present in more than six ARDS populations (both
clinical trials and clinical cohorts) and associated with differential treatment
response in numerous ARDS clinical trials, including approach to positive
end-expiratory pressure (PEEP) level, fluid management, and statin therapy (Calfee
et al. 2014, 2018; Famous et al. 2017).
Fewer studies have focused on using metabolomics to identify high-risk subsets
of ARDS. Metwaly et al. tested whether orthogonal projections to latent structures
discriminant analysis (OPLS-DA) could differentiate previously identified subsets of
ARDS, including direct pulmonary risk factor for developing ARDS (e.g., pneumo-
nia) vs indirect risk factors (e.g., pancreatitis or non-pulmonary sepsis), and
hyperinflammatory subphenotype using latent class (Metwaly et al. 2021). Indeed,
using metabolites identified via high-VIP score for OPLS-DA, these subsets could
be separated in both testing and training sets; metabolomics outperformed protein
analyses in separating the direct and indirect classes.
The Rogers et al. study of 197 patients included 75 patients with ARDS (Rogers
et al. 2021). As discussed above, PLS-DA was not able to differentiate patients with
sepsis from ARDS. However, unbiased metabolic clustering of the entire cohort did
reveal three subsets of patients, with differential metabolites marked by lipid classes
driving the subclasses, with markedly different risks of mortality, regardless of
ARDS status.

10.3 Pulmonary Specific Samples for Metabolomic Profiling

Perhaps one of the most promising aspects of metabolomic profiling is the ability to
examine trace amounts of a lung-specific sample which could be captured without
need for a tissue biopsy to use for metabolic profiling (e.g., bronchoalveolar lavage
fluid, expired breath, or free-flowing pulmonary edema fluid). Such non-blood
samples may overcome the non-specific metabolic changes that occur with critical
illness and allow for a more lung injury-specific signature to emerge.
380 A. E. Pacheco-Navarro and A. J. Rogers

To date, lung-specific metabolomics studies have been limited by similar

challenges that have been seen in blood cohorts: (1) fairly small sample size,
particularly given the inherent heterogeneity of ARDS, (2) varying control popula-
tion, and (3) differing lung sample fluids. Not surprisingly, given such disparities in
sampling, there are no common metabolites emerging that consistently differentiate
ARDS from non-ARDS fluid.
Only two studies to date examined the unbiased identification of high-risk subsets
within ARDS using non-blood samples. In a very small study by Rogers et al.
investigating the metabolome of pulmonary edema fluid in ARDS versus congestive
heart failure, ARDS edema could not be reliably distinguished from cardiogenic
edema via PLS-DA, but hierarchical clustering did separate a metabolically distinct
subset of one-third of the sample which was at markedly increased risk of mortality
(Rogers et al. 2017).
Viswan and colleagues performed a large study, examining both blood and BAL
fluid in >150 ARDS patients to examine metabolites associated with three
definitions of high-risk subsets of ARDS: those defined by (1) ARDS severity by
PaO2:FIO2 ratio, (2) direct vs indirect ARDS, and (3) survivors and nonsurvivors
(Viswan et al. 2019). They did identify numerous associations between subsets,
including overlap between the BAL and blood metabolites identified in high-risk
subsets, though some methodology is unclear (Rogers et al. 2021; Investigators et al.
2021).
ARDS is a clinically-defined syndrome which carries substantial risk of mortality.
It is marked by both clinical and temporal heterogeneity, suggesting that a focus on
metabolomics of blood and lung biomarkers could be fruitful. Though numerous
metabolomics studies have been published in the last 2 decades, more work in large
cohorts is needed to identify reproducible signatures that provide targetable meta-
bolic subgroups of ARDS. Similarly, to what extent readily available plasma
samples provide an adequate metabolic sample for lung biology, versus reflect
systemic illness regardless of ARDS status requires ongoing study.

11 Metabolomics of Critical Illness: Future Directions

Both sepsis and ARDS are heterogenous and life-threatening clinical syndromes in
which a dysregulated immune response triggers dramatic changes in metabolism.
Identifying targeted and effective treatments for these syndromes remains a chal-
lenge. The current therapeutic paradigm hinges on treating underlying infections and
otherwise providing supportive care; no effective therapies have been found to treat
the metabolic and inflammatory derangements that drive these syndromes.
Metabolomics offers a potential tool for better identifying the biological
mechanisms underlying these syndromes and has already shown promise in prog-
nostication of both sepsis and ARDS. Further work will be needed to more cleanly
differentiate the lung-specific injury that occurs in ARDS from the frequently
overlapping changes seen in sepsis. Ultimately, metabolomics offers the opportunity
to uncover important biologic pathways that may serve as therapeutic targets and
The Metabolomics of Critical Illness 381

identify subgroups of patients in both sepsis and ARDS who may respond differen-
tially to current and future therapeutics.

References
Allingstrup M, Afshari A (2015) Selenium supplementation for critically ill adults. Cochrane
Database Syst Rev (7):CD003703. https://doi.org/10.1002/14651858.CD003703.pub3
Annane D, Sébille V, Charpentier C et al (2002) Effect of treatment with low doses of hydrocorti-
sone and fludrocortisone on mortality in patients with septic shock. JAMA 288(7):862–871.
https://doi.org/10.1001/jama.288.7.862
Annane D, Renault A, Brun-Buisson C et al (2018) Hydrocortisone plus fludrocortisone for adults
with septic shock. N Engl J Med 378(9):809–818. https://doi.org/10.1056/NEJMoa1705716
Bellani G, Laffey JG, Pham T et al (2016) Epidemiology, patterns of care, and mortality for patients
with acute respiratory distress syndrome in intensive care units in 50 countries. JAMA 315(8):
788–800. https://doi.org/10.1001/jama.2016.0291
Besecker BY, Exline MC, Hollyfield J et al (2011) A comparison of zinc metabolism, inflammation,
and disease severity in critically ill infected and noninfected adults early after intensive care unit
admission. Am J Clin Nutr 93(6):1356–1364. https://doi.org/10.3945/ajcn.110.008417
Calfee CS, Gallagher D, Abbott J, Thompson BT, Matthay MA, Network NA (2012) Plasma
angiopoietin-2 in clinical acute lung injury: prognostic and pathogenetic significance. Crit Care
Med 40(6):1731–1737. https://doi.org/10.1097/CCM.0b013e3182451c87
Calfee CS, Delucchi K, Parsons PE et al (2014) Subphenotypes in acute respiratory distress
syndrome: latent class analysis of data from two randomised controlled trials. Lancet Respir
Med 2(8):611–620. https://doi.org/10.1016/S2213-2600(14)70097-9
Calfee CS, Delucchi KL, Sinha P et al (2018) Acute respiratory distress syndrome subphenotypes
and differential response to simvastatin: secondary analysis of a randomised controlled trial.
Lancet Respir Med 6(9):691–698. https://doi.org/10.1016/S2213-2600(18)30177-2
Casaer MP, Van den Berghe G (2014) Nutrition in the acute phase of critical illness. N Engl J Med
370(25):2450–2451. https://doi.org/10.1056/NEJMc1404896
Casaer MP, Mesotten D, Hermans G et al (2011) Early versus late parenteral nutrition in critically ill
adults. N Engl J Med 365(6):506–517. https://doi.org/10.1056/NEJMoa1102662
Cooper MS (2003) Corticosteroid Insufficiency in acutely Ill patients. N Engl J Med:8
Dolinay T, Kim YS, Howrylak J et al (2012) Inflammasome-regulated cytokines are critical
mediators of acute lung injury. Am J Respir Crit Care Med 185(11):1225–1234. https://doi.
org/10.1164/rccm.201201-0003OC
Eckerle M, Ambroggio L, Puskarich MA et al (2017) Metabolomics as a driver in advancing
precision medicine in sepsis. Pharmacotherapy 37(9):1023–1032. https://doi.org/10.1002/phar.
1974
Englert JA, Rogers AJ (2016) Metabolism, metabolomics, and nutritional support of patients with
sepsis. Clin Chest Med 37(2):321–331. https://doi.org/10.1016/j.ccm.2016.01.011
Famous KR, Delucchi K, Ware LB et al (2017) Acute respiratory distress syndrome subphenotypes
respond differently to randomized fluid management strategy. Am J Respir Crit Care Med
195(3):331–338. https://doi.org/10.1164/rccm.201603-0645OC
Ferguson ND, Fan E, Camporota L et al (2012) The Berlin definition of ARDS: an expanded
rationale, justification, and supplementary material. Intensive Care Med 38(10):1573–1582.
https://doi.org/10.1007/s00134-012-2682-1
Fleischmann C, Scherag A, Adhikari NKJ et al (2016) Assessment of global incidence and
mortality of hospital-treated sepsis. Current estimates and limitations. Am J Respir Crit Care
Med 193(3):259–272. https://doi.org/10.1164/rccm.201504-0781OC
382 A. E. Pacheco-Navarro and A. J. Rogers

Gaieski DF, Edwards JM, Kallan MJ, Carr BG (2013) Benchmarking the incidence and mortality of
severe sepsis in the United States. Crit Care Med 41(5):1167–1174. https://doi.org/10.1097/
CCM.0b013e31827c09f8
Hermans G, Casaer MP, Clerckx B et al (2013) Effect of tolerating macronutrient deficit on the
development of intensive-care unit acquired weakness: a subanalysis of the EPaNIC trial.
Lancet Respir Med 1(8):621–629. https://doi.org/10.1016/S2213-2600(13)70183-8
ICU Outcomes. Philip R. Lee Institute for Health Policy Studies Accessed 24 Apr 2022. https://
healthpolicy.ucsf.edu/icu-outcomes
Investigators RC, Gordon AC, Mouncey PR et al (2021) Interleukin-6 receptor antagonists in
critically ill patients with Covid-19. N Engl J Med 384(16):1491–1502. https://doi.org/10.1056/
NEJMoa2100433
Jansen TC, van Bommel J, Schoonderbeek FJ et al (2010) Early lactate-guided therapy in intensive
care unit patients: a multicenter, open-label, randomized controlled trial. Am J Respir Crit Care
Med 182(6):752–761. https://doi.org/10.1164/rccm.200912-1918OC
Jones AE, Shapiro NI, Trzeciak S et al (2010) Lactate clearance vs central venous oxygen saturation
as goals of early sepsis therapy: a randomized clinical trial. JAMA 303(8):739–746. https://doi.
org/10.1001/jama.2010.158
Kalil AC, Patterson TF, Mehta AK et al (2021) Baricitinib plus Remdesivir for hospitalized adults
with Covid-19. N Engl J Med 384(9):795–807. https://doi.org/10.1056/NEJMoa2031994
Khardori R, Castillo D (2012) Endocrine and metabolic changes during sepsis. Med Clin North Am
96(6):1095–1105. https://doi.org/10.1016/j.mcna.2012.09.005
Langley RJ, Tsalik EL, van Velkinburgh JC et al (2013) An integrated clinico-metabolomic model
improves prediction of death in sepsis. Sci Transl Med 5(195):195ra95. https://doi.org/10.1126/
scitranslmed.3005893
Langley RJ, Tipper JL, Bruse S et al (2014) Integrative “omic” analysis of experimental bacteremia
identifies a metabolic signature that distinguishes human sepsis from systemic inflammatory
response syndromes. Am J Respir Crit Care Med 190(4):445–455. https://doi.org/10.1164/
rccm.201404-0624OC
Manzanares W, Dhaliwal R, Jiang X, Murch L, Heyland DK (2012) Antioxidant micronutrients in
the critically ill: a systematic review and meta-analysis. Crit Care Lond Engl 16(2):R66. https://
doi.org/10.1186/cc11316
Marik PE, Pastores SM, Annane D et al (2008) Recommendations for the diagnosis and manage-
ment of corticosteroid insufficiency in critically ill adult patients: consensus statements from an
international task force by the American College of Critical Care Medicine. Crit Care Med
36(6):1937–1949. https://doi.org/10.1097/CCM.0b013e31817603ba
Masiero E, Agatea L, Mammucari C et al (2009) Autophagy is required to maintain muscle mass.
Cell Metab 10(6):507–515. https://doi.org/10.1016/j.cmet.2009.10.008
Matthay MA, Zemans RL, Zimmerman GA et al (2019) Acute respiratory distress syndrome. Nat
Rev Primer 5(1):18. https://doi.org/10.1038/s41572-019-0069-0
Metwaly SM, Winston BW (2020) Systems biology ARDS research with a focus on metabolomics.
Metabolites 10(5). https://doi.org/10.3390/metabo10050207
Metwaly S, Cote A, Donnelly SJ et al (2021) ARDS metabolic fingerprints: characterization,
benchmarking, and potential mechanistic interpretation. Am J Physiol Lung Cell Mol Physiol
321(1):L79–L90. https://doi.org/10.1152/ajplung.00077.2021
Mickiewicz B, Vogel HJ, Wong HR, Winston BW (2013) Metabolomics as a novel approach for
early diagnosis of pediatric septic shock and its mortality. Am J Respir Crit Care Med 187(9):
967–976. https://doi.org/10.1164/rccm.201209-1726OC
Mickiewicz B, Duggan GE, Winston BW et al (2014) Metabolic profiling of serum samples by 1H
nuclear magnetic resonance spectroscopy as a potential diagnostic approach for septic shock.
Crit Care Med 42(5):1140–1149. https://doi.org/10.1097/CCM.0000000000000142
Mikkelsen ME, Miltiades AN, Gaieski DF et al (2009) Serum lactate is associated with mortality in
severe sepsis independent of organ failure and shock. Crit Care Med 37(5):1670–1677. https://
doi.org/10.1097/CCM.0b013e31819fcf68
The Metabolomics of Critical Illness 383

Piel DA, Gruber PJ, Weinheimer CJ et al (2007) Mitochondrial resuscitation with exogenous
cytochrome c in the septic heart. Crit Care Med 35(9):2120–2127. https://doi.org/10.1097/01.
ccm.0000278914.85340.fe
Pool R, Gomez H, Kellum JA (2018) Mechanisms of organ dysfunction in sepsis. Crit Care Clin
34(1):63–80. https://doi.org/10.1016/j.ccc.2017.08.003
Preiser JC, Ichai C, Orban JC, Groeneveld ABJ (2014) Metabolic response to the stress of critical
illness. Br J Anaesth 113(6):945–954. https://doi.org/10.1093/bja/aeu187
Puskarich MA, Finkel MA, Karnovsky A et al (2015) Pharmacometabolomics of l-carnitine
treatment response phenotypes in patients with septic shock. Ann Am Thorac Soc 12(1):
46–56. https://doi.org/10.1513/AnnalsATS.201409-415OC
Puskarich MA, Jennaro TS, Gillies CE et al (2021) Pharmacometabolomics identifies candidate
predictor metabolites of an L-carnitine treatment mortality benefit in septic shock. Clin Transl
Sci. https://doi.org/10.1111/cts.13088
RECOVERY Collaborative Group, Horby P, Lim WS et al (2021) Dexamethasone in hospitalized
patients with Covid-19. N Engl J Med 384(8):693–704. https://doi.org/10.1056/
NEJMoa2021436
Rogers AJ, McGeachie M, Baron RM et al (2014) Metabolomic derangements are associated with
mortality in critically ill adult patients. PLoS One 9(1):e87538. https://doi.org/10.1371/journal.
pone.0087538
Rogers AJ, Contrepois K, Wu M et al (2017) Profiling of ARDS pulmonary edema fluid identifies a
metabolically distinct subset. Am J Physiol Lung Cell Mol Physiol 312(5):L703–L709. https://
doi.org/10.1152/ajplung.00438.2016
Rogers AJ, Guan J, Trtchounian A et al (2019) Association of elevated plasma interleukin-18 level
with increased mortality in a clinical trial of statin treatment for acute respiratory distress
syndrome. Crit Care Med 47(8):1089–1096. https://doi.org/10.1097/CCM.0000000000003816
Rogers AJ, Leligdowicz A, Contrepois K et al (2021) Plasma metabolites in early sepsis identify
distinct clusters defined by plasma lipids. Crit Care Explor 3(8):e0478. https://doi.org/10.1097/
CCE.0000000000000478
Rubenfeld GD, Caldwell E, Peabody E et al (2005) Incidence and outcomes of acute lung injury. N
Engl J Med 353(16):1685–1693. https://doi.org/10.1056/NEJMoa050333
Rudd KE, Johnson SC, Agesa KM et al (2020) Global, regional, and national sepsis incidence and
mortality, 1990-2017: analysis for the global burden of disease study. Lancet Lond Engl
395(10219):200–211. https://doi.org/10.1016/S0140-6736(19)32989-7
Rydzak T, Groves RA, Zhang R et al (2022) Metabolic preference assay for rapid diagnosis of
bloodstream infections. Nat Commun 13:2332. https://doi.org/10.1038/s41467-022-30048-6
Sair M, Etherington PJ, Peter Winlove C, Evans TW (2001) Tissue oxygenation and perfusion in
patients with systemic sepsis. Crit Care Med 29(7):1343–1349. https://doi.org/10.1097/
00003246-200107000-00008
Schmidt C, Höcherl K, Schweda F, Bucher M (2007) Proinflammatory cytokines cause down-
regulation of renal chloride entry pathways during sepsis. Crit Care Med 35(9):2110–2119.
https://doi.org/10.1097/01.ccm.0000281447.22966.8b
Shapiro NI, Howell MD, Talmor D et al (2005) Serum lactate as a predictor of mortality in
emergency department patients with infection. Ann Emerg Med 45(5):524–528. https://doi.
org/10.1016/j.annemergmed.2004.12.006
Singer M, De Santis V, Vitale D, Jeffcoate W (2004) Multiorgan failure is an adaptive, endocrine-
mediated, metabolic response to overwhelming systemic inflammation. Lancet 364(9433):
545–548. https://doi.org/10.1016/S0140-6736(04)16815-3
Singer M, Deutschman CS, Seymour CW et al (2016) The third international consensus definitions
for sepsis and septic shock (Sepsis-3). JAMA 315(8):801–810. https://doi.org/10.1001/jama.
2016.0287
Sinha P, Delucchi KL, McAuley DF, O’Kane CM, Matthay MA, Calfee CS (2020) Development
and validation of parsimonious algorithms to classify acute respiratory distress syndrome
384 A. E. Pacheco-Navarro and A. J. Rogers

phenotypes: a secondary analysis of randomised controlled trials. Lancet Respir Med 8(3):
247–257. https://doi.org/10.1016/S2213-2600(19)30369-8
Sprung CL, Annane D, Keh D et al (2008) Hydrocortisone therapy for patients with septic shock. N
Engl J Med 358(2):111–124. https://doi.org/10.1056/NEJMoa071366
Thompson BT, Chambers RC, Liu KD (2017) Acute respiratory distress syndrome. N Engl J Med
377(6):562–572. https://doi.org/10.1056/NEJMra1608077
Tomazini BM, Maia IS, Cavalcanti AB et al (2020) Effect of dexamethasone on days alive and
ventilator-free in patients with moderate or severe acute respiratory distress syndrome and
COVID-19: the CoDEX randomized clinical trial. JAMA 324(13):1307–1316. https://doi.org/
10.1001/jama.2020.17021
Vanhorebeek I, Gunst J, Derde S et al (2011) Insufficient activation of autophagy allows cellular
damage to accumulate in critically ill patients. J Clin Endocrinol Metab 96(4):E633–E645.
https://doi.org/10.1210/jc.2010-2563
Venkatesh B, Finfer S, Cohen J et al (2018) Adjunctive glucocorticoid therapy in patients with
septic shock. N Engl J Med 378:797. https://doi.org/10.1056/NEJMoa1705835
Villar J, Ferrando C, Martínez D et al (2020) Dexamethasone treatment for the acute respiratory
distress syndrome: a multicentre, randomised controlled trial. Lancet Respir Med 8(3):267–276.
https://doi.org/10.1016/S2213-2600(19)30417-5
Viswan A, Ghosh P, Gupta D, Azim A, Sinha N (2019) Distinct metabolic endotype mirroring acute
respiratory distress syndrome (ARDS) subphenotype and its heterogeneous biology. Sci Rep
9(1):2108. https://doi.org/10.1038/s41598-019-39017-4
Wasyluk W, Wasyluk M, Zwolak A (2021) Sepsis as a pan-endocrine illness-endocrine disorders in
septic patients. J Clin Med 10(10):2075. https://doi.org/10.3390/jcm10102075
Zinter MS, Spicer A, Orwoll BO et al (2016) Plasma angiopoietin-2 outperforms other markers of
endothelial injury in prognosticating pediatric ARDS mortality. Am J Physiol Lung Cell Mol
Physiol 310(3):L224–L231. https://doi.org/10.1152/ajplung.00336.2015