Methods in

Molecular Biology 2891

Olga Deda · Helen G. Gika

Ian D. Wilson Editors

Metabolic
Profiling
Methods and Protocols
Second Edition
METHODS IN MOLECULAR BIOLOGY

Series Editor
John M. Walker
School of Life and Medical Sciences
University of Hertfordshire
Hatfield, Hertfordshire, UK

For further volumes:

http://www.springer.com/series/7651
For over 35 years, biological scientists have come to rely on the research protocols and
methodologies in the critically acclaimed Methods in Molecular Biology series. The series was
the first to introduce the step-by-step protocols approach that has become the standard in all
biomedical protocol publishing. Each protocol is provided in readily-reproducible step-by-
step fashion, opening with an introductory overview, a list of the materials and reagents
needed to complete the experiment, and followed by a detailed procedure that is supported
with a helpful notes section offering tips and tricks of the trade as well as troubleshooting
advice. These hallmark features were introduced by series editor Dr. John Walker and
constitute the key ingredient in each and every volume of the Methods in Molecular Biology
series. Tested and trusted, comprehensive and reliable, all protocols from the series are
indexed in PubMed.
 Metabolic Profiling

Methods and Protocols

Second Edition

Edited by

Olga Deda and Helen G. Gika

Laboratory of Forensic Medicine & Toxicology, Aristotle University of Thessaloniki, Thessaloniki, Greece

Ian D. Wilson
Division of Systems Medicine, Imperial College London, London, UK
Editors
Olga Deda Helen G. Gika
Laboratory of Forensic Medicine Laboratory of Forensic Medicine
& Toxicology & Toxicology
Aristotle University of Thessaloniki Aristotle University of Thessaloniki
Thessaloniki, Greece Thessaloniki, Greece

Ian D. Wilson
Division of Systems Medicine
Imperial College London
London, UK

ISSN 1064-3745 ISSN 1940-6029 (electronic)

Methods in Molecular Biology
ISBN 978-1-0716-4333-4 ISBN 978-1-0716-4334-1 (eBook)
https://doi.org/10.1007/978-1-0716-4334-1
© The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Science+Business Media, LLC, part
of Springer Nature 2025

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 Humana imprint is published by the registered company Springer Science+Business Media, LLC, part of Springer
Nature.
The registered company address is: 1 New York Plaza, New York, NY 10004, U.S.A.

If disposing of this product, please recycle the paper.

Preface

This volume is the second edition of Metabolic Profiling: Methods and Protocols, which was
originally published in 2018 as Vol. 1738 in the Methods in Molecular Biology series. The
current edition provides a mixture new and updated methods and protocols from the first
edition. These new contributions are designed to reflect some of the changes in the
development and practice of metabolic phenotyping.
With the chapters in this second edition, we have tried to maintain coverage of impor-
tant topics from the previous edition (e.g., quality control in untargeted metabolic pheno-
typing) while expanding the topics discussed to include other important aspects of this type
of research such as the implementation of quality assurances processes, as discussed by
Kirwan et al (without which any data generated is of doubtful value). Of course, once
good quality data has been acquired, effort has to be put into converting it into useful
information and an important prerequisite for this is careful bio- and chemoinformatic data
analysis as detailed in two contributions by Witting and Rainer and Garcia-Aloy et al.,
respectively.
For those interested in practical aspects of the art of metabolic phenotyping, the volume
contains a number of updated protocols that are focused on untargeted metabolic pheno-
typing (using reversed-phase and ion-pair LC-MS) that can be used for holistic, hypothesis
free, sample analyses. The rapidly maturing field of lipidomics is covered in two contribu-
tions, one on quantitative lipid analysis using supercritical fluid chromatography (SFC-MS)
by Takeda et al and a second, employing UHPLC-TIMS-PASEF-MS, by Denti et al. In
addition there are protocols for metabolite profiling of feces (Deda et al.), the extraction of
plant material for comprehensive metabolome analysis by NMR spectroscopy (Schripsema
and Dagnino), and an LC-ion mobility-MS-based protocol for acquiring metabolome data
on wine and grapes by Sáez et al.
In addition to untargeted methodologies, often used for “biomarker discovery” in the
last few years, we have seen an increasing trend toward the development and application of
targeted metabolite profiling methods. In this volume we have therefore sought to include
protocols that provide access to such methods. The targeted analysis methods described
cover the determination of metanephrines in urine via LC-MS/MS (Thaitumu and Frank), a
“multi-targeted” HILIC- MS/MS method (Virgiliou et al.) and a procedure employing
solid-state high-resolution “Magic Angle Spinning” (HR-MAS) NMR spectroscopy applied
to profiling olive fruit phenolics (Manolopoulou and Spyros).
Metabolic phenotyping (metabonomics/metabolomics) is now well established in all
areas of biological, environmental, and clinical research. In recent years, this has resulted in
an exponential increase in the number of publications, employing both untargeted and
targeted analytical methods, now appearing in journals every year. This enthusiasm for
metabolomics/metabonomics is gratifying to many of those who have watched the field
grow. The rapid expansion in its user base, however, has been much faster than the increase
in the number of trained practitioners of the art who have the necessary expertise in this
metabolomics. Sadly, because of this, many of these newly published articles have suffered

v
vi Preface

from problems such as poor study design, inadequate or unsuitable analytical strategies and
techniques, and subsequent faulty data analysis. Valuable studies are poorly executed,
analyzed, and reported and that is a great shame. The Volume Editors hope that the
contents of the first and second editions of Metabolic Profiling: Methods and Protocols will
provide researchers with pointers toward good practice in metabolic phenotyping.

Thessaloniki, Greece Olga Deda

Thessaloniki, Greece Helen G. Gika
London, UK Ian D. Wilson
Contents

Preface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . v
Contributors. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ix

1 Metabolic Profiling: A Perspective on the Current Status,

Challenges, and Future Directions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
Helen G. Gika, Georgios Theodoridis, and Ian D. Wilson
2 Quality Assurance in Metabolomics and Metabolic Profiling . . . . . . . . . . . . . . . . 15
Jennifer A. Kirwan, Ulrike Bruning, and Jonathan D. Mosley
3 Quality Control and Validation Issues in LC-MS-Based Metabolomics . . . . . . . 53
Olga Begou, Helen G. Gika, Georgios Theodoridis, and Ian D. Wilson
4 Bio- and Chemoinformatic Approaches for Metabolomics Data Analysis . . . . . . 67
Michael Witting and Johannes Rainer
5 Data Treatment for LC-MS Untargeted Analysis. . . . . . . . . . . . . . . . . . . . . . . . . . . 91
Mar Garcia-Aloy, Johannes Rainer, and Pietro Franceschi
6 Untargeted Metabolic Phenotyping by LC-MS . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109
Ian D. Wilson and Elizabeth Want
7 Quantitative Lipidomics of Biological Samples Using Supercritical
Fluid Chromatography Mass Spectrometry. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 131
Hiroaki Takeda, Yoshihiro Izumi, and Takeshi Bamba
8 Rat Fecal Metabolomics-Based Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 153
Olga Deda, Helen G. Gika, and Georgios Theodoridis
9 Ion Pair Chromatography for Endogenous Metabolite LC-MS
Analysis in Tissue Samples Following HGH Resolution Untargeted
Acquisition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 165
Filippos Michopoulos
10 HILIC-MS/MS Multi-targeted Method for Metabolomics Applications. . . . . . 181
Christina Virgiliou, Helen G. Gika, and Georgios Theodoridis
11 A Protocol for GC-MS Profiling of Chiral Secondary Amino Acids . . . . . . . . . . 205
Stanislav Opekar, Helena Zahradnı́čková, Petr Vodrážka,
Lucie Řimnáčová, Martin Moos, and Petr Šimek
12 UHPLC-TIMS-PASEF®-MS for Lipidomics: From Theory to Practice . . . . . . . 221
Vanna Denti, Simone Serrao, Eleonora Bossi, and Giuseppe Paglia
13 Advanced LC-IMS-MS Protocol for Holistic Metabolite Analysis
in Wine and Grape Samples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 239
Vania Sáez, Sara Ferrero-del-Teso, Fulvio Mattivi,
Urska Vrhovsek, and Panagiotis Arapitsas
14 Analysis of Urinary Metanephrines Using Liquid Chromatography
Tandem Mass Spectrometry . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 257
Marlene N. Thaitumu and Elizabeth L. Frank

vii
viii Contents

15 Two-Phase Extraction for Comprehensive Analysis of the

Plant Metabolome by NMR. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 269
Jan Schripsema and Denise Dagnino
16 Olive Fruit Phenolic Profiling Using High Resolution-Magic
Angle Spinning (HR-MAS) Solid-State NMR Spectroscopy . . . . . . . . . . . . . . . . . 277
Efstathia Manolopoulou and Apostolos Spyros

Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 285
Contributors

PANAGIOTIS ARAPITSAS • Department of Food Quality and Nutrition, Research and

Innovation Center, Fondazione Edmund Mach, San Michele all’Adige, Italy; Department
of Wine, Vine, and Beverage Sciences, School of Food Science, University of West Attica,
Athens, Greece
TAKESHI BAMBA • Division of Metabolomics, Medical Research Center for High Depth Omics,
Medical Institute of Bioregulation, Kyushu University, Fukuoka, Japan
OLGA BEGOU • Department of Chemistry, Aristotle University, Thessaloniki, Greece; Biomic
Auth, Bioanalysis and Omics Laboratory, Center for Interdisciplinary Research and
Innovation, Aristotle University, Thessaloniki, Greece; ThetaBiomarkers, Center for
Interdisciplinary Research, and Innovation (CIRI-AUTH), Aristotle University, Balkan
Center, Thessaloniki, GR, Greece
ELEONORA BOSSI • Department of Medicine and Surgery, Proteomics and Metabolomics
Unit, University of Milano-Bicocca, Vedano al Lambro, Italy
ULRIKE BRUNING • Metabolomics, Berlin Institute of Health at Charité Universitatsmedizin,
Berlin, Germany
DENISE DAGNINO • Grupo Metabolômica, Universidade Estadual do Norte Fluminense,
Campos dos Goytacazes, RJ, Brazil
OLGA DEDA • School of Medicine, Aristotle University Thessaloniki, Thessaloniki, Greece;
Biomic Auth, Bioanalysis and Omics Laboratory, Centre for Interdisciplinary Research of
Aristotle, University of Thessaloniki, Innovation Area of Thessaloniki, Thermi, Greece
VANNA DENTI • Department of Medicine and Surgery, Proteomics and Metabolomics Unit,
University of Milano-Bicocca, Vedano al Lambro, Italy
SARA FERRERO-DEL-TESO • Instituto de Ciencias de la Vid y del Vino (ICVV) (Universidad
de La Rioja-CSIC-Gobierno de La Rioja), Logroño, La Rioja, Spain
PIETRO FRANCESCHI • Research and Innovation Centre, Fondazione E. Mach, Trento, Italy
ELIZABETH L. FRANK • Analytic Biochemistry, Calculi and Manual Chemistry, Mass
Spectrometry, ARUP Laboratories, Inc., Salt Lake City, UT, USA
MAR GARCIA-ALOY • Research and Innovation Centre, Fondazione E. Mach, Trento, Italy
HELEN G. GIKA • Department of Medicine, Aristotle University, Thessaloniki, Greece; Biomic
Auth, Bioanalysis and Omics Laboratory, Center for Interdisciplinary Research and
Innovation, Aristotle University, Thessaloniki, Greece
YOSHIHIRO IZUMI • Division of Metabolomics, Medical Research Center for High Depth
Omics, Medical Institute of Bioregulation, Kyushu University, Fukuoka, Japan
JENNIFER A. KIRWAN • Metabolomics, Berlin Institute of Health at Charité
Universitatsmedizin, Berlin, Germany
EFSTATHIA MANOLOPOULOU • NMR Laboratory, Chemistry Department, University of Crete,
Heraklion, Crete, Greece
FULVIO MATTIVI • Department of Food Quality and Nutrition, Research and Innovation
Center, Fondazione Edmund Mach, San Michele all’Adige, Italy

ix
x Contributors

FILIPPOS MICHOPOULOS • Bioscience, Research and Early Development, Oncology,

AstraZeneca, Cambridge, Cambridgeshire, UK
MARTIN MOOS • Laboratory of Analytical Biochemistry & Metabolomics, Biology Centre,
Czech Academy of Sciences, České Budějovice, Czech Republic
JONATHAN D. MOSLEY • Center for Environmental Measurement and Modeling,
Environmental Protection Agency, Athens, GA, USA
STANISLAV OPEKAR • Laboratory of Analytical Biochemistry & Metabolomics, Biology Centre,
Czech Academy of Sciences, České Budějovice, Czech Republic
GIUSEPPE PAGLIA • Department of Medicine and Surgery, Proteomics and Metabolomics
Unit, University of Milano-Bicocca, Vedano al Lambro, Italy
JOHANNES RAINER • Institute for Biomedicine, Eurac Research, Bolzano, Italy
LUCIE ŘIMNÁČOVÁ • Laboratory of Analytical Biochemistry & Metabolomics, Biology Centre,
Czech Academy of Sciences, České Budějovice, Czech Republic
VANIA SÁEZ • Department of Food Quality and Nutrition, Research and Innovation Center,
Fondazione Edmund Mach, San Michele all’Adige, Italy
JAN SCHRIPSEMA • Grupo Metabolômica, Universidade Estadual do Norte Fluminense,
Campos dos Goytacazes, RJ, Brazil
SIMONE SERRAO • Department of Medicine and Surgery, Proteomics and Metabolomics Unit,
University of Milano-Bicocca, Vedano al Lambro, Italy
PETR ŠIMEK • Laboratory of Analytical Biochemistry & Metabolomics, Biology Centre, Czech
Academy of Sciences, České Budějovice, Czech Republic
APOSTOLOS SPYROS • NMR Laboratory, Chemistry Department, University of Crete,
Heraklion, Crete, Greece
HIROAKI TAKEDA • Department of Biotechnology and Life Science, Tokyo University of
Agriculture and Technology, Tokyo, Japan; RIKEN Center for Brain Science, Saitama,
Japan
MARLENE N. THAITUMU • Department of Pathology, University of Utah School of Medicine,
Salt Lake City, UT, USA; ARUP Laboratories, Inc., Salt Lake City, UT, USA
GEORGIOS THEODORIDIS • Biomic Auth, Bioanalysis and Omics Laboratory, Centre for
Interdisciplinary Research of Aristotle, University of Thessaloniki, Innovation Area of
Thessaloniki, Thermi, Greece; Department of Chemistry, Aristotle University Thessaloniki,
Thessaloniki, Greece; ThetaBiomarkers, Balkan Center B1.4, Center for Interdisciplinary
Research, and Innovation (CIRI-AUTH) Aristotle University, Balkan Center,
Thessaloniki, GR, Greece
CHRISTINA VIRGILIOU • Department of Chemical Engineering, Aristotle University,
Thessaloniki, Greece; Biomic Auth, Bioanalysis and Omics Laboratory, Center for
Interdisciplinary Research and Innovation, Aristotle University, Thessaloniki, Greece
PETR VODRÁŽKA • Laboratory of Analytical Biochemistry & Metabolomics, Biology Centre,
Czech Academy of Sciences, České Budějovice, Czech Republic
URSKA VRHOVSEK • Department of Food Quality and Nutrition, Research and Innovation
Center, Fondazione Edmund Mach, San Michele all’Adige, Italy
ELIZABETH WANT • Division of Systems Medicine, Department of Metabolism, Digestion and
Reproduction, Imperial College, London, UK
IAN D. WILSON • Division of Systems Medicine, Department of Metabolism, Digestion and
Reproduction, Imperial College, London, UK; Division of Systems Medicine, Department of
 Contributors xi

Metabolism Department of Metabolism, Digestion and Reproduction, Imperial College,

London, UK
MICHAEL WITTING • Metabolomics and Proteomics Core, Helmholtz Munich, Munich,
Germany; Chair of Analytical Food Chemistry, TUM School of Life Sciences, Technical
University of Munich, Munich, Germany
HELENA ZAHRADNÍČKOVÁ • Laboratory of Analytical Biochemistry & Metabolomics, Biology
Centre, Czech Academy of Sciences, České Budějovice, Czech Republic
 Chapter 1

Metabolic Profiling: A Perspective on the Current Status,

Challenges, and Future Directions
Helen G. Gika, Georgios Theodoridis, and Ian D. Wilson

Abstract
Metabolic profiling continues to develop, and research is now conducted on this topic globally in hundreds
of laboratories, from small groups up to national centers and core facilities. Here we briefly provide a
perspective on the current status and challenges facing metabolic phenotyping (metabonomics/metabo-
lomics) and consider future directions for this important area of biomarker and systems biology research.

Key words Metabolic profiling, Metabolic phenotyping, Metabolomics, Metabonomics, Biomarker

discovery, Metabolic markers

1 Introduction

Untargeted metabolic phenotyping (metabonomics/metabolo-

mics [1, 2] or metabotyping [3]) involves the characterization of
the metabolites present in samples such as biological fluids (plasma,
serum, urine), cells, organs, and whole organisms using both untar-
geted and targeted analytical methodologies. Some of the earliest
examples of the use of untargeted metabolite profiling to determine
disease-based metabolic phenotyping can be traced to the late
1940s and the use, by Dent and Dalgliesh [4, 5], of
two-dimensional paper chromatography for urinalysis in studies
on inborn errors of metabolism. Later, gas chromatography
(GC) was employed by Pauling and colleagues in the early 1970s
to characterize urinary volatiles for the diagnosis of disease [6–8],
and elsewhere liquid chromatography (LC)-based investigations on
metabolic phenotyping were also being undertaken (e.g., [9, 10]).
Thus, a case can be made, if one were minded to do so, that
recognizably untargeted metabolic phenotyping studies, of the
sort now classified as metabolomics, were in use well before geno-
mic (and proteomic) profiling became both practicable and fash-
ionable. This belies the idea that metabotyping is only the most

Olga Deda et al. (eds.), Metabolic Profiling: Methods and Protocols, Methods in Molecular Biology, vol. 2891,
https://doi.org/10.1007/978-1-0716-4334-1_1,
© The Author(s), under exclusive license to Springer Science+Business Media, LLC, part of Springer Nature 2025

1
2 Helen G. Gika et al.

recent latest addition to the “omics” family. However, despite

perhaps being the first omic, it was only when advanced analytical
systems such as 1H nuclear magnetic resonance (NMR) spectros-
copy and mass spectrometry (MS) (particularly hyphenated with
GC or LC), providing both wide metabolite coverage and struc-
tural information, combined with multivariate statistical analysis
(MVA), became available that rapid advances took place. These
techniques have enabled the rapid development of metabolic phe-
notyping to a high level of sophistication and facilitated the “holis-
tic” analysis of samples in the hunt for biomarkers and has resulted
in a large rise in the number of researchers entering the field. This
increased use of metabolic phenotyping has resulted in an ever-
increasing number of publications on the topic with many
hundreds per year being produced and rivalling the numbers seen
for genomics/ transcriptomics and proteomics (see Fig. 1). This
expansion has increased the desire and the need for careful and
appropriate study design and encouraged moves toward the stan-
dardization of methods wherever possible. In addition, a more
nuanced understanding of the relative advantages and limitations
of the “holistic” approach compared to more “targeted” (and semi-
targeted) approaches to metabolic phenotyping has developed.
Simply stated, untargeted methods ideally carry with them no
assumptions or preconceptions about the likely nature of any bio-
markers that may, or may not, be present. They are thus more likely
to highlight novel, and unexpected, metabolites and are thus
hypothesis-free, but potentially hypothesis generating. However,
while the intention might be to undertake “unbiased” analysis, the

Fig. 1 Publication trends in the areas of genomics/transcriptomics, proteomics, and metabolic phenotyping
from 2003 to 2022 generated from references contained within SCOPUS using the search terms genomics,
proteomics, metabolomics, or metabonomics or metabolic profiling and transcriptomics in the title, abstract,
or keywords. (Search made June 2023)
 Metabolic Profiling: A Perspective on the Current Status, Challenges. . . 3

results will inevitably be “skewed” by the capabilities of the tech-

nologies used to investigate the samples. Targeted assays on the
other hand are limited only to the metabolites for which they have
been developed, all of which will be known compounds, and carry
the assumption that some of them are expected to be biomarkers
(which they may well turn out to be). A further, and by no means
trivial, question is how comprehensively do untargeted methods
cover the metabolome? And currently this is a difficult question to
give any real answer to as unlike, e.g., the human genome, the
metabolome is not yet fully mapped. Also, the concentrations
of these metabolites can range over several orders of magnitude,
from zeptomoles to micromoles, making demands on detection
and quantification somewhat challenging. So, while we believe we
know the key biochemical pathways, the very significant, and
continuing, advances in analytical technologies have only served
to highlight the limitations of our knowledge. A problem that
becomes more evident every day is our inability to provide unequiv-
ocal identifications for most of the compounds detected by LC-MS
in metabolic phenotyping studies.

2 Methodology

In attempting to maximize the recovery of metabolite information,

it has become clear that obtaining the most comprehensive meta-
bolic profiles currently requires the use of multiple methods of
analysis. Broadly speaking, at the present time the bulk of metabo-
nomic phenotyping studies use either 1H NMR spectroscopy or
MS-based methods for sample analysis. In the case of 1H NMR
spectroscopy, 500–600 MHz instruments are generally the “work-
horses” of this methodology (although much higher field strengths
have been employed), offering a good compromise for features
such as cost, field strength, dynamic range, and sensitivity. 1H
NMR spectroscopy provides an excellent untargeted methodology
for the analysis of biofluids, culture media, cells, food and tissue
extracts, etc. and is well suited to profiling samples such as urine and
plasma. In addition, the high information content of the spectra,
and good databases, often enable the characterization and identifi-
cation of unknowns. Another benefit of 1H NMR spectroscopy is
that it is equally sensitive for all protons irrespective of compound
structure, and although it may be employed as an untargeted
method, this property enables the quantification of metabolites
even in the absence of an authentic standard [11]. This latter
property is in stark contrast to MS-based methods where metabo-
lite detection depends on structure-based ionization properties,
which can only confidently be determined from an authentic stan-
dard. 1H NMR spectroscopic analysis of liquid samples normally
requires only minimal sample preparation (e.g., filtering and pH
4 Helen G. Gika et al.

adjustment). Samples such as cells and tissues are most often ana-
lyzed after extraction into a suitable solvent, drying down, and
redissolution in a 1H NMR compatible solvent. Solid
(or semisolid) samples can be analyzed without extraction using
“magic angle spinning” NMR spectroscopy (although uptake of
this technology in metabolic profiling applications has been lim-
ited) [11]. A noteworthy property of 1H-NMR-spectroscopy-
based methods is that they are generally easily transferable between
laboratories such that, e.g., a one-dimensional 1H NMR spectrum
acquired at 600 MHz in one laboratory should be identical with
one obtained for the same sample, analyzed under the same experi-
mental conditions in another laboratory, irrespective of the
manufacturer.
An illustrative example of the use of 1H NMR spectroscopy to
compare the effects of the liver toxin acetaminophen (APAP) and a
less toxic structural analogue, the meta-isomer AMAP is shown in
Fig. 2. This example shows the 1H NMR spectra obtained for
“aqueous” liver extracts obtained from control (untreated)
AMAP and APAP dosed mice [12]. As these spectra show, it is
not difficult to see the differences in composition of the extracts
resulting from the different dosing regimens, with both drug-
related material and endogenous metabolites clearly present.
MS, like 1H NMR spectroscopy, can also be employed for the
direct analysis of liquid (or gaseous) samples via directly infusing
(DI) suitably prepared samples into the ion source of the mass
spectrometer (DIMS) [13]. DIMS-based analysis, using either
flow injection analysis (FIA), interfaces such as the “nanomate,”
desorption ESI-MS (DESI), or Rapid Evaporative Ionization Mass
Spectrometry (REIMS), can have advantages in terms of simplicity
and speed. However, matrices such as, e.g., tissues, cell extracts,
food, urine, or blood-derived samples can defy simple methods of
characterization because of their sheer complexity and confounding
effects on analyte ionization performance such as ion-suppression/
enhancement. Furthermore, isobaric or isomeric compounds are
difficult to determine confidently if they are present as mixtures
using methods that do not employ a separation step. This has
resulted in the increasing adoption of the use of hyphenated tech-
niques incorporating a chromatographic, electrophoretic, or ion
mobility (IM) separation prior to MS. Thus, LC, GC, IM, and
capillary zone electrophoresis (CZE-MS, or CE-MS) are very
widely used for metabolic phenotyping, with LC-MS (or LC-IM-
MS)-based techniques currently the most popular.
GC-MS-based methods are a priority technology selection for
the analysis of volatiles in samples such as exhaled breath [14] but
can be also used for involatile metabolites following derivatization
[15]. For the GC analysis of involatile analytes, the derivatization
step needs careful optimization as various metabolites can react
with the derivatization reagents at different rates thus resulting in
 Metabolic Profiling: A Perspective on the Current Status, Challenges. . . 5

Fig. 2 Representative 1H-NMR spectra of hepatic extract metabolic profiles of the acetaminophen (APAP),
AMAP, and control groups at 1 h. Resonances assigned to drug-related molecules have been colored in
red. Key: APAP/AMAP-G = APAP/AMAP glucuronide; APAP-SG, APAP glutathionyl; APAP-NAC, APAP-N-
acetylcysteinyl; APAP/AMAPNHCOCH3, APAP/AMAP N-acetyl resonance; GSH, reduced glutathione; GSSG,
oxidized glutathione; Phe, phenylalanine; d-3-HB, d-3-hydroxybutyrate; AMP, adenosine monophosphate,
overlapped resonances from glucose/glycogen/maltose labeled. From reference [12] reprinted with
permission
6 Helen G. Gika et al.

different reaction kinetics even for molecules of the same class

[16]. Suitable reagents and routes to the preparation of GC-MS
compatible derivatives are available for most of the functional
groups on the many analyte classes such as, e.g., amino acids,
sugars, acids, etc., likely to be encountered in biological samples.
These methods can then be adapted to derivatize metabolites pres-
ent in a broad range of sample types (or indeed in their extracts)
including blood plasma/serum, urine, tissue and plant extracts,
food, etc. The requirement for, often, extensive sample preparation
and derivatization prior to the analysis of the wide range of invola-
tile metabolites present in biological matrices can make their analy-
sis by GC-MS analysis both time consuming and labor intensive.
However, this disadvantage is more than compensated for by the
fact that GC-MS, with either electron impact (EI) or atmospheric
pressure chemical ionization (APCI), represents a reliable and well-
researched technique for metabolite analysis. Thus, the availability
of both robust GC-MS instrumentation and high-quality MS data-
bases (e.g., NIST, Agilent Fiehn, etc.) can enable the confident
identification of potential biomarkers [14–16].
Current practice for the LC-MS-based metabolic phenotyping
of biological samples generally involves the use of ultra (high)
performance LC separations (UPLC, UHPLC), which have gradu-
ally displaced conventional HPLC methods. U(H)PLC is based on
the use of sub 2 μm stationary phases [17–19] which, although
requiring higher operating pressures than HPLC, offers highly
efficient, high resolution, chromatographic separations. U
(H)PLC analyses are typically performed over timescales in the
region of 5–15 min, on columns of 1–2.5 mm internal diameter
and 5–15 cm in length and flow rates from 200 to 900 μL/min.
Even faster analysis is possible, and if the loss of some metabolome
coverage is considered acceptable in pursuit of high throughput,
metabolic phenotyping methods of less than 5 min can be used
[20]. In Fig. 3, an example of the results obtained using a gradient
RP-UPLC-MS method is illustrated [19], and while this 3D mass
chromatogram indicates that many metabolites are eluted early on,
further optimization of the gradient could be used to maximize the
use of the “chromatographic space” available in the separation. A
newer version of U(H)PLC has also been described in the form of
vacuum jacketed U(H)PLC [21] which provides significantly
enhanced performance via greatly reduced band broadening.
As well as conventional LC, the related technique of supercriti-
cal fluid chromatography (SFC) has also been used for metabolic
phenotyping (in both metabolomic and lipidomic studies). While
the specialist nature of the equipment required has somewhat
limited its application, SFC nevertheless shows considerable poten-
tial being able to deal with both polar metabolites and nonpolar
lipids [22].
 Metabolic Profiling: A Perspective on the Current Status, Challenges. . . 7

Fig. 3 Representation of a 3D mass chromatogram obtained from the reversed-phase analysis of rat urine by
UPLC–TOF-MS. Reprinted from reference [19] with permission

However, prior to any form of chromatographic analysis, sam-

ple preparation plays a key role in ensuring successful, reproducible,
and robust analyses. In the case of simple liquid samples (e.g.,
urine), sample preparation can often be limited to dilution (with
water or an organic solvent as appropriate) followed by the removal
of particulates by centrifugation. With proteinaceous samples such
as, e.g., plasma/serum, the removal of proteins is an essential
precursor to analysis. This is in order to preserve the integrity of
the chromatographic column, which would otherwise suffer con-
tamination and degradation in performance. Protein removal is
readily achieved via precipitation using solvents such as methanol
or acetonitrile and centrifugation. In the case of solid samples, such
as tissues, solvent extraction is generally performed.
Following sample preparation metabolites can be profiled using
generic chromatographic methods optimized for different classes of
metabolites. Thus, reversed-phase (RP) LC separations are
8 Helen G. Gika et al.

considered to be suitable for medium polar to nonpolar analytes

[19, 20]. However, for accessing the polar metabolome (which
contains, e.g., sugars and polar ionic metabolites), hydrophilic
interaction liquid chromatography (HILIC) is a more appropriate
profiling approach, e.g [19, 20]. However, HILIC provides a par-
tial solution for polar/ionic compounds, and for those for which it
is not well suited, then the options of ion pair (IP) LC or derivati-
zation may be required. In the case of IP analysis, reagents such as,
e.g., tributyl ammonium can be added to the mobile phase to retain
acidic metabolites by providing a positively charged counter ion to
“pair” with them [23, 24]. IP methods have been shown to be very
effective, but IPLC contaminates the system with the reagent and
may result in the need for its permanent dedication to that meth-
odology (decontaminating of the LC and MS instruments is not
trivial). A different approach is to use “class-based” derivatization
to modify chromatographic (and often MS) properties such that
retention on RP-columns is possible, and this, while slightly more
time consuming than injecting unmodified analytes, avoids the
need for dedicating an instrument to IPLC.
If an alternative to LC-based methods for analyzing the polar
(ionic) metabolome is required, then capillary electrophoresis
represents an obvious method of separation prior to MS [25]. How-
ever, one thing is clear, and this is that in order to obtain the
most comprehensive metabolite profiles for a set of samples using
untargeted methods, it is likely that several different chro-
matographic systems, with MS analysis using both +ve and -ve
ESI (and also possibly APCI), will be needed.
An exciting development that has been made possible by the
more widespread availability of commercially instrumentation capa-
ble of performing it is the application of ion mobility (both as a
separation technique and a spectrometry (IMS)). IMS can be either
used as a standalone technique with or without MS (IMS/MS) or
inserted, as a second dimension of separation, between GC, LC,
SFC, or CE. This IM separation, which unlike LC takes microse-
conds rather than minutes, has brought interesting, and very
important, new capabilities to metabolic profiling. Firstly, IM can
provide a significant improvement in the MS data obtained by
resolving co-eluting metabolites based on what is effectively
gas-phase electrophoresis. Secondly, by determining the “collision
cross section” (CCS) value of a molecule, an additional molecular
descriptor is obtained which can help with metabolite identification
[26]. Thus, IM-based advances may therefore improve several
aspects of metabolic phenotyping methods and protocols.

2.1 Quality Control in Developing a suitable DIMS or hyphenated MS analysis is by no

MS- and LC-MS-Based means the end of method development for robust metabolic phe-
Analysis notyping, and to ensure that any analytical data generated is useful
requires further work. This is because, unlike, e.g., 1H NMR
 Metabolic Profiling: A Perspective on the Current Status, Challenges. . . 9

spectroscopy-based metabolite profiling, which is inherently

robust, MS and hyphenated-MS (GC, LC, SFC, CE, etc.) methods
suffer from challenges resulting from system degradation during
use. Thus, over the course of an analytical run, both the column
(if used) and ion source of the MS may become degraded by the
build-up of contamination, thus resulting in “run order” effects.
For the column, such contamination can result in everything from
changes in retention time, peak shape, or back pressure right up to
outright column failure. For the MS, contamination can result in
loss of sensitivity, and there can also potentially be changes in mass
accuracy. Monitoring these effects for potentially thousands of
analytes requires the use of appropriate quality control procedures
which allow the monitoring, assessment, and acceptance (or not) of
the analytical data being generated. Such systems can also, if there is
a need, be used to correct analytical drift (e.g., of retention time or
response) to improve the outcome of the analysis. Many methods
could be adopted for ensuring the quality of untargeted metabo-
lomics/metabonomics to ensure the validity of data obtained when
performing this type of analysis. However, one commonly adopted
quality control (QC) strategy involves the use of so-called pooled
QC samples. These pooled QCs are usually prepared by taking
aliquots of the samples under analysis and mixing them to form a
representative bulk sample. This sample can then be used to pre-
condition/equilibrate the LC system, which is achieved by repeat-
edly injecting the sample until stable retention and peaks shapes are
achieved. In many ways, this means that the pooled QC acts as the
“ultimate,” system suitability test, as analysis should not be under-
taken until good results are achieved. Once the system is seen to be
acceptable for use, the analytical run can proceed with the QC
sample which is then injected at regular intervals to monitor analyt-
ical performance [27, 28]. On completion of the batch, the QC
data are examined using, e.g., multivariate statistical procedures
such as PCA (principle components analysis) to obtain a rapid
overview of the analytical batch and reveal any trends/anomalies
in the QC data indicative of time-related (or other) effects that
suggest problems had occurred that need examination.

2.2 Data Processing Assuming that the analytical data passes such preliminary scrutiny,
further measures to optimize, e.g., peak alignment, can be per-
formed and the data can be examined for the presence of potential
biomarkers. This part of the process relies heavily on the correct
choice and the correct function of software tools. Some software
may operate as black boxes, not allowing much information on the
theory or the tools used and/or not allowing much freedom in the
selection of parameters. This kind of software also often provides
different levels of multivariate statistics for data analysis, such as
PCA, and offers options for advanced visualization plots. Open-
source software (including web-based data treatment servers) is
10 Helen G. Gika et al.

also available that can be used to perform these tasks, and its use
continues to increase [29]. All such open-source tools necessitate at
least some basic-level knowledge of the use of software language in
R or MATLAB environment, but they offer unparalleled freedom
in optimizing and tailoring the data treatment and data scrutiny
process. They also offer advanced control and functionalities in
developing visualization of the findings and in generating plots,
tables, and illustrations. Such tools can offer impressive outputs;
however, statistical analysis is not yet fool proof, and the famous
Benjamin Disraeli quote concerning statistics is still timely
[30]. Hence, researchers are advised to always verify the validity
of their findings by the use of different statistical analysis tools.
Researchers who are novices to the field are advised to consult
with fellow researchers who are more experienced in the statistical
analysis of metabolomic data. Indeed the concept and the needs of
metabolic phenotyping are different from those applied in statistics
for genomics. Hence, the statistical analysis tools to be selected
and/or their fine-tuning for effective data treatment in metabolo-
mics may vary to a great extent from the treatment of genomic data.

2.3 Metabolite Having completed the data analysis, the structural identification of
Identification peaks that correlate (or appear to be correlated) with whatever
biological problem is being investigated may then be pursued.
However, unequivocal metabolite identification (to level 1 of the
MSI) is not a trivial exercise and is not often achievable [31].
Techniques like 1H NMR spectroscopy or GC-MS provide good
structural information and benefit from large databases of spectral
data that can be search for potential structures. Indeed, in many
cases, positive assignments of identity to the detected metabolites
can be made. However, with LC-MS, it still remains the case that,
while large databases are available that can provide information to
assist in achieving some level of annotation, these do not represent
identifications, and on examination many of the suggestions that
are returned are obviously incorrect having no possible biological
plausibility for the species/sample type being investigated [32]. A
further problem is that even when the putative annotations appear
plausible, it may not be possible to obtain the authentic standards
necessary to unequivocally confirm identity. An increasing con-
cern is that many research groups are outsourcing metabolomic
profiling to central laboratories/commercial providers that do not
disclose information on their proprietary methods and thus provide
only lists of molecules without raw data and methodological pro-
tocols. In such case, the onus is on the recipient of such data to
ensure its validity so that only metabolite identification that is
correct is finally reported [32].
Where a number of candidates for identity are returned from
such searches, the unknowns should be characterized to the extent
possible using the wide range of MS techniques available via
Metabolic Profiling: A Perspective on the Current Status, Challenges. . . 11

modern MS instruments, including scan analyses such as MS/MS

(MSn) in data-dependent acquisition (DDA) mode and MS/MS
with different collision energies (such as MSE). Via the combina-
tion of different levels of MS information (including accurate mass
and isotope ratio measurements obtained from both precursor and
product ions and the use of in silico predictions of the fragment
ions), it is likely to generate a list of tentatively annotated
unknowns. If IM-generated CCS data have also been obtained, it
may be possible to use it to increase confidence in the annotation
(or alternatively exclude potential structures). The proposed struc-
ture should also have the chromatographic properties that would
be expected in the LC system used (i.e., compounds annotated as
“lipids” that eluted close to the solvent front in RPLC are unlikely,
but eluting in the high organic portion of the gradient may be
possible). The putatively identified metabolite should also be “plau-
sible” [32] so, e.g., the “identification” of a drug or pesticide as a
“biomarker” in a cell culture is not plausible, but a change in the
amount of an amino acid may well be. However, if metabolite
identification still remains uncertain after the investigations sug-
gested above, but is essential as the hypothesis that has been gen-
erated is based on it, it may be that only its isolation followed by
further analysis with, e.g., NMR spectroscopy or total synthesis will
be sufficient to prove identity.
Once potential biomarkers have been identified (or annotated
if unambiguous identification is not possible), a variety of other
considerations must be addressed depending upon what the aims of
the study were. So, if the investigation was aimed at achieving a
mechanistic understanding of a biological system, then how do
these metabolites fit in? Are the metabolites directly involved in
the pathways responsible for the observed changes? or are they the
result of a “global” response as the system adapts to a new bio-
chemical state? On the other hand, if the question was rather can we
reliably identify the likelihood of a patient responding to a particu-
lar drug or treatment regime (pharmacometabonomics), then the
function of the metabolites and, therefore, the need for certainty in
metabolite identification are, at least initially, less important than
the reliability of the prediction of effect.
Whatever the research question at some point, it is likely that an
attempt will be made to obtain a deeper understanding of the
biochemistry being affected using tools such as pathway analysis.
Fortunately, much effort has been devoted to the development
of software that facilitates pathway analysis (both commercial and
internet-based freeware). Many of these software tools use data
from public databases such as the Human Metabolome Database
(HMDB) and/or KEGG (Kyoto Encyclopedia of Genes and Gen-
omes). Their overall aim is the provision of methods to systemati-
cally place the metabolites of interest in some biological context by
highlighting relevant metabolic pathways showing some degree of
12 Helen G. Gika et al.

perturbation in the studied data set [29]. It will be interesting to

see how the increasingly rapid developments in the area of artificial
intelligence contribute to this (and indeed other) areas of
metabolomics.
Once there is confidence that the “biomarkers” do actually help
to provide a direct, mechanistic, link to elucidating the mechanism
responsible for the problem being studied, general, untargeted,
discovery methods are no longer appropriate, and what is really
required are specific (and validated) targeted methods for the bio-
marker’s accurate quantification. These assays can then be used to
reanalyze the samples and confirm their utility before applying the
assay to subsequent sample sets so as to begin the process of
confirming any hypothesis generated based on them.

3 Conclusions

Acquiring good quality information about the metabolome, using

either targeted or untargeted methods, is not an easy task, and best
practices are still evolving. However, performing and disseminating
methodology that can provide reliable, repeatable results that lead
to the discovery of new biomarkers and that provide a better
understanding of biological systems is essential to maintain prog-
ress in many areas of the life sciences, with benefits to, e.g., agricul-
ture, the environment, sport, and medicine. This is because the
metabolome reflects to a direct extent the state of the biological
system under investigation. Another reason for performing such
studies is the fact that changes in the metabolome are the final result
of the gene and protein function and as such are often amplified
relative to, e.g., the single nucleotide polymorphisms which caused
their perturbation. Yet another reason is that (for example) for
certain food products such as olive oil, wine, or honey, there is
hardly anything else to look for: the content of such samples is
massively dominated by small molecule metabolites.
Publication numbers in metabolic phenotyping continue to
increase, year on year, and we see no sign of this decreasing.
However, while many of the technological issues concerning meta-
bolomics/metabonomics are slowly being addressed, the influx of
new researchers into the area, with sometimes little experience of
the practical issues (particularly those surrounding metabolite iden-
tification), represents an ever-present challenge. It is our belief that
the availability of sources of good quality protocols will help with
standardization and encourage best practice in this important and
rapidly growing field of biology.
 Metabolic Profiling: A Perspective on the Current Status, Challenges. . . 13

References
1. Nicholson JK, Lindon JC, Holmes E (1999) 15. Kopka J (2006) Current challenges and devel-
“Metabonomics”: understanding the meta- opments in GC–MS based metabolite profiling
bolic responses of living systems to pathophys- technology. J Biotechnol 124:312–322
iological stimuli via multivariate statistical 16. Moros G, Chatziioannou AC, Gika HG et al
analysis of biological NMR spectroscopic data. (2017) Investigation of the derivatization con-
Xenobiotica 29:1181–1189 ditions for GC–MS metabolomics of biological
2. Fiehn O, Kopka J, Dörmann P et al (2000) samples. Bioanalysis 9:53–65
Metabolite profiling for plant functional geno- 17. Gika HG, Theodoridis GA, Plumb RS et al
mics. Nat Biotechnol 18:1157–1161 (2014) Current practice of liquid
3. Gavaghan CL, Holmes E, Lenz E et al (2000) chromatography–mass spectrometry in meta-
FEBS Lett 484:169–174 bolomics and metabonomics. J Pharm Biomed
4. Dent CE (1952) Lectures on the scientific basis Anal 87:12–25
of medicine, vol 2. Athlone Press, London 18. Rainville PD, Theodoridis G, Plumb RS et al
5. Dalgliesh CE (1956) Two-dimensional paper (2014) Advances in liquid chromatography
chromatography of urinary indoles and related coupled to mass spectrometry for metabolic
substances. Biochem J 64:481–485 phenotyping. TrAC Trends Anal Chem 61:
6. Teranishi R, Mon TR, Robinson AB et al 181–191
(1972) Gas chromatography of volatiles from 19. Theodoridis GA, Gika HG, Want EJ et al
breath and urine. Anal Chem 44:18–20 (2012) Liquid chromatography–mass spec-
7. Pauling L, Robinson AB, Teranishi R et al trometry based global metabolite profiling: a
(1971) Quantitative analysis of urine vapor review. Anal Chim Acta 711:7–16
and breath by gas-liquid partition chromatog- 20. Lioupi A, Marinaki M, Virgiliou C, Begou O,
raphy. Proc Natl Acad Sci USA 68:2374–2376 Gika H, Wilson I, Theodoridis G (2023) Prob-
8. Robinson AB, Pauling L (1974) Techniques of ing the polar metabolome by UHPLC-MS.
orthomolecular diagnosis. Clin Chem 20:961– TrAC Trends Anal Chem. https://doi.org/
965 10.1016/j.trac.2023.117014
9. Scott CD, Chilcote DD, Lee NE (1972) Cou- 21. Plumb RS, Isaac G, Rainville PD, Hill J,
pled anion and cation-exchange chromatogra- Gethings LA, Johnson KA, Lauterbach J, Wil-
phy of complex biochemical mixtures. Anal son ID (2022) High throughput UHPLC-MS-
Chem 44:85–89 based lipidomics using vacuum jacketed col-
umns. J Proteome Res 21:691–701
10. Scott CD, Chilcote DD, Katz S et al (1973)
Advances in the application of high resolution 22. van de Velde B, Guillarme D, Kohler I (2020)
liquid chromatography to the separation of Supercritical fluid chromatography – mass
complex biological mixtures. J Chromatogr spectrometry in metabolomics: past, present,
Sci 11:96–100 and future perspectives. J Chromatogr
B. https://doi.org/10.1016/j.jchromb.2020.
11. Lenz EM, Wilson ID (2007) Analytical strate- 122444
gies in metabonomics. J Proteome Res 6:443–
458 23. Gika HG, Wilson ID, Theodoridis GA (2014)
LC–MS-based holistic metabolic profiling.
12. Kyriakides M, Maitre L, Stamper BD et al Problems, limitations, advantages, and future
(2016) Comparative metabonomic analysis of perspectives. J Chromatogr B 966:1–6
hepatotoxicity induced by acetaminophen and
its less toxic meta-isomer. Arch Toxicol 90: 24. Michopoulos F, Whalley N, Theodoridis G
3073–3085 et al (2014) Targeted profiling of polar intra-
cellular metabolites using ion-pair-high perfor-
13. Han J, Danell RM, Patel JR et al (2008) mance liquid chromatography and-ultra high
Towards high-throughput metabolomics performance liquid chromatography coupled
using ultrahigh-field Fourier transform ion to tandem mass spectrometry: applications to
cyclotron resonance mass spectrometry. Meta- serum, urine and tissue extracts. J Chromatogr
bolomics 4:128–140 A 1349:60–68
14. Theodoridis G, Gika HG, Wilson ID (2011) 25. Mamani-Huanca M, Villaseñor A, Gonzalez-
Mass spectrometry-based holistic analytical Riano C, López-López A, López-
approaches for metabolite profiling in systems Gonzálvez A, Barbas C (2023) Capillary elec-
biology studies. Mass Spectrom Rev 30:884– trophoresis mass spectrometry-based untar-
906 geted metabolomics to approach disease
diagnosis. TrAC Trends Anal Chem:117049
14 Helen G. Gika et al.

26. Zhou Z, Luo M, Chen X et al (2020) Ion 29. Xia J, Wishart DS (2010) MetPA: a web-based
mobility collision cross-section atlas for metabolomics tool for pathway analysis and
known and unknown metabolite annotation visualization. Bioinformatics 26:2342–2344
in untargeted metabolomics. Nat Commun 3 0 . https :// en.w ikiped i a . o r g / w i k i / L i e s , _
11:4334. https://doi.org/10.1038/s41467- damned_lies,_and_statistics
020-18171-8 31. Kodra D, Pousinis P, Vorkas PA et al (2022) Is
27. Gika HG, Theodoridis GA, Wingate JE et al current practice adhering to guidelines pro-
(2007) Within-day reproducibility of an posed for metabolite identification in LC-MS
HPLC- MS-based method for metabonomic untargeted metabolomics? A meta-analysis of
analysis: application to human urine. J Prote- the literature. J Proteome Res 21:590–598
ome Res 6:3291–3303 32. Theodoridis G, Gika H, Raftery D,
28. Gika HG, Zisi C, Theodoridis G, Wilson ID Goodacre R, Plumb RS, Wilson ID (2023)
(2016) Protocol for quality control in meta- Ensuring fact-based metabolite identification
bolic profiling of biological fluids by U in liquid chromatography–mass spectrometry-
(H)PLC-MS. J Chromatogr B 1008:15–25 based metabolomics. Anal Chem 95:3909–
3916
 Chapter 2

Quality Assurance in Metabolomics and Metabolic Profiling

Jennifer A. Kirwan, Ulrike Bruning, and Jonathan D. Mosley

Abstract
Metabolic profiling (untargeted metabolomics) aims for a global unbiased analysis of metabolites in a cell or
biological system. It remains a highly useful research tool used across various analytical platforms. Incre-
mental improvements across multiple steps in the analytical process may have large consequences for the
end quality of the data. Thus, this chapter concentrates on which aspects of quality assurance can be
implemented by a lab in the (pre-)analytical stages of the analysis to improve the overall end quality of their
data. The scope of this chapter is limited to liquid-chromatography-mass spectrometry (LC-MS)-based
profiling, which is one of the most widely utilized platforms, although the general principles are applicable
to all metabolomics experiments.

Key words Quality assurance, Quality control, Quality management, Validation, Robust method,
Metabolomics, Untargeted, Mass spectrometry, NMR, Data quality

1 Introduction

Quality management (QM) is the practice of managing a process

such that it produces robust and reproducible results within a
defined tolerance. QM is used throughout most industries and is
defined by the ISO9001 standard [1, 2], with specific guidance on
testing and calibration provided by ISO/IEC 17025 [3]. It is
divided into quality assurance (QA) and quality control (QC). QA
sets out the processes which should be followed in order to assure
reliable quality, whereas quality control is the undertaking of mea-
surements to show that these standards have been met [1, 4, 5].
Typically, QA is what is done before or during a process, whereas
QC tends to be measured at the end of a process or subprocess
[4]. As a concept, QC is both well understood and well covered in
the laboratory, and its role in metabolomics has been firmly estab-
lished, though consensus continues to elude the metabolomics
community [4, 6, 7]. QA is perceived to be equally important;
indeed, it is the bedrock that underlies the principles of good
QM, and the results of QC typically depend on how well the QA

Olga Deda et al. (eds.), Metabolic Profiling: Methods and Protocols, Methods in Molecular Biology, vol. 2891,
https://doi.org/10.1007/978-1-0716-4334-1_2,
© The Author(s), under exclusive license to Springer Science+Business Media, LLC, part of Springer Nature 2025

15
16 Jennifer A. Kirwan et al.

process has been designed and implemented. Despite this, there is

no comprehensive approach to QA in metabolomics that is cur-
rently recognized by the community. While there are existing docu-
ments that cover the major aspects of QM [2, 3], it is questionable
whether they have been applied to metabolic profiling to any great
extent, especially outside of the commercial world. In addition,
there are guidance documents on analytical testing written by the
FDA and EMA, respectively [8, 9], which are aimed at precise
measurement and quantification of single analytes. However,
these guidelines do not account for either complex mixtures or
for the measurement of unknown signals. A recent initiative by
the Metabolomics Quality Assurance and Control Consortium
(mQACC) [10] is in the process of providing living documents
on good quality management for metabolomics. This chapter aims
to offer brief guidance on steps which may improve overall quality
in an untargeted metabolomics experiment, and some aspects may
also be useful for individuals undertaking targeted metabolic
profiling.
Metabolomics is a science that generates large amounts of data,
normally with the goal of biological interpretation. Technical varia-
tion in the analysis of these datasets is a frequent cause of reduced
statistical power directly resulting in increased false positive and
negative results [11, 12]. Such variation can be due to random
effects or systematic variation [13]. One of the benefits of stringent
quality assurance is its role in reducing and defining the technical
variability that may lead to a statistically significant but a biologically
unimportant difference between two or more populations of inter-
est. The specific extra advantages of a strong QA process to a
metabolomics lab include the following:
– Confidence that data is robust and reproducible and that results
are biologically meaningful.
– Increased time and cost efficiency in the lab.
– Better confidence in planning and executing experiments, with
reduced risk of loss of potentially irreplaceable samples.
– Better record keeping and assessment of progress over time.
– Greater staff engagement.
– Easier management of on-boarding of new individuals or trans-
fer of methods to new laboratories.
– Greater confidence and compliance in meeting open science
requirements, including publication of raw data and steps in
data processing.
It is also prudent to consider the potential risks of not having an
adequate quality assurance process in place, including increased risk
of false discovery, increased risk of missing an important finding
because the technical variation exceeds the biological variation, and
 Quality Assurance in Metabolomics and Metabolic Profiling 17

a higher frustration level in the lab as experiments are not successful

and results are ambiguous. The reader is particularly drawn to the
need for impartiality [3] as part of a quality management strategy. In
the ISO standards, this normally refers to maintaining objectivity
and avoiding financial or other conflicts of interest. From a scientific
perspective, widening this definition to objectively assessing the
quality of the resulting data can only improve the scientific
robustness.

2 Background

2.1 Benefits of The major reasons for good quality management in a lab are to
Quality Management improve experimental quality and robustness and to have good
in the Lab traceability of what was done, including any deviations from pro-
tocols or unexpected occurrences. This in turn should improve the
repeatability of experimental work, including in other labs. In
addition, multiple studies across different industries have shown
that implementing a quality standard such as ISO9001 leads to
better financial performance [14]. While the current ISO quality
standards are referred to here for guidance, certification is not
required to implement the quality management guidelines. Good
quality management will be beneficial in any lab, and each lab
should implement practices that best match their size and budget.
In generic terms, the ISO9000 series defines the benefits of a
good quality management system (QMS) [15] as follows:
1. Defining, improving, and controlling processes.
2. Reducing waste.
3. Preventing mistakes.
4. Lowering costs.
5. Facilitating and identifying training opportunities.
6. Engaging staff.
7. Setting organization-wide direction.
Having a defined QMS communicates a readiness to produce
consistent results to both staff and “customers.” The term “cus-
tomer” will be used in the broadest sense throughout this chapter.
Most QMS were developed for industrial objectives where profit
and customer satisfaction are key objectives. In academic, govern-
mental, and nonprofit labs, the customer may be either a collabo-
rator, a program office and their stakeholders, or the individual
themselves requiring high quality data for their research project.

2.2 Defining Needs Before a QMS can be implemented, the requirements of all stake-
and Expectations holders need to be determined. In a metabolomics context, the
requirement of an international company offering pay-for-service
18 Jennifer A. Kirwan et al.

metabolomics services has very different requirements, and budget,

compared to a small lab in a university carrying out cutting edge
research. Stakeholders can include paying customers, but may also
include students in the lab, grant funders, university strategy teams,
and wider organizations, e.g., regulatory bodies for medical
research. Identifying external and internal issues which will impact
quality results, both positively and negatively, should also be con-
sidered at this stage. The scope of each individual study should also
be carefully determined to ensure that the planned experiments
match the expectations.

2.3 Risks and A formal QMS communicates a willingness to address both risks
Opportunities and opportunities associated with the organizational or individual
laboratory’s objectives. Risk-based thinking encourages a lab to
reduce risk to the overall results and to seek and implement oppor-
tunities to improve. It therefore involves risk management, i.e., to
think in advance of which steps, in which processes, could go wrong
and to plan in advance to mitigate or avoid these risks and their
effects. In addition, it also uses this assessment process, and the data
monitoring inherent to QA and QC processes, to identify areas
where changes in the process could improve the results in some
way, either through time, cost, or quality improvements. Risk-
based thinking and risk management are central to the aims of
QM and should be a major focus of any metabolomics laboratory.
The two major risks in a metabolomics lab tend to center around
people and instruments. People are our greatest asset, and knowl-
edge loss from the lab is difficult to replace. Instruments are our
most expensive laboratory investments and are central to the qual-
ity of the results, and the risk management of both project time and
quality is ultimately dependent on keeping them operating at their
best. People management starts with good leadership and a clear
vision of how to recruit, retain, and promote the best staff. In
academic environments, the individual lab can do much—even
within an unstable financial environment—to make their key mem-
bers of staff feel valued and appreciated. Readers are referred to the
wide variety of resources on people management and leadership,
details of which are outside the scope of this chapter. Regarding
instruments, thorough research when purchasing new instrumen-
tation, and a structured program of training staff in best practice
maintenance will help to mitigate this risk.
Good risk management includes detailing the critical risks in
each process: running out of the correct consumables, contami-
nated solvents, contamination from plasticizers or other chemicals
from consumables such as Eppendorf tubes, human error (e.g.,
mixing up samples), methods not being robust, etc. Risk manage-
ment strategies to minimize such occurrences include, but are not
limited to, the following:
 Quality Assurance in Metabolomics and Metabolic Profiling 19

– Detailed planning, especially regarding ordering and inventory

management, record keeping, time management of experimen-
tal workflow, and people management.
– Buying consumables in bulk (as far as use-by dates allow) such
that single batches can be used for single projects [16].
– Use of high quality solvents (LC-MS grade) and using known
suppliers [17].
– Checking whether solvents or chemicals are still in date and
routine testing of solvents and consumables to check for
contaminants.
– Rules on labelling samples, e.g., type written only, or quality
checking that individuals are labelling with legible numbers
(1 and 7 and 6 and 9 are particularly easy to mix up). Be
particularly aware of different countries‘ conventions in number
writing when working in an international lab.
– Four-eye principle for large or complex experiments to reduce
sample mix-up.
An active approach to identifying and adopting opportunities
for quality improvement is important. Indeed, appropriate risk
management ensures that quality thresholds can be met and
improved upon. For example, since many metabolites are poten-
tially labile in light, we have recently instituted a “dark room” for
sample extraction to protect light sensitive metabolites from degra-
dation. This additional step can reduce variability introduced dur-
ing sample preparation, and in doing so, it helps to ensure that
undesirable effects are minimized, while desirable effects are max-
imized. To improve quality, there also needs to be a willingness to
constantly refine and update processes as new information is
learned.

2.4 Quality Science works on the principle of setting out objectives and finding
Objectives and ways to achieve them. Quality assurance is no different, and a lab
Planning for Success should invest time in defining the quality objectives it is seeking to
meet and how they correspond with the lab’s wider strategic goals,
aims, and budget. Setting quality objectives is similar to setting out
scientific objectives. ISO 9001 particularly emphasizes that quality
objectives should be consistent with the quality policy, measurable,
and relevant to the work and needs of the lab and customers and
consider any applicable requirements.
Implementing a robust QMS requires good leadership, a clear
vision, adequate support and resources to implement the quality
plan, and identifying who the key individuals are for each stage and
how it should be monitored. The timelines and evaluation criteria
also need to be identified. Each individual in the lab should have a
clear understanding of what the lab is trying to achieve and how
their own part plays a role. To this end, formalized protocols should
20 Jennifer A. Kirwan et al.

be maintained and updated for every process. Additionally, “orga-

nizational knowledge” should be preserved as much as possible by
retention of key members of staff, and individuals new to a process
should receive adequate training. Good training and an optimal
working environment will assist with this goal.
In metabolomics, a broad definition of a quality result would be
one where:
• The biological question of interest has been clearly defined and
answered.
• The dataset(s) are measurably accurate, precise, and reproduc-
ible with either a broad range of biologically relevant signals
detected in the linear range (non-targeted analysis) or the meta-
bolites of interest can be precisely and accurately quantified in a
defined percentage of samples (targeted).
• The data analysis provides sufficiently and measurably robust
identifications and uses appropriate statistics for the experimen-
tal design. Confounding factors or bias are not seen to play a
major role in the results or can be measured and addressed using
correction factors.
• The experiment, when repeated with different samples, in a
different lab (and preferably on a different instrument) yields
similar results, i.e., the science itself is reproducible.

3 Table of Definitions

3.1 Definitions Definitions are based on ISO/IEC guide 99:2007. Readers are
directed to their websites [18] for a full list. Specific terms listed
here are discussed in ISO 17025 (2017) [3].
Decision rule: Accounts for the measurement uncertainty in a
measurement when trying to show that a test result conforms
with a specified requirement.
Verification: Objective evidence that a given method fulfils speci-
fied requirements. Verification is not synonymous with
validation.
Validation: Where a verification proves that the specified require-
ments are fit for intended use. While both the FDA and EMA
have well recognized guidelines [8, 9] for the validation of
analytical methods for single compounds, validation is more
complex where multiple compounds or untargeted metabolo-
mic methods are concerned. In addition, a method is only
validated for the biological matrix on which the original valida-
tion was conducted, especially for accurate absolute quantifica-
tion [19, 20].
 Quality Assurance in Metabolomics and Metabolic Profiling 21

4 A Framework of Quality Assurance Activities in Metabolomics

4.1 Organizational The metabolomics analytical pipeline is both long and complex
Context (Fig. 1) and increasingly involves multiple persons, from multiple
scientific disciplines. Technical variation can occur at all stages of
the pipeline, either from biases in sample collection (e.g., smoking,
sex imbalances, different collection clinics, etc.) to
non-standardized protocols or multiple instruments being used or
due to random variation.
While the analytical chemist may be involved in both sample
collection and data processing, their principal role focuses on the
sample handling and chemical analysis which occurs after samples
have been collected, processed, and stored. This chapter will focus
therefore on the laboratory analytical factors which the analyst can
control. While we recognize that multiple instrumentation and
methods are used for metabolomics, we will concentrate here on
mass spectrometry-based analyses. For readers interested in the
sample collection stages, they are referred to the wide variety of
literature on the subject [21–23]. Indeed, sample collection, pro-
cessing, and storage are worth a chapter in their own right, and
these steps need to be done well for subsequent steps to be success-
ful. The bioinformatical and data analysis steps also require strin-
gent quality management. A good understanding of how individual
processes build into a system, and the management of this opera-
tion, contributes to a laboratory’s overall effectiveness and
efficiency.

4.2 Defining Metabolomics takes place as a series of individual processes which

Individual Processes combine to produce the end result. The success of each process is
in the Lab highly dependent on the quality of the processes which took place
before it. Likewise, the quality of the current process will normally
directly contribute to the success of the subsequent processes.
Thus, each process as well as its exact beginning and end needs to
be defined, and the inputs and outputs need to be determined
(Fig. 2). The sequence of several processes, and how different
processes interact, also needs to be determined. In addition to
natural start and end points of processes, it is useful to identify

Fig. 1 The metabolomics analytical pipeline. Steps in blue are those where the analytical chemist tends to be
most involved
22 Jennifer A. Kirwan et al.

Fig. 2 Diagram detailing the sequence of metabolomics QA/QC processes and their interactions. (Adapted
from ISO 9001)

some QC points where the quality success of a particular process

can be measured. Of course, some processes lend themselves to
more QC points than others. Importantly, it should be decided
before beginning who the responsible individuals are, how pro-
cesses will be overseen and monitored, and which resources are
required, thus determining whether a process has the necessary
support to succeed. Health and safety should also be considered
when planning these steps. Furthermore, labs should seek to con-
trol external factors that feed into their lab processes wherever
possible, e.g., use of external providers or purchase of highest-
grade consumables. Understanding where the risks are in a process
(i.e., where it can go wrong) and the opportunities (where it can be
improved) can be done as part of an ongoing process to improve
the QMS.

4.3 Monitoring and ISO 9001 and ISO/IEC 17025:2017(E) both specifically mandate
Measuring of that organizations monitor and measure the resources they have or
Resources require to ensure valid results when monitoring. This includes that
they are suitable for purpose, properly maintained to ensure this
suitability, and that they are regularly calibrated. In a metabolomics
lab, this starts with an appropriate laboratory environment, i.e.,
with suitable space, gases, electricity, and other resources to main-
tain a high analytical quality. Critically, maintaining and retaining
good records is seen as an essential part of this process by ISO 9001
and is also an essential part of good scientific practice. To this end,
this chapter provides some essential checklists and templates that
can assist the metabolomics lab in record maintenance.
 Quality Assurance in Metabolomics and Metabolic Profiling 23

4.3.1 Instrument Commercial analytical instruments should be selected for their

Installation/Operational suitability for the function they are expected to fulfil. Vendor
Qualifications suitability is also important. It is highly recommended to have
vendors with quality assurance processes in place, e.g., ISO 9001
certification, to assess design, development, manufacturing, ser-
vice, and support [24, 25]. On arrival at the lab, instruments should
be installed by a qualified individual and should be verified for both
functional and performance specifications (equipment qualification
or EQ). EQ can be divided into design operation (largely the
responsibility of the instrument manufacturer), installation, opera-
tional, and performance qualifications (IQ, OQ, and PQ). The lines
between OQ and PQ are gray and can be different between instru-
ments [26]. Ultimately, the user needs to ensure that it was
installed properly, it is performing to a desired set of key perfor-
mance indicators previously agreed with the manufacturer (OQ),
and it performs to a specification showing it is fit for its regularly
intended use (PQ). In metabolomics, system suitability tests (SST)
largely serve the latter purpose. Instruments should be maintained
according to manufacturer’s instructions including regular servic-
ing and calibrations. Logs of these should be maintained (often a
simple Excel spreadsheet can serve this purpose). Users should
undergo comprehensive training from an appropriate person in
how to use the instrument in your specific laboratory setting (see
section on training).

4.3.2 Maintenance ISO 9001 requires organizations to ensure valid results when mon-
Schedules itoring. In practice, this means regular testing and calibration of all
equipment in the analytical laboratory involved in the monitoring
process. This includes, but is not limited to, pipettes, centrifuges,
balances, centrifugal evaporators, cold rooms, refrigerators, and
freezers, as well as the larger analytical instruments themselves
(chromatography instrumentation, MS, NMR, CE, etc.). Monitor-
ing includes calibration, verification, or both, and this should be
done to international measurement standards whenever possible. If
an international measurement standard is not available, the stan-
dard against which an instrument is being assessed must be formally
documented and retained as part of good quality assurance.
Labs that are already accredited under various national clinical
laboratory accreditation schemes, especially where they conform to
ISO 15189 (2022) (Medical Laboratories–Requirements for qual-
ity and competence), are probably already meeting these require-
ments [27]. However, even for labs that do not seek accreditation,
regular documented maintenance and checking of all systems are a
must for ongoing confidence that results are both precise and
accurate. For more advanced analytical instrumentation, checklists
can serve well (Table 1). At least some of these checks can be carried
out by trained in-house personnel (e.g., calibration accuracy of
pipettes), although we would recommend for safety reasons that
24 Jennifer A. Kirwan et al.

Table 1
A template for a checklist for the regularly scheduled maintenance of a Q-Exactive (QE) mass
spectrometer coupled with an Accela LC system

Monthly maintenance:

 Check QE cooling fan filters and clean as needed

 Check gas ballast valve on fore pump

Date Initials Date Initials Date Initials Date Initials Date Initials

 Flush LC pump piston guide bushings

 Flush LDA/columns with IPA for 20 CV at a low flow rate (200 uL/min or less)
 Replace aqueous solvents

Date Initials Date Initials Date Initials Date Initials Date Initials

Annual maintenance:
 Change fore pump oil (6-12 months)  Replace autosampler valve rotor
 Check anti-suckback valve on fore pump seals (6-12 months)
 Clean fore pump fan guard  Replace DLW holding loop
 Clean and grease the rolls and rails
of cold stack

Date Initials Date Initials Date Initials Date Initials

(continued)
 Quality Assurance in Metabolomics and Metabolic Profiling 25

Table 1
(continued)

 Replace primary piston seals on LC

pump LDAs (6-12 months)
 Leak check QE gas lines  Clean check valves and pistons
when replacing piston seals on LC
pump LDAs (6-12 months)

Date Initials Date Initials Date Initials Date Initials

As needed:

 Clean API source exit lens and S-lens

Date Initials Date Initials Date Initials Date Initials Date Initials

 Replace HESI-II probe needle insert

Date Initials Date Initials Date Initials Date Initials Date Initials

 Replace DLW syringe plunger and needle

Date Initials Date Initials Date Initials Date Initials Date Initials

 Replace LC-valve Teflon tube/ferrule

Date Initials Date Initials Date Initials Date Initials Date Initials

 Rebuild the N2 generator compressors (max 3x per compressor)

 Change valves and filters
 Reset service timers

Date Initials Date Initials Date Initials Date Initials Date Initials

This can be adapted to most high-resolution LC-MS instruments with minimal changes
26 Jennifer A. Kirwan et al.

other equipment, e.g., centrifuges, should be tested only by some-

one formally trained in both the testing and the safety procedures.
Normally, an individual from a company specializing in such pro-
cedures is required.
Under ISO 9001, it is also a requirement that each individual
instrument is formally identified to allow their status to be checked,
and all instruments must be safeguarded from accidents which may
damage or invalidate their calibration. For mass spectrometry, vent-
ing an instrument is well known to disturb the calibration status. It
is not always appreciated that other environmental issues such as
unstable temperatures will also affect mass accuracy. This has
proven to be a problem in the past, e.g., in institutes where
higher-level decisions have been taken to switch off air condition-
ing overnight for environmental reasons.
Mass accuracy, mass precision, and reproducibility across ana-
lyses should be measured and documented as part of a formal SST
for MS and NMR instruments on a regular basis. Where comple-
mentary instruments such as chromatography or ion mobility are
being used, retention time or collision cross-section (CCS) values
should also be tested for reproducibility as part of this procedure.
More information, including on reporting standards, can be found
in Kirwan et al. [4].

4.3.3 Temperature Temperature is a critical parameter for sample storage, analytical

Monitoring Systems preparation, and measurement integrity. Sample storage tempera-
ture will affect both the range of detectable metabolites and the
technical variability for particular metabolites. Temperature stabil-
ity is often equally, or sometimes more, critical than absolute tem-
perature. A detailed and implementable temperature action plan is
ideally in place, including plans to move samples to spare storage if
required, with which staff can respond in the event of an emer-
gency. Refrigerator and freezer temperature monitoring is highly
recommended, ideally with an emergency notification system to
automatically alert named lab staff when a temperature moves out
of an optimal range. Temperature gradients and temperature
changes in a sample storage facility should be avoided where possi-
ble; temperature gradients used to be a particular issue with liquid
nitrogen gas phase storage, although changes to freezer design have
vastly reduced this risk [28]. Moving samples between freezers or
sites remains a considerable risk for subjecting samples to unwanted
temperature gradients or changes and should be undertaken
with care.
Temperature is also critical for sample preparation and analysis.
Technical reproducibility during sample preparation can often vary
dependent on both the absolute temperature and the variation in
temperature throughout the process. For this reason, it is best both
to accurately control and state temperatures in the methods.
 Quality Assurance in Metabolomics and Metabolic Profiling 27

Statements such as “extraction occurred at room temperature”

should be avoided as too imprecise. Room temperature and its
monitoring are often also critical in maintaining optimum perfor-
mance of instruments. Commercial instruments all have a stated
preferred temperature range for operation. Even small variations in
temperature can jeopardize the instrument calibration, leading to
less accurate data and potential batch drift. A clear understanding of
which instruments are susceptible to small temperature fluctuations
will improve lab planning and design for environmental control.
Particular attention should be paid to temperature gradients across
a room, where large instruments may hinder air circulation and
reduce the efficiency of any air conditioning.

4.3.4 Documentation and There are two reasons to keep good documentation:
Audit Trail
(i) Most importantly, to have good, documented records of every
process procedure, results, and improvements. This provides
an overview of how the lab is performing, both against its own
quality objectives and also in relation to international labs with
similar aims. It is critical for monitoring where processes may
require optimization to maintain or improve quality.
(ii) To comply with good scientific practice and the international
standard laid out in ISO 9001.
All formal documentation created within a lab should have
appropriate identification and description including, but not lim-
ited to, title, date, and author or reference number. In many cases, a
version number may also be appropriate. Control of documenta-
tion, including who has access to it, and any data protection issues
associated with confidential documentation are also important
[2]. Version control of documents is important, and it is strongly
recommended that fully audited software (e.g., GitLab) is used to
maintain the most important documentation such that no changes
can be implemented without them being recorded. Audit trails,
such as those implemented on many vendor proprietary software
programs, enable formal records of who has made which changes at
which timepoints and are easy to implement as a formal record.
Documentation, both external and internal, may be required both
to show that certain conformity standards have been met and for
certification purposes. It is important that such documentation is
stored in a way that does not allow unintentional alterations.
While often considered synonymous, there is a technical differ-
ence between documents and records which determines how they
should be treated. ISO9000:2015(E) [1] defines a document as
both the information and the medium on which it is stored. Exam-
ples of documents include specifications (including the quality
management plan), documented information, and formal records.
It can include written policies, manufacturers’ technical
28 Jennifer A. Kirwan et al.

specifications and plans, posters, and charts. Documented informa-

tion is any information including processes and operational infor-
mation, including standard operating protocols, that must be
properly stored and controlled by the organization. ISO/IEC
17025:2017(E) [3] specifically emphasizes the requirement for
management system documentation for a testing and calibration
laboratory, including ensuring that all personnel have access to
those documents that are required to fulfill their responsibilities.
In addition, authorized personnel should check and approve docu-
ments prior to their release, and documents should be regularly
reviewed and updated, uniquely identified, and versioned. Ideally, a
system should be established to prevent the use of obsolete versions
of documents. Documented information also includes records.
Records refer to any document which contains evidence of the
results achieved by a process, including traceability, verification,
and both preventative and corrective action. This would include
both lab books and additional records such as instrument logbooks.
Different procedures need to be adopted with records compared to
other documented information. As with all documents, records
should be readable (both legible and stored on appropriate
media), appropriately stored, and backed up. However, while pro-
cesses and operational information should be controlled, versioned,
and updated as required by contrast, records should not be ver-
sioned or altered. This is because of the role records play in
providing a provable record of the evidence of results.
All-important documentation, especially records, should be subject
to good data management practices to ensure they are suitably
backed up and archived for a reasonable length of time after the
project is completed; general rules of good scientific practice are
country and specialty specific but are normally around 10 years
[29, 30]. Practically, these requirements are fulfilled by considering
the recommendations for good scientific practice on data backup
and data management [31–33]. Specific examples of individual
documents or records are dealt with below.
1. Protocols
Poor preanalytical or analytical procedures can introduce arte-
facts or unwanted variability into the resulting dataset, includ-
ing potentially destroying or modifying metabolites of interest
[34–37]. Well written (or filmed) protocols (also known as
standard operating protocols or SOPs) should be available
and validated for both their effectiveness in achieving their
aims and in being both detailed and easy to follow by a trained
person. Protocols should be documented for every process
undertaken in the metabolomics lab on a routine basis: balance
usage, pipetting, preparation of buffers, preparation of accu-
rately diluted standards, health and safety procedures, cleaning
procedures, etc. [38]. Protocols for sample extraction and
 Quality Assurance in Metabolomics and Metabolic Profiling 29

analysis should also be available per method, and details on how

to assess the quality of a process and assess when a process has
failed can either be included as part of a process SOP or
designated as a process in its own right with its own protocol.
Training individuals in each protocol and regular auditing and
reviewing of protocols and other documentation are impor-
tant. Versioning prevents out-of-date protocols being used and
should include details on what has changed and why a new
version is useful. Consider publishing protocols openly in a
centralized repository such as Protocol Exchange (https://
protocolexchange.researchsquare.com/). Checklists can be
useful for particularly complex protocols.
2. Instrument Logbooks
Logbooks record changes or occurrences on instrumentation
and can be important in both tracking how an instrument is
performing and in explaining unexpected results, including
sometimes many weeks after data was collected. Exactly what
to collect in a logbook may be instrument specific; we would
recommend calibration results and details of interventions
including cleaning, venting, installation of new parts (including
consumables such as columns or liners), and any observed
instrument issues (including how they were resolved).
3. Lab Notebooks
In addition to protocols and logbooks, well kept, legible lab
books detailing all experimental work should be kept according
to good scientific practice. The ALCOA+ principle is a good
guide: attributable, legible, contemporaneous, original, and
accurate and ensuring it is complete, consistent, enduring,
and available [39, 40]. Enough details need to be recorded in
a lab book to enable another individual from that scientific field
to replicate what the individual did. To ensure that no impor-
tant tasks are forgotten, a checklist is useful when heading into
the lab each day (see Table 2). With increasing use of electronic
lab books, templates (to enable easy recording of experimental
parameters and QC results) make the well-organized scientist’s
life much easier. An example template of a scientific mass
spectrometry experiment is given in Table 3. It has the benefit
both of ensuring all important information is recorded and that
it is easy to find if the lab book is later used by another individ-
ual. Good lab practice (GLP) mandates that lab books should
be co-signed every day. Instituting a signature process in the lab
ensures that record keeping is thorough (and handwriting is
legible!). Appropriate backups of data and records should be
regularly made. Implementing the 3-2-1 rule of different cop-
ies on different media forms in different locations is a good
starting point for approaching data management. Data backup
strategies are equally important for paper lab books and may
30 Jennifer A. Kirwan et al.

Table 2
A template for a lab book checklist

Version X.X: [Lab book owner’s name]

Aims: for each day
□ Are the aims of the work done today clearly defined?
□ Is the sample owner clearly defined?
□ Is the project clearly defined?
□ Have you confirmed there is a quote and a purchase order in place?
□ Are the samples clearly defined – type, number, class membership
□ Is any re-numbering of samples clearly defined?
Sample preparation
□ Is the protocol that was used clear, with a version number and a copy stored in the lab book which is
easy to find? (this can either be an electronic link to a protocol (must be an official version of a protocol
stored in the protocol folder), or a copy of the protocol copied and pasted into the day’s lab book).
Deviations from official protocols should be written out in full for that day, including the reasons for
any deviations.
Where details are included in the protocol, these do not have to be included specifically for that day’s lab book,
except where there is a deviation from protocol.
□ Are the samples that were prepared and the order in which they were prepared clearly stated?
□ Is the origin of the samples clear (especially if aliquot numbers from a biobank)? Is there storage history
clear, including any freeze/thaw cycles?
□ If samples were taken from frozen storage, what temperature were they stored at and how were they
defrosted and for how long?
□ Is the amount of sample used clearly stated, along with any official observations, e.g., 20 μL of plasma
was taken, avoiding fibrin clots when pipetting?
□ Is the assistance of anybody else in the lab properly noted, including who did what?
□ Is the order of solvent addition and the amount of solvent added in each step clear (including if
solvents were added individually or mixed); if mixed, has this been properly validated? Are the batch
numbers and makes of solvents and chemicals properly noted?
□ Are full details of any internal standards noted, including amounts, concentrations, and when they
were added? Is the date the standards were prepared noted on the official records?
□ Is the addition of any other steps, e.g., derivatization steps, fully noted, including processes, amounts
of chemicals, temperatures, and times?
□ Are any particular conditions (e.g., prepared in dark room, all samples were prepared within 60 min)
noted? This is particularly important to note if denoted high risk to quality on protocol not to follow a
certain step.
□ Details of preparation of any blank samples; how prepared, which processing steps they underwent and
amounts, and type of solvents or solutions used to mimic sample.
□ Details of preparation of QC samples, including how prepared and from which original samples. Also
include details of commercial QC samples, e.g., LTR QCs here.

(continued)
 Quality Assurance in Metabolomics and Metabolic Profiling 31

Table 2
(continued)

□ Were samples dried or otherwise stabilized? What happened to samples after drying (i.e., immediately
analyzed or stored—If so at what temperature)?
Sample analysis
□ Details and results of instrument SST including method and chemical composition of SST. Full details
of how the SST standard was prepared should be included.
□ Exact details of samples used, to include how they have been stored up to now, and for how long, e.g..
the 24 Baker project human serum samples prepared on August 2, 2023, and stored at -80C for 1 week
were taken from freezer for analysis. Full sample details are provided on that date.
□ Which instrument has been used?
□ Which method was used?
□ Which column and batch number?
□ For LC-MS: which buffers were used, how were they made up, and when?
□ For GC-MS: which liner type and gas was used?
□ For other analysis type, please denote specific details.
□ Which chromatography method was used: name, gradient, injection amount, and all details that allow
the experiment to be repeated on another instrument of the same specification?
□ Which MS method was used: all details of method that allow the experiment to be repeated on another
instrument of the same specification?
Data preprocessing
□ Initial check of raw data; were any rejected due to data quality concerns? Which and why?
□ Which software was used (including version number) to carry out which steps in data preprocessing?
Inputs and parameters used should be noted, as is the location of the original data and subsequent data
steps.
□ Are all data preprocessing steps clearly named, and is the order in which they were carried out clear?
Are outputs of each step clearly recorded and easily accessible to authorized individuals?
□ For steps that require a reference dataset, e.g., pqn normalization, is the data that was used as the
reference dataset clear?
□ What filtering steps were carried out on the data; which parameters were used?
□ Which quality control checks have been undertaken on the data? Which samples were excluded as a
result?
Data analysis
□ Is a data analysis plan in place before starting?
(NB: Certain datasets carry a lot of information in what is not there, e.g., there may be biological
information in the missing value patterns seen between genetic knockouts and their wild-type
counterparts; consider this in the data analysis plan)
□ Which dataset is being used for which data analytical step?
□ How have missing values been handled for each data analytical step (e.g., imputed with knn—k = 5 or
statistical analysis set to ignore missing values, etc.)?

(continued)
32 Jennifer A. Kirwan et al.

Table 2
(continued)

□ Are all software packages, version numbers, and parameters recorded?

□ Is the data analysis recorded somewhere where it is correctly versioned, commented, and stored, e.g.,
Gitlab, centrally stored R studio project, or is otherwise correctly documented so that it can be
accessed later? Is this linked to in the lab book?
□ Is a clear summary available in lab book?
Identification
□ How was identification carried out? What level of identification does it achieve?
□ Software and version number recorded, along with parameters
□ Results stored centrally for access by appropriate members of lab and link provided in lab book
□ Summary recorded in lab book
Metabolite pathway mapping and related commercial services
□ Which software tool was used, which version number and which parameters?
□ Which dataset was used? How does this affect interpretation, e.g., if only statistically significant results
were used, interpretation is very different to where an enrichment analysis was attempted? Joint
pathway or multi-omics methods results are different again
□ Link to full, centrally stored results
□ Summary of results recorded in lab book
□ Clear explanation of what was done and why to allow data analysis to be repeated

require additional thought to process monitoring where the

process cannot be automated [41]. Also consider how software
code is used and stored to ensure data traceability. This includes
vendor software as well as code developed in-house. Versioning
tools such as Gitlab or Anaconda can be beneficial to store
coding histories and versions in an organized way.
4. Data and Metadata Management
Where possible, keeping copies of all relevant information,
including metadata in a lab book, is still recommended as a
secure way of saving the data. For example, if links are used to
original data, storage should be implemented so the links do
not become unusable as data is archived. Furthermore, it is
generally a good idea to implement a triply redundant backup
system to preserve important data. For example, we have found
it useful to keep original data on the instrument or PI’s com-
puter, backed up to a local network drive that has its own offsite
backup in case of power outages that can compromise the
fidelity of the data. Regardless of the specific data management
system, electronic data should be regularly backed up accord-
ing to best practice guidelines, including managing traceability,
confidentiality, and fidelity of data [40, 42].
 Quality Assurance in Metabolomics and Metabolic Profiling 33

Table 3
A template for an MS untargeted measurement suitable for use and adaptation across a range of
mass spectrometers

Project name: Max SUPER

Folder: S:/2023/Max Super

File name: Untargeted work LC-MS measurement

Mass spectrometer: Waters Synapt

Complementary instrumentation (make, type, ID) Waters Vanquish 1D LC-MS: “Walter”

Sample list

Machine details:

Machine/Mass Mode (include Analysis MS-Method

spectrometer details of ion type
mobility and
MSMS here)

Waters Synapt DIA Positive LC-MS C:\MassLynx\methods\2023_06_08_untargeted.mat

mode, mass
range: 70-1000

If targeted MSMS is used, include list of targets and collision energies, or link to list of targets.
A text copy of the method should be kept in an appropriate place in the lab folder.

Liquid chromatography

LC method name C:\MassLynx\methods\2023_06_08_untargeted.mat

Buffer A 0.2% FA in water (LC-MS grade, Sigma):

(Solvent mix, exact composition, grade
and manufacturers)

Buffer B 0.2% FA in methanol (LC-MS grade, Sigma):

(Solvent mix, exact composition, grade
and manufacturers)

Buffer C N/A

Seal wash 10% iso-propanol in water (water and iso-propanol

(Solvent mix, exact composition, both LC-MS grade, Sigma)
grade and manufacturers)

Injection wash(es) Acetonitrile 10% one wash followed by Acetonitrile

(solvent mix, exact composition, 90% one wash (water and iso-propanol both LC-MS
grade and manufacturers) grade, Sigma)

Chip-based Electrospray Ionization Method name

NanoMate (Triversa) N/A

(continued)
34 Jennifer A. Kirwan et al.

Table 3
(continued)

Maintenance:

Machine details operator Last calibration:

SST details

Date run

Composition of SST

Method used

Buffer 1

Buffer 2

Column used (for LC) Exact type Serial number

Results stored Full file and folder name

RMS retention time drift

RMS mass error

Overall result Pass/Fail

Notes (detail previous calibrations or SSTs

that failed and required repeating.
Include details of why they failed if known)

(SST and calibration results should also be entered on the respective instrument log tables for
ongoing monitoring)

Type, size, serial number, batch number

Column: VisionHT C18 Basic; L × I.D. 150 mm × 4.6 mm, 3 μm particle size
(Dr. Maisch) part number 230189

Column: N/A

Type, size, serial number, batch number

Chip: N/A

During measurement:
no problems reported:
Injection problems:
Spray problems:
Mass inaccuracy:
Column problems:
Leaks:
Other:
 Quality Assurance in Metabolomics and Metabolic Profiling 35

5. Additional Documentation
Additional documentation is any documentation that acts to
improve QA and can include, for example, tracking of projects’
metrics over time to see how interventions improve results,
laboratory management information systems (LMIS) or
freezer records to track where samples are stored and how
they have been processed so far, project management tools to
track project progress, etc. Tracking metrics such as number of
signals, number of missing values, and relative standard devia-
tions of QC samples across multiple projects can be particularly
useful for monitoring how incremental changes over time may
be affecting results. Project management tools have proved
useful in our hands in monitoring progress of projects and
better time management of multiple projects leading to better
efficiency.

4.3.5 Staff Training Training An individual carrying out a process must have the com-
(Including Records) petence to undertake this process. The leader of a lab, and ulti-
mately the organization, needs to monitor this.

Competence is arranged around four major themes:

1. The initial competence that an individual has in a particular
process. This normally requires an assessment of the difficulty
of the process to be undertaken, the background of the indi-
vidual tasked with the process, and the ability of the individual
to be trained in the process.
2. Where required, the leader of the lab is then tasked with
organizing the training of the individual in that task.
3. Where required, an evaluation of the necessary competence.
This can be done by shadowing an individual in a lab, evaluat-
ing the results of quality measures, or both.
4. Documentation reporting training and testing. The purpose of
the documentation is for the lab, although an additional
benefit of such formal documentation is for the individual to
prove to future employers their competence in certain areas.
We are fortunate in the analytical laboratory that they are
normally staffed with highly motivated, highly intelligent indivi-
duals who enjoy learning new tasks. Training in use of complex
instrumentation is routine in many labs and ideally should be both
structured and documented. Rather surprisingly, in a recent survey,
over half of respondents (65%) stated that they had no in-house
training program [38]. Labs can choose to outsource their training
to external companies (instrument manufacturers or training
schools) or create their own structured program. It is useful to
properly plan who will perform the training, how many hours it
will take to reach each level, how competency is shown that the level
36 Jennifer A. Kirwan et al.

has been achieved, and how this will be documented. Who requires
training and how the training lessons will be broken down and
structured are also important. The trainer is ideally both competent
and confident in their own knowledge of the instrument or process
that they are teaching. A checklist is useful, both to ensure that
training is systematic and so that both the trainer and the trainee are
confident that all relevant knowledge has been passed on for that
level (see Table 4). In many cases, it is helpful to have more than one
trainer training an individual since different individuals can stress
different aspects when training. It can also highlight any differences
in practices in the trainers which can then be addressed.
In the area of quality assurance, improving performance in
highly routine (potentially boring) tasks should also not be
neglected, e.g., pipetting, making of standard solutions and buf-
fers, and weighing. The idea of marginal gains is well known in both
sport and surgery and is extremely important in the field of quality
assurance [43–45]. It can often highlight differences in practice
(including health and safety practice) and encourage dialogue.
Indeed, a culture of regularly refreshing these skills for all lab
members in a fun training session encourages both teamwork and
communication. Assessments of such skills should be done in ways
that encourage confidence and for individuals to seek help where
required, e.g., anonymized competitions. Since many labs also have
robots, the robot should also be assessed on the accuracy, precision,
and repeatability of its pipetting, including across different solvent
types.

Motivation Ultimately, a quality assurance process is only as good

as the person who least adheres to it. It is no accident that many of
the quality assurance models put personnel at the center of their
models. Giving individuals at the bench more power to improve
their own quality assurance protocols both enhances motivation
and often leads to insights about quality improvement which are
not apparent to managers. Having quality improvement champions
throughout the lab raises enthusiasm for the process. Training is an
essential part of this recipe. Time spent training is reaped in less
time spent problem-solving. Keeping training records enables each
individual to have training needs assessed one or more times a year.
Training can be internal and external. For example, one of our best
technicians trains everybody in the lab on their pipetting technique
(QA). They are then assessed after training (QC). This will be
assessed once a year in a fun anonymized assessment. If any individ-
ual performs badly, they can be discretely retrained. This ensures
that everyone new to the lab learns an accurate way to pipette,
regardless of previous background or training, and reduces the
likelihood of technical variation resulting from poor pipetting style.
 Quality Assurance in Metabolomics and Metabolic Profiling 37

Table 4
A template for recording training of new lab staff

Start date of training: End date of

training:
Name of trainee:
Name of trainer:
Document ID # and title (SOP, etc.)
Document ID # and title (SOP, etc.)
Software (name and version #) (if applicable)
Software (name and version #) (if applicable)
Trainer has completed the following:
□ Provided the above documents to the trainee
□ Demonstrated the process, technique, procedure, or software
□ Demonstrated required routine maintenance, instrument verification, and/or □ Not applicable
function checks
□ Demonstrated data processing (use of software, calculations, etc.) □ Not applicable
□ Provided information on records management requirements for the process, technique, procedure, or
software
Trainee has completed the following components:
□ Read and understood the documents provided
□ Successfully performed the procedure with known samples, materials, or □ Not applicable
standards
□ Demonstrated problem-solving skills □ Not applicable
□ Performed required routine maintenance, instrument verification, and/or □ Not applicable
function checks
□ Successfully performed data processing (use of software, calculations, etc.) □ Not applicable
Comments:
I have completed the training and competency components above and feel confident that I can accurately
perform the procedure(s)
Trainee signature/Date:
Trainee has successfully completed all training and competency components listed above and has been deemed
competent to perform the above procedures independently.
Trainer signature/Date:

4.3.6 Method ISO17025 [3] specifically recommends both using appropriate

Development/Validation methods and validating any methods used in the lab including
laboratory-based methods, non-standard methods, or methods
applied outside of their original scope (which would include using
a method validated on one matrix for another matrix). Both the
38 Jennifer A. Kirwan et al.

EMA and FDA provide comprehensive advice on how to validate an

analytical method for a single compound. In addition to their
recommendations, confirming that there is no chemical interaction
or mistaken identities, including in a calibration curve, will improve
quantification accuracy [46].

4.3.7 Quality System Performance Evaluation

Audits Performance evaluation requires that an individual lab monitors its
key metrics to see if it is performing to the required standard. These
metrics should have been decided as part of the QMS process, and
labs are recommended to look at the international situation for
their methods, techniques, and technologies to set appropriate
metrics. In addition, while quality is important, overall output
should also be assessed as a quality measure to ensure that attention
to quality is improving rather than disabling productivity. Auditing
of metabolomicsMetabolomics procedures has previously been
mentioned by expert consortiums to improve QA [5].
As part of a performance evaluation, labs should determine
who performs the evaluation, how often, and how it will be moni-
tored and measured. Performance evaluation can be done as part of
a formal accreditation scheme such as by International Organiza-
tion for Standardization (ISO) or College of American Pathologists
(CAP); in practice, only around 10% of labs reported have formal
accreditation [5]. Internal audits are low cost and can be regularly
performed. Performance evaluation feedback to fully explain results
enables performance in audits to improve in the future. Perfor-
mance evaluation should evaluate not only the performance of the
quality standards within the lab and whether international stan-
dards on quality management have been met but also external
provider services, the implementation and functioning of the
QMS and planning, effectiveness of evaluating risks and opportu-
nities, and any required improvements. The leadership team should
define the scope, criteria, and regularity of each audit, including
who should be involved. Results should be documented and
reported to the local management, and importantly, any corrective
action should occur in a timely manner. The management team
should also regularly review the QMS to assess its continuing
suitability and effectiveness and to confirm it fits with the strategic
goals of the lab and wider institute.
“Customer satisfaction” is an important part of performance
evaluation. Were they satisfied with the quality of the data, the
reporting, and the time taken? Customers who did not have their
“expected” results due to scientific reasons (rather than perceived
poor performance on the behalf of the lab) should have their
feedback properly evaluated to see if they had understood why
results were not as anticipated and there are good lessons to be
learned in evaluating likely project success at the beginning, proper
sample collection and delivery, and communication success.
 Quality Assurance in Metabolomics and Metabolic Profiling 39

Nonconformity and Corrective Action

Clear guidelines are required as to course of action if a quality
parameter is not achieved. Whether in between studies or during
the acquisition of metabolomics analytical data, this normally
means a go/no-go decision, i.e., whether to keep going, revise,
or stop and review. If the decision for a failed quality parameter is to
keep going, this suggests that either the quality parameter is not
important enough that it should be used or there is not enough
focus on quality. Revising may include either e.g., cleaning an
instrument if an SST has not reached expected criteria or repeating
an extraction again if there is clear contamination in the process
blanks with a known cause. Stop and review occurs when a quality
parameter is not met, and the reasons are not clear. This requires
individuals to review what has gone wrong and to try and deter-
mine the cause. If, for example, a mass spectrometry result has
intensities less than anticipated, this may be due to a degraded
standard, or a tuning or calibration issue, or that the mass spec-
trometer needs cleaning. It could also indicate something has
fundamentally gone wrong with the instrument. Much like a medi-
cal diagnosis, it is helpful to list the differential diagnoses that may
be causing the observed symptom and systematically work through
the list, reviewing the effect of each diagnostic test or corrective
action. Things that occur frequently should be at the top of the list,
and things that are inexpensive to check (in both temporal and
financial terms) should also be high on the list. Rare occurrences,
especially when they are expensive to diagnose, should be left until
last. The old medical adage “when you hear the sound of hooves,
don’t think first of zebras” also applies to analytical chemistry. Lists
of common problems and their solutions should be documented in
each lab per instrument type to enable smoother problem-solving
the next time the problem occurs. Documenting all corrective
action and its effects is important, both for future problem-solving
and to have a history of the instrument. Marginal gains also apply
here, and sometimes a combination of solutions is required to reach
the required standard again. Risk analyses should be updated where
it is perceived that certain actions may have contributed to the
quality failure, e.g., requiring extra sample cleanup steps for partic-
ularly dirty samples such as fecal samples. In this way, continual
improvement becomes not just an aim but an actuality in a
metabolomics lab.

5 Putting Quality Assurance into Practice

A checklist is provided in Table 5 describing many of the steps in an

average analytical experiment along with key questions to consider
from a quality assurance perspective. The checklist is meant to be
used during all steps of the metabolomics study in which the
40 Jennifer A. Kirwan et al.

Table 5
A template for a checklist for an average analytical experiment doing metabolomics

Workflow stage Steps Quality assurance points

Aims Define wider study aims □ Do all experimentalists understand the wider aims of the study?
□ Have the major objectives and biological questions been defined?

Scope Define scope of □ Are the planned experiment(s) sufficient to answer the biological question(s)?
metabolomics □ Are there sufficient resources available?
□ Are the wider processes well aligned with the metabolomics (e.g., sampling
protocols)?
□ Have appropriate controls been discussed?
□ Are sufficient number of samples planned for the required statistical power?

Collection Sample collection □ Have samples been collected using a metabolomics-specific protocol?
□ Have samples been stored using a metabolomics-specific protocol?

Transfer Sample acceptance □ Are the number of delivered samples correct?

□ Are there enough samples per statistical class?
□ Are all samples useable?
□ Are samples traceable?
□ Have samples been entered/updated on the tracking system?
□ Have field blanks been collected?

Training Training □ Are individuals trained in the techniques to be used?

□ Are robots validated in their methods?
Method validation □ Are method(s) taken from published paper validated in-house?
□ Are method(s) developed in-house validated?

Quality Required QC points □ Are QC points established throughout the workflow?

controls Required QC samples □ Are QC samples properly considered and used in each study?
(QCs) Required blank samples □ Are all routes of contamination properly considered and measurable where
possible?
Number of QC samples □ Are a sufficient number of QC samples used for later statistical (or other)
purposes?
Internal standards □ Are internal standards appropriate for the experimental aims?
□ Are internal standards appropriate for the chemistry of the experiment?
□ Is a written protocol for internal standards available?
□ Do the internal standards cover the m/z and retention time range of interest?
□ Have the appropriate addition points for internal standards been considered?
□ Are the internal standards in stock?

Batch length □ Will the batch length compromise the integrity of the samples?
Pilot dilution series □ Has the sample matrix behavior been tested?
□ Are the samples appropriately diluted for the study?
□ Are different classes tested separately to avoid bias?
Pilot internal standard □ Has the optimal ratio of internal standard to sample been tested?
test □ Are the correct solvents being used to dissolve the internal standards?
□ Are the best concentrations of internal standards being used?
Dilution series of QCs □ Is a dilution series of pooled QCs planned for every analytical batch?

Preanalytical Equipment condition □ Has all equipment been calibrated and properly maintained?
□ Are appropriate vials being used for the experiment (contaminants, size,
batch, etc.)?
□ Are samples kept at constant low temperature (e.g., with ice bath)?
□ What type of pipette is required for accurate pipetting?
□ Which solvents are required and at which grade?
□ Are particular solvent brands preferred?
□ Which sources and grades of water are required?

(continued)
 Quality Assurance in Metabolomics and Metabolic Profiling 41

Table 5
(continued)

Workflow stage Steps Quality assurance points

Sample extraction □ Are solvents at the appropriate temperature?

□ Are appropriate grades of solvents used?
□ Are internal standards added at the appropriate point(s)?
□ Are appropriates steps taken to completely denature enzymes in the sample?
□ Are samples kept at appropriate temperatures prior to denaturation?
Sample cleanup □ Are samples appropriately centrifuged/filtered to remove particulates?
□ Are additional sample cleanup steps (e.g., solid-phase extraction) required?
□ have the potential effects of sample cleanup on instrument behavior and/or
sample reproducibility been evaluated?
Sample drying □ Are all samples fully dried?
□ Do storage conditions require additional drying steps?
Sample extract storage □ Are storage temperatures monitored?
□ What are allowable storage times for sample extract (e.g., tissue and
metabolite considerations)?

Instrument Check instrumentation □ Is the instrument turned on and correctly connected?

setup □ Are the vacuum levels within the proper range?
□ Have the temperature(s) reached the proper set point(s)?
Instrument calibration □ Are the instrument(s) fully calibrated?
□ Are the instrument(s) ready to run?
Priming the LC □ Has the LC flowed for enough time to equilibrate the system for a reliable
SST test?
Priming the MS □ Is the tune page correct for the analytical method?
□ Is the correct polarity set?

Analysis Remove samples from - □ Have the samples warmed sufficiently to prevent condensation from
80 °C freezer affecting them?
□ Have the samples sufficiently equilibrated to operating temperature?
Reconstitution □ Have the internal standards been correctly made up and added?
□ Is the reconstituted internal standard sufficiently new?
□ Is an appropriate reconstitution solvent used?
Mobile phases □ Are the mobile phases sufficiently new?
□ Have the mobile phases been appropriately sonicated?
□ Have the mobile phases been appropriately filtered?
□ Is the pH at an appropriate level?
□ Are the solvents sufficiently protected from light?
Laboratory environment □ Is the lab room temperature within an appropriate range?
□ Is the lab relative humidity within an appropriate range?
Autosampler temperature □ Has the autosampler reached the correct set point prior to loading samples?
Chromatographic □ Is a pre-column correctly installed (if being used)?
column □ Has the column(s) been appropriately conditioned?
□ Is the column(s) being used in an appropriate temperature and pH operating
range?
□ Is the column temperature being maintained in an appropriate range?
□ Has the column been monitored for degradation?
Flow rate and gradients □ Have the flow rates and gradients been optimized for reproducibility?
□ Are the correct flow rates and gradients being applied as stipulated in the
method protocol?
Injection volume □ Are injection volumes accurate and reproducible?
□ Are injection volumes within the autosampler’s working specifications?
□ Has the injection volume been optimized for mass loaded on column, sharp
chromatography, etc.?

(continued)
42 Jennifer A. Kirwan et al.

Table 5
(continued)

Workflow stage Steps Quality assurance points

Sample sequence □ Is the correct sample run list/sequence being used?

□ Are the sample positions on the autosampler correct?
□ Are the correct instrument methods being used?
□ Is a true or solvent blank being analyzed first?
□ Are appropriate samples being used and with sufficient number of injections
to stabilize column conditions?
Real-time instrument □ Is the autosampler picking up the correct sample?
checks □ Is the LC starting and correctly timed with the MS?
□ Are all instruments functioning as anticipated?
□ Are the correct starting conditions being applied?
□ Is data being correctly collected?
□ Is data correctly stored and retrievable (e.g., sufficient disk space)?
□ Are data being checked regularly through an analytical batch?

Documentation Protocols □ Are protocols being clearly recorded?

□ Are protocols being clearly versioned and stored in a central location?
□ Are all lab members using the latest protocol version?
Lab books □ Are lab books being updated daily?
□ Are lab books being checked/signed daily?

analytical chemist is most involved (see Fig. 1). Full details of the
steps covered in the checklist, including details of key quality assur-
ance points, the reasoning behind them, and actions to be taken to
improve the steps, are provided below.

Aims and Scope

The wider aims of the study and the scope of the metabolomics
should be understood by all researchers involved in the study.
Understanding the aims of the entire study ultimately allows better
design for sample collection and of individual experiments to
ensure that the major objectives and biological questions of the
study can be answered. Discussions over appropriate controls, sam-
ple numbers, and resources should take place at this stage. Defining
the scope of the metabolomics enables experiments to be properly
designed to meet the goals of the wider study and the metabolo-
mics objectives in particular. It requires consideration that the
scope is properly resourced (including staff, finances, and instru-
mentation) and ensures that wider processes, both before and after
the metabolomics, are fully aligned with the metabolomics. Major
benefits include ensuring that experiments are designed to answer
the major objectives and are properly statistically powered. Addi-
tional advantages for metabolomics include making sure that sam-
ples are collected using a protocol suitable for metabolomics and
for the particular study or analytical needs and that the metabolo-
mics data is processed in a format that enables planned
 Quality Assurance in Metabolomics and Metabolic Profiling 43

bioinformatics to be conducted in a streamlined manner, e.g.,

normalization and data cleanup approaches are synergistic with
parallel work packages working on proteomics, genetic, or other
data sources.

Transfer
The aims of quality assurance in transfer are to ensure that the
samples are delivered in a state fit for the intended purpose; that
the number, type, and identity of samples are not in doubt; and that
any concerns about the transfer process are recorded for proper
interpretation of any data collected from these samples. In addition,
there may be a legal necessity when taking possession of physical
samples from other laboratories that tracking and ownership of
these samples is clear. Useful points to specifically check are as
follows:
(i) Number of delivered samples and that they correspond to
what was previously agreed.
(ii) That there are sufficient numbers of samples per statistical
class. Where numbers deviate from the original plan, it is
good to record in writing what the potential implications of
this are for the aims and objectives of the study and to com-
municate this to the important stakeholders before experi-
mental work starts.
(iii) Are all samples useable. Samples sometimes have less volume
than recorded, particularly after long periods of storage where
sublimation can occur. Certain sample types are also suscepti-
ble to other issues, e.g., blood products sometime have large
amounts of fibrous material which can reduce the effective
volume available for use (and may also selectively bind specific
metabolites affecting results).
(iv) Are samples traceable: each sample should ideally come with a
full sample history including temperature, storage, and other
tracking information.
(v) Have samples been entered/updated on the tracking system;
this contributes to good record keeping and ensures the integ-
rity of the tracking system in point (iv).
(vi) Have field blanks been collected; this enables the origin of
contaminants to be properly assessed.
The specific metabolomics aims of the study should have been
defined within the metabolomics scope.
44 Jennifer A. Kirwan et al.

6 Training

A period of preexperimental work often precedes the analytical

experiment. This is the time where both any training needs of
staff and any method validation work required for the study are
undertaken. The aims of method validation are to ensure that the
method is fit for purpose and works as anticipated for the matrix
that is intended to be analyzed.
Quality control deserves its own chapter. The number and
type of quality control points and samples that will be employed in
the study should be determined. Where protocols have been clearly
laid out, quality control points are often routine and stipulated in
the protocols, although extra points may be added as necessitated
by the needs of the study. Quality control samples are often also
stipulated in standard protocols, although may be adjusted as
required due to the requirements of the study. More information
on quality control samples can be found in Kirwan et al. [4]. Other
factors essential to the study can also be assessed and planned in this
time period:
(i) Whether and which internal standards are required. Select
internal standards according to needs of project and protocol
requirement. Internal standards should be appropriate to the
chemistry of the experiment and the requirements of the
experimental aims. In general, for untargeted metabolomics,
they should ideally cover the full range of the chemical space
being analyzed and the m/z range of interest and spread the
full range of the LC retention times. Consider when it is
appropriate to add them in the sample preparation protocol.
Ensure that they are in stock before the start of the experiment
and a written protocol is available for their use.
(ii) Maximum batch length. Validating the maximum batch
length ensures that the batch length does not compromise
the integrity of the samples while minimizing the number of
batches required for large-scale experiments. Optimizing this
parameter improves technical reproducibility overall. This
parameter is normally determined as part of normal method
validation.
(iii) Pilot dilution series. Where untargeted metabolomics is con-
ducted, a pilot experiment of a dilution series of pooled QCs
or an otherwise representative sample of the matrix to be
analyzed enables the sample matrix behavior to be tested.
This enables an appropriate dilution ratio to be selected for
the samples used in the real study. Ideally, each class to be
tested is run separately to ensure that results are not biased
toward individual classes. Also, an appropriate range of differ-
ent dilution ratios should be tested, with a suitable number of
 Quality Assurance in Metabolomics and Metabolic Profiling 45

replicates per dilution, per class. The optimal dilution is par-

tially dependent on the study aims, for example, if there are
metabolites of interest that must be detected. In general, the
optimal dilution ratio enables the maximum number of
detected signals at a concentration which is within the linear
range of the detector, for the greatest number of classes. This
is differentiated from the dilution series of QCs in point
(v) below, which is run for every batch to determine which
signals are within the linear range.
(iv) Pilot internal standard test. This enables the internal standards
to be added at an appropriate concentration and should be
performed for both internal standards added at the extraction
phase and for those added at the reconstitution phase. Aims
are to determine the optimal concentration of standards to
use, appropriate solvents to use to dissolve them, and the
optimal time to add them. This step may be omitted when
this has previously been done with the same matrix for the
same analytical method.
(v) Dilution series of QCs. As part of every batch, a dilution series
of pooled QCs can be analyzed to determine which signals are
potentially contaminants and which signals are within the
linear range and thus more robust for statistical analysis. This
additionally gives a clear independent assessment of the sensi-
tivity difference between batches.

7 Preanalytical

The preanalytical phase encompasses all of the processes that occur

prior to sample analysis. The sample handling procedures including
collection, denaturation, and extraction should all have been appro-
priately validated for that matrix prior to an experiment starting.
(i) Planning: planning ensures that all relevant factors have been
thought of before experimental work begins. It is essential to
good quality assurance.
(ii) Equipment condition: all equipment should be calibrated
and maintained and their status in this regard checked before
each experiment. All equipment should be regularly moni-
tored to ensure this occurs.
(iii) Equipment and conditions; consider how each individual
consumable or equipment may be affecting the technical
reproducibility or the overall validity of your results. Individ-
ual points are provided in a checklist (see Table 5); users are
encouraged to draw up their own lists suitable for their own
scientific aims.
46 Jennifer A. Kirwan et al.

(iv) Materials: Chemicals of the highest grade are normally

required. Maintaining brand loyalty allows scientists to
understand origin of contaminant signals and reduce the
risk of experiments needing to be repeated due to inferior
batches of chemicals from less reliable suppliers.
(v) Preanalytical method: Sample extraction and clean
up. Control factors include the temperature and grade of
solvent, internal standards, denaturation, and temperature
control of samples to reduce technical variation. Temperature
and grade of solvent will influence the effectiveness of
quenching and limit contamination. Internal standards
added at the extraction stage will enable the efficiency of
extraction to be determined as well as enable better normali-
zation, identification, and other quality control steps to be
applied. The denaturation step should be carefully chosen
including important parameters of the starting material,
e.g., how much water a biological tissue contains, or its
physical size or accessibility, to ensure that the denaturation
method chosen is effective for the sample type. Lastly, good
temperature control of samples during this procedure will
improve technical reproducibility.
(vi) A sample cleanup stage is not always used, but, where
required, is useful for reducing particulate matter and
unwanted matrix effects. Samples should be checked after
this step for any visible particulate matter. Coloration in the
sample can also indicate whether additional cleanup
(or dilution) is required. Sometimes, poor sample reproduc-
ibility or strange analytical instrumentation behavior can also
be a sign that this stage is not sufficient for the sample type.
(vii) A drying step is especially important for certain methods
which require derivatization. Incomplete drying both leads
to sample dilution and may directly impact the efficiency of
derivatization methods. Check samples are completely dry,
and consider repeating this step if samples are stored dry
where condensation may occur when samples are transferred
to different temperature zones, e.g., when removing from a
freezer to room temperature.
(viii) Storage temperature and time should be carefully monitored
to reduce storage degradation effects.

8 Instrument Setup

Check instrumentation: This step ensures basic functioning of

instruments before starting. Ensure all instrumentation is on and
correctly connected and vacuums and temperatures have reached
set points.
 Quality Assurance in Metabolomics and Metabolic Profiling 47

Instrument calibration should be conducted at this point to

ensure the instrument is properly functioning and calibrated, and
the LC should be started with a suitable flow rate and solvent mix,
to equilibrate it before starting LC-MS. A system suitability test
should be conducted regularly to ensure the instrument is
operating fit for the purpose, and lastly, the instrument should be
set to the starting conditions of the analytical method (LC and MS)
to ensure that both are equilibrated prior to starting.

Analysis
Analysis is a multiple step procedure which is highly dependent on
the analytical procedure to be followed. Like sample extraction, all
steps should have been previously validated to ensure that they are
fit for the purpose for the matrix of interest. Table 5 gives some
examples of steps in this procedure including:
(i) Remove samples from -80 freezer or place of storage; allow-
ing to equilibrate to working temperature will reduce the risk
of condensation affecting samples and reduce batch effects.
(ii) Reconstitution: reconstituting in an appropriate solvent mix,
with an appropriate mix of internal standards, at an appropri-
ate dilution is essential to ensuring that metabolites of inter-
est are properly and fully dissolved and can be fully analyzed
as part of the experiment.
(iii) Mobile phases should be made up freshly, filtered, and soni-
cated to reduce air or particulate matter and pH checked.
They should also be protected from light to reduce the risk of
polymerization (a particular issue with acetonitrile). Where
instruments are used for multiple analytical methods, espe-
cially where buffers are used, an extra step to ensure that
crystallization within the instrument lines does not occur if
solvents are changed out can save large amounts of time in
troubleshooting instruments, e.g., the crystallization limit of
ammonium acetate at 95% is 10 mM.
(iv) Lab temperature and humidity monitoring ensures that
instruments are operating at their optimum. Unexpected
changes in these values also enable action to be taken quickly
before too many analyses need to be repeated.
(v) Autosampler temperatures should also be monitored regu-
larly to ensure that the autosampler temperature control is
working and sample integrity can be maintained throughout
analysis. Most modern instruments have an automated mon-
itoring system for this.
(vi) Column integrity: use of a pre-column reduces long-term
costs by increasing the time between full column changes.
Both pre-column and column should be fully conditioned
48 Jennifer A. Kirwan et al.

when new to reduce batch drift. Column temperature and

use should be regularly monitored to enable preemptive
action to be taken and reduce the number of rejected
analyses.
(vii) Flow rates and gradients should be as stipulated in the
method protocol (and should have been optimized as part
of method development). Checking that the correct method
is applied is a useful point to have on a sample setup checklist
to reduce the risk of human error resulting in an entire batch
reanalysis being required.
(viii) Injection volume directly impacts the chromatography and
should be optimized as part of method development. Use the
minimum volume possible, in a resuspension solvent close to
the LC starting conditions (where possible), and ensure that
it is operating in the working range of the autosampler. Since
there is often a trade-off between solvent LC conditions and
the best conditions to fully dissolve a sample, optimizing
injection conditions is an underrated skill in method
development.
(ix) Sample run list: the sample run list should be checked in
detail that all samples are correct, their positions in the auto-
sampler are correct, and the correct LC and MS methods are
being used.
(x) To confirm the instruments and methods are running as
intended, after starting the method, the autosampler should
be checked to confirm it is correctly picking up a sample, and
the full method should be checked to see that it is fully
functioning, the methods are correct, and the data is being
correctly saved on the system. It is recommended that a true
blank or a solvent blank is thus used as the first sample in the
sequence since if there is a problem, if there is no loss of
sample, and, in addition, if there is major instrument contam-
ination, it is picked up early.
(xi) Column equilibration: columns normally show some drift in
retention time for the first few analyses. This leads to batch
drift and can be avoided by running several “sacrificial”
pooled QC samples prior to the real analysis. This column
equilibration is most successful when a complex sample simi-
lar to the matrix of interest is analyzed. Analyzing a blank or
solvent sample typically changes the column conditions
again.
(xii) The analysis should then be checked at regular intervals
throughout a batch to ensure that it is correctly proceeding
and to identify major issues early.
 Quality Assurance in Metabolomics and Metabolic Profiling 49

9 Documentation and Record Keeping

Documentation is the last major checklist item. There should be

clear planning on which documents need to be developed, updated,
and reviewed. Which records are required and their format should
ideally be planned at the start of each project. Readers are directed
to Subheading 4 for more information. Protocols should be vali-
dated and versioned and kept in a centralized place, lab books
should be updated and signed daily, and important additional
information, e.g., instrument methods, sample analysis order,
etc., should have a dedicated centralized storage place for easy
finding. Recording templates are also documents and should be
regularly versioned, updated, and reviewed alongside other docu-
mentation. A well-planned, written data storage and access policy
should be used to protect documentation and related records.

10 Templates for Tracking and Recording QA Metabolomics Activities

Disclaimer
The views expressed in this book chapter are those of the authors
and do not necessarily represent the views or policies of the
U.S. Environmental Protection Agency.

Acknowledgments

Quality management is a complex topic. The authors would like to

acknowledge and thank the many individuals from multiple disci-
plines who have contributed to our understanding on this subject.
Too many to name individually, we know who you are and are very
grateful.

References

1. International Standards Organization (2015) laboratories. https://www.iso.org/obp/ui/

2015, posting date. ISO 9000:2015 #iso:std:iso-iec:17025:ed-3:v1:en
(en) Quality management systems—Funda- 4. Kirwan JA, Gika H, Beger RD, Bearden D,
mentals and vocabulary. International Stan- Dunn WB, Goodacre R, Theodoridis G,
dards Organization. [Online] Witting M, Yu L-R, Wilson ID (2022) Quality
2. International Organization for Standardization assurance and quality control reporting in
(2015) ISO 9001: 2015 Quality management untargeted metabolic phenotyping: mQACC
systems—Requirements. https://www.iso. recommendations for analytical quality man-
org/obp/ui/#iso:std:iso:9001:ed-5:v1:en agement. Metabolomics 18:70
3. International Organization for Standardization 5. Evans AM, O’Donovan C, Playdon M,
(2017) ISO/IEC 17025 General requirements Beecher C, Beger RD, Bowden JA,
for the competence of testing and calibration Broadhurst D, Clish CB, Dasari S, Dunn WB
50 Jennifer A. Kirwan et al.

(2020) Dissemination and analysis of the qual- metrology—basic and general concepts and
ity assurance (QA) and quality control associated terms (VIM)
(QC) practices of LC–MS based untargeted 19. Raposo F, Barceló D (2021) Challenges and
metabolomics practitioners. Metabolomics strategies of matrix effects using
16:1–16 chromatography-mass spectrometry: an over-
6. Hibbert DB, Korte E-H, Örnemark U (2021) view from research versus regulatory view-
Metrological and quality concepts in analytical points. TrAC Trends Anal Chem 134:116068
chemistry (IUPAC recommendations 2021). 20. Chamberlain CA, Rubio VY, Garrett TJ (2019)
Pure Appl Chem 93:997–1048 Impact of matrix effects and ionization effi-
7. Broadhurst D, Goodacre R, Reinke SN, ciency in non-quantitative untargeted metabo-
Kuligowski J, Wilson ID, Lewis MR, Dunn lomics. Metabolomics 15:1–9
WB (2018) Guidelines and considerations for 21. Dagher G (2022) Quality matters: interna-
the use of system suitability and quality control tional standards for biobanking. Cell Prolif
samples in mass spectrometry assays applied in 55:e13282
untargeted clinical metabolomic studies. Meta- 22. González-Domı́nguez R, González-Domı́n-
bolomics 14:1–17 guez Á, Sayago A, Fernández-Recamales Á
8. Administration UFaD (2018) Bioanalytical (2020) Recommendations and best practices
method validation guidance for industry. In: for standardizing the pre-analytical processing
Center for Drug Evaluation and Research of blood and urine samples in metabolomics.
CfVM (ed) US Food and Drug Administration Meta 10:229
9. European Medicine Agency (2011). Guideline 23. Biais B, Bernillon S, Deborde C, Cabasson C,
on bioanalytical method validation. In Rolin D, Tadmor Y, Burger J, Schaffer AA,
(CHMP) CfMPfHU (ed) Moing A (2012) Precautions for harvest, sam-
10. Consortium MQAaQC 2023, posting date. pling, storage, and transport of crop plant
[Online] metabolomics samples. In: Plant metabolo-
11. Dudzik D, Barbas-Bernardos C, Garcı́a A, Bar- mics: methods and protocols, pp 51–63
bas C (2018) Quality assurance procedures for 24. Huber L (1996) Quality assurance and instru-
mass spectrometry untargeted metabolomics. mentation. Accred Qual Assur 1:24–34
A review. J Pharm Biomed Anal 147:149–173 25. Lam H (2004) Procurement, qualification, and
12. Dumas-Mallet E, Button KS, Boraud T, calibration of laboratory instruments: an
Gonon F, Munafò MR (2017) Low statistical overview. In: Analytical method validation and
power in biomedical science: a review of three instrument performance verification, pp
human research domains. R Soc Open Sci 4: 139–151
160254 26. Bedson P, Sargent M (1996) The development
13. Barraza F, Arancibia M, Madrid E, Papuzinski and application of guidance on equipment
C (2019) General concepts in biostatistics and qualification of analytical instruments. Accred
clinical epidemiology: random error and sys- Qual Assur 1:265–274
tematic error. Medwave 19 27. Grochau IH, Leal DKB, ten Caten CS (2020)
14. Manders B, de Vries H, Blind K (2013) Acad- European current landscape in laboratory
emy of management proceedings, p 12255 accreditation. Accred Qual Assur 25:303–310
15. Quality AAf 2023, posting date. What is a 28. Hunt CJ (2019) Technical considerations in
Quality Management System (QMS). [Online] the freezing, low-temperature storage and
16. Dunn WB, Broadhurst D, Begley P, Zelena E, thawing of stem cells for cellular therapies.
Francis-McIntyre S, Anderson N, Brown M, Transfus Med Hemother 46:134–150
Knowles JD, Halsall A, Haselden JN (2011) 29. Nussbeck SY, Weil P, Menzel J, Marzec B,
Procedures for large-scale metabolic profiling Lorberg K, Schwappach B (2014) The labora-
of serum and plasma using gas chromatography tory notebook in the 21st century: the elec-
and liquid chromatography coupled to mass tronic laboratory notebook would enhance
spectrometry. Nat Protoc 6:1060–1083 good scientific practice and increase research
17. Anderson HE, Santos IC, Hildenbrand ZL, productivity. EMBO Rep 15:631–634
Schug KA (2019) A review of the analytical 30. Deutsche Forschungsgemeinschaft e.V. (2022)
methods used for beer ingredient and finished 2022, posting date. Guidelines for Safeguard-
product analysis and quality control. Anal ing Good Research Practice (English language
Chim Acta 1085:1–20 translation). Deutsche Forschungsge-
18. Commission IOfSIE (2007) ISO/IEC Guide meinschaft e.V. 1.1. [Online]
99:2007(en) International voacbulary of
 Quality Assurance in Metabolomics and Metabolic Profiling 51

31. Wilkinson MD, Dumontier M, Aalbersberg IJ, community questionnaire. Metabolomics 13:
Appleton G, Axton M, Baak A, Blomberg N, 1–6
Boiten J-W, da Silva Santos LB, Bourne PE 39. Charoo NA, Khan MA, Rahman Z (2022)
(2016) The FAIR guiding principles for scien- Data integrity issues in pharmaceutical indus-
tific data management and stewardship. Sci try: common observations, challenges and
Data 3:1–9 mitigations strategies. Int J Pharm:122503
32. Kumar R, Venkatesh K (2022) Proceedings of 40. World Health Organization (2019) Guideline
international conference on deep learning. In on data integrity. WHO Drug Inf 33:773–793
Computing and Intelligence: ICDCI 2021, pp 41. Perkel JM (2019) Ways to avoid a data-storage
687–698 disaster. Nature 568:131–132
33. Hart EM, Barmby P, LeBauer D, 42. Sharma K, Shingatgeri VM, Pal S (2021) Role
Michonneau F, Mount S, Mulrooney P, of data digitization on data integrity. In: Qual-
Poisot T, Woo KH, Zimmerman NB, Hollister ity assurance implementation in research labs,
JW (2016) Ten simple rules for digital data pp 221–245
storage, vol vol. 12. Public Library of Science,
San Francisco, p e1005097 43. Fleming IO, Garratt C, Guha R, Desai J,
Chaubey S, Wang Y, Leonard S, Kunst G
34. Gika HG, Theodoridis GA, Wilson ID (2008) (2016) Aggregation of marginal gains in car-
Liquid chromatography and ultra-performance diac surgery: feasibility of a perioperative care
liquid chromatography–mass spectrometry fin- bundle for enhanced recovery in cardiac surgi-
gerprinting of human urine: sample stability cal patients. J Cardiothorac Vasc Anesth 30:
under different handling and storage condi- 665–670
tions for metabonomics studies. J Chromatogr
A 1189:314–322 44. Tan WS, Tan MY, Lamb BW, Sridhar A,
Mohammed A, Baker H, Nathan S, Briggs T,
35. Gika HG, Theodoridis GA, Wingate JE, Wil- Tan M, Kelly JD (2018) Intracorporeal robot-
son ID (2007) Within-day reproducibility of an assisted radical cystectomy, together with an
HPLC- MS-based method for metabonomic enhanced recovery programme, improves post-
analysis: application to human urine. J Prote- operative outcomes by aggregating marginal
ome Res 6:3291–3303 gains. BJU Int 121:632–639
36. Yin P, Lehmann R, Xu G (2015) Effects of 45. Gilbert D, Henry A, Choudhury A (2022)
pre-analytical processes on blood samples used Marginal gains and clinical trials–improving
in metabolomics studies. Anal Bioanal Chem and influencing practice. Clin Oncol 34:419
407:4879–4892
46. Garcı́a-Rivera MA, Fernández-Ochoa Á,
37. Gil A, Siegel D, Permentier H, Reijngoud DJ, Brüning U, Fritsche-Guenther R, Kirwan JA
Dekker F, Bischoff R (2015) Stability of energy (2022) Identification and validation of small
metabolites—an often overlooked issue in molecule analytes in mouse plasma by liquid
metabolomics studies: a review. Electrophore- chromatography–tandem mass spectrometry:
sis 36:2156–2169 a case study of misidentification of a short-
38. Dunn WB, Broadhurst DI, Edison A, chain fatty acid with a ketone body. Talanta
Guillou C, Viant MR, Bearden DW, Beger 242:123298
RD (2017) Quality assurance and quality con-
trol processes: summary of a metabolomics
 Chapter 3

Quality Control and Validation Issues in LC-MS-Based

Metabolomics
Olga Begou, Helen G. Gika, Georgios Theodoridis, and Ian D. Wilson

Abstract
Metabolic profiling performed using untargeted metabolomics of different, complex biological samples
aims to apply agnostic/holistic, hypothesis-free, analysis of the small molecules that are present in the
analyzed sample. This approach has been the center of major investments and dedicated efforts from the
research community for many years. However, limitations and challenges remain, particularly with regard to
the validation and the quality of the obtained results. This has led to increasing community engagement,
with the formation of think tanks, the establishment of working groups, and the many seminars on quality
control (QC) in metabolomics. Here we describe a quality control (QC) protocol used to monitor LC-MS-
based metabolomics analysis. A key target is the monitoring of analytical precision. This methodology is
described for the analysis of urine but can be applied to different biological matrices, such as various
biofluids, cell, and tissue extracts.

Key words Quality control, Untargeted metabolomics, Biological samples, Validation, LC-MS

1 Introduction

Untargeted metabolomics represents an expanding research disci-

pline, dealing with the holistic analysis of small molecule metabo-
lites. Typically, metabolomics analysis deals with compounds of
molecular weight (MW) < 1500 Da. Application of such holistic
analysis can offer a snapshot of the ongoing biochemistry and
phenotype (metabotype) and can reveal and monitor end-points
and changes occurring in living systems as a result of biological
stimuli or genetic manipulation [1].
The field has shown much growth in the last decade, mainly as a
result of technological advances in (high-resolution) mass spec-
trometry (MS) and NMR spectroscopy [3] (mainly H1 NMR).
Chromatographic techniques are also central to the field being
coupled to MS with both liquid and gas chromatography (LC and
GC) representing major technologies for metabolomics [3–5]. This

Olga Deda et al. (eds.), Metabolic Profiling: Methods and Protocols, Methods in Molecular Biology, vol. 2891,
https://doi.org/10.1007/978-1-0716-4334-1_3,
© The Author(s), under exclusive license to Springer Science+Business Media, LLC, part of Springer Nature 2025

53
54 Olga Begou et al.

application of high-resolution spectroscopic modes results in the

generation of massive datasets; these are subsequently treated with
advanced, sophisticated, multivariate statistical analysis tools. The
combination of the abovementioned analytical techniques facili-
tates the simultaneous detection and qualitative/relative measure-
ment of many hundreds/thousands of molecules, which are
typically present in samples of, e.g., blood, urine, cells, different
tissues, and so forth [2, 3]. This type of comprehensive analysis is
not yet free of problems. Despite the impressive steps forward in
technology, the simultaneous detection and relative quantification
of hundreds of molecules of diverse physicochemical properties,
chemical classes, and concentration ranges remains a challenging
topic [4–7].
Metabolomic research is typically categorized as untargeted or
targeted approaches. Targeted methods fall into two main classes,
either covering a moderate number of analytes (ca. 10–150) for
which concentrations are obtained or alternatively simply covering
a large number of known metabolites for which standards are
available. Often targeted assays are employed in a hypothesis-driven
study, as the analysis is focused on the measurement of preselected
molecules (as described in more detail in Chaps. 9 and 10 of the
present book). Targeted analyses have been the topic of research
and development from various standpoints for decades. So in these
efforts, key validation points have been well addressed and proto-
cols and guidelines have been produced. These document how new
methods should determine report analytical figures of merit, such
as repeatability, analytical accuracy, and precision as well as impor-
tant features such as stability, storage, etc. (for a review on targeted
metabolomics see Ref. [8]). Therefore, the validation of targeted
methods, although not easy or trivial, can be based on existing
guidelines and, thus, can be considered easier to address, compared
to untargeted methods.
In contrast, untargeted metabolomics, by its very nature, aims
toward achieving holistic and unbiased analysis of the sample of
interest. The scope is to allow for the discovery of new, unexpected
biomarkers [9] and to gain mechanistic information on the bio-
chemistry of the system under study. In these settings, validation
and quality assurance are significantly harder as the analyst does not
follow specific signals (or molecules) [3, 10]. Hyphenated techni-
ques, such as LC-HRMS, generate high-dimensional data (3D or
4D). Such data require special software and extensive, multifaceted,
complex data processing. As a result and despite the wide applica-
bility, validation strategies for global metabolic profiling are not
advanced to the same level as those applied to targeted analysis or
bioanalysis [11].
Over the last two decades “partial” solution to the QA/QC,
this problem has been developed via the use of sample-based
“quality control” (QC) samples to evaluate analytical system
Quality Control and Validation Issues in LC-MS-Based Metabolomics 55

stability and analytical precision [12–15]. Analysis of QC samples

can offer information about the performance of the analytical sys-
tem and, by doing so, can reinforce confidence in the quality of
acquired data. QC samples are typically used for the equilibration of
the analytical system (where they can also provide a system suitabil-
ity test), for monitoring analytical signals, and to facilitate the
monitoring and calculation of intra/inter day precision. QC sam-
ples have also been used to allow for signal correction (normaliza-
tion), e.g., to correct for batch and run-order effect and also for
method standardization [12, 14, 15]. QC data have been used as
quantitative indicators of random errors or fluctuations during the
analytical run, which in large-scale facilities is even performed
online during data acquisition. Hence, the analysis of QC samples
(especially in LC-MS-based metabolomics) has a greater value than
just the evaluation of chromatographic and mass spectrometry
setting.
QC samples are typically prepared by mixing small, equal ali-
quots from the real test samples; in this way, the resulting pool
sample is expected to contain all molecules present in the real
samples at an averaged content. This is an easy, convenient
approach that can be applied during sample handling at a minimal
cost in time/resources and can function effectively for small/
medium studies of up to ca 300 samples. For larger-scale studies,
such schemes may become impractical or simply not possible; for
practical reasons a bulk sample of the same matrix may be substi-
tuted [16, 17]. In some circumstances, sometimes both a bulk
matrix and the study pool samples are used, with the former used
to enable analyses to be compared over long time periods/different
batches [18]. The utility of this concept, or its variants, has gained
acceptance and is now widely applied in metabolomics-based stud-
ies for a variety of specimens matrices, liquid (such as urine, plasma,
serum), or cells and tissues.
The needs for dissemination and wider use of QC measures in
untargeted metabolomics has recently led to new initiatives, such as
the formation of mQACC (The metabolomics Quality Assurance
and quality Control Consortium), which aims toward the develop-
ment and dissemination of collaborative efforts among researchers
and stakeholders in academia, industry, and government to address
key quality assurance (QA) and quality control issues in untargeted
metabolomics. The consortium has published a series of papers
advocating the use of QC strategies in metabolomics, including
but not limited to reporting minimum standards [17], use of
reference materials for QC [18], and other aspects of dissemination
of the need for QC of metabolomics [19, 20]. In the pursuit of
better and more trustworthy metabolomics research, important
aspects include the active application of the FAIR (findability,
accessibility, interoperability, and reusability) principles [21]. In
the same line of thought, an important area for further
56 Olga Begou et al.

improvement of QC in untargeted metabolomics is metabolite

identification, where reporting of quality data remains a challenge.
Despite the publication of reporting guidelines, common practice is
publishing without reporting the necessary level of identification
and often without the necessary proof [23, 24]. This can result in
erroneous identification [23], leading to incorrect assumptions on
the underlying biochemistry. Improving the application of QC
measures is thus of paramount importance in untargeted
metabolomics.
In the present protocol, we describe the use of QC samples
during the analytical procedure involved in LC-MS-based global
metabolic profiling of biological samples and the application of the
resulting data to monitor the systems performance. However, the
ideas covered here could also be applied for examining the data
obtained when using many other types of chromatography in
untargeted metabolomics.

2 Materials

All solvents (methanol, acetonitrile, formic acid) used should be of

LC-MS analytical grade. Water should be of Millipore quality
(18.0 MΩ, at 25  C). Standards should be of analytical or higher
grade. All standards should be stored at 20  C or 80  C (see
Note 1).

2.1 Stock, Working, Stock solutions of all metabolites should be prepared at a concen-
and Calibration tration of 1 mg/mL (or appropriate concentration depending on
Solutions analyte solubility) in methanol or mixtures of methanol with water.
Working standards at concentration of 10 μg/mL are prepared
from the stock solution by appropriate dilution. From the stock
solutions of the standards, prepare a mixture (concentration
1–5 μg/mL) of the compounds of interest by mixing appropriate
volumes.
All solutions should be stored at 20  C.

2.2 Mobile Phase 1. Mobile Phase A: Water +0.1 vol. % formic acid: add 100 μL of
formic acid to 1 L of Millipore water.
2. Mobile Phase B: Acetonitrile +0.1 vol. % formic acid: add
100 μL of formic acid to 1 L of LC-MS grade acetonitrile. In
the case of plasma/serum samples, use methanol in mobile
phase B instead of acetonitrile, or a mixture of methanol with
acetonitrile.
3. Wash solvent: Use a “strong” solvent containing acetonitrile/
water, 80:20 v/v for post injection cleaning cycles and a weak
solvent of water/acetonitrile, 80:20 v/v for pre-injection
washes (see Note 2).
 Quality Control and Validation Issues in LC-MS-Based Metabolomics 57

2.3 Chromatographic Chromatographic analysis can be performed on a HSS T3 C18

Materials and column (Waters, 2.1 mM  150 mM, < 2 μm i.d.) or similar type
Instrumentation of column. Online filters and pre-columns/guard columns of the
same material as the analytical column should be also used.
An ultra-high performance liquid chromatography (U
(H)PLC) system coupled to a high-resolution mass spectrometer
with an ESI source can be used.

2.4 Software Appropriate software includes the following:

Data acquisition and processing software (e.g., Excalibur, Mas-
sLynx, Analyst or other depending on the MS instrument of
choice). Special software for peak-picking, such as Marker Lynx,
Sieve, MarkerView (or other vendor software), or open source/free
software (XCMS, MzMine, MSdial or other). Microsoft Excel,
Statistica, and other advanced spreadsheet software. SIMCA-P or
other software for multivariate statistical analysis. MATLAB,
Python, R programming language, or associated software packages.

3 Methods

3.1 Analytical 1. In the beginning, analysts must make sure that the chro-
System Preparation matographic system and the mass spectrometer are in suitable
condition for the analysis. Necessary steps: Check the mass
accuracy of the mass spectrometer; if necessary, recalibrate the
mass spectrometer following the appropriate procedure recom-
mended by the manufacturer to achieve maximum mass accu-
racy and resolution. It is advisable that the calibration
procedure should take place every 3–4 months and that the
temperature of the laboratory should not vary significantly.
Detection sensitivity should also be checked comparing analyt-
ical signals with signals obtained from the analysis of a known
sample (system suitability test).
2. Load the required liquid chromatographic method. This would
include adjusting parameters, such as flow rate, column and
autosampler’s temperature, wash cycles, and gradient elution
program.
3. Allow the system to equilibrate and run a no-injection gradient
in order to aid column equilibration.
4. Depending on the matrix of the samples being analyzed, run a
suitable number of pooled QC/bulk matrix samples at the
beginning of the analytical batch to achieve system stability. It
has been observed that for urine analysis 5 QC injections are
needed; for serum or plasma extract this may need to be
increased to 10–20 injections (see Note 3 and 4).
58 Olga Begou et al.

5. The initial 5–10 injections of the pooled QC that are run prior
to the commencement of the analysis of the batch in order to
equilibrate the system also represent a useful “system suitability
test” (SST). The data from these samples should be examined
to confirm that peak shapes are acceptable and that the
expected number of peaks for a sample of that type of matrix
are being seen over the expected retention time range. Evi-
dence of aberrant behavior for the conditioning QCs should
result in the analytical batch not being run until the problem
has been corrected.

3.2 Sample Handling 1. It is advisable to aliquot all test samples on collection into an
and QC Preparation appropriate number of sub-aliquots for later storage and in
order to avoid unnecessary freeze/thaw cycles.
2. Ideally, samples should be stored in freezers as soon as possible
after sampling/collection, and at the lowest available tempera-
ture, at least at 20  C but preferably at 80  C. If possible,
store sub-aliquots in different freezers, in case a freezer mal-
function takes place.
3. Make sure all samples, stock, and working solutions are cor-
rectly and fully labeled. Thaw stock/working solutions and real
samples shortly before use and sample preparation, respectively.
Use appropriate labels that are not affected by freeze-thaw
cycles.

3.3 Sample 1. Vigorously vortex every sample, after thawing, prior to sample
Preparation preparation.
2. Transfer an appropriate volume (e.g., 50–200 μL) of each
sample into a 1.5 mL Eppendorf tube.
3. Depending on the matrix (urine, plasma, serum), a different
sample preparation procedure is recommended (see Note 5 and
6), as explained below in Subheadings 3.3.1, 3.3.2, and 3.3.3.

3.3.1 For Urine Samples 1. For human urine, dilute the sample in ratio 1:3 v/v with
ultrapure water. For animal (e.g., rodent), urine dilution with
methanol rather than water may be required to precipitate
protein.
2. Vortex for 5 min and centrifuge at 12,000 g or higher speeds
for 5–10 min at 4  C.
3. Transfer the clear supernatant into an LC-MS vial (n.b. 96 well
plates or similar can be substituted for LC-MS vials).
4. Place the vial (well plate) into a precooled (4  C) autosampler
for analysis. If analysis is not being undertaken, immediately
store the vial (well plate) in a fridge (0–4  C) if analysis will start
shortly or in a freezer ( 20/80  C) if analysis will be delayed
for longer than a few hours.
 Quality Control and Validation Issues in LC-MS-Based Metabolomics 59

3.3.2 For Plasma or 1. Add three times the sample volume of ice-cold methanol or
Serum Samples acetonitrile for protein precipitation. (see Note 7).
2. Vortex for 5 min and centrifuge at 12,000 g or higher speeds
for 5–10 min.
3. Transfer the clear supernatant into Eppendorf tubes.
4. Evaporate to dryness and reconstitute with water/acetonitrile,
95:5 v/v to the initial volume.
5. Repeat step 2.
6. Transfer the clear supernatant into an LC-MS vial (well plate).
7. Place the vial (well plate) into a precooled (4  C) autosampler
for analysis. If analysis is not taking place immediately, follow
the procedure described for urine in Subheading 3.3.1, Step 4.
When taking plasma/serum samples out of the freezer, it is
advisable to centrifuge samples again before analysis.

3.3.3 For Quality Control 1. Create a pool sample by mixing equal volumes of each sample
Samples (QC) analyzed. Make an appropriate number of aliquots (a minimum
of 1 QC for each 10 samples to be analyzed and possibly up to
1 for every 5, see below) and store them at a freezer. If there are
clear groups (e.g., test/patient and control), it may also be
worth preparing “phenotypic” QC s made up of samples
from these groups. Such phenotypic QCs can then be analyzed
alongside the pooled QC to highlight any differences in analyt-
ical properties due to differences in sample composition.
2. Handle each QC as a real sample and follow the sample prepa-
ration procedure developed for the type of matrix to be
analyzed.
3. If the number of the samples being analyzed is large (e.g.,
>1000) and the creation of a pooled sample from all the sam-
ples is impractical, consider making QC samples from only the
samples contained within each analytical run to provide a
within-batch QC. Have a separate “bulk matrix” QC sample
also analyzed within every batch to enable between batch com-
parisons and normalization [16–18].
4. If the acquired samples are only available in small amounts,
such that making a pooled QC of any sort is impractical, use a
bulk sample of the same matrix (ideally from the same species).

3.3.4 For Test Mix 1. Spike a QC sample with an appropriate volume of the metha-
Sample nolic mixture containing appropriate standard metabolites as
mentioned in Subheading 2.1, in order to achieve a final con-
centration of 5–10 μg/mL. It is advisable to perform this
procedure by evaporating the methanolic stock solution and
then reconstituting by dissolution in the QC matrix.
60 Olga Begou et al.

3.4 Sequence 1. Make sure that the order of the samples is randomized, to avoid
Preparation introducing bias due to changes with individual analytical runs
and between run “batch” effects. Randomization of large
sample-sets can be performed using the specific commands in
spreadsheet programs, such as Microsoft excel. QC samples
should not be randomized but inserted regularly in the run
sequence (e.g., one QC sample every 5–10 real samples).
2. The number of QC samples placed in a sequence depends on
the total number of samples analyzed and on the duration of
the analysis. For a small number of samples (<100), one QC
every five sample injections is recommended. For larger batches
of samples (100 to several thousand), the number QC samples
should represent at least 10% of the total real samples analyzed.
3. After column equilibration, in the middle and at the end of
the sequence, insert in the analytical sequence injections of
the spiked test mix solution. Injections of standards solutions
or blanks should be avoided within the batch of test samples,
as the system can be disequilibrated by the injection of
non-matrix solutions (see Notes 3 and 8). Similarly, blank
sample injections should be avoided in the middle of the ana-
lytical batch.
4. A good way to improve the relative quantification aspects in
such analysis is to implement a sequence of injection of serial
dilutions of the QC sample. This enables the detection of
saturating peaks, which would be better detected in dilute
samples (e.g., 1:10 v/v). At the same time, this series of ana-
lyses facilitates the study of the response of non-saturating
peaks in response to dilution. The dilution series can be run
at the end of each batch of samples.

3.5 Data Analysis In untargeted metabolomics-based studies, data analysis strategies

require certain steps involving raw data acquisition, normalization,
scaling, feature and peak detection before multivariate/univariate
statistical analysis, biomarker identification, and biochemical
interpretation.

3.5.1 Data Processing From the raw data acquired, observe the peak width, retention
time, peak intensity/signal, mass accuracy, and noise intensity
from indicative chromatographic peaks located at the beginning,
middle, and end of the chromatogram. These can help researchers
to define carefully the optimum parameters for the software
(XCMS, MassLynx etc.) to be used for peak alignment, peak peak-
ing, and integration.
Data from the QC samples are then used to evaluate the quality
of analysis via the following steps: Treat QC samples as a separate
group and process them alone using exactly the same parameters
 Quality Control and Validation Issues in LC-MS-Based Metabolomics 61

selected for processing of the whole sample set. Peaks that appear in
less than 70 or 80% of the QC samples should be omitted from the
dataset. Apply Principal Component Analysis (PCA), using
SIMCA-P or other statistical software, to the processed data from
the QC samples in order to observe any trends, such as time-related
drift during the analytical run of these specific samples.
More details on data preprocessing can be found in Chaps. 4
and 5 of this book.

3.5.2 Data Quality For data quality evaluation, there are some steps that should be
Evaluation taken into consideration. It is advisable to subject the data to the
test against preset criteria of quality, as explained below.
Firstly, assess the performance of the analytical run by inspect-
ing an overlay of full scan chromatograms of all the replicates of the
test mix samples, placed at the beginning, middle, and end of the
sequence. Calculate chromatographic data, such as the retention
time (tR), peak area, and height and determine the precision (RSD
%) of each value. The acceptable limits should be less than 2% for
retention time variation and less than 20–30% for peak area varia-
tion. In cases where these conditions are not met, reexamine the
data carefully to find any trend that might indicate system under-
performance or malfunction. MS-derived data should also be
acceptable in terms of mass accuracy.
The next step is to assess the data acquired from QC samples
placed at regular intervals among the samples, in order to
strengthen and corroborate the initial assessment of the perfor-
mance of the analytical run. Evaluate the full scan chromatograms
of all QC samples. It is expected that all QC samples show good
consistency, reproducibility, and small variability with respect to tR,
peak area, and height, with the possible exception of the first
5 to 10 injections (conditioning QCs). An overlay of full scan
chromatograms can provide only an estimation of analytical repeat-
ability and system stability. A thorough check should be done on
extracted ion chromatograms of all peaks along the whole length of
the analysis. Precision of the tR, the mass detected, and the signal
intensity should be unchanged across the chromatogram and
should not show, e.g., time-related effects. These can be assessed
against the preset criteria of, e.g., RSD% should be less than 2% for
tR, less than 25% for the area of abundant peaks, and less than 30%
of low signal peaks), as mentioned above. Peaks that fail such
criteria should be excluded from the dataset (including data from
the test samples). Evidence of poor repeatability for the QC sam-
ples should be taken as a possible reason for not accepting a run. A
recommended scheme of QC in the analytical data acquired by
untargeted profiling is shown in Fig. 1.
If there are any problems concerning the reproducibility of the
data, try to estimate where the inconsistency is appearing. If it is
62 Olga Begou et al.

Fig. 1 Suggested flow chart of QC in untargeted mass spectrometry-based metabolic profiling based on QC
pooled samples analysis [17]

located in a specific time window, investigate if it is justifiable to

exclude this window in the peak picking and alignment process. If
the data is poor and fails the acceptance criteria, work to find the
reason of the failure and repeat the analysis.
Providing that the run passes the predetermined acceptance
criteria, perform PCA analysis on all samples and QC replicates, in
order to observe variations in the acquired data. PCA will immedi-
ately show similarities or differences among real and QC samples.
The latter should ideally form a tight cluster. For untargeted data,
investigate the effect of different scaling modes. Applying pareto
scaling and different visualization tests helps in identifying potential
underlying issues. The following steps are suggested:
1. Color QC and real samples according to analytical run order
and try to identify time trends in the score plots.
 Quality Control and Validation Issues in LC-MS-Based Metabolomics 63

2. Check for any QC outliers or for batch effect due to run order.
3. Examine the QC data for any unexpected trends. If such trends
are found, study additional visualization plots, such as time
series plots, and see for samples that are out of 2 SD or 3 SD
thresholds. Test samples appearing as such “outliers” should
not be discarded outright but rather be thoroughly scrutinized.
QC samples out of 2 SD indicate that the data from the
neighboring test samples should be thoroughly scrutinized
and probably discarded with the samples being reanalyzed.
Thereinafter, in the exported peak table data, look for any
similarities in the pairs (tR and mass) among the peaks reported,
in order to assess the success of peak alignment. Then, using
appropriate statistical software, create graphs with the data
exported from the QC samples to evaluate the analytical perfor-
mance and precision.
Finally, when preparing publications/reports where QCs have
been used, ensure that all relevant information is provided (guide-
lines and proformas for this are given [17]).
Figure 2 presents a schematic of the proposed validation
scheme. It starts with the QC and sequence preparation, by initially
mixing equal volumes of all the samples to be analyzed (formed
form either urine or plasma/serum) in a vial and then injecting the
QC sample at the beginning of the analytical batch (system equili-
bration) and then periodically, e.g., every 5–10 real samples during
analysis. From the resulting dataset, an extracted-ion chromato-
gram (XIC) of all QC samples can be created, via data processing or
data-mining, where data are collected only for specific m/z values
of the compounds of interest. Those results can also be used for the
formation of an analysis chart, showing total sequence order (both
QC and real samples). Data processing is complemented by statis-
tical analysis. Repeatability and run order trends can be assessed by
evaluating %RSD values of QC’s ion intensities. Features presented
in most samples (~70%) with %RSD values <30% represent a good
dataset on which to carry out further work. Finally, PCA score plots
can be created from the examined results, providing information
for the accuracy and precision of the analysis, as well as, for any time
trends according order analysis of both QC and real samples.

4 Notes

1. The storage period for each standard at the freezer is dependent

on the nature of the analyte and the solvent used for dissolu-
tion. Standard solutions should be labeled with an expiry date.
2. Make sure mobile phases are filtered and sonicated to degas
them properly before use.
64 Olga Begou et al.

Fig. 2 Schematic representation of the acquisition and acceptance procedures followed during and after
analytical batch is analyzed. The process begins with quality control (QC) samples and sequence preparation
(equal volumes of all real samples analyzed are mixed to provide the pooled QC). The QC sample is then
injected 5–10 times at the beginning of the analytical batch, for system equilibration, and typically every 5–10
real samples, throughout the analytical batch(s). From the resultant dataset, an extracted-ion chromatogram
(XIC) can be generated for all QC samples for selected m/z values, where the peak width, retention time, peak
intensity/signal, and mass accuracy can be observed. The final step is the statistical evaluation of the results.
To assess repeatability and trends in the run order, %RSD values of QC’s peak areas are evaluated. A robust
dataset is indicated by features present in most samples (> 70%) with %RSD values lower than 30%.
Additionally, a PCA score plot can be generated, in order to provide insights regarding the robustness of the
analysis. QCs should cluster together

3. Avoid making injections with standard solutions during the

sequence so that the system is not de-conditioned.
4. Always monitor and record the exact column pressure (bar or
psi) at the beginning and end of a run. Rising pressure can be
an indication that the column is becoming blocked and so
provides an indication of the need for either replacement or
cleaning.
5. Apply proper cleanup steps during sample preparation. Filter-
ing samples improves the medium /long-term integrity of the
separation system and as a result may safeguard high quality
results if necessary.
6. During sample preparation, all necessary safety precautions
must be taken. Make sure gloves and goggles are used and
that samples are prepared under fume hoods.
 Quality Control and Validation Issues in LC-MS-Based Metabolomics 65

7. It has been noticed that ice-cold solvents (methanol, acetoni-

trile) at three times the sample volume help to provide better
protein precipitation.
8. In the case of large numbers of samples being analyzed, divide
them into randomized batches. Carefully monitor the system
performance during each batch and have appropriate cleaning
cycles in between (column, cone, source, etc.). Before starting
a new sequence, make sure that the system is re-equilibrated.
9. Bespoke scripts, data preprocessing, and quality control in
metabolomics datasets have also been developed by indepen-
dent groups, such as the National Phenome Center UK; an
exemplar report, tools and guidelines can be found in
reference [25].

Acknowledgments

This work was financially supported by the project “Biomic_AUTh,

Center of Excellence in Metabolomics research (BiACEM)” (Proj-
ect number: 101079370), which is implemented under the Action
“Horizon Europe Twinning grant.”

References
1. Gika HG, Wilson ID, Theodoridis GA (2019) spectrometry-based metabolic phenotyping
Omics | metabolomics: an analytical (metabolomics/metabonomics) studies in
perspective. In: Encyclopedia of analytical sci- publications. J Chromatogr B 1164:122515
ence. https://doi.org/10.1016/B978-0-12- 7. Gika HG, Theodoridis GA, Earll M et al
409547-2.14003-X (2010) Does the mass spectrometer define the
2. Theodoridis GA, Gika HG, Plumb R, Wilson marker? A comparison of global metabolite
ID (2019) Liquid chromatographic methods profiling data generated simultaneously via
combined with mass spectrometry in UPLC-MS on two different mass spectro-
metabolomics. In: Proteomic and metabolo- meters. Anal Chem 82:8226–8234. https://
mic approaches to biomarker discovery, pp doi.org/10.1021/ac1016612
149–169. https://doi.org/10.1016/B978-0- 8. Rainville PD, Theodoridis G, Plumb RS, Wil-
12-818607-7.00009-8 son ID (2014) Advances in liquid chromatog-
3. Wilson I (2016) Methods and techniques for raphy coupled to mass spectrometry for
metabolic phenotyping. Bioanalysis 9:1–3. metabolic phenotyping. TrAC Trends Anal
https://doi.org/10.4155/bio-2016-4985 Chem 61:181–191. https://doi.org/10.
4. Gika HG, Wilson ID, Theodoridis GA (2014) 1016/j.trac.2014.06.005
LC–MS-based holistic metabolic profiling. 9. Begou O, Gika HG, Wilson ID et al (2017)
Problems, limitations, advantages, and future HyphenatedMS–based targeted approaches in
perspectives. J Chromatogr B 966:1–6. metabolomics. Analyst
https://doi.org/10.1016/j.jchromb.2014. 10. Gika HG, Theodoridis GA, Earll M, Wilson ID
01.054 (2012) A QC approach to the determination of
5. Gika H, Virgiliou C, Theodoridis G, Plumb day-to-day reproducibility and robustness of
RS, Wilson ID (2019) Untargeted LC/MS- LC-MS methods for global metabolite
based metabolic phenotyping metabonomics/ profiling in metabonomics/metabolomics.
metabolomics: the state of the art. J Chroma- Bioanalysis 4:2239–2247. https://doi.org/
togr B 1117:136–147 10.4155/bio.12.212
6. Wilson ID, Theodoridis G, Virgiliou C (2021) 11. Gika HG, Zisi C, Theodoridis G et al (2016)
A perspective on the standards describing mass Protocol for quality control in metabolic
66 Olga Begou et al.

profiling of biological fluids by U(H)PLC-MS. untargeted metabolomics and lipidomics: a

J Chromatogr B Analyt Technol Biomed Life review by the metabolomics quality assurance
Sci 1008:15–25. https://doi.org/10.1016/j. and quality control consortium (mQACC).
jchromb.2015.10.045 Metabolomics 18:24
12. Want EJ, Wilson ID, Gika H et al (2010) 19. Evans AM, O’Donovan C, Playdon M (2020)
Global metabolic profiling procedures for Dissemination and analysis of the quality assur-
urine using UPLC-MS. Nat Protoc 5:1005– ance (QA) and quality control (QC) practices
1018. https://doi.org/10.1038/nprot. of LC-MS based untargeted metabolomics
2010.50 practitioners. Metabolomics 16:113
13. Gika HG, Theodoridis GA, Wingate JE et al 20. Beger RD, Dunn WB, Bandukwala A,
(2007) Within-day reproducibility of an Bethan B, Broadhurst D, Clish CB, Dasari S,
HPLC-MS-based method for metabonomic Derr L, Evans A, Fischer S, Flynn T,
analysis: application to human urine. J Prote- Hartung T, Herrington D, Higashi R, Hsu
ome Res 6:3291–3303. https://doi.org/10. P-C, Jones C, Kachman M, Karuso H,
1021/pr070183p Kruppa G, Lippa K, Maruvada P, Mosley J,
14. Zelena E, Dunn WB, Broadhurst D et al Ntai I, O’Donovan C, Playdon M, Raftery D,
(2009) Development of a robust and repeat- Shaughnessy D, Souza A, Spaeder T, Salholz B,
able UPLC MS method for the long-term Tayyari F, Ubhi B, Verma M, Walk T, Wilson I,
metabolomic study of human serum. Anal Witkin K, Bearden DW, Zanetti KA (2019)
Chem 81:1357–1364. https://doi.org/10. Towards quality assurance and quality control
1021/ac8019366 in untargeted metabolomics studies. Metabo-
15. Dunn WB, Broadhurst D, Begley P, Human lomics 15(1):4
Serum Metabolome (HUSERMET) Consor- 21. Savoi S, Arapitsas P, Duchêne É et al (2021)
tium (2011) Procedures for large-scale meta- Grapevine and wine metabolomics-based
bolic profiling of serum and plasma using gas guidelines for fair data and metadata manage-
chromatography and liquid chromatography ment. Meta 11:757
coupled to mass spectrometry. Nat Protoc 6: 22. Theodoridis G, Gika H, Raftery D et al (2023)
1060–1083. https://doi.org/10.1038/nprot. Ensuring fact-based metabolite identification
2011.335 in liquid chromatography-mass spectrometry-
16. Lewis MR, Pearce JTM, Spagou K et al (2016) based metabolomics. Anal Chem 95:3909–
Development and application of ultra- 3916
performance liquid chromatography-TOF MS 23. Kodra D, Pousinis P, Vorkas PA et al (2022) Is
for precision large scale urinary metabolic phe- current practice adhering to guidelines pro-
notyping. Anal Chem 88:9004–9013. https:// posed for metabolite identification in LC-MS
doi.org/10.1021/acs.analchem.6b01481 untargeted metabolomics? A meta-analysis of
17. Kirwan JA, Gika H, Beger RD et al (2022) the literature. J Proteome Res 21:590–598
Quality assurance and quality control reporting 24. Lewis M, Chekmeneva E, Camuzeaux S,
in untargeted metabolic phenotyping: Sands C, Yuen A, David M et al (2022) An
mQACC recommendations for analytical qual- open platform for large scale LC-MS-based
ity management. Metabolomics 18:70 metabolomics. ChemRxiv. Cambridge Open
18. Lippa KA, Aristizabal-Henao JJ, Beger RD et al Engage, Cambridge
(2022) Reference materials for MS-based
 Chapter 4

Bio- and Chemoinformatic Approaches for Metabolomics

Data Analysis
Michael Witting and Johannes Rainer

Abstract
Metabolomics data analysis includes, next to the preprocessing, several additional repetitive tasks that can
however be heavily dataset dependent or experiment setup specific due to the vast heterogeneity in
instrumentation, protocols, or also compounds/samples that are being measured. To address this, various
toolboxes and software packages in Python or R have been and are being developed providing researchers
and analysts with bioinformatic/chemoinformatic tools to create their own workflows tailored toward their
specific needs. This chapter presents tools and example workflows for common tasks focusing on the
functionality provided by R packages developed as part of the RforMassSpectrometry initiative. These
tasks include, among others, examples to work with chemical formulae, handle and process mass spectrom-
etry data, or calculate similarities between fragment spectra.

Key words R, RforMassSpectrometry, Formula handling, Mass spectra handling, Spectra similarity
calculation

1 Introduction

Non-targeted metabolomics data analysis is very diverse in its

nature and ranges from primary data preprocessing, e.g., peak
picking, retention time alignment, and normalization, to biological
interpretation of obtained results. Some of these tasks are highly
repetitive, e.g., when preparing data or libraries for metabolite
annotation. However, different steps in data preparation or anno-
tation can be performed in an automated manner using scripting
languages like Perl, Python, or R. These languages often require
only a minimal understanding of programming to perform basic
tasks of data manipulation. R is a prominent example, extensively
used in metabolomics, e.g., the preprocessing package XCMS was
developed in R [1, 2] (see Chapter 5).
Automated data preparation should not only focus on the
initial steps of preprocessing but should also be included in later

Olga Deda et al. (eds.), Metabolic Profiling: Methods and Protocols, Methods in Molecular Biology, vol. 2891,
https://doi.org/10.1007/978-1-0716-4334-1_4,
© The Author(s), under exclusive license to Springer Science+Business Media, LLC, part of Springer Nature 2025

67
68 Michael Witting and Johannes Rainer

stages of an analysis, especially for time-consuming, repetitive tasks.

In case of metabolite database creation for annotation purposes,
this can include curation of metabolite lists and calculation of
masses and mass-to-charge ratios (m/z), among others. The use
of standardized tools and functions allows reproducible generation
of such lists and reduces human error in curation. This also includes
the use of novel approaches for normalization of retention and
migration times. Furthermore, handling of mass spectral data for
library matching in an automated fashion is an important task. All
together allow to develop reproducible scripts and workflows for
metabolite annotation and identification.
Different tools and packages have been developed as part of the
RforMassSpectrometry initiative, which can be seamlessly
integrated with each other, as well as with other R packages or
packages from the Bioconductor project, and enable users to build
customized analysis workflows through the provided rich infra-
structure. In this chapter the use of functions from different
packages is presented all selected toward the task of standardized
and reproducible metabolite annotation.

2 Materials

All presented calculations and computations can be performed on a

standard Windows PC (e.g., authors performed execution on Win-
dows 10 64-bit machine using R 4.3.1 within RStudio Version
2023.06.0). All used R libraries can be downloaded from the
Comprehensive R Archive Network or Bioconductor. The follow-
ing packages and their respective version are used: Spectra (1.10.2),
MetaboAnnotation (1.4.2), AnnotationHub (3.8.0), MsBack-
endMgf (1.8.0), msdata (0.40.0), MetaboCoreUtils (1.8.0), pheat-
map (1.0.12), msentropy (0.1.4), MobilityTransformR (1.4.0),
BridgeDbR (2.11.1).

2.1 R Scripts All scripts can be used directly in R or more comfortable with an
IDE-like RStudio, which allows more convenient work without the
need to remember all variables, since they are displayed in a separate
window which also allows manual inspection of their content. The
presented code snippets work with different packages, which are
not part of the base R distribution and have to be therefore installed
manually. A version of R > = 4.3 and Bioconductor > = 3.18 is
required for the R scripts. The required packages, along with their
dependencies, can be installed directly within R issuing the com-
mands below:

install.packages("BiocManager")
library(BiocManager)
BiocManager::install(c("Spectra", "MetaboAnnotation",
 Metabolomics Data Analysis 69

"AnnotationHub", "MsBackendMgf",
"msdata", "MetaboCoreUtils",
"pheatmap", "msentropy",
"MobilityTransformR",
"BridgeDbR"))

For the work with R scripts, a basic understanding of R func-

tionalities and data structures is needed (see Note 1). Most results
in the example are returned as a list of specific datatypes, e.g.,
vectors of strings or numeric values or more specific objects (e.g.,
Spectra).

2.2 Data Files The file “GNPS-NIH-NATURALPRODUCTSLIBRARY .mgf”

used in Subheading 3.3 is available on the GNPS library page and
can be downloaded from https://gnps-external.ucsd.edu/
gnpslibrary. The mzML file used in Subheading 3.3.4 is provided
through the msdata R package. The file “metabolites_20220707.
bridge” used in Subheading 3.4 is available from the BridgeDb
project homepage can be downloaded from https://www.
bridgedb.org/mapping-databases/metabolite-mappings.html.

3 Methods

3.1 Formula and Handling of chemical formulae is a recurring task in metabolomics

Mass Utility Functions data processing. Chemical formulae can be standardized in certain
ways to allow a harmonized representation. Different databases
might use different systems to write chemical formulae, but the
Hill notation is currently the most widespread system. The Meta-
boCoreUtils package provides several functions to handle, parse, and
modify chemical formulae [3].

3.1.1 Parsing of 1. All functions from MetaboCoreUtils support either a single

Chemical Formulae formula (as a character string) or multiple formulae (as a char-
acter vector) as input.

chem_formula <- c("C6H12O6", "C6H13NO2")

2. The function countElements() can parse chemical formulae

into individual elements and counts their occurrence. A named
vector is returned in which the names represent the element
and the value the count for that respective element. In case of a
vector with multiple formulae, a list of named vectors is
returned.

chem_formula_count <- countElements(chem_formula)

70 Michael Witting and Johannes Rainer

3. The reverse function is pasteElements(), which converts a

named vector or a list of named vectors into a character vector
with valid chemical formulae (according to the Hill notation).

chem_formula_pasted <-
pasteElements(chem_formula_count)

4. It is also possible to include isotopes in chemical formulae, e.g.,

[13C6]. Both functions, countElements() and pasteEle-
ments(), accept input containing isotopes.

iso_formula <- c("[13C6]H12O6", "[13C3]C3H12O6")

iso_formula_count <- countElements(iso_formula)
iso_formula_pasted <-
pasteElements(iso_formula_count)

5. Lastly, chemical formulae can be standardized to the Hill nota-

tion using the standardizeFormula() function:

chem_formula_nonstd <- c("H12O6C6")

chem_formula_std <-
standardizeFormula(chem_formula_nonstd)

3.1.2 Calculation of 1. Exact masses can be calculated from chemical formulae without
Exact Masses, Kendrick or with isotopes. Either strings or named lists are accepted as
Masses, or Kendrick Mass input in the function calculateMass().
Defect
calculateMass(chem_formula)
calculateMass(chem_formula_pasted)
calculateMass(iso_formula)
calculateMass(iso_formula_pasted)

2. Kendrick masses (KM) and Kendrick mass defects (KMD) are

useful tools for the identification of homologues series, e.g., for
the analysis of lipids [4]. As an example, we create below a data
frame with example lipids as input.

lipids <- data.frame(

name = c("PC 32:0", "PC 34:0", "PC 36:0",
"PC 36:1", "PC 36:2", "PS 36:1"),
formula = c("C40H80NO8P", "C42H84NO8P", "C44H88NO8P",
"C44H86NO8P", "C44H84NO8P", "C42H80NO10P"))
 Metabolomics Data Analysis 71

3. From the provided formula, the mass and the m/z are calcu-
lated using the function mass2mz() (see below). This returns a
named vector with the m/z values and the input formula as
names.

lipid_mass <- calculateMass(lipids$formula)

lipid_mz <- mass2mz(lipid_mass, adduct = "[M+H]+")

4. From the m/z value, the Kendrick mass (KM) and Kendrick
mass defect (KMD) are calculated using the functions calcu-
lateKm() and calculateKmd(). By default, the KM and
KMD are calculated for CH2, but also other KMD are possible
(see Note 2).

lipid_km <- calculateKm(lipid_mz)

lipid_km

lipid_kmd <- calculateKmd(lipid_mz)

lipid_kmd

5. The referenced KMD normalizes the KMD to a specific back-

bone of lipids, e.g., PC headgroup and glycerol, and is calcu-
lated using the lipid_rkmd() function. This allows to filter
for members of the same lipid class as the RKMD results in
negative integer values corresponding to the number of double
bonds.

lipid_rkmd <- calculateRkmd(lipid_mz,

rkmd = 0.749206)
lipid_rkmd

6. The function isRkmd() checks if a calculated RKMD is close

to an integer value within a specific tolerance and returns either
TRUE or FALSE.

isRkmd(lipid_rkmd,
rkmdTolerance = 0.02)

3.1.3 Working with 1. From obtained exact masses, the exact m/z can be calculated
Adducts for different adducts. The function mass2mz() performs the
calculation and accepts a vector of exact masses and adduct
definitions as input. Results are returned as a matrix with one
row per input mass and one column per defined adduct.

exact_mass <- calculateMass(chem_formula)

mass2mz(exact_mass, c("[M+H]+", "[M+Na]+")
72 Michael Witting and Johannes Rainer

2. The function adductNames() returns all currently supported

adducts.

adductNames()
adductNames(polarity = "positive")
adductNames(polarity = "negative")

3. The mz2mass() function performs the opposite operation by

converting a vector of m/z values into neutral masses.

mz2mass(c(181.0707, 203.0526,
132.1019, 154.0838),
c("[M+H]+", "[M+Na]+"))

3.1.4 Working with 1. Isotopic pattern can be used for calculation of potential sum
Isotopic Patterns formulae. The package Rdisop allows the decomposition of
masses, m/z, and isotope patterns. Similar to MetaboCoreUtils,
Rdisop also provides a function to calculate masses from chemi-
cal formulae. Additionally, if these formulae represent ions, the
charge can be defined.

getMass(getMolecule("C6H12O6")
getMass(getMolecule("C6H13O6", z = 1)

2. The results are virtually the same for MetaboCoreUtils and

Rdisop.

getMass(getMolecule("C6H12O6")) -
calculateMass("C6H12O6")

3. From a given formula, the function getIsotope() calculates

the isotopic pattern. This function returns a matrix containing
the m/z values of the different isotopes and their abundance.

getIsotope(getMolecule("C6H13O6", z = 1))
getIsotope(getMolecule("C6H12O6Na", z = 1))

4. As input for the decomposition either a single mass, an m/z or

an isotopic pattern is required. In the example below, a data
frame containing the isotopic pattern of tryptophan is used as
input and plotted for inspection.

iso_pattern <- data.frame(masses = c(205.09715,

206.100189, 207.102574),
intensities = c(100.000,
12.854, 1.170))
plot(iso_pattern$masses, iso_pattern$intensities, type = "h")
 Metabolomics Data Analysis 73

5. From the monoisotopic mass, possible chemical formulae can

be calculated using decomposeMass(). The code below
returns in total 11 different formulae, and the correct formula
can be found on the fifth rank.

decomposeMass(iso_pattern$masses[1], z = 1, ppm = 5.0)

6. Likewise, decomposeIsotopes() uses a vector of masses and

a vector of intensities for the calculation of possible chemical
formulae. The use of an isotope pattern increased the accuracy,
and the correct formula is found on the first rank.

decomposeIsotopes(iso_pattern$masses,
iso_pattern$intensities,
z = 1,
ppm = 5.0)

3.1.5 Matching of 1. The MetaboAnnotation package allows to match two vectors or

Chemical Formulae data frames based on chemical formulae. As an example, we
define below two vectors with chemical formulae that we want
to match against each other.

queries <- c("C6H12O6", "C11H12N2O2")

targets <- c("C6H12O6", "C6H13NO2")

2. The matching is performed with the matchFormula() func-

tion, which returns a Matched object. Data can be retrieved
with the matchedData() function which returns a data.
frame with the matching results.

matchFormula(queries, targets)
matchedData(matchFormula(queries, targets))

3. Internally, the function first standardizes the formulae to Hill

notation prior to the matching, which allows to also match
nonstandard formulae.

queries <- c("H12O6C6", "C11H12N2O2")

targets <- c("C6H12O6", "C6H13NO2")
matchFormula(queries, targets)
matchedData(matchFormula(queries, targets))

3.2 Retention Time Metabolite identification on the highest level requires, according to
Indexing and Mobility the Metabolomics Standard Initiative, information orthogonal to
Transformation MS1, MS2, or MSn. Retention or migration times represent such an
orthogonal information, but separation conditions in LC and CE
are far from being standardized. Even when using nominally the
74 Michael Witting and Johannes Rainer

same conditions (e.g., in case of LC using the same column and

eluents), differences in RTs are observed based on differences in
instrumentation, dead volumes, etc. In GC, retention time
indexing is used to normalize for differences in chromatographic
conditions. Likewise, different systems for retention time indexing
in LC have been described [5–7], and for CE effective mobility
transformation can be used for normalizing CE migration times
[8, 9]. The packages MetaboCoreUtils and MobilityTransformR
offer functions to perform these normalizations.

3.2.1 Retention Time 1. Retention time indexing requires a number of index markers,
Indexing e.g., N-alkyl pyridinium-3-sulfonates (NAPS), as described by
Stoffel et al. In the example below we create a data frame with
retention times and retention indices of these markers. Typi-
cally, the retention index is the number of carbons in the side
chain of NAPS multiplied by 100.

naps <- data.frame(rtime = c(0.456, 0.497, 0.595,

0.874, 2.500, 3.736,
4.414, 4.957,
5.432, 5.875),
rindex = c(100, 200, 300, 400,
500, 600, 700, 800, 900, 1000))

2. As input for retention time indexing, a vector of retention

times is required. The data.frame below contains names
and retention times of example metabolites.

metabolites <- data.frame(name = c("O-

acetylcarnitine", "2-methylbutyrylcarnitine",

"tauroursodeoxycholic acid"),
rtime = c(0.46, 2.759,
5.222))

3. The indexing is performed with the indexRtime() function,

which accepts a vector of retention times and the data.frame
containing the indices. By default, the indexing is performed
using linear interpolation, but other interpolation methods can
be used such as splines or Akima splines. The function returns a
vector of indices with the same length as the input. In case that
retention times of substances fall outside the range, the stan-
dard linear interpolation returns NA. This behavior can be
overwritten for custom functions, e.g., allowing extrapolation.

rindex <- indexRtime(metabolites$rtime, naps)

rindex
 Metabolomics Data Analysis 75

4. Accuracy of retention indexing can be improved by using sub-

stances as secondary corrections. A data.frame with
measured indices and corrected indices is required for the
function correctRindex(). The other input is a vector of
retention indices, which shall be corrected.

ref <- data.frame(rindex = c(109.7561, 855.7895),

refindex = c(110, 850))
correctRindex(rindex, ref)

3.2.2 Effective Mobility 1. MetaboCoreUtils also offers the possibility to perform effective
Transformation mobility transformation for CE-MS-based metabolomics
experiments. Markers of the electroosmotic flow are required
to perform this conversion.

eof <- data.frame(rtime = c(840.796),

mobility = c(0))

2. As input for effective mobility transformation, a vector of

migration times (for consistency also called rtime) is required.
The data.frame below contains the migration time of
procaine.

metabolites <- data.frame(name = c("procaine"),

rtime = c(450.239))

3. The transformation is performed using the function con-

vertMtime(). Since the formulae for effective mobility trans-
formation are using minutes as time unit, the provided
retention times in seconds need to be divided by 60. The
function returns a vector of the same length as the input. For
more complex data, e.g., entire CE-MS runs, see Note 3.

eff <- convertMtime(metabolites$rtime / 60,

rtime = eof$rtime / 60,
mobility = eof$mobility,
tR = 3 / 60,
U = 30,
L = 800)
eff

3.3 Handling and Handling of different kinds of MS spectra, e.g., MS1 isotope pat-
Processing Mass tern or MS2 fragmentation pattern, is an important task for meta-
Spectra Data bolomics data analysis and metabolite identification workflows. The
Spectra package provides a flexible and expandable infrastructure
for loading, handling, and analyzing MS data in R. By default, the
package allows import/export of data from/to files in mzML,
76 Michael Witting and Johannes Rainer

mzXML, or CDF format, but support for a variety of other file

formats is provided through additional R packages such as the
MsBackendMgf (support for files in Mascot Generic Format
(MGF)) or the MsBackendMsp (support for files in NIST MSP
format). Further, the Spectra package provides functions to process,
filter, and clean MS data as well as functionality to compare frag-
ment spectra using a flexible framework that allows, besides stan-
dard and established similarity calculations, also the use of similarity
measures provided by other packages or even user-provided simi-
larity scores. Finally, the package allows export of MS data in a
variety of formats.

3.3.1 Importing 1. The MGF file format (http://www.matrixscience.com/help/

Fragment Spectra from data_file_help.html) is a very simple and loosely defined plain
MGF Files text file format used by many databases and laboratories for
exchange of their MS2 data. Besides some required fields, such
as PEPMASS for the precursor m/z of a spectrum, it allows to
store additional metadata information, and thus a variety of
MGF dialects have been evolved. To map such custom fields to
standard attributes in a Spectra object, we define below a
variable map. We get the default mapping of fields using the
spectraVariableMapping() function and add additional
mappings for the MS level (defined with the field MSLEVEL
in the GNPS MGF file) and SMILES (provided with the field
SMILES).

library(Spectra)
library(MsBackendMgf)
map <- spectraVariableMapping(MsBackendMgf())
map <- c(map, c(msLevel = “MSLEVEL”, smiles = “SMILES”))

2. To import MS data from a MGF file, we use the Spectra

function providing the file name and specify source =
MsBackendMgf(). In addition, we pass our mapping defini-
tion using the mapping parameter.

gnps <- Spectra("GNPS-NIH-NATURALPRODUCTSLIBRARY.mgf",

source = MsBackendMgf(), mapping = map)

3. The various available fields/spectra variables imported from

the file can be listed using the spectraVariables() func-
tion. The MS levels of the spectra reported in the MGF file will
be available, because of the variable mapping used above, as
spectra variable msLevel.

spectraVariables(gnps)

4. The Spectra object now contains all fragment spectra from

the MGF file. Individual spectra variables can be extracted
 Metabolomics Data Analysis 77

either using one of the dedicated accessor functions or using $

and the name of the variable. In both cases, the values for all
spectra will be returned as a vector of length equal to the
number of spectra in the object.

msLevel(gnps)
gnps$msLevel

5. Peak data, i.e., the m/z and intensity values of the individual
mass peaks per spectrum, can be extracted separately using
either the mz() and intensity() functions or together in a
two-dimensional array using the peaksData() function. For
the former case, a list of numeric values is returned and for the
latter a list of numerical (two-column) matrices.

## Get the m/z values for the 5th spectrum

mz(gnps)[[5]]
## Get the peaks matrix for the 5th spectrum
peaksData(gnps)[[5]]

3.3.2 Processing and 1. Fragment spectra (and also MS1 spectra) can contain a large
Cleaning of Spectra Data amount of low-intensity peaks that might represent in most
cases noise. The filterIntensity() function can be used
to remove such low-signal peaks. In addition to a fixed intensity
cutoff, also a function defining a spectrum-specific cutoff can
be provided. This function is supposed to take intensity values
of one spectrum as an input (parameter x) and to return a
logical vector of length equal to the number of peaks in a
spectrum defining whether a mass peak should be kept or
removed. The function below returns TRUE for all peaks with
an intensity larger than 1% of the maximum intensity of that
spectrum.

int_cut <- function(x, ...) {

x > max(x, na.rm = TRUE) * 0.01
}

2. Apply this intensity filter on the data. To evaluate the impact of

this filter on the data, we also determine the median number of
peaks before and after filtering (the function lengths()
reports the number of mass peaks of each spectrum in the
Spectra object).

median(lengths(gnps))
gnps <- filterIntensity(gnps, intensity = int_cut)
median(lengths(gnps))
78 Michael Witting and Johannes Rainer

3. In addition to the many other filtering and processing func-

tions available in the Spectra package, it is also possible to apply
any user-defined data manipulation function to a Spectra
object using the addProcessing() function. To illustrate
this, we define below a function that takes a peak matrix as
input (i.e., a two-dimensional array with the m/z and intensity
values of a spectrum), scales the intensity values such that the
intensities of all peaks sum up to 1, and returns this modified
peak matrix. This is a common data manipulation operation for
fragment spectra.

scale_int <- function(x, ...) {

x[, "intensity"] <- x[, "intensity"] /
sum(x[, "intensity"], na.rm = TRUE)
x
}

4. This function can be applied to the data using the addPro-

cessing() function. To evaluate the result of this operation,
we sum the intensities of all peaks for all spectra. This sum
should be 1 for all spectra after applying the data manipulation
function.

gnps <- addProcessing(gnps, scale_int)

sum(intensity(gnps))

3.3.3 Exporting MS Data 1. Data from any Spectra object can be exported to a file in
in MGF Format MGF format using the export() function and defining the
output format by setting backend = MsBackendMgf(). We in
addition pass the mapping between spectra variables and MGF
fields defined above with parameter mapping. Finally, with
optional parameter exportTitle = FALSE, we disable the
default export of a title (name) for each spectrum. The gener-
ated MGF file will then be in GNPS-flavor format. Note that it
would be similarly possible to export data in, e.g., MSP format
by loading the MsBackendMsp package and using back-
end = MsBackendMsp() instead.

export(gnps, file = "gnps_exported.mgf", mapping = map,

backend = MsBackendMgf(), exportTitle = FALSE)

3.3.4 Spectra Similarity Calculating similarities between fragment spectra is a core concept
Calculations for most annotation/identification approaches. While a large num-
ber of software and similarity measures exist, we focus here on
functionality provided by the MsCoreUtils, Spectra, and MetaboAn-
notation Bioconductor packages [3]. Spectra similarity calculation
 Metabolomics Data Analysis 79

is in general a two-step approach in which first (fragment) peaks

from the two compared spectra need to be matched against each
other using their m/z values, to then calculate a similarity score
between them considering also their intensities. Here we show how
similarity calculations can be efficiently performed in R on a simple
example in which we compare fragment spectra, after some initial
cleaning, using different similarity calculation approaches and simi-
larity algorithms. In particular, we focus on the capability of the
infrastructure provided by the Spectra package to create highly
customized analysis pipelines through combination of pre- and
user-defined analysis steps/functions.
1. We load data from an mzML file with spectra from a single
LC-MS/MS run. MS2 spectra were measured using a data-
dependent acquisition strategy. The data file is available in the
msdata R package, and we thus provide the path to this data file
in the Spectra() constructor call. In addition, to enable
import of MS data from an mzML file, we set
source = MsBackendMzR().

fl <- system.file("TripleTOF-SWATH",
"PestMix1_DDA.mzML",
package = "msdata")
dda <- Spectra(fl, source = MsBackendMzR())

2. We next restrict the data to MS2 spectra and apply the intensity
filter defined in the previous example that removes all peaks
with an intensity below 1% of the maximum peak signal in a
spectrum.

dda <- filterMsLevel(dda, 2)

dda <- filterIntensity(dda, int_cut)

3. Eventual fragment peaks representing the precursor ion as well

as peaks with an m/z value larger than the precursor m/z are
subsequently removed using the filterPrecursorPeaks()
function. We specify mz = ">=" and tolerance = 0.05 to
remove all peaks with an m/z value larger or equal to the
spectrum’s precursor m/z minus a value of 0.05 (which also
removes the eventually present precursor peak from each frag-
ment spectrum). Finally, we reduce the dataset to fragment
spectra with at least three peaks.

Dda <- filterPrecursorPeaks(dda, tolerance = 0.05,

mz = “>=”)
dda <- dda[lengths(dda) >= 3]
80 Michael Witting and Johannes Rainer

4. For spectra similarity calculations, relative intensities between

the individual fragment peaks are more important than abso-
lute intensity values. Thus, we scale all intensities within each
spectrum to a total intensity sum of 1 using the scale_int()
function defined in the previous example.

dda <- addProcessing(dda, scale_int)

5. For the present example we in addition focus only on fragment

spectra recorded in a restricted retention time window and
filter thus the dataset to spectra measured between 450 and
550 s.

dda <- filterRt(dda, rt = c(450, 550))

6. We next calculate pairwise similarities between all spectra. As

detailed above, spectra similarity scoring is a two-step approach
that first needs a mapping of peaks between the compared
spectra followed by the actual similarity calculation between
these aligned spectra. The compareSpectra() function
allows users, with parameters MAPFUN and FUN, to specify a
function (either built-in or user provided) for each of these two
steps. The default for MAPFUN is the joinPeaks() function
that maps peaks between spectra based on parameters ppm and
tolerance as well as a third parameter type that defines how
peaks are joined [3]: with the default type = "outer" all peaks
from both spectra (whether mapped or not) are passed to the
similarity calculation function. A similarity calculation function
can be specified with parameter FUN with the default being
ndotproduct() from the MsCoreUtils package that calculates
the (normalized) dot product similarity. Generally, compare-
Spectra() is used to calculate similarities between fragment
spectra from two different Spectra objects, but in the present
example we submit only a single Spectra object, hence calcu-
lating pairwise similarities between all of its spectra. See Note 4
for performance considerations of this operation.

sim <- compareSpectra(

dda, MAPFUN = joinPeaks, tolerance = 0, ppm = 50,
FUN = MsCoreUtils::ndotproduct)

7. The result of this call is a similarity matrix with numbers of rows

and columns being equal to the number of compared spectra.
We below visualize this matrix using the pheatmap() function
disabling the default clustering of rows and columns to ensure
the spectra similarities are shown in the order of input spectra
and hence ordered by their retention time. As can be seen in the
 Metabolomics Data Analysis 81

Fig. 1 Heatmap visualizing the pairwise similarity matrix between all fragment spectra of the example dataset.
Spectra are ordered by retention time

resulting plot (Fig. 1), fragment spectra recorded at about the

same retention time (thus likely representing the same ion)
have a higher similarity.

library(pheatmap)
pheatmap(sim, cluster_rows = FALSE, cluster_cols = FALSE)

8. In this similarity matrix, we next identify spectra with very high

similarities (dot product similarity higher than 0.9). Due to its
symmetric nature, we first restrict the similarity matrix to the
82 Michael Witting and Johannes Rainer

Fig. 2 Mirror plot of two fragment spectra with high similarity. Matching peaks between the two spectra are
indicated with a blue color

upper triangle excluding also self-self similarities present in the

diagonal.

sim[lower.tri(sim, diag = TRUE)] <- NA

high_sim <- which(sim > 0.9, arr.ind = TRUE)

9. To validate similarity scoring results, it is generally good to also

visually inspect the compared spectra using a mirror plot.
Below we create a mirror plot (Fig. 2) between two spectra
with a reportedly high similarity (spectrum numbers 2 and 3 in
the data subset).

plotSpectraMirror(dda[2], dda[3], ppm = 50, tolerance = 0)

10. While a mirror plot is ideal for pairwise comparison, it does not
allow to visualize multiple spectra in a single plot. One possi-
bility for such a visualization would be a simple intensity heat-
map. To create such a plot for the first ten spectra, in our
dataset, we generate next the required two-dimensional repre-
sentation of our data by first binning the spectra along the m/z
dimension with a bin size of 0.1 and then extract and combine
the intensity values for these binned spectra into an intensity
matrix.

dda_bin <- bin(dda[1:10], binSize = 0.1)

imat <- do.call(rbind, intensity(dda_bin))

11. After removing columns (m/z bins) without intensities, we can

again use the pheatmap function to visualize the spectra (see
Fig. 3).
 Metabolomics Data Analysis 83

Fig. 3 Visualization of multiple spectra. Rows in this plot represent different spectra and columns (binned) of
m/z values with color-coded intensities of fragment peaks in the individual spectra

z <- colSums(imat) == 0
imat <- imat[, !z]
colnames(imat) <- mz(dda_bin)[[1]][!z]
rownames(imat) <- seq_along(dda_bin)
pheatmap(imat, cluster_rows = FALSE, cluster_cols = FALSE)

12. The default spectra similarity calculation algorithm (or more

precisely the default peak mapping function) considers frag-
ment peaks from both spectra (i.e., peaks are mapped/joined
using an outer join). Alternatively, it is also possible to calculate
a reverse score that considers all peaks from the second spectrum
and only those from the first spectrum that can be mapped to
peaks of the second spectrum. For that we select
type = “right” in the compareSpectra() call in which
case the above described right join of the peaks will be
performed.

## Calculate reverse score between the spectrum 2 and 3

compareSpectra(dda[2], dda[3], ppm = 50, tolerance = 0,
type = "right")
## As a comparison, the score considering all peaks
compareSpectra(dda[2], dda[3], ppm = 50, tolerance = 0)

13. Also, compareSpectra() supports the use of different simi-

larity algorithms to assess spectra similarity. Next to the simi-
larity functions provided by MsCoreUtils, also user-defined
functions or algorithms from external R packages can be
used. An MS entropy-based [10] similarity can, for example,
be calculated by passing the msentropy_similarity()
function from the msentropy R package with parameter FUN
to the compareSpectra() function. Since this function per-
forms the peak mapping internally, we need to disable the
default peak mapping from compareSpectra() by setting
84 Michael Witting and Johannes Rainer

MAPFUN = joinPeaksNone. With parameters ms2_toleran-

ce_in_ppm and ms2_tolerance_in_da, we configure the
peak mapping of msentropy_similarity, and with
clean_spectra = FALSE, we disable its default internal spec-
trum cleaning since we performed already some spectrum
cleaning above.

library(msentropy)
compareSpectra(dda[2], dda[3], MAPFUN = joinPeaksNone,
FUN = msentropy_similarity,
ms2_tolerance_in_ppm = 50,
ms2_tolerance_in_da = 0,
clean_spectra = FALSE)

3.3.5 Comparing As a first step in the compound identification process, experimental

Experimental Fragment fragment spectra of the unknows are usually first compared against
Spectra Against a Public publicly available fragment spectra. Such comparisons are however
Reference Database hampered by the lack of common data structures and formats and
different ways to access the available resources. The infrastructure
provided by the Spectra package is designed to alleviate this task by
supporting a variety of input formats and sources simplifying hence
the integration of external data resources into annotation work-
flows. A very easy, and also reproducible, workflow is to match
experimental fragment spectra against reference spectra from one
of the MassBank data releases provided through Bioconductor’s
AnnotationHub.
1. Connect to the Bioconductor AnnotationHub and list avail-
able MassBank data releases.

library(AnnotationHub)
ah <- AnnotationHub()
query(ah, "MassBank")

2. Retrieve the MassBank data for a specific release. This dataset,

identified by its stable identifier listed by the call above, will be
downloaded and cached locally. Any future call will load the
resource from the local cache instead of downloading it again.

mb <- ah[["AH107049"]]

3. The data is provided as a CompDb object (SQLite database)

defined in the CompoundDb package, and its full fragment
spectra data (along with additional compound annotations)
can be accessed as a Spectra object.

mbs <- Spectra(mb)

 Metabolomics Data Analysis 85

4. Similar to the experimental data above, we process also the MS2

data from MassBank using the same cleaning steps, i.e., by
removing peaks with an intensity lower than 1% of the maxi-
mum peak signal or with an m/z greater or equal to the
spectra’s precursor m/z, followed by exclusion of spectra with
less than three peaks and finally scaling all intensities to a total
intensity sum of 1.

mbs <- filterIntensity(mbs, int_cut)

mbs <- filterPrecursorPeaks(mbs, tolerance = 0.05,
mz = ">=")
mbs <- mbs[lengths(mbs) >= 3]
mbs <- addProcessing(mbs, scale_int)

5. Pairwise similarities between experimental fragment spectra

and reference spectra could again be calculated with the com-
pareSpectra() function, which would however result in a
very large similarity matrix which would also be difficult to
handle and further process. Instead, we use the matchSpec-
tra() function from the MetaboAnnotation package which
simplifies the task by allowing to restrict results to hits with a
similarity higher than a provided threshold. A simple pairwise
spectra similarity scoring can be performed and configured
with the CompareSpectraParam parameter object. To use
MS entropy similarity scores, we specify, like in the example
above, MAPFUN = joinPeaksNone and FUN = msentropy_-
similarity, configuring this similarity function with the
additional parameters ms2_tolerance_in_ppm, ms2_to-
lerance_in_da, and clean_spectra. With the parameter
THRESHFUN we define which hits to report. In our case we keep
all hits with a spectra similarity score higher than 0.7.

res <- matchSpectra(

dda, mbs, CompareSpectraParam(
MAPFUN = joinPeaksNone,
FUN = msentropy_similarity,
ms2_tolerance_in_ppm = 50,
ms2_tolerance_in_da = 0,
clean_spectra = FALSE,
THRESHFUN = function(x) x > 0.7))

6. The result object contains, in addition to the matching results,

also the query and target Spectra objects. The results are
organized along the query (i.e., the length of the result object
is equal to the number of query spectra), and each element of
the result object contains the results from one query spectrum.
To inspect the results, we below subset the object to the first
query spectrum which could be matched to at least one
86 Michael Witting and Johannes Rainer

fragment spectra from MassBank. We use the whichQuery()

function to identify such query spectra and subset the result
object to the first query spectrum with matching target spectra.
In addition, we use the pruneTarget() function on this
result subset to keep only target fragment spectra matching
the selected query spectrum.

idx <- whichQuery(res)

res_a <- res[idx[1]]
res_a <- pruneTarget(res_a)

7. Get the name of the compound(s) in MassBank with reported

similar fragment spectra as well as the similarity score for the
matches. In our example, the name of the compound is avail-
able in column "target_name". colnames(res_a) could
be used to list all available column names (from the query and
target Spectra objects); column names from the target object
can be identified with the prefix "target_". Note that the full
results for the present query spectrum (i.e., all information
from the matching target spectra) could be extracted with
matchedData(res_a)).

## Get the name of the matching compound(s)

res_a$target_name
## Get the similarity score for each matching spectra
res_a$score

8. Finally, it is also advisable to visually validate the matches.

Below we create a mirror plot between the query spectrum
and the first matching target spectrum from MassBank (Fig. 4).
plotSpectraMirror(query(res_a), target(res_a)[1])

Fig. 4 Mirror plot between an experimental fragment spectrum (top) and a reference spectrum from MassBank
(bottom). Matching peaks are indicated by a blue color
 Metabolomics Data Analysis 87

3.4 Mapping/ Depending on the further downstream data analysis tool, e.g.,
Conversion of MetaboAnalyst or others, different identifiers are required for path-
Metabolite Identifiers way mapping or similar tasks. Manual search for different identifiers
in database is a tedious task since no harmonized metabolite data-
base exists. Metabolite structures represent their most unique iden-
tifier, but often identifiers from different databases such as KEGG,
HMDB, and ChEBI are preferred. Such identifiers can be con-
verted into each other, for example, using the chemical translation
service. Another possibility that can be easily integrated into R
workflows is BridgeDB via the BridgeDbR package [11, 12].
1. The BridgeDbR package requires the most current bridge
database, which can be found on the project website (see
above). This database is loaded into a variable called mapper
used for the entire conversion process.

mapper <- loadDatabase("metabolites_20220707.bridge")

2. BridgeDbR used specific system codes for each metabolite

database. They can be retrieved by the getSystemCode()
function.

getSystemCode("InChIKey")
getSystemCode("ChEBI")
getSystemCode("KEGG Compound")

3. Mapping or conversion of different identifiers is performed

using the map() function. As an example, the KEGG identifier
of ATP (C00002) is used as input and converted to different
other identifiers, e.g., from ChEBI or HMDB. This function
returns a data frame.

map(mapper,
"C00002",
source = getSystemCode("KEGG Compound"),
target = getSystemCode("ChEBI"))

4. Mapping of multiple identifiers to the same target database can

be performed with the function maps(). It requires as input a
data.frame containing the system code for the source data
base and the actual identifier.

input <- data.frame(source = c("Ck", "Ch"),

identifier = c("C00002", "HMDB0000538"))

5. However, this function requires the target database to be

known. When working with identifier from different databases,
it is helpful to automatically recognize their source. This can be
88 Michael Witting and Johannes Rainer

achieved through regular expressions. The web page

identifiers.org lists identifiers used in biochemical research as
well as the respective regular expressions that can be used. The
example only detects KEGG and HMDB identifiers but can be
customized. With the function the input data frame can be
generated automatically.

identifyDb <- function(x) {

if(grepl("^C\crd+$", x)) {
return(getSystemCode("KEGG Compound"))
} else if (grepl("^HMDB\crd+$", x)) {
return(getSystemCode("HMDB"))
} else {
return(NA_character_)
}

ids <- c("C00002", "HMDB0000538")

input <- data.frame(identifier = ids,

source = unlist(lapply(ids, identifyDb)))

4 Notes

1. Consult a book or web page on working with R if not familiar,

e.g., https://cran.r-project.org/doc/manuals/r-release/R-
intro.pdf.
2. This approach can be used even further to construct three-
dimensional KMD plot for the annotation of lipids [4].
3. Mobility transformation can be also performed on more com-
plex data structures using the MobilityTransformR package.
4. In most cases, especially if MS data from mzML, mzXML, or
CDF files is analyzed, the peak data (m/z and intensity values of
the individual peaks) is only loaded upon request into memory.
Also, data manipulation operations are not propagated imme-
diately to the MS data of a Spectra object but only applied
when the respective (m/z and intensity) values are requested.
Loading the full MS data into memory and forcing all data
manipulation operations to be applied on the data can increase
 Metabolomics Data Analysis 89

the performance of many analysis steps, given sufficient avail-

able main memory. Data can be loaded into memory, and data
manipulations can be applied immediately to the MS data with
the function calls below.

dda <- setBackend(dda, backend = MsBackendMemory())

dda <- applyProcessing(dda)

References
1. Smith CA, Want EJ, O’Maille G, Abagyan R, 7. Hao J-D, Chen Y-Y, Wang Y-Z, An N, Bai P-R,
Siuzdak G (2006) XCMS: processing mass Zhu Q-F et al (2023) Novel peak shift correc-
spectrometry data for metabolite profiling tion method based on the retention index for
using nonlinear peak alignment, matching, peak alignment in untargeted metabolomics.
and identification. Anal Chem 78(3): Anal Chem 95(35):13330–13337. https://
7 7 9 – 7 8 7 . h t t p s : // d o i . o r g / 1 0 . 1 0 2 1 / doi.org/10.1021/acs.analchem.3c02583
ac051437y 8. Drouin N, Pezzatti J, Gagnebin Y, González-
2. Benton HP, Wong DM, Trauger SA, Siuzdak G Ruiz V, Schappler J, Rudaz S (2018) Effective
(2008) XCMS2: processing tandem mass spec- mobility as a robust criterion for compound
trometry data for metabolite identification and annotation and identification in metabolomics:
structural characterization. Anal Chem 80(16): toward a mobility-based library. Anal Chim
6382–6389 Acta 1032:178–187. https://doi.org/10.
3. Rainer J, Vicini A, Salzer L, Stanstrup J, Badia 1016/j.aca.2018.05.063
JM, Neumann S et al (2022) A modular and 9. González-Ruiz V, Gagnebin Y, Drouin N,
expandable ecosystem for metabolomics data Codesido S, Rudaz S, Schappler J (2018)
annotation in R. Metabolites 12(2):173 ROMANCE: a new software tool to improve
4. Korf A, Vosse C, Schmid R, Helmer PO, data robustness and feature identification in
Jeck V, Hayen H (2018) Three-dimensional CE-MS metabolomics. Electrophoresis
Kendrick mass plots as a tool for graphical 39(9-10):1222–1232. https://doi.org/10.
lipid identification. Rapid Commun Mass 1002/elps.201700427
Spectrom 32(12):981–991. https://doi.org/ 10. Li Y, Kind T, Folz J, Vaniya A, Mehta SS, Fiehn
10.1002/rcm.8117 O (2021) Spectral entropy outperforms
5. Stoffel R, Quilliam MA, Hardt N, Fridstrom A, MS/MS dot product similarity for small-
Witting M (2022) N-Alkylpyridinium sulfo- molecule compound identification. Nat Meth-
nates for retention time indexing in reversed- ods 18(12):1524–1531. https://doi.org/10.
phase-liquid chromatography-mass spectrome- 1038/s41592-021-01331-z
try-based metabolomics. Anal Bioanal Chem 11. van Rijswijk M, Beirnaert C, Caron C,
414(25):7387–7398. https://doi.org/10. Cascante M, Dominguez V, Dunn W et al
1007/s00216-021-03828-0 (2017) The future of metabolomics in ELIXIR
6. Aalizadeh R, Nikolopoulou V, Thomaidis NS [version 2; peer review: 3 approved].
(2022) Development of liquid chro- F1000Research 6:1649. https://doi.org/10.
matographic retention index based on coca- 12688/f1000research.12342.2
mide diethanolamine homologous series (C 12. Batchelor C, Brenninkmeijer CYA,
(n)-DEA). Anal Chem 94(46):15987–15996. Chichester C, Davies M, Digles D, Dunlop I
https://doi.org/10.1021/acs.analchem. et al (2014) Scientific lenses to support multi-
2c02893 ple views over linked chemistry data. Springer,
Cham, pp 98–113
 Chapter 5

Data Treatment for LC-MS Untargeted Analysis

Mar Garcia-Aloy, Johannes Rainer, and Pietro Franceschi

Abstract
Liquid Chromatography-Mass Spectrometry (LC-MS) untargeted experiments require complex bioinfor-
matic strategies to extract information from the experimental data. Here we discuss the “data preproces-
sing,” the set of procedures performed on the raw data to produce a data matrix which will be the starting
point for the subsequent statistical analysis. Data preprocessing is a crucial step on the path to knowledge
extraction, which should be carefully controlled and optimized in order to maximize the output of any
untargeted metabolomics investigation.

Key words Preprocessing, Peak picking, Retention time correction, Metadata, Quality check, Missing
values

1 Introduction

Liquid Chromatography-Mass Spectrometry (LC-MS) is an estab-

lished and widely used analytical technique and combines the sepa-
ration potential of LC with the ability of MS to quantify the ion
intensity [1]. From a data scientist point of view, LC-MS produces
“three-dimensional” datasets where each ion generated in the ioni-
zation source is characterized by m/z value, retention time (rt), and
intensity.
While in targeted applications selected ions are used to quantify
the analytes of interest, untargeted investigations deal with the full
set of ionic signals trying to characterize all detectable analytes,
including also chemical unknowns [2]. However, this potential
comes at the price of a more complicated and heavily dataset-
dependent knowledge extraction process, which requires advanced
bioinformatics tools and strategies [3].
A schematic representation of the data analysis workflow to
maximize the process of knowledge mining in an untargeted
LC-MS investigation is presented in Fig. 1: the path involves a
constant feedback between experimental design, data processing,
and statistical analysis. Data inspection and evaluation of the results

Olga Deda et al. (eds.), Metabolic Profiling: Methods and Protocols, Methods in Molecular Biology, vol. 2891,
https://doi.org/10.1007/978-1-0716-4334-1_5,
© The Author(s), under exclusive license to Springer Science+Business Media, LLC, part of Springer Nature 2025

91
92 Mar Garcia-Aloy et al.

Fig. 1 Data analysis workflow of a typical LC-MS untargeted experiment. Only the colored boxes are detailed
in the text

of each processing step are mandatory to ensure an optimal data

processing and to avoid misleading results. Analysts should thus
take the time to inspect data and results and avoid the use of soft-
wares or pipelines relying on a single out-of-the-box algorithm and
do not allow customization of individual analysis steps. In this
chapter, we focus on the so-called data preprocessing that aims to
summarize the raw mass spectrometry (MS) data into a
two-dimensional matrix with samples on the rows and ions—usu-
ally referred to as LC-MS features or simply features—on the col-
umns. Our discussion will also touch on data organization, because
this step plays a fundamental role in guaranteeing the interopera-
bility and reproducibility of the whole data analysis process
[4]. Considering the high pace of evolution of bioinformatics, a
pure “how-to” guide would be outdated quite fast. Thus, we try to
also include some general considerations valid in a wider variety of
contexts. In parallel, to give practical guidance to the researcher, we
illustrate how the workflow can be implemented in the R environ-
ment [5], which is obviously not the only possible solution [6–
11]. We believe that a general understanding of the key aspects of
the problem can be of value beyond the specific solution chosen to
perform data analysis.
 Data Treatment for LC-MS Untargeted Analysis 93

2 Materials

1. Metadata organization. Spreadsheets to provide sample infor-

mation and specify which raw data files contain measurements
of which sample. These can be simple tabular text files or Excel
spreadsheets assigning samples to data files and ideally provide
already the full phenotypic as well as experimental information
(such as batch, column used, date of measurement) for each
sample/file. Alternatively, the ISA metadata model can be used
to encode such information [4, 12], in particular, if one wants
to store the raw data in the Metabolights repository.
2. Data conversion. Vendor software allowing data export in open
file format (see Note 1) or the cross-platform ProteoWizard
toolkit [13, 14].
3. Processing (chromatographic peak picking, grouping and
retention time correction, imputation of missing peaks, quality
check). The xcms [15–17] Bioconductor package, which is a
popular solution to process raw LC-MS data.
4. Annotation. The MetaboAnnotation [18] Bioconductor pack-
age, which provides a set of functions to perform compound
annotation in a user-friendly manner.
5. Example dataset. Spiked-apple dataset available from Metabo-
Lights (MTBLS59) [19], downloaded and unzipped in a direc-
tory named “MTBLS59” within the R working directory.
Briefly, the dataset contains the results of the analysis of
40 apple extracts divided in four classes that were spiked with
a set of known compounds. The extracts were analyzed by
UPLC-QTOF-MS in positive and negative mode.

3 Methods

3.1 Metadata The term “experiment metadata” refers to the set of information
Organization describing the samples and the analytical pipeline (see Note 2).
Organizing and storing a detailed record of the sample metadata
(see Note 3) is crucial to perform a reliable interpretation of the
results, evaluating the presence (and the impact) of possible con-
founding factors.
1. Collect and organize (possibly with the help of a spreadsheet)
all the important properties of the samples (e.g., treatment
type, day of collection, origin, etc.). Keep track also of the
sample-specific analytical details such as batch, column, mea-
surement date, and time.
2. It is recommended to define a spreadsheet that contains the file
names of the raw data files along with descriptions of the
94 Mar Garcia-Aloy et al.

samples for each file as additional columns. Such a file could

then be imported in R and used as metadata information along
the experimental data enabling to, e.g., label or color data
points in plots based on file names or provided phenotypic
information.
3. Another, more automated, option would be to create a table
directly in R describing the sample properties based on the file
names.

3.2 Data Conversion In general, the vendor software of the MS instrument manufacturer
(See Note 4) can be used to perform this task (see Note 5). Alternatively, or if the
vendor software does not support data export in the open mzML
file format, the open source ProteoWizard “msconvert” software
(see Note 6), which is available as a command line tool or as a point-
and-click application for Windows (MSconvertGUI) (see Note 7),
can be used. Also, a msconvert docker image is available that allows
reproducible batch processing and conversion of large sets of files
on, e.g., a linux cluster.
1. Open a terminal or a command window and run “msconvert”
into the raw data directory. As an example, to convert vendor-
specific raw data (here .RAW data files) into mzML files and
store them into the open_data directory, use:

msconvert *.RAW -o open_data

Alternatively, the use of the GUI application is self-

explanatory. Also in this case batch mode conversion of a set
of files is possible.

3.3 Preprocessing Each metabolite produces at least one peak in the (m/z, rt) plane.
These are sometimes also referred to as chromatographic peaks to
distinguish them from the mass peaks, which are individual signals
in a single mass spectrum measured by the MS instrument. To
quantify metabolites, a dedicated peak detection algorithm has to
be applied to the raw MS data that is able to distinguish real signal
of an ion from electrical or chemical noise [7, 20]. This process of
chromatographic peak detection (peak picking) has to be per-
formed (separately) on all data files and needs to be customized
depending on the LC-MS setup used for the experiment. Subse-
quently, peaks representing signals from the same type of ions (with
similar m/z and rt) need to be grouped across samples to define the
final list of features (correspondence analysis or grouping). In order
to do that, it is necessary to account for any potential shifts of m/z
and rt between samples (retention time correction) [21]. The
measured m/z values can differ between samples (or even between
scans of the same sample) because even the most advanced mass
spectrometer instrument is not absolutely precise. Rt shifts, instead,
 Data Treatment for LC-MS Untargeted Analysis 95

are unavoidable characteristics of chromatography. Technically

speaking, a “retention time correction” step aims to compensate
for the shifts in rt, and in the subsequent correspondence step,
chromatographic peaks found in the different samples are
“grouped” (across samples) into features to result in a final consen-
sus list. A large number of algorithms for these tasks are available in
commercial or open source software products [22, 23].
At the end of each step in the processing workflow, it is
extremely important to check the outcomes of the overall process
and to eventually adjust and adapt the settings. Such quality checks
can be easily done using a series of plots and simple visualizations.
Below, within each of these steps, we suggest some practices for
such quality checks on the generated data.

3.4 Preprocessing: First it is necessary to load the MS data, ideally along with the
Data Import corresponding associated metadata. Below we create a data frame
with sample annotation based on the file names of the experiment
and load the MS data into R using the readMSData() function
from the MSnbase R package. With the parameter mode =
"onDisk", only general spectra information is imported into
memory, while m/z and intensity values will be retrieved on-the-
fly from the raw data files allowing also processing of very large MS
experiments on standard computer systems [24].

library(xcms)
cdffiles <- list.files(path = "MTBLS59",
pattern = "CDF",
full.names = TRUE)
myfiles <- grep("neg", cdffiles, value = TRUE)[1:40]
pd <- data.frame(sample_name = basename(myfiles),
do.call(rbind, strsplit(gsub(".CDF", "",
basename(myfiles)), "_")))
colnames(pd) <- c("sample_name", "apple", "group",
"polarity", "sample")
raw_data <- readMSData(
files = myfiles,
pdata = new("NAnnotatedDataFrame", pd),
mode = "onDisk")

3.5 Preprocessing: The main characteristics that distinguish a real peak (representing
Peak Picking signal of an ion) from noise are its intensity, width, and shape.
Different algorithms handle them in different ways [7, 20], but
most algorithms for detection of chromatographic peaks in LC-MS
data assume the peak shape to be nearly gaussian. Most importantly,
not only the algorithm but also the specific parameters of each
algorithm greatly affect the results (see Note 8).
96 Mar Garcia-Aloy et al.

1. Choose the peak-picking algorithm on the basis of the char-

acteristics of the analytical pipeline [25, 26]. In xcms, several
choices are available (see Note 9). The type of instrumental
setting used to analyze MTBLS59 suggests using the centWave
algorithm [16].
2. Identify the critical parameters of the algorithm and possibly
link them with the characteristics of the analytical pipeline. For
centWave, for example, the most important parameters to be
considered are ppm, which is determined by the maximum
expected deviation of m/z values of mass peaks of the specific
ion across consecutive scans (spectra) (note that this is usually
much larger than the ppm specified by the manufacturer used
to determine the mass accuracy of the instrument); prefil-
ter, which has to be set considering the minimum intensity
(I) of a true signal (at the single spectrum level) and also in how
many consecutive scans (k) that signal should be detected;
peakwidth, which defines the expected range of chro-
matographic peak-widths (in seconds); and snthresh, which
sets the cut-off to distinguish signal from noise (signal-to-noise
ratio).
3. Quality check: Check the results of the peak picking on a
representative sample (see Note 10). For instance, to run cen-
tWave with default parameters on a specific file:

test_sample <- filterFile(raw_data, 1)

test_peaks <- findChromPeaks(test_sample,
param = CentWaveParam())

The results of this step can then be visually inspected by

plotting the location of identified chromatographic peaks in the
m/z, rt plane:

plotChromPeaks(test_peaks)

In addition, it is advisable to evaluate the peak detection

results for known compounds that should be present in the
data. For the present example, we extract the ion chromato-
gram for quercetin-3-glucoside and quercetin-3-galactoside
and plot that data. Identified chromatographic peaks would
be indicated/highlighted in that plot.

chr <- chromatogram(

raw_data,
mz = 463.0882 + 0.01 * c(-1, 1),
rt = c(405, 430))
xchr <- findChromPeaks(chr, param = CentWaveParam())
plot(xchr, peakType = "rectangle", peakBg = NA)
 Data Treatment for LC-MS Untargeted Analysis 97

Fig. 2 Signal for deprotonated ions of quercetin-3-glucoside and quercetin-3-

galactoside. Peak areas of identified chromatographic peaks should be high-
lighted with a rectangle. The fact that no rectangles are observed indicates that
no chromatographic peaks have been identified using the default parameters

4. Optimize the peak-picking parameters until the result is satis-

factory (see Note 11).
Default parameters failed to identify peaks for the above
compounds (Fig. 2). Peak detection parameters can be evalu-
ated and optimized by performing the analysis on the extracted
ion chromatogram (Fig. 3):

xchr <- findChromPeaks(chr,

param = CentWaveParam(
prefilter = c(0, 0),
peakwidth = c(2, 15),
snthresh = 1))
plot(xchr, peakType = "rectangle", peakBg = NA)

5. Perform the peak picking on the whole dataset with the

selected set of parameters. The final result object will contain
the full list of the peaks detected in all the samples, with some
additional information on the mass and retention time ranges.
The complete analysis (with the parameters set as suggested in
[26]) can be performed using.

xdata <- findChromPeaks(raw_data,

param = CentWaveParam(
ppm = 15,
prefilter = c(0, 0),
peakwidth = c(2, 15),
snthresh = 1))
98 Mar Garcia-Aloy et al.

Fig. 3 Signal for deprotonated ions of quercetin-3-glucoside and quercetin-3-

galactoside. Peak areas of identified chromatographic peaks are highlighted
with a rectangle

3.6 Preprocessing: 1. Choose the retention time correction algorithm among the
Retention Time available options (see Note 9). In this example, we use obiwarp
Correction and [26], which is based on Dynamic Time Warping, an algorithm
Grouping (See Note 12) used to find the best stretching of the time dimension of two
time series to make them as similar as possible [27]. LC-MS
data are multidimensional so in obiwarp the similarity is deter-
mined based on the full spectral information.
2. Determine the more relevant parameters of the algorithm. In
obiwarp, the relevant parameters are the choice of the reference
sample for the rt warping (center); the m/z bins used for
retrieving the spectra (profStep); the spectral similarity func-
tion (distFunc); and the penalizations on the warping opti-
mization (gapInit and gapExtend). With the default
parameters (which work reasonably well in most of the cases),
the reference sample is the one containing the largest number
of peaks. The retention time correction can be performed with
the following command:

xdata <- adjustRtime(xdata, param = ObiwarpParam())

3. In some experiments, it might be helpful to perform the align-

ment based on only a subset of the samples, e.g., if quality
control samples were injected at regular intervals or if the
experiment contains blanks. The parameter subset can be
used to define the subset of samples on which the alignment
of the full data set will be based on (see Note 9).
 Data Treatment for LC-MS Untargeted Analysis 99

Fig. 4 Outcome of the retention time correction performed on the MTBLS59 data
as obtained by plotAdjustedRtime (), after minor visual parameters
customization. Each line represents one sample and is colored by sample class.
The retention time correction is different for each sample and it is not linear in
time

4. Quality check: A global evaluation of the retention time cor-

rection procedure helps in detecting unwanted extreme defor-
mations of the original time scale.
Visually inspect the retention time correction. In xcms, this
can be done using the dedicated function:

plotAdjustedRtime(xdata)

The situation for the MTBLS59 data is presented in Fig. 4

which shows the retention time deviation (in seconds) for each
sample along the chromatographic retention time. Usually, the
first and the last part of the chromatography requires the
heaviest corrections. In the example, the greatest correction is
about 4 s. Considering that here we are dealing with UPLC,
this adjustment is acceptable and speaks of a satisfactory repro-
ducibility in the chromatography. As a rule of thumb, the
maximum correction should be comparable to the chro-
matographic peak width in that particular rt range.
5. Choose the “grouping” algorithm or, in other terms, the pro-
cedure which matches peaks with similar m/z and rt across
samples of an experiment (correspondence analysis; see Note
9). Among the possible choices in xcms, we rely on the default
density-based solution [15].
6. Choose the grouping parameters according to the analytical
pipeline. The peak density method used for the present exam-
ple iterates over slices along the m/z dimension (which size can
be configured using the parameter mzwid) and groups identi-
fied chromatographic peaks within each of these across sam-
ples, if their retention time is close. For that, within each m/z
slice, the density distribution of the chromatographic peaks’
retention times is calculated and peaks are grouped into the
same feature if they fall within the same peak in this density
100 Mar Garcia-Aloy et al.

distribution curve. The density distribution of the peaks’ reten-

tion times is calculated using the base R density() function,
and its parameter bw sets the smoothing bandwidth to use:
larger values result in a strongly smoothed density distribution
that can result in peaks with different retention times to be
grouped into the same feature. Finally, parameter minFrac-
tion allows defining a proportion of samples (of at least one
sample class) in which a chromatographic peak needs to be
detected in order to define a feature. Again, we set the para-
meters as suggested in [26] and minFraction equal to 0.5:

xdata <- groupChromPeaks(xdata,

param = PeakDensityParam(
sampleGroups = xdata$group,
bw = 3,
binSize = 0.025,
minFraction = 0.5))

7. Quality check: Compounds known to be present in the samples

should be correctly detected and aligned on the whole dataset.
Inspect the Extracted Ion Traces (EICs) for compounds
known to be present in the samples. For these, the preproces-
sing should have identified chromatographic peaks and should
have defined a feature in the final list of features. Also, the
chromatographic peaks should ideally be nicely aligned after
preprocessing. In the case of the apple extract in MTBLS59,
possible metabolites to check are quercetin-3-glucoside and
quercetin-3-galactoside, both showing a strong ionic signal at
m/z 463.0870 (in negative ion mode) [19] at around 415 s.
The EICs in the vicinity of their peak and the position of the
corresponding feature(s) in the groups can be extracted with
the following commands:

mzr <- c(463.0882 + 0.01 * c(-1, 1))

rtr <- c(405, 430)
pdp <- processHistory(xdata, type = "Peak grouping")
pdp <- processParam(pdp[[1]])
myeic <- chromatogram(xdata, mz = mzr, rt = rtr)
plotChromPeakDensity(myeic, param = pdp)

The resulting plot is shown in Fig. 5. Looking at the upper

part of the plot, it is clear that the EICs are correctly aligned (see
Note 13). However, with the settings used for the correspon-
dence analysis, xcms merged peaks for both compounds into
the same feature (see Note 14).
In this example, ions from both compounds are then trea-
ted as a single entity, and this could depend on the specific
choice of the parameter bw, which is related to the acceptable rt
 Data Treatment for LC-MS Untargeted Analysis 101

Fig. 5 Quality check on the processing procedure using the quercetin-3-

glucoside and quercetin-3-galactoside in negative ionization mode, which are
expected to be present. Minor visual parameters customization has been
performed compared to the base command reported in the text. Samples are
colored by sample class. The upper panel shows the EIC for the m/z slice
corresponding to the ion [M-H]- of quercetin-3-glucoside and quercetin-3-
galactoside, whereas the lower panel shows the position of the identified
chromatographic peaks at their retention time (x-axis) and index within
samples of the experiments (y-axis), using the black line to represent the peak
density estimate and the gray rectangle(s) to show the defined feature(s) based
on the settings applied

shift for the same chromatographic peak across the complete

set of samples. bw is in fact one of the most important para-
meters in peak grouping using the peak density method. If
instead of using bw = 3, we apply bw = 1, the algorithm success-
fully groups the peaks into two different features (Fig. 6):

bw(pdp) <- 1
plotChromPeakDensity(myeic, param = pdp)

3.7 Preprocessing: It could happen that peaks belonging to a subset of samples are
Imputation of Missing missing from “their” groups. This is correct if the corresponding
Peaks analyte is not present in a specific sample, but it could also be the
result of a failed identification of the chromatographic peak during
peak detection (e.g., if the signal in that particular sample was too
low or too noisy). These missing peaks result in missing values in
the final feature data matrix, which are usually tricky to handle
during statistical analysis. xcms provides fillChromPeaks(), a
102 Mar Garcia-Aloy et al.

Fig. 6 Peak grouping with reduced bw setting on m/z slice and rt window
containing signals from quercetin-3-glucoside and quercetin-3-galactoside in
negative ionization mode

function to rescue such signals by integrating all intensities

measured by the MS instrument in samples in which no peak was
detected in the m/z and rt area in which a signal for the respective
feature (ion) is expected. This area is defined using the m/z and rt
areas of the chromatographic peaks in samples where signal for the
feature was detected. If the peak was wrongly missed, fillChrom-
Peaks() will recover it, otherwise that missing value will be
replaced with the sum of the noise in the respective m/z and rt
area. Subsequently, one of the many data imputation approaches
common also to other omic technologies can be used to handle still
missing values, if needed.
1. Run the imputing function issuing:

xdata <- fillChromPeaks(xdata,

param = ChromPeakAreaParam())

3.8 Data Matrix Extract the final two-dimensional data matrix with samples on the
Extraction rows and features on the columns, using as intensity the integral of
the area under the peak:

dt <- t(featureValues(xdata,
value = "into",
method = "sum"))
 Data Treatment for LC-MS Untargeted Analysis 103

The m/z and rt details about the features can be extracted using
the featureDefinitions() function:

ft <- featureDefinitions(xdata)

Alternatively, the full results can also be extracted as a Summar-

izedExperiment using the quantify() function. The Sum-
marizedExperiment is a standard container for biological data
in Bioconductor and would allow an easier integration with other
Bioconductor packages.

se <- quantify(xdata, method = “sum”)

3.9 Annotation of At this point, features are only defined by a specific m/z and rt
MS-Based value, which does not provide sufficient information to make sense
Metabolomics Data of the generated data. Only by assigning a name (identity) to the
signals of interest (annotation) enables the biological interpretation
of the obtained results. Several software tools for metabolite anno-
tation and identification are available [28]. Here we use the Meta-
boAnnotation R package [18] because it provides a customizable
infrastructure that can be easily applied to xcms output. MetaboAn-
notation provides a set of functions that allow matching between
query and target entities (such as m/z and/or rt values).
1. Create a reference database containing theoretical values. Such
reference values are usually defined by measuring pure stan-
dards on the same LC-MS setup.

target_df <- data.frame(

name = c("quercetin-3-galactoside",
"quercetin-3-glucoside"),
mz = 463.0882,
rt = c(412, 420))

2. Match the measured m/z and rt values against reference data-

base and extract the full matching table:

library(MetaboAnnotation)
matched_feat <- matchValues(ft, target_df,
param = MzRtParam(tolerance = 0.005,
ppm = 0, toleranceRt = 5),
rtColname = c("rtmed", "rt"),
mzColname = c("mzmed", "mz"))
matchedData(matched_feat)[whichQuery(matched_feat), ]
104 Mar Garcia-Aloy et al.

which produces the following output:

mzmed rtmed target_name

FT322 463.0871 413.6519 quercetin-3-galactoside
FT323 463.0871 418.2810 quercetin-3-glucoside

4 Notes

1. Raw data are usually stored in vendor proprietary formats that

can generally only be used by proprietary, vendor-specific,
software. A close data format hampers data comparison and
sharing, which is instead made possible by adopting “open”
data formats with publicly available specifications. Examples of
open formats available for LC-MS are mzML [29], which is
supported by the Proteomics Standards Initiative, mz5 [30],
mzDB [31], and netCDF.
2. Sample metadata often mirror the factors included in the exper-
imental design (type of treatment, gender, etc.) but they also
include key characteristics of the samples. They have to be
considered during the design of the analytical workflow, in
order to ensure proper randomization of the samples [32–38].
3. The use of a standardized language and/or ontologies is essen-
tial to profit as much as possible from the experiment, allowing
result comparison, integration, and sharing [4, 12]. The com-
pliance with ISA-standards is also a requirement for submitting
experimental data (both raw data and metadata) to Metabo-
Lights, an open-access and general purpose repository for
metabolomics data [39].
4. It is worth mentioning that the data conversion to an open data
format is an optional step. In particular, it is not mandatory
when the analyses are performed with the vendor software. On
the contrary, it is usually necessary in most cases when an open
source solution is preferred. It is important to know, though,
that not all the relevant analytical metadata are usually included
in the open source version of the data (such as the analytical
column temperature). To avoid information loss, it is then
necessary to store a copy of the original “closed” files which
can be inspected only with the vendor software. This fact has to
be taken into consideration in the planning of the data storage
infrastructure.
5. Examples of data conversion can be found in [9].
6. The list of open and closed formats handled by ProteoWizard is
available on the project website (http://proteowizard.
sourceforge.net/tools.shtml).
 Data Treatment for LC-MS Untargeted Analysis 105

7. To perform the conversion, the vendor-specific Windows

Dynamic Link Libraries (DLLs) are needed for reading the
raw data files and the sample metadata. Such DLLs are
distributed either as part of the specific vendor software or
within the Windows version of the ProteoWizard suite. Also,
an official docker image including all required DLLs for con-
version is available (https://hub.docker.com/r/chambm/
pwiz-skyline-i-agree-to-the-vendor-licenses) which allows con-
version also on non-Windows systems.
8. Automatic peak detection usually leads to sub-optimal solu-
tions when compared to visual inspection (e.g., it could happen
that a peak is instead just noise). Nevertheless, visual inspection
will introduce subjectivity and not reproducible results (besides
the fact that it is a very time consuming task). The advantage of
automatic systems is that they always apply the same
“approach” impartially. Thus given the “approach” (or, more
precisely, the algorithm and its parameters), the outcome of the
procedure is determined. The algorithms used to perform all
the preprocessing tasks should be flexible, but an excess of
flexibility can produce unwanted artefacts (wrongly matched
features). Therefore, to get the maximum from the data, it is
necessary to match this flexibility with the real analytical char-
acteristics of the experiment (which should also be optimized!).
Given the importance of the parameter tuning phase, auto-
matic optimization tools have also been developed [40], but
it is nevertheless strongly suggested to also manually evaluate
the results of the peak detection step on at least a set of
compounds known to be present in the analyzed samples.
9. The characteristics of the different algorithms are discussed in
the xcms vignette and manual, both available from the Biocon-
ductor website (https://bioconductor.org/packages/release/
bioc/html/xcms.html).
10. Since the chosen values for the pick picking parameters will be
applied to the whole dataset, it is important that the reference is
representative of the sample complexity. The more representa-
tive the sample is, the better the final outcome of the analysis
will be. If quality control samples are a pool of samples, they are
usually a good choice.
11. The signature of a good peak picking is a structured organiza-
tion of points in the space defined by the retention time (rt) on
the abscissa and m/z on the ordinate. Since each chemical
produces several m/z’s signals at the same retention time, the
points should roughly show a pattern characterized by “vertical
stripes”. Dots spread uniformly in the plane or clear horizontal
patterns are the symptoms of a problem in peak picking or in
the chromatography itself. The number of detected peaks can
106 Mar Garcia-Aloy et al.

Fig. 7 Pick picking outcome on the MTBLS59 dataset: (left) the peak detection is performed using the default
settings for centWave, (right) or the set of parameters defined in Methods Subheading 3.5, step 5

also be an interesting indicator of a sensible choice of the

parameters, as shown in Fig. 7. The situation on the left
panel of Fig. 7 is obviously not satisfactory.
12. In some cases, an additional grouping step can be necessary
before retention time correction.
13. Actually a minor retention time correction has been
applied here: the compound elutes within 405 and 430 s
which correspond to a correction smaller than 2 s as can be
noted from Fig. 4. The EICs for the uncorrected measures can
be obtained using.

myeic <- chromatogram(raw_data, mz = mzr, rt = rtr)

14. The choice of the peak-picking algorithm and of its parameters

obviously limits the range of compounds that can be detected.
In many cases, the samples will contain analytes eluting with
diverse chromatographic peak shapes. An automatic algorithm
will never be able to tackle all the possible analytical situations.
The optimal preprocessing finds the “best” and less critical
compromise between flexibility and reliability. The case pre-
sented in Fig. 5 is particularly critical because the two isomers
elute almost at the same time and their chromatographic peaks
are not well separated. The presence of a single feature for these
distinct metabolites would make it impossible to rely on these
settings to study a biological process which affects the balance
between these two metabolites.
 Data Treatment for LC-MS Untargeted Analysis 107

References
1. Ardrey RE (2003) Liquid chromatography–mass 13. Kessner D, Chambers M, Burke R et al (2008)
spectrometry: an introduction. https://online ProteoWizard: open source software for rapid
l i b r a r y. w i l e y. c o m / d o i / b o o k / 1 0 . 1 0 0 2 proteomics tools development. Bioinformatics
/0470867299 24:2534–2536. https://doi.org/10.1093/bio
2. Patti GJ, Yanes O, Siuzdak G (2012) Innova- informatics/btn323
tion: metabolomics: the apogee of the omics 14. Chambers MC, MacLean B, Burke R et al
trilogy. Nat Rev Mol Cell Biol 13:263–269. (2012) A cross-platform toolkit for mass spec-
https://doi.org/10.1038/nrm3314 trometry and proteomics. Nat Biotechnol
3. Alonso A, Marsal S, Julià A (2015) Analytical 3010(30):918–920. https://doi.org/10.
methods in untargeted metabolomics: state of 1038/nbt.2377
the art in 2015. Front Bioeng Biotechnol 3:23. 15. Smith CA, Want EJ, O’Maille G et al
https://doi.org/10.3389/fbioe.2015.00023 (2006) XCMS: processing mass spectrometry
4. Sansone S-A, Rocca-Serra P, Field D et al data for metabolite profiling using nonlinear
(2012) Toward interoperable bioscience data. peak alignment, matching, and identification.
Nat Genet 44:121–126. https://doi.org/10. Anal Chem 78:779–787. https://doi.org/10.
1038/ng.1054 1021/ac051437y
5. R Core Team (2016) R: a language and envi- 16. Tautenhahn R, Bottcher C, Neumann S
ronment for statistical computing. https:// (2008) Highly sensitive feature detection for
www.r-project.org/ high resolution LC/MS. BMC Bioinformatics
6. De Vos RCH, Moco S, Lommen A et al (2007) 9. https://doi.org/10.1186/1471-2105-
Untargeted large-scale plant metabolomics 9-504
using liquid chromatography coupled to mass 17. Benton HP, Want EJ, Ebbels TMD (2010)
spectrometry. Nat Protoc 2. https://doi.org/ Correction of mass calibration gaps in liquid
10.1038/nprot.2007.95 chromatography-mass spectrometry metabolo-
7. Rafiei A, Sleno L (2014) Comparison of peak- mics data. Bioinformatics 26. https://doi.org/
picking workflows for untargeted liquid 10.1093/bioinformatics/btq441
chromatography/high-resolution mass spec- 18. Rainer J, Vicini A, Salzer L et al (2022) A
trometry metabolomics data analysis. Rapid modular and expandable ecosystem for meta-
Commun Mass Spectrom 29. https://doi. bolomics data annotation in R. Metabolites 12:
org/10.1002/rcm.7094 1 7 3 . h t t p s : // d o i . o r g / 1 0 . 3 3 9 0 /
8. Yu T, Park Y, Johnson JM, Jones DP (2009) METABO12020173
apLCMS-adaptive processing of high- 19. Franceschi P, Masuero D, Vrhovsek U et al
resolution LC/MS data. Bioinformatics 25. (2012) A benchmark spike-in data set for bio-
https://doi.org/10.1093/bioinformatics/ marker identification in metabolomics. J Che-
btp291 mom 26. https://doi.org/10.1002/cem.
9. Tauler R, Gorrochategui E, Jaumot J, Tauler R 1420
(2015) A protocol for LC-MS metabolomic 20. Zhang J, Gonzalez E, Hestilow T et al (2009)
data processing using chemometric tools. Pro- Review of peak detection algorithms in liquid-
toc Exch. https://doi.org/10.1038/protex. chromatography-mass spectrometry. Curr
2015.102 Genomics 10. https://doi.org/10.2174/
10. Katajamaa M, Orešič M (2007) Data proces- 138920209789177638
sing for mass spectrometry-based metabolo- 21. Åberg KM, Alm E, Torgrip RJO (2009) The
mics. J Chromatogr A 1158:318–328 correspondence problem for metabonomics
11. Lange E, Tautenhahn R, Neumann S, Gröpl C datasets. Anal Bioanal Chem 394:151–162
(2008) Critical assessment of alignment proce- 22. Smith R, Ventura D, Prince JT (2015) LC-MS
dures for LC-MS proteomics and metabolo- alignment in theory and practice: a comprehen-
mics measurements. BMC Bioinformatics 9: sive algorithmic review. Brief Bioinform 16:
375. https://doi.org/10.1186/1471-2105- 104–117. https://doi.org/10.1093/bib/
9-375 bbt080
12. Rocca-Serra P, Salek RM, Arita M et al (2015) 23. Koch S, Bueschl C, Doppler M et al (2016)
Data standards can boost metabolomics MetMatch: a semi-automated software tool for
research, and if there is a will, there is a way. the comparison and alignment of LC-HRMS
Metabolomics 12:14. https://doi.org/10. data from different metabolomics experiments.
1007/s11306-015-0879-3 Metabolites 6. https://doi.org/10.3390/
metabo6040039
108 Mar Garcia-Aloy et al.

24. Gatto L, Gibb S, Rainer J (2021) MSnbase, Mol Cell Proteomics 14. https://doi.org/10.
efficient and elegant R-based processing and 1074/mcp.O114.039115
visualization of raw mass spectrometry data. J 32. Krzywinski M, Altman N (2014) Points of sig-
Proteome Res 20:1063–1069. https://doi. nificance: designing comparative experiments.
org/10.1021/acs.jproteome.0c00313 Nat Methods 11:597–598
25. Brodsky L, Moussaieff A, Shahaf N et al (2010) 33. Krzywinski M, Altman N (2014) Points of sig-
Evaluation of peak picking quality in LC-MS nificance: analysis of variance and blocking. Nat
metabolomics data. Anal Chem 82:9177– Methods 11:699–700
9187. https://doi.org/10.1021/ac101216e 34. Krzywinski M, Altman N, Blainey P (2014)
26. Forsberg EM, Huan T, Rinehart D et al (2018) Points of significance: nested designs. Nat
Data processing, multi-omic pathway Methods 11. https://doi.org/10.1038/
mapping, and metabolite activity analysis nmeth.3137
using XCMS online. Nat Protoc 13:633–651. 35. Krzywinski M, Altman N (2014) Points of sig-
https://doi.org/10.1038/nprot.2017.151 nificance: two-factor designs. Nat Methods 11:
27. Prince JT, Marcotte EM (2006) Chro- 1187–1188
matographic alignment of ESI-LC-MS proteo- 36. Altman N, Krzywinski M (2014) Points of sig-
mics data sets by ordered bijective interpolated nificance: sources of variation. Nat Methods
warping. Anal Chem 78. https://doi.org/10. 12:5–6
1021/ac0605344
37. Blainey P, Krzywinski M, Altman N (2014)
28. Stanstrup J, Broeckling C, Helmus R et al Points of significance: replication. Nat Meth-
(2019) The metaRbolomics toolbox in Bio- ods 11:879–880
conductor and beyond. Metabolites 9:200.
https://doi.org/10.3390/metabo9100200 38. Altman N, Krzywinski M (2015) Points of sig-
nificance: split plot design. Nat Methods 12:
29. Martens L, Chambers M, Sturm M et al (2011) 165–166
mzML – a community standard for mass spec-
trometry data. Mol Cell Proteomics 10: 39. Haug K, Salek RM, Conesa P et al (2013)
R110.000133 MetaboLights – an open-access general-pur-
pose repository for metabolomics studies and
30. Wilhelm M, Kirchner M, Steen JAJ, Steen H associated meta-data. Nucleic Acids Res 41.
(2012) mz5: space- and time-efficient storage https://doi.org/10.1093/nar/gks1004
of mass spectrometry data sets. Mol Cell Pro-
teomics 11. https://doi.org/10.1074/mcp. 40. Libiseller G, Dvorzak M, Kleb U et al
O111.011379 (2015) IPO: a tool for automated optimization
of XCMS parameters. BMC Bioinformatics 16:
31. Bouyssié D, Dubois M, Nasso S et al 118. https://doi.org/10.1186/s12859-015-
(2015) MzDB: a file format using multiple 0562-8
indexing strategies for the efficient analysis of
large LC-MS/MS and SWATH-MS data sets.
 Chapter 6

Untargeted Metabolic Phenotyping by LC-MS

Ian D. Wilson and Elizabeth Want

Abstract
Untargeted analysis by LC-MS is a valuable tool for metabolic profiling (metabonomics/metabolomics),
and applications of this technology have grown rapidly over the past decade. LC-MS offers advantages of
speed, sensitivity, relative ease of sample preparation, and large dynamic range compared to other platforms
in this role. However, like any analytical approach, there are still drawbacks and challenges that have to be
overcome, some of which are being addressed by advances in both column chemistries and instrumentation.
In particular, the combination of LC-MS with ion mobility offers many new possibilities for improved
analyte separation, detection, and structural identification. There are many untargeted LC-MS approaches
which can be applied to metabolic phenotyping, and these usually need to be optimized for the type of
sample, the nature of the study, or the biological question. Some of the main LC-MS approaches for
untargeted metabolic phenotyping are described in detail in the following protocol.

Key words LC-MS, Mass spectrometry, Ion mobility, Liquid chromatography, Untargeted metabolic
profiling

1 Introduction

Metabolic profiling, or phenotyping, is the determination of the low

molecular weight metabolites, usually defined as <1 kDa, present in
biological samples. This can be performed on biofluids (e.g., urine,
serum), tissues (e.g., brain, liver etc.), plants, cells or culture media,
etc. These measurements can offer valuable mechanistic insights
into the response of biological systems (microorganisms, plants,
and animals) to drugs or toxins, disease diagnosis and progression,
as well as the effects of aging, diet, stress, and exercise on an
organism [1–3].
The metabolite profiles generated for biological samples are
generally complex, and currently, no single analytical approach
will provide comprehensive metabolome coverage. Typically, a
combination of analytical techniques, including NMR spectros-
copy, and mass spectrometry combined with separation approaches
(mainly GC-MS and LC-MS) will be needed to ensure

Olga Deda et al. (eds.), Metabolic Profiling: Methods and Protocols, Methods in Molecular Biology, vol. 2891,
https://doi.org/10.1007/978-1-0716-4334-1_6,
© The Author(s), under exclusive license to Springer Science+Business Media, LLC, part of Springer Nature 2025

109
110 Ian D. Wilson and Elizabeth Want

Fig. 1 Improved spectral quality due to enabling IMS for two co-eluting lipid species. The elevated collision
energy spectra of these co-eluting lipids (DG [36:5] and SM [32:1]) without the incorporation of IM separation
(a and b) contain the same collection of fragment ions. IM separation prior to CID produces cleaner elevated
collision energy spectra (c and d) improving assignment of fragment ions to the lipid precursor ions increasing
confidence in possible lipid identifications. (From King et al. [32]. With permission)

comprehensive metabolome coverage [4–6]. Improvements in

instrument sensitivity and resolution, as well as expanding data-
bases and software that, if used with care, can help in the structural
identification of metabolites, mean that ever greater coverage of the
metabolome can now be achieved [7–9].
When used properly, LC combined with high accuracy MS/MS
provides a sensitive means of metabolic profiling. Separation of
metabolites in a sample prior to mass spectrometric (MS) analysis
helps to reduce the complexity of the mixture entering the ion
source and, as a result, the risk of ion suppression. Chro-
matographic separation can also aid in discriminating between
many isobaric species. This serves to improve the sensitivity and
quantification of metabolites, as well as to aid structural identifica-
tion. The recent introduction of ion mobility (IM) systems, which
can provide an orthogonal second dimension of separation, when
combined with LC-MS/MS in an LC-IM-MS/MS configuration
can considerably improve the spectral data obtained [10] (Fig. 1).
The improvements in the resulting MS spectra, plus the retention
time (tR) and the IM-derived collision cross-section data, can
greatly ease the task of the structural elucidation of potential bio-
markers. The breadth of LC column chemistries and types of mass
spectrometers available means that LC-MS is an immensely valu-
able approach for untargeted metabolic profiling and also targeted
metabolite analyses. Further, the advent of ultra (high) perfor-
mance liquid chromatography (UPLC or UHPLC) has further
improved chromatographic resolution and hence metabolite cover-
age [10–13]. Here, we will focus on the application of LC-MS for
the untargeted analysis of biofluids and tissues.
Unlike targeted LC-MS-based metabolite analyses, which tend
to focus on either particular classes of molecules or metabolic path-
ways, often providing sensitive, accurate, and quantitative measure-
ments for the targeted analytes [14, 15], untargeted methods
enable the investigator to “prospect” for novel/unexpected
 Metabolic Phenotyping by LC-MS 111

changes in metabolic profiles. These untargeted analyses can prove

to be invaluable for hypothesis generation and the discovery of
novel biomarkers [16]. The key to these methods is that they aim
to employ (a) unbiased sample preparation, which does not favor
the extraction of any a particular type or class of metabolite, and
(b) to the extent possible, they offer a method that allows com-
pounds from a range of classes and polarities to be profiled using
appropriate analyses.
The most common approach to the initial untargeted profiling
of biofluids and tissues in metabolic phenotyping is reversed-phase
(RP) LC-MS [17–21]. RP LC-MS enables the separation and
detection of a wide range of moderately polar to nonpolar meta-
bolites. However, a different approach is required in order to gain
access to the more polar metabolites present in many biofluids and
tissue samples which are either unretained or elute close to the
solvent front in RPLC. Typically, hydrophilic interaction liquid
chromatography (HILIC) is the first choice for this type of mole-
cule, and the number of applications of this type of separation has
increased greatly in metabolic profiling studies over the last decade
[22, 23]. The mode of separation in HILIC resembles that tradi-
tionally associated with normal phase LC, with polar compounds
more retained than apolar ones. This results in better retention and
detection of polar analytes than seen with RPLC. Thus, HILIC and
RPLC provide separations that are complementary and are often
used in parallel for metabolic profiling studies in order to provide
more comprehensive coverage of the metabolome. Currently, ultra-
high performance liquid chromatography (UHPLC) methods,
based on the use of sub 2 μm particles and higher pressures than
conventional HPLC, are favored for metabolic phenotyping.
The typical mass spectrometers of choice for these untargeted
LC-MS analyses are time-of-flight (ToF), quadrupole-time-of-
flight (Q-ToF), or “Orbitrap” instruments. Such instruments can
provide high mass resolution and mass accuracy, high sensitivity,
fast scan speeds, and a good dynamic range. The ability of Q-ToF
and orbitrap instruments to acquire MS/MS spectra, which can
greatly aid the identification of possible biomarkers, makes these
instruments invaluable when, as in the case of untargeted metabolic
profiling, the identity of the metabolites is largely unknown (unlike
targeted metabolomics).
Here, we provide details of untargeted LC-MS protocols
designed for the analysis of biological mammalian samples using
HILIC for polar metabolites and RPLC to cover mid- to nonpolar
compounds. Methods for the extraction of metabolites from dif-
ferent biological matrices are also described. The importance and
implementation of data quality assessment, such as through the use
of quality control (QC) samples, is introduced, but covered in more
detail in Chapter 3. The overall process to be followed in this type
of analysis is illustrated in the flow diagram shown in Fig. 2.
112 Ian D. Wilson and Elizabeth Want

Fig. 2 LC-MS untargeted workflow showing the main steps of the process. These steps are described in this
protocol

2 Materials

2.1 Samples In metabolic profiling, biological samples are commonly analyzed,

e.g., urine, serum/plasma, tissue, cell extracts, and cell culture
media (see Notes 1–3).

2.2 Standards and 1. System suitability/test mix

Chemicals (a) It is advisable to include a mixture of compounds, which
can be termed “System suitability mix or test mix” (see
Subheading 3.3) in order to assess chromatographic and
mass spectrometric performance. This can be a commer-
cially sourced test mixture, or one designed in-house to
encompass the metabolite classes of interest.
2. Metabolite extraction solvents
(a) Plasma/serum: methanol, acetonitrile (isopropyl alcohol
is often used if lipidomic analysis is to be undertaken).
(b) Urine (HILIC analysis): acetonitrile, methanol.
(c) Tissue: methanol/dichloromethane/methyl tertiary butyl
ether/water.
 Metabolic Phenotyping by LC-MS 113

3. UPLC-MS mobile phases

(a) LC-MS grade water, acetonitrile, methanol,
isopropanol—exact compositions depend on nature of
analysis (see Subheading 3.3; LC gradients) (see Note 15).
(b) Mobile phase additives: formic acid, ammonium acetate,
ammonium formate.
4. Lockmass and calibration solutions: leucine encephalin,
sodium formate.

2.3 Common 1. Pipettes.

Equipment 2. Pipette tips.
3. Eppendorf tubes.
4. Bead beater polypropylene tubes.
5. MS vials.
6. MS well plates.
7. Sealing cap mats.
8. Glass bottles.

2.4 LC-MS Systems 1. HPLC or UHPLC chromatography systems, e.g., Acquity

Ultra Performance LC.
2. Analytical columns—specific to chromatographic method, e.g.,
C18, C8, HILIC, HSS.
3. Mass spectrometer, e.g., Q-ToF, Orbitrap or ion mobility-
enabled MS, e.g., Synapt or TIMS TOF etc., if IM is used.

2.5 Other 1. Bench top centrifuge.

Instrumentation 2. Precellys bead beater or similar.
3. Vacuum evaporator.
4. Ultrasonic bath.
5. Vortex mixer.

2.6 Data Processing 1. Vendor supplied software or freeware (e.g., XCMS, MZMine
Software 3, MSDial) for pre-processing of raw data. This includes peak
detection, alignment, and normalization (Subheading 3.6).
2. R, Python and associated software packages.
3. Excel or similar.
4. SIMCA, MATLAB, or similar for multivariate analysis.
5. PRISM or JMP or similar for univariate analysis.
114 Ian D. Wilson and Elizabeth Want

3 Methods

LC-MS-based untargeted analysis can be divided into multiple key

steps (Fig. 1) as follows: sample preparation, chromatographic
separation (plus the option of IM if available), and mass spectro-
metric detection, followed by data analysis. Sample preparation to
extract metabolites is dependent on sample type. The most com-
mon samples used in untargeted LC-MS analyses for e.g., animals
are biofluids, e.g., urine, serum, plasma, followed by tissue (liver,
heart, brain etc.), feaces, cell extracts, and cell culture media. Meta-
bolites are extracted from these samples and the extracts analyzed
by HPLC-MS or UHPLC-MS, often using Q-ToF or Orbitrap
mass spectrometers, ideally employing both positive and negative
electrospray ionization (ESI) to maximize metabolome coverage.
Key to sample preparation is the removal of particulates and pro-
teins from the samples, as these can affect chromatographic perfor-
mance and may even result in column blockage, reducing column
lifespan. Sample preparation for untargeted LC-MS analysis ideally
should not favor specific types of molecules but rather should aim
for a broad extraction of metabolite classes, unless there is a specific
interest in a particular class of molecules (e.g., lipids). It may be that
the preparation method extracts polar metabolites into an aqueous
fraction and nonpolar metabolites into an organic fraction, which
are then analyzed using separate chromatographic methods, as in
the case of tissue samples. This has benefit of reducing sample
complexity and may improve detection, quantification, and
identification.
Important aspects of an untargeted LC-MS analysis are as
follows:

System Suitability or Test Samples Typically, 10–15 compounds

will be used in a system suitability mix, which will be run at the start
of the analysis (see step 9, Subheading 3.2) as a “system suitability
test” to assess instrument performance (e.g., chromatographic peak
shape and retention time, mass accuracy, and detector response)
before the analysis of study samples. Parameters such as detector
response are prone to change over the analytical run, and so this
mixture can also be run at the end of the sample analysis to assess
these changes. Ideally, the compounds chosen for this mixture will
span a range of metabolite classes and chromatographic retention
times and may be tailored to the specific biological sample being
analyzed.

Quality Control Samples Quality control (QC) samples are the

key to successful, robust untargeted LC-MS analyses. These sam-
ples are representative of the study sample set and are usually made
by mixing small aliquots (e.g., 10–50 μL) to produce a “pooled”
 Metabolic Phenotyping by LC-MS 115

Fig. 3 Selected portion of a BPI chromatogram showing serum QC samples overlaid. Inset is an example of
monitoring ion intensity of a specific metabolite over the QC samples

QC from all study samples. QC samples can be/are used for

(a) conditioning/passivation of the LC column prior to the run
and (b) assessment of the reproducibility/stability of the run and
resulting data quality, and (c) correction of (signal) drifts within and
between batches. Subheading 3.2 step 10 describes the setup for
QC samples in the run. For a more detailed description, the reader
is directed to Chapter 3. The type of data that can be obtained from
the QC samples to support the quality of data arising from an
untargeted assay is shown in Fig. 3. Here the total ion current
mass chromatograms for a number of QCs are overlaid, and a
particular ion has been isolated to show the lack of variation in
peak shape and retention time over the time course of the study.

Sample Randomization Study samples should be randomized

within a batch to avoid bias which may arise from changes in the
system over the course of the run, e.g., decrease in detector sensi-
tivity, drifts in chromatographic retention times, or mass accuracy.
Samples can be randomized using in-house scripts, online software
(https://www.randomizer.org/; https://www.random.org/lists/
), or a randomized block design in the case of larger studies. For a
more detailed description, the reader is directed to Chapter 2.
116 Ian D. Wilson and Elizabeth Want

Sample Preparation
1. Urine [24]
(a) RP LC-MS.
– Aliquot an appropriate volume of urine, e.g.,
50–100 μL into a labeled Eppendorf tube.
– Dilute urine sample with appropriate volume of
LC-MS grade water (1:1 for human samples, 1:3 for
rodent and canine samples).
– Centrifuge at 12,000×g for 10 min to remove
precipitate.
– Aliquot sample into clean Eppendorf tube.
– Prepare quality control sample as described in step 6.
– -Aliquot appropriate volume of sample into well plate
or MS vials.
(b) HILIC LC-MS.
– Aliquot appropriate volume of urine, e.g., 50–100 μL
into a labeled Eppendorf tube.
– Dilute urine sample with appropriate volume of aceto-
nitrile (1:1 for human samples, 1:3 for rodent and
canine samples).
– Centrifuge at ~13,000×g for 10 min to remove
precipitate.
– Aliquot sample into clean Eppendorf tube.
– Prepare quality control sample as described in step 6.
– Aliquot appropriate volume of sample into well plate or
MS vials.
2. Serum/Plasma
(a) The key to extracting metabolites from serum/plasma is
the precipitation of proteins. Here, a simple method
involving protein precipitation with methanol is
suggested [25].
Method 1: Preparation for RP LC-MS using methanol
extraction
(b) Aliquot an appropriate volume of serum/plasma into a
labeled Eppendorf tube.
(c) Add cold methanol in a ratio of at least 3:1, e.g., 150 μL
methanol: 50 μL plasma, and vortex for 15 s.
(d) Leave at -20 °C for at least 20 min.
(e) Centrifuge at 14–16,000×g for 10 min.
(f) Remove supernatant to clean labeled Eppendorf tube.
 Metabolic Phenotyping by LC-MS 117

(g) Dry in vacuum concentrator.

(h) Redissolve in water or RP starting mobile phase.
3. Tissue Sample Preparation for RP and HILIC Analysis
[20, 26]
1. Weigh 50 mg (+/- 5 mg) tissue section and place in
(a) 2 mL polypropylene bead-beater tubes if using a Pre-
cellys bead beater (or similar)
(b) 2 mL Eppendorf tubes if using a Qiagen Tissue lyser
(or similar)
Keep tubes on ice before, during, and after the extraction
process.
2. Homogenize the sample with 1.5 mL prechilled methanol/
water (1:1) using either a Precellys bead beater, or Qiagen
Tissue lyser following the instructions below. The number
of bead-beating cycles will vary depending on the tissue type
(see Notes 10–11).

Method Requirements Instruction

Precellys Add 1 mm zirconium 40 s per cycle, cool on ice; 40 s
bead beads—100 μL in tube for subsequent cycles
beater
Qiagen Stainless steel beads 25 Hz speed, 5 min cycle
tissue
lyser

3. Centrifuge the sample at 4 °C for 10 min at 10,000×g.

4. To prepare the aqueous extracts, aliquot the supernatant into
a clean Eppendorf tube. Retain the pellet in the bead-
beating tube on ice for subsequent organic extraction (see
step 9).
5. Take separate 250 μL aliquots for RPLC and HILIC
analyses.
6. Take 150 μL of each sample to form the QC sample.
7. Dry aqueous extracts in a Savant vacuum concentrator for
~180 min at 45 °C using the V-AQ (vacuum
aqueous) mode.
8. The sample can then be treated in one of the following ways:
(a) Resuspension in starting mobile phase or similar and
immediate analysis.
(b) Storage at -40 °C or lower until analysis.
9. To prepare the organic extracts, add 1.5 mL prechilled solu-
tion of dichloromethane/methanol (1:3) to the pellet from
step 4.
118 Ian D. Wilson and Elizabeth Want

10. Homogenize the sample using either a Precellys bead

beater, or Qiagen Tissue lyser, following the instructions
in step 2 (see Notes 11–12).
11. Take separate 250 μL aliquots for RPLC (lipid) and HILIC-
MS analyses.
12. Take 150 μL of each sample to form the QC sample.
13. Store at -40 °C or lower until analysis.

3.1 Preparation of Prepare a fresh stock solution of sodium formate; 0.1 mg/mL in
LC-MS Calibration water. Add 1 mL of stock solution to 9 mL of isopropanol (IPA) to
Solution obtain a working solution of 0.01 mg/mL. This solution can be
stored at 4 °C for a number of weeks. Note that an alternative
calibration solution can be used instead, and many instrument
vendors will recommend one.

3.2 LC-MS Analysis 1. Transfer samples to LC-MS vials or polypropylene 96-well

plates. If using well plates, heat-seal foil lids or use polypropyl-
ene cap mats to reduce evaporation and contamination. If
samples have been stored in vials or well plates prior to analysis,
centrifuge at 1350×g for 5 min at 4 °C (ideally) or room
temperature.
2. Place vials or well plates in the LC autosampler. The autosam-
pler temperature (see Note 21) should be maintained at 4 °C
for all analyses except for lipid analysis, where the temperature
should be 8–10 °C to avoid precipitation of the lipids. Samples
should be stable in the autosampler for 24–48 h, but stability of
individual metabolites over this time period will naturally vary.
Stability of metabolites can be assessed in the different
biological sample types, e.g., by using the QC samples.
3. Create the sample analysis list. Choose the LC gradient and MS
settings as described in Subheadings 3.3 and 3.4, respectively.
Column temperature will be dependent on the analysis (e.g.,
column dimensions and mobile phase composition) and will
vary from 40 °C for urine samples to 55 °C for lipid analysis.
Column temperature has been optimized based on the mobile
phases employed for the LC gradient.
4. Select electrospray ionization mode.
5. Ensure that the instrument is set up for the optimum accurate
mass by infusing a solution of a reference compound such as
leucine encephalin into the mass spectrometer. Follow the
instrument-specific instructions for setup and acceptance
criteria.
6. Calibrate the instrument prior to analysis using a solution of
the appropriate concentration of sodium formate as described
in Subheading 3.1 or similar calibration solution. Follow the
 Metabolic Phenotyping by LC-MS 119

instrument-specific instructions for calibration and acceptance

criteria. Calibration points should be spread as evenly as possi-
ble over the mass range that is being acquired for the analysis
(typically 50–1000 or 1200 m/z for metabolic profiling). In
general, the residual mDa for each individual calibration point
should be <1.5 mDa, with the majority of calibration points
having residuals of <0.5 mDa. With most instruments, this can
be performed using a setup wizard or manually depending on
the preferences of the instrument operator.
7. If IM is to be used, ensure that the instrument has also been
calibrated according to the manufacturer’s instructions, i.e.,
using appropriate calibration solution and settings.
8. Start with a blank sample to assess the condition of the column.
This blank sample can be water or a solvent mixture similar to
the starting mobile phase conditions. If the LC column appears
“dirty,” i.e., shows a high background signal, then a sequence
of blank samples is recommended. Background signal levels can
be monitored until an acceptable level is reached.
9. Run the system suitability/test mixture(s) to assess that peak
intensity, retention times, and mass accuracy are within the
acceptance criteria (and similarly if, e.g., IM is also enabled
that the instrument is providing the appropriate expected
drift times and CCS values from the appropriate SST).
10. Inject pooled quality control (QC) samples (see Notes 17
and 21) in order to condition/passivate the column before
study sample analysis. At least 10 QC samples are recom-
mended for RP analysis of urine and plasma/serum. This may
need to be increased to ~15 QC samples for RP analysis of
tissue samples and for HILIC analysis in general. These QC
samples can be assessed by visually overlaying and monitoring
peak intensity, retention times, and mass accuracy. Condition-
ing QC samples can also be used to assess the appropriate
injection volume needed in order to obtain peaks that are not
overloaded and can act as a sample-based system suitability test
for that type of sampler. In addition, the pooled QC can be
used to assess sample dilution factors. An example of QC
sample chromatograms is shown in Fig. 3.
11. Add a QC dilution series to assess the linear response of ana-
lytes. Typically dilutions from the original QC concentration of
1/2, 1/4, 1/8, 1/16, etc. are used. Inject the QC dilution
series from low to high concentration at the start and end of
the analytical batch.
12. In addition, set up untargeted MS/MS experiments, e.g., data-
dependent acquisition (DDA) experiments using at least one
conditioning QC sample (see Subheading 3.5).
120 Ian D. Wilson and Elizabeth Want

13. Randomize the study samples, add to the sample analysis list,
and inject a QC sample every 5–10 samples. Aim for at least
10% QC samples per study batch in order to be able to perform
appropriate statistics. End the sample batch with a QC sample,
followed by a blank sample to assess carryover, the QC dilution
series, and a final system suitability/test mix injection to assess
system changes over the run.
14. Injection volume will depend on sample type and instrumenta-
tion, but usually ranges from 2 to 10 μL. Use the QC samples
to assess the appropriate injection volume.
15. When confident that the performance of the system is accept-
able, based on the data from the SST(s) and the conditioning
QC samples, start the analysis of the study samples, including
within-run QC samples.

3.2.1 LC Gradients 1. Example UHPLC gradients for urine, serum, and tissue sam-
ples by both RP and HILIC analysis are shown in Tables 1, 2, 3,
4, 5, 6, and 7. The column dimensions and details are also
provided. Note that these gradients will require optimization
for different column chemistries, lengths, and particle sizes.

3.3 MS Settings Example MS settings, including cone and capillary voltage, are
shown in Table 8 below. Note that these settings are for Waters
Xevo Q-ToF mass spectrometers and will require optimization for
each specific mass spectrometer. However, they can be used as
guidelines, based on an ultraperformance liquid chromatography
(UHPLC) setup with a flow rate of 400–500 μL/min, and a Q-ToF
mass spectrometer.

Table 1
Example gradient for the RP analysis of urine samples

Time (min) A (%) B (%) Comment

0 99 1
1 99 1
3 85 15
6 50 50
9 5 95
10 5 95 Column washing
10.1 99 1 Column re-equilibration
12 99 1
Taken from ref. 24
Column = UPLC Acquity HSS T3. Flow rate = 0.5 mL/min. Mobile phases; A = 0.1%
formic acid in water, B = 0.1% formic acid in acetonitrile
 Metabolic Phenotyping by LC-MS 121

Table 2
Example gradient for the HILIC analysis of urine samples

Time (min) A (%) B (%) Comment

0 99 1
1 99 1
12 0 100 Column washing
12.1 99 1 Column re-equilibration
15 99 1
Taken from ref. 24
Column = UPLC Acquity HILIC. Flow rate = 0.4 mL/min. Mobile phases; A = 95%
acetonitrile, 5% water + ammonium acetate (10 mM final concentration), B = 50%
acetonitrile, 50% water + ammonium acetate (10 mM final concentration)

Table 3
Example gradient for the RP analysis of plasma/serum samples

Time (min) A (%) B (%) Comment

0 99.9 0.1
2 99.9 0.1
6 75 25
10 20 80
12 10 90
21 0.1 99.9 Column washing
23 0.1 99.9
24 99.9 0.1 Column re-equilibration
26 99.9 0.1
Taken from ref. 25
Column = UPLC Acquity HSS T3. Flow rate = 0.4 mL/min. Mobile phases; A = 0.1%
formic acid in water, B = 0.1% formic acid in methanol

3.4 MS/MS DDA experiments: Tandem mass spectrometry experiments can be

Experiments set up whereby a specific number of precursor ions are selected
using predetermined rules and thresholds, such as exceeding a
certain ion intensity. These MS/MS experiments can provide infor-
mation which may aid in the structural identification of discrimina-
tory metabolites. Although it is often the case that further MS/MS
experiments need to be performed, once such potential biomarkers
have been selected from data analysis, performing DDA experi-
ments at this time does not impact negatively on the study setup.
122 Ian D. Wilson and Elizabeth Want

Table 4
Example gradient for the HILIC analysis of plasma/serum samples and
aqueous tissue extracts

Time (min) A (%) B (%) Comment

0 99 1
2 99 1
8 45 55
9 1 99
9.1 1 99 Flow rate increased to 0.8 mL/min
11 1 99 Column washing
11.1 99 1 Column re-equilibration
19 99 1
19.1 99 1 Flow rate decreased to 0.4 mL/min
23 99 1
Taken from ref. 26
Flow rate = 0.4 mL/min unless stated. Column = UPLC Acquity HILIC. Mobile
phases; A = acetonitrile/water (95:5), B = acetonitrile/water (50:50)

Table 5
Example gradient for the RP analysis of tissue samples—aqueous extract

Time (min) A (%) B (%) Comment

0 99.9 0.1
2 99.9 0.1
6 75 25
10 20 80
12 10 90
21 0.1 99.9 Column washing
23 0.1 99.9
24 99.9 0.1 Column re-equilibration
26 99.9 0.1
Taken from ref. 26
Column = UPLC Acquity HSS T3. Flow rate = 0.4 mL/min. Mobile phases; A = 0.1%
formic acid in water, B = 0.1% formic acid in methanol

3.5 Data It is beyond the scope of this protocol to provide a detailed descrip-
Preprocessing tion of data analysis. However, the key data analysis steps pertaining
to LC-MS untargeted analysis for metabolic profiling studies are
described briefly in steps 1–6. There are many different commercial
 Metabolic Phenotyping by LC-MS 123

Table 6
Example gradient for the RP analysis of tissue samples—organic extract

Time (min) A (%) B (%) Comment

0 60 40
2 57 43
2.1 50 50
12 46 54
12.1 30 70
18 1.0 99 Column washing
18.1 60 40 Column re-equilibration
20 60 40
Taken from ref. 26
Column = UPLC Acquity CSH. Flow rate = 0.4 mL/min. Mobile phases; A = acetoni-
trile (ACN): water (60:40), B = isopropanol: ACN (90:10). In both mobile phases
ammonium formate was diluted to 10 mM and formic acid to 0.1%

Table 7
Example gradient for the HILIC analysis of tissue samples—aqueous
extracts

Time (min) A (%) B (%) Comment

0 99 1
2 99 1
8 45 55
9 1 99
9.1 1 99 0.8 mL/min column washing
11 1 99 0.8 mL/min column washing
11.1 99 1 0.8 mL/min column washing
19 99 1 Column re-equilibration
23 99 1
Taken from ref. 26
Column = Temperature = 35 °C. Flow rate = 0.4 mL/min unless stated. Mobile phases;
A = acetonitrile (ACN): water (95:5), B = ACN: water (50:5). In both A and B, the
concentration of ammonium acetate is 10 mM and formic acid is present at 1%

and freeware available to perform some or all of these steps. These

are listed in the Materials section of this protocol.
1. The first step in data preprocessing of untargeted LC-MS data
is peak picking. Define chromatographic processing regions—it
may be that it is not desirable to include the solvent front, e.g.,
124 Ian D. Wilson and Elizabeth Want

Table 8
Example MS settings for positive and negative mode ESI analysis using a Q-ToF mass spectrometer

Parameter Setting
Capillary voltage 1–3 kV electrospray (ESI)+, 1–2.5 kV ESI-mode
Cone voltage 30 V
Source temperature e.g. 120 °C
Desolvation temperature e.g. 350 °C
Cone gas flow 25 L/h
Desolvation gas flow 900 L/h
The main parameters are shown. These will need to be optimized for each instrument and ESI mode and are to some
extent dependent on solvent system and mobile phase flow rate

0–1 min, and the re-equilibration portion of the data. Some

software may allow the user to omit these regions from further
analysis.
2. This is followed by alignment of the peaks to correct for any
retention shifts that have occurred between samples during the
analysis.
3. Peaks are then integrated using peak area through a selected
algorithm.
4. Data is often normalized during this process. A common
approach is to use median fold change [27] to account for
differences in sample dilutions. These are particularly common
in biological samples such as urine samples.
5. QC filtering is key in untargeted LC-MS metabolic profiling
studies. This is explained in Subheading 3.7.
6. Output of metabolite feature table for multivariate analysis.

3.6 Data Analysis Once the metabolite table has been produced, this can be exported
into appropriate software, e.g., excel for further analysis, e.g., QC
CV filtering (covered in more detail in Chapter 3).

3.7 Database Still a bottleneck in LC-MS untargeted analysis, structural identifi-

Searching for cation is key in metabolic profiling studies. Databases such as
Structural HMDB (http://www.hmdb.ca/), the Metabolomics Workbench
Identification Metabolite database:-(https://www.metabolomicsworkbench.
org/databases/metabolitedatabase.php) and LIPIDMAPS
(http://www.lipidmaps.org/) are growing in size as research
groups and individual researchers add to them. These databases
can be searched in a few ways in order to attempt to identify
candidate biomarkers, such as MS searches, MS/MS searches, or
text searches. Readers are directed to reviews in this area
 Metabolic Phenotyping by LC-MS 125

[28, 29]. However, such databases must be used with care, as many
of the database “hits” that are suggested based on, e.g., molecular
mass alone may, on close inspection, turn out to be biologically
implausible (e.g., pharmaceuticals, pesticides, or unlikely
phytochemicals) [30].
Structural identification is, initially, usually based on MS/MS
experiments that will, hopefully, result in the production of charac-
teristic fragmentation patterns. These fragments will provide high
mass accuracy information in addition to that obtained from the
MS spectrum alone. Such MS/MS spectra can then also be used in
database searches, helping to narrow the list of potential candidates
down from those suggested using MS data alone. In addition,
chromatographic properties and ion mobility-derived collision
cross-section data can be used to support (or disprove) putative
identifications. Untargeted MS/MS data (e.g., DDA experiments)
can be collected on QC conditioning samples or any QC sample
injected throughout the run.
The level of identification required also needs to be considered
against the suggested criteria of, e.g., the Metabolomics Society
Initiative (MSI) which offers a number of levels of increasing cer-
tainty of metabolite identity supported by experimental data [31].
Targeted MS/MS is usually performed on selected samples
containing the metabolite feature(s) of interest in high abundance.
These targeted assays would be performed after univariate or mul-
tivariate analysis has been performed and potential biomarkers
selected. Many of the commonly used databases allow for upload-
ing of MS/MS fragment ion information, e.g., HMDB and
METLIN. HMDB will output predicted MS/MS spectra which
can be visualized against the experimental MS/MS spectra
obtained during a study.

4 Notes

1. Check that the containers/vials used to collect and store sam-

ples are free of contaminants that may interfere with
subsequent analysis (and recheck when new stocks are
obtained). Aim to use the same batch of containers for a single
study, where possible.
2. Exercise caution when preparing samples from humans, as
there is a risk of infection. Samples should be handled and
prepared according to appropriate Biosafety protocols.
3. Check that any filtration devices that are used to remove parti-
culates from samples (e.g., syringe filters etc.) and are free of
contaminants that may interfere with subsequent analysis (and
recheck when new stocks are obtained).
126 Ian D. Wilson and Elizabeth Want

4. Tissues should be collected on to ice and snap frozen as soon as

possible after collection or processed for analysis immediately.
This is to avoid further metabolism from occurring or metabo-
lite losses.
5. Note that tissue samples can be prone to contamination from
sources such as anesthetics, surgical tubes, or instruments, e.g.,
cleaning solutions.
6. Tissue samples should be stored at the lowest possible temper-
ature, ideally -80 °C but otherwise -40 °C. This will help to
minimize further metabolite losses.
7. At the time of collection, sub-aliquot samples, including tissue
samples into appropriate storage containers to avoid freeze-
thaw issues.
8. If samples are to be transferred between sites, they should be
transported on dry ice if possible.
9. Exercise caution when handling solvents for sample prepara-
tion and the preparation of mobile phases, and check each
batch of these for interfering contaminants prior to use.
10. For the extraction of tissue samples, the volume of extraction
solvent stated in this protocol was optimized for 50 mg (+/-
5 mg) of tissue. This solvent volume can therefore be scaled up
or down depending on the amount of tissue to be processed. It
must be noted that the bead beater tubes or Eppendorf tubes
are 2 mL and so there is a risk of leakage through the lid of the
tube during extraction if >1.5 mL of liquid is placed into them
along with the tissue sample.
11. The number and speed of the bead beating cycles can be
adjusted according to the nature of the tissue. Thus, e.g.,
fibrous tissue (such as veins, gut or placenta samples) may
require cycles to be performed using higher frequency, e.g.,
3 cycles at 6500 rather than 4000 mHz.
12. There is also a risk that adding stainless steel beads to organic
solvents contained in Eppendorf may result in the tubes
degrading during tissue lysis, leading to damage and/or loss
of sample due to leakage.
13. It is important to use good quality, reinforced tubes for bead
beating; otherwise, there is risk of tube breakage in the Tissue
lyser, leading to damage to the instrument and/or loss of
sample.
14. Check all solvents (including water) for contaminants before
use, especially if suppliers/batches are different. Aim to use the
same batch of solvents for a single study.
15. When preparing the mobile phases, it is important to note that
the addition of salts, particularly to mobile phases with a high
 Metabolic Phenotyping by LC-MS 127

content of an organic modifier, will benefit from more careful

preparation. This is the case for mobile phases for both HILIC
and lipidomic analyses. The key to successful preparation is
to prepare solvents in advance, e.g., the day before sample
analysis commences and check that there are e.g., no precipi-
tates/cloudiness etc.
16. Sonicate the bottles containing the mobile phases during prep-
aration to ensure salts have dissolved (ideally in a sonicating
water bath at >40 °C). Leave the bottles at room temperature
for at least 30 min after sonication to assess whether precipita-
tion occurs. If precipitation is seen, repeat the sonication pro-
cedure until the mobile phases are clear.
17. QC samples are generally the result of pooling small aliquots of
the individual study samples. In some cases, the limited
volumes of the study samples make this approach impractical,
and in this situation, either an “in house” or commercially
prepared, external, QC sample of the same matrix could be
used, e.g., the NIST “human plasma 1950.” Although such
samples may not be an exact match to the study samples, they
offer a means of conditioning the column and assessing factors
such as instrument performance, peak shapes, and instrument
drift.
18. Column temperature should be optimized for mobile phase
composition—using methanol-based solvents results in higher
back pressures than acetonitrile as result of the former’s higher
viscosity. By increasing the column temperature (to, e.g.,
50–55 °C), the viscosity of methanol-based mobile phases
will be reduced resulting in lower column backpressures.
19. The storage of organic extracts in 96 well plates is best avoided
if possible, as contaminants may leach into the solvent and
affect the subsequent analysis. For organic extracts, sample
storage in glass MS vials is to be preferred. However, it must
be noted that lipids can interact with glass, which may be
dependent on, e.g., temperature and pH and so must be
assessed.
20. In the case of MS vials, it is best, where possible, to use vials and
caps from the same batch (there are 100 vials in a box). This is
because there may be inter-batch differences in terms of purity,
and these differences can affect the analysis. It is good practice
to screen new batches of vials and caps to ensure that they are fit
for purpose.
21. With studies of more than, e.g., >100 samples, the time
required for analysis can be more than 24 h. In this situation,
even if in an autosampler is maintained at ca. 4 °C, there is the
potential for less stable metabolites to undergo some
128 Ian D. Wilson and Elizabeth Want

degradation. It is therefore important that the analytical data

are scrutinized for this instability, which can be assessed by
examining, e.g., the QC data for trends indicating a decline
in intensity during analysis not shown by other components in
the sample. It may also be possible to use the system suitability
test mixture to assess stability.

References
1. Holmes E, Loo RL, Stamler J et al (2008) 95(22):8433–8442. https://doi.org/10.
Human metabolic phenotype diversity and its 1021/acs.analchem.2c04343
association with diet and blood pressure. 10. Letertre M, Munjoma MC, Slade SE et al
Nature 453(7193):396–400. https://doi. (2020) Metabolic phenotyping using UPLC–
org/10.1038/nature06882 MS and rapid microbore UPLC–IM–MS:
2. Newgard CB (2017) Metabolomics and meta- determination of the effect of different dietary
bolic diseases: where do we stand? Cell Metab regimes on the urinary metabolome of the rat.
25(1):43–56. https://doi.org/10.1016/j. Chromatographia 83:853–861
cmet.2016.09.018 11. Nassar AF, Wu T, Nassar SF, Wisnewski AV
3. Uthe H, van Dam NM, Hervé MR et al (2021) (2017) UPLC-MS for metabolomics: a giant
Chapter Six – A practical guide to implement- step forward in support of pharmaceutical
ing metabolomics in plant ecology and biodi- research. Drug Discov Today 22(2):463–470.
versity research. In: Pétriacq P, Bouchereau A https://doi.org/10.1016/j.drudis.2016.
(eds) Advances in botanical research, vol 98. 11.020
Elsevier, Amsterdam, pp 163–203 12. Zhao YY, Lin RC (2014) UPLC-MS(E) appli-
4. Saurina J, Sentellas S (2017) Strategies for cation in disease biomarker discovery: the dis-
metabolite profiling based on liquid chroma- coveries in proteomics to metabolomics. Chem
tography. J Chromatogr B 1044–1045:103– Biol Interact 215:7–16. https://doi.org/10.
111. https://doi.org/10.1016/j.jchromb. 1016/j.cbi.2014.02.014
2017.01.011 13. Plumb RS, Gethings LA, Rainville PD et al
5. Haggarty J, Burgess KE (2016) Recent (2023) Advances in high throughput LC/MS
advances in liquid and gas chromatography based metabolomics: a review. TrAC Trends
methodology for extending coverage of the Anal Chem 160:116954
metabolome. Curr Opin Biotechnol 43:77– 14. Siskos AP, Jain P, Römisch-Margl W et al
85. https://doi.org/10.1016/j.copbio.2016. (2016) Interlaboratory reproducibility of a tar-
09.006 geted metabolomics platform for analysis of
6. Markley JL, Brüschweiler R, Edison AS et al human serum and plasma. Anal Chem 89(1):
(2016) The future of NMR-based metabolo- 656–665
mics. Curr Opin Biotechnol 43:34–40. 15. Begou O, Gika HG, Wilson ID et al (2017)
https://doi.org/10.1016/j.copbio.2016. Hyphenated MS-based targeted approaches in
08.001 metabolomics. Analyst 142:3079–3100
7. Chen Y, Xu J, Zhang R, Abliz Z (2016) Meth- 16. Gika H, Virgiliou C, Theodoridis G et al
ods used to increase the comprehensive cover- (2019) Untargeted LC/MS-based metabolic
age of urinary and plasma metabolomes by phenotyping (metabonomics/metabolomics):
MS. Bioanalysis 8(9):981–997. https://doi. the state of the art. J Chromatogr B 1117:
org/10.4155/bio-2015-0010 136–147
8. Wang S, Jiang X, Ding R et al (2022) MS-IDF: 17. Dunn WB, Broadhurst D, Begley P et al (2011)
a software tool for nontargeted identification Human serum metabolome (HUSERMET)
of endogenous metabolites after chemical iso- consortium. Procedures for large-scale meta-
tope labelling based on a narrow mass defect bolic profiling of serum and plasma using gas
filter. Anal Chem 94(7):3194–3202. https:// chromatography and liquid chromatography
doi.org/10.1021/acs.analchem.1c04719 coupled to mass spectrometry. Nat Protocol
9. Liu Y, Zhang Y, Vennekens T et al (2023) 6(7):1060–1083. https://doi.org/10.1038/
MeRgeION: a multifunctional R pipeline for nprot.2011.335
small molecule LC-MS/MS data processing, 18. Gray N, Lewis MR, Plumb RS et al (2015)
searching, and organizing. Anal Chem High-throughput microbore UPLC-MS
 Metabolic Phenotyping by LC-MS 129

metabolic phenotyping of urine for large-scale 26. Want EJ, Masson P, Michopoulos F et al
epidemiology studies. J Proteome Res 14(6): (2013) Global metabolic profiling of animal
2714–2721. https://doi.org/10.1021/acs. and human tissues via UPLC-MS. Nat Protoc
jproteome.5b00203 8(1):17–32. https://doi.org/10.1038/nprot.
19. Wilson ID (2015) Metabolic phenotyping by 2012.135
liquid chromatography-mass spectrometry to 27. Veselkov KA, Vingara LK, Masson P et al
study human health and disease. Anal Chem (2011) Optimized preprocessing of ultra-
87(5):2519. https://doi.org/10.1021/acs. performance liquid chromatography/mass
analchem.5b00409. No abstract available spectrometry urinary metabolic profiles for
20. Vorkas PA, Isaac G, Anwar MA et al (2015) improved information recovery. Anal Chem
Untargeted UPLC-MS profiling pipeline to 83(15):5864–5872. https://doi.org/10.
expand tissue metabolome coverage: applica- 1021/ac201065j
tion to cardiovascular disease. Anal Chem 28. Nash WJ, Dunn WB (2019) From mass to
87(8):4184–4193. https://doi.org/10.1021/ metabolite in human untargeted metabolo-
ac503775m mics: recent advances in annotation of metabo-
21. Gika HG, Theodoridis GA, Plumb RS et al lites applying liquid chromatography-mass
(2014) Current practice of liquid spectrometry data. TrAC Trends Anal Chem
chromatography-mass spectrometry in meta- 120:115324. https://doi.org/10.1016/j.
bolomics and metabonomics. J Pharm Biomed trac.2018.11.022
Anal 87:12–25. https://doi.org/10.1016/j. 29. Bocker S (2016) Searching molecular structure
jpba.2013.06.032 databases using tandem MS data: are we
22. Spagou K, Wilson ID, Masson P et al (2011) there yet? Curr Opin Chem Biol 36:1–6.
HILIC-UPLC-MS for exploratory urinary https://doi.org/10.1016/j.cbpa.2016.
metabolic profiling in toxicological studies. 12.010
Anal Chem 83(1):382–390. https://doi.org/ 30. Theodoridis G, Gika H, Raftery D et al (2023)
10.1021/ac102523q Ensuring fact-based metabolite identification
23. Virgiliou C, Sampsonidis I, Gika HG et al in liquid chromatography–mass spectrometry-
(2015) Development and validation of a based metabolomics. Anal Chem 95(8):
HILIC-MS/MS multi-targeted method for 3909–3916
metabolomics applications. Electrophoresis 31. Sumner LW, Amberg A, Barrett D et al (2007)
36:2215–2225 Proposed minimum reporting standards for
24. Want EJ, Wilson ID, Gika H et al (2010) chemical analysis Chemical Analysis Working
Global metabolic profiling procedures for Group (CAWG) Metabolomics Standards Ini-
urine using UPLC-MS. Nat Protocol 5(6): tiative (MSI). Metabolomics 3:211–221.
1005–1018. https://doi.org/10.1038/nprot. https://doi.org/10.1007/s11306-007-
2010.50 0082-2
25. Dunn WB, Broadhurst D, Begley P et al (2011) 32. King AM, Trengove RD, Mullin LG et al
Procedures for large-scale metabolic profiling (2020) Rapid profiling method for the analysis
of serum and plasma using gas chromatography of lipids in human plasma using ion mobility
and liquid chromatography coupled to mass enabled-reversed phase-ultra-high perfor-
spectrometry. Nat Protoc 6(7):1060–1083 mance liquid chromatography/mass spectrom-
etry. J Chromatogr A 1611:460597
 Chapter 7

Quantitative Lipidomics of Biological Samples Using

Supercritical Fluid Chromatography Mass Spectrometry
Hiroaki Takeda, Yoshihiro Izumi, and Takeshi Bamba

Abstract
Lipidomics has attracted attention in the discovery of unknown biomolecules and for capturing the changes
in metabolism caused by genetic and environmental factors in an unbiased manner. However, obtaining
reliable lipidomics data, including structural diversity and quantification data, is still challenging. Supercrit-
ical fluid chromatography (SFC) is a suitable technique for separating lipid molecules with high throughput
and separation efficiency. Here, we describe a quantitative lipidomics method using SFC coupled with mass
spectrometry. This technique is suitable for characterizing the structural diversity of lipids (e.g., phospho-
lipids, sphingolipids, glycolipids, and glycerolipids) with high quantitative accuracy to understand their
biological functions.

Key words Lipidomics, Phospholipids, Sphingolipids, Glycerolipids, Glycolipids, Liquid-liquid

extraction, Quantitative analysis, Supercritical fluid chromatography, Mass spectrometry, Plasma,
Lipoproteins, Cells, Extracellular vesicles, Organs, Tissues

1 Introduction

Lipidomics is a powerful tool for evaluating the changes in lipid

metabolism and for identifying novel candidate biomarkers of vari-
ous diseases [1–3]. Lipids exhibit wide structural diversity (Fig. 1)
and vary their distributions across species, organs, cell types, and
organelles [4]. With the progress in untargeted lipidomics, approx-
imately 50,000 lipid structures are currently registered in the
LIPID MAPS database and have been classified based on their
specific structures (e.g., phosphatidylcholine (PC), phosphatidyl-
ethanolamine (PE), phosphatidylglycerol (PG), phosphatidylserine
(PS), phosphatidylinositol (PI), phosphatidic acids (PA), lyso-PC
(LPC), lyso-PE (LPE), lyso-PG (LPG), lyso-PS (LPS), lyso-PI
(LPI), lyso-PA (LPA), sphingomyelin (SM), ceramide (Cer),
monoacylglycerol (MG), diacylglycerol (DG), triacylglycerol
(TG), cholesteryl ester (CE), and fatty acid (FA)) (Table 1)

Olga Deda et al. (eds.), Metabolic Profiling: Methods and Protocols, Methods in Molecular Biology, vol. 2891,
https://doi.org/10.1007/978-1-0716-4334-1_7,
© The Author(s), under exclusive license to Springer Science+Business Media, LLC, part of Springer Nature 2025

131
132 Hiroaki Takeda et al.

Fig. 1 Chemical diversity of glycerophospholipids and sphingolipids. The abbreviations of lipid classes are
summarized in Table 1

[5, 6]. Mass spectrometry (MS) is widely used in omics science

because of its sensitivity and selectivity [7]. However, the discrimi-
nation of lipid structures and accurate quantification of lipids
remain inadequate [8]. Theoretically, various isomers such as cis/-
trans isomers of unsaturated bonds (e.g., PC 18:1(9Z)/18:1
(9Z) vs. PC 18:1(9E)/18:1(9E)), positions of unsaturated bonds
(e.g., PC 18:0/18:3(Δ9,12,15) vs. PC 18:0/18:3(Δ6,9,12)),
positions of fatty acyl chains (e.g., PC 18:0/18:2 vs. PC 18:2/
18:0), combinations of different fatty acyl chains (e.g., PC 16:
0_20:4 vs. PC 18:2_18:2), and different lipid classes (e.g., PC 15:
0_18:1 vs. PE 18:0_18:1 (both, exact mass 745.5622,
C41H80NO8P)) may be present in a lipid structure (Fig. 2a)
[2, 6]. In addition, some isomers based on analytical features such
as adduct ions (e.g., [PC 32:0 + H]+ vs. [PA 37:1 + NH4]+ (both,
m/z 734.5694, C40H81NO8P) and [PS 32:0 + H]+ vs. [PG 32:
2 + NH4]+ (both, m/z 736.5123, C38H75NO10P)), can be present
and in-source fragmentation (e.g., generation of PA by neutral loss
of head group of PC) may occur. Depending on the mass resolu-
tion, it is necessary to consider the isobaric ions derived from
isotopes (e.g., PC 32:1 + 13C2 (exact mass 733.5532,
12
C3813C2H78NO8P) vs PC 32:0 (exact mass 733.5622,
12
C40H80NO8P)) and different lipid classes (e.g., PC 32:0 vs. PS
32:1 (exact mass 733.4894, C38H72NO10P) and CE 18:0 (exact
mass 652.6158, C45H80O2) vs DG 38:0 (exact mass 652.6006,
C41H80O5)). Because the structural depth of lipid molecules that
can be determined varies depending on the analytical strategy, lipid
annotation is recommended according to international standards
(Subheading 2.4.1) [9]. The quantification of constituent lipids by
MS is also challenging because of their different signal responses
depending on the structure and sample matrix. Because not all lipid
Table 1
Summary of lipid classes and their effective fragment ions for annotation

Positive ion mode Negative ion mode

Lipid class Abbreviation Adduct ion Fragment Adduct ion Fragment

Lyso-phosphatidylcholine LPC [M + H]+ 184.1 [M + CH3COO]- [FA - H]-
Lyso-phosphatidylethanolamine LPE [M + H]+ Loss of 141.0 Da [M - H]- [FA - H]-
Phosphatidylcholine PC [M + H]+ 184.1 [M + CH3COO]- [FA - H]-
Phosphatidylethanolamine PE [M + H]+ Loss of 141.0 Da [M - H]- [FA - H]-
Phosphatidylglycerol PG [M + NH4]+ Loss of 189.0 Da [M - H]- [FA - H]-
Bis(monoacylglycero)phosphate BMP [M + NH4]+ [M + H - LPA]+ [M - H]- [FA - H]-
Phosphatidylserine PS [M + H]+ Loss of 185.0 Da [M - H]- [FA - H]-
Phosphatidylinositol PI [M + NH4]+ Loss of 277.1 Da [M - H]- [FA - H]-
Sphingomyelin SM [M + H]+ 184.1 [M + CH3COO]-
Ceramide Cer [M + H]+ 264.3 (18:1;O2) [M + CH3COO]-
Hexosylceramide HexCer [M + H]+ 264.3 (18:1;O2) [M + CH3COO]-
Sulfatide SHexCer [M + H]+ 264.3 (18:1;O2) [M - H]- 97.0
Cholesteryl ester CE [M + NH4]+ 369.4 Not detected
+ +
Diacylglycerol DG [M + NH4] [M + H - FA] Not detected
+ +
Triacylglycerol TG [M + NH4] [M + H - FA] Not detected
Quantitative Lipidomics Using SFC/MS
133
134 Hiroaki Takeda et al.

Fig. 2 Characterization of SFC/QqQMS introduced in the chapter. (a) Nomenclature and annotation levels of
each lipid class. (b) Quantification level of each lipid class. (c) Characteristics of the main analytical strategies.
Each performance was determined by three levels. (d) Characteristics of mass spectrometers. The analytical
performance of each instrument was determined by three levels. The characteristic of structural depth is
evaluated when CID is applied

molecules with a stable isotope label can be prepared for lipidome

analysis, it is recommended to use internal standards (ISDs) that are
not present in the target samples (i.e., deuterium-labeled or
odd-chain ISDs, at least one per lipid class) (Fig. 2b) [9]. Although
the ionization efficiency depends primarily on the polar head
groups, the properties of the fatty acyl chains (i.e., the number of
carbons and unsaturated bonds) also affect the MS signal response.
For more accurate quantification, a fragmentation model was
developed to normalize the fragmentation efficiency of different
fatty acyl properties toward accurate quantification [10]. Thus, it is
necessary to determine an appropriate lipidomics strategy that
depends on the biological objectives.
Various analytical methods have been developed by combining
chromatography and MS techniques (Fig. 2c, d) [7, 11, 12]. Direct
infusion MS (DI-MS) can quantify individual lipid molecules
because the ionization suppression of each lipid class can be cor-
rected by introducing a sample with the ISDs of the corresponding
lipid class directly into the MS. To further separate isomeric and
isobaric compounds while maintaining quantitative performance,
normal-phase liquid chromatography (LC) and hydrophilic inter-
action chromatography (HILIC), which can recognize differences
 Quantitative Lipidomics Using SFC/MS 135

Fig. 3 Features of SFC. (a) Phase diagram. (b) Critical point of each substance. (c) Characteristics of SFC/MS/
MS introduced in the chapter. (d) Schematic diagram of SFC/QqQMS. (e) Lineup of 23 fatty acyl chains
contained in the in-house lipid MRM library [20]

between the polar head groups rather than between the fatty acyl
chains, are widely used. In particular, HILIC has been widely
applied for the separation of lipid classes, including isomers of
glycolipids, such as diastereomers of hexosylceramides (e.g., gluco-
sylceramide (GlcCer) vs. galactosylceramide (GalCer)) [13] and
positional isomers of gangliosides (e.g., GD1a vs. GD1b)
[14]. However, the analytical throughput is low because of the
time-consuming formation of a water-rich layer. Reversed-phase
LC (RPLC) is suitable for untargeted lipidomics because it can
separate lipid molecules based primarily on the differences in the
fatty acyl chains (i.e., the number of carbons and unsaturated
bonds) and slightly polar head groups. However, normalization of
ionization efficiency by ISDs in RPLC/MS is difficult unless lipi-
dome isotope labeling of yeast is used [15], because the degree of
ionization suppression is different for each lipid molecule.
Recently, supercritical fluid chromatography (SFC) has been
used to separate a wide range of lipid molecules [16–18]. In partic-
ular, supercritical carbon dioxide (CO2) is suitable as a mobile
phase for lipid analysis because of its low polarity (similar to that
of n-hexane), low cost, ease of handling, non-toxicity, and
non-flammability (Fig. 3a, b). Because the polarity of the mobile
phase is changed by gradually increasing the composition of the
modifier solvent (e.g., methanol), columns for which the stationary
phase has a high polarity are suitable for SFC. Dissolution solvent
and ionization efficiency result in unique behaviors compared to
that of HILIC because the major mobile phase composition is
supercritical CO2 [16, 19]. Because the weak solvent for HILIC-
based lipid separation has a high content of acetonitrile, it is neces-
sary to dissolve the sample extract in a weak solvent to prevent peak
136 Hiroaki Takeda et al.

broadening for the compounds with weak retention. However,

acetonitrile is insufficient for dissolving a wide range of lipids. In
contrast, a mixture of methanol and chloroform can be effectively
used as a sample solvent in SFC to dissolve and retain lipids
[20]. Compounds with proton adduct ions increase MS sensitivity
by generating methoxylcarbonic acid, which acts as a proton donor
in the positive-ion mode in a mixture of CO2 and methanol [19]. In
addition, the characteristics of supercritical fluids (e.g., low viscosity
and high diffusivity) provide high separation efficiency and short
equilibration times, enabling high-throughput analysis (Fig. 3c). In
our previous study, a diethylamine (DEA)-bonded silica column
was used to separate a wide variety of lipid classes while maintaining
good peak shapes [20]. By adding ISDs for each lipid class, the lipid
classes and their constituent molecules were quantified with reliable
accuracy (Fig. 3c) [20, 21]. Although some additional devices (e.g.,
CO2 pumps, back pressure regulators, and makeup pumps) are
required in SFC operation (Fig. 3d) [22], the basic usability of
SFC is similar to that of the conventional LC. Therefore, SFC is a
powerful tool, particularly for the normal-phase mode, with the
advantages of throughput, lipid coverage, and separation
performance.
Because the constituent lipids of the same lipid class are eluted
at similar times in the normal-phase mode, the corresponding peaks
of the lipid classes can be easily identified by checking the ISD
chromatograms. However, annotation of detailed structures (e.g.,
fatty acyl composition and double bond position) may be difficult
depending on the mass spectrometer. Untargeted lipidomics has
been widely applied using high-resolution tandem mass spectrom-
etry (HRMS) in the data-dependent acquisition (DDA) or data-
independent acquisition (DIA) modes. Lipid class separation by LC
or SFC in the normal-phase mode is not sufficient for DDA to
obtain MS/MS spectra because more isomeric and isobaric com-
pounds co-elute in this mode than in the RPLC mode. DIA can
acquire MS/MS spectra of the observed precursor ions without
missing any information. However, when using the normal-phase
mode of LC or SFC, various isomeric and isobaric compounds are
included in the MS/MS chromatogram. Although it is difficult to
apply modern fast-scanning triple quadrupole mass spectrometers
(QqQMS) to untargeted lipidomics because of their low mass
resolution, multiple reaction monitoring (MRM) approaches can
generate MS/MS chromatogram peaks with sufficient data points
(i.e., more than 10 points) because of the very short acquisition
time per compound. Therefore, the structural isomers of lipid
molecules with different fatty acyl chains (e.g., PC 16:0_20:
4 vs. PC 18:2_18:2) can be quantified by selecting fatty acyl frag-
ments in the negative-ion mode of MRM. SFC/QqQMS was the
best combination for quantifying individual lipid molecules while
 Quantitative Lipidomics Using SFC/MS 137

Fig. 4 Lipid separation using the DEA column. (a) Separation of lipid classes using the DEA column. The dot
plot represents the gradient condition of the modifier solvent. (b) MS separation of the constituent PC
molecules using the MRM mode. Fragments of fatty acyl chains were selected in the negative-ion mode.
(c) Separation of positional isomers using the DEA column. (The figures are re-used from ref. [20] with minor
modifications)

separating structural isomers with different fatty acyl chains

(Fig. 3c).
This chapter describes the detailed protocols and precautions
for SFC/QqQMS-based lipidomics of biological samples. Because
QqQMS is a targeted analysis tool, an in-house lipid MRM library
for efficient screening of lipids including 22 lipid classes with 23 dif-
ferent fatty acyl chains (i.e., approximately 2500 lipids) was present
in our previous paper (Fig. 3e) [20]. Using scheduled MRM based
on the retention time of each lipid class, the initial lipid screening
can be completed in approximately 5 h (16 MRM methods). If
necessary, MRM screening methods can be easily expanded using a
bioinformatics tool (Subheading 2.4.4). In addition to screening
lipid molecules in targeted samples, lipidomics databases (Subhead-
ing 2.4.2) and data repositories (Subheading 2.4.3) are also effec-
tive for smooth study progress. The separation and quantification
by SFC/QqQMS are summarized in detail in Figs. 4 and 5
[20, 21]. As the SFC/QqQMS-based lipidomics method has
been applied to various biological samples [21, 23–25], a sampling
protocol is also present in this chapter.
138 Hiroaki Takeda et al.

Fig. 5 Quantification of lipid classes and their constituent molecules. (a) Strategy for lipid quantification. CF
refers to the correction factor determined by calibration curve of charged aerosol detector (CAD). (b)
Separation of lipid molecules by using the DEA column in the rabbit plasma extracts. The number in
parentheses represents the number of lipids detected. (c) Evaluation of quantitative performance by spike-
and-recovery test. (d) Chromatogram of the HepG2 cell extracts by SFC-CAD. Each peak was annotated by
QqQMS. Ceramide phosphoethanolamine (Cer PE) was used as the ISD. (e) Comparison of quantitative levels
between QqQMS and CAD. Lipid extract of HepG2 cells was used for comparison. The figures are re-used from
ref. [20, 21] with minor modifications

2 Materials

2.1 Sample 1. Salt solution (d = 1.006 g/mL): Add sodium chloride (11.4 g)
Preparation and ethylenediaminetetraacetic acid (EDTA) disodium salt
(0.1 g) to a volumetric flask and dissolve them by distilled
2.1.1 Lipoproteins
water (500 mL) and 1 mol/L sodium hydroxide solution
(1 mL). Add distilled water to bring the volume to 1000 mL,
then fine-tune the solution by adding small increments of
distilled water.
2. Salt solution (d = 1.182 g/mL): Add sodium bromide
(24.98 mg) to the salt solution (d = 1.006 g/mL) (100 mL).
3. Polypropylene tube for ultracentrifugation (4.7 mL).
4. Pasteur pipette.
5. Beckman CentriTube Slicer for micro ultracentrifuge tubes.
6. Ultracentrifuge with a fixed-angle rotor.

2.1.2 Cells 1. Phosphate-buffered saline (PBS).

2. Cell counter.
 Quantitative Lipidomics Using SFC/MS 139

2.1.3 Extracellular 1. 0.22 μm filter.

Vesicles 2. Filtrated PBS (-).
3. Centrifuge.
4. Ultracentrifuge with a swing rotor.

2.1.4 Tissues and Organs 1. Liquid nitrogen (N2).

2. Ball-mill grinder.
3. Zirconia balls (diameter of 5 mm) for homogenization.
4. (Option) Freeze dryer.

2.1.5 Tissue Sections 1. Dry ice powder.

2. Membrane glass slide for laser microdissection (e.g., polyethyl-
ene naphthalate): It is recommended to check the MS back-
ground noise caused using the plastic material before sample
analysis. Alternatively, contamination by plastic membrane can
be removed using DIRECTOR Laser Microdissection Slides
(AMR Inc., Tokyo, Japan).
3. Polymerase chain reaction tube (0.2 or 0.5 mL) or 96-well
plate.
4. PBS.
5. Cryostat.
6. Laser capture microdissection system.

2.2 Lipid Extraction 1. Methanol for LC/MS: Because methanol is used for lipid
extraction and as a mobile phase for SFC, LC/MS-grade meth-
2.2.1 Liquid-Liquid
anol is recommended to reduce the MS background noise.
Extraction
2. Distilled water for LC/MS: Same precautions as for methanol.
3. Chloroform for high performance LC (HPLC): As chloroform
is converted to toxic phosgene in the presence of light, it
should be stored in the dark.
4. Eppendorf Safe-Lock Tube (2.0 mL): Because Eppendorf
tubes are made of polypropylene, they exhibit chemical resis-
tance and temperature stability. In addition, plasticizers that
increased the MS background were not detected when organic
solvents were used.
5. Micropipettes (20, 200, and 1 mL): Micropipettes with O-ring
materials, which undergo small volumetric expansion when
dispensing organic solvents, and internal springs, which are
resistant to organic solvents, are suitable for accurate dispens-
ing and for avoiding unnecessary contamination.
6. Cooling rack for microtubes: Prior to lipid extraction, the
sample rack should be cooled in a freezer at -30 °C.
140 Hiroaki Takeda et al.

7. Vortex mixer with microtube attachment: This mixer can be

used for the simultaneous stirring of up to 12 samples during
lipid extraction.
8. Ultrasonication bath.
9. Centrifuge with an angle rotor.
10. Centrifugal evaporator: It is recommended to prepare glass
material in the gas-contacting part because it is necessary to
evaporate the solvent for liquid-liquid extraction (e.g., metha-
nol, chloroform, and acetic acid).
11. Spray concentrator with N2 generator: Although the drying
time is shorter than that of the centrifugal evaporator when
evaporating chloroform, care should be taken not to contami-
nate the compounds from the spray nozzle.

2.2.2 Standards 1. Mouse SPLASH LIPIDOMIX Mass Spec Standard (Avanti

Polar Lipids Inc., Alabaster, AL): 45 μM of LPC 18:1(d7);
2 μM of LPE 18:1(d7); 100 μM of PC 15:0_18:1(d7); 7 μM
of PE 15:0_18:1(d7); 5 μM of PG 15:0_18:1(d7); 20 μM of PS
15:0_18:1(d7); 20 μM of PI 15:0_18:1(d7); 10 μM of PA 15:
0_18:1(d7); 20 μM of PC plasmalogen (PC P-) 18:0_18:1(d9);
5 μM of PE plasmalogen (PE P-) 18:0_18:1(d9); 20 μM of SM
18:1;O2_18:1(d9); 250 μM of CE 18:1(d7); 15 μM of DG 15:
0_18:1(d7); and 35 μM of TG 15:0_18:1(d7)_15:0.
2. Cer 18:1;O2(d7)_15:0 (Avanti Polar Lipids Inc.).
3. MG 18:1(d7) (Avanti Polar Lipids Inc.).
4. The National Institute of Standards and Technology (NIST)
Standard Reference Material for Human Plasma (SRM 1950)
(NIST, Gaithersburg, MD) [8, 26]: SRM 1950 is suitable for
the standardization and quality assessment of lipidomics data
across instruments, facilities, and analysis dates.

2.3 Lipidome 1. Brown vial for autosampler: Brown vials are useful for protect-
Analysis ing light-sensitive lipids.
2.3.1 Sample 2. Inert glass inserts for sample vials: The silanol groups on the
Preparation glass were converted to hydrophobic groups to prevent the
adsorptive interaction of polar compounds. Because the
amount of affected lipids depends on the contact area with
the glass, the inert glass insert can effectively reduce adsorp-
tion, particularly for the analysis of sample extracts with low
concentrations.
3. Vial cap and slit septum: The slit septum was effective for
inserting the injection needle. When storing the sample vial
after lipid analysis, it is recommended to replace it with a vial
cap without a slit to prevent the volatilization of the sample
elution.
 Quantitative Lipidomics Using SFC/MS 141

2.3.2 SFC 1. Nexera UC system (Shimadzu Co., Kyoto, Japan). The back
pressure regulator can be operated without splitting the lines
after passing through the column.
2. CO2 (99.9% grade).
3. Body scale: Because a heavy CO2 bomb is placed on the scale, a
thin scale is recommended.
4. Torus DEA column (3.0 × 100 mm, 1.7 μm, Waters Co.,
Milford, MA): The DEA column can be used for the simulta-
neous analysis of different phospholipids by effectively setting
up scheduled MRM transitions, because it can separate phos-
pholipid classes at the baseline while maintaining good peak
shapes of acidic phospholipids (Fig. 4a) [20].
5. Methanol/water (95/5, v/v) with 0.1% (w/v) ammonium
acetate as a modifier and a makeup pump solvent: LC/MS-
grade methanol and distilled water is recommended to reduce
the MS background noise.

2.3.3 MS 1. LCMS-8060 (Shimadzu Co.) with an electrospray ionization

ion source: Because the maximum scan speed is 3000 μ/s, it is
possible to set a large number of MRM transitions for simulta-
neous analysis (dwell time, 1 ms; and pause time, 2 ms).
2. N2 generator.
3. Argon gas bomb.

2.4 Data Analysis 1. Lipidomics Standards Initiative (https://lipidomicstandards.

org/) [9]: Guidelines for qualitative and quantitative analyses
2.4.1 Guideline
and the resources for isomeric and isobaric overlaps, including
in-source fragmentation, have been summarized. Based on the
analytical approach, the shorthand notation of lipid structures
is as follows: PC 36:1 (e.g., MS/MS of polar head group), PC
18:0_18:1 (e.g., MS/MS of fatty acyl chains), PC 18:0/18:1
(e.g., differential mobility spectrometry [27, 28] and electron
activated dissociation [29]), PC 18:0_18:1(Δ9) (e.g., Patern-
ò-Büchi reaction [30], HRMS with electron activated dissocia-
tion [29], and oxygen attachment dissociation [31]), and PC
18:0/18:1(9Z) (e.g., electron activated dissociation [32]). In
addition to the analytical approach, the selection of ISDs affects
the reliability of quantitative performance as follows: level
1, absolute quantification (use of 13C labeled ISD); level
2, semi-quantification (DI-MS, HILIC/MS, or SFC/MS
with use of odd chain or deuterium labeled ISDs); and level
3, relative quantification (RPLC/MS or use of ISDs that are
not matched with the targeted lipid classes). The appropriate
notation of lipid classes and the quantitative reliability of this
methodology are summarized in Fig. 2a, b.
142 Hiroaki Takeda et al.

2.4.2 Database 1. LIPID MAPS (https://www.lipidmaps.org/) [5, 6]: This data-

base can be used to obtain detailed information on lipid mole-
cules (e.g., InChiKey, SMILES, structure, formula, exact mass
of lipids with various adduct ions, and predicted MS/MS
spectrum).
2. KEGG PATHWAY Database (https://www.genome.jp/kegg/
pathway.html) [33]: This database can be searched for whole
lipid metabolism in a variety of animals. It is useful for char-
acterizing the changes in lipid metabolism based on the analyt-
ical results.
3. RIKEN LIPIDOMICS (http://prime.psc.riken.jp/menta.
cgi/lipidomics/index): This database can be searched for the
results obtained by untargeted lipidomics approach in
biological samples (e.g., mouse tissues and organs, mouse
cultured cells, and human plasma).
4. MetaboAtlas21 (https://metaboatlas21.metabolomics.fgu.cas.
cz/): This database can be searched for a comprehensive atlas
of the mouse metabolome and lipidome (e.g., various tissues
and organs of mice fed a normal or high-fat diet).

2.4.3 Data Repository 1 . M e t a b o l o m i c s W o r k b e n c h ( h t t p s : // w w w .

metabolomicsworkbench.org/) [34]: A wide variety of lipi-
dome data (e.g., organs and animals) are available free of
charge. For example, raw data from 10 regions (i.e., basal
ganglia, cerebellum, cerebral cortex, hippocampus, hypothal-
amus, medulla, midbrain, olfactory bulb, pons, and thalamus)
obtained using the Thermo Q Exactive HF hybrid Orbitrap
for the construction of the mouse brain metabolome atlas
have been published [35]. Using these data repositories,
MRM lists for quantitative lipidomics can be generated rap-
idly, without sample analysis.

2.4.4 Software and 1. LabSolutions LCMS (Shimadzu Co.).

Bioinformatics Tool 2. MSconvertGUI (https://proteowizard.sourceforge.io/): This
software can convert MS files into .txt files.
3. MS-DIAL (http://prime.psc.riken.jp/compms/msdial/main.
html) [36]: This software is an automated annotation tool for
metabolites analyzed by HRMS. Various types of data can be
analyzed without data format conversion (e.g., ABF, mzML,
netCDF, IBF, WIFF, RAW, LCD, and QCD).
4. LipidCreator (https://lifs-tools.org/lipidcreator.html) [37]:
This software is an automated tool for calculating m/z of
precursor and product ions. It is useful for expanding the
MRM screening methods based on previous studies.
 Quantitative Lipidomics Using SFC/MS 143

5. Multi-ChromatoAnalysT (Beforce, Fukuoka, Japan) (https://

www.be-force.co.jp/en): This software is an automated pro-
gram for the quantification of lipid molecules obtained in the
MRM mode of QqQMS. It has the advantage of performing
peak picking of different samples simultaneously by inserting
retention time information determined by previous analysis
and of easy handling of peak alignment.
6. MetaboAnalyst (https://www.metaboanalyst.ca/) [38]: This
website can support various data analyses such as enrichment
analysis, pathway analysis, statistical analyses such as t-test (two
groups), analysis of variance (more than two groups), principal
component analysis, and heat map for visualization.
7. LION/web (http://lipidontology.com/) [39]: This website
can support enrichment analysis of lipidomics data.
8. Programming tools such as R and Python: The ggplot2 pack-
age is convenient for creating various graphs, including histo-
grams, box plots, beeswarm plots, and receiver operating
characteristic curves.

3 Method

3.1 Sample 1. Using a micropipette, transfer plasma or serum (10 μL) into a
Preparation clean tube. Ensure that fibrin is not contaminated (see Notes 1
and 2).
3.1.1 Blood
2. Use the volume to normalize the lipidomics data.

3.1.2 Lipoproteins [23, 1. Add plasma (1.5 mL) to the tubes for ultracentrifugation and
24] spin down at 700 g for 3 min at 4 °C.
2. Add the salt solution (3.2 mL, d = 1.006 g/mL) with a Pasteur
pipette slowly along the tubes and centrifuge the tubes at
375,000 g for 5 h at 4 °C.
3. Collect the upper and lower layers in clean tubes (upper very
low-density lipoprotein (VLDL) fractions, 1.5 mL; lower other
fractions, 3.2 mL). The microslicer is recommended to split the
tubes without contaminating the fractions.
4. If necessary, the fractions are centrifuged again using the salt
solution with the same density (d = 1.006 g/mL) to obtain the
desired fractions.
5. Collect low-density lipoprotein (LDL) and high-density lipo-
protein (HDL) fractions sequentially by repeating the above
steps (VLDL, d < 1.006 g/mL; LDL, 1.019 g/
mL < d < 1.063 g/mL; HDL, d > 1.063 g/mL). Add the
salt solution (d = 1.182 g/mL) until the desired concentration
144 Hiroaki Takeda et al.

is reached, and scale up the solution to the corresponding

concentration.
6. Use the amount of lipoproteins to normalize the lipidomics
data. Estimate the number of VLDL and LDL particles using
apolipoprotein B-100 (apoB-100), because a single apoB-100
molecule is contained in the lipoprotein particle.

3.1.3 Cells 1. Cultivate cells of interest (e.g., cell lines and primary cells) at
37 °C in an incubator. Among dishes, use one dish for cell
counting.
2. Count the number of cells using the cell counter.
3. Wash the cells with PBS.
4. Quench by adding cold methanol and collect the cells in a clean
tube. Care should be taken to not evaporate the methanol by
performing this step promptly.
5. Use the number of cells to normalize the lipidomics data.

3.1.4 Extracellular 1. Centrifuge the conditioned medium at 2000 g for 10 min at

Vesicles [25] 4 °C and filter the supernatant through a 0.22 μm filter unit.
2. Centrifuge the filtered medium at 150,000 g for 70 min at
4 °C.
3. Wash the pellet once with filtrated PBS (-) and resuspend it in
the remaining PBS (-).
4. Use the number of extracellular vesicles to normalize the
lipidomics data.

3.1.5 Tissues and Organs 1. Dissect the animals as soon as possible and immediately freeze
the collected organs using liquid N2. Be aware that the lipid
metabolism may change depending on the time since death
(see Note 3).
2. If necessary, lyophilize the organs in the dark.
3. Homogenize the dried organs by ball milling at 30 ~ 50 Hz for
2 min after freezing in liquid N2.
4. Collect a sample (5 mg) of the crushed organs into a clean tube.
Attention should be paid to the precision of the electronic
balance.
5. Use dry or wet weight to normalize the lipidomics data.

3.1.6 Tissue Sections 1. Prepare a thin frozen section (approximately 10–20 μm) and
[40] mount it on the glass slide for laser microdissection. Use dry ice
powder to freeze whole brains while preventing sample
cracking.
2. Collect the region of interest (ROI) by laser microdissection on
a cap of either a polymerase chain reaction tube or a 96-well
 Quantitative Lipidomics Using SFC/MS 145

plate. It is recommended to add PBS to the tube or plate before

obtaining the ROI (e.g., 50–60 μL for 0.5 mL tube or
20–30 μL for 0.2 mL tube) because in some cases the sample
unexpectedly sticks to the plate or tube because of static
electricity.
3. Use the area to normalize the lipidomics data.

3.2 Lipid Extraction 1. Add methanol/chloroform/water (1000 μL, 10:5:3, v/v/v)

to the samples on ice and vortex mix for 1 min at room
3.2.1 Liquid-Liquid
temperature (r.t.) using a maximum speed vortex mixer.
Extraction [41]
2. Extract lipids by using an ultrasonic bath for 5 min. Float for
microtube is useful for extraction of multiple samples.
3. Centrifuge the samples at 16,000 g for 5 min at 4 °C and collect
the supernatant (700 μL) into the clean tube. Care must be
taken not to collect precipitated protein pellets.
4. Add chloroform (195 μL) and water (195 μL) to the collected
supernatant on ice and vortex mix for 1 min at r.t. using a
maximum speed vortex mixer.
5. Centrifuge the samples at 16,000 g for 5 min at 4 °C and collect
aliquots of the aqueous and organic layers into clean tubes.
Care must be taken not to collect the middle layer of the
protein pellet (aqueous layer, gangliosides and lysophospholi-
pids [14, 42]; organic layer, phospholipids, sphingolipids, and
neutral lipids) (see Notes 4–6).
6. Prepare dry aliquots of the aqueous and organic layers using a
centrifuge evaporator and a spray concentrator with N2 genera-
tor, respectively (see Note 7).
7. Dissolve the dried extracts in methanol/chloroform (1:1, v/v)
containing the Mouse SPLASH LIPIDOMIX Mass Spec Stan-
dard that was diluted according to the dynamic range of MS.

3.3 Lipidome 1. Deaerate the mobile phase solvents sufficiently before analysis
Analysis to prevent unexpected pump problems or signal noise caused
by air bubbles. Replace the old mobile phase with a new one
3.3.1 Equilibration of SFC
within 2 weeks to prevent compositional changes caused by
volatilization or microbial contamination.
2. Prepare methanol as a needle wash solvent. Precautions are
same as for modifier and makeup pump solvents (see Note 8).
3. Measure the weight of the CO2 bomb periodically and note
that the remaining amount of CO2 should not be less than
10 kg. An unstable retention time of compounds may be
caused if the correct pressure is not reached.
4. Purge the autosampler and pumps to remove air.
146 Hiroaki Takeda et al.

5. Cool the CO2 pump head at 4 °C and check the temperature

behavior. It may be necessary to maintain the density of CO2
for a stable retention time.
6. Place the analytical column in the oven and maintain the oven
temperature at 50 °C.
7. Activate the back pressure regulator to 10 MPa and a total flow
of 1.0 mL/min. Activate the back pressure regulator before
activating the CO2 pump to prevent accidents by sending the
solution at low pressure.
8. Equilibrate the SFC system to the initial modifier solvent con-
ditions for at least 5 min. The SFC conditions are as follows:
flow rate of mobile phase, 1.0 mL/min; flow rate of makeup
pump, 0.1 mL/min; mobile phase (A), CO2; mobile phase (B),
methanol/water (95/5, v/v) with 0.1% (w/v) ammonium
acetate; gradient condition of mobile phase (B), 1% (1 min),
1–65% (11 min), 65% (6 min), 65–1% (0.1 min), 1% (1.9 min);
temperature of column oven, 50 °C; active back pressure regu-
lator, 10 MPa; injection volume, 1 μL (partial loop) (see Note
9).
9. Monitor the pump pressure to evaluate the SFC condition and
the degree of column exhaustion. For early detection of pro-
blems, visually check the spray of the mobile phase under the
initial condition.

3.3.2 Equilibration of MS 1. Equilibrate the MS status by using the MS method for lipi-
dome analysis. The QqQMS conditions are as follows: polarity,
both positive- and negative-ion modes; electrospray voltage,
4.0 kV; desolvation line temperature, 250 °C; heat block tem-
perature, 400 °C; nebulizing gas (N2) flow, 3.0 L/min; drying
gas (N2) flow, 10.0 L/min; collision-induced dissociation
(CID) gas (argon) pressure, 0.23 MPa, detector voltage,
2.16 kV, dwell time, 1 ms; pause time, 2 ms; polarity switching
time, 15 ms.
2. Perform autocalibration in both the positive- and negative-ion
modes to check the resolution, mass accuracy, and sensitivity if
problems occur. This will be helpful in troubleshooting if it is
unclear whether the problems originate from the SFC or
the MS.

3.3.3 Sample Analysis 1. Prior to lipid analysis, optimize collision energy for each lipid
class using the internal standards. Flow injection mode is effi-
cient for MRM optimization.
2. Analyze the ISD mixture to check the data quality (e.g., reten-
tion time, peak width, and intensity). Replace the column if the
retention time and peak width are not reproducible. Clean the
 Quantitative Lipidomics Using SFC/MS 147

MS instrument according to the operating manual if the MS

sensitivity is drastically reduced.
3. Transfer the sample extracts to vials and place them in an
autosampler. Keep the autosampler unit at less than 10 °C to
prevent the volatilization of the sample solvent.
4. Inject the quality control sample (1 μL), which is the mixture of
equal amounts of lipid extracts, using in-house lipid MRM
library including 22 lipid classes with a variety of 23 fatty acyl
chains (i.e., total of approximately 2500 lipids) (Fig. 3e)
[20]. The lipid screening should be completed in approxi-
mately 5 h (16 MRM methods). Use the fragment ions in
Table 1 as a reference when expanding the MRM screening
methods (see Note 10).
5. Determine the dilution ratio based on the result for the quality
control sample and create an MRM library for quantification
(see Note 11).
6. Inject the extract (1 μL) for every sample for quantitative
lipidome analysis (see Notes 12–14).

3.4 Data Analysis 1. Prepare untargeted lipidomics data.

3.4.1 MS-DIAL (for 2. Set the parameters for automatic annotation of lipids, such as
Untargeted Lipidomics) the measurement parameters (e.g., chromatography type, frag-
mentation, and ion mode), mass tolerance (MS1 and MS2),
minimum peak height for peak picking, retention time toler-
ance for peak alignment, targeted lipid classes, and adduct ions.
3. Confirm the alignment result to remove misannotated lipids.
Note the retention time, fragmentation efficiency, and peak
height of each lipid molecule.
4. Output the data matrix to a text file for data interpretation.

3.4.2 Multi- 1. When using the MS data except for .lcd files, convert them to .
ChromatoAnalysT (for txt files using the MSconvertGUI software.
Targeted Lipidomics) 2. Drag and drop the .txt files to Multi-ChromatoAnalysT and
read the data.
3. Perform peak alignment for every sample and select the start
and end points of the peaks for the calculation of the peak area
(see Notes 15–17).
4. Starting from the second time, insert the retention time infor-
mation determined by the preceding analysis to pick the peaks
automatically. Alternatively, add new data to the previous
data set.
5. Input the limit of quantification (peak intensity) and the con-
centrations of ISDs to quantify lipid molecules automatically.
6. Output the data matrix to an Excel sheet to interpret the data.
148 Hiroaki Takeda et al.

3.4.3 Interpretation of 1. Calculate the approximate concentrations of the constituent

Data molecules by normalizing the lipidomics data by the sample
volume and Mouse SPLASH LIPIDOMIX Mass Spec Standard
(Fig. 5c). Estimate the concentrations of lipid classes by sum-
ming the concentrations of the constituent molecules (Fig. 5e).
2. Extract useful information by using different data analyses with
the MetaboAnalyst, LION/web, and programming tools. For
example, visualize the overall picture of lipidomics data using
principal component analysis and/or hierarchical cluster analy-
sis. In addition, consider the biological importance hidden in
the lipidomics data by observing the behavior of lipid metabo-
lism using enrichment analysis.

4 Notes

1. Maintain r.t. and time when leaving blood still to prepare the
serum sample.
2. Use the same anticoagulant (e.g., EDTA, citrate, and heparin)
throughout the study. EDTA or heparin is recommended when
analyzing polar metabolites because citrate is a metabolite in
the tricarboxylic acid cycle.
3. Store the dissected biological samples in a freezer at -80 °C
prior to lipid extraction.
4. Maintain the acidic (pH 4) and cold (<4 °C) conditions to
eliminate acyl migration of lysophospholipids during the
extraction process [43]. The extraction of lipids under alkaline
conditions accelerates the acyl migration of lysophospholipids.
5. Non-polar lipid classes such as CEs and TGs should be
extracted using biphasic liquid extraction, because the solubil-
ity of these classes is insufficient when using monophasic liquid
extraction [44].
6. In the analysis of gangliosides, taking into account the dynamic
range of MS, it may be appropriate to use biphasic liquid
extraction, which can further enrich the aqueous layer, rather
than monophasic liquid extraction.
7. Store the dried extract in a freezer at -30 °C until immediately
before lipid analysis.
8. Check carryover levels before lipid analysis. The carryover
levels should be less than 1% to quantify lipids with reliable
accuracy. A mixture of isopropanol, methanol, acetonitrile, and
water (1:1:1:1, v/v/v/v) may be effective for washing the
needle if the carryover levels are serious for some lipids.
 Quantitative Lipidomics Using SFC/MS 149

9. Set the minimum modifier pumping ratio to 10% when using

more than 10% water in the modifier solvent because supercrit-
ical CO2 and water are prone to phase separation.
10. Peak tailing of acidic lipids may be prevented using a mobile
phase additive such as EDTA [45]. Alternatively, poor peak
shapes can be avoided in the presence of trace metals using poly
(ether ether ketone) line tubes and metal-free columns.
11. Special care must be taken with the quantitative data of the
lipid, which is the first compound in the MRM transition. The
chromatogram of the first compound may be affected by falling
behind in the cyclic process when setting a wide m/z range
between the first and last MRM transition. To avoid this prob-
lem, an unnecessary compound can be set at the first transition.
12. Increasing the injection volume is effective when analyzing
samples with low concentrations. Confirm the peak widths of
the compounds with low retention in advance.
13. Analyze the negative control sample to remove the lipids
derived from experimental materials such as micropipette tips
and microtubes. Detect some FA, MG, DG, and TG molecules
when using the above materials. Remove these artifacts whose
peak intensity is less than tenfold that of the negative control
sample from the data matrix.
14. Analyze the quality control lipid molecules every few times to
normalize the MS data to avoid the effect of changes in instru-
ment sensitivity over time.
15. Determine lipid structure by confirming the specific fragments
in both the positive- and negative-ion modes (Table 1). Exact
annotation can be facilitated by checking the retention time of
analytes in both positive- and negative-ion modes.
16. Care must be taken not to misannotate the lipid molecules via
in-source fragmentation. For example, observe some peaks at
the same m/z as LPA and PA because some lysophospholipids
and phospholipids (e.g., LPC and PC) are converted to LPA
and PA during electrospray ionization.
17. Care must be taken to identify PG and bis(monoacylglycero)-
phosphate isomers when two peaks are detected. Confirm the
neutral loss fragments of the polar head group (i.e., glycerol) in
the positive-ion mode to discriminate PG (Table 1).

References

1. Han X (2016) Lipidomics for studying metab- spectrometry-based lipidomics. Anal Chim
olism. Nat Rev Endocrinol 12(11):668–679. Acta 1137:156–169. https://doi.org/10.
https://doi.org/10.1038/nrendo.2016.98 1016/j.aca.2020.09.060
2. Xu T, Hu C, Xuan Q, Xu G (2020) Recent 3. Masoodi M, Gastaldelli A, Hyötyl€ainen T,
advances in analytical strategies for mass Arretxe E, Alonso C, Gaggini M, Brosnan J,
150 Hiroaki Takeda et al.

Anstee QM, Millet O, Ortiz P, Mato JM, Yuan M, Zhao XH, Zhou S (2017) Harmoniz-
Dufour JF, Orešič M (2021) Metabolomics ing lipidomics: NIST interlaboratory compari-
and lipidomics in NAFLD: biomarkers and son exercise for lipidomics using SRM 1950-
non-invasive diagnostic tests. Nat Rev Gastro- Metabolites in Frozen Human Plasma. J Lipid
enterol Hepatol 18(12):835–856. https://doi. Res 58(12):2275–2288. https://doi.org/10.
org/10.1038/s41575-021-00502-9 1194/jlr.M079012
4. Harayama T, Riezman H (2018) Understand- 9. Liebisch G, Ahrends R, Arita M, Arita M, Bow-
ing the diversity of membrane lipid composi- den JA, Ejsing CS, Griffiths WJ, Holčapek M,
tion. Nat Rev Mol Cell Biol 19(5):281–296. Köfeler H, Mitchell TW, Wenk MR, Ekroos K
https://doi.org/10.1038/nrm.2017.138 (2019) Lipidomics needs more standardiza-
5. Fahy E, Subramaniam S, Brown HA, Glass CK, tion. Nat Metab 1(8):745–747. https://doi.
Merrill AH Jr, Murphy RC, Raetz CRH, Rus- org/10.1038/s42255-019-0094-z
sell DW, Seyama Y, Shaw W, Shimizu T, 10. Schuhmann K, Moon HK, Thomas H, Acker-
Spener F, van Meer G, VanNieuwenhze MS, man JM, Groessl M, Wagner N, Kellmann M,
White SH, Witztum JL, Dennis EA (2005) A Henry I, Nadler A, Shevchenko A (2019)
comprehensive classification system for lipids. J Quantitative fragmentation model for
Lipid Res 46(5):839–861. https://doi.org/ bottom-up shotgun lipidomics. Anal Chem
10.1194/jlr.E400004-JLR200 91(18):12085–12093. https://doi.org/10.
6. Liebisch G, Fahy E, Aoki J, Dennis EA, 1021/acs.analchem.9b03270
Durand T, Ejsing CS, Fedorova M, 11. O’Donnell VB, Ekroos K, Liebisch G, Wake-
Feussner I, Griffiths WJ, Köfeler H, Merrill lam M (2019) Lipidomics: current state of the
AH Jr, Murphy RC, O’Donnell VB, art in a fast moving field. Wiley Interdiscip Rev
Oskolkova O, Subramaniam S, Wakelam Syst Biol Med 12(1):1466. https://doi.org/
MJO, Spener F (2020) Update on LIPID 10.1002/wsbm.1466
MAPS classification, nomenclature, and short- 12. Han X, Gross RW (2022) The foundations and
hand notation for MS-derived lipid structures. development of lipidomics. J Lipid Res 63(2):
J Lipid Res 61(12):1539–1555. https://doi. 100164. https://doi.org/10.1016/j.jlr.2021.
org/10.1194/jlr.S120001025 100164
7. Cajka T, Fiehn O (2014) Comprehensive anal- 13. von Gerichten J, Schlosser K, Lamprecht D,
ysis of lipids in biological systems by liquid Morace I, Eckhardt M, Wachten D,
chromatography-mass spectrometry. Trends Jennemann R, Gröne HJ, Mack M, Sandhoff
Analyt Chem 61:192–206. https://doi.org/ R (2017) Diastereomer-specific quantification
10.1016/j.trac.2014.04.017 of bioactive hexosylceramides from bacteria
8. Bowden JA, Heckert A, Ulmer CZ, Jones CM, and mammals. J Lipid Res 58(6):1247–1258.
Koelmel JP, Abdullah L, Ahonen L, Alnouti Y, https://doi.org/10.1194/jlr.D076190
Armando AM, Asara JM, Bamba T, Barr JR, 14. Hájek R, Jirásko R, Lı́sa M, Cı́fková E, Holča-
Bergquist J, Borchers CH, Brandsma J, Breit- pek M (2017) Hydrophilic interaction liquid
kopf SB, Cajka T, Cazenave-Gassiot A, chromatography-mass spectrometry character-
Checa A, Cinel MA, Colas RA, Cremers S, ization of gangliosides in biological samples.
Dennis EA, Evans JE, Fauland A, Fiehn O, Anal Chem 89(22):12425–12432. https://
Gardner MS, Garrett TJ, Gotlinger KH, doi.org/10.1021/acs.analchem.7b03523
Han J, Huang Y, Neo AH, Hyötyl€ainen T, 15. Rampler E, Criscuolo A, Zeller M, El Abiead Y,
Izumi Y, Jiang H, Jiang H, Jiang J, Schoeny H, Hermann G, Sokol E, Cook K,
Kachman M, Kiyonami R, Klavins K, Klose C, Peake DA, Delanghe B, Koellensperger G
Köfeler HC, Kolmert J, Koal T, Koster G, (2018) A novel lipidomics workflow for
Kuklenyik Z, Kurland IJ, Leadley M, Lin K, improved human plasma identification and
Maddipati KR, McDougall D, Meikle PJ, Mel- quantification using RPLC-MSn methods and
lett NA, Monnin C, Moseley MA, isotope dilution strategies. Anal Chem 90(11):
Nandakumar R, Oresic M, Patterson RE, 6494–6501. https://doi.org/10.1021/acs.
Peake D, Pierce JS, Post M, Postle AD, analchem.7b05382
Pugh R, Qiu Y, Quehenberger O, Ramrup P,
Rees J, Rembiesa B, Reynaud D, Roth MR, 16. Nováková L, Perrenoud AGG, Francois I,
Sales S, Schuhmann K, Schwartzman ML, Ser- West C, Lesellier E, Guillarme D (2014) Mod-
han CN, Shevchenko A, Somerville SE, ern analytical supercritical fluid chromatogra-
St. John-Williams L, Surma MA, Takeda H, phy using columns packed with sub-2 μm
Thakare R, Thompson JW, Torta F, Triebl A, particles: a tutorial. Anal Chim Acta 824:18–
Trötzmüller M, SJK U, Vuckovic D, Weir JM, 35. https://doi.org/10.1016/j.aca.2014.
Welti R, Wenk MR, Wheelock CE, Yao L, 03.034
 Quantitative Lipidomics Using SFC/MS 151

17. Lesellier E, West C (2015) The many faces of Shevchenko A, Torta F, Wakelam MJO, Wheel-
packed column supercritical fluid ock CE, Wenk MR (2018) MS-based lipido-
chromatography—a critical review. J Chroma- mics of human blood plasma: a community-
togr A 1382:2–46. https://doi.org/10.1016/ initiated position paper to develop accepted
j.chroma.2014.12.083 guidelines. J Lipid Res 59(10):2001–2017.
18. Shulaev V, Isaac G (2018) Supercritical fluid https://doi.org/10.1194/jlr.S087163
chromatography coupled to mass 27. Maccarone AT, Duldig J, Mitchell TW,
spectrometry—a metabolomics perspective. J Blanksby SJ, Duchoslav E, Campbell JL
Chromatogr B 1092:499–505. https://doi. (2014) Characterization of acyl chain position
org/10.1016/j.jchromb.2018.06.021 in unsaturated phosphatidylcholines using dif-
19. Fujito Y, Izumi Y, Nakatani K, Takahashi M, ferential mobility-mass spectrometry. J Lipid
Hayakawa Y, Takayama M, Bamba T (2023) Res 55(8):1668–1677. https://doi.org/10.
Understanding the mechanism of CO2-assisted 1194/jlr.M046995
electrospray ionization for parameter optimiza- 28. Šala M, Lı́sa M, Campbell JL, Holčapek M
tion in supercritical fluid chromatography mass (2016) Determination of triacylglycerol regioi-
spectrometry. Anal Chim Acta 1246:340863. somers using differential mobility spectrome-
https://doi.org/10.1016/j.aca.2023.340863 try. Rapid Commun Mass Spectrom 30(2):
20. Takeda H, Izumi Y, Takahashi M, Paxton T, 256–264. https://doi.org/10.1002/rcm.
Tamura S, Koike T, Yu Y, Kato N, Nagase K, 7430
Shiomi M, Bamba T (2018) Widely-targeted 29. Baba T, Campbell JL, Le Blanc JCY, Baker
quantitative lipidomics method by supercritical PRS, Ikeda K (2018) Quantitative structural
fluid chromatography triple quadrupole mass multiclass lipidomics using differential mobil-
spectrometry. J Lipid Res 59(7):1283–1293. ity: electron impact excitation of ions from
https://doi.org/10.1194/jlr.D083014 organics (EIEIO) mass spectrometry. J Lipid
21. Takeda H, Takahashi M, Hara T, Izumi Y, Res 59(5):910–919. https://doi.org/10.
Bamba T (2019) Improved quantitation of 1194/jlr.D083261
lipid classes using supercritical fluid chroma- 30. Zhu Y, Wang W, Yang Z (2020) Combining
tography with a charged aerosol detector. J mass spectrometry with Paternò–Büchi reac-
Lipid Res 60(8):1465–1474. https://doi. tion to determine double-bond positions in
org/10.1194/jlr.D094516 lipids at the single-cell level. Anal Chem
22. Berger TA (2022) The evolution and current 92(16):11380–11387. https://doi.org/10.
state of instrumentation for analytical super- 1021/acs.analchem.0c02245
critical fluid chromatography. J Chromatogr B 31. Takahashi H, Shimabukuro Y, Asakawa D,
1211:123478. https://doi.org/10.1016/j. Yamauchi S, Sekiya S, Iwamoto S, Wada M,
jchromb.2022.123478 Tanaka K (2018) Structural analysis of phos-
23. Takeda H, Koike T, Izumi Y, Yamada T, pholipid using hydrogen abstraction dissocia-
Yoshida M, Shiomi M, Fukusaki E, Bamba T tion and oxygen attachment dissociation in
(2015) Lipidomic analysis of plasma lipopro- tandem mass spectrometry. Anal Chem
tein fractions in myocardial infarction-prone 90(12):7230–7238. https://doi.org/10.
rabbits. J Biosci Bioeng 120(4):476–482. 1021/acs.analchem.8b00322
https://doi.org/10.1016/j.jbiosc.2015. 32. Baba T, Campbell JL, Le Blanc JCY, Baker PRS
02.015 (2017) Distinguishing cis and trans isomers in
24. Takeda H, Izumi Y, Tomita A, Koike T, intact complex lipids using electron impact
Shiomi M, Fukusaki E, Matsuda F, Bamba T excitation of ions from organics mass spec-
(2017) Lipoprotein profiling methodology trometry. Anal Chem 89(14):7307–7315.
based on determination of apolipoprotein con- https://doi.org/10.1021/acs.analchem.
centration. Bioanalysis 9(1):9–19. https://doi. 6b04734
org/10.4155/bio-2016-0234 33. Kanehisa M, Araki M, Goto S, Hattori M,
25. Nishida-Aoki N, Izumi Y, Takeda H, Hirakawa M, Itoh M, Katayama T,
Takahashi M, Ochiya T, Bamba T (2020) Lipi- Kawashima S, Okuda S, Tokimatsu T, Yama-
domic analysis of cells and extracellular vesicles nishi Y (2007) KEGG for linking genomes to
from high- and low-metastatic triple-negative life and the environment. Nucleic Acids Res 36:
breast cancer. Metabolites 10(2):67. https:// 480–484. https://doi.org/10.1093/nar/
doi.org/10.3390/metabo10020067 gkm882
26. Burla B, Arita M, Arita M, Bendt AK, 34. Sud M, Fahy E, Cotter D, Azam K, Vadivelu I,
Cazenave-Gassiot A, Dennis EA, Ekroos K, Burant C, Edison A, Fiehn O, Higashi R, Nair
Han X, Ikeda K, Liebisch G, Lin MK, Loh KS, Sumner S, Subramaniam S (2016) Meta-
TP, Meikle PJ, Orešič M, Quehenberger O, bolomics workbench: an international
152 Hiroaki Takeda et al.

repository for metabolomics data and meta- 8(6):giz061. https://doi.org/10.1093/

data, metabolite standards, protocols, tutorials gigascience/giz061
and training, and analysis tools. Nucleic Acids 40. Knittelfelder O, Traikov S, Vvedenskaya O,
Res 44:463–470. https://doi.org/10.1093/ Schuhmann A, Segeletz S, Shevchenko A,
nar/gkv1042 Shevchenko A (2018) Shotgun lipidomics
35. Ding J, Ji J, Rabow Z, Shen T, Folz J, Brydges combined with laser capture microdissection:
CR, Fan S, Lu X, Mehta S, Showalter MR, a tool to analyze histological zones in cryosec-
Zhang Y, Araiza R, Bower LR, Lloyd KCK, tions of tissues. Anal Chem 90(16):
Fiehn O (2021) A metabolome atlas of the 9868–9878. https://doi.org/10.1021/acs.
aging mouse brain. Nat Commun 12(1): analchem.8b02004
6021. https://doi.org/10.1038/s41467- 41. Bligh EG, Dyer WJ (1959) A rapid method of
021-26310-y total lipid extraction and purification. Can J
36. Tsugawa H, Ikeda K, Takahashi M, Satoh A, Biochem Physiol 37(8):911–917. https://doi.
Mori Y, Uchino H, Okahashi N, Yamada Y, org/10.1139/o59-099
Tada I, Bonini P, Higashi Y, Okazaki Y, 42. Wang C, Wang M, Han X (2015) Comprehen-
Zhou Z, Zhu ZJ, Koelmel J, Cajka T, sive and quantitative analysis of lysophospholi-
Fiehn O, Saito K, Arita M, Arita M (2020) A pid molecular species present in obese mouse
lipidome atlas in MS-DIAL 4. Nat Biotechnol liver by shotgun Lipidomics. Anal Chem 87(9):
38(10):1159–1163. https://doi.org/10. 4879–4887. https://doi.org/10.1021/acs.
1038/s41587-020-0531-2 analchem.5b00410
37. Peng B, Kopczynski D, Pratt BS, Ejsing CS, 43. Okudaira M, Inoue A, Shuto A, Nakanaga K,
Burla B, Hermansson M, Benke PI, Tan SH, Kano K, Makide K, Saigusa D, Tomioka Y,
Chan MY, Torta F, Schwudke D, Meckelmann Aoki J (2014) Separation and quantification
SW, Coman C, Schmitz OJ, MacLean B, of 2-acyl-1-lysophospholipids and 1-acyl-2-
Manke MC, Borst O, Wenk MR, lysophospholipids in biological samples by
Hoffmann N, Ahrends R (2020) LipidCreator LC-MS/MS. J Lipid Res 55(10):2178–2192.
workbench to probe the lipidomic landscape. https://doi.org/10.1194/jlr.D048439
Nat Commun 11(1):2057. https://doi.org/ 44. Höring M, Stieglmeier C, Schnabel K,
10.1038/s41467-020-15960-z Hallmark T, Ekroos K, Burkhardt R, Liebisch
38. Pang Z, Chong J, Zhou G, de Lima Morais G (2022) Benchmarking one-phase lipid
DA, Chang L, Barrette M, Gauthier C, Jacques extractions for plasma Lipidomics. Anal Chem
PÉ, Li S, Xia J (2021) MetaboAnalyst 5.0: 94(36):12292–12296. https://doi.org/10.
narrowing the gap between raw spectra and 1021/acs.analchem.2c02117
functional insights. Nucleic Acids Res 49(1): 45. Hsiao JJ, Potter OG, Chu TW, Yin H (2018)
388–396. https://doi.org/10.1093/nar/ Improved LC/MS methods for the analysis of
gkab382 metal-sensitive analytes using medronic acid as
39. Molenaar MR, Jeucken A, Wassenaar TA, van a mobile phase additive. Anal Chem 90(15):
de Lest CHA, Brouwers JF, Helms JB (2019) 9457–9464. https://doi.org/10.1021/acs.
LION/web: a web-based ontology enrichment analchem.8b02100
tool for lipidomic data analysis. Gigascience
 Chapter 8

Rat Fecal Metabolomics-Based Analysis

Olga Deda, Helen G. Gika, and Georgios Theodoridis

Abstract
The gut’s symbiome, a hidden metabolic organ, has gained scientific interest for its crucial role in human
health. Acting as a biochemical factory, the gut microbiome produces numerous small molecules that
significantly impact host metabolism. Metabolic profiling facilitates the exploration of its influence on
human health and disease through the symbiotic relationship. Fecal metabolomics-based analysis is an
indisputably valuable tool for elucidating the biochemistry of digestion and absorption in the gastrointesti-
nal system, serving as the most suitable specimen to study the symbiotic relationship between the host and
the intestinal microbiota. It is well-established that the balance of the intestinal microbiota changes in
response to various stimuli, both physiological, such as gender, age, diet, and exercise, and pathological,
such as gastrointestinal and hepatic diseases. Fecal samples have been analyzed using widely adopted
analytical techniques, including NMR spectroscopy, GC-MS, and LC-MS/MS. Rat fecal samples are
frequently used and particularly useful substrates for metabolomics-based studies in related fields.
The complexity and diversity of fecal samples necessitate careful and skillful handling to extract metabo-
lites, while avoiding their deterioration, effectively and quantitatively. Several determinative factors, such as
the fecal sample weight to extraction solvent solution volume, the nature and pH value of the extraction
solvent, and the homogenization process, play crucial roles in achieving optimal extraction for obtaining
high-quality metabolic fingerprints, whether for untargeted or targeted metabolomics.

Key words Metabolomics, Metabolic profiling, Metabonomics, Sample preparation, Fecal samples,
Rats, Fecal extract, NMR, GC-MS, LC-MS/MS, Metabolites, Metabolome, Gut microbiome, Gut
microbiota

1 Introduction

The large reservoir of microbes that inhabit the intestinal epithe-

lium creates a distinctive symbiotic relationship with the host called
the “symbiome.” This symbiome functions as a hidden metabolic
organ [1], often referred to as a “forgotten” [2] or “invisible”
organ [3], due to its essential role in performing vital metabolic
processes. Recent scientific interest has dramatically increased,
highlighting the profound significance of the gut bacterial popula-
tion in maintaining human health, echoing Hippocrates’ wisdom

Olga Deda et al. (eds.), Metabolic Profiling: Methods and Protocols, Methods in Molecular Biology, vol. 2891,
https://doi.org/10.1007/978-1-0716-4334-1_8,
© The Author(s), under exclusive license to Springer Science+Business Media, LLC, part of Springer Nature 2025

153
154 Olga Deda et al.

from 2500 years ago that the root of good and bad health lies in the
gut [4].
Gut microbiome is literally a biochemical factory producing
numerous small molecules, derived either from the metabolism of
dietary compounds or from de novo biosynthesis. It contributes to
host metabolism in a great scale since it is estimated that approxi-
mately 50% of fecal and urine metabolites [5, 6] and 10% of the
blood metabolites originated from the gut microbiota [7].
Metabolomics extensively applies to studying how the gut
microbiome affects the host through their symbiotic relationship.
Various analytical methods investigate the metabolic effects result-
ing from gut microbiome disequilibrium (dysbiosis) for biomarker
discovery. Combining metabolomics with high-throughput DNA
and RNA sequencing reveals the link between the metabolome and
gut microbiome in host (human) health and disease [7–21].
While numerous biomarkers associated with the pathophysiol-
ogy of the gastrointestinal tract have been identified through meta-
bolomics analysis [22], there are instances where different
biomarkers were proposed, even when studying similar factors
such as biospecimen, population, or animal models. Several factors
could account for this variation, including experimental design
details, variations in sample collection and preparation, the selec-
tion of analytical parameters, and disparities in data handling and
statistical analysis. Human studies are further complicated by
genetic variations, diverse lifestyles, potential medications, and
other confounding factors, adding to the complexity of disease
biomarker discovery.

1.1 Metabolomics of Fecal samples serve as a valuable bio-specimen from a chemical/

Fecal Samples biochemical perspective, offering a powerful tool for the diagnosis
and prognosis of numerous gastrointestinal diseases [1, 23]. How-
ever, to derive reliable conclusions, raw fecal samples necessitate
appropriate preparation processes before analysis [24]. Unlike com-
mon samples, fecal samples exhibit complexity, heterogeneity, and a
high concentration of non-digested macromolecule [24–26], while
also reflecting the nutrients metabolism by gut microbiota
[27]. Furthermore, fecal samples should undergo special treat-
ment, due to their nature, for the avoidance of negative effects on
the analytical systems, regardless of the analytical technique subse-
quently applied. Sample preparation is a crucial aspect in
metabolomics-based analysis; it significantly impacts the quality of
the obtained results [28–30]. Inappropriate sample preparation
could lead to ineffective, non-repeatable, and poor extraction,
thus affecting the obtained metabolic profile. The golden rule is a
fast and repeatable, generic (non-selective) sample preparation pro-
cess, capable of extracting different chemical-class metabolites. The
perspective of “the less, the better” in sample preparation is highly
relevant and appealing in metabolomics, but it may not hold
 Rat Fecal Metabolomics-Based Analysis 155

entirely true for fecal samples. Fecal sample preparation should

prioritize effectiveness and precision to ensure reliable results.
Depending on the analytical method to be employed (MS or
NMR; untargeted or targeted approach) and the potential presence
of specific compounds of interest, the fecal sample preparation
process may vary. Even minor variations in sample preparation
procedures and conditions, even among similar methods, can sig-
nificantly impact the results. This could jeopardize the value of the
obtained results and hinder the establishment of metabolomics-
based biomarkers in clinical investigations. Usually, no special men-
tion is given to the reasons for which the specific protocol is
applied.
Undoubtedly, the ratio of fecal sample weight to extract solvent
volume has the greatest impact [26, 31, 32]. However, this does
not necessarily mean that denser samples lead to better results,
because there is a possibility that high density will create interfer-
ences in the analytical system.
Solvent extraction buffer is also a very strong factor affecting
the fecal metabolic profiling obtained by the most used analytical
techniques, namely, nuclear magnetic resonance (NMR) spectros-
copy, gas chromatography mass spectrometry (GC–MS), and liquid
chromatography mass spectrometry (LC–MS/MS). The nature of
the extraction solvent and the pH value of extraction solvent solu-
tion strengthen or weaken the extraction of metabolites. For exam-
ple, inappropriate choice of extraction solvent solution pH value
may lead to deterioration of fecal extract through hydrolysis of
some metabolites and affect the extraction capability of ionizable
metabolites.
Furthermore, homogenization, filtration, and centrifugation
are also parameters which can prove critical [33]. Various homoge-
nization methods are utilized, including sonication, mechanical
smashing using a mechanical crusher, or tissuelyser, proving partic-
ularly advantageous in NMR-based metabolomics [32, 33]. Another
effective option is bead-beating technology, which can swiftly lyse,
homogenize, and grind challenging samples in just a few seconds.
Disaggregating fecal sample prior to homogenization step could
enhance extraction efficacy [26].
While filtration may not be as critical as the previously men-
tioned processes in influencing fecal metabolic profiling, it remains
a proper practice to remove particulates and safeguard the analytical
system, particularly in LC-MS/MS-based metabolic profiling
[34]. To remove particulates, usually two cycles of centrifugation
are required: the first is in the early steps of the process and the
second immediately before the subsequent analysis.
Another practice, rather controversial, in fecal sample prepara-
tion is lyophilization. Although lyophilization eliminates the water
content [35], its use is not recommended as it has been proven to
affect the obtained fecal metabolic profiling. For example, it could
156 Olga Deda et al.

lead to reduced levels of short chain fatty acids (SCFAs) and poten-
tially of other volatile compounds [33, 36], thus editing the profile
especially in GC-MS mode.
Among the innumerous metabolite classes, SCFAs are high-
lighted to be of utmost importance and highly correlated to the gut
microbiome symbiosis and its alterations. The determination of
fecal SCFAs by an independent method provides the advantage of
a distinct process, both in sample preparation and analysis, most
suitable for the specific analytes; however, it has the cost of a second
analysis in addition to fecal metabolic profiling.
In addition, more than one extraction cycle could be per-
formed, and combination of extracts could take place [32]. This
practice is not proposed to keep the simplicity of the sample prepa-
ration process.
Finally, depending on the technique by which fecal samples are
analyzed, other conditions and parameters may affect the analysis
performance. For example, in NMR metabolomics-based analysis,
PBS buffer preparation (preparation in D2O, pH value, and salt
addition) appreciably results in collected spectra. The pH value of
PBS noticeably affects the behavior of the chemical shift reference
used. The use of deuterated solvents is not mandatory, but
improves the shimming, mainly in older spectrometers. Further-
more, the oversized signal of a non-deuterated solvent in 1H-NMR
could deteriorate the spectra’s quality. Sample preparation often
includes the addition of sodium azide as preservative and TSP as
chemical shift reference.
Undoubtedly, the derivatization process of the fecal sample is a
determining factor, regarding the GC-MS analysis. Critical factors
such as residual moisture, insufficient derivatization reagents, and
shorter or longer incubation, at lower or higher temperature, could
greatly degrade the obtained chromatograms [37, 38].
Choosing the right extraction solvent, like the mobile phase,
can lead to improved peaks in liquid chromatography. However,
using highly concentrated fecal extracts may cause column blockage
and break down the analytical system.

2 Materials

2.1 Laboratory 1. -80 °C freezer.

Equipment 2. Millipore Purification System for water.
3. Electronic analytical balance.
4. Mini shaker laboratory vortex.
5. Tissuelyser.
6. Ultrasonic tissue processor.
7. Micro refrigerated centrifuge.
8. Pure nitrogen gas generator.
 Rat Fecal Metabolomics-Based Analysis 157

9. Adjustable variable volume pipettes of 1000 mL, 100 mL,

20 mL, and 10 mL and pipette tips.
10. Eppendorf tubes of 2 mL.
11. Screw cap glass vials of 1.5 mL.
12. Syringes of 2.5 mL.
13. 25 mM diameter sterile syringe filters PTFE with 0.22 μm
pore siz.
14. NMR sample tubes, 5 mM, 7″ length (or appropriate size for
the spectrometer probe used).
15. GC-MS vials, caps, and inserts vials 250 μL.
16. LC-MS/MS glass vials, caps with PTFE/Silicone Septa and
low volume inserts.

2.2 Instrumentation 1. 500 MHz (or higher) NMR spectrometer equipped with a
Software 5 mM triple resonance probe at 300 K (or similar). The appro-
priate software to control acquisition of fecal sample (match-
ing, tuning, shimming) to set pulse calibration parameters and
to process.
2. Agilent 7890A GC coupled to a 5975C inert XL EI/CI MSD
with Triple-Axis Detector MS and a CTC-CH 4222 autosam-
pler with an Agilent HP-5 ms (29 m × 250 μm × 0.25 μm) in
split/splitless mode. MSD ChemStation (Agilent Technologies,
California, USA) to acquire and process GC-MS data.
3. ACQUITY UPLC coupled to a Xevo TQD MS System
(Waters, Massachusetts, USA) with an ACQUITY HILIC,
BEH amide column (2.1 × 150 mM, 1.7 μm). Waters Mas-
sLynx® software to collect and process LC-MS/MS data.

2.3 Reagents 1. 1-propanol (LC-MS grade).

2. Acetonitrile (LC-MS grade).
3. Methanol (LC-MS grade).
4. Methanol (HPLC grade) (see Note 1).
5. Deuterium Oxide (D2O) 99.96%.
6. Sodium dihydrogen phosphate monohydrate
(H2O•NaH2PO4).
7. Disodium hydrogen phosphate anhydrous (Na2HPO4) (see
Note 2).
8. Sodium chloride, granular/USP/FCC (see Note 3).
9. Pyridine anhydrous 99.8%.
10. Methoxyamine hydrochloride (MeOX).
11. N-methy-N-(trimethylsilyl)trifluoroacetamide (MSTFA).
12. Trimethylchlorosilane (TMCS).
158 Olga Deda et al.

13. Trimethylsilyl propanoic acid (TSP), (sodium salt of deuterated

molecule).
14. Ammonium formate, for MS ≥ 99.0% (see Note 4).
15. Formic acid eluent additive for LC-MS/MS (see Note 5).
16. H2O (LC-MS/MS grade).

3 Methods

3.1 Sample 1. Scientists should receive personal protective equipment before

Collection and Storage handling experimental animals.
2. Each animal should be placed in a separate cage during sample
collection, in case that the animals are kept in nonindividual
cage (see Note 6).
3. Fecal pellet should be collected immediately after defecation to
limit sample deterioration.
4. Fecal pellets should be collected using tweezers, which should
be cleaned from sample to sample.
5. More than one fecal pellet should be collected, to ensure an
adequate volume of fecal extract that can pass through the filter
during sample preparation.
6. Fecal pellets should be placed in a 2 mL Eppendorf tube.
7. Sodium azide could be added to protect fecal samples from
contamination.
8. Immediately after the collection, fecal samples should be placed
at -80 °C, for long-term storage.
9. Fecal pellets could be subjected to snap freezing with liquid
nitrogen, as this process better inhibits biochemical interactions.
10. When collecting samples directly from the intestine of rats,
proper procedures should be followed, including the use of
appropriate anesthesia and humane euthanasia, in compliance
with the relevant ethical regulations. For example, in EU
countries, this requires adherence to Directive 2010/63/EU,
which governs the protection of animals used for scientific
purposes. Ethical practices in keeping, handling, and treating
experimental animals should align with the principles of the
3Rs (Replacement, Reduction, and Refinement) as well as
institutional and national guidelines.

3.2 Sample 1. 500 mg of smashed stool material are placed in a 2 mL

Preparation Eppendorf tube.
3.2.1 Analysis Using 2. 1 mL Phosphate Buffer Saline (PBS: 1.9 mM Na2HPO4,
NMR Spectroscopy 8.1 mM NaH2PO4, 150 mM NaCl, pH 7.4) is added, in a
ratio of 1:2 fecal sample weight to extraction buffer volume (see
Note 7) [26]. PBS could be prepared either in D2O or in
 Rat Fecal Metabolomics-Based Analysis 159

distilled water. D2O should be added at least in a ratio of 1/5

(v/v) and then fill up to final volume with distilled water. NaCl
is added to facilitate the extraction of fecal samples. Preparing
PBS, dissolution of the salts requires vortexing and sonication.
3. Vortex-mixing is followed for 2 min.
4. The mixture is homogenized by ultrasonic homogenizer for
15 min.
5. Fecal slurry is centrifuged for 20 min at 4 °C (18,000 × g).
6. 400 μL of the clear supernatant are diluted with 150 μL of
D2O.
7. 50 μL of 0.1% 3-(Trimethylsilyl) propionic-2,2,3,3-d4 acid
sodium salt (TSP) in D2O is added in the extract.
8. Centrifugation for 18 min at 4 °C (15,000 × g).
9. Finally, 550 μL of the clear extract are placed in a 5 mM NMR
tube for analysis with methods as those described in Chap. 8, in
the previous book edition (ISSN 1940-6029).

3.2.2 Analysis Using GC- 1. 500 mg of smashed stool material are placed in a 2 mL
MS Eppendorf tube.
2. 1 mL of MeOH-CHCl3 1:1 (v/v) is added, in a ratio of 1:
2 fecal sample weight to extraction buffer volume.
3. Vortex-mixing is followed for 2 min.
4. The mixture is homogenized by sonication for 10 min.
5. Organic fecal extract is centrifuged for 20 min at 4 °C
(18,000 × g).
6. 100 μL of the supernatant are placed in a glass GC-MS vial and
evaporated under nitrogen stream.
7. 20 μL methoxyamine hydrochloride in pyridine (40 mg/mL)
and 180 μL MSTFA (1% trimethylchlorosilane [TMCS]) are
added followed by incubation for 90 min at 28 °C and for
30 min at 37 °C, respectively (see Note 8).
8. 5 μL of 1 mM solution of 2-fluorobiphenyl are added, as
internal standard. An alternative option is the addition of
5 μL pentadecan (in pyridine) solution. Before the addition of
internal standard, the vial should be allowed to fall to ambient
temperature (see Note 9).
9. The gas chromatographic analysis (GC-MS) of the fecal
extracts is performed as described in detail, in a previously
published paper from our group [26].

3.2.3 Sample 1. 250 mg of smashed stool material are placed in a 2 mL

Preparation for HILIC-MS Eppendorf tube.
2. Extraction solvent, 1-propanol or acetonitrile, is added in a
ratio of 1:4 fecal sample weight to extraction solvent volume
160 Olga Deda et al.

(see Note 10) [26, 34]. Acetonitrile better extracts a great wide
of metabolites, but 1-proponol preferably extracts amino acids
related to bowel disease.
3. Vortex-mixing is followed for 2 min.
4. The mixture is homogenized by sonication for 10 min.
5. Ultracentrifugation for 30 min at 4 °C (20,000 × g) is
followed.
6. Supernatants are filtered through syringe filters PTFE 0.22 μm.
7. The filtrate should be received in an Eppendorf tube and not
immediately in LC-MS vial which has a narrower opening.
8. The required sample volume, depending on the number of
subsequent injections, is added in the LC-MS vial insert prior
to analysis.
9. Analyze by targeted HILIC-MS/MS with a multi-analyte
method (e.g., as the one described in Chap. 10, in the present
book and in a previous published paper [39]) or an untargeted
method.
The applied protocols for analytical techniques used are
presented briefly in Fig. 1.

Fig. 1 Rat fecal sample preparation protocols for metabolomics-based analysis using NMR spectroscopy,
GC-MS, and LC-MS/MS
 Rat Fecal Metabolomics-Based Analysis 161

4 Notes

1. Used only for cleaning the analytical equipment.

2. Preparation of PBS buffer, as already described in Chaps. 8 and
14, in the previous book edition (ISSN 1940-6029).
3. 25 mg NaCl could be added for every 10 mL PBS buffer to
maximize extraction efficacy.
4. Used for the preparation of mobile phases in LC-MS/MS
system as described in Chap. 10, in the present book.
5. Used for the preparation of Wash Solvent and Purge Solvent of
LC-MS/MS systems as described in Chap. 10, in the
present book.
6. The rats of each cage should be appropriately marked so that
they are effectively identified, which is necessary for the collec-
tion of their samples. Herbal hair dye could be used for the
marking and should be renewed at regular time approximately
every 15 days.
7. Higher extraction ratios can be used (such as 1: 3) in case of
insufficient fecal sample quantity, to ensure sufficient aliquot of
fecal extract.
8. For a more effective derivatization process, the sample should
be placed in a glass insert in GC-MS glass vial.
9. Quick handling is required during the addition of the internal
standard, since the volatile derivatives could be evaporated
while opening the cap.
10. Lower extraction ratios can be used, such as 1: 3 or 1: 2, in
combination with supervision of the analytical system to pre-
vent problems in case of analysis of concentrated extracts.

References

1. Guinane CM, Cotter PD (2013) Role of the 5. Zheng X, Xie G, Zhao A et al (2011) The
gut microbiota in health and chronic gastroin- footprints of gut microbial-mammalian co-me-
testinal disease: understanding a hidden meta- tabolism. J Proteome Res 10(12):5512–5522
bolic organ. Ther Adv Gastroenterol 6(4): 6. Wang G, Huang S, Wang Y et al (2019) Bridg-
295–308 ing intestinal immunity and gut microbiota by
2. O’Hara AM, Shanahan F (2006) The gut flora metabolites. Cell Mol Life Sci 76(20):
as a forgotten organ. EMBO Rep 7(7): 3917–3937
688–693 7. Lyu Q, Hsu C-C (2018) Can diet influence our
3. Li X, Liu L, Cao Z et al (2020) Gut microbiota health by altering intestinal microbiota-derived
as an “invisible organ” that modulates the fecal metabolites? mSystems 3(2)
function of drugs. Biomed Pharmacother 8. Nicholson JK, Holmes E, Kinross J et al (2012)
121:109653 Host-gut microbiota metabolic interactions.
4. Sekirov I, Russell SL, Antunes LCM, Finlay BB Science 336(6086):1262–1267
(2010) Gut microbiota in health and disease.
Physiol Rev 90(3):859–904
162 Olga Deda et al.

9. Sircana A, Framarin L, Leone N et al (2018) 21. Huart J, Leenders J, Taminiau B et al (2019)

Altered gut microbiota in type 2 diabetes: just a Gut microbiota and fecal levels of short-chain
coincidence? Curr Diab Rep 18(10) fatty acids differ upon 24-hour blood pressure
10. Qu W, Nie C, Zhao J et al (2018) Microbiome- levels in men. Hypertens 74(4):1005–1013
metabolomics analysis of the impacts of long- 22. De Preter V, Verbeke K (2013) Metabolomics
term dietary advanced-glycation-end-product as a diagnostic tool in gastroenterology. World
consumption on C57BL/6 mouse fecal micro- J Gastrointest Pharmacol Ther 4(4):97–107
biota and metabolites. J Agric Food Chem 23. Holmes E, Li JV, Athanasiou T, Ashrafian H,
66(33):8864–8875 Nicholson JK (2011) Understanding the role
11. Zhao L, Lou H, Peng Y, Chen S, Zhang Y, Li X of gut microbiome-host metabolic signal dis-
(2019) Comprehensive relationships between ruption in health and disease. Trends Microbiol
gut microbiome and faecal metabolome in 19(7):349–359
individuals with type 2 diabetes and its compli- 24. Deda O, Gika HG, Wilson ID, Theodoridis
cations. Endocrine 66(3):526–537 GA (2015) An overview of fecal sample prepa-
12. Ayeni FA, Biagi E, Rampelli S et al (2018) ration for global metabolic profiling. J Pharm
Infant and adult gut microbiome and metabo- Biomed Anal 113:137–150
lome in rural bassa and urban settlers from 25. Bollard ME, Stanley EG, Lindon JC, Nichol-
Nigeria. Cell Rep 23(10):3056–3067 son JK, Holmes E (2005) NMR-based meta-
13. Dhakan DB, Maji A, Sharma AK et al (2019) bonomic approaches for evaluating
The unique composition of Indian gut micro- physiological influences on biofluid composi-
biome, gene catalogue, and associated fecal tion. NMR Biomed 18(3):143–162
metabolome deciphered using multi-omics 26. Deda O, Chatziioannou AC, Fasoula S et al
approaches. GigaScience 8(3) (2017) Sample preparation optimization in
14. Houttu N, Mokkala K, Laitinen K (2018) fecal metabolic profiling. J Chromatogr B Ana-
Overweight and obesity status in pregnant lyt Technol Biomed Life Sci 1047:115–123
women are related to intestinal microbiota 27. Hooper LV, Midtvedt T, Gordon JI (2002)
and serum metabolic and inflammatory pro- How host-microbial interactions shape the
files. Clin Nutr 37(6):1955–1966 nutrient environment of the mammalian intes-
15. Alexander K, van den Munckhof Inge CL, tine. Annu Rev Nutr 22:283–307
Lianmin C et al (2019) Gut microbial associa- 28. Theodoridis G, Gika H, Franceschi P et al
tions to plasma metabolites linked to cardiovas- (2012) LC-MS based global metabolite
cular phenotypes and risk. Circ Res 124(12): profiling of grapes: solvent extraction protocol
1808–1820 optimisation. Metabolomics 8(2):175–185
16. Maya-Lucas O, Murugesan S, Nirmalkar K et al 29. Gika H, Theodoridis G (2011) Sample prepa-
(2019) The gut microbiome of Mexican chil- ration prior to the LC-MS-based metabolo-
dren affected by obesity. Anaerobe 55:11–23 mics/metabonomics of blood-derived
17. Garcia-Mazcorro JF, Pedreschi R, Yuan J et al samples. Bioanalysis 3(14):1647–1661
(2019) Apple consumption is associated with a 30. Gika HG, Wilson ID, Theodoridis GA (2014)
distinctive microbiota, proteomics and meta- LC–MS-based holistic metabolic profiling.
bolomics profile in the gut of Dawley Sprague Problems, limitations, advantages, and future
rats fed a high-fat diet. PLoS One 14(3) perspectives. J Chromatogr B 966:1–6
18. Garcia-Mazcorro JF, Pedreschi R, Chew B, 31. Lamichhane S, Yde CC, Schmedes MS, Jensen
Dowd SE, Kawas JR, Noratto G (2018) Die- HM, Meier S, Bertram HC (2015) Strategy for
tary supplementation with raspberry extracts nuclear-magnetic-resonance-based metabolo-
modifies the fecal microbiota in obese diabetic mics of human Feces. Anal Chem 87(12):
db/db mice s. J Microbiol Biotechnol 28(8): 5930–5937
1247–1259 32. Wu J, An Y, Yao J, Wang Y, Tang H (2010) An
19. Chen Y-T, Yang N-S, Lin Y-C et al (2018) A optimised sample preparation method for
combination of lactobacillus Mali APS1 and NMR-based faecal metabonomic analysis. Ana-
dieting improved the efficacy of obesity treat- lyst 135(5):1023–1030
ment via manipulating gut microbiome in 33. Saric J, Wang Y, Li J et al (2008) Species varia-
mice. Sci Rep 8(1):1–14 tion in the fecal metabolome gives insight into
20. Kisuse J, La-ongkham O, Nakphaichit M et al differential gastrointestinal function. J Prote-
(2018) Urban diets linked to gut microbiome ome Res 7(1):352–360
and metabolome alterations in children: a com- 34. Deda O, Gika H, Panagoulis T, Taitzoglou I,
parative cross-sectional study in Thailand. Raikos N, Theodoridis G (2017) Impact of
Front Microbiol 9(June) exercise on fecal and cecal metabolome over
 Rat Fecal Metabolomics-Based Analysis 163

aging: a longitudinal study in rats. Bioanalysis ethyl chloroformate derivatization. Anal Bio-
9(1):21–36 chem 393(2):163–175
35. Bezabeh T, Somorjai RL, Smith ICP (2009) 38. Gao X, Pujos-Guillot E, Sébédio J-L (2010)
MR metabolomics of fecal extracts: applica- Development of a quantitative metabolomic
tions in the study of bowel diseases. Magn approach to study clinical human fecal water
Reson Chem 47(S1):S54–S61 metabolome based on trimethylsilylation deriv-
36. Monleon D, Garcia-Valles R, Morales JM et al atization and GC/MS analysis. Anal Chem
(2014) Metabolomic analysis of long-term 82(15):6447–6456
spontaneous exercise in mice suggests 39. Virgiliou C, Sampsonidis I, Gika HG,
increased lipolysis and altered glucose metabo- Raikos N, Theodoridis GA (2015) Develop-
lism when animals are at rest. J Appl Physiol ment and validation of a HILIC-MS/MS mul-
1985 117(10):1110–1119 titargeted method for metabolomics
37. Gao X, Pujos-Guillot E, Martin J-F et al (2009) applications. Electrophoresis 36(18):
Metabolite analysis of human fecal water by gas 2215–2225
chromatography/mass spectrometry with
 Chapter 9

Ion Pair Chromatography for Endogenous Metabolite LC-MS

Analysis in Tissue Samples Following HGH Resolution
Untargeted Acquisition
Filippos Michopoulos

Abstract
A protocol for the preparation of tissue extracts for the targeted analysis ca. 150 polar metabolites,
including those involved in central carbon metabolism, is described, using a reversed phase ion pair U
(H)PLC-MS method. Data collection enabled in high-resolution mass spectrometry detection provides
highly specific and sensitive acquisition of metabolic intermediates with wide range physicochemical
properties and pathway coverage. Technical aspects are discussed for method transfer along with the
basic principles of sample sequence setup, data analysis, and validation. At last general comments are
given to help the assessment of data quality and system performance.

Key words Polar metabolites, Ion pair chromatography, Mass spectrometry, Metabolite profiling

1 Introduction

Late 1980s investigations that would lead to the formation of a new

omic technology, subsequently called metabonomics, was estab-
lished predominately based on 1H NMR spectroscopy [1, 2]. The
unparalleled reproducibility and quantitative nature of NMR spec-
troscopy, combined with its inherently high structural information
content in combination with pattern recognition informatic tools,
opened up new horizons into understanding biological phenom-
ena. Despite the relative lack of sensitivity that limits the use of the
NMR technology to relatively highly abundant metabolites, the
technique remained the workhorse of the fast growing metabo-
nomics field until early in the twenty-first century [3–7]. Advances
in analytical instrumentation and particularly at the interface
between mass spectrometry with separation techniques (liquid or
gas chromatography, capillary electrophoresis, etc.) raised scientific
interest in these hyphenated technologies as they offered increased
coverage of the metabolome and generally much greater sensitivity

Olga Deda et al. (eds.), Metabolic Profiling: Methods and Protocols, Methods in Molecular Biology, vol. 2891,
https://doi.org/10.1007/978-1-0716-4334-1_9,
© The Author(s), under exclusive license to Springer Science+Business Media, LLC, part of Springer Nature 2025

165
166 Filippos Michopoulos

than 1H NMR spectroscopy [8–12]. The opportunities offered by

MS in combination with the plethora of available chromatographic
substrates and formats open a new horizon in metabolite detection
producing information rich data set that required the development
of new bioinformatic tools to deconvolute the resulting complex
data sets and visualize metabolic phenotypes [13–15]. Hypothesis
free, untargeted, metabolic profiling approaches greatly primarily
benefited from these technological advances in hyphenated analyti-
cal tools which enabled the establishment of differential metabolic
phenotypes across many research areas. However, successful
deployment of LC-MS-based tools had to overcome challenges
such as structure elucidation, metabolite ID determination, stan-
dardization, and reproducibility in untargeted metabonomics.
In the last decade, an increasing number of publications have
reported the development of targeted methodologies for metabo-
lite detection [16–19]. These assays utilize tandem mass spectro-
meters to selectively detect hundreds of known metabolic
intermediates and are reproducible, amenable to standardization
without requiring the often time-consuming structure elucidation
required by untargeted methodologies. The ability to customize
the selection of metabolite to target-specific biological process and
pathways (e.g., central carbon metabolism) has enabled analysts to
better optimize chromatographic parameters and improve the
detection of highly polar metabolites such as phosphorylated com-
pounds and carboxylic acids. The depth of information provided
from such “tailor-made” approaches has been driven by the need to
address more mechanistic biological questions. In the case of tar-
geted approaches for polar hydrophilic analytes, reversed-phase
chromatography is poorly suited to their analysis, and chro-
matographic strategies employing hydrophobic interaction liquid
chromatography (HILIC) [19–21] and ion pair chromatography
using [16–18] and to a lesser extend derivatization protocols
[22, 23] have been widely utilized. However, all these approaches
“bias” the detected metabolic phenotype, while a large proportion
of metabolic intermediates remain undetected. The development of
new high-resolution data-independent fragment acquisition (DIA)
methods such as SWATH has improved metabolite identification,
and in conjunction with retention time, libraries of authentic com-
pounds enable a more in-depth metabolic assessment of biological
systems.
Here we describe a high-resolution untargeted protocol using
ion pair chromatographic separation to resolve ~150 metabolic
intermediates on a C18 column. The protocol has been optimized
to provide semiquantitative data on tissue specimens although it
has also been used to analyze urine, plasma, serum, and samples
obtained from cell-based in vitro studies. The proposed methodol-
ogy can also be adapted to address demands for absolute quantifi-
cation; however, for such a purpose, we advise the development of a
 Ion Pair Chromatography for Endogenous Metabolite LC-MS Analysis in Tissue. . . 167

separate assay using the chromatographic separation outlined on

this chapter with additional careful evaluation of accuracy, preci-
sion, stability, and linear range of detection prior to providing
concentration values for the metabolites of interest.

2 Materials

2.1 Preparation of Water of chromatographic grade purity (18.2 MΩ), organic sol-
Analytical Standards, vents (acetonitrile, methanol, isopropanol), and mobile phase addi-
Test Mixture, and tives (acetic acid, tributylamine) of LC MS grade should be used.
Infusion Solution The same quality criteria must be applied for analytical standards to
obtain the highest available purity.
1. For all analytical standards listed in Table 1, the appropriate
amount is weighed in a 1.5 mL Eppendorf tube to result in a
1 mL 50 mM solution in MeOH/H2O 50/50 v/v (stock
solution). For amino acids and nucleotides, solubility is
improved by the addition of small amounts of HCl, while
other classes of metabolite may require small amounts of
1 M NaOH.
2. Stock solutions must be diluted 1/100 v/v with HPLC grade
water to produce a solution (dilution A) with a nominal con-
centration of 500 μM which will be further diluted 1/50 with
either HPLC water (dilution B) or MeOH/H2O 50/50 v/v
(infusion solution) to result to a final standard concentration of
10 μM.
3. Stock solutions, dilutions A and B should be transferred to
1.8 mL cryovials and stored at -20 °C as reference material
for future use.
4. An aliquot of 30 μL of the appropriate stock solution standards
(see Table 1) can be mixed in a 15 mL Falcon tube and
concentration adjusted to 100 μM with HPLC water before
dividing into 10 μL aliquots (test mixture 1–6) and stored at -
20 °C for in batch validation.

2.2 Liquid 1. To mix mobile phase A modifying reagents, use a 5 or 15 mL

Chromatography Falcon tube mix, 0.860 mL of acetic acid, 2.380 mL tributyla-
mine, and 1.0 mL of MeOH (see general remark 1). After a
quick manual shake, the content of the tube is added to a
990 mL of HPLC water in a 1 L solvent bottle, and then the
falcon tube is rinsed with 5.68 mL of HPLC water before
added to mobile phase A bottle. Shake mobile phase A bottle
manually for 1 min to ensure good solvent homogenization
before connecting to chromatographic system.
2. Prepare mobile phase B in a 1 L solvent bottle by mixing
800 mL of MeOH with 200 mL of isopropanol and manually
168 Filippos Michopoulos

Table 1
List of metabolites measured with the current methodology, mass spectrometer parameters, and text
mixtures composition

Molecular Q1 mass RT Test

Metabolite name formula (Da) (min) mixture
2’-Deoxyadenosine diphosphate C10H15N5O9P2 410.03 12.1 4
2’-Deoxyuridine triphosphate C9H15N2O14P3 466.97 13.4 5
2’-Deoxyadenosine monophosphate C10H14N5O6P 330.07 10 2
2’-Deoxyadenosine triphosphate C10H16N5O12P3 490.00 13.5 1
2’-Deoxycytidine diphosphate C9H15N3O10P2 386.02 11.4 2
2’-Deoxycytidine monophosphate C9H14N3O7P 306.0.05 7.6 5
2’-Deoxycytidine triphosphate C9H16N3O13P3 465.99 13.2 4
2’-Deoxyguanosine monophosphate C10H14N5O7P 346.06 9.3 6
2’-Deoxyuridine monophosphate C9H13N2O8P 307.04 8.9 1
3-(2-hydroxyphenyl) propionic acid C9H10O3 165.06 13 5
6-Phosphogluconic acid C6H13O10P 275.02 11.3 1
Acetoacetyl CoA C25H40N7O18P3S 424.56 14.1 5
Acetyl CoA C23H38N7O17P3S 808.12 14.2 1
a
Adenine C5H5N5 134.05 1.2 2
Adenosine C10H13N5O4 266.09 3.15
Adenosine diphosphate C10H15N5O10P2 426.03 11.9 2
Adenosine monophosphate C10H14N5O7P 346.06 9.6 1
Adenosine triphosphate C10H16N5O13P3 505.99 13.4 4
Adenosine-3′5’- cyclic monophosphate C10H12N5O6P 328.05 10.3 1
Adipic acid C6H10O4 145.06 10.8 3
ADP ribose C15H23N5O14P2 558.07 11.6 4
Arginine C6H14N4O2 173.11 0.8 4
Arginosuccinate C10H18N4O6 289.12 2.75
Asparagine C4H8N2O3 131.05 0.8 4
Aspartic acid C4H7NO4 132.04 3 4
Benzoic acid C7H6O2 121.04 11.9 4
Butyryl CoA C25H42N7O17P3S 417.58 15.1 5
cis aconitic acid C6H6O6 173.02 11.9 3
Citramalic acid C5H8O5 147.04 10.6 6
Citric acid C6H8O7 191.03 12 5

(continued)
 Ion Pair Chromatography for Endogenous Metabolite LC-MS Analysis in Tissue. . . 169

Table 1
(continued)

Molecular Q1 mass RT Test

Metabolite name formula (Da) (min) mixture
Citrulline C6H13N3O3 174.09 0.8 4
Coenzyme A (CoA) C21H36N7O16P3S 766.11 14
Coumaric acid C9H8O3 163.10 10.8 6
Creatine C4H9N3O2 130.07 0.8 4
Creatinine C4H7N3O 112.06 0.8 2
Crotonyl CoA C25H40N7O17P3S 834.14 14.4 4
Cystine C6H12N2O4S2 239.02 0.8 2
Cytidine monophosphate C9H14N3O8P 322.05 6.6 3
Cytosine C4H5N3O 110.04 0.8 4
Dihydroxy acetone phosphate C3H7O6P 168.99 13.7
Dinitrosalicylic acid C7H4N2O7 227.00 14.9 5
Ferulic acid C10H10O4 193.06 11.1 6
Flavin-adenine-dinucleotide C27H33P2N9O15 784.16 13.3 6
Folic acid C19H19N7O6 440.14 11.9 3
Fructose 1 phosphate C6H13O9P 259.03 6.1 5
Fructose 1,6 bisphosphate C6H14O12P2 339.00 11.8 1
Fructose 6 phosphate C6H13O9P 259.03 4.9 6
Fumaric acid C4H4O4 115.01 11.3 4
Galactose 1 phosphate C6H13O9P 259.03 4.7 4
Glucosamine 6 phosphate C6H14NO8P 258.05 0.8 5
Glucose 1 phosphate C6H13O9P 259.03 5.3 3
Glucose 6 phosphate C6H13O9P 259.03 4.3 1
Glucuronic acid C6H10O7 193.04 3.7 5
Glutamic acid C5H9NO4 146.05 2.6 5
Glutamine C5H10N2O3 145.07 0.8 5
Glutaric acid C5H8O4 131.04 10.4 5
Glutathione ox C20H32N6O12S2 611.15 10.4 6
Glutathione red C10H17N3O6S 306.08 5.9 5
Glyceraldehyde-3-phosphate C3H7O6P 169.00 11.4 5
Glycerate 1,3 bisphosphate C3H8O10P2 264.96 13.3 1
Glycerate 3 phosphate C3H7O7P 184.99 11.4 1

(continued)
170 Filippos Michopoulos

Table 1
(continued)

Molecular Q1 mass RT Test

Metabolite name formula (Da) (min) mixture
Glyoxylic acid C2H2O3 91.00 11.1 5
Guanine C5H5N5O 150.05 1.6 5
Guanosine C10H13N5O5 282.09 1.6 3
Guanosine diphosphate C10H15N5O11P2 442.02 11.5 5
Guanosine monophosphate C10H14N5O8P 362.06 8.7 3
Guanosine triphosphate C10H16N5O14P3 521.99 13.3 2
Guanosine-3′5’-cyclic monophosphate C10H12N5O7P 344.05 9.8 4
Histidine C6H9N3O2 154.07 0.8 4
Hydroxy glutaric acid C5H8O5 147.04 10.7 4
Hydroxy phenyl acetic acid C8H8O3 151.05 9.9 2
Hydroxymethyl glutaryl CoA C27H44N7O20P3S 910.16 14.4 6
Indole-2-carboxylic acid C9H7NO2 160.05 13.5 5
Inosine C10H12N4O5 267.08 1.5 3
Inosine monophosphate C10H13N4O8P 347.05 8.7 3
Isobutyryl CoA C25H42N7O17P3S 836.16 15.1 3
Isocitric acid C6H8O7 191.03 12.1 3
Isoleucine C6H13NO2 130.09 1.2 2
Isovaleryl CoA C26H44N7O17P3S 424.58 15.6 2
Itaconic acid C5H6O4 129.03 10.6 6
Kynurenic acid C10H7NO3 188.04 11.8
Kynurenine C10H12N2O3 207.08 1.8
Lactic acid C3H6O3 89.03 4.7 1
Leucine C6H13NO2 130.09 1.3 4
Maleic acid C4H4O4 115.01 9.9 3
Malic acid C4H6O5 133.02 10.7 2
Malonic acid C3H4O4 103.01 9.7 3
Malonyl CoA C24H38N7O19P3S 425.55 14.3 2
Mannose 6 phosphate C6H13O9P 259.03 4.4 2
Melatonin C13H16N2O2 231.12 11.2
Mercapturic acid C5H9NO3S 162.03 9
Mesaconic acid C5H6O4 129.03 11.3 2

(continued)
 Ion Pair Chromatography for Endogenous Metabolite LC-MS Analysis in Tissue. . . 171

Table 1
(continued)

Molecular Q1 mass RT Test

Metabolite name formula (Da) (min) mixture
Methionine C5H11NO2S 148.05 0.9 4
Methyl malonyl CoA C25H40N7O19P3S 432.06 14.4 4
Methylxanthine C6H6N4O2 165.05 2 6
N-acetylglucosamine C8H15NO6 220.09 0.8 6
Nicotinamide adenine dinucleotide C21H27N7O14P2 662.11 7.8 1
Nicotinamide adenine dinucleotide phosphate C21H29N7O17P3 743.08 11.7 5
Nicotinamide adenine dinucleotide phosphate C21H30N7O17P3 744.08 13.4 1
reduced
Nicotinamide adenine dinucleotide reduced C21H29N7O14P2 664.12 12 2
Nicotinic acid C6H5NO2 122.03 9.5 3
Nitrophenol C6H5NO3 138.03 10.4 6
OH Butyryl CoA C25H42N7O18P3S 852.15 14.1 6
Ophthalmic acid C11H19N3O6 288.13 6.4
Orotic acid C5H4N2O4 155.02 6.2 3
Palmitic acid C16H32O2 255.24 17.8
Pantothenic acid C9H17NO5 218.11 9.6 3
P-creatine C4H10N3O5P 210.03 10.5 2
Phenylalanine C9H8O3 164.05 2.5 6
Phosphoenolpyruvic acid C3H5O6P 166.98 11.8 1
Phthalic acid C8H6O4 165.03 12.5 6
Pimelic acid C7H12O4 159.07 11.5 1
Proline C5H9NO2 114.06 0.8 2
Proprionyl CoA C24H40N7O17P3S 822.14 14.6 6
Phosphoserine C3H8NO6P 184.00 6.7 1
Pyroglutamic acid C5H7NO3 128.04 5
Pyruvic acid C3H4O3 87.02 6.4 1
Quinolinic acid C7H5NO4 166.02 11.9
Riboflavin C17H20N4O6 375.14 9.9 5
Ribose 5 phosphate C5H11O8P 229.02 4.7 1
Ribulose 5 phosphate C5H11O8P 229.02 5.8 2
S-5-adenosyl-L-cysteine C13H18N6O5S 369.10 1.3

(continued)
172 Filippos Michopoulos

Table 1
(continued)

Molecular Q1 mass RT Test

Metabolite name formula (Da) (min) mixture
S-5-adenosyl-L-homocysteine C14H20N6O5S 383.12 2.1
Salicylic acid C7H6O3 137.03 13 6
Serine C3H7NO3 104.04 0.8 1
Shikimic acid C7H10O5 173.05 3.4 6
Sorbitol/mannitol C6H14O6 181.08 0.8 3
Succinic acid C4H6O4 117.03 9.9 1
Succinyl CoA C25H40N7O19P3S 866.13 14.5 6
Threonine C4H9NO3 118.06 0.8 2
Thymidine diphosphate C10H16N2O11P2 401.02 11.9 3
Thymidine monophosphate C10H15N2O8P 321.05 9.6 4
Thymidine triphosphate C10H17N2O14P3 480.99 13.5 2
a
Thymine C5H6N2O2 125.04 1.5 6
Tryptophan C11H12N2O2 203.09 3.8 6
Tyrosine C9H11NO3 180.07 1.1 6
UDP glucose C15H24N2O17P2 565.05 11 3
UDP glucuronic acid C15H22N2O18P2 579.03 13.1 3
Uracil C4H4N2O2 111.03 0.9 6
Uridine C9H12N2O6 243.07 1.10 3
Uridine diphosphate C9H14N2O12P2 403.00 11.5 3
Uridine monophosphate C9H13N2O9P 323.03 7.90 3
a
Valine C5H11NO2 116.08 0.80 4
Xanthine C5H4N4O2 151.03 1.00
Xanthurenic acid C10H7NO4 204.04 11.5
Xylulose 5 phosphate C5H11O8P 229.02 5.8 5
α-ketoglutaric acid C5H6O5 145.02 11.1 1
a
Weak ionization in negative ESI

shake for 1 min before connecting to chromatographic system

(see general remark 2).
3. In a 0.5 L solvent bottle, mix equal volumes of isopropanol
with acetonitrile and shake to ensure mixing (syringe and
needle wash).
 Ion Pair Chromatography for Endogenous Metabolite LC-MS Analysis in Tissue. . . 173

4. In a 0.5 L solvent bottle, mix 450 mL of HPLC water with

50 mL of isopropanol and shake to ensure mixing (Seal wash).

2.3 Tissue Extraction 1. In a 0.5 L solvent bottle, add 100 mL of HPLC water, 200 mL
Solvent of MeOH, and 200 mL acetonitrile before manual shake to
ensure mixing.

2.4 Reagents 1. Acetonitrile (VWR, 83642.320).

2. Methanol (VWR, 85800.320).
3. Isopropanol (VWR, 84881.320).
4. Acetic acid (Sigma Aldrich, A6283).
5. Tributylamine (Sigma Aldrich, 90781).
6. Negative calibration solution (SCIEX, 4463277 or 5042913).
7. HPLC vials (VWR, 548-0120A).
8. HPLC vials caps (VWR, 548-3206A).
9. Acquity premier HSS T3 1.8 μm 2.1*100 mM (Waters,
186009468).
10. Precellys extraction tubes (Bertin, soft tissue: P000918-
LYSK1-A.0, hard tissue: P000922-LYSK0-A).

3 Methods

3.1 Tissue Extraction Tissue samples must be processed and extracted from frozen to
and Sample reduce endogenous metabolite degradation. For the present pro-
Preparation tocol, we used a combined extraction homogenization approach
that is performed using a Precellys 24 system with an attached
temperature control unit. For soft tissue samples, we advise the
use of CK14 tubes, while for harder tissue the CK28R format is
more appropriate. Extraction and homogenization is achieved in
the appropriate tube format from a 50 mg of frozen tissue with
1 mL ACN/MeOH/H2O 40/40/20 v/v/v. Tubes must be
shaken at 5000 rpm for 20 sec, and the process must be repeated
three times with an intermittent pause of 30 sec between each
repeat. While the extraction and homogenization is taking place,
the extraction chamber unit must be at the lowest possible temper-
ature and not more than 10 °C. A clear supernatant is obtained after
centrifugation at 10691 g for 5 min at 0 °C, and this is transferred
to cryovial for storage at a minimum -20 °C until analysis. The
extraction-homogenization procedure is repeated with fresh sol-
vent as described above, and the resulted clear supernatant is com-
bined int the same cryovial.
Prior to LC-MS analysis, a minimum of a 1 in 10 dilution with
HPLC water must be performed to avoid excess carryover of highly
abundant metabolites across injections. For different tissues, an
174 Filippos Michopoulos

exploratory dilution experiment can be employed using different

dilution extract/total sample volume ratios (1/10, 1/20 v/v etc.)
to define the optimal dilution for a given study. Samples after water
dilution must be quickly vortexed and centrifuged at 2250 g for
10 min at 4 °C.
For batch validation and metabolite confirmation, a pooled
sample (QC) is prepared by mixing equal aliquots of the individual
extracts, and this is treated the same way as the study sample
extracts. The QC sample is also used to condition the analytical
platform at the beginning of the sample sequence as well as being
injected at regular intervals during the analytical sequence to assess
metabolite analytical reproducibility.
Metabolite identification and retention time confirmation is
obtained by cross comparison between text mixture, spiked QC,
and individual samples. The test mixture is an aqueous sample
consisting of test mixtures 1 to 6 diluted 1/20 v/v with HPLC
water. The spiked QC sample is a test mixture sample prepared in
the matrix (QC) of interest.

3.2 U(H)PLC Sample analysis according to the present protocol was performed
Separation on Thermo Ultimate 3000 RS pump combined with an Ultimate
3000 autosampler operating at 4 °C using a Acquity HSS T3 U
(H)PLC column (Waters Corp, 2.1 × 100 mM, 1.8 μm particle
size) with a column temperature maintained at 60 ± 0.5 °C. The
analysis was performed using a 5 μL injection of sample with
gradient elution at flow rate of 400 μL/min with a binary solvent
mixing schedule of 0 min, 0% B 0.5 min, 0% B; 4 min, 5% B; 6 min,
5% B; 6.5 min, 20% B; 8.5 min, 20% B; 14 min, 55% B; 15 min,
100% B; 17 min, 100% B; 18 min, 0% B; and 21 min 0% B.

3.3 Mass Spectrometric data was acquired on ABSCIEX 6600 triple TOF
Spectrometry instrument operating in a negative electrospray ionization acquisi-
tion mode. With high-resolution acquisition, there is no need to
optimize metabolite detection parameters such as declustering
potential, entrance potential, collision energy, and collision exit
potential. The accurate mass information given in Table 1 is calcu-
lated based on the molecular formula for each metabolite. Injection
(5 μL) of an authentic standard at concentration of 5 μM is suffi-
cient to confirm adequate retention on the chromatographic col-
umn and elution time (RT). Interface source parameters were
optimized at the flow rate (0.4 mL/min) of the intended chro-
matographic separation. For this purpose, 10 μL/min of each
metabolite (dilution A) was infused via the syringe pump to a
T-connector where it was combined with 0.39 mL/min 90%
mobile phase A via the UHPLC system delivering to ESI source a
final metabolite concentration of 12.5 μM. The optimal ESI source
parameters were as follows: ion spray voltage -4.5 kV, temperature
500 °C, collision energy -5 V, curtain gas 35, ion source gas(1) 60,
 Ion Pair Chromatography for Endogenous Metabolite LC-MS Analysis in Tissue. . . 175

ion source gas(2) 50, and declustering potential -10 V. Gas values
are arbitrary units. For co-eluting and isobaric species where the
TOF MS scan does not provide sufficient resolving capacity, a
multiple reaction monitoring high-resolution scan (MRMHR)
can be used to explore potential differential fragmentation and
resolve the molecules of interest.

3.4 Sample The sample sequence should be set up appropriately to provide a

Sequence clear measure of the data quality and system performance. Gradient
blank, solvent blank, test mixture, conditioning samples, quality
controls (QC), and spiked QC are additional to individual sample
injections that help in assessing the analysis quality. The use of these
injections and a brief description of the sample sequence is as
given here:
1. A gradient blank at the beginning of the samples sequence
helps to identify any signals/impurities/contaminants related
to the solvents used for chromatography or prolonged carry-
over in the system. The gradient blank is an acquisition where
no injection is performed and the chromatographic system
runs the gradient solvent schedule. As contaminants these can
be specific to different batches of solvent; this is an important
pre-analysis check as poor quality solvents can adversely affect
the analysis.
2. The use of two solvent blank injections after the gradient blank
helps the identification of signals related to solvent in which the
samples have been reconstituted/diluted. Consecutive injec-
tions of solvent blank injections enable the assessment of the
severity and persistence of any carryover.
3. System conditioning, which is essential to stabilize analyte
retention times, is performed with the same matrix as that of
the analysis sample set. The pooled sample, described in the
tissue extraction and preparation section above, is recom-
mended to be used unless volume limitations due to small
sample size prohibit its use. In these circumstances, a matrix
sample of the same origin from a different study can be sub-
stituted. System conditioning improves reproducibility of
detected signals, and the number of conditioning injections is
matrix and instrument saturation related. A minimal of 7 to 10
conditioning injections is recommended.
4. Test mixture and spiked QC injections are the first steps of
batch quality validation. After system conditioning, two injec-
tions of the test mixture followed by a spiked QC injection
enable retention time confirmation and assessment of the over-
all system performance by examination of the resulting metab-
olite resolution in the ion chromatogram. The first text mixture
176 Filippos Michopoulos

injection does not always provide the optimal chromatographic

separation; so 2 to 3 consecutive injections are recommended.
5. Due to the relatively high metabolite concentrations (5μM for
each) in both test mixture and spiked QC, which may result in
some modest carryover on the following injection, a further
two to three system conditioning injections (as per point
3 above) are recommended to re-equilibrate the chro-
matographic system for analysis of the sample set under
investigation.
6. QC injections are the foundation of the statistical analysis and
the assessment of the analytical reproducibility for each metab-
olite detected. The sample analysis should start and finish with
a QC injection to enable quality assessment at the beginning
and end of the individual sample analysis. A minimum of five
QC injections are required to obtain sufficient robustness for
the statistical analysis. Therefore, QC injections are performed
at regular intervals, interspersed between the individual test
samples.
7. Automatic in batch calibration every 5 injections using the
calibration delivery system (CDS) at 100 μL/min of negative
calibration solution is infused for 2 min at the mass spectrome-
ter in order to maintain mass accuracy during the analysis time.
8. Samples are analyzed in random order to reduce data bias, in
blocks of five to ten, sandwiched between two consecutive QC
injections. The number of injections in a sample set defines the
length of the sample block required to achieve the minimum of
five QC injections needed for robust statistical results.
9. Upon completion of the last QC injection, a spiked QC and
test mixture injection should be obtained to assess any metab-
olite retention time drift and intensity loss during the analysis
along the sample sequence.
10. Finishing the sample sequence with two injections of the sol-
vent blank and a blank gradient helps to explore carryover
across the analytical batch as well as any increase or decrease
on the background signal (threshold) essential for peak
integration.

3.5 Data Analysis Raw spectrometric data are processed with MuliQuant software to
obtain peak areas for each of the detected metabolites across the
sample set using the TOF MS high-resolution data with 20mDa
extraction window. The first step prior to peak integration is the
visual confirmation of the retention time for each metabolite peak
by comparing the trace obtained for the test mixture, spiked QC,
and the individual samples. Given the inability of obtain biological
matrix free from the endogenous metabolites being determined,
the solvent blank injection data can be used to define the
 Ion Pair Chromatography for Endogenous Metabolite LC-MS Analysis in Tissue. . . 177

background threshold values for each metabolite to ensure signals

are associated with actual metabolite presence and not the back-
ground. Smoothing factor and peak splitting are two more para-
meters that must be customized using the sample injections in
order to obtain accurate and reproducible peak integration. The
peak integration report can be exported as a text file from the
MultiQuant software and further processed in customer optimized
visualization and statistical software packages. For example, in Excel
univariate statistical analysis (f-test, t-test, anova) can be performed
to validated significant differences across meaningful group com-
parison on metabolite level very simply. It is also essential to extend
the univariate analysis and combine with analytical reproducibility
data (coefficient of variation values, CV) to reinforce statistical
significance. Data normalization is also essential to reduce bias
and trends due to analysis order or loss of signal intensity since no
internal standards are used in this analysis protocol. Median value
normalization is highly regarded as an adequate approach for this
type of data set. Typical validation criteria for detecting important
differences in metabolite concentrations across two group of sam-
ples can be regarded as the complete accomplishment of the fol-
lowing three: (1) Absolute natural logarithmic fold change >0.5,
coefficient of variation <30%, and p-value<0.05. Figure 1 provides
a representative overlay of extracted ion chromatograms of a text
mixture injection following the gradient elution sequence of the
proposed protocol.

4 General Remarks

1. Addition of 1 mL MeOH is critical to improve tributylamine

aqueous solubility. Add first tributylamine into an falcon tube,
followed by MeOH, and then the acetic acid before quick
manual shake. The homogenized Falcon tube content is then
added to water in mobile phase A.
2. Use glass cylinders to accurately measure solvent volume in
preparation of all solvent mixtures. This is to ensure consistent
mobile phase composition and greatly improve retention time
reproducibility of the chromatographic separation.
3. Prepare fresh mobile phase A before every analytical batch. This
helps in reducing background signals as well as bacterial growth
in the aqueous LC solvent.
4. Replace mobile phase B monthly to avoid retention time shift
due evaporation induced changes in solvent composition.
5. Monitor pressure buildup across the analytical batch to avoid
deterioration of chromatographic performance.
178 Filippos Michopoulos

Fig. 1 Typical extracted ion chromatograms of a reference metabolite mixture injection obtained following the
proposed chromatographic separation

6. Always prime solvent lines before commencing a new batch

analysis. This helps to reduce the formation of air bubbles in
the LC system.
7. Ensure enough needle and syringe wash is available to reduce
carryover between sample injections.
8. Clean the ESI source and Qjet after prolonged periods of use to
restore mass spectrometer sensitivity.
9. Data quality assessment is essential prior to data analysis steps.
Visual cross comparison of ion chromatograms (overlay them)
of two test mix injections—one acquired at the beginning and
the second one at the end of the sample sequence—must be
utilized to assess signal intensity lose or retention time drift and
reproducibility of the chromatographic resolution. Similar
examination must be performed using the first and last QC
sample injections. If signal intensity drops by more than 30%, it
is advised not to proceed further the data analysis. Similarly,
measures must be taken if retention time drifts more than 20 s
across the analysis time. Signal lose or retention time drift must
not be assessed using conditioning injection data.
 Ion Pair Chromatography for Endogenous Metabolite LC-MS Analysis in Tissue. . . 179

10. Data visualization with multivariate statistical tools as such of

principal component analysis (PCA) can be utilized to assess
more comprehensively data quality by examining QC group
injections forming a tight cluster on scores plot.

References
1. Bales JR, Bell JD, Nicholson JK, Sadler PJ, 10. Lenz EM, Wilson ID (2007) Analytical strate-
Timbrell JA, Hughes RD et al (1988) Meta- gies in Metabonomics. J Proteome Res 6(2):
bolic profiling of body fluids by proton NMR: 443–458
self-poisoning episodes with paracetamol (acet- 11. Lindon JC, Nicholson JK (2008) Analytical
aminophen). Magn Reson Med 6(3):300–306 technologies for metabonomics and metabolo-
2. Nicholson JK, Wilson ID (1989) High resolu- mics, and multi-omic information recovery.
tion proton magnetic resonance spectroscopy TrAC Trends Anal Chem 27(3):194–204
of biological fluids. Prog Nucl Magn Reson 12. Theodoridis G, Gika HG, Wilson ID (2008)
Spectrosc 21(4–5):449–501 LC-MS-based methodology for global metab-
3. Barzilai A, Horowitz A, Geier A, Degani H olite profiling in metabonomics/metabolo-
(1991) Phosphate metabolites and steroid hor- mics. TrAC Trends Anal Chem 27(3):251–260
mone receptors of benign and malignant breast 13. Shulaev V (2006) Metabolomics technology
tumors. A nuclear magnetic resonance study. and bioinformatics. Brief Bioinform 7(2):
Cancer 67(11):2919–2925 128–139
4. Gavaghan CL, Holmes E, Lenz E, Wilson ID, 14. Smith CA, Want EJ, O’Maille G, Abagyan R,
Nicholson JK (2000) An NMR-based metabo- Siuzdak G (2006) XCMS: processing mass
nomic approach to investigate the biochemical spectrometry data for metabolite profiling
consequences of genetic strain differences: using nonlinear peak alignment, matching,
application to the C57BL10J and Alpk: and identification. Anal Chem 78(3):779–787
ApfCD mouse. FEBS Lett 484(3):169–174 15. Xia J, Mandal R, Sinelnikov IV, Broadhurst D,
5. Kurhanewicz J, Vigneron DB, Nelson SJ, Wishart DS (2012) MetaboAnalyst 2.0 – a
Hricak H, MacDonald JM, Konety B et al comprehensive server for metabolomic data
(1995) Citrate as an in vivo marker to discrimi- analysis. Nucleic Acids Res 40(W1):W127–
nate prostate cancer from benign prostatic WW33
hyperplasia and normal prostate 16. Lu W, Bennett BD, Rabinowitz JD (2008)
peripheral zone: detection via localized proton Analytical strategies for LC–MS-based targeted
spectroscopy. Urology 45(3):459–466 metabolomics. J Chromatogr B 871(2):
6. Lynch MJ, Nicholson JK (1997) Proton MRS 236–242
of human prostatic fluid: correlations between 17. Buescher JM, Moco S, Sauer U, Zamboni N
citrate, spermine, and myo-inositol levels and (2010) Ultrahigh performance liquid chroma-
changes with disease. Prostate 30(4):248–255 tography- tandem mass spectrometry method
7. Marx A, de Graaf AA, Wiechert W, Eggeling L, for fast and robust quantification of anionic and
Sahm H (1996) Determination of the fluxes in aromatic metabolites. Anal Chem 82(11):
the central metabolism of Corynebacterium 4403–4412
glutamicum by nuclear magnetic resonance 18. Michopoulos F, Whalley N, Theodoridis G,
spectroscopy combined with metabolite bal- Wilson ID, Dunkley TP, Critchlow SE (2014)
ancing. Biotechnol Bioeng 49(2):111–129 Targeted profiling of polar intracellular meta-
8. Gika HG, Theodoridis GA, Earll M, Snyder bolites using ion-pair-high performance liquid
RW, Sumner SJ, Wilson ID (2010) Does the chromatography and-ultra high performance
mass spectrometer define the marker? A com- liquid chromatography coupled to tandem
parison of global metabolite profiling data gen- mass spectrometry: applications to serum,
erated simultaneously via UPLC-MS on two urine and tissue extracts. J Chromatogr A
different mass spectrometers. Anal Chem 1349:60–68
82(19):8226–8234 19. Gika HG, Theodoridis GA, Vrhovsek U, Mat-
9. Dunn WB, Ellis DI (2005) Metabolomics: cur- tivi F (2012) Quantitative profiling of polar
rent analytical platforms and methodologies. primary metabolites using hydrophilic interac-
TrAC Trends Anal Chem 24(4):285–294 tion ultrahigh performance liquid
180 Filippos Michopoulos

chromatography–tandem mass spectrometry. J 22. Kloos D, Lingeman H, Mayboroda OA,

Chromatogr A 1259:121–127 Deelder AM, Niessen WMA, Giera M (2014)
20. Schiesel S, L€ammerhofer M, Lindner W (2010) Analysis of biologically-active, endogenous car-
Multitarget quantitative metabolic profiling of boxylic acids based on chromatography-mass
hydrophilic metabolites in fermentation broths spectrometry. TrAC Trends Anal Chem 61:
of β-lactam antibiotics production by HILIC– 17–28
ESI–MS/MS. Anal Bioanal Chem 396(5): 23. Kloos D, Derks RJE, Wijtmans M,
1655–1679 Lingeman H, Mayboroda OA, Deelder AM
21. Yuan M, Breitkopf SB, Yang X, Asara JM et al (2012) Derivatization of the tricarboxylic
(2012) A positive/negative ion-switching, tar- acid cycle intermediates and analysis by online
geted mass spectrometry-based metabolomics solid-phase extraction-liquid chromatography–
platform for bodily fluids, cells, and fresh and mass spectrometry with positive-ion electro-
fixed tissue. Nat Protoc 7(5):872–881 spray ionization. J Chromatogr A 1232:19–26
 Chapter 10

HILIC-MS/MS Multi-targeted Method for Metabolomics

Applications
Christina Virgiliou, Helen G. Gika, and Georgios Theodoridis

Abstract
Metabolomics aims at identification and quantitation of key end point metabolites, basically polar, in order
to study changes in biochemical activities in response to pathophysiological stimuli or genetic modifica-
tions. Targeted profiling assays enjoying a growing popularity over the last years with LC-MS/MS as a
powerful tool for development of such (semi-)quantitative methods for a large number of metabolites.
Here we describe a method for absolute quantitation of ca. 100 metabolites belonging to key metabolite
classes such as sugars, amino acids, nucleotides, organic acids, and amines with a hydrophilic interaction
liquid chromatography (HILIC) system comprised with ultra (high) performance liquid chromatography
(UHPLC) with detection on a triple quadrupole mass spectrometer operating in both positive and negative
modes.

Key words HILIC-MS/MS, Metabolic profiling, Targeted metabolomics, Polar analytes

1 Introduction

Metabonomics or metabolomics, often described as the holistic

metabolic profile of complex matrices such as biological fluids,
tissue, and cell extracts, represents with genomics and proteomics
the major platforms in system biology [1]. Developments in analyt-
ical chemistry and particular advances in analytical technologies
made metabolomics a rapidly developed research field over the
past few decades [2, 3]. Although mass spectrometry and NMR
are the most popular analytical platforms used for metabolomics-
based studies, emphasis is lately placed on LC/MS approaches due
to wide range coverage of metabolites and its high efficiency
[4, 5]. Typically, MS-studies for metabolite profiling can be cate-
gorized in targeted and untargeted/holistic approaches [6]. Tar-
geted analysis is usually a hypothesis-driven strategy. It focuses
mainly on the measurement (identification and quantification) of
selected metabolites with known chemical properties; thus sample

Olga Deda et al. (eds.), Metabolic Profiling: Methods and Protocols, Methods in Molecular Biology, vol. 2891,
https://doi.org/10.1007/978-1-0716-4334-1_10,
© The Author(s), under exclusive license to Springer Science+Business Media, LLC, part of Springer Nature 2025

181
182 Christina Virgiliou et al.

preparation can be adapted in order to minimize limitations includ-

ing matrix effect [7, 8]. On the other hand, holistic approach
investigates the whole metabolic complement of the analyzed sam-
ple. However, it has been realized that there is a number of issues
that need to be resolved for untargeted MS studies including
standardization, robustness, and reproducibility that effect the
validity of the results [9–11]. Limitations of MS methods and
current developments on triple quadrupole instrumentation along
with advanced software capabilities resulted in the development of
tailored-made targeted metabolomics methods able to (semi)-
quantify tens of analytes of specific interest in a single injection
and provide solid, quantitative, and unambiguous data [12–16].
The development of a comprehensive method for targeted
metabolomics represents a challenge. The samples of interest usu-
ally contain highly polar metabolites with different physicochemical
properties, coexisting in samples at different concentration ranges.
Additional efforts are required for fine-tuning of all analytical para-
meters in order to find the optimum for most of the analytes
measured within the method. According to the literature, a number
of multi-analytes methods use hydrophilic interaction liquid chro-
matography (HILIC) separation for a simultaneous analysis of large
numbers of polar metabolites [17–23]. Reversed phase
(RP) chromatography is not able to retain polar metabolites while
although ion pair chromatography has been reported as the most
powerful option, ion-pair reagent contaminates the MS system at
major extend [14, 24–27]. However, innovations in column tech-
nologies/stationary phases have introduced increased options
available for the chromatography of polar metabolites [28, 29].
In the present protocol, the operation procedure for the iden-
tification and quantification of ca. 100 metabolites via (HILIC)
UPLC-MS/MS method is described. The aim of the method is
the identification and quantitation of key end point metabolites
known to exist in biological fluids (serum, plasma, urine, amniotic
fluid) in order to study their metabolic profile. Mass spectrometer
and chromatographic condition were optimized in order to achieve
satisfactory detection, quantification, maximum peak capacity, and
retention for as many metabolites as possible in a single run
[30, 31]. Absolute quantitation was performed using the standard
addition approach. Matrix effect and recovery were estimated, and
sample preparation procedure was optimized in order to reduce
matrix effect particularly for blood samples. Additionally, the
method was adapted and optimized for the absolute quantitation
of a subset of polar analytes, amino acids (AAs), and their deriva-
tives in urine, using isotope-labeled internal standards
(IS) [32]. Amino acids have been popular in targeted metabolomics
studies; however, due to their polar nature, they are poorly retained
by conventional RPLC. For that reason, various chemistries of
stationary phases such as HILIC, ion exchange, “mixed mode”
 HILIC-MS/MS Multi-targeted Method for Metabolomics Applications 183

chromatography, and modification of polar groups (e.g., derivati-

zation) have been investigated for the accurate quantitation of
amino acids in different matrices [15, 33–37].

2 Materials

All solvents used should be of LC/MS analytical grade. Use pur-

ified water (18.2 MΩ, at 25 °C). All used standards should be of
analytical or higher grade. Stock and working standard solutions
should be kept at -20 °C, and solvents containing >50% of water
should be replaced after 20 days (see Note 1).

2.1 Preparation of Classify compounds in different concentration groups: group A,

Stock and Working group B, and group C (details in Table 1) (see Note 2 and 3).
Standard Solutions Stock solutions of the analytes should be prepared in concen-
trations of (group A) 1000, (group B) 5000, and (group C)
10,000 mg/L in methanol/water, 1:1 (v/v), or water, depending
on analyte solubility (see Note 4). Prepare working standards mix-
tures from the stock solution by appropriate dilution with acetoni-
trile/water, 95:5 (v/v).
Calibration standards (9 standards) should be prepared by
serial dilution of the highest concentration standard.
For the standard mixture with the highest concentration (std
9): Mix 0.0032 mL of each standard in group A with 0.052 mL of
each standard in group B and 0.152 mL of each standard in group
C. In order to reach the final volume of 8 mL, dilute with 95:
5 MeCN:H2O. Repeat this step depending on the desired final
volume of std 9.

2.2 Preparation of Mobile Phase A

Mobile Phase for
Stock Buffer Solution Preparation
Liquid
Chromatography In a 50 mL beaker, weight 0.631 g ammonium formate and add
water (<50 mL) (MilliQ). Sonicate until salt is fully dissolved. Pour
the sonicated solvent into a 50 mL volumetric flash and add H2O
MilliQ till the calibration mark.
Final ammonium formate buffer concentration: 200 mM.

Mobile Phase A, 95:5 MeCN:H2O, 10 mM Ammonium

Formate
Pour the 50 mL buffer into a clean 1000 mL Schott bottle. Add
gradually 950 mL MeCN. Shake and sonicate for homogenization
and place in ultrasonic for degassing (see Note 5).
184 Christina Virgiliou et al.

Table 1
Concentration group and concentration (mg/L) of each group in standards 1–9

Metabolites Group Concentration Standard Mixture (mg/L)

2-Hydroxyisobutyric A STD 1 A 0.01
2-Hydroxyisovaleric acid A B 0.09
2-Methylhippuric acid A C 0.475
3-Methylhistidine B STD 2 A 0.02
4-Hydroxyphenyllactate B B 0.18
a-Ketoglutaric acid B C 0.95
Acetylcarnitine A STD 3 A 0.08
Adenine A B 0.72
Adenosine A C 3.8
Adipic acid A STD 4 A 0.2
Alanine B B 1.8
Arabitol B C 9.5
Arginine A STD 5 A 0.5
Ascorbic acid A B 4.5
Asparagine B C 23.75
Aspartic acid B STD 6 A 1
Benzoic acid A B 9
Betaine A C 47.5
Biotin A STD 7 A 1.2
Cadaverine A B 10.8
Caffeine A C 57
Choline A STD 8 A 1.5
Cotinine A B 13.5
Creatine A C 71.25
Creatinine A STD 9 A 2
Cystine B B 18
Cytidine A C 95
Cytosine A
Dimethylamine A
Folic acid A
Fructose B

(continued)
 HILIC-MS/MS Multi-targeted Method for Metabolomics Applications 185

Table 1
(continued)

Metabolites Group Concentration Standard Mixture (mg/L)

Fucose B
Fumaric acid A
g- aminobutyric A
Galactosamine A
Galactose B
Glucose C
Glutamic acid B
Glutamine B
Glycine B
Guanine B
Hippuric acid B
Histamine A
Homocysteine B
Hypotaurine A
Hypoxanthine A
Inosine A
Inositol B
Isoleucine A
Itaconic acid A
Kynurenic acid A
Lactic acid C
Lactose B
Leucine A
Lysine B
Malic acid A
Malonic acid B
Maltose B
Mannitol B
Methionine A
Methylamine A
Monoisoamylamine A

(continued)
186 Christina Virgiliou et al.

Table 1
(continued)

Metabolites Group Concentration Standard Mixture (mg/L)

N-Acetylaspartate A
Nicotinamide A
Nicotinic acid A
Ornithine B
Pantothenic acid A
Phenylalanine B
Picolinic acid B
Proline B
Putrescine A
Pyridoxine A
Pyroglutamic acid B
Pyruvic acid B
Ribose B
Riboflavine A
Sarcosine A
Serine B
Sorbitol B
Suberic acid B
Sucrose A
Taurine B
Theobromine A
Thiamine A
Threonine B
Thymidine A
Thymine B
Trimethylamine A
Trimethylamine-n-oxide A
Tryptamine A
Tryptophan B
Tyrosine B
Uracil C

(continued)
 HILIC-MS/MS Multi-targeted Method for Metabolomics Applications 187

Table 1
(continued)

Metabolites Group Concentration Standard Mixture (mg/L)

Uric acid C
Uridine A
Valine B
Vitamin B12 A
Xanthine B
Xanthurenic acid A
Xylitol B
Xylose B

Mobile Phase B
Stock Buffer Solution Preparation
In a 50 mL beaker, weight 0.2254 g ammonium formate and add
water (<250 mL) (MilliQ). Sonicate until salt is fully dissolved.
Pour dissolved buffer into a 250 mL volumetric flash and add H2O
MilliQ till the calibration line.
Final ammonium formate buffer concentration: 14.2 mM.

Mobile Phase B, 30:70 MeCN:H2O, 10 mM Ammonium

Formate
With the use of a clean volumetric cylinder, measure 210 mL buffer
and pour into a 500 mL Schott bottle. Measure 90 mL MeCN and
pour it gradually to the buffer. Shake and sonicate for homogeni-
zation and degassing.

Wash, Weak Solvent

Measure 420 mL water and pour into a clean 1 L Schott bottle;
measure 180 mL of methanol and add to water. Finally, add 0.1%
formic acid (HCOOH, 600 μL).

Purge, Strong Solvent

Measure 380 mL acetonitrile and pour into a clean 1 L Schott
bottle; measure 20 mL of methanol and add to water. Finally, add
0.1% formic acid (HCOOH, 400 μL).

Seal Wash Solvent

Measure 900 mL of water and pour into a 1 L Schott bottle;
measure 100 mL of MeCN and add to water. Mix thoroughly and
degas shortly in ultrasonic.
188 Christina Virgiliou et al.

Equipment and Colum

A Waters Acquity H-Class UPLC system coupled to Xevo TQD MS
spectrometer under control of MassLynx 4.1 was used for the
present protocol. In case where sample volume is limited, total
recovery vials (or similar type) with pre-slit silicone septa screw
caps 9 mM were used.
Chromatography was performed on an acquity BEH amide
column (2.1 mM × 150 mM, 1.7 μm) protected by an acquity
UPLC VanGuard pre-column.

3 Methods

Methods for LC-MS/MS are prepared and stored in the centrally

shared hard drive.

3.1 UPLC Method With regard to UPLC conditions, the parameters for binary solvent
manager are as follows: flow rate is kept constant throughout the
whole analysis 0.5 mL/min and following gradient is programmed:
4 min isocratic step at 100% A, then rising to 40% B linearly over the
next 21 min, and finally reaching 85% B over 5 min.The column
was equilibrated for 10 min in the initial conditions. Regarding the
Sample Manager, flow through needle system is applied in the
present protocol. Injection volume is set at 5 μL and sample tem-
perature at 6 °C. Injection system was subjected to two washing
cycles with a strong solvent and a weak solvent prior to injection
and one cycle of 6 s with strong solvent for post wash. Column
temperature is set to 40 °C.

3.2 MS Method In order to edit the MS method, find the optimum parameters for
MRM transition of each metabolite. For that protocol, manual
optimization was performed in order to find precursor and product
ions and the optimum cone voltage and collision energy. With
regard to capillary voltage, the best possible for most metabolites
detected in positive and negative ionization mode was applied. All
the other parameters were set according to tune page and the linked
calibration file.
Apply multiple reaction monitoring (MRM) mode for the
detection and quantification of all the compounds. Operate elec-
trospray ionization at polarity switching mode. Set capillary voltage
at +3.5 kV or -3.5 kV, block, and desolvation temperatures at 150 °
C and 350 °C, respectively. Set desolvation gas flow rate at 650 L/
h and cone gas at 50 L/h. Optimum cone voltage and collision
energy for each analyte after direct infusion and optimum time
window and dwell times are presented in Table 2.
System start-up and pretests.
Table 2
Analytes and their MRM conditions and retention time

Cone
Monoisotopic Precursor Product Voltage Colission Rt Molecular Dwell
A/A Name Formula Mass Ion Ion (V) Energy (V) Polarity (min) Weight Time
1 2-Hydroxyisobutyric C4H8O3 104.04 103 57 30 10 - 8.0 104.10 0.005
2 2-Hydroxyisovaleric acid C5H10O3 118.06 117 71 30 12 - 6.0 118.13 0.005
3 2-Methylhippuric acid C10H11NO3 193.20 192 148 35 12 - 8.2 193.20 0.005
4 3-Methylhistidine C7H11N3O2 169.09 170 109 30 10 + 19.0 169.18 0.003
5 4-Hydroxyphenyllactate C9H10O4 182.05 181 63 33 12 - 10.0 182.17 0.02
6 a-Ketoglutaric acid C5H6O5 146.01 145 101 20 9 - 16.0 146.11 0.02
7 Acetylcarnitine C9H17NO4 203.12 204 85 30 10 + 14.4 203.23 0.003
8 Adenine C5H5N5 135.05 136 119 40 20 + 3.6 135.13 0.003
9 Adenosine C10H13N5O4 267.10 268 136 20 15 + 4.4 267.24 0.003
10 Adipic Acid C6H10O4 146.06 145 101 25 12 - 16.0 146.14 0.005
12 Alanine C3H7NO2 89.05 90 44 20 10 + 16.0 89.09 0.005
13 Arabitol C5H12O5 152.07 151 89 25 10 - 9.9 152.14 0.02
14 Arginine C6H14N4O2 174.11 175 70 30 19 + 21.9 174.20 0.005
15 Ascorbic acid C6H8O6 176.03 176 70 20 15 + 3.6 176.12 0.02
16 Asparagine C4H8N2O3 132.05 133 74 20 14 + 18.2 132.11 0.02
17 Aspartic acid C4H7NO4 133.04 134 74 18 16 + 21.8 133.11 0.02
18 Benzoic acid C7H6O2 122.04 121 77 25 11 - 1.8 122.12 0.032
HILIC-MS/MS Multi-targeted Method for Metabolomics Applications

19 Betaine C5H11NO2 117.07 118 59 38 18 + 12.5 117.15 0.005

20 Biotin C10H16N2O3S 244.09 245 227 25 14 + 8.1 244.31 0.003
189

(continued)
Table 2
190

(continued)

Cone
Monoisotopic Precursor Product Voltage Colission Rt Molecular Dwell
A/A Name Formula Mass Ion Ion (V) Energy (V) Polarity (min) Weight Time
21 Cadaverine C5H14N2 102.12 103 86 15 8 + 20.4 102.17 0.003
22 Caffeine C8H10N4O2 194.08 195 138 38 18 + 0.9 194.19 0.032
23 Choline C5H14NO 104.11 104 60 40 22 + 7.0 104.17 0.003
Christina Virgiliou et al.

25 Cotinine C10H12N2O 176.09 117 80 30 20 + 1.1 176.22 0.032

26 Creatine C4H9N3O2 131.07 132 90 28 10 + 16.3 131.11 0.003
27 Creatinine C4H7N3O 113.06 114 88 30 10 + 4.8 113.11 0.003
29 Cystine C6H12N2O4S2 240.02 241 152 26 12 + 24.6 240.30 0.02
30 Cytidine C9H13N3O5 243.09 244 112 15 10 - 11.0 243.22 0.005
31 Cytosine C4H5N3O 111.04 112 95 40 14 + 7.5 111.10 0.003
32 Dimethylamine C2H7N 45.06 46 30 30 30 + 8.2 45.08 0.005
33 Folic acid C19H19N7O6 441.14 442 295 22 13 + 22.4 441.39 0.005
34 Fructose C6H12O6 180.06 181 140 5 8 + 11.8 180.16 0.01
35 Fucose C6H12O5 164.06 163 59 22 12 - 7.0 164.16 0.005
36 Fumaric acid C4H4O4 116.01 115 71 25 8 - 20.3 116.07 0.02
37 g-Aminobutyric C4H9NO2 103.06 104 69 22 15 + 17.3 103.12 0.005
38 Galactosamine C6H13NO5 179.07 180 72 18 18 + 17.8 179.17 0.01
39 Galactose C6H12O6 180.06 179 89 15 8 - 11.9 180.15 0.02
40 Glucose C6H12O6 180.06 179 59 25 16 - 14.6 180.15 0.02
41 Glutamic acid C5H9NO4 147.05 130 84 25 16 + 21.0 147.13 0.02
42 Glutamine C5H10N2O3 146.07 148 84 20 15 + 17.8 146.14 0.005
43 Glycine C2H5NO2 75.03 76 30 35 6 + 17.0 75.06 0.02
44 Guanine C5H5N5O 151.05 152 135 35 17 + 10.0 151.13 0.005
45 Hippuric cid C9H9NO3 179.06 178 134 32 11 - 9.4 179.17 0.005
46 Histamine C5H9N3 111.08 112 95 23 12 + 13.7 11.15 0.003
48 Homocysteine C4H9NO2S 135.03 134 88 10 8 - 16.4 135.19 0.005
49 Hypotaurine C2H7NO2S 109.02 110 92 22 18 + 15.8 109.14 0.015
50 Hypoxanthine C5H4N4O 136.04 137 110 40 18 + 4.8 136.11 0.003
51 Inosine C10H12N4O5 268.08 269 137 15 10 + 9.2 268.22 0.005
52 Inositol C6H12O6 180.06 181 109 15 10 + 17.6 180.16 0.02
53 Isoleucine C6H13NO2 131.09 132 86 25 12 + 13.3 131.17 0.003
54 Itaconic acid C5H6O4 130.03 129 85 20 8 - 14.4 130.09 0.02
55 Kynurenic acid C10H7NO3 189.04 190 172 32 12 + 9.8 189.16 0.005
56 Lactic Acid C3H6O3 90.03 89 43 30 10 - 11.7 90.08 0.02
57 Lactose C12H22O11 342.12 343 163 10 10 + 18.5 342.30 0.02
58 Leucine C6H13NO2 131.09 132 86 20 10 + 13.4 131.17 0.003
59 Lysine C6H14N2O2 146.11 147 84 14 14 + 22.5 146.19 0.01
60 Malic acid C4H6O5 134.02 133 115 22 10 - 20.2 134.08 0.02
61 Malonic acid C3H4O4 104.01 103 59 15 9 - 14.4 104.06 0.02
62 Maltose C12H22O11 342.12 341 161 25 8 - 18.1 342.30 0.02
63 Mannitol C6H14O6 182.08 183 69 15 11 + 13.5 182.17 0.02
HILIC-MS/MS Multi-targeted Method for Metabolomics Applications

65 Methionine C5H11NO2S 149.05 150 104 22 9 + 14.0 149.21 0.005

66 Methylamine CH5N 31.04 32 32 25 3 + 10.0 31.05 0.005
191

(continued)
Table 2
192

(continued)

Cone
Monoisotopic Precursor Product Voltage Colission Rt Molecular Dwell
A/A Name Formula Mass Ion Ion (V) Energy (V) Polarity (min) Weight Time
67 Monoisoamylamine C5H13N 87.10 88 43 18 11 + 4.6 87.10 0.003
68 N-Acetylaspartate C6H9NO5 175.04 176 134 18 10 + 20.1 175.14 0.005
69 Nicotinamide C6H6N2O 122.05 123 96 40 15 + 1.1 122.12 0.032
Christina Virgiliou et al.

70 Nicotinic acid C6H5NO2 123.03 124 80 38 18 + 10.5 123.10 0.005

71 Ornithine C5H12N2O2 132.09 133 70 45 10 + 22.7 132.16 0.02
72 Pantothenic acid C9H16NO5 218.10 220 90 25 13 + 12.2 219.23 0.005
73 Phenylalanine C9H11NO2 165.08 166 120 22 12 + 12.6 165.19 0.003
74 Picolinic acid C6H5NO2 123.03 124 106 25 10 + 28.9 123.10 0.005
75 Proline C5H9NO2 115.06 116 70 20 20 + 14.5 115.13 0.02
76 Putrescine C4H12N2 88.10 89 72 15 8 + 21.0 88.15 0.003
77 Pyridoxin C8H11NO3 169.07 170 152 28 12 + 2.0 169.18 0.003
78 Pyroglutamic acid C5H7NO3 129.04 130 84 30 12 + 15.0 129.11 0.005
79 Pyruvic acid C3H4O3 88.02 89 48 20 12 + 7.1 88.06 0.02
81 Ribose C5H10O5 150.05 149 89 22 10 - 4.3 150.13 0.005
82 Riboflavine C17H20N4O6 376.14 377 243 38 22 + 9.6 376.36 0.005
83 Sarcosine C3H7NO2 89.05 90 44 20 8 + 15.3 89.09 0.005
84 Serine C3H7NO3 105.04 106 60 20 10 + 17.9 105.09 0.005
85 Sorbitol C6H14O6 182.08 181 101 35 10 - 13.5 182.17 0.02
87 Suberic acid C8H14O4 174.09 173 111 35 12 - 10.7 174.20 0.005
88 Sucrose C12H22O11 342.12 341 179 40 14 - 16.8 342.29 0.02
89 Taurine C2H7NO3S 125.01 126 108 25 10 + 14.4 125.14 0.005
90 Theobromine C7H8N4O2 180.06 181 163 35 17 + 1.1 180.16 0.032
91 Thiamine C12H17N4OS 265.11 265 122 22 12 + 11.8 300.81 0.003
92 Threonine C4H9NO3 119.06 120 74 20 10 + 10.0 119.11 0.005
93 Thymidine C10H14N2O5 242.09 243 127 11 9 + 1.6 242.22 0.032
94 Thymine C5H6N2O2 126.04 127 110 40 19 + 1.2 126.11 0.032
95 Trimethylamine C3H9N 59.07 60 45 31 10 + 5.6 59.11 0.005
96 Trimethylamine-n-oxide C3H9NO 75.07 76 59 28 10 + 13.0 75.10 0.003
97 Tryptamine C10H12N2 160.10 161 144 15 11 + 4.1 160.21 0.003
98 Tryptophan C11H12N2O2 204.09 205 146 20 15 + 12.7 204.22 0.005
99 Tyrosine C9H11NO3 181.07 182 136 22 13 + 14.5 181.19 0.005
100 Uracil C4H4N2O2 112.03 113 70 40 15 + 1.9 112.08 0.02
101 Uric acid C5H4N4O3 168.03 169 141 35 15 + 16.3 168.11 0.02
102 Uridine C9H12N2O6 244.07 243 110 35 16 - 4.7 244.12 0.005
103 Valine C5H11NO2 117.08 118 72 20 10 + 14.4 117.15 0.003
104 Vitamin B12 C63H89CON14O14P 1355.58 678 147 42 27 + 19.4 1335.37 0.005
105 Xanthine C5H4N4O2 152.03 153 136 33 15 + 7.4 152.11 0.01
106 Xanthurenic acid C10H7NO4 205.04 206 188 35 12 + 12.8 205.17 0.02
107 Xylitol C5H12O5 152.07 151 89 30 10 - 9.9 152.15 0.02
108 Xylose C5H10O5 150.13 149 89 25 7 - 8.0 150.13 0.005
HILIC-MS/MS Multi-targeted Method for Metabolomics Applications
193
194 Christina Virgiliou et al.

Follow each step here to ensure high quality of your data. For
the system start-up, the system is prepared for analysis and checked
for system suitability.
1. Install mobile phase A, B, wash, purge, and seal wash, and in
the two left channels, install neat acetonitrile and water.
2. Prime eluents for 3 min each.
3. Connect the BEH amide column.
4. Set the column temperature to 40 °C and the temperature of
sample manage to 6.
5. Flush the column with 95% MeCN and 5% water for 15 min at
flow of 0.2 mL/min.
6. Increase flow rate gradually until 0.5 mL/min.
7. Flush column until equilibration (psi delta < 20) (see Note 6).
8. Load UPLC method.
For MS, load the appropriate AcquityDB file with the extension
.ipr and the most recent calibration file.

3.3 Sample Extraction of samples may vary between different matrices. So far,
Preparation the described method or its variants were tested with serum, urine,
amniotic fluid, intra/extra-cellular content, feces, and various types
of animal tissue but also in foods such as honey, muscle tissue, and
flour. The sample preparation procedure for blood serum samples is
presented as follows:
1. Allow samples to thaw in room temperature.
2. Mix 50 μL of sample with 130 μL MeCN, 10 μL H2O, and
10 μL MeOH in an Eppendorf vial of 1 mL by the use of
variable volume pipette 20–200 μL (see Note 7)
3. Vortex for 10 min and centrifuge at 7000rpm for 10 min.
4. Transfer supernatant to an LC-MS with glass insert vial.
5. Immediately transfer to precooled Sample Manager.
For standard addition approach, sample preparation of spiking
samples is as follows (see Note 8):
1. Mix 50 μL of sample with 100 μL of standard calibration
mixture 1 and 50 μL of MeCN in an Eppendorf vial of 1 mL
by the use of variable volume pipette 20–200 μL.
2. Vortex for 10 min and centrifuge at 7000rpm for 10 min.
3. Transfer supernatant to an LC-MS with glass insert vial.
4. Repeat steps 1–3 using increased concentration standard cali-
bration mixtures (at least 3 in total).
5. Immediately transfer to precooled Sample Manager.
 HILIC-MS/MS Multi-targeted Method for Metabolomics Applications 195

In case of external calibration, transfer standard curve to

LC-MS vial with glass insert (see Note 9 and 10).
For quality control sample (QC sample), in order to evaluate
systems stability, mix equal volumes of all samples of the dataset and
follow the sample preparation procedure.

3.4 UPLC-MS/MS The following steps describe how to setup a sample table for data
Analysis acquisition. Following a sample order has been evaluated to be
optimal for both throughput and quality controls:
1. Run a gradient without injection in order to evaluate column
performance and have a measure of the minimum and maxi-
mum pressure during analysis (see Note 11).
2. Perform six replicate injections of a standard mixture in order
to pretest systems performance (retention time, signal) and to
equilibrate system (see Note 12).
During analysis with external calibration approach
1. Run a before-batch calibration curve.
2. Run injections of a QC sample in order to perform matrix
equilibration of the system. The number of equilibration injec-
tions depends on the analyzed matrix. For cell media and urine,
five injections may be adequate; for blood and tissue samples,
higher numbers may be necessary, depending also on the use of
the column (new columns need more injections to saturate
active sites/equilibrate).
3. Samples are injected in random order in blocks of ten samples.
4. After each block of ten samples, injection of a standard mixture
and a QC sample is performed.
5. Repeat steps 2 and 3 for a maximum of 100 samples (see Note
13).
6. Run an after-batch calibration curve.
During analysis with standard addition approach
1. Run injections of a QC sample in order to perform matrix
equilibration of the system.
2. Samples followed by their paired spiking samples are injected in
order in block of ten samples.
3. After each block of ten samples, injection of a standard mixture
and a QC sample is performed.
4. Repeat steps 2 and 3 for a maximum of 100 samples.
After finishing the sample, batch analysis column and the MS
have to be cleaned for further use.
196 Christina Virgiliou et al.

1. Flush the column with 50% MeCN and 50% H2O for 50 min,
column temperature 50 °C, flow 0.2 mL/min.
2. Flush the column with 50% MeCN and 50% H2O for 30 min,
column temperature 40 °C, flow rate 0.5 mL/min.
3. Clean MS cone as recommended.

3.5 Data Treatment: Quantitation can be performed in both manually and automated
Quantification of ways using vendor software. In the present protocol, TargetLynx
Metabolites (Waters) was used. A Quantify Method must be created before
integration or quantification can be performed. Related software
from other vendors are MutliQuant (Sciex), Xcalibur, and Mas-
sHunter WorkStation - Quantitative Analysis.
TargetLynx data can be saved as .qld files for further manipula-
tion. Complete the summary of the results and .qld files (area,
response, concentration, S/N, SD, measured concentration etc.)
can be exported as .txt files and open with excel Microsoft program
for further treatment.
In case of external calibration curve approach, once you find
the optimum method parameters, calibration of standards and
quantification can be performed directly with only one process
(see Note 14).
When standard addition method is applied, automated quanti-
tation is not available. Integrated results for each sample and
spiked/fortified samples together with corresponding spiking
levels must be imported in a spreadsheet program (excel or similar)
where unknown concentrations of metabolites in samples can be
calculated as intercept/slope.

4 Absolute Quantitation of AAs in Urine

The described method was adapted and optimized for the absolute
quantitation of 20 AAs (alanine, arginine, asparagine, aspartic acid,
cysteine, glutamic acid, glutamine, glycine, histidine, isoleucine,
leucine, lysine, methionine, phenylalanine, proline, serine, threo-
nine, tryptophan, tyrosine, valine) and 9 AA derivatives (betaine,
choline, hypotaurine, pyroglutamic acid, taurine,
3-methylhistidine, 4-hydroxyproline, sarcosine, creatine).
Preparation of calibration standards includes the division of
AAs into five separate groups according to their concentration in
urine, namely, A, B, C, D and E. Standard working solutions were
freshly prepared before the analysis by dilution with ACN: H2O
(95:5, v/v) at nine concentration levels. Spike the calibrators in
50 μL of synthetic urine (see Note 15) to get a final concentration
range of 0.01–2 μg/mL for group A, 0.02–4 μg/mL for group B,
0.09–18 μg/mL for group C, 0.15–30 μg/mL for group D,
0.425–95 μg/mL for group E, and 1–200 μg/mL for group F
(Table 3). Add, as IS, in all calibrators or QCs an equal volume
HILIC-MS/MS Multi-targeted Method for Metabolomics Applications 197

Table 3
Concentration group and concentration (mg/L) of each group in standards
1–9

A B C D E F
std1 0.02 0.04 0.18 0.3 0.95 2
std2 0.04 0.08 0.36 0.6 1.9 4
std3 0.16 0.32 1.44 2.4 7.6 16
std4 0.4 0.8 3.6 6 19 40
std5 1 2 9 15 47.5 100
std6 2 4 18 30 95 200
std7 2.4 4.8 21.6 36 114 240
std8 3 6 27 45 142.5 300
std9 4 8 36 60 190 400
Metabolites Group concentration
Alanine C
Arginine A
Asparagine D
Aspartate C
Choline B
Creatine E
Cysteine A
Glutamate C
Glutamine F
Glycine F
Histidine E
Hypotaurine A
Isoleucine A
Leucine B
Lysine F
Methionine A
Betaine B
Phenylalanine C
Proline D
Pyroglutamic acid F
Sarcosine A

(continued)
198 Christina Virgiliou et al.

Table 3
(continued)

A B C D E F
Serine D
Taurine F
Threonine E
Tryptophan C
Tyrosine D
Valine C
3-methylhistisine E
4-hydroxyproline A

(30 μL) of isotope-labeled AA mixture solution containing 20 AAs.

Prepare fresh the IS solution by a 1:100 (v/v) dilution of the cell
free amino acid mixture—13C,15N in ACN:H2O (95:5, v/v).
Adjust the mobile phases A and B pH at 3 by the addition of
formic acid (1 mL for A and 700 μL for b) in the aqueous phase
before mixing with the organic solvent. To shorten the analysis
time, you can apply the following 10 min gradient with a linear
increase from 20% to 50% B; return the system to the initial condi-
tions within 0.1 min and re-equilibrate to the initial conditions for
4 min. The injection volume can be reduced to 3 μL.
Add to the MS method the MRM transitions of the isotope-
labeled amino acids after optimization of parent and daughter ion,
cone voltage, collision energy, and dwell time. The optimum values
after manual optimization are presented in Table 4.
The sample preparation procedure for urine samples is pre-
sented as follows:
1. Allow samples to thaw in room temperature.
2. Mix 50 μL of sample with 120 μL MeCN:H2O (95:5, v/v) and
30 μL IS in an Eppendorf vial of 1.5 mL by the use of variable
volume pipette 20–200 μL.
3. Vortex for 10 min and centrifuge at 6700 × g for 10 min.
4. Transfer supernatant to an LC-MS with glass insert vial.
5. Immediately transfer to precooled Sample Manager.

Calibration Curve and Quantitation of AAs

For the 20 AA analytes that isotope-labeled IS is available, prepare
the calibration curve based on the response factor (analyte area/ IS
area), while for the AA derivatives, construct the calibration based
on the peak areas of the analytes. Chemically relevant labeled ISs
 HILIC-MS/MS Multi-targeted Method for Metabolomics Applications 199

Table 4
MRM conditions and retention time of amino acids and amino acid derivatives

Precursor
Name ion Product ion Cone voltage (V) Collision energy (V) Rt (min)
Glycine 76 30 35 6 3.8
Glycine 13C 15N 79 32 35 6 3.9
Sarcosine 90 44 20 8 3.1
Alanine 90 44 20 10 3.4
Alanine 13C 15N 94 47 20 10 3.4
Choline 104 59 10 10 1.5
Serine 106 60 20 10 4.3
Hypotaurine 110 92 22 8 3.4
Serine 13C 15N 110 92 20 10 3.3
Proline 116 70 20 25 2.7
Betaine 118 58 10 10 2.2
Valine 118 72 18 9 2.5
Threonine 120 74 20 10 3.6
Proline 13C 15N 122 76 20 25 2.6
Cysteine 122 76 18 12 3.7
Valine 13C 15N 124 77 18 9 2.5
Threonine 13C 15N 125 78 20 10 3.7
Taurine 126 108 25 10 2.7
Pyroglutamic 130 84 30 12 2.3
4OH Proline 132 68 30 25 3.4
Isoleucine 132 69 25 12 2.1
leucine 132 86 20 10 1.9
Creatine 132 90 28 10 3.4
Asparagine 133 74 20 14 4.4
Aspartic acid 134 74 18 16 5.9
Isoleucine 13C 15N 139 92 25 12 2.1
Leucine 13C 15N 139 92 20 10 1.9
Asparagine 13C 15N 139 77 20 14 4.5
Aspartic acid 13C 15N 139 77 18 16 5.9
Glutamine 147 84 20 15 4.2
Lysine 147 84 14 14 6.5

(continued)
200 Christina Virgiliou et al.

Table 4
(continued)

Precursor
Name ion Product ion Cone voltage (V) Collision energy (V) Rt (min)
Glutamic acid 148 130 20 13 4.8
Methionine 150 104 22 9 2.3
Glutamine 13C 15N 154 89 20 15 4.2
Glutamic acid 13C 15N 154 136 20 13 4.8
Lysine 13C 15N 155 90 14 14 6.5
Histidine 156 110 22 13 6.2
Methionine 13C 15N 156 109 22 9 2.4
Histidine 13C 15N 165 118 22 13 6.6
Phenylalanine 166 120 22 12 1.9
3-Methylhistidine 170 109 30 10 6.2
Arginine 175 70 30 19 6.2
Phenylalanine 13C 15N 176 129 22 12 1.9
Tyrosine 182 136 22 13 2.6
Arginine 13C 15N 185 75 30 19 6.1
Tyrosine 13C 15N 192 145 22 13 2.6
Tryptophan 205 146 20 15 1.9
Tryptophan 13C 15N 205 188 20 15 1.8

have been found to be effective as surrogate IS; thus in the case of

pyroglutamate (5-oxoproline), 3-methylhistidine and
4-hydroxyproline use 13C15N L-glutamic acid, 13C15N histidine,
and 13C15N proline, respectively (Fig. 1). Fit all the standard
curves to a 1/x2 weighted linear regression.

5 Notes

1. This will help to avoid growth of bacteria.

2. Classification of compounds in the current report is based on
their reference concentration in serum according to the litera-
ture and HMDB database.
3. Compounds are present in biological fluids at different concen-
tration ranges so their grouping is important in terms of
quantitation.
 HILIC-MS/MS Multi-targeted Method for Metabolomics Applications 201

Fig. 1 Calibration curves comparing parallelism for 3-methylhistidine (a: area on y-axis, b response factor
area 3-methylhistidine/area 13C15N histidine) and pyroglutamic acid (c: area on y-axis, d response factor
area pyroglutamic acid/area 13C15N glutamic acid)

4. In certain cases, addition of minor amounts of base NaOH,

KOH (for Riboflavine, Uric Acid, Xanthine, Threonine, Aspar-
tate) or HCl acid (for 2-Hydroxyisobutyric, Inositol,
3-methylhistidine, Cystine, Xylose, Tyrosine) and heating (for
Thymine, Uracil) and/or sonication (for Cysteine) is needed to
assist dissolution.
5. Due to high concentration of buffer, addition of MeCN turns
the solvent cloudy. Ultrasonic at room temperature assists in
dissolvation.
6. Column pressure should not exceed 5300 psi, otherwise sys-
tem may be over pressure when eluent B reach 85%.
7. The choice of extraction solvent is based on average solvent
content in MeCN, MeOH, and H2O of standard mixtures.
8. Since there is no available analyte free matrix, standard addition
approach is performed in order to avoid limitation such as
matrix effect, in cases where absolute concentration is required.
9. External calibration curve can be applied when only compari-
son between samples of the same or different batches and not
absolute quantitation is required.
202 Christina Virgiliou et al.

10. For the external calibration standards, standard 9 should be

diluted 1:1 with 95:5 MeCN:H20, and then proceed to serial
dilutions.
11. Optimum column pressure ranges during the gradient: mini-
mum 5000–5300 psi, maximum 11,000–12,500 psi. In case of
higher pressure range, clean the column as recommended.
12. According to our groups tests that have been performed for
system equilibration, four injections are the minimum for
retention time and signal stabilization for serum.
13. According to the tests that have been performed for system
performance, 100 serum samples are the maximum in order to
avoid variations due to column and cone contaminations from
the matrix.
14. Always use mean calibration curve (before and after batch
calibration curve) for quantification. In case when TargetLynx
is used, just select standard samples of both curves together
with real samples, QCs, and standard mixtures, and perform
integration and quantitation by a single process.
15. Prepare synthetic urine by dissolving 14.1 g of NaCl, 2.8 g of
KCl, 17.3 g of urea, 1.9 mL of aqueous ammonia (25%), 0.60 g
of CaCI2, and 0.43 g of MgSO4 in 1 L of 0.02 M HCl.

References

1. Nicholson JK, Lindon JC, Holmes E (1999) 6. Cajka T, Fiehn O (2016) Toward merging
“Metabonomics”: understanding the meta- untargeted and targeted methods in mass
bolic responses of living systems to pathophys- spectrometry-based metabolomics and lipido-
iological stimuli via multivariate statistical mics. Anal Chem 88:524–545. https://doi.
analysis of biological NMR spectroscopic data. org/10.1021/acs.analchem.5b04491
Xenobiotica 29:1181–1189. https://doi.org/ 7. Zhou B, Xiao JF, Tuli L, Ressom HW (2012)
10.1080/004982599238047 LC-MS-based metabolomics. Mol BioSyst 8:
2. Nicholson JK, Connelly J, Lindon JC, Holmes 4 7 0 – 4 8 1 . h t t p s : // d o i . o r g / 1 0 . 1 0 3 9 /
E (2002) Metabonomics: a platform for study- c1mb05350g
ing drug toxicity and gene function. Nat Rev 8. Serafimov K, L€ammerhofer M (2022) Meta-
Drug Discov 1:153–161. https://doi.org/10. bolic profiling workflow for cell extracts by
1038/nrd728 targeted hydrophilic interaction liquid
3. Nicholson JK, Lindon JC (2008) Systems biol- chromatography-tandem mass spectrometry. J
ogy: metabonomics. Nature 455:1054–1056. Chromatogr A 1684:463556. https://doi.
https://doi.org/10.1038/4551054a org/10.1016/j.chroma.2022.463556
4. Theodoridis G, Gika HG, Wilson ID (2011) 9. Verpoorte R, Choi YH, Kim HK (2010) Meta-
Mass spectrometry-based holistic analytical bolomics: will it stay? Phytochem Anal 21:2–3.
approaches for metabolite profiling in systems https://doi.org/10.1002/pca.1191
biology studies. Mass Spectrom Rev 30:884– 10. Schrimpe-Rutledge AC, Codreanu SG, Sher-
906. https://doi.org/10.1002/mas.20306 rod SD, McLean JA (2016) Untargeted meta-
5. Gika HG, Theodoridis GA, Plumb RS, Wilson bolomics strategies – challenges and emerging
ID (2014) Current practice of liquid directions. J Am Soc Mass Spectrom 27:1897–
chromatography–mass spectrometry in meta- 1905. https://doi.org/10.1007/s13361-
bolomics and metabonomics. J Pharm Biomed 016-1469-y
Anal 87:12–25. https://doi.org/10.1016/j. 11. Gertsman I, Barshop BA (2018) Promises and
jpba.2013.06.032 pitfalls of untargeted metabolomics. J Inherit
 HILIC-MS/MS Multi-targeted Method for Metabolomics Applications 203

Metab Dis 41:355–366. https://doi.org/10. Anal Test 4:140–162. https://doi.org/10.

1007/s10545-017-0130-7 1007/s41664-020-00126-z
12. Roberts LD, Souza AL, Gerszten RE, Clish CB 21. Lavarello C, Barco S, Bartolucci M, Panfoli I,
(2012) Targeted metabolomics. Curr Protoc Magi E, Tripodi G, Petretto A, Cangemi G
Mol Biol Chapter:Unit30.2. https://doi.org/ (2021) Development of an accurate mass
10.1002/0471142727.mb3002s98 retention time database for untargeted meta-
13. Griffiths WJ, Koal T, Wang Y, Kohl M, Enot bolomic analysis and its application to plasma
DP, Deigner H-P (2010) Targeted metabolo- and urine pediatric samples. Molecules 26:
mics for biomarker discovery. Angew Chem Int 4 2 5 6 . h t t p s : // d o i . o r g / 1 0 . 3 3 9 0 /
Ed Engl 49:5426–5445. https://doi.org/10. molecules26144256
1002/anie.200905579 22. Iturrospe E, Da Silva KM, Talavera Andújar B,
14. Michopoulos F, Whalley N, Theodoridis G, Cuykx M, Boeckmans J, Vanhaecke T,
Wilson ID, Dunkley TPJ, Critchlow SE Covaci A, van Nuijs ALN (2021) An explor-
(2014) Targeted profiling of polar intracellular atory approach for an oriented development of
metabolites using ion-pair-high performance an untargeted hydrophilic interaction liquid
liquid chromatography and -ultra high perfor- chromatography-mass spectrometry platform
mance liquid chromatography coupled to tan- for polar metabolites in biological matrices. J
dem mass spectrometry: applications to serum, Chromatogr A 1637:461807. https://doi.
urine and tissue extracts. J Chromatogr A org/10.1016/j.chroma.2020.461807
1349:60–68. https://doi.org/10.1016/j. 23. Nakatani K, Izumi Y, Takahashi M, Bamba T
chroma.2014.05.019 (2022) Unified-hydrophilic-interaction/
15. Roggensack T, Merz B, Dick N, Bub A, Krüger anion-exchange liquid chromatography mass
R (2020) Targeted ultra-performance liquid spectrometry (Unified-HILIC/AEX/MS): a
chromatography/tandem mass spectrometric single-run method for comprehensive and
quantification of methylated amines and simultaneous analysis of polar metabolome.
selected amino acids in biofluids. Rapid Com- Anal Chem 94:16877–16886. https://doi.
mun Mass Spectrom 34:e8646. https://doi. org/10.1021/acs.analchem.2c03986
org/10.1002/rcm.8646 24. Buescher JM, Moco S, Sauer U, Zamboni N
16. Virgiliou C, Kanelis D, Pina A, Gika H, (2010) Ultrahigh performance liquid
Tananaki C, Zotou A, Theodoridis G (2020) chromatography-tandem mass spectrometry
A targeted approach for studying the effect of method for fast and robust quantification of
sugar bee feeding on the metabolic profile of anionic and aromatic metabolites. Anal Chem
Royal Jelly. J Chromatogr A 1616:460783. 82:4403–4412. https://doi.org/10.1021/
https://doi.org/10.1016/j.chroma.2019. ac100101d
460783 25. Coulier L, Bas R, Jespersen S, Verheij E, van
17. Schiesel S, L€ammerhofer M, Lindner W (2010) der Werf MJ, Hankemeier T (2006) Simulta-
Multitarget quantitative metabolic profiling of neous quantitative analysis of metabolites using
hydrophilic metabolites in fermentation broths ion-pair liquid chromatography-electrospray
of β-lactam antibiotics production by HILIC- ionization mass spectrometry. Anal Chem 78:
ESI-MS/MS. Anal Bioanal Chem 396:1655– 6573–6582. https://doi.org/10.1021/
1679. https://doi.org/10.1007/s00216- ac0607616
009-3432-2 26. Mateos-Vivas M, Rodrı́guez-Gonzalo E, Dom-
18. Gika HG, Theodoridis GA, Vrhovsek U, Mat- ı́nguez-Álvarez J, Garcı́a-Gómez D, Carabias-
tivi F (2012) Quantitative profiling of polar Martı́nez R (2016) Determination of nucleo-
primary metabolites using hydrophilic interac- sides and nucleotides in baby foods by hydro-
tion ultrahigh performance liquid philic interaction chromatography coupled to
chromatography-tandem mass spectrometry. J tandem mass spectrometry in the presence of
Chromatogr A 1259:121–127. https://doi. hydrophilic ion-pairing reagents. Food Chem
org/10.1016/j.chroma.2012.02.010 211:827–835. https://doi.org/10.1016/j.
19. Qing G, Yan J, He X, Li X, Liang X (2020) foodchem.2016.05.091
Recent advances in hydrophilic interaction liq- 27. Lioupi A, Virgiliou C, Walter TH, Smith KM,
uid interaction chromatography materials for Rainville P, Wilson ID, Theodoridis G, Gika
glycopeptide enrichment and glycan separa- HG (2022) Application of a hybrid zwitter-
tion. TrAC Trends Anal Chem 124:115570. ionic hydrophilic interaction liquid chromatog-
https://doi.org/10.1016/j.trac.2019.06.020 raphy column in metabolic profiling studies. J
20. Hu Q, Tang H, Wang Y (2020) Challenges in Chromatogr A 1672:463013. https://doi.
analysis of hydrophilic metabolites using chro- org/10.1016/j.chroma.2022.463013
matography coupled with mass spectrometry. J
204 Christina Virgiliou et al.

28. Lioupi A, Marinaki M, Virgiliou C, Begou O, acid analysis by liquid chromatography-tandem

Gika H, Wilson I, Theodoridis G (2023) Prob- mass spectrometry: a tutorial review. TrAC
ing the polar metabolome by UHPLC-MS. Trends Anal Chem 131:116018. https://doi.
TrAC Trends Anal Chem 161:117014. org/10.1016/j.trac.2020.116018
https://do i.org/10 .10 16/j.trac.2023. 34. Bruno C, Veyrat-Durebex C, Lumbu Lukun-
117014 tonda CH, Andres CR, Moreau C, Bendavid C,
29. Hosseinkhani F, Huang L, Dubbelman A-C, Homedan C, Labarthe F, Tardieu M, Bigot A,
Guled F, Harms AC, Hankemeier T (2022) Maillot F, Benz-de Bretagne I, Blasco H
Systematic evaluation of HILIC stationary (2021) Validation of plasma amino acid profile
phases for global metabolomics of human using UHPLC-mass spectrometer (QDa) as a
plasma. Metabolites 12:165. https://doi.org/ screening method in a metabolic disorders ref-
10.3390/metabo12020165 erence centre: Performance and accreditation
30. Sampsonidis I, Witting M, Koch W, concerns. Clin Biochem 92:34–45. https://
Virgiliou C, Gika HG, Schmitt-Kopplin P, doi.org/10.1016/j.clinbiochem.2021.03.004
Theodoridis GA (2015) Computational analy- 35. Maciel LS, Marengo A, Rubiolo P, Leito I,
sis and ratiometric comparison approaches Herodes K (2021) Derivatization-targeted
aimed to assist column selection in hydrophilic analysis of amino compounds in plant extracts
interaction liquid chromatography–tandem in neutral loss acquisition mode by liquid
mass spectrometry targeted metabolomics. J chromatography-tandem mass spectrometry. J
Chromatogr A 1406:145–155. https://doi. Chromatogr A 1656:462555. https://doi.
org/10.1016/j.chroma.2015.06.008 org/10.1016/j.chroma.2021.462555
31. Virgiliou C, Sampsonidis I, Gika HG, 36. Chen X, Wu H, Cao Y, Yao X, Zhao L, Wang T,
Raikos N, Theodoridis GA (2015) Develop- Yang Y, Lv D, Chai Y, Cao Y, Zhu Z (2013)
ment and validation of a HILIC-MS/MS mul- Ion-pairing chromatography on a porous gra-
titargeted method for metabolomics phitic carbon column coupled with time-of-
applications. Electrophoresis 36:2215–2225. flight mass spectrometry for targeted and
https://doi.org/10.1002/elps.201500208 untargeted profiling of amino acid biomarkers
32. Virgiliou C, Theodoridis G, Wilson ID, Gika involved in Candida albicans biofilm forma-
HG (2021) Quantification of endogenous tion. Mol BioSyst 10:74–85. https://doi.org/
amino acids and amino acid derivatives in 10.1039/C3MB70240E
urine by hydrophilic interaction liquid chroma- 37. Vilches AP, Norström SH, Bylund D (2017)
tography tandem mass spectrometry. J Chro- Direct analysis of free amino acids by mixed-
matogr A 1642:462005. https://doi.org/10. mode chromatography with tandem mass spec-
1016/j.chroma.2021.462005 trometry. J Sep Sci 40:1482–1492. https://
33. Violi JP, Bishop DP, Padula MP, Steele JR, doi.org/10.1002/jssc.201601097
Rodgers KJ (2020) Considerations for amino
 Chapter 11

A Protocol for GC-MS Profiling of Chiral Secondary Amino

Acids
Stanislav Opekar, Helena Zahradnı́čková, Petr Vodrážka,
Lucie Řimnáčová, Martin Moos, and Petr Šimek

Abstract
A simple analytical workflow is described for gas chromatographic-mass spectrometric (GC-MS)-based
chiral profiling of secondary amino acids (AAs) in biological matrices. The sample preparation is carried out
directly in aqueous biological sample extracts and involves in situ heptafluorobutyl chloroformate (HFBCF)
derivatization—liquid-liquid microextraction of nonpolar products into hexane phase followed by
subsequent formation of the corresponding methylamides from the HFB esters by direct treatment with
methylamine reagent solution. The (O, N) HFB-butoxycarbonyl–methylamide AA products (HFBOC-
MA) are separated on a Chirasil-L-Val capillary column and quantitatively measured by GC-MS operated in
selected ion monitoring (SIM) mode. The protocol includes 12 simple pipetting steps and covers the
quantitative analysis of 8 L, D pairs of secondary amino acids, including proline, isomeric 3-, 4-hydro-
xyprolines, pipecolic acid, nipecotic acid, azetidine-2-carboxylic acid, and cis- and trans-5-hydroxy-L-
pipecolic acid using 13C5 –L-proline as an internal standard. The individual analytical steps are commented
on and explained, with emphasis on the chiral GC-MS analysis of secondary amino acids in human urine,
serum, and peptide hydrolysate samples.

Key words Secondary amino acids, Amino acids, GC-MS, Alkyl chloroformate derivatization, Chiral
analysis, Urine, Serum, Peptide hydrolysates, Quantitative analysis

1 Introduction

Secondary amino acids are an important subcategory of AAs and

represent essential circulating, highly polar metabolites and build-
ing blocks of protein formation in all living organisms. Secondary
AAs are characterized by four-, five-, and six-membered cyclic
structure, conformational rigidity, and occurrence in various iso-
meric and optically active chiral forms. Their detailed identification,
comprehensive separation, and, most importantly, quantitative chi-
ral analysis are therefore a challenging task that requires an
advanced method for their profiling.

Olga Deda et al. (eds.), Metabolic Profiling: Methods and Protocols, Methods in Molecular Biology, vol. 2891,
https://doi.org/10.1007/978-1-0716-4334-1_11,
© The Author(s), under exclusive license to Springer Science+Business Media, LLC, part of Springer Nature 2025

205
206 Stanislav Opekar et al.

Enantiomeric analysis of a comprehensive set of secondary AAs

requires either an efficient chiral separation environment [1] or
separation of their actual diastereomers obtained by AA treatment
with an appropriate chiral derivatization reagent [2]. Comprehen-
sive profiling by the direct or indirect method is difficult to access
and tedious and usually involves high-cost analytical materials. In
addition, secondary AAs are highly polar and usually occur in a
complex biological material, so interfering components must be
minimized by an appropriate extraction step. In practice, derivati-
zation of the highly polar AA functional groups is a valuable
approach that improves the analytical properties of the target AA
analytes, including their separation. For chiral analysis, the GC-MS
technique is a viable alternative because it is inexpensive, and con-
ventional chiral capillary columns such as those with the Chirasil Val
phase are well established [3, 4]. Among the derivatization meth-
ods, two-step acylation-esterification [5] and the use of alkyl chlor-
oformates (RCFs) are popular approaches for chiral AA analysis
[6]. We have continuously investigated the latter approach as it
involves favorable in situ derivatization of protic groups within
seconds and can be easily combined with simultaneous liquid-liquid
microextraction of the resulting nonpolar derivatives into an
organic phase that can be injected into a GC-MS system [7].
Esterification of carboxyl by RCFs in aqueous media is a com-
plex reaction, where under pyridine catalysis the RCFs decompose
by involvement of carbon dioxide [8]. The released CO2 is partly
dissolved in the whole sample medium; it enhances the effective
surface area between the immiscible organic and aqueous phase and
renders thus a powerful dispersive phase enabling to reach the final
equilibrium in less than 5 s [9, 10].
Our initial attention was focused on the chiral analysis of amino
acids based on their in situ derivatization with fluoroalkyl chloro-
formates (FCF) coupled with simultaneous liquid–liquid microex-
traction (LLME) of the arising low polar derivatives into
chloroform or isooctane and final L-Chirasil Val column separation
[11]. Using the HFBCF and LLME into isooctane, extraordinarily
clean extracts were obtained enabling efficient GC–MS separation
of more than 35 amino acid enantiomeric pairs as their (N,O,S)-
heptafluorobutoxycarbonyl O-heptafluorobutyl esters on a Chirasil
Val in human serum, except the D,L enantiomers of arginine and
cysteine (not eluted) and proline (the derivatives not separated)
[6]. For the efficient separation of the AA enantiomers by GC,
the Chirasil-Val is a versatile and most commonly used stationary
phase. It is a statistical copolymer of dimethylsiloxane and 2-(car-
boxypropyl)-siloxane, typically coupled with L-valine L- or D-valine-
t-butylamide [12]. Efficient chiral recognition of enantiomeric AA
solutes by Chirasil-Val requires effective hydrogen bonding
between the stationary phase and the analyte. The absence of the
hydrogen in the derivatized secondary amino group, therefore,
 A Protocol for GC-MS Profiling of Chiral Secondary Amino Acids 207

Fig. 1 Derivatization reaction scheme of 4-trans-L-hydroxyproline (4 t-L-Hyp) using 2,2,3,3,4,4,4-

heptafluorobutyl chloroformate followed by methylamidation

makes their enantiomeric separation impossible. However, we

found that heptafluorobutyl esters are sensitive to the basic envi-
ronment and upon action of sterically unhindered primary amines
can be easily converted to the corresponding amides. The amida-
tion of the trifluoroethyl ester group with propyl-, isopropyl-, and
isobutylamine for chiral AA analysis under harsh conditions was
earlier studied by Abe and coworkers and improved the chiral
separation of primary D,L AAs [13] and also DL-Pro [14].
Fluoroalkyl chloroformates (FCFs) exhibit some advanced
properties over the traditionally used alkyl chloroformates
(RCFs), providing highly volatile and less polar derivatives extract-
able in nonpolar hydrocarbon solvents [10]. Recently, we applied
the well-established sample preparation strategy based on fluor-
oalkyl chloroformates for a comprehensive GC-MS analysis of sec-
ondary AAs after their subsequent HFBCF derivatization/LLME
and methylamine treatment [1]. The reaction scheme for the reac-
tion of HFBCF with proline followed by methylamidation of the
ester group is shown in Fig. 1.
Here we describe a protocol for the HFBCF-mediated sample
preparation [10, 11, 15] with the amidation step for the chiral GC–
MS analysis of the secondary AAs. In the first step, a HFBCF-
mediated LLME protocol was performed as described previously
[10]. In the second step, the arising (N,O)-heptafluorobutoxycar-
bonyl O-heptafluorobutyl (HFBOC-HFB) esters are converted
with methylamine into the corresponding N-methylamides. The
reaction mixture is then evaporated by a mild stream of nitrogen
and redissolved in chloroform for GC–MS analysis on the L-Chir-
asil-Val phase. The developed method enables simultaneous chiral
analysis of the four-, five- and six-membered secondary AAs in
various complex biological matrices. The method was examined in
the chiral AAs analysis in blood serum, urine, collagen hydrolysates,
and biologically active peptide hydrolysates. The workflow of the
GC-MS-based secondary amino acid analysis is depicted in Fig. 2.
The GC–MS method enables excellent separation of eight second-
ary AA enantiomeric pairs and one pair of HyPip stereoisomers of
secondary amino acids as HFBOC-MA in a single run, Fig. 3.
The elaborated sample preparation protocol is simple and
involves gradual pipetting of uniform small volumes of a sample
208 Stanislav Opekar et al.

Fig. 2 The analytical workflow (from Ref. [1], Springer Nature with permission)

and necessary liquid media in 12 steps: (1) a sample; (2) an internal

standard solution; (3) a pH adjustment; (4) an organic reaction
medium containing the HFBCF reagent; (5–6) a repeated addition
of a catalytic medium with pyridine; (7) an organic extraction
medium; (8) an acidification medium; (9) an upper extraction
phase transfer into a GC autosampler vial; (10) addition of methyl-
amine, keeping at 40
C, 15 min followed by nitrogen stream
drying; (11) an organic medium of chloroform; and finally
(12) the sample extract injection into a GC-MS spectrometer.
The protocol was initially developed for chiral GC-MS analysis
of secondary amino acids in urine, serum, or peptide hydrolysate
[1]. The urine and serum levels of the detected amino acids in
healthy control subjects (n ¼ 5) determined by this method are
shown in Table 1. The absolute concentration of Pro in the pooled
serum sample was estimated to be 195.7  8.6 μmolL1. The
relative ratio of D-Pro was 0.10%. In addition, total Pip was also
quantified and reached 5.23  0.30 μmolL1. D-Pip was not
detected in the serum; therefore, the value of total Pip belongs
exclusively to L-Pip. Interestingly, 4 t-L-Hyp and L-Nip were found
in blood serum, but they were not quantified. Quantification of
D, L-Pro, and D, L-Pip was based on analysis of HFBCF mediated
derivatives (SPS1), and the enantiomeric ratio was calculated based
on analysis of HFBOC-MA derivatives (SPS1 and SPS2). Total Pro
in the pooled urine sample was found at a concentration of
28.6  0.9 μmolL1, and the enantiomeric ratio of D-Pro reached
 A Protocol for GC-MS Profiling of Chiral Secondary Amino Acids 209

Fig. 3 GC-SIM-MS separation of D and L isomers obtained for the target secondary amino acids; (a) if 4c-Hyp
is not present, (b) if 4 t-Hyp is not present, (c) resolution between 5 t-L-HyPip and 5c-L-HyPip (from Ref. [1],
Springer Nature with permission)

Table 1
The content of secondary amino acids in human serum and urine (n = 5)

Secondary acid Human serum [μmolL1] Human urine [μmolL1]

Pro 195.7  8.6 28.6  0.9
D-Pro a
0.20  0.01 0.51  0.02
L-Proa 195.5  8.6 28.1  0.9
Pip 5.23  0.30 3.34  0.10
D-Pip a
0 2.76  0.08
L-Pip a
5.23  0.30 0.58  0.02
a
The value was calculated from the enantiomeric ratio
210 Stanislav Opekar et al.

1.8%, corresponding to a value of 0.51  0.02 μmolL1 (Table 1).

The majority of Pip was observed in urine as D-Pip (82.5%), in
contrast to the exclusive presence of L-Pip in blood serum, consis-
tent with previous findings [16]. Most likely, this preferential excre-
tion of D-Pip is caused by the activity of intestinal bacteria
[17, 18]. Total Pip was also quantified and a value of
3.34  0.10 μmolL1 was observed in urine. Interestingly, 4 t-L-
Hyp and L-Nip were found in trace amounts in urine, but they were
not quantified.
Nevertheless, the procedure can easily be adapted to work-up
of any aqueous biological material with low protein content. If the
content of biopolymers is high (> 2 mg/mL), then a precipitation
step is preferred prior to the described workflow (see Note 1).

2 Materials

Use of an appropriate laboratory wear, glasses, and other personal

protective equipment is recommended; follow standard laboratory
precautions and local guidelines, especially waste disposal regula-
tions. Use a functional fume hood for the sample work-up.

2.1 Samples Samples containing no or little protein and cell residues (urine, cell
culture media) or cell and tissue extracts can be analyzed by this
method. Moreover, this method can be applied to precipitated
serum/plasma or peptide hydrolysates.

2.2 Chemicals, 1. The analytical grade chemicals should be used. All solutions,
Solutions, except that containing the HFBCF reagent, should be prepared
Reaction Media in distilled deionized water (DI water, < 1.5 μS.cm1, 25
C)
and stored them at 4
C (unless otherwise indicated).
2. 100 mM NaHCO3 (99,998% purity, Alfa Aesar) solution: Dis-
solve 840 mg in 100 mL of DI water.
3. 1 M NaOH (99.99% purity, Alfa Aesar) solution: Dissolve 4 g
in 100 μL of DI water.
4. Heptafluorobutanol (HFBOH) (99% purity, Chromservis, Pra-
gue, Czech Republic).
5. Heptafluorobutyl chloroformate (HFBCF, 98%, Chromser-
vis), (see Note 2).
6. Solvents: Isooctane (2,2,4-trimethylpentane, 99.5%), pyridine
(p.a., 99.0%), isopropanol (2-propanol, 99.5% purity) (all
Sigma-Aldrich).
7. The organic reaction medium: prepare isooctane, HFBCF, and
HFBOH in a volume ratio 15:4:1 (v/v/v) in a Teflon-capped,
well-tightened 4-mL glass vial. Store in a refrigerator, where
the mixture remains stable for several months.
 A Protocol for GC-MS Profiling of Chiral Secondary Amino Acids 211

8. The catalytic medium: mix 1 M NaOH with pyridine in a

volume ratio of 24:1 (v/v).
9. For peptide hydrolysis prepare: 6 M HCl by dilution of 37%
HCl (2 μL) with 2 mL of DI water.

2.3 Analytical 1. Internal standard solution (IS): 13C5-L-Pro (Cambridge Iso-

Standards, Stock tope Laboratories, MW ¼ 120.1 g/moL). Prepare a stock
Solutions solution in 100 mM HCl with a final concentration of
1 mmol/L, i.e., 5 nmoL in 5 μL of the applied internal stan-
dard solution.
2. Secondary amino acid (AA) standard solutions in 100 mM HCl
containing: D-Aze, L-Aze, D- Pro, L- Pro, L-Pip, D-Pip, L-Nip,
D-Nip, 3c-D-Hyp, 3c-L-Hyp, 4c-L-Hyp, 4c-D-Hyp, 3 t-L-Hyp,
3 t-D-Hyp; 4 t-L-Hyp, 4 t-D-Hyp, 5 t-L-HyPip; 5c-L-HyPip at a
concentration of 10 mmol/L. For the complete metabolite list,
refer to Table 2.

2.4 GC-MS 1. Agilent 7890A GC system equipped with G4513A autosam-

Instrumentation pler (Agilent), multimode injector (MMI), equipped with a
10 μL syringe (CTC Analytics, P/N PAL3-SYH-207807).
2. Sky® 4 mM I.D. cyclo double taper inlet liner (Restek, P/N
23310).
3. CP-Chirasil-L-Val column (25 m  0.25 mM, 0.12 μm, Agilent
Technologies, USA, P/N CP7495).
4. Single quadrupole mass triple-axis detector (5975 MSD Inert
XL, Agilent) equipped with an inert EI ion source.
5. Autosampler 2 mL vials (12  32 mM) with 9 mM PP open
hole caps (Labicom, P/N 5310F-09) and 0.040 “PTFE/sili-
cone/PTFE Septa (Labicom, P/N 604060-09).
6. Inert conical glass insert, 200 μL volume, (Chromacol, P/N
02-MTV).

2.5 Additional 1. The sample preparation glass culture tubes 6  50 mM, mate-
Equipment rial: sodium-potassium silicate.
2. glass (Merci, P/N Z1632000605010), or borosilicate glass
(Kimble-Kontes, P/N 73500-650).
3. Common screw cap Teflon-lined 2 and 4 mL amber vials for
the reagent solutions.
4. An adjustable 50 and 100 μL Transferpettor pipette with a glass
capillary (Brand, P/N 701868 and 701873) for manipulation
with the reagents and their mixtures in isooctane. The pipette
tips with 25 mM capillary (gel-loading type, VWR Int.) for
aspirating the upper organic phase in the 6  50 mM vial.
Table 2
212

The list of secondary AAs determined by the described GC-MS protocol

HFBOC-MA
HFBOC-HFB derivatives
derivatives (SPS1) (SPS1 + SPS2) Metabolite Database Coding

RT Diagnostic RT Diagnostic
No. Traditional Name IUPAC Name Abbreviation [min] (m/z) [min] (m/z) CAS KEGG HMDB
1 L-Azetidine-2- (2S)-Azetidine-2-carboxylic L-Aze 5.65 282 5.88 282 2133-34-8 C08267 HMDB29615
Stanislav Opekar et al.

carboxylic acid acid

2 D-Azetidine-2- (2R)-Azetidine-2- D-Aze 5.65 282 5.52 282 7729-30-8 – –
carboxylic acid carboxylic acid
3 L-Proline (2S)-Pyrrolidine-2- L-pro 7.2 296 7.51 296 147–85-3 C00148 HMDB0000162
carboxylic acid
4 D- Proline (2R)-Pyrrolidine-2- D-pro 7.2 296 6.71 296 344–25-2 C00763 HMDB0003411
carboxylic acid
5 L-Pipecolic acid (2S)-Piperidine-2- L-pip 8.17 310 7.83 310 3105-95-1 C00408 HMDB0000716
carboxylic acid
6 D-Pipecolic acid (2R)-Piperidine-2- D-pip 8.17 310 7.58 310 1723-00-8 – HMDB0005960
carboxylic acid,
D-Homoproline
7 L-Nipecotic acid (3S)-Piperidine-3- L-nip 11.74 310 11.74 309 59,045–82-8 – –
carboxylic acid
8 D-Nipecotic acid (3R)-Piperidine-3- D-nip 11.74 310 12.29 309 25,137–00-2 – –
carboxylic acid
9 trans-3-Hydroxy- (2S,3S)-3- 3 t-L-Hyp 19.47 521 19.96 312 876,067 C05147 –
L-proline Hydroxypyrrolidine-2-
carboxylic acid
 10 trans-3-Hydroxy- (2R,3R)-3- 3 t-D-Hyp 19.47 521 19.6 312 – – –
D-proline Hydroxypyrrolidine-2-
carboxylic acid
11 cis-3-Hydroxy-L- (2S,3R)-3- 3c-L-Hyp 21.98 538 26.74 312 567–35-1 C19706 –
proline Hydroxypyrrolidine-2-
carboxylic acid
12 cis-3-Hydroxy-D- (2R,3S)-3- 3c-D-Hyp 21.84 538 26.35 312 – – –
proline Hydroxypyrrolidine-2-
carboxylic acid
13 trans-4-Hydroxy- (2S,4R)-4- 4 t-L-Hyp 22.38 294 22.91 294 51–35-4 C01157 HMDB0000725
L-proline Hydroxypyrrolidine-2-
carboxylic acid
14 trans-4-Hydroxy- (2R,4S)-4- 4 t-D-Hyp 22.24 294 22.01 294 – – –
D-proline Hydroxypyrrolidine-2-
carboxylic acid
15 cis-4-Hydroxy-L- (2S,4S)-4- 4c-L-Hyp 23.68 294 23.83 294 618–27-9 C01015 HMDB0240251
proline Hydroxypyrrolidine-2-
carboxylic acid
16 cis-4-Hydroxy-D- (2R,4R)-4- 4c-D-Hyp 23.61 294 22.91 294 2584-71-6 – –
proline Hydroxypyrrolidine-2-
carboxylic acid
17 trans-5-Hydroxy- (2S,5R)-5- 5 t-L-HyPip 23.34 308 22.3 308 824,943–40- – HMDB0029426
L-pipecolic acid Hydroxypiperidine-2- 0
carboxylic acid
18 cis-5-Hydroxy-L- (2S,5S)-5- 5c-L-HyPip 24.89 308 24.66 308 63,088–78-8 – –
pipecolic acid Hydroxypiperidine-2-
carboxylic acid
SPS1 Sample preparation step 1 (preparation of HFBOC-HFB derivatives)
A Protocol for GC-MS Profiling of Chiral Secondary Amino Acids

SPS2 Sample preparation step 2 (preparation of HFBOC-MA derivatives)

The traditional name, the IUPAC name, abbreviation, retention data, diagnostic SIM ions, and metabolite database coding are listed
213
214 Stanislav Opekar et al.

5. Alternatively, a common pipette (10–100 μL) such as a Biohit

Prolin® mechanical pipette (Sartorius, P/N 720050) equipped
with an Optifit tip 200 (P/N 4059.9002) can be used.
6. A common vortex for sample mixing and a minicentrifuge such
as MySpin 6 (Thermo Scientific, 2000  g) for a complemen-
tary separation of immiscible layers in sample vials.

3 Methods

3.1 Sampling and 1. A serious attention should be paid to the sample collection,
Storage transport, and storage, because any omission may result in false
results. For urine, collection of the morning second-void sam-
ples is the most common practice and the easiest sampling
method.
2. For urine analysis, store freshly collected samples at 4
C within
2 h. For longer than 48 h storage, freeze the samples and keep
at 20
C (see Note 3).

3.2 Preparation of 1. Prepare a calibration mixture following the procedure

Calibration Solutions described in [1] by adding the same volume of each stock
solution standard, Table 2 (only L-forms for quantification),
in a vial. The concentration of calibration mixture solution is
1 mmol/L (10 AAs mixed).
2. Prepare the middle and lower calibration mixtures; i.e.,
10 times diluted (0.1 mmol/L) and 100 times diluted
(0.01 mmol/L) to the original 1 mmol/L level. (see Note 9).
3. Prepare an appropriate pooled sample by mixing an equal small
volume of all samples included in the study for the verification
of the average EI MS response for each target metabolite and
for the quality control (QC) analysis.

3.3 Sample 1. Transfer 50 μL of aqueous biological sample into a 6  50 mM

Preparation Procedure culture tube. For serum/plasma, transfer 50 μL of plasma/
for Serum/Plasma serum suppernatant after precipitation (see Note 1) or 50 μL
of peptide hydrolysate (see Note 8). Urine can be used directly.
2. Spike the sample with the 5 μL of the internal standard
solution.
3. Adjust pH to ca 9 with 25 μL 100 mM NaHCO3 solution and
vortex gently.
4. Add 50 μL of the organic reaction medium (isooctane, HFBCF
and HFBOH, 15:4:1, v/v/v) (see Note 4).
5. Add 25 μL of the catalytic medium (1 M NaOH-pyridine, 24:
1, v/v) and vortex the content for ca 3 s leaving the organic
phase milky.
 A Protocol for GC-MS Profiling of Chiral Secondary Amino Acids 215

6. Add a second portion (25 μL) of the catalytic medium and

shake the content for 5 s until the milky phase becomes
clarified.
7. Add 50 μL of the isooctane extraction medium, mix for about
1–2 s.
8. Add of 25 μL of 1 M aqueous HCl and vortex briefly; if the
phases are not well separated, a minicentrifuge may be a conve-
nient option (see Note 5).
9. Aspirate 15 μL of the upper organic phase into a vial insert
(150–200 μL volume).
10. Inject a sample extract aliquot (0.5 μL) by using a pulsed
splitless injection into a GC injector; start the GC-MS acquisi-
tion (see Note 6).
11. Aspirate 50 μL of the upper organic phase into another vial
insert (150–200 μL volume). Add 50 μL 8 M methylamine
solution in ethanol, mix for about 2–3 s.
12. Keep the mixture at 40
C for 15 min in an oven, and after the
completion of the amidation reaction, remove all volatile com-
ponents by a gentle stream of nitrogen.
13. Add 15 μL of chloroform and mix for about 10 s.
14. Inject a sample solution aliquot (0.5 μL) by using a pulsed
splitless injection into a GC injector; start the GC-MS acquisi-
tion (see Note 6).

3.4 GC-MS Analysis 1. The instrument GC–MS conditions are summarized in Table 3.
2. First, analyze the standard mixtures to check the separation
performance, retention times, the analyte peak shape, and
acquisition of the employed fragment ions in the obtained EI
spectra, Table 2.
3. Using a single quadrupole MS analyzer, selected-ion monitor-
ing (SIM) mode is commonly used for quantification. Use the
characteristic m/z ions listed in Table 2.
4. Prepare an appropriate analysis sequence of a sample series
consisting of repeated blank, standard, the QCs (see Subhead-
ing 3.2, step 3), calibration, and real samples. Measure regu-
larly the blanks, QC samples, and standards (at least every
10 sample runs). Analyze samples in a random order to avoid
a systematic error.
5. Change the GC injector liner approximately after 150 samples
depending on the sample matrix. Before use, condition each
new liner by running the following sequence: solvent blank,
standard mixtures, the pooled QC sample extract (twice), and
solvent blank (twice).
216 Stanislav Opekar et al.

Table 3
GC-EI-SIM-MS operating conditions

GC Injector Mode Pulsed splitless (elevated head pressure from 110 to 220 kPa)
Liner Sky® 4 mM I.D. cyclo double taper inlet liner (Restek, P/N 23310)
Temperature 200
C
Injection 0.5 μL
volume
Temperature 200
C
GC Oven_1a Initial 89
C
temperature
Ramp Hold for 1.5 min, 2
C/min to 150
C, hold for 6 min then back
20
C/min to 89
C
Run time 42 min
GC Oven_2 b
Initial 110
C
temperature
Ramp Hold for 1 min, 5
C/min to 180
C, 1
C/min to 200
C hold for
8 min then back 20
C/min to 110
C
Run time 47.5 min
GC column Capillary CP-Chirasil-L-Val (25 m  0.25 mM, 0.12 μm, Agilent
column Technologies, USA, P/N CP7495)
Carrier gas Helium
Flow rate 1.2 mL/min
Mode Constant flow
Outlet pressure Vacuum
GC-MS transfer Temperature 200
C
line
MS quadrupole Ion source 200
C
temperature
Full scan mode m/z 50–900 Da
SIM mode Metabolite SIM m/z ions, see Table 1
Software MSD ChemStation (version E.02, Agilent)
a
HFBOC-HFB analytes after SPS1
b
HFBOC-MA analytes after SPS 1 + 2

3.5 Data Analysis 1. Peak areas for corresponding ions are integrated. Their ratio is
calculated to test for potential interferences.
2. The peaks of particular metabolites are normalized by the peak
area of the internal standard 13C5 proline.
 A Protocol for GC-MS Profiling of Chiral Secondary Amino Acids 217

3. Use an appropriate vendor data processing software for data

calibration and metabolite quantification.
4. Check metabolite responses in the QC samples measured reg-
ularly throughout the whole sample set. If the analyte’s relative
response to the IS fluctuates with RSD > 30%, then even a
semiquantitative measurement of such metabolite is difficult,
and it should be excluded from the study (see Note 7).
5. Determine the secondary AAs levels in samples from non-chiral
quantification analysis of HFBOC-HFB derivatives, and recal-
culate the levels for individual enantiomers according to d/l
ratio. Export the analytical data matrix into a Microsoft Excel®
spreadsheet or other format suitable to explore for further
chemometric analysis.
6. Use an appropriate statistical software to recognize differences
among the studied metabolite sample sets. The calculation of
p-values by means of a t-test helps to determine significance of
the obtained results. The data set can be conveniently exam-
ined graphically by means of box plots that display patterns of
quantitative data and thus facilitate interpretation of observed
metabolite changes in the studied organism.

4 Notes

1. For serum/plasma, (lipo)proteins must be precipitated by a

mixture of methanol/acetonitrile 1:1 prior to the application
of this protocol. Serum/plasma (for instance, 1 mL) is treated
with 4 mL methanol/acetonitrile mixture (1:1, v/v), vortexed,
and centrifuged (4
C, 5500 rpm, 10 min). The supernatant
was collected in a new vial and stored at 20
C. The appropri-
ate aliquot of supernatant (usually 50 μL) is concentrated to
dryness in a glass culture tube and reconstituted in 50 μL
deionized water prior to sample derivatization.
2. The 2,2,3,3,4,4,4-heptafluorobutyl chloroformate (HFBCF)
reagent is a liquid with a boiling point 105–107
C and density
1.6 g/cm3 [15]. The reagent must be stored in tightly closed
Teflon-lined cap glass vials at 4
C and thus is stable for at least
24 months.
WARNING! Manipulation with HFBCF must be per-
formed in a well-ventilated area (fume-hood).
3. Urine like other important biological matrices is a rich metab-
olite mixture. Be careful and always take into account proper-
ties of each metabolite of interest in the studied matrix. Check
carefully the metabolite stability by sample measurement
within a convenient time period before making final conclu-
sions from the measured data.
218 Stanislav Opekar et al.

3. The used HFBC reagent volume (10 μL, 60 μmoL) is efficient

for work-up urine volumes below 50 μL.
4. The acidification step substantially further decreases the pyri-
dine catalyst content in the arising upper organic layer and thus
contamination of a used GC-MS system. Consequently, the
liner change follows typically after 120–150 samples, less fre-
quently than in earlier methods [19].
5. If the prepared organic sample extracts are not measured imme-
diately, they can be stored in tightly closed Teflon-lined auto-
sampler vials up to 2 weeks at 20
C.
6. Note that demands on metabolite profiling do always not
conform strict guidelines requested for instance by guidelines
in drug analysis, and data showing a higher uncertainty may be
useful in the study. Moreover, metabolite levels observed
between two studied models rarely change by more than
1 order of magnitude, and thus narrower calibration ranges
can be used throughout a metabolomic study with respect to
the amount estimated in the pooled QC sample [10].
7. Peptide hydrolysate: Dissolve appropriate amount of peptide
(1–500 μg) in 100 μL of 6 molL1 HCl and hydrolyze in sealed
vessel in the inert argon atmosphere at 110
C for 24 h. Evap-
orate the sample to dryness under a stream of nitrogen in a
6  50 mM culture tube and follow the sample preparation
protocol.
8. These solutions are used for preparation of 9 calibration levels
in triplicates by pipetting 5 μL of 100 times diluted solution
(0.05 nmoL); 10 μL of 100 times diluted solution (0.1 nmoL);
25 μL of 100 times diluted solution (0.25 nmoL); 5 μL of
10 times diluted solution (0.5 nmoL); 10 μL of 10 times
diluted solution (1 nmol); 25 μL of 10 times diluted solution
(2.5 nmoL); 5 μL of undiluted solution (5 nmoL); 10 μL of
undiluted solution (10 nmoL); 25 μL of undiluted solution
(25 nmoL).

Acknowledgments

This work was supported by the Grant Agency of the Czech

Republic, project No. 23-06600S, and the E.U. Fund Interreg,
project No. BYCZ118 is greatly appreciated.

References
1. Opekar S, Zahradnickova H, Vodrazka P, chloroformate & methylamine derivatization.
Rimnacova L, Simek P, Moos M (2021) A Amino Acids 53(3):347–358
chiral GC-MS method for analysis of secondary 2. Langrock T, Garcia-Villar N, Hoffmann R
amino acids after heptafluorobutyl (2007) Analysis of hydroxyproline isomers
 A Protocol for GC-MS Profiling of Chiral Secondary Amino Acids 219

and hydroxylysine by reversed-phase HPLC preparation and gas chromatography-mass

and mass spectrometry. J Chromatogr B Analyt spectrometry. J Chromatogr A 1443:211–232
Technol Biomed Life Sci 847(2):282–288 11. Zahradnickova H, Husek P, Simek P (2009)
3. Levkin PA, Levkina A, Czesla H, Schurig V GC separation of amino acid enantiomers via
(2007) Temperature-induced inversion of the derivatization with heptafluorobutyl chlorofor-
elution order of enantiomers in gas chromatog- mate and Chirasil-L-Val column. J Sep Sci
raphy: N-ethoxycarbonyl propylamides and 32(22):3919–3924
N-trifluoroacetyl ethyl esters of alpha-amino 12. Koppenhoefer B, Bayer E (1984) Chiral recog-
acids on Chirasil-Val-C-11 and Chirasil-Dex nition in the resolution of enantiomers by Glc.
stationary phases. Anal Chem 79(12): Chromatographia 19:123–130
4401–4409 13. Abe I, Kawazuma M, Fujimoto N, Nakahara T
4. Zahradnickova H, Husek P, Simek P, (1995) N-alkyloxycarbonyl isobutylamides as
Hartvich P, Marsalek B, Holoubek I (2007) readily prepared diamide derivatives of amino-
Determination of D- and L-amino acids pro- acids for separation of enantiomeric isomers by
duced by cyanobacteria using gas chromatog- chiral phase capillary gas-chromatography.
raphy on Chirasil-Val after derivatization with Chem Lett 4:329–330
pentafluoropropyl chloroformate. Anal Bioanal 14. Abe I, Nakahara T (1996) Enantiomer separa-
Chem 388(8):1815–1822 tion of amino acids as their N-alkyloxycarbonyl
5. Gehrke CW, Kuo KCT, Zumwalt RW (1987) alkylamide derivatives by chiral phase capillary
Amino acid analysis by gas chromatography, GC. Hrc J High Resolut Chromatogr 19(9):
vol II. CRC Press, Boca Raton 511–514
6. Simek P, Husek P, Zahradnickova H (2019) 15. Simek P, Husek P, Zahradnickova H (2012)
Heptafluorobutyl chloroformate-based sample Heptafluorobutyl chloroformate-based sample
preparation protocol for nonchiral and chiral preparation protocol for chiral and nonchiral
amino acid analysis by gas chromatorgraphy- amino acid analysis by gas
mass spectrometry. In: Alterman MA chromatography. In: Alterman MA, Hunziker
(ed) Amino acid analysis methods and proto- P (eds) Amino acid analysis: methods and pro-
cols, vol 2030, 2nd edn. Springer, Silver tocols, vol 828, pp 137–152
Spring, pp 237–251 16. Armstrong DW, Gasper M, Lee SH,
7. Husek P (1998) Chloroformates in gas chro- Zukowski J, Ercal N (1993) D-amino-acid
matography as general purpose derivatizing levels in human physiological fluids. Chirality
agents. J Chromatogr B 717(1–2):57–91 5(5):375–378
8. Opekar S, Kvicala J, Moos M, Pejchal V, Simek 17. Fujita T, Amuro Y, Hada T, Higashino K
P (2021) Mechanism of alkyl chloroformate- (1999) Plasma levels of pipecolic acid, both L-
mediated esterification of carboxylic acids in and D-enantiomers, in patients with chronic
aqueous media. J Org Chem 86(23): liver diseases, especially hepatic encephalopa-
16293–16299 thy. Clin Chim Acta 287(1–2):99–109
9. Rimnacova L, Husek P, Simek P (2014) A new 18. Fujita T, Hada T, Higashino K (1999) Origin
method for immediate derivatization of of D- and L-pipecolic acid in human physiolog-
hydroxyl groups by fluoroalkyl chloroformates ical fluids: a study of the catabolic mechanism
and its application for the determination of to pipecolic acid using the lysine loading test.
sterols and tocopherols in human serum and Clin Chim Acta 287(1–2):145–156
amniotic fluid by gas chromatography-mass 19. Dettmer K, Stevens AP, Fagerer SR, Kaspar H,
spectrometry. J Chromatogr A 1339:154–167 Oefner PJ (2012) Amino acid analysis in physi-
10. Husek P, Svagera Z, Hanzlikova D, ological samples by GC-MS with propyl chlor-
Rimnacova L, Zahradnickova H, Opekarova I, oformate derivatization and iTRAQ-LC-MS/
Simek P (2016) Profiling of urinary amino- MS. In: Alterman MA, Hunziker P (eds)
carboxylic metabolites by in-situ heptafluoro- Amino acid analysis: methods and protocols,
butyl chloroformate mediated sample vol 828, pp 165–181
 Chapter 12

UHPLC-TIMS-PASEF®-MS for Lipidomics: From Theory

to Practice
Vanna Denti, Simone Serrao, Eleonora Bossi, and Giuseppe Paglia

Abstract
Trapped ion mobility spectrometry (TIMS) using parallel accumulation serial fragmentation (PASEF®) is
an advanced analytical technique that offers several advantages in mass spectrometry (MS)-based lipido-
mics. TIMS provides an additional dimension of separation to mass spectrometry and accurate collision
cross-section (CCS) measurements for ions, aiding in the structural characterization of molecules. This is
especially valuable in lipidomics for identifying and distinguishing isomeric or structurally similar com-
pounds. On the other hand, PASEF technology allows for fast and efficient data acquisition by accumulat-
ing ions in parallel and then serially fragmenting them. This accelerates the analysis process and improves
throughput, making it suitable for high-throughput applications. Moreover, the combination of TIMS and
PASEF reduces co-elution and ion coalescence issues, leading to cleaner and more interpretable mass
spectra. This results in higher data quality and more confident identifications. In this chapter, a data-
dependent TIMS-PASEF® workflow for lipidomics analysis is presented.

Key words Lipidomics, PASEF, TIMS, Ion mobility, Mass spectrometry

1 Introduction

1.1 Lipidomics by Ultra-high-performance liquid chromatography (UHPLC) enables

LC-IM-MS the separation of lipids prior to MS analysis. Lately, UHPLC-MS
has become the technique of choice for untargeted lipidomic
experiments due to its broad coverage of lipid species, convenient
sample preparation, and high sensitivity [1]. The two most widely
used approaches for chromatographic separation in UHPLC-MS
for lipidomic analyses are hydrophilic interaction liquid chromatog-
raphy (HILIC) [2] and reversed phase chromatography
(RP) [3]. While HILIC separates lipids according to their polar
head groups (which result in distinct lipid class separation), RP
chromatography separates lipids by the composition of their fatty
acyl chains. In RP separation, lipid retention time increases with an
increasing fatty acyl carbon number and decreases with an increased
number of double bonds. Therefore, it is possible to separate

Olga Deda et al. (eds.), Metabolic Profiling: Methods and Protocols, Methods in Molecular Biology, vol. 2891,
https://doi.org/10.1007/978-1-0716-4334-1_12,
© The Author(s), under exclusive license to Springer Science+Business Media, LLC, part of Springer Nature 2025

221
222 Vanna Denti et al.

different lipid species from the same class by their cumulative

double bond index.
Recent studies have shown the benefits of combining tradi-
tional LC-MS-based lipidomics methods with ion mobility spec-
trometry [4]. The greatest difficulties in accurately identifying and
quantifying lipids is due to the wide variety and complexity of their
molecular structures, particularly from the presence of isomeric and
isobaric lipids [5, 6]. For instance, overlapping isotopologues of
identical lipids from the same lipid class that only differ by one
double bond frequently co-elute, resulting in the co-isolation and
co-fragmentation of precursor ions and eventually impacting the
validity of isotopic and MS/MS spectra [5, 6]. The separation and
distinction of isobaric and isomeric lipids in MS-based lipidomics
must therefore be improved.

1.2 TIMS-PASEF® in IM-MS offers a cutting-edge analytical platform that can separate
Lipidomics ions beyond the mass-to-charge ratio (m/z). This technique per-
mits to resolve isobaric and isomeric compounds using the mobility
of an ion, which is directly related to its size, shape, and charge as
well as to the properties of the buffer gas [7]. In this chapter, we
will focus on trapped ion mobility spectrometry (TIMS) technol-
ogy that was first introduced in 2011 by Fernandez Lima et al.
TIMS reverses the concept of the classical drift tube IMS [8].
In fact, TIMS holds ions in a stationary position against the gas
flow and then releases them depending on their mobility. In partic-
ular, it allows work to be performed in the parallel accumulation
serial fragmentation (PASEF®) operation mode in combination
with conventional LC-MS [9, 10]. In the dual TIMS design, ions
can be accumulated and analyzed simultaneously because the ion
accumulation and ion analysis functions are spatially separated.
The ions are first gathered in the trap; then they are swiftly
delivered to the analyzer for the mobility examination. To prevent
the loss of the ions, the trap is simultaneously loaded with the
following batch of ions.
TIMS-PASEF®, working in data-dependent (DDA) mode, can
extend the dynamic range of mass spectrometry, enabling the
detection of both highly abundant and low-abundance ions in the
same analysis. Therefore, TIMS-DDA-PASEF analysis can be used
to obtain level 2 lipid annotation [11] (see Notes 1 and 2).
In this modality, TIMS and MS/MS precursor selection are
synchronized. This system selects numerous precursors for MS/MS
acquisition in a single TIMS scan by using the temporal separation
of ions from the TIMS instrument.
The precursor ions in TIMS-QTOF equipment pass through
the analytical quadrupole serially rather than in parallel, as they do
in traditional MS/MS. When in PASEF® mode, an ion group is
collected in the upstream trap while another ion group that has
already been accumulated is examined simultaneously in the
 UHPLC-TIMS-PASEF®-MS for Lipidomics: From Theory to Practice 223

Fig. 1 Online Parallel Accumulation—Serial Fragmentation (PASEF®) with the timsTOF. (a) Analytes eluting
from the chromatographic column are ionized and enter the mass spectrometer through a glass capillary. (b)
In the dual TIMS analyzer, the first TIMS section traps and stores ion packets, and the second resolves them by
mobility. (c. d) Ion mobility-separated ions are released sequentially from the second TIMS analyzer as a
function of decreasing electrical field strength and yield mobility-resolved mass spectra. (e) In PASEF® MS/MS
scans, the TIMS analyzer and the quadrupole are synchronized, and the quadrupole isolation window switches
within sub-milliseconds between mobility-resolved precursor ions of different m/z. (f) This yields multiple ion
mobility-resolved MS/MS spectra from a single TIMS scan and ensures that multiple trapped precursor ion
species are used for fragmentation. (g) Cleaner MS/MS spectra obtained by the fragmentation of TIMS-
resolved ions

downstream TIMS analyzer (Fig. 1b). Accordingly, this approach

has shown improved molecular coverage and is crucial for compre-
hensive profiling of complex samples. The buildup takes place while
the TIMS analytical scan is running [12].

2 Materials

2.1 Equipment A liquid chromatography system such as an Elute UHPLC (Bruker

Daltonik GmbH, Germany)
1. TimsTOF fleX mass spectrometer (Bruker Daltonik GmbH,
Germany) or other mass spectrometers with TIMS-PASEF®
technology, such as timsTOF Pro (Bruker Daltonik GmbH,
Germany).
224 Vanna Denti et al.

2.2 Reagents 1. EquiSPLASH™ LIPIDOMIX® Quantitative Mass Spec Inter-

nal Standard (Avanti Polar Lipids, Alabaster, Alabama, USA).
2. Tuning mix MMI-L low concentration tuning mix (Agilent
Technologies, USA).
3. Sodium formate (Sigma-Aldrich®).
4. Acetic acid (Sigma-Aldrich®).
5. Ammonium acetate (Sigma-Aldrich®).
6. Solvents: isopropanol, methanol, acetonitrile, and water
(UHPLC-MS LiChrosolv®, Supelco).

2.3 Supplies 1. UHPLC column: Charged surface Hybrid (CSH™)

ACQUITY UPLC C18 (2.1 mM
100 mM) column,
1.7 μm (Waters).
2. Amber glass vial with screw top 2 mL, 12
32 mM (12 mM
cap), 5188-6535 (Agilent).
3. Screw cap with PTFE/red silicone septa (12 mM cap), 5185-
5820 (Agilent).
4. Vial insert, 150 μL, glass with polymer feet, for 2 mL standard
opening screw-top vials, 5183-2088 (Agilent).

2.4 Software 1. timsControl (Bruker Daltonik GmbH, Germany) is used to

control the timsTOF fleX instrument.
2. Hystar 6.0 (Bruker Daltonik GmbH, Germany) is used to
control the elute UHPLC system.
3. Data Analysis 6.0 (Bruker Daltonik GmbH, Germany) is used
to have a general view of the data acquired.
4. MetaboScape® (Bruker Daltonik GmbH, Germany) is used for
the qualitative analysis of 4D-lipidomics data (TIMS-DDA-
PASEF® data).

3 Methods

3.1 TIMS-MS All the following steps require the timsControl software.
Settings
1. Turn the instrument into the “Operate state.”
2. Select the appropriate acquisition method.
3. Wait for the instrument set-up (~ 20 min).
In this chapter, the steps for a DDA-PASEF® acquisition are
described. Therefore, the parameters to setup involve both the MS
and IM dimensions. To ensure accurate measurements both in the
MS and the IM dimensions, the instrument must be calibrated
 UHPLC-TIMS-PASEF®-MS for Lipidomics: From Theory to Practice 225

prior to any TIMS-MS acquisition experiment and post-

acquisition.

3.1.1 Data-Dependent For the purpose of this chapter, a DDA-PASEF® was performed
Acquisition (DDA) Coupled both in positive and negative ion mode. The main parameters of the
with TIMS PASEF® methods applied are reported in Tables 1, 2, 3 and 4. The para-
meters selected for the method must be used for calibration. If the
IM or the m/z range changes after the calibration, the timsControl
software will highlight with a red color the dimension that needs to
be calibrated. For the protocol presented in this chapter, the ana-
lyses were performed with 100–1350 m/z mass range and
0.55–1.90 V*s/cm2 mobility range (see Note 3).
Ensure that the MS method consists of three segments:
1. 0–0.05 min—The instrument operates in full scan and TIMS
on mode. The flow from the UHPLC is infused through a
divert valve to the ESI source.
2. 0.05–0.3 min—The instrument operates in full scan and TIMS
on mode. The UHPLC flow goes to waste and a calibrant
solution (see next paragraph Subheading 4.) is infused through
the divert valve.
3. 0.3–21 min—The instrument operates in PASEF® mode. The
UHPLC flow enters in the ESI source.

3.1.2 MS and TIMS 1. Prepare 1 mL of the following solutions:

Calibration – Sodium formate.
– Tuning mix MMI-L.
– Solution of sodium formate:tuning mix (1:1 v/v).

Table 1
TIMS setting and source parameters for lipidomics analysis in ESI(+) and (2) for a timsTOF fleX
instrument

TIMS settings

Polarity Mode Ramp time Accumulation time Duty cycle Ramp rate
ESI () Custom 100 ms 100 ms 100% 9.42 HZ
ESI (+) Custom 100 ms 100 ms 100% 9.42 HZ
Principal source parameters
Polarity Transfer
End plate offset Capillary Nebulizer Dry gas Dry temp
ESI () 500 V 3600 V 2.2 Bar 10 l/min 220  C
ESI (+) 500 V 4500 V 2.2 Bar 10 l/min 220  C
226 Vanna Denti et al.

Table 2
Tune parameters for lipidomics analysis in ESI(+) and (2) for a timsTOF fleX instrument

MS tune parameters

Transfer

Polarity Deflection 1 delta Funnel 1RF isCID energy Funnel 2RF Multiple RF
ESI () 80 V 500Vpp 0 eV 250 Vpp 200Vpp
ESI (+) +80 V 500Vpp 0 eV 250 Vpp 200Vpp
Polarity Quadrupole Focus pre TOF Collision cell
Ion energy Transfer time Pre-pulse storage Collision energy Collision RF
ESI () 5 eV 65 μs 5 μs 10 eV 1100 Vpp
ESI (+) 5 eV 65 μs 5 μs 10 eV 1100 Vpp
TIMS tune parameters
Offset
Polarity Deflection transfer Funnel 1In Collision cell in
To capillary exit To deflection discard To deflection transfer
ESI () 20 V 120 V 80 220 V
ESI (+) 20 V 120 V 80 220 V
Polarity Accumulation trap Accumulation exit Ramp start
Funnel 1 in To accumulation transfer To accumulation exit
ESI () 100 V 0 100 V
ESI (+) 100 V 0 100 V
Polarity Ion charge control Advanced parameters
Enable Target intensity Lock accumulation to mobility range
ESI () X 7.50 M X
ESI (+) X 7.50 M X

Make sure to have two Hamilton glass syringes of 500 μL, one
for the first solution and one for the other two.
2. The pre-acquisition calibration is a two-step procedure:
– External m/z calibration is performed by direct infusion
using a divert valve at 180 μL/h of a sodium formate
solution using the high-order polynomial calibration
(HPC) calibration mode.
External mobility calibration is performed by direct
infusion using a divert valve at 180 μL/h of tuning mix
MMI-L using the linear calibration mode.
 UHPLC-TIMS-PASEF®-MS for Lipidomics: From Theory to Practice 227

Table 3
PASEF® parameters for lipidomics analysis in ESI(+) and (2) for a timsTOF fleX instrument

PASEF® parameters

Precursor ions
Collision energy settings Active exclusion
Polarity # PASEF ramps Total cycle time Collision energy Release after
ESI () 2 0.32 s 40 eV 0.10 min
ESI (+) 2 0.32 s 40 eV 0.10 min
Polarity Precursor repetition Isolation width
Target intensity Intensity threshold
ESI () 4000 100 2 m/z
ESI (+) 4000 100 2 m/z

Table 4
MetaboScape® import parameters

T-ReX 4D workflow

Peak detection Ranges MS/MS import

parameters
Intensity Minimum 4D peak
Polarity threshold size RT m/z Import method
ESI () 1000 100 0.3–20 min 50–1000 Max sum
ESI (+) 1000 100 0.3–20 min 50–1000 Max sum
Ion deconvolution
Polarity Primary ion Seed ions Common ions EIC correlation
ESI () [M-H] [M + CHOO] [M + Cl] 0.8
+ + +
ESI (+) [M-H] [M + Na] [M + H–H2O] 0.8
[M + NH4]+ [M + K]+

3. Save the calibrated TIMS-PASEF® method (see Note 4).

Additionally, for a post-acquisition recalibration of both
the MS and IM, an external calibrant is infused during the
acquisition.
4. Fill the syringe used with the tuning mix with the sodium
formate:tuning mix (1:1 v/v) and connect it to a divert valve.
5. The calibrant mix is infused using a divert valve in a calibration
segment from 0.05 to 0.3 min at the start of each analysis run.
228 Vanna Denti et al.

3.2 UHPLC For the work presented in this chapter, the extracted lipidome was
Preparation separated by LC prior to the TIMS-PASEF® measurement using an
elute UHPLC system. RP chromatographic separation was
achieved using an ACQUITY UPLC CSH C18 Column (Waters).
1. Prepare the following solutions:
– Mobile phase A: 10 mM ammonium acetate:acetonitrile
(40:60 v/v) with 0.1% acetic acid.
– Mobile phase B: Isopropanol:phase A (90:10 v/v).
– Seal wash: Water:isopropanol (80:20 v/v).
2. Mount the column in the column compartment.
3. Install the solutions in the corresponding lines.
4. Insert the wash lines as follows:
(a) Wash 1: in Mobile phase A.
(b) Wash 2: in Mobile phase B.
5. Open the Hystar 6.0 software to control the UHPLC system:
– Prime the seal wash.
– Purge solvent lines A and B.
– Perform the initial wash of the autosampler (Wash
2 + Wash 1).
6. Set the parameters of the chromatography:
– Set the column temperature at 55  C and turn on the
column oven.
– Set the flow rate to 0.250 mL/min.
– Set the injection mode to “μL-pickup.”
– Set the autosampler temperature to 10  C.
– Set the elution gradient timetable:
– 0 min 99% A, 1 min 99% A, 1.10 min 60% A, 5 min 20% A,
11 min 20% A, 12 min 1% A, 18 min 1% A, 18.10 min
60% A, 20 min 99%. Add 1 min of equilibration time after
each run.
7. Turn on the flow and start at least three runs with no injection
to perform column conditioning.
8. Save the method.
The injection volume will be defined when setting the sample
list, as described in the next paragraph.

3.2.1 Sample List Using Hystar 6.0 software, it is possible to set up the sample list
Settings table.
1. Indicate the name and position of the sample.
2. Set the number of injections for each sample.
 UHPLC-TIMS-PASEF®-MS for Lipidomics: From Theory to Practice 229

3. Set the MS method (set the TIMS-PASEF® method saved and

calibrated in the timsControl software).
4. Set the LC method (the one created and saved using
Hystar 6.0).
5. Set the injection volume: 4 μL for acquisition in negative ion
mode; 2 μL for acquisition in positive ion mode.
6. Set the system directory to save data files.
7. Save the sample list.
8. Select the first sample in the list and load the parameters.

3.2.2 Sample Analysis 1. Perform the lipid extraction. There are numerous techniques
and complete procedures available in the literature (see
Note 5).
2. Prepare a pooled sample using aliquots of the lipid extracts
from each sample. This will act as a quality control (QC)
sample.
3. Your samples should be put into glass vials with screw caps and
PFTE septa. In order to use less sample volume for the analysis,
glass adapters can be used.
4. Place the vials in the autosampler and proceed with the sample
list setting, as described in the previous paragraph.
5. Perform at least four equilibration runs using the QC sample at
the start of each batch of analyses.
6. The order of sample injection should be randomized, and QC
samples should be analyzed during the run. Use these to test
system suitability.
7. Start the data acquisition.

3.2.3 Data Extraction and Once the lipidomic analysis is complete, it is possible to have a quick
Data Processing check on the quality of the data acquisition using DataAnalysis
software (see Note 6).
In Figs. 2 and 3 are illustrated some examples of the output
visible with this software. The base peak chromatogram and the
total ion count of the MS2 sample analyzed in positive and negative
ion modes are shown. Lipidomics analysis obtained with
LC-TIMS-PASEF® mode can be represented as four-dimensional-
data: time, mobility, m/z, and MS/MS.
Checking these results, you can have a qualitative overview of
the analysis. This step is not mandatory but can be a first screening.
To proceed with the lipidomic data analysis, it is necessary to obtain
a list of identified lipids and their intensities in each sample. To do
so, it is possible to use different software: MetaboScape®, (Bruker
Daltonik GmbH, Germany), or Mzmine3 (http://mzmine.github.
io/).
230 Vanna Denti et al.

Fig. 2 The base peak chromatogram of a lipid extract from plasma analyzed as described in the previous
paragraphs, in ESI(+) a and ESI() c. The corresponding total ion chromatogram (TIC) of the MSn ions is
reported in b and d

It is possible to find more about mzmine3 methods for the

elaboration of TIMS-PASEF® data in the work by Schmid et al.
[13] (see Note 7).

Data Import
For 4D lipidomic data, we need to define import parameters for all
the data dimensions. In general, it is necessary to take care of the
following steps:
– Set the intensity threshold and the data point expected for
each peak.
– Set RT and m/z ranges. Mobility ranges will be automatically
detected.
– Set the import method for the MS/MS.
– Perform ion deconvolution, setting the type of ion adducts
expected.
– Perform data recalibration considering the calibration segment
at 0–0.3 min.
 UHPLC-TIMS-PASEF®-MS for Lipidomics: From Theory to Practice 231

Fig. 3 An example of high quality MS/MS spectra obtained with TIMS-PASEF® analysis in ESI(+) and ESI()
mode (a and b, respectively). The ion ID, m/z of the precursor, RT, and mobility are reported. The MS/MS
fragments can be used to confirm the ion identity of a PC(15:0/18:1)(d7)
232 Vanna Denti et al.

In Table 4 are reported the main parameters used in MetaboS-

cape® to process the data acquired as described in the previous
paragraphs (see Note 8).

Lipids Annotation
The 4D feature list imported after the previous step can be anno-
tated matching the experimental data with experimental lipid
libraries or with computational algorithms (i.e., Lipid Species
Annotation, LipiDex Library Forge [14]).
MetaboScape® enables the annotation of lipid species consid-
ering accurate m/z ratio, isotopic pattern, and CCS (if available) of
the precursor, as well as MS/MS spectra. An annotation is assigned
if the score for each of these parameters is higher than a defined
threshold. A final total score for the match is assigned for each
identification. If necessary, lipid identification can be curated man-
ually, examining the MS/MS spectra of the feature.
One of the advantages of TIMS-PASEF® analysis is the possi-
bility to obtain clean spectra, as described in the introduction. In
Fig. 3 is possible to observe the MS/MS spectra (A. ESI+, B. ESI)
of the feature annotated as a PC(15:0/18:1)[d7]. The feature has
the same RT and mobility, and the fragments are those specific for
this lipid fragmentation, both in positive and negative mode.
Additionally, an overall view of the lipids identified in an ana-
lytical batch can be offered by plotting a Kendrick Mass-Defect
Plot, and it can be used to further validate the level 2 annotations
assigned by Lipid Species Annotation. In the example in Fig. 4, it is
possible to see the KMD for CH2 defect plotted against the m/z
obtained from the positive and negative (A and B, respectively).
Another advantage of using ion mobility for lipidomics analysis
is that it can potentially discriminate isomeric molecules. In Fig. 5,
it is reported the example of two isomers of a cholesterol ester
(CE) 18:3 putatively annotated with Lipid Species Annotation. In
panel A is possible to see the extracted ion chromatogram (EIC) of
CE(18:3) + NH4+ (664.6031 m/z  0.005 Da) in green. The EICs
of the two isomers (R1 at 1/K0 ¼ 1.394–1.421 and R2
1/K0 ¼ 1.424–1.454. The regions isolated from the heatmap in
panel B allowed to extract the ion mobilograms (EIMs) of R1 and
R2 (panel C), respectively. In panel D, it is possible to observe that
the two species share the same m/z precursor (indicated as a blue
square) and some characteristic peaks (i.e. 369.3516 m/z: -FA 18:3
(+HO) -Cholesterol), despite some differences in the lower mass
range (i.e., 311.2732 m/z and 313.2890 m/z).
Fig. 4 Kendrick Mass-Defect (KMD) Plot showing the level 2 annotated lipids of the two datasets obtained as
described above. The figure above (a) reports the data of the ESI(+) analysis, whereas b reports the data of the
ESI() analysis. As shown in the legend on the right, the colors correspond to the lipid classes identified, while
the radius of the dots is proportional to their CCS. CCS and RT outlier symbols are also reported
234 Vanna Denti et al.

Fig. 5 Example of two isomers of a compound identified as a CE(18:3) by Lipids Species Annotation algorithm.
In panel a, the extracted ion chromatogram (EIC) of the [M + NH4]+ 664.6031 m/z: in green, the overall EIC; in
blue and red, the EICs of the R1 and R2 isomers, respectively. In panel b are indicated the two isomers ions
isolated in the regions of interest R1 and R2. The extracted ions mobilograms (EIMs) of the isolated regions are
reported in panel c. In panel d, the MS/MS spectra of the two isomers of the CE(18:3), in blue the R1 and in red
R2 (see Note 3)

Data Export for Statistical Analysis

– Export the annotated feature lists as csv.
– Perform normalization, scaling, and data quality check
using QCs.
– Perform univariate and multivariate statistical analysis as needed.
Many free and open source software programs are available for
lipid functional data analysis, including pathway and enrichment
analysis [15–18].
 UHPLC-TIMS-PASEF®-MS for Lipidomics: From Theory to Practice 235

4 Notes

1. To reach higher sensitivity in quantitative analysis, a possibility

could be to operate in full scan with TIMS on. Following this,
the analytes can be matched (using m/z, RT, isotopic cluster,
and CCS) with experimental libraries obtained after the TIMS-
DDA PASEF® analysis of pooled samples using the same LC
separation.
2. The TIMS-PASEF® technology can also work in
data-independent acquisition (DIA) and parallel reaction mon-
itoring (PRM) modes. Despite the possible advantages in sen-
sitivity and in MS/MS spectra quality, at the moment, the data
processing of lipid analysis for this type of data is troublesome
and time-consuming [12].
3. The IM resolution of a TIMS instrument can be improved by
increasing the ramp time, which defines the length of the TIMS
separation in both TIMS cells. This will also decrease propor-
tionally the amount of ions reaching the analyzer and, conse-
quently, the sensitivity. Thus, it can result in a lower number of
MS/MS events.
4. The calibrated ranges used for the acquisition must be the same
as those used for the IM calibration, which is related to the m/z
calibration. A new calibration is required for the new m/z and
IM ranges if either one of the two parameters changes.
5. The sample used as an example in this chapter is a lipid extract
obtained after protein precipitation of plasma with cold metha-
nol (1:5 ratio). In this case, the methanol solution contained
5% of deuterated EquiSPLASH® LIPIDOMIX® Mass Spec
Standards. Samples were vortexed for 20 s before incubation
at 20  C for 1 h. The extracted samples were centrifuged in an
Eppendorf centrifuge 5424 R (Eppendorf, Hamburg, Ger-
many) at 15000 rpm for 15 min at 4  C before the supernatant
was transferred to glass vials.
6. This data check with DataAnalysis software can be useful when
optimizing acquisition method parameters and comparing dif-
ferent analyses.
7. mzmine3 is an open-source software for mass spectrometry
data processing. TIMS-PASEF® data can be elaborated with
this software by defining parameters such as peak picking,
smoothing, mobility, and RT alignment [13]. Tims-tof raw
data (.d files) can be imported directly, and the processing can
be scheduled, defining a batch workflow. The aligned feature
lists can be merged and used for lipids annotation. This can be
performed by matching with a spectral library or using a plug-
in for in silico lipids fragmentation. The aligned mass list can
236 Vanna Denti et al.

also be exported for annotation with SIRIUS [13] or GNPS

[19]. The processed feature list annotated can be exported as .
csv for external statistical analysis or can be exported directly to
Metaboanalyst 5.0 [20].
8. When multiple sample groups are analyzed, the data import
using MetaboScape® must take into account the sample size of
each group. Accordingly, the software requires a “minimum
number of features for extraction” and the “presence of fea-
tures in a minimum number of analyses.” To define this num-
ber in a proper way, it is recommended to activate the option
“Filter features by occurrences in groups.” Consequently,
select a presence of features in sample group of 80%. Then,
access the corresponding number of features. This number will
define the setting value for the “minimum number of features
for extraction” and “presence of features in a minimum num-
ber of analyses.”

Acknowledgments

Research funded by Regione Lombardia, regional law n 9/2020,

resolution n 3776/2020: Programma degli interventi per la
ripresa economica: sviluppo di nuovi accordi di collaborazione
con le università per la Ricerca, l’Innovazione e il Trasferimento
tecnologico: NephropaThy.

References
1. Züllig T, Trötzmüller M, Köfeler HC (2020) 5. Ridgway N, McLeod R (2021) Biochemistry of
Lipidomics from sample preparation to data lipids, lipoproteins and membranes. Elsevier
analysis: a primer. Anal Bioanal Chem 412: 6. Wojcik R, Webb IK, Deng L, Garimella SV,
2191–2209 Prost SA, Ibrahim YM, Baker ES, Smith RD
2. Zhang Z, Singh M, Kindt A, Wegrzyn AB, (2017) Lipid and glycolipid isomer analyses
Pearson MJ, Ali A, Harms AC, Baker P, Han- using ultra-high resolution ion mobility spec-
kemeier T (2023) Development of a targeted trometry separations. Int J Mol Sci
hydrophilic interaction liquid 7. Paglia G, Smith AJ, Astarita G (2022) Ion
chromatography-tandem mass spectrometry mobility mass spectrometry in the omics era:
based lipidomics platform applied to a corona- challenges and opportunities for metabolomics
virus disease severity study. J Chromatogr A and lipidomics. Mass Spectrom Rev 41:722–
1708:464342 765
3. Konjevod M, Sáiz J, Nikolac Perkovic M, 8. Fernandez-Lima F, Kaplan DA, Suetering J
Nedic Erjavec G, Tudor L, Uzun S, et al (2011) Gas-phase separation using a
Kozumplik O, Barbas C, Zarkovic N, Pivac N, trapped ion mobility spectrometer. Int J Ion
Strac DS (2022) Plasma lipidomics in subjects Mobil Spectrom 14
with combat posttraumatic stress disorder. Free 9. Meier F, Park MA, Mann M (2021) Trapped
Radic Biol Med 189:169–177 ion mobility spectrometry and parallel
4. Camunas-Alberca SM, Moran-Garrido M, Sáiz accumulation-serial fragmentation in proteo-
J et al (2023) Integrating the potential of ion mics. Mol Cell Proteomics 20:100138
mobility spectrometry-mass spectrometry in 10. Di Poto C, Tian X, Peng X et al (2021) Meta-
the separation and structural characterisation bolomic profiling of human urine samples
of lipid isomers. Front Mol Biosci 10:1112521
 UHPLC-TIMS-PASEF®-MS for Lipidomics: From Theory to Practice 237

using LC-TIMS-QTOF mass spectrometry. J mammalian lipidome metabolic pathways on

Am Soc Mass Spectrom 32:2072–2080 LIPID MAPS. F1000Res 10:4
11. Sumner LW, Amberg A, Barrett D et al (2007) 16. Molenaar MR, Jeucken A, Wassenaar TA et al
Proposed minimum reporting standards for (2019) LION/web: a web-based ontology
chemical analysis. Metabolomics 3:211–221 enrichment tool for lipidomic data analysis.
12. Rudt E, Feldhaus M, Margraf CG et al (2023) Gigascience 8
Comparison of data-dependent acquisition, 17. Li F, Song J, Zhang Y et al (2021) LINT-web: a
data-independent acquisition, and parallel web-based lipidomic data mining tool using
reaction monitoring in trapped ion mobility intra-omic integrative correlation strategy.
spectrometry-time-of-flight tandem mass Small Methods 5:e2100206
spectrometry-based lipidomics. Anal Chem 18. Bhattacharya SK (2023) Lipidomics: methods
95:9488–9496 and protocols. Springer
13. Schmid R, Heuckeroth S, Korf A et al (2023) 19. Wang M, Carver JJ, Phelan VV et al (2016)
Integrative analysis of multimodal mass spec- Sharing and community curation of mass spec-
trometry data in MZmine 3. Nat Biotechnol trometry data with global natural products
41:447–449 social molecular networking. Nat Biotechnol
14. Hutchins PD, Russell JD, Coon JJ (2018) 34:828–837
LipiDex: an integrated software package for 20. Pang Z, Chong J, Zhou G et al (2021) Meta-
high-confidence lipid identification. Cell Syst boAnalyst 5.0: narrowing the gap between raw
6:621–625.e5 spectra and functional insights. Nucleic Acids
15. Gaud C, Sousa CB, Nguyen A et al (2021) Res 49:W388–W396
BioPAN: a web-based tool to explore
 Chapter 13

Advanced LC-IMS-MS Protocol for Holistic Metabolite

Analysis in Wine and Grape Samples
Vania Sáez, Sara Ferrero-del-Teso, Fulvio Mattivi, Urska Vrhovsek,
and Panagiotis Arapitsas

Abstract
The final aim of metabolomics is the comprehensive and holistic study of the metabolome in biological
samples. Therefore, the use of instruments that enable the analysis of metabolites belonging to various
chemical classes in a wide range of concentrations is essential, without compromising on robustness,
resolution, sensitivity, specificity, and metabolite annotation. These characteristics are crucial for the analysis
of very complex samples, such as wine, whose metabolome is the result of the sum of metabolites derived
from grapes, yeast(s), bacteria(s), and chemical or physical modification during winemaking. In recent
years, a big advantage, in this direction, was the hardware developments on hyphenated instruments that
enable the integration of liquid chromatography (LC), ion mobility spectrometry (IMS), and mass spec-
trometry (MS). This chapter describes an LC-IMS-MS protocol for the analysis of wine and grape samples
as well as the use of IMS data in metabolite annotation.

Key words Vitis, Metabolomics, Metabolite annotation, Traveling wave ion mobility, CCS, Mass
spectrometry, HDMSE

1 Introduction

Ion mobility spectrometry (IMS) is an analytical technique based

on ion separation by their mobility, while they traverse a gas-filled
drift region under the influence of an electric field. The separation
of ions is based on their charge, size, and shape [1] within a time
scale of milliseconds, whereas mass spectrometer detection occurs
within microseconds [2]. This distinction facilitates the hyphen-
ation of IMS and MS, typically with a quadrupole time-of-flight
(QToF) mass spectrometer, adding a new post-ionization separa-
tion, partly orthogonal to mass spectrometry, without interfering
with the timescale of the MS analyzer [2, 3].
The development of instruments that combine ion mobility
spectrometry and mass spectrometry (IMS-MS) commenced in

Olga Deda et al. (eds.), Metabolic Profiling: Methods and Protocols, Methods in Molecular Biology, vol. 2891,
https://doi.org/10.1007/978-1-0716-4334-1_13,
© The Author(s), under exclusive license to Springer Science+Business Media, LLC, part of Springer Nature 2025

239
240 Vania Sáez et al.

1970 and has been commercially available since 2006

[2, 3]. According to the technology, the commercially available
IMS-MS analyzers can be classified as drift tube ion mobility spec-
trometry (DTIMS), traveling wave ion mobility spectrometry
(TWIMS), trapped ion mobility spectrometry (TIMS), field asym-
metric waveform ion mobility spectrometry (FAIMS), and cyclic
ion mobility spectrometry (cIMS) [4].
Similar to DTIMS, TWIMS instruments are considered linear
and time-dispersive methods [3, 5]. In contrast to DTIMS, which
uses uniform static voltage gradients, TWIMS uses traveling waves,
a continuous train of voltage pulses driving packets of ions through
an ion guide device full filled with a gas. The ions are periodically
overtaken by the waves, where the ions with lower mobility are
overtaken by traveling waves more often than those with higher
mobility and are therefore retained for longer time periods in the
mobility device, achieving ion separation by mobility [6].In
TWIMS instruments, the collision cross-section (CCS) value,
which is considered a physicochemical identifier, is determined by
instrumental calibration using reference compounds under well-
established and very specific conditions [3, 4, 6].
The incorporation of IMS into an LC-IMS-MS instrument
should enable researchers to address the challenge of analyzing
complex biological matrices more effectively in an untargeted
approach, because it simultaneously increases metabolite coverage
and annotation confidence. The use of LC-IMS-MS analytical plat-
form data-dependent (DDA) or data-independent (DIA) acquisi-
tion facilitates the acquisition of fourth-dimensional data in a single
injection (retention time, CCS, MS, MS/MS, or MSE), thereby
improving peak capacity and enhancing fragmentation specificity.
As has been demonstrated, the experimental IMS-CCS-derived
value for a determined adduct or ion is a useful compound descrip-
tor and is highly reproducible [7–11]. Indeed, the inclusion of the
CCS value to determine a compound, together with the mass-to-
charge ratio (m/z), increases the confidence level for metabolite
annotation and decreases the number of candidates [3, 10, 11].
In wine and grape metabolomics, LC-MS is the most prevalent
analytical platform to provide an unbiased and comprehensive view
of the metabolome. The application of the LC-MS untargeted
metabolomics approach in wine science enables us to gain a greater
understanding of the wine metabolome and how it is influenced by
diverse conditions, such as terroir, quality, storage conditions, and
variety discrimination, among others [12–16]. Very recently, wine
research has begun to transition to LC-IMS-MS for multidimen-
sional characterization of wine according to variety, origin, and
vintage [17, 18]. This chapter presents a protocol for the analysis
of wine and grape samples using an LC-IMS-MS analytical platform
for an untargeted metabolomics workflow with DIA acquisition
method (HDMSE) (Fig. 1). The protocol was developed on a
 Advanced LC-IMS-MS Protocol for Holistic Metabolite Analysis in Wine. . . 241

Fig. 1 Schematic overview of the LC-IMS-MS protocol for wine and grape samples. The UPLC-TWIMS-QToF
protocol has two separations: the first is pre-ionization in the UPLC column, and the second is post-ionization
in the TWIMS device. In the HDMSE DIA method, the trap acts as an ion guide, the TWIMS device achieves the
separation by ion mobility, and the transfer device operates as collision cell. In the HDMSE DIA, the precursor
ion is associated with drift time data, and the fragments are matched with the precursor by retention time
correlation data. This strategy allows to obtain multidimensional data in a single injection

quadrupole-TWIMS-ToF geometry mass spectrometer (Synapt XS,

Waters), in which the TWIMS device is part of the so-called triwave
region with Trap and Transfer devices, which can function as ion
guides or collision cells. The chapter includes a protocol for sample
preparation, a description of data analysis and visualization using
Progenesis QI, and the construction, management, and use of an
in-house database to support confident annotation.

2 Materials

2.1 Reagents and Leucine-enkephalin acetate salt hydrate (≥ 95% HPLC).

Solvents Major MIX IMS/ToF Calibration Kit, Waters™ (see Note 1).
MiliQ water (18.2 MΩ) and water for chromatography
(LC-MS grade).
Methanol, acetonitrile, 2-propanol, and formic acid for
UPLC/MS (99%) (LC-MS grade, ≥ 99.9%) (see Note 1).
Ar(g) as API gas, He(g) for helium cell, N2(g) for IMS, and
TRAP gas (> 99.9% vol/vol).
Commercially available standards (≥ 95%).

2.2 Devices, Borosilicate glass syringe of 0.5 mL with stainless steel and PTFE
Equipment, and luer lock (see Note 2).
Instruments UPLC Column Acquity TM Premier HSS T3 with VanGuard
Fit, 1.8 μm, 2.1 × 100 mM, Waters™ (SKU:186009471).
Acquity I-class UPLC with a binary solvent manager, sample
manager FTN-1, column manager, and sample organizer (Waters
242 Vania Sáez et al.

Corporation) in tandem with a Synapt-XS high-resolution mass

spectrometer with a ZSpray™ LockSpray™ II ESI source (Waters
Corporation).

2.3 Software DriftScope v2.9, Waters Corporation.

MassLynx v4.2 software, Waters Corporation.
Progenesis QI v3.0, Nonlinear Dynamics, Waters Corporation.
EZinfo v30.3, Umetrics.
UNIFI Scientific Information System v1.9.4.053 (Waters
Corporation).

3 Methods

3.1 Solvent Prepare a leucine enkephalin (C28H37N5O7, MW: 555.6 mg/

Preparation (See moL) solution of 1 mg/mL. Weigh 6.1 mg of leucine-enkephalin
Note 2) acetate salt hydrate (C30H43N5O10, MW: 633.7 g/moL) and
transfer it to a 5 mL volumetric flask. Add 4.5 mL of MiliQ water
3.1.1 Preparation of Lock and 0.5 mL of LC-MS-grade methanol. Keep this solution at 4 °C.
Mass (LM) Solution In a 500 mL volumetric flask, add 250 mL of LC-MS-grade
methanol, 50 μL of leucine enkephalin solution (1 mg/mL), and
0.5 mL formic acid LC-MS grade, and fill with LC-MS-grade water
up to 500 mL. Pour the 100 pg/μL leucine enkephalin solution
into a dark 500 mL certified bottle and sonicate for 5 min. Label
this solution as a lock mass solution (LM).

3.1.2 CCS Major MIX The CCS calibration mix, referred to as the CCS Major MIX
IMS/ToF Calibration IMS/ToF solution, allows automated or manual mass and CCS
Solution calibration for high-resolution mass spectrometers with ion mobil-
ity. The CCS Major MIX IMS/ToF content (see Note 3) enables
calibration up to 2000 Da in negative and positive acquisition
modes. The solution was prepared using a glass Pasteur pipette
following the detailed procedure described by the manufacturer
[19]. After calibration, the solution should be stored at 4 °C.

3.1.3 Mobile Phase and The chromatographic separation required two mobile phases:
Wash Solution Preparation (A) water with 0.1% formic acid and (B) methanol with 0.1% formic
acid. For mobile phase A in a 2 L volumetric flask, fill half the flask
with LC-MS-grade water, add 2 mL of formic acid, and fill the
volumetric flask with LC-MS-grade water. For mobile phase B in a
2 L volumetric flask, fill 1 L with LC-MS-grade methanol, add
2 mL of formic acid, and fill to its total volume with LC-MS-
grade methanol. Pour the mobile phase into a certified class A
borosilicate bottle.
In addition to mobile phases, this method also requires the
preparation of a syringe wash, seal wash, and purge and wash
solutions. The seal wash solution is composed of water and aceto-
nitrile (90:10). To prepare, mix 450 mL of MS-grade water and
 Advanced LC-IMS-MS Protocol for Holistic Metabolite Analysis in Wine. . . 243

50 mL of MS-grade acetonitrile in a 500 mL volumetric flask. Pour

the solution into a certified bottle. To prepare the wash solution
(water, methanol, 2-propanol, and acetonitrile (25:25:25:25)), add
125 mL of each MS-grade solvent (water, methanol, 2-propanol,
and acetonitrile, respectively) to a 500 mL volumetric flask. Pour
into a certified 500 mL bottle-labeled wash solution. To prepare
the purge solution (water and methanol (90:10)), add 450 mL of
LC-MS-grade water into a 500 mL volumetric flask, and fill to its
total volume with MS-grade methanol. Transfer the solution to a
certified bottle labeled purge solution. The syringe wash solution is
composed of methanol and water (50:50). Pour 250 mL of
MS-grade methanol into a 500 mL volumetric flask and fill to its
total volume with MS-grade water, mix, and pour into a labeled
certified borosilicate bottle.

3.2 LC-IMS-MS The Synapt XS high-resolution mass spectrometer (Waters Corpo-

Method ration) can operate in four distinct resolution modes: sensitivity,
resolution, high resolution, and enhanced resolution. Before start-
3.2.1 IMS (TWIMS)-MS
ing, calibrate the mass spectrometer (QToF) using assisted or man-
(QToF) Calibration
ual calibration in IntelliStart with a reference solution and LM
solution (see Note 4).
For the next steps of calibration, select “resolution” and “MS
mode” acquisition in “mobility ToF” mode.
Tune and set the instrument in accordance with the parameters
listed in Table 1. Before beginning the calibration, the instrument
must operate for 1 h in “mobility ToF” mode (see Note 4).
Sonicate the CCS Major MIX IMS/ToF solution for 5 min,
purge the analyte probe with the solution, and initiate infusion at
20 μL/min. Ensure a good spectral signal intensity with a stable
beam in the range of 50–2000 Da prior to starting the calibration
process (see Note 3).
CCS calibration for TWIMS using a reference solution (CCS
Major MIX IMS/ToF solution) is the first calibration step. For this
purpose, select the “create CCS calibration” option on the config-
uration mode in the Intellistart feature from “Synapt XS” tab in the
MassLynx console. Acquire the calibration data manually (the
assisted option is also available) with the drift time function selected
in continuum mode for 4 min with a scan time of 0.2 s in the range
of 50–2000 Da. Insert the created calibration data file for the
selected reference compound profile with the correct acquisition
mode used (Resolution, ESI+, or ESI-). Intellistart generates a
summary report that includes the determination of the adjusted
CCS equation (Ω adjusted). Verify the monoisotopic peaks (with a
single charge) in the reference and data file matches. The calibration
is acceptable if the root-mean-square (RMS %CCS) value is
below 1.5.
244 Vania Sáez et al.

Table 1
IMS-MS operating parameters for grape and wine UPLC-TWIMS-QToF protocol

ESI + acquisition ESI - acquisition

mode mode
Acquisition setting Mass range 50–2000 Da
Scan time 0.2 s
Type of acquisition Continuum
Source conditions Capillary 3.00 kV 2.00 kV
Sampling cone 40 30
Source offset 4
Source temperature 120 °C
Desolvation temperature 450 °C
Cone gas flow 50 L/h
Desolvation gas 800 L/h
Nebulizer (Bar) 6.5
IMS gas configuration Nitrogen IMS buffer gas 90 mL/min
flow
Helium cell gas flow 180 mL/min
Lock mass infusion LockSpray capillary (kV) 3.0 2.6
setting LockSpray flow 10.0 μL/min
Triwave DC Trap DC entrance 3.0
Trap DC bias 45.0
Trap DC 0
Trap DC exit 0
IMS DC entrance 20.0
IMS helium cell DC 50.0
IMS helium exit -20
IMS bias 3.0
IMS exit 0
Transfer DC entrance 5.0
Transfer DC exit 15.0
Triwave Trap wave velocity 311 m/s
Trap wave height 6.0 V
IMS wave velocity 650 m/s
IMS wave height 40 V
Transfer wave velocity 141 m/s
Transfer wave height 4.0 V

The second calibration step corresponds to the “LockCCS

setup” using the LM solution (Subheading 3.1.1). This step must
always be performed after CCS calibration. Purge the analytical
probe with the LM solution and initiate the infusion at 10 μL/
min. Before running the calibration, initiate the infusion and wait
for the beam to stabilize. Start the “LockSpray source setup” in
Intellistart using the tune page settings (Table 1). Intellistart auto-
matically acquires LM data and generates a report.
 Advanced LC-IMS-MS Protocol for Holistic Metabolite Analysis in Wine. . . 245

3.2.2 LC Analytical The column manager is set at 40 °C, the injection volume is 2 μL,
Parameters the flow rate is 0.3 mL/min, and the multistep gradient is as
follows: 0.0–1.0 min (0% mobile phase B), 1.0–3.0 min (0–10%
mobile phase B), 3.0–12.0 min (10–40% mobile phase B),
12.0–16.0 min (40–100% mobile phase B), 16.0–18.0 min (100%
mobile phase B), 18.0–18.10 min (100–0% mobile phase B),
18.10–20.0 min (0% mobile phase B). Before inserting the rack
with the sample vials in the autosampler and/or sample organizer,
activate the temperature control and set to 4 °C in the sample
manager of the UPLC system. Place the vials once the temperature
is stable.

3.2.3 TWIMS-QToF The DIA HDMSE method was developed for the resolution acqui-
Method sition mode for both polarities (ESI+ and ESI-), with a normal
dynamic range in continuum acquisition mode ranging from
50 to 2000 Da with a scan time of 0.2 s. The source conditions,
IMS settings, Triwave, and Triwave DC for ESI+ and ESI- para-
meters are described in Table 1 (Subheading 3.2.1).
The collision energies for the functions in the HDMSE method
differ. For ESI+ and ESI-, the low-energy function is disabled (<
5 V). The collision energy was ramped from 20 to 50 V in ESI-
mode acquisition and from 20 to 40 V in ESI+ in the transfer
collision cell for the high energy function. An LM solution (Sub-
heading 3.1.1) was infused throughout the entire data acquisition
process (Table 1). The acquisition interval for LM is 15 s, with a
scan average of 3. The HDMSE method acquires the LM but does
not apply the correction (see Subheading 3.4, Note 5).

3.3 Sample Prior to sample preparation and/or data acquisition using analytical
Preparation and Data methodology, the untargeted metabolomic workflow must include
Acquisition relevant steps, such as experimental design and pre-analytical pro-
cess [20, 21].
The collection and organization of metadata must begin with
experimental design and sampling. The documented metadata
must be clear and accessible to any researcher and must track any
changes or observations made to the samples during the metabo-
lomic workflow (see Note 6). Codify the samples according to a
random sequence for sample preparation, extraction, and analysis
[22]. Label each sample with the relevant information. It is prefer-
able to use printed labels rather than handwritten labels to avoid
unreliable readings and/or ink smears during the metabolomic
workflow.

3.3.1 Preparation of Wine To prevent undesirable oxygen-dependent reactions, aliquot the

Samples wine into 10 mL labeled dark vials and fill to the total volume
under an N2 atmosphere in a random sequence. Prepare a pooled
quality control (QC) by mixing equal aliquots of each sample in a
dark vial under an N2 environment (see Note 7).
246 Vania Sáez et al.

Always under an N2 atmosphere, dilute each wine sample with

degasified MiliQ water (from 2 to 5 times, depending on the
experimental design) and filter into certified 2 mL dark vials using
a 0.22 μm PDVF filter. Include a blank (MiliQ water) employing
the same procedure. The wine sample bottles and aliquots must be
stored at 4 °C after sample preparation. Avoid unnecessary thawing
cycles from 4 °C to room temperature throughout the metabolo-
mic workflow.

3.3.2 Preparation of Weight 100 mg of frozen ground grape into a 2 mL labeled

Grape Samples Eppendorf tube. Under a fume hood, add 200 μL of methanol
and 200 μL of water and 500 μL of dichloromethane (see Note 1),
and vortex the tube for 1 min. After centrifugation for 5 min at 4 °C
and 13,000 × g, return the samples to the fume hood, transfer the
aqueous phase of each sample to a new labeled Eppendorf tube, and
add 400 μL of dichloromethane and 400 μL of MiliQ water to the
aqueous extract. Repeat the centrifugation for 5 min at 4 °C and
13,000 × g. The methanolic/aqueous phase is filtered into certified
dark vials using a 0.22 PVDF filter. Prepare a dark vial containing a
pooled QC mixture of equal volumes of each grape extract. If the
sample set contains more than 50 samples, it is recommended that
grape extraction be performed in batches using a random sequence
to avoid prolonged maceration of the grape sample in the extrac-
tion solvent. Store extract samples of freeze-dried ground grape
samples at -20 °C and grape samples at -80 °C. Avoid unnecessary
samples or extract thawing cycles throughout the metabolomic
workflow.

3.3.3 Build a Sample For data acquisition, injections must follow the previously estab-
Sequence lished random sequence of the sample preparation step. The first
injection of a data batch sample must be a blank, followed by the
injection of 5 pooled QC from the same vial. Begin sample injec-
tions in random order from injection number 6. A pooled QC must
be injected for every 6 sample injections. Add an injection of the
pooled QC at the end of the sample batch.

3.3.4 LC-IMS-MS Start- Prior to initiating data acquisition, clean the source of the IMS-MS
Up and Pretest analyzer (ESI) in accordance with the manufacturer’s instructions.
Load and purge the mobile phases A and B and the wash solutions,
and then prime the mobile phases for 8 min at a ratio of 50% mobile
phase A to 50% mobile phase B. Proceed with the calibration of the
IMS-MS analyzer as described in Subheading 3.2.1. Install the
column, turn on the temperature (40 °C), and gradually bring
the flow rate to 0.3 mL/min. Wait until the pressure difference is
less than 10 psi before starting the data acquisition.
 Advanced LC-IMS-MS Protocol for Holistic Metabolite Analysis in Wine. . . 247

3.4 Data Analysis, The LC-IMS-MS data format is supported by MZmine [23],
Processing, and MS-DIAL [24], MetaboScape, Skyline [4, 25], and Progenesis
Visualization QI [8]. This section describes the analysis and processing of data
using Progenesis QI commercial software [26] (see Note 8).
To demonstrate data analysis processing and visualization, a
small training dataset of 18 runs (including 1 blank, 11 samples,
and 7 pooled QC files) of wine samples (4 Cabernet Sauvignon,
4 Barbera, and 3 Timorasso samples) in ESI+ and ESI- was
acquired using the described UPLC-TWIMS-QToF methodology
(see Note 9). The 36 raw data files (ESI+ and ESI-) are accessible
in the MetaboLights (https://www.ebi.ac.uk/metabolights/
MTBLS8329) open repository [27]. Figure 2 shows a multidimen-
sional visualization of the HDMSE data for a wine sample, includ-
ing the chromatograms at low and high energy applied in the
transfer (Fig. 2a, b) and drift time in milliseconds v/s m/z plot
(Fig. 2c).
For the analysis of the training dataset with Progenesis QI,
initially it is necessary to select the data format and ionization
mode in order to import correctly the raw data into the software
(see Note 10) [26]. To start the processing, a pooled QC must be
selected as an alignment reference (see Note 11). If the CCS
calibration is performed (Subheading 3.2.1), the software

Fig. 2 HDMSE DIA chromatogram of Cabernet Sauvignon wine in ESI- (a) low energy, (b) high energy, and (c)
drift time vs. m/z plot for all features with different charges in the range of m/z 50–2000, with a drift time
range of 1–13 milliseconds in ESI- acquisition mode
248 Vania Sáez et al.

Fig. 3 (a) PCA biplot with no filter applied; (b) PCA biplot with a filter p-value <0.01, max abundance >2 for
HDMSE DIA of Cabernet Sauvignon, Timorasso, and Barbera wine samples acquired in ESI+

automatically converts drift time (ms) to a CCS value in square

ångström (Å2) for each feature. Deconvolution, alignment, and
peak picking (the retention time limits for peak picking for the
protocol are from 0.03 to 19.00 min) are automatically performed
by Progenesis QI. Visually inspect the alignment once processing is
complete.
For comparison, group the run files according to the experi-
mental design groups and conditions in the study. Control the
normalization method and examine the deconvolution of the
adducts. Perform peak picking after defining the groups in the
experimental setup.
Control the quality of the data sample set by grouping all the
groups of the study and the pooled QC acquired. Go to the
compound statistic tab and visually examine whether the pooled
QC group are tightly clustered in the principal component analysis
(PCA) biplot (Fig. 3).
Explore the grouped samples according to experimental groups
or conditions in the PCA biplot. Apply tags and filters such as
p-value and maximum fold change to select the features of interest.
Tag and select the relevant features t (see Note 12). PCA biplot
visualization enables the outlier’s detection.
To perform an orthogonal projection to latent structures dis-
criminant analysis (OPLS-DA), export the results from Progenesis
QI to the EZinfo software. The desired features can be selected and
imported to Progenesis QI from the normalized model.
 Advanced LC-IMS-MS Protocol for Holistic Metabolite Analysis in Wine. . . 249

3.5 Metabolite The strategies for metabolite annotation process of the selected
Annotation features include resources such as in-house libraries (RT, MSE spec-
tra, and TWCCSN2 value) built with analytical standards and man-
aged by UNIFI software (Subheading 3.5.1), spectral public or
published databases/databanks, the Metabolic Profiling CCS
Library, and the curated ion mobility atlas for experimental and
predicted CCS values [10].

3.5.1 In-House Library UNIFI commercial software enables the construction, manage-
ment, and organization of information for the creation of a custo-
mized scientific in-house library for attaining confident annotation
(Rt, m/z, MSE, and CCS value) in real samples.
The first step in creating a customized library is to acquire the
raw data from the reference standard(s). The concentration of
standard compounds can range from 1 ppm to 10 ppm. Import
the HDMSE DIA raw file(s) and create a sample set from the
UNIFI main menu.
The second step is to create the library from the tools tab in the
scientific library section. Each compound in the library is stored as
an “item” and can be created, moved, or copied from the “Manage
Library” section. Each item includes data from ESI+ and ESI-
mode acquisitions. The metabolite structure must be uploaded in
“. mol” format.
The third step is to create an analysis method for the output
HDMSE DIA raw data (select the option to generate a process-only
method for accurate mass screening on IMS data). Through the
menu path Create and Analysis Method, the user can store data in a
library from standard analysis or to analyze samples by comparing a
standard from a library with the selected features. After the creation
of the Analysis Method, a new menu should be displayed including
the tabs: “purpose,” “processing,” “reporting,” and “history.”
Import and filter the ionization technique of the item(s)/com-
pounds from the libraries in the purpose task.
Select the processing parameters; we recommend using a toler-
ance of 0.1 min for the retention time, < 10 ppm for m/z,
and < 2% for the CCS value. The identification of the adducts is
based on the specific setting in the “processing” section, including
the expected adduct and the possible transformations (glucuroni-
dation, glutathione S-conjugation, oxidation, etc.). Configure the
correct LM settings in the “LockMass” tab. Check the calibration
report (Subheading 3.2.1.) for LM and input the reference value
(see Note 5) and the drift time from the calibration report. After
analysis, the settings can be modified for further processing.
The fourth step is to run the analysis method on the sample set
generated in the first step. The processing includes peak and chan-
nel processing, identification, extract ion chromatograms (XIC),
and comparison. The processing step can last from minutes to
250 Vania Sáez et al.

Fig. 4 (a) Component summary of polyphenols identified in a Cabernet Sauvignon wine sample by UNIFI from
raw HDMSE data in ESI-. Four metabolites were annotated at the first level using the in-house library with a Δ
ppm < 5 and Δ CCS% < 1%; (b) HDMSE spectra data for catechin precursor ion; and (c) ion fragments, with
fragment prediction indicated in blue

hours, depending on the resolution mode and type of data acquisi-

tion, sample complexity, the number of target compounds in the
method, and the generation of XIC.
The review of the processed data will reveal the identified,
unobserved, and unknown compounds. The matched compounds
will display the XIC (optional) and high- and low-energy spectral
information, as well as a comparison of retention time, m/z, and
CCS value (Fig. 4). In the high energy spectrum, the fragmentation
match algorithm will indicate accurate m/z fragmentation. To
automatically save the experimental processing data for a standard
or isolated compound in a customized library, select the compound
and send the results, including the fragmentation results, to the
specified library.
The in-house library provides confident annotation for known
compounds of interest. However, UNIFI elucidation tool should
be highlighted to review the fragmentation of known compounds,
as well as shed light on the annotation of unknown compounds or
unmatched compounds. Elucidation tool results are based on the
elemental composition from the adduct and fragment ions (exact
mass and abundance ratio of the isotopic peak) as well as fragmen-
tation prediction. The elucidation tool displays an organized list of
possible candidates associated with the selected libraries.

3.5.2 Reporting an Curated database/libraries, in silico data spectra and CCS predic-
Annotated Feature tion [4, 10, 28], and scoring algorithms are used as resources for
the group of features which are not covered with the in-house
library to support the metabolite annotation.
 Advanced LC-IMS-MS Protocol for Holistic Metabolite Analysis in Wine. . . 251

Progenesis QI can perform a broad screening of the selected

features (Subheading 3.4) using mass spectral database/libraries
from the “identify compounds tab” according to a selected toler-
ance for the m/z precursor and for theoretical fragmentation,
libraries, elemental composition, and isotope similarity score.
To report the annotated features, should be indicated the
confidence levels of annotation for each feature (see Note 13). To
report a CCS value, the drift gas used, the ion separation mecha-
nism, and whether it is a primary or secondary value [3] should be
included in the annotation (see Note 14). Specify the calibrant
employed in the case of a secondary value. Include the Δ ppm for
m/z and the Δ CCS% for each feature in the annotated features
report.

4 Notes

1. Before proceeding, it is strongly advised to read and consult the

material safety data sheets (MSDS) of all reagents, adhere to the
laboratory safety instructions and protocol, and wear protective
equipment. Dichloromethane can cause irritation, is volatile,
and is considered to be carcinogenic. The substances acetoni-
trile, methanol, 2-propanol, and formic acid are harmful, toxic,
and highly flammable. The reagents must be safely handled in a
fume hood.
The hazardous composition of the CCS Major MIX
IMS/ToF solution necessitates preparation and safe handling
in a fume hood.
2. For solvent preparation, it is highly recommended to avoid
materials made of plastic (e.g., tips) and to use glass materials
instead. In order to avoid contamination, adsorption, or carry-
over effects, all labware (syringes, volumetric flasks, graduated
cylinders, etc.) should be cleaned with the solvent at least three
consecutive times before use.
3. CCS Major MIX IMS/ToF solution calibration solution is a
commercial mixture of polyalanine, the commercial mixture
“Ultramark 1621,” low molecular weight organic acids (sal-
ycylic acid, succinic acid, perfluorononanoic acid, perfluorode-
canoic acid, perfluorododecanoic acid, stearic acid, and
theophylline), and the commercial LCMS QC Reference Stan-
dard solution. In negative acquisition mode (ESI-), the mix-
ture has a range from m/z 93.0345 to 1965.9368, and in
positive acquisition mode (ESI+), the range is m/z
152.0706 to 1921.9459.
4. The mass spectrometer calibration protocol is described in
Arapitsas and Mattivi et al. [22] and can be adapted to diverse
mass spectrometers. The parameters to tune the IMS-MS
252 Vania Sáez et al.

analyzer for calibration and HDMSE methodology adhere to

the manufacturer’s recommendations for TWIMS-QToF
instruments. Following the manufacturer’s recommendations,
the described protocol can be transferred to an IMS-MS ana-
lyzer with the same technology or adapted to another IMS-MS
analyzer.
5. The CCS LM report will display both the theoretical m/z value
and drift time, as well as the measured drift time. In the report,
the ESI+ theoretical values for leucine enkephalin are m/z
556.277 [M+H] and a drift time of 6.185 ms, while for ESI-
, the values are m/z 554.262 [M-H] and a drift time of
6.002 ms. The output raw data do not have the LM correction,
for the data analysis is required to perform a LM correction.
Progenesis QI and UNIFI software can correct the multidi-
mensional LC-IMS-MS data [29].
6. The FAIR guidelines for grapevine and wine metabolomics
published by the COST Action CA 17111 INTEGRAPE con-
sortium [30] include helpful tables that describe in detail the
sample collection, sample preparation, methodology, data
transformation, and post-processing. It is highly recommended
that these guidelines be consulted for organizing the metadata
description. The sample description should use the terminol-
ogy, ontology, and standards of the OIV (International Orga-
nization of Vine and Wine).
7. Use the pooled QC to test the factor dilution that will be
applied to the samples for data acquisition. If the sequence
includes more than 100–150 injections, a good strategy could
be the division of sample set into smaller subsample sets with an
equal number of conditions and control treatment samples
[22] using a partial random sequence [21, 30].
8. Data analysis in metabolomics follows a general workflow that
includes steps such as data preprocessing (deconvolution, peak
picking, filtering, and alignment), data pretreatment (normali-
zation, centering, scaling, missing value imputation, and
removal of outliers), data processing (statistical modeling and
data visualization), validation, selection of features as markers,
annotation, and biological interpretation [22]. The processing
of LC-IM-MS data is possible with free or commercial software
using the workflow described above. The software/informatic
tool of choice for data analysis are determined by the research-
er’s experience and available resources. The output raw data
following this protocol is in continuum mode, data-
independent acquisition, without LockMass correction (leu-
cine enkhephalin), and with a “CCS calibration” file. After
calibration (Subheading 3.2.1), each acquired raw data file
contains a “mob_cal.csv” file. Progenesis QI software is
Advanced LC-IMS-MS Protocol for Holistic Metabolite Analysis in Wine. . . 253

appropriate for these LC-IMS-MS raw data because it is capa-

ble of performing the LockMass correction and recognizing
the CCS calibration file during data processing.
9. The small dataset is intended to demonstrate the UniFi and
Progenesis QI software tools for protocol description and is
available to academics for practicing data processing but not for
research purposes.
10. The data acquired in continuum mode is recognized by Pro-
genesis QI as profile data format.
11. The alignment process is performed to compensate the drifts
due variability in the chromatography analysis between runs
during data acquisition. A poor selection of reference data file
can lead to problematic alignment and normalization. The file
used as a reference in the alignment process should be a pooled
QC acquired in the middle of the data sequence; avoid using a
standard mix as QC, a sample file, and not use a blank as a
reference file.
12. The “review compound” tab in Progenesis QI displays the
retention time (min), CCS value, m/z, p-value, and q-value
(FDR-adjusted p-value). When a feature is selected as a bio-
marker, consider the q-value in addition to the p-value.
13. The definition of levels of confidence for annotation was intro-
duced in 2007 by the Metabolomics Standards Initiative
(MSI). The classification is defined by basic requirements or
orthogonal parameters [11, 31, 32]. Initially, the levels ranged
from 1 to 4 [31, 32], but levels from 1 to 5 [32] were proposed
later. In 2017, the compound identification group of the
Metabolomic Society annual meeting [11] introduced level
0 for 5 confidence levels for annotation, ranging from 0 to 4
(level 0: unambiguous 3D structure; level 1: confident 2D
structure; level 2: probable structure; level 4: unknown feature
of interest). The most recent classification includes the use of
the CCS value as orthogonal information for metabolite anno-
tation from LC-IMS-MS data [11].
14. CCS is a derived value of mobility and depends in the ion
structure, buffer gas, temperature, and the ratio between the
electric field (E) and the gas number density (N) [3, 4]. The
experimental CCS values reported in databases and libraries
can be measured using various ion separation mechanism (drift
tube DTCCS, traveling wave TWCCS, and trapped ion
TIM
CCS), with nitrogen serving as the buffer gas for the
majority of measurements. Consider the mechanism of ion
mobility separation and buffer gas for annotations not covered
by an in-house library. If the experimental value is a secondary
value, the calibrant used in the determination must be
known [4].
254 Vania Sáez et al.

Acknowledgments

The research was funded by the ERDF 2014–2020 Program of the

Autonomous Province of Trento (Italy) with EU co-financing
(Fruitomics).

References

1. Cumeras R, Figueras E, Davis CE, Baumbach lipid fingerprinting and identification. Anal
JI, Gràcia I (2015) Review on ion mobility Chem 87(2):1137–1144. https://doi.org/10.
spectrometry. Part 2: hyphenated methods 1021/ac503715v
and effects of experimental parameters. Analyst 8. Paglia G, Astarita G (2017) Metabolomics and
140:1391–1410. https://doi.org/10.1039/ lipidomics using traveling-wave ion mobility
c4an01101e mass spectrometry. Nat Protoc 12:797–813.
2. Lapthorn C, Pullen F, Chowdhry BZ (2012) https://doi.org/10.1038/nprot.2017.013
Ion mobility spectrometry-mass spectrometry 9. Righetti L, Dreolin N, Celma A,
(IMS-MS) of small molecules: separating and McCullagh M, Barknowitz G, Sancho JV, Dall
assigning structures to ions. Mass Spectrom ‘Asta C (2020) Travelling wave ion mobility-
Rev 32(1):43–71. https://doi.org/10.1002/ derived collision cross section for mycotoxins:
mas.21349 investigating interlaboratory and interplatform
3. Gabelica V, Shvartsburg AA, Afonso C, reproducibility. J Agric Food Chem 68(39):
Barran P, Benesch JLP, Bleiholder C, Bowers 10937–10943. https://doi.org/10.1021/acs.
MT, Bilbao A, Bush MF, Campbell JL, Cam- jafc.0c04498
puzano IDG, Causon T, Clowers BH, Creaser 10. Zhou Z, Luo M, Chen X, Yin Y, Xiong X,
CS, De Pauw E, Far J, Fernandez-Lima F, Wang R, Zhu ZJ (2020) Ion mobility collision
Fjeldsted JC, Giles K, Groessl M, Hogan CJ, cross-section atlas for known and unknown
Hann S, Kim HI, Kurulugama RT, May JC, metabolite annotation in untargeted metabo-
McLean JA, Pagel K, Richardson K, Ridgeway lomics. Nat Commun 11. https://doi.org/10.
ME, Rosu F, Sobott F, Thalassinos K, Valen- 1038/s41467-020-18171-8
tine SJ, Wyttenbach T (2019) Recommenda- 11. Blaženović I, Kind T, Ji J, Fiehn O (2018)
tions for reporting ion mobility mass Software tools and approaches for compound
spectrometry measurements. Mass Spectrom identification of LC-MS/MS data in metabo-
Rev 38(3):291–320. https://doi.org/10. lomics. Meta 8:31. https://doi.org/10.3390/
1002/mas.21585 metabo8020031
4. Paglia G, Smith AJ, Astarita G (2022) Ion 12. Diamantidou D, Zotou A, Theodoridis G
mobility mass spectrometry in the omics era: (2018) Wine and grape marc spirits metabolo-
challenges and opportunities for metabolomics mics. Metabolomics 14. https://doi.org/10.
and lipidomics. Mass Spectrom Rev 41:722– 1007/S11306-018-1458-1
765. https://doi.org/10.1002/mas.21686
13. Arapitsas P, Ugliano M, Marangon M,
5. May JC, McLean JA (2015) Ion mobility-mass Piombino P, Rolle L, Gerbi V, Versari A, Mat-
spectrometry: time-dispersive instrumentation. tivi F (2020) Use of untargeted liquid
Anal Chem 87(3):1422–1436. https://doi. chromatography-mass spectrometry metabo-
org/10.1021/ac504720m lome to discriminate italian monovarietal red
6. Richardson K, Giles K (2021) Chapter 4. wines, produced in their different terroirs. J
Travelling wave ion mobility separation: basics Agric Food Chem 68(47):13353–13366.
and calibration. In: Ion mobility-mass spec- https://doi.org/10.1021/acs.jafc.0c00879
trometry, fundamentals and applications, pp 14. Sáez V, Schober D, González Á, Arapitsas P
8 3 – 1 0 4 . h t t p s : // d o i . o r g / 1 0 . 1 0 3 9 / (2021) LC–MS-based metabolomics discrimi-
9781839162886-00083 nates premium from standard chilean
7. Paglia G, Angel P, Williams JP, Richardson K, cv. Cabernet sauvignon wines from different
Olivos HJ, Thompson JW, Menikarachchi L, valleys. Meta:11. https://doi.org/10.3390/
Lai S, Walsh C, Moseley A, Plumb RS, Grant metabo11120829
DF, Palsson BO, Langridge J, Geromanos S, 15. Arapitsas P, Della CA, Gika H, Narduzzi L,
Astarita G (2015) Ion mobility-derived colli- Mattivi F, Theodoridis G (2016) Studying the
sion cross section as an additional measure for effect of storage conditions on the metabolite
 Advanced LC-IMS-MS Protocol for Holistic Metabolite Analysis in Wine. . . 255

content of red wine using HILIC LC-MS https://doi.org/10.1007/978-1-4939-7643-

based metabolomics. Food Chem 197. 0_16
https://doi.org/10.1016/j.foodchem.2015. 23. Schmid R, Heuckeroth S, Korf A, Smirnov A,
09.084 Myers O, Dyrlund TS, Bushuiev R, Murray KJ,
16. Tzachristas A, Dasenaki ME, Aalizadeh R, Hoffmann N, Lu M, Sarvepalli A, Zhang Z,
Thomaidis NS, Proestos C (2021) Develop- Fleischauer M, Dührkop K, Wesner M, Hoog-
ment of a wine metabolomics approach for stra SJ, Rudt E, Mokshyna O, Brungs C,
the authenticity assessment of selected greek Ponomarov K, Mutabdžija L, Damiani T, Pud-
red wines. Molecules 26. https://doi.org/10. ney CJ, Earll M, Helmer PO, Fallon TR,
3390/molecules26102837 Schulze T, Rivas-Ubach A, Bilbao A,
17. Causon TJ, Ivanova-Petropulos V, Richter H, Nothias LF, Wang M, Orešič M,
Petrusheva D, Bogeva E, Hann S (2019) Fin- Weng JK, Böcker S, Jeibmann A, Hayen H,
gerprinting of traditionally produced red wines Karst U, Dorrestein PC, Petras D, Du X, Plus-
using liquid chromatography combined with kal T (2023) Integrative analysis of multimodal
drift tube ion mobility-mass spectrometry. mass spectrometry data in MZmine 3. Nat Bio-
Anal Chim Acta 1052:179–189. https://doi. technol 41(4):447–449. https://doi.org/10.
org/10.1016/j.aca.2018.11.040 1038/s41587-023-01690-2
18. Wang Z, Chen X, Liu Q, Zhang L, Liu S, Su Y, 24. Tsugawa H, Ikeda K, Takahashi M, Satoh A,
Ren Y, Yuan C (2023) Untargeted metabolo- Mori Y, Uchino H, Okahashi N, Yamada Y,
mics analysis based on LC-IM-QTOF-MS for Tada I, Bonini P, Higashi Y, Okazaki Y,
discriminating geographical origin and vintage Zhou Z, Zhu ZJ, Koelmel J, Cajka T,
of Chinese red wine. Food Res Int 165. Fiehn O, Saito K, Arita M, Arita M (2020) A
https://doi.org/10.1016/j.foodres.2023. lipidome atlas in MS-DIAL 4. Nat Biotechnol
112547 38:1159–1163. https://doi.org/10.1038/
19. https://support.waters.com/KB_Inst/Mass_ s41587-020-0531-2
Spectrometry/WKB22696_What_is_the_sam 25. Kirkwood KI, Pratt BS, Shulman N, Tamura K,
ple_preparation_procedure_for_Major_Mix_ MacCoss MJ, MacLean BX, Baker ES (2022)
solution_used_on_Vion_IMS_QTof. Accessed Utilizing skyline to analyze lipidomics data
25.07.2023 containing liquid chromatography, ion mobil-
20. Dudzik D, Barbas-Bernardos C, Garcı́a A, Bar- ity spectrometry and mass spectrometry
bas C (2018) Quality assurance procedures for dimensions. Nat Protoc 17:2415–2430.
mass spectrometry untargeted metabolomics. https://doi.org/10.1038/s41596-022-
A review. J Pharm Biomed Anal 147:149– 00714-6
173. https://doi.org/10.1016/j.jpba.2017. 26. https://www.nonlinear.com/progenesis/qi/
07.044 v2.1/user-guide/. Accessed 02.08.2023
21. Ulaszewska MM, Weinert CH, Trimigno A, 27. Haug K, Cochrane K, Nainala VC, Williams M,
Portmann R, Andres Lacueva C, Chang J, Jayaseelan KV, O’Donovan C (2020)
Badertscher R, Brennan L, Brunius C, Bub A, MetaboLights: a resource evolving in response
Capozzi F, Cialiè Rosso M, Cordero CE, to the needs of its scientific community.
Daniel H, Durand S, Egert B, Ferrario PG, Nucleic Acids Res 48. https://doi.org/10.
Feskens EJM, Franceschi P, Garcia-Aloy M, 1093/nar/gkz1019
Giacomoni F, Giesbertz P, González- 28. Cai Y, Zhou Z, Zhu ZJ (2023) Advanced ana-
Domı́nguez R, Hanhineva K, Hemeryck LY, lytical and informatic strategies for metabolite
Kopka J, Kulling SE, Llorach R, Manach C, annotation in untargeted metabolomics. TrAC
Mattivi F, Migné C, Münger LH, Ott B, Trends Anal Chem 158. https://doi.org/10.
Picone G, Pimentel G, Pujos-Guillot E, 1016/j.trac.2022.116903
Riccadonna S, Rist MJ, Rombouts C, 2 9. https://suppor t.waters.com/ KB_Inf/
Rubert J, Skurk T, Sri Harsha PSC, Van MassLynx/WKB27224_What_is_the_m_z_
Meulebroek L, Vanhaecke L, Vázquez- of_Leucine_Enkephalin_and_Glu-Fibrinopep
Fresno R, Wishart D, Vergères G (2019) Nutri- tide_B_used_as_a_lockmass_on_Xevo_G2-
metabolomics: an integrative action for meta- XS_or_SYNAPT_G2-Si. Accessed 17.08.2023
bolomic analyses in human nutritional studies. 30. Savoi S, Arapitsas P, Duchêne É,
Mol Nutr Food Res 63. https://doi.org/10. Nikolantonaki M, Ontañón I, Carlin S,
1002/mnfr.201800384 Schwander F, Gougeon RD, Ferreira ACS,
22. Arapitsas P, Mattivi F (2018) LC-MS untar- Theodoridis G, Töpfer R, Vrhovsek U, Adam-
geted protocol for the analysis of wine. In: Blondon AF, Pezzotti M, Mattivi F (2021)
Methods in molecular biology, pp 225–235. Grapevine and wine metabolomics-based
256 Vania Sáez et al.

guidelines for fair data and metadata manage- analysis: chemical analysis working group
ment. Meta 11. https://doi.org/10.3390/ (CAWG) metabolomics standards initiative
metabo11110757 (MSI). Metabolomics 3:211–221. https://
31. Sumner LW, Amberg A, Barrett D, Beale MH, doi.org/10.1007/s11306-007-0082-2
Beger R, Daykin CA, Fan TWM, Fiehn O, 32. Schymanski EL, Jeon J, Gulde R, Fenner K,
Goodacre R, Griffin JL, Hankemeier T, Ruff M, Singer HP, Hollender J (2014) Iden-
Hardy N, Harnly J, Higashi R, Kopka J, Lane tifying small molecules via high resolution mass
AN, Lindon JC, Marriott P, Nicholls AW, Reily spectrometry: communicating confidence.
MD, Thaden JJ, Viant MR (2007) Proposed Environ Sci Technol 48(4):2097–2098.
minimum reporting standards for chemical https://doi.org/10.1021/es5002105
 Chapter 14

Analysis of Urinary Metanephrines Using Liquid

Chromatography Tandem Mass Spectrometry
Marlene N. Thaitumu and Elizabeth L. Frank

Abstract
Metanephrines (metanephrine [MN] and normetanephrine [NMN]) are O-methylated metabolites derived
from the catecholamines, epinephrine, and norepinephrine, respectively. High concentrations of metane-
phrines have been observed in individuals with pheochromocytoma, a neuroendocrine tumor. Measure-
ment of metanephrines in urine is used to screen for the tumor. Analysis using liquid chromatography-
tandem mass spectrometry (LC-MS/MS) is recommended due to the high sensitivity, specificity, and
throughput of the technique. Herein, we describe an optimized LC-MS/MS assay for the analysis of
urinary metanephrines.

Key words Liquid chromatography-tandem mass spectrometry, Metanephrines, Pheochromocytoma

1 Introduction

Pheochromocytoma (PHEO) is a rare neuroendocrine tumor that

forms in the chromaffin cells of the adrenal medulla. High concen-
trations of metanephrines have been observed in individuals with
PHEO. It is challenging to diagnose the tumor given the
non-specificity of its symptoms, which include sweating, panic
attacks, and headaches. Chronic production of metanephrines due
to lack of treatment has been associated with hypertension and
cardiovascular disease, which can be fatal [1–4] and emphasizes
the need for accurate and precise analytical assays for measurement
of these metabolites.
Metanephrines are O-methylated metabolites derived from
catecholamines by catechol-O-methyltransferase (COMT) cataly-
sis. MN is derived from epinephrine and NMN from norepineph-
rine (see Fig. 1). It has been reported that plasma and urinary
metanephrines are the benchmark for PHEO screening given
their high sensitivity as biomarkers [2, 5–7]. This is because meta-
nephrines continuously enter the circulation in comparison to

Olga Deda et al. (eds.), Metabolic Profiling: Methods and Protocols, Methods in Molecular Biology, vol. 2891,
https://doi.org/10.1007/978-1-0716-4334-1_14,
© The Author(s), under exclusive license to Springer Science+Business Media, LLC, part of Springer Nature 2025

257
258 Marlene N. Thaitumu and Elizabeth L. Frank

Fig. 1 Metabolism of NMN and MN from norepinephrine and epinephrine by

COMT [19]

catecholamines, which are released episodically [5]. If PHEO is

suspected due to high metanephrines concentrations, imaging or
histopathology of the tumor may be performed for
confirmation [8].
Metanephrines in circulation exist as free or sulfate-conjugated
metabolites. The enzyme sulfate transferase 1A3 (SULT1A3),
found in the gastrointestinal tract, acts on the free metanephrines
to form conjugates that are excreted renally. Before analysis, the
conjugated metanephrines are typically converted to the free form
by acid hydrolysis or enzyme catalysis [9–12]. It has been reported
that free metanephrines are rapidly removed from circulation in
comparison with conjugated metanephrines and are, therefore, in
lower concentrations [13]. However, Grouzmann et al. reported
that there was no difference in free or deconjugated plasma meta-
nephrines in screening for PHEO [14].
Free metanephrines are usually measured in plasma, while total
(free and deconjugated) metanephrines are assayed in urine
[15, 16]. Although deconjugation of metanephrines makes the
urinary metanephrines assay less specific, it is still highly recom-
mended for multiple reasons. Metanephrines concentrations are
higher in urine, which makes the assay less challenging, and collec-
tion of urine is noninvasive, making it more convenient and less
costly. Additionally, the assay can be used as a confirmatory test
given the low prevalence of the disease (<1:100,000 occurrence),
thus avoiding possible false positives of a plasma free metanephrines
assay [9, 17] (see Note 1).
Analysis of urinary metanephrines has previously been con-
ducted using assays such as enzyme linked immunosorbent assay
(ELISA), gas chromatography-mass spectrometry (GC-MS), and
high-performance liquid chromatography with electrochemical
detection (HPLC-ECD). However, liquid chromatography tan-
dem mass spectrometry (LC-MS/MS) is preferred for higher selec-
tivity and throughput, improved sensitivity, and shorter sample
preparation time [1, 4–6, 18].
 Analysis of Urinary Metanephrines Using Liquid Chromatography Tandem Mass. . . 259

An in-use urinary metanephrines LC-MS/MS assay was opti-

mized and modified for higher accuracy and efficiency. In this
chapter, we describe analysis of urinary metanephrines (MN and
NMN) using the optimized protocol [19].

2 Materials

Solvents should be of LC-MS/MS analytical grade and standards of

a similar grade or higher. The materials section was referenced from
Zlatuse Clark et al. [19] and Mass Spectrometry Laboratory, ARUP
Laboratories, Inc., Salt Lake City, UT, USA.

2.1 Samples Timed (24-h) or random urine. Timed specimens should be refri-
gerated during collection. Specimens are stable for 2 weeks refri-
gerated or 1 month frozen at or below
20  C (see Note 2).

2.2 Reagents and Regarding preparation volumes, see Note 3.

Solutions
1. Solid phase extraction (SPE) elution solvent (2% formic acid,
5% MeOH in deionized water): Using a pipette, add 49 mL
deionized water into a 100 mL glass bottle. Add 1 mL formic
acid and 2.5 mL methanol (MeOH). Cap the bottle and mix by
inversion. The elution solvent should be prepared before
each run.
2. SPE wash solvent (methanol 70:30 (v/v)): Using a graduated
cylinder, measure 700 mL MeOH and 300 mL deionized water
and add to a 1 L glass bottle. Let stand at room temperature for
at least 2 hours. The solution is stable for 1 year at room
temperature (20–25  C).
3. Sodium phosphate buffer, 0.4 M: Separately weigh 18.93 g
sodium monophosphate and 9.2 g sodium phosphate (dibasic)
(Sigma-Aldrich, St. Louis, Missouri, USA) and add both to a
500 mL volumetric flask containing 400 mL deionized water.
Bring the solution to volume with deionized water. Adjust to
pH 7  0.1 using 6 M NaOH. Stir the solution on a plate mixer
for 10 min. Allow to equilibrate for 2 h at room temperature,
then transfer the solution to a labeled glass bottle. The solution
is stable for 1 year at room temperature (20–25  C).
4. Mobile phase A—0.2% formic acid in water: Using a graduated
cylinder, measure 1 L of deionized water and add to a labeled
glass bottle. Using a pipette, add 2 mL of formic acid. Cap the
bottle and mix by inversion. The mobile phase is stable for
2 days at room temperature (20–25  C) or 2 weeks refrigerated
(2–8  C).
5. Mobile phase B—0.2% formic acid in methanol: Using a
graduated cylinder measure 1 L of methanol and add to a
260 Marlene N. Thaitumu and Elizabeth L. Frank

labeled glass bottle. Using a pipette, add 2 mL of formic acid.

Cap the bottle and mix by inversion. The mobile phase is stable
for 5 days at room temperature (20–25  C).
6. Needle 1—Methanol:water, 80:20 (v/v): Using a graduated
cylinder, measure 800 mL of MeOH and add to a labeled glass
bottle; measure separately 200 mL of deionized water. Cap and
mix by inversion.
7. Needle wash 2—Formic acid (0.2%) in deionized water: Using
a graduated cylinder, measure 500 mL deionized water and add
to a labeled glass bottle. Using a pipette, measure and add 1 mL
of formic acid. Cap and mix by inversion.

2.3 Standards and Regarding preparation volumes, see Note 3.

Controls

2.3.1 Standards: 0.5 mM 1. Prepare 0.1 L of 0.5 mM combined stock solution of MN (()
MN and NMN Combined metanephrine hydrochloride—Toronto Research Chemicals,
Stock Solution North York, Ontario, Canada) and NMN (DL-normetanephr-
ine hydrochloride—Sigma-Aldrich, St. Louis, Missouri, USA).
The metanephrine standards are in the form of hydrochloride
salts. Prepare stock solutions using information from Table 1.
2. Determine the metanephrines weights for a 0.5 mmoL solu-
tion using the following formulas:
(a) Weight ¼ Concentration ðmMÞ  Volume ðLÞ 
mg
Molecularweight mmoL =Purity mg
233:69
(b) MN : Weight ¼ 0:500 mM  0:100 L  mmoL ¼ 11:92 mg
0:98
mg
219:67
(c) NMN : Weight ¼ 0:500 mM  0:100 L  mmoL ¼ 11:21 mg
0:98
3. Using an analytical balance accurate to 5 decimal places, accu-
rately weigh MN and NMN in separate 10 mL glass screw cap
tubes.
4. Transfer both standards to the same 100 mL volumetric flask
by dissolving and washing out the glass tubes with the equiva-
lent of three tube volumes using deionized water to ensure
complete transfer.
5. Using a volumetric pipette, add 10 mL of 0.05 mol/L hydro-
chloric acid (HCL) prepared in house to the flask. Final HCL
concentration in the stock is 5 mM.

Table 1
Values for calculation of weights for a combined MN and NMN 0.5 mM stock standard

Compound Concentration (mM) Vol (L) HCL salt MW (g/mol) Purity

() metanephrine HCl 0.5 0.1 233.69 0.98
DL-normetanephrine HCl 0.5 0.1 219.67 0.98
 Analysis of Urinary Metanephrines Using Liquid Chromatography Tandem Mass. . . 261

Table 2
Volume of 0.5 mM stock standard (MN and NMN) solution required for
250 mL calibrators preparation

Calibrators (nmol/L) Volume of 0.5 mM stock required (μL)

25 12.5
100 50
500 250
2000 1000
7000 3500

6. Bring the stock standard solution to 100 mL volume with

deionized water. Mix the stock solution thoroughly by invert-
ing the volumetric flask (capped with a stopper) three times.
7. The final concentration of the stock solution for both metane-
phrines is 0.5 mM.

2.3.2 Standards: 1. Dilute the stock to five combined MN and NMN calibration
Combined MN and NMN standards: 25, 100, 500, 2000, and 7000 nmol/L using the
Calibrator Standards (25, information in Table 2.
100, 500, 2000, and 2. Using a graduated cylinder, add 200 mL of deionized water to
7000 nmol/L) five 250 mL volumetric flasks correctly labeled with the calibra-
tor concentration.
2. Using a volumetric pipette, add 25 mL of 0.05 mol/L HCL
(prepared in house) to the flasks. Final concentration of HCL
in each flask is 5 mM.
3. Using a pipette, add the appropriate volume of the combined
stock solution for each calibrator per information in Table 2.
Bring the calibrators to 250 mL volume with deionized water.
4. Mix the calibrator solutions thoroughly by inverting the volu-
metric flask (capped with a stopper) three times (see Note 4).

2.3.3 Combined Internal 1. Prepare combined 12.5 μM MN (rac-Metanephrine-d3·HCl

Standards (ISTD) Working (α1, β2)—Cambridge Isotope Laboratories Inc., Andover,
Solution Massachusetts, USA) and NMN (rac-Normetanephrine-
d3·HCl (α1, β2)—Medical Isotopes Inc., Pelham, New Hamp-
shire, USA) internal standard. The metanephrine ISTDs are in
the form of hydrochloride salts. See Table 3 for values required
for calculations of weights. Determine the metanephrines
ISTD weights for 500 mL of 12.5 μM combined ISTD work-
ing solution using the following calculations:
(a) Weight ¼ Concentration ðmMÞ  Volume ðLÞ
g
Molecular Weight moL =Purity:
262 Marlene N. Thaitumu and Elizabeth L. Frank

Table 3
Information for preparation of a combined MN and NMN 0.5 mM ISTD stock solution

Compound Conc (mM) Vol (L) ISTD HCL salt MW (g/moL) Purity
rac-Metanephrine-d3·HCl (α1, β2) 0.0125 0.5 236.71 0.98
rac-Normetanephrine-d3·HCl (α1, β2) 0.0125 0.5 222.68 0.98

mg
236:71
(b) MN
d3 weight ¼ 0:125 mM  0:5 L  mmoL ¼ 1:51 mg
0:98
mg
222:68
(c) NMN
d3 weight ¼ 0:0125 mM  0:5 L  mmoL ¼ 1:42 mg
0:98
2. Using an analytical balance accurate to 5 decimal places, accu-
rately weigh MN and NMN ISTDs in a separate 10 mL glass
screw cap tubes.
3. Transfer both MN and NMN ISTDs to the same 0.5 L volu-
metric flask by dissolving and washing out the glass tubes with
the equivalent of three tube volumes using deionized water to
ensure complete transfer.
4. Using a graduated cylinder, add 50 mL of 0.05 mol/L HCL
solution (prepared in house) to the volumetric flask (final
concentration of HCL in solution is 5 mM).
5. Bring the combined ISTD working solution to 0.5 L volume
with deionized water. Mix the ISTD working solutions thor-
oughly by inverting the volumetric flask (capped with a stop-
per) three times (see Note 5).

2.3.4 Quality 1. For Lyphochek Quantitative Urine Controls I and II (Bio-Rad

Controls (QC) Clinical Division, Hercules, California, USA) preparation, fol-
low manufacturer instructions.
2. The negative control is deionized water.

2.4 Equipment and 1. Masterblock Polypropylene, 96 Deep-Well, 1.2 mL.

Supplies 2. Collection Plate, 96-well, 2 mL, Square/Round Conic.
2.4.1 Supplies 3. Single Well Waste Plate, 96-well, 300 mL Pyramid Bottom.
4. Solid Phase Extraction (SPE) Plate, Plate RCH PWCX 1 mL
96-well, 20 mg (Tecan, San Jose, California, USA).
5. 10 mL glass tubes.
6. Graduated cylinders.
7. Volumetric flasks.
8. Glass bottles.
9. Pipettes.
 Analysis of Urinary Metanephrines Using Liquid Chromatography Tandem Mass. . . 263

10. Pipette tips.

11. 10 mL plastic tubes.
12. Multicolored 1.5 mL microcentrifuge tubes.

2.4.2 Equipment 1. SPEware Corp. 96-well positive pressure extraction manifold

(SPEware Corp., Baldwin Park, California, USA).
2. High speed centrifuge.
3. Analytical balance.
4. Aluminum dry heat block (VWR, Radnor, Pennsylvania, USA).
5. Tandem Mass Spectrometer: AB/Sciex API 3200; Ion Source:
AB/Sciex TurboIonspray® interface; Software; Analyst
(SCIEX, Toronto, Canada).
6. Agilent HPLC system with a Leap Technologies Pal autosam-
pler, 1200 series pump (Agilent, Santa Clara, California, USA).
7. Phenomenex Kinetex pentafluorophenyl propyl (PFP) column;
2.1 mM  50 mM, 5 μm particles (Phenomenex, Torrance,
California, USA).

3 Methods

3.1 Sample 1. Add 50 μL of the working internal standard to 250 μL of urine

Extraction sample, calibrators, controls, and blanks.
2. Acidify the samples by adding 30 μL of 6 mol/L HCL.
3. Incubate the samples in an aluminum dry heat block at 90  C
for 15 min.
4. Centrifuge the samples at a cold temperature (5  C) at 2000 g
for 5 min.
5. Add 0.5 mL of 0.4 M sodium phosphate buffer (pH 7) and
30 μL of 6 mol/L NaOH to neutralize the samples.
6. On a 96-well positive pressure extraction manifold, condition
the RCH weak cation exchange (PWCX) SPE plate with
1.0 mL methanol and equilibrate with 1.0 mL of deionized
water.
7. Briefly vortex the samples and transfer to the conditioned and
equilibrated RCH PWCX SPE plate.
8. Wash the samples with 1.8 mL deionized water followed by an
equivalent amount of 70% MeOH.
9. Elute the metanephrines in 500 μL of 2% formic acid in 5%
methanol.
10. Inject 15 μL onto the LC column.
264 Marlene N. Thaitumu and Elizabeth L. Frank

Table 4
LC gradient for separation of urinary metanephrines

Time (min) Mobile phase A (%) Mobile phase B (%)

0.0 95.0 5.0
1.5 95.0 5.0
2.5 50.0 50.0
2.8 50.0 50.0
3.0 95.0 5.0
4.5 95.0 5.0

3.2 LC Method 1. Mobile phase A—0.2% formic acid in water.

2. Mobile phase B—0.2% formic acid in methanol.
3. The injection volume is 15 μL; flow rate 500 μL/min; column
temperature 30  C.
4. Mobile phase gradient—see Table 4.
5. See Note 6.

3.3 MS Method 1. Multiple reaction monitoring (MRM) is used in positive mode;

a quantifier (Quant) peak is used for quantification and quali-
fier (Qual) peak for confirmation.
2. Other MS parameters: collision energy (25 V), entrance poten-
tial (11 V), nitrogen gas for nebulizing and auxiliary (517 kPa,
75 psi), curtain (172 kPa, 25 psi), collision (34.5 kPa, 5 psi),
interface heater temperature (600  C), and capillary voltage
(1500 V).
3. Transitions and declustering potentials (DP) are given in
Table 5. MS parameters must be optimized for each analyte
and instrument.

3.4 Data Analysis 1. Data is analyzed using appropriate software, e.g., SCIEX Ana-
lyst software.
2. Calibration for each analyte with 1/x weighting should pro-
duce a correlation coefficient > 0.99.
3. Evaluate chromatography of calibrators and controls for reten-
tion time, peak shape, IS area counts, and ion ratios (see Fig. 2,
from Clarke and Frank [19]). Minimum IS peak area counts
must be established during method validation.
4. QC concentrations must fall within 2 standard deviations of
the set mean (see Note 7).
5. Data can be analyzed using an alternative set of transitions for
MN, if necessary due to interference (see Note 8).
 Analysis of Urinary Metanephrines Using Liquid Chromatography Tandem Mass. . . 265

Table 5
Metanephrines MRMs and DPs

Compound Precursor (m/z) Product (m/z) Transition DP (V)

MN 180 165 Quant 42
MN 180 120 Qual 42
MN-d3 183 168 Quant 42
MN-d3 183 123 Qual 42
NMN 166 134 Quant 33
NMN 166 106 Qual 33
NMN-d3 169 137 Quant 33
NMN-d3 169 109 Qual 33

4 Notes

1. Both plasma and urinary metanephrines assays present advan-

tages and challenges. Plasma free metanephrines have lower
concentrations, but their analysis is more specific compared to
urinary total metanephrines. For this reason, it is recom-
mended to have both assays performed to give the clinician
detailed and confirmatory information on the patients’ condi-
tion before proceeding to more expensive confirmatory tests.
2. Drugs which are structurally similar to metanephrines (e.g.,
hydralazine, cimetidine, amphetamine, levodopa, and isoethar-
ine among others) can possibly cause interference; therefore, it
is important for the patient to avoid the medications before
urine collection.
3. The reagents and solutions volumes should be tailored to the
demands and specifics of the assay volume, turnaround time,
storage capacity and conditions, and throughput among other
laboratory factors.
4. Calibrators can be aliquoted into multicolored microcentrifuge
tubes (to identify each concentration) and stored frozen
(
60  C to
90  C) for up to 1 year.
5. ISTD working solution can be aliquoted into 10 mL plastic
tubes and stored frozen for at least 3 months at
70  C.
6. For enhanced specificity, consider optimizing LC baseline sep-
aration with other catecholamines (dopamine, norepinephrine,
and epinephrine) and its metabolites such as
3-methoxytyramine as they are structurally similar and can
cause interference. This is important for epinephrine and
NMN which are isobaric (MW ¼ 183.2 g/moL) [19].
Fig. 2 Urinary metanephrines chromatography. (a) a 200 nmol/L calibrator, (b) a healthy patient urine sample,
(c) an abnormal patient urine sample. Diagram referenced from Clark and Frank [19]. Reprinted from
Publication title, 879/31, Zlatuse D. Clark and Elizabeth L. Frank, Urinary metanephrines by liquid chromatog-
raphy tandem mass spectrometry: Using multiple quantification methods to minimize interferences in a high
throughput method, 3673–3680., Copyright (2023), with permission from Elsevier
 Analysis of Urinary Metanephrines Using Liquid Chromatography Tandem Mass. . . 267

7. In addition to the acceptance criteria mentioned in Subheading

3.4, Quant and Qual peaks can be used to analyze accuracy
(interference) [19].
(a) The qualitative ion ratio (QIR) is defined as the ratio of
the analyte concentration of the Quant peak to the analyte
concentration of the Qual peak.
(b) The acceptance criterion for QIR is 30% of the QIR of
the calibration standards.
8. MN can be assessed using an alternate transition if interference
is present. Alternate Quant transitions are 180/148 for MN
and 183/151 for MN-d3 [19].

References
1. Joyce CM, Melvin A, O’Shea PM, Costelloe recommendations for clinical practice from
SJ, O’Halloran DJ (2020) Case report of a the first international symposium. Nat Clin
phantom pheochromocytoma. Biochem Med Pract Endocrinol Metab 3(2):92–102
30(2):325–330 9. Pamporaki C, Därr R, Bursztyn M, Glöckner S,
2. Zhang W, Li X, Li W, Zhang Y, Cai J, Feng S, Bornstein SR, Lenders JWM, Pacak K,
Sun Z (2024) Clinical diagnosis of pheochro- Krinner A, Eisenhofer G (2013) Plasma-free
mocytoma and paraganglioma-induced sec- vs deconjugated metanephrines for diagnosis
ondary hypertension through UPLC-MS/MS of phaeochromocytoma. Clin Endocrinol
analysis of plasma catecholamines and their 79(4):476–483
metabolites. J Clin Hypertens 26(4):416–424 10. Eisenhofer G, Prejbisz A, Peitzsch M,
3. Jain A, Baracco R, Kapur G (2019) Pheochro- Pamporaki C, Masjkur J, Rogowski-Lehmann-
mocytoma and paraganglioma—an update on N, Langton K, Tsourdi E, Pe˛czkowska M,
diagnosis, evaluation, and management. Fliedner S et al (2018) Biochemical diagnosis
Pediatr Nephrol 35(4):581–594 of chromaffin cell tumors in patients at high
4. Ceccato F, Mantero F (2019) Monogenic and low risk of disease: plasma versus urinary
forms of hypertension. Endocrinol Metab free or deconjugated O-methylated catechol-
Clin N Am 48(4):795–810 amine metabolites. Clin Chem 64(11):
5. Lenders JWM, Duh Q-Y, Eisenhofer G, 1646–1656
Gimenez-Roqueplo A-P, Grebe SKG, Murad 11. Eisenhofer G, Åneman A, Hooper D,
MH, Naruse M, Pacak K, Young WF (2014) Rundqvist B, Friberg P (2002) Mesenteric
Pheochromocytoma and paraganglioma: an organ production, hepatic metabolism, and
endocrine society clinical practice guideline. J renal elimination of norepinephrine and its
Clin Endocrinol Metabol 99(6):1915–1942 metabolites in humans. J Neurochem 66(4):
6. Eisenhofer G, Peitzsch M, McWhinney BC 1565–1573
(2016) Impact of LC-MS/MS on the labora- 12. Eisenhofer G, Coughteree MW, Goldstein DS
tory diagnosis of catecholamine-producing (1999) Dopamine sulphate: an enigma
tumors. TrAC Trends Anal Chem 84 resolved. Clin Exp Pharmacol Physiol Suppl
(PB):106–116 26:S41–S53
7. Kim HJ, Lee JI, Cho YY, Lee SY, Kim JH, Jung 13. Eisenhofer G, Lenders J (2012) Rapid circula-
BC, Kim SW, Chung JH, Min Y-K, Lee M-S tory clearances and half-lives of plasma free
et al (2015) Diagnostic accuracy of plasma free metanephrines. Clin Endocrinol 77(3):
metanephrines in a seated position compared 484–485
with 24-hour urinary metanephrines in the 14. Grouzmann E, Drouard-Troalen L, Baudin E,
investigation of pheochromocytoma. Endocr J Pierre-François P, Müller B, Grand D, Buclin T
62(3):243–250 (2010) Diagnostic accuracy of free and total
8. Pacak K, Eisenhofer G, Ahlman H, Bornstein metanephrines in plasma and fractionated
SR, Gimenez-Roqueplo A-P, Grossman AB, metanephrines in urine of patients with pheo-
Kimura N, Mannelli M, McNicol AM, Tischler chromocytoma. Eur J Endocrinol 162(5):
AS (2007) Pheochromocytoma: 951–960
268 Marlene N. Thaitumu and Elizabeth L. Frank

15. Jeyaraman K, Natarajan V, Nicolai T, Jacob P, pheochromocytoma: measurement of fractio-

Nair A, Shanthly N, Oommen R, Varghese G, nated plasma metanephrines compared with
Joseph FJ, Seshadri MS et al (2013) The role of the combination of 24-hour urinary metane-
urinary fractionated metanephrines in the diag- phrines and catecholamines. J Clin Endocrinol
nosis of phaeochromocytoma. Indian J Med Metabol 88(2):553–558
Res 137(2):316–323 18. Grebe SK, Singh RJ (2011) LC-MS/MS in the
16. van Duinen N, Corssmit EPM, Jong BD, Kema clinical laboratory—where to from here? Clin
IP, Romijn JA (2013) Plasma levels of free Biochem Rev 32(1):5–31
metanephrines and 3-methoxytyramine indi- 19. Clark ZD, Frank ES (2011) Urinary metane-
cate a higher number of biochemically active phrines by liquid chromatography tandem
HNPGL than 24-h urinary excretion rates of mass spectrometry: using multiple quantifica-
catecholamines and metabolites. Eur J Endo- tion methods to minimize interferences in a
crinol 169(3):377–382 high throughput method. J Chromatogr B
17. Sawka AM, Jaeschke R, Singh RJ, Young WF Analyt Technol Biomed Life Sci 879(31):
(2003) A comparison of biochemical tests for 3673–3680
 Chapter 15

Two-Phase Extraction for Comprehensive Analysis

of the Plant Metabolome by NMR
Jan Schripsema and Denise Dagnino

Abstract
Metabolomics is the area of research, which strives to obtain complete metabolic fingerprints, to detect
differences between them and to provide hypothesis to explain those differences (Schripsema J, Dagnino D,
Handbook of chemical and biological plant analytical methods. Wiley, New York, 2015). However,
obtaining complete metabolic fingerprints is not an easy task. Metabolite extraction is a key step during
this process, and much research has been devoted to finding the best solvent mixture to extract as much
metabolites as possible.
Here a procedure is described for analysis of both polar and apolar metabolites using a two-phase
extraction system. D2O and CDCl3 are the solvents of choice, and their major advantage is that, for the
identification of the compounds, standard databases can be used because D2O and CDCl3 are the solvents
most commonly used for pure compound NMR spectra. The procedure enables the absolute quantification
of components due to the addition of suitable internal standards. The extracts are also suitable for further
analysis with other systems like LC-MS or GC-MS.

Key words Two-phase extraction, NMR, Metabolic fingerprints, Plants, Identification,

Quantification

1 Introduction

Good planning is essential in any research, more even in metabo-

lomic experiments due to the large numbers of samples typically
analyzed. Every step should be made clear, from the question(s) to
be answered to the methods and analytical techniques selected to
achieve the goal. The whole procedure should be able to extract a
maximum of information in a time and cost-effective way. Once
defined, the procedure should be strictly followed in order to allow
the proper comparison of the many samples [1].
NMR and MS, coupled to various chromatographic techni-
ques, are today the main analytical methods used in metabolomics.
Both methods have specific advantages and disadvantages and con-
tinue to develop and find useful applications.

Olga Deda et al. (eds.), Metabolic Profiling: Methods and Protocols, Methods in Molecular Biology, vol. 2891,
https://doi.org/10.1007/978-1-0716-4334-1_15,
© The Author(s), under exclusive license to Springer Science+Business Media, LLC, part of Springer Nature 2025

269
270 Jan Schripsema and Denise Dagnino

Advantages of NMR over MS-based metabolomics are the

reproducibility of NMR spectra and the possibility of direct quanti-
fication [2]. Further, a large data set exists, available in the literature
or specific data banks, containing spectra of pure compounds.
Nevertheless, since the spectra vary according to the solvent used,
direct comparison is only possible if the spectra to be compared
have been obtained in the same solvent.
The extraction of the samples is a critical step in the metabo-
lomics experiment, and much literature has been devoted to this
subject, e.g., [3–5].
Most of the literature concerning NMR-based metabolomics
uses mixtures of solvents in order to maximize the number of
extracted metabolites [6–8]. However, when using this strategy, it
is not possible to compare the spectra obtained directly with the
readily available literature or data bases.
In the present paper, we focus on the extraction of plant
material that takes into account the need to obtain complete meta-
bolic fingerprints of the samples and the qualitative and quantitative
determination of the compounds therein. We demonstrate the
usefulness of a two-phase extraction system using water and chlo-
roform. Both solvents are commonly used for pure compound
NMR identification. Thus, the spectra obtained in metabolomic
experiments can be readily compared with the available literature
for the identification of the compounds therein. The spectra
obtained with the suggested two-phase method are completely
complementary and very rarely a compound appears in both phases.
The separation in two phases diminishes superposition of signals
(Fig. 1) and makes identification and quantification of the com-
pounds in the extracts easier and preciser. It is a quick and extremely
simple extraction procedure that can be carried out in targeted or
untargeted approaches, and if necessary, the extracts can be further
analyzed by LC-MS or GC-MS.
After obtaining the NMR spectra, the next steps in the meta-
bolomics experiment involve the preprocessing of the obtained
experimental results to make them suitable for the subsequent
analysis. Generally, this involves multivariate data analysis, preceded
by the process of binning or bucketing to simplify the alignment of
spectra. To avoid the problems associated with these multivariate
data analysis such as data reduction and suppression of minor
signals, an alternative method was introduced involving similarity
calculations and differential NMR [9]. This methodology permit-
ted the analysis of components over five orders of magnitude
[10]. The intention of both multivariate analysis and differential
NMR is to reveal the differences between sample groups, indicating
the peaks or signals which are specifically increased or decreased in
certain datasets. In the final step of the experiments, these results
should be interpreted and the compounds responsible for the
signals identified.
 Two-Phase Extraction for Comprehensive Analysis of the Plant Metabolome by NMR 271

Fig. 1 1H NMR spectra from the (a) D2O and (b) CDCl3 extracts of coffee powder obtained with the two-phase
extraction. In the CDCl3 phase, signals from caffeine (marked with C) and pyridine (marked with P) are
indicated. In the D2O phase. Signals from TMSP and trigonelline (marked with T) are indicated

In fact, one of the main difficulties in metabolomics is the

discovery of the identity of the compounds detected. Peaks or
signals are found which are specifically increased or decreased in
certain datasets, but linking those peaks to specific compounds
needs a lot of precaution, and frequently additional experiments
are required. For instance, in metabolomics based on LC-MS,
major difficulties arise in the comparison of chromatograms
(because of shifts in retention times) and/or spectra (reproducibil-
ity of fragmentation), when data obtained by different instruments
or research groups are compared and when databases are consulted.
In NMR, the reproducibility is not such a problem, which turns the
comparison of datasets easier.

2 Materials

2.1 Internal A known amount of internal standard solution is added to each of

Standards the phases so as to allow the quantification of the compounds
272 Jan Schripsema and Denise Dagnino

Table 1
Calculations with the NMR data shown in Fig. 1. For each spectrum, the
calculation of the absolute quantity of a specific compound is illustrated

D2O sample CDCl3 sample

Trigonelline Caffeine
9.11 ppm 3.59 ppm
IS 0.00 ppm IS 8.61 ppm
IS solution 47.0 mg/100.0 g 151 mg/99.73 g
Actual weight IS sol. 110 mg 170 mg
Weight 100 uL IS sol 110 mg 150 mg
Actual quant. plant material 105 mg 105 mg
Quant. extraction solvent 875 mg 1206 mg
Quant. extr. Solvent transferred 524 mg 532 mg
Multiplication factor 15.90331 24.46831
Quant. IS in 100 uL IS sol. (mg) 0.0517 mg 0.227 mg
Area compound signal 62.54 252.29
Area IS signal 78.11 159.66
No. Hs IS signal 9 2
No. Hs compound signal 1 3
Mol. Weight compound 137.14 194.19
Mol. Weight IS 172.27 79.1
Abs. Quant. (mg/g) 4.72 14.36

extracted. For the aqueous phase, TMSP [3-(Trimethylsilyl)-

proprionic-2,2,3,3-d4 acid, sodium salt] is used and for the organic
phase pyridine. The TMSP spectrum has only one signal at
0.00 ppm that is suitable for both the quantification and as a
reference for the chemical shift.
For the organic phase, pyridine is suggested since it provides
some advantages. The spectrum of pyridine contains three signals
of which two (at 8.61 ppm and 7.69 ppm) are clearly separated
from other signals. The third signal (7.29 ppm) is close to the
residual solvent signal and is more difficult to integrate. TMS,
often used as chemical shift reference in chloroform, is not advised
for quantification purposes, because it is more volatile and its signal
is quite narrow. Furthermore, its relaxation time is higher than
most other compounds [11].
In metabolomics experiments, many signals are not identified,
but the peak areas should be corrected to permit quantitative
 Two-Phase Extraction for Comprehensive Analysis of the Plant Metabolome by NMR 273

comparison between samples. First of all instead of the areas, the

relative areas in relation to the area of the IS are used (area/area IS
signal). This relative area should then be corrected for eventual
deviations of the actual quantities of the IS added to the tube, the
quantity of plant material extracted, and the quantity of solvent
used to extract the plant material and of the liquid transferred to the
NMR tube. This leads to the following multiplication factor (MF):
MF = ðActual Weight IS sol:=Weight 100 uL IS sol:Þ
× ð1000 mg of plant material=actual Quant:plant materialÞ
× ðQuant:extraction solvent=Quant:extraction solvent transferredÞ:
This factor is the same for all signals in the NMR spectrum of a
sample. In this way direct comparison between samples is possible.
If the compound of interest has well-resolved signals which can
be integrated, the absolute quantity of the compound can be
calculated using the following formula:
Quant:IS in 100 uL IS sol:ðmgÞ
× ðarea compound signal=area IS signalÞ × MF
× ðno:Hs IS signal=no:Hs compound signalÞ
× ðmol:weight compound=mol:weight ISÞ:
This formula provides the absolute quantity per gram of plant
material. These calculations are illustrated in Table 1 for specific
signals in the spectra shown in Fig. 1.

2.2 Preparation of Deuterated solvents should be at least 99.8% deuterated. The inter-
the Internal Standard nal standards should be of analytical grade or higher. Weighing of
Solution (ISS) standards and solvents should be carried out with maximum
precision.
Internal standard Solution 1 (ISS1): Weigh 150 mg of pyridine
(MW: 79.10 g/moL) in an empty bottle and add 100 g
CDCl3.
Internal standard Solution 2 (ISS2): Weigh 50 mg of TMSP (MW:
172.27 g/moL) in an empty bottle and add 100 g D2O.
Internal standard solutions can be kept for up to 1 year at room
temperature, if properly handled. Flasks should be kept tightly
closed when not in use. In case of handling a large number of
samples, distribute the prepared ISS in smaller bottles so as to
minimize the time each flask is open.

3 Methods

3.1 Extraction and The extraction described here has been used for dried plant material
Analysis (freeze dried) and also for the analysis of dried bacteria and food
stuff such as teas, coffee, butter, and cheese. The procedure is
274 Jan Schripsema and Denise Dagnino

Fig. 2 Schematic representation of the two-phase extraction for comprehensive analysis of the plant
metabolome by NMR

schematically presented in Fig. 2. All pipetting should be carried

out on a precision balance and the weights registered (see Note 1).
When pipetting CDCl3 with conventional pipettes, before the
actual pipetting, aspire the liquid a few times to avoid dripping
during pipetting.
1. Weigh 100 mg of dried plant material in a 2 mL Eppendorf
tube (see Note 2 and 3).
2. Add 0.80 mL of D2O to the plant material and take care all
material has been wetted (see Note 4).
3. Add 0.80 mL of CDCl3 to the wetted plant material (see
Note 5).
4. Thoroughly vortex the mixture.
5. Speed the extraction by ultrasonic mixing for 10 min.
6. Centrifuge at 13,000 rpm for 5 min.
7. Take out the Eppendorf tubes carefully from the centrifuge.
8. Add 0.10 mL of ISS1 to an NMR tube.
9. Take out 0.5 mL of the D2O phase (upper phase) from the
Eppendorf tube and transfer it to the same NMR tube.
10. Seal the NMR tube, vortex to mix the extract and the ISS1.
11. Take another NMR tube and add to it 0.10 mL of ISS2.
12. Take out 0.5 mL of the CDCl3 phase (lower phase) with the
pipette adjusted to 0.800 mL and add to the same NMR tube
(see Notes 6, 7, 8, 9, and 10).
13. Seal the NMR tube, vortex to mix the extract and the ISS2.
14. Acquire the 1H NMR spectra (100–128 scans) of the extracts
in both NMR tubes. For D2O, use presaturation of the residual
solvent signal (see Notes 11, 12, 13, 14, and 15).

4 Notes

1. Weighing increases the precision since pipetting is less precise,

especially when chloroform is pipetted using conventional
pipettes.
Two-Phase Extraction for Comprehensive Analysis of the Plant Metabolome by NMR 275

2. The proportion of sample to solvent can be changed depending

on the concentration of the extracted compounds. The amount
of sample chosen should preferably not saturate the solvents.
3. To determine the best proportion of sample to solvent
volumes, re-extract the sample with the same procedure and
determine whether the extract contains around the predicted
value. If not the proportion of sample to solvent can be
adapted.
4. When the material is well wetted before the addition of
CDCL3, the extraction efficiency is increased.
5. CDCl3 is added after the water to avoid evaporation.
6. Plant material stays between the two phases. When taking out
the lower phase, the layer should be carefully traversed with the
tip of the pipette. It is usually possible to displace the pellet
formed between the phases by pushing it with the pipette tip
and thus gain direct access to the lower phase.
7. When taking out the lower layer, chloroform vapors decrease
the actual volume of the liquid. One should therefore adjust
the pipette to 0.800 mL and take out as much of CDCl3 as
possible without taking along any of the aqueous phase.
8. When the other phase enters the pipette, care should be taken
not to transfer this to the NMR tube. The actual quantity
which is transferred is determined by the weight of the liquid
transferred to the NMR tube.
9. If a drop of the CDCl3 extract falls on the balance plate, just
allow it to evaporate (usually very quick) and register the
weight after.
10. Minimum amount of volume in the NMR tube for a good
quality measurement is 0.50–0.60 mL. If for any reason the
volume is insufficient, record the weight of the extract and just
add more deuterated solvent to complete the volume.
11. 100 or 128 scans is generally sufficient to obtain good quality
spectra, but the quantity can be increased (or decreased)
according to the need.
12. After obtaining the spectra, they can be processed manually or
automatically. This involves phasing and calibration. The peak
shape should be verified to check if the sample was correctly
shimmed. In the water samples, the TMSP signal can be
checked for symmetry. In the chloroform samples, the TMS
or residual solvent signal can be used for this purpose. If not
adequate the samples should be remeasured or discarded.
13. In the measurement of the aqueous extracts, presaturation of
the residual solvent signal is used. In our experiments the
Bruker program zgcppr was used. Typical parameters include
276 Jan Schripsema and Denise Dagnino

the relaxation delay of 5.0 s. Acquisition of 64 K data points

and 100 or 128 scans.
14. To analyze the extracts by mass spectrometric methods, it is
necessary to dilute the extract in the deuterated solvents at least
a factor 1000 with non-deuterated solvent. This will also revert
eventual exchange of H with D.
15. The accuracy of the absolute quantifications might be increased
by using a longer relaxation delay (e.g., 30 s, about five times
the relaxation time of the slowest relaxing nuclei); however,
this leads to very long experimental times. Alternatively, a
calibration can be done in relation to a single experiment
with a long relaxation delay.

References
1. Schripsema J, Dagnino D (2015) 7. Beltran A, Suarez M, Rodrı́guez MA et al
Metabolomics. In: Hostettmann K, (2012) Assessment of compatibility between
Stuppner H, Marston A, Chen S (eds) Hand- extraction methods for NMR and LC/MS-
book of chemical and biological plant analytical based metabolomics. Anal Chem 84:5838–
methods, 1st edn. Wiley, New York 5844. https://doi.org/10.1021/ac3005567
2. Schripsema J (2010) Application of NMR in 8. Kim HK, Choi YH, Verpoorte R (2010)
plant metabolomics: techniques, problems and NMR-based metabolomic analysis of plants.
prospects. Phytochem Anal 21:14–21. https:// Nat Protoc 5:536–549. https://doi.org/10.
doi.org/10.1002/pca.1185 1038/nprot.2009.237
3. Deda O, Gika HG, Wilson IA, Theodoridis GA 9. Schripsema J (2019) Similarity and differential
(2015) An overview of fecal preparation for NMR spectroscopy in metabolomics: applica-
global metabolic profiling. J Pharm Biomed tion to the analysis of vegetable oils with H-1
Anal 113:137–150. https://doi.org/10. and C-13 NMR. Metabolomics 15:39.
1016/j.jpba.2015.02.006 https://doi.org/10.1007/s11306-019-
4. Kim H-S, Park SJ, Hyun S-H et al (2011) 1502-9
Biochemical monitoring of black raspberry 10. Schripsema J, Merlim RD, Parvan LG et al
(Rubus coreanus Miquel) fruits according to (2022) Towards a holistic view of tablet quality,
maturation stage by 1H NMR using multiple an extensive study on paracetamol tablets with
solvent systems. Food Res Int 44:1977–1987. nuclear magnetic resonance using similarity
https://doi.org/10.1016/j.foodres.2011. calculations, differential NMR and hierarchical
01.023 cluster analysis. J Pharm Biomed Anal 215:
5. Heyman HM, Meyer JJM (2012) NMR-based 114773. https://doi.org/10.1016/j.jpba.
metabolomics as a quality control tool for 2022.114773
herbal products. S Afr J Bot 82:21–32. 11. Schripsema J (2008) Comprehensive analysis
https://doi.org/10.1016/j.sajb.2012.04.001 of polar and apolar constituents of butter and
6. Kim HK, Verpoorte R (2010) Sample prepara- margarine by nuclear magnetic resonance,
tion for plant metabolomics. Phytochem Anal reflecting quality and production processes. J
21:4–13. https://doi.org/10.1002/pca.1188 Agric Food Chem 56:2547–2552. https://doi.
org/10.1021/jf072274c
 Chapter 16

Olive Fruit Phenolic Profiling Using High Resolution-Magic

Angle Spinning (HR-MAS) Solid-State NMR Spectroscopy
Efstathia Manolopoulou and Apostolos Spyros

Abstract
High Resolution-Magic Angle Spinning (HR-MAS) solid-state NMR spectroscopy is finding increasing
application in the analysis of solid foods, bypassing the need for complicated solvent extraction procedures.
In the present protocol, we report a simple analytical approach based on HR-MAS NMR spectroscopy for
the phenolic profiling of olive fruits, flesh, or skin. This approach allows the facile characterization of
phenolic compounds in olive fruits cultivated for extra-virgin olive oil production as a function of matura-
tion and variety, in addition to processing technology for table olives.

Key words NMR spectroscopy, Solid state, Phenolic profiling, Food analysis, Authentication

1 Introduction

High-Resolution–Magic Angle Spinning (HR-MAS) solid-state

NMR spectroscopy is a powerful analytical spectroscopic technique
that can provide high-resolution NMR spectra of solid amorphous
materials by swelling the sample in minimal amounts of a deuter-
ated solvent (10–40 μL) and rotating it at the magic angle
(θ = 54.7°), using a specialized rotor/probe system designed for
HR-MAS spectroscopy [1, 2]. Apart from swelling, the deuterated
solvent allows NMR spectral acquisition under deuterium lock
conditions, which affords “liquid-like” 1D and 2D NMR spectra
of very high resolution, possessing natural linewidths of less than
1 Hz. This methodology thus allows the extension of the powerful
analytical capabilities of liquid-state NMR, such as chemical com-
position determination and profiling to solid amorphous samples.
As in liquids NMR, compound identification and verification can be
achieved by comparison with literature data and/or publicly avail-
able NMR spectroscopy databases. Unknown compounds can also
be detected and identified by performing more complex standard
1D and 2D NMR experiments prior to quantification, which are all

Olga Deda et al. (eds.), Metabolic Profiling: Methods and Protocols, Methods in Molecular Biology, vol. 2891,
https://doi.org/10.1007/978-1-0716-4334-1_16,
© The Author(s), under exclusive license to Springer Science+Business Media, LLC, part of Springer Nature 2025

277
278 Efstathia Manolopoulou and Apostolos Spyros

available for the HR-MAS platform, as in liquid-state NMR. Due to

its practical advantages, HR-MAS NMR spectroscopy in recent
years has found increasing application to the study of several solid
foods [3–6], including meat [7, 8], cheese [9], fruits [10–13],
plants [14–16], olive leaves [17], etc. Apart from avoiding compli-
cated sample pretreatment steps, such as extraction with organic
solvents or water to concentrate analytes present in very small
amounts, HR-MAS NMR offers the advantage of minimal sample
quantity, as, for example, usually less than 10 mg are needed for
4 mM rotor inserts with 50 μL available volume, while inserts with
even smaller volumes (12 μL) are available, for sample sizes as low
as 1 mg.
In the present protocol, we describe an experimental HR-MAS
NMR procedure that can be used for the direct profiling of pheno-
lic compounds in olive fruits, without the need for phenol solvent
extraction prior to analysis [18, 19]. The consumption of phenolic
compounds present in olives has been linked to beneficial effects on
human health [20, 21] and is an important quality factor for both
table olives and olive oil. The determination of the phenolic con-
tent of olives is very important since it changes with the maturation
index of the fruit and has a significant effect on the phenolic load of
the olive oil subsequently produced from the fruit [22, 23]. Further-
more, fruit variety and the processing technology have a significant
effect on the phenolic profile of edible (table) olives, which is linked
to their nutritional properties and quality [24]. The protocol devel-
oped can be applied to characterize the phenolic composition either
of the whole olive fruit, or the olive flesh and skin separately, as it is
well known that their chemical composition is different [25].

2 Materials

1. Screw cap glass vials (4 mL).

2. Analytical balance.
3. Variable volume pipette (10–100 μL).
4. Pestle and mortar.
5. Liquid nitrogen (0.5 Lt).
6. Glass flasks (50 or 100 mL).
7. Freeze Dryer (Telstar Cryodos).
8. MeOD containing 0.05% TMSP (trimethyl-silyl propionic acid
sodium salt-d4, internal chemical shift standard).
9. NMR spectrometer equipped with a High Resolution-Magic
Angle Spinning (HR-MAS) probe and MAS accessory.
 Olive Fruit Phenolic Profiling Using High Resolution-Magic Angle. . . 279

10. 4 mM zirconia rotor equipped with a 50 μL Teflon insert

(or appropriate diameter rotors for the NMR spectrometer
probe used).
11. Insert screw, rotor cap, loading funnel, and other tools for
rotor filling/preparation provided by the spectrometer
manufacturer.

3 Methods

3.1 Sample 1. Store 1–2 g of olive fruit, flesh, or skin at -18 °C for 24 h (see
Preparation Note 1).
2. Cut frozen sample to small pieces and weigh.
3. Place sample in 50 mL glass flask, connect to freeze-drier, and
freeze-dry for at least 16 h.
4. Weigh sample again to calculate the amount of moisture.
5. Grind sample with a pestle and mortar under liquid nitrogen
(see Note 2).
6. Pre-weigh the empty rotor and introduce 5–10 mg of frozen
olive fruit/flesh/skin powder using the rotor loading funnel
accordingly (see Note 3).
7. Add carefully 35–40 μL of deuterated methanol into the rotor/
insert via pipette (see Note 4).
8. Completely seal the rotor by placing the insert screw and the
rotor cap using the provided rotor tools.
9. Insert the rotor in the probe either directly or using an auto-
sampler, if available.
10. Run NMR experiment protocol.

3.2 1H NMR 1. Set the probe temperature to 298 K, spin the sample at the
Spectroscopy magic angle at 4 kHz, and wait (3–5 min) until the sample
Experimental Protocol temperature is equilibrated.
with Water 2. Lock, tune, and shim the sample according to standard NMR
Presaturation spectrometer procedures.
3. Load a standard solvent presaturation (Bruker: zgpr or zgcppr)
pulse program with default spectrometer parameters, and set
the frequency of the residual water signal exactly on resonance.
4. Record a water-suppressed 1H NMR spectrum with para-
meters: SW = 20 ppm, NS = 160 scans, DS = 16 dummy
scans, AQ = 4.0 s, D1 = 1 s.
5. Perform Fourier transformation, phase correction, and base-
line correction according to standard NMR spectrometer
(or processing software) procedures.
280 Efstathia Manolopoulou and Apostolos Spyros

3.3 NMR Data Initial phenolic profiling may be achieved by comparing the 1H
Analysis NMR spectra of olives with those of standard olive phenols
obtained in the same solvent, and it should be noted that liquid-
state NMR spectra of phenol standards may also be used for identi-
fication, since chemical shift differences between liquid-state and
HR-MAS NMR spectra of olives, if present, are usually minimal.
Furthermore, the standard arsenal of homonuclear and heteronuc-
lear 2D NMR experiments for structure elucidation (gCOSY,
gHSQC, gHMBC, etc.) is available also for the HR-MAS NMR
spectroscopy platform and may be used for compound identifica-
tion and verification. Other means of phenol identification include
the use of publicly available NMR spectral databases, such as
FoodDB [26] and the Biological Magnetic Resonance Bank [27].
Figure 1 depicts the aromatic region of the HR-MAS solid-
state 1H NMR spectra of olive flesh obtained from four different
olive varieties, along with the NMR spectra of four standard phe-
nolic compounds (verbascoside 1, oleocanthal 2, oleacin 3, oleur-
opein 4) that have been reported as constituents of olive oil and
olives [18, 28–30]. Koroneiki, Chondrolia, and Lianolia samples

Fig. 1 High Resolution-Magic Angle Spinning (HR-MAS) solid-state 1H NMR spectra (aromatic region) of model
compounds (verbascoside 1, oleocanthal 2, oleacin 3, oleuropein 4) and of olive fruit (flesh) of different
varieties (Koroneiki, Chondrolia, Kavala, Lianolia) in MeOD-d4, obtained by applying the method developed in
this study at a proton frequency of 400.2 MHz. Numbers 5 and 6 correspond to the compounds ligstroside and
cornoside, respectively
Olive Fruit Phenolic Profiling Using High Resolution-Magic Angle. . . 281

were obtained from Crete, Greece, while the Kavala sample was a
local olive variety obtained from Kavala, Greece. The phenolic
composition of the olive flesh samples of Fig. 1 differs significantly,
as evident by comparison with the NMR spectra of the standard
compounds. Oleacin 3 is an important and well-known polyphenol
also present in high-quality extra-virgin olive oil and is found to be
present in Chondrolia, Kavala, and Lianolia olive fruit. (see Note 5)
Koroneiki variety on the other hand is found to contain mainly
oleuropein 4, ligstroside 5, and verbascoside 1. Chondrolia and
Kavala olives also contain another interesting phenolic compound,
cornoside 6, [31] and smaller amounts of hydroxytyrosol and
tyrosol, while the doublet at δ 5.95 in the NMR spectrum of
Chondrolia is assigned to halleridone, which is a well-known deg-
radation product of cornoside [29, 31].
As a second example, Fig. 2 depicts the HR-MAS solid-state
1 1
H- H gCOSY 2D NMR spectrum (aromatic region) of olive skin
obtained from processed table olives of Chalkidiki variety. Corno-
side and halleridone are identified as components of the olive skin

Fig. 2 High Resolution-Magic Angle Spinning (HR-MAS) solid-state 1H-1H gCOSY

2D NMR spectrum (aromatic region) of olive skin from table olives of Chalkidiki
variety, at a proton frequency of 400.2 MHz. Crosspeaks assigned to specific
phenolic compounds are indicated as follows: Lut Luteolin, pC p-coumaric acid,
Ty Tyrosol, Hty Hydroxytyrosol, Corn Cornoside, Hal Halleridone
282 Efstathia Manolopoulou and Apostolos Spyros

in this 2D NMR spectrum, along with the simple phenols tyrosol,

hydroxytyrosol, luteolin, and p-coumaric acid, indicating the dif-
ference in phenolic composition between olive flesh and skin.
Thus, we conclude that HR-MAS solid-state NMR can provide
fast, direct, and efficient profiling of the phenolic compounds pres-
ent in olive fruit, flesh, and skin by employing an easy-to-use
experimental protocol that avoids complicated solvent extraction
procedures, apart from the prior removal of water from the olive
sample. Furthermore, this protocol may be easily extended to the
phenolic profiling of other fruits (walnut, almond) or dry plant
material (mint, peppermint, etc.).

4 Notes

1. If only the flesh or skin are the focus of analysis, they must be
separated carefully before freezing.
2. Grinding was not used for olive skin samples, instead these
were just cut in very small pieces after freeze drying.
3. It is important to complete the rotor loading procedure as fast
as possible, since partial thawing of the olive sample will make
the use of the funnel ineffective. In this respect, the loading
funnel may also be cooled right before use.
4. If the sample quantity is less than 5 mg, the solvent volume may
be slightly increased to 45 μL, in order to make sure that the
active volume of the rotor after sealing does not contain any air
pockets that will make shimming inefficient.
5. Oleacin and oleocanthal are present in their hemiacetal form in
CD3OD solutions, as already reported [32].

Acknowledgments

We acknowledge support of this work by the project FoodO-

micsGR “Comprehensive Characterization of Foods” (MIS
5029057) which is implemented under the Action “Reinforcement
of the Research and Innovation Infrastructure,” funded by the
Operational Program “Competitiveness, Entrepreneurship and
Innovation” (NSRF 2014-2020) and co-financed by Greece and
the European Union (European Regional Development Fund).

References
1. Valentini M, Ritota M, Cafiero C, Cozzolino S, 2. Spyros A, Dais P (2012) NMR spectroscopy in
Leita L, Sequi P (2011) The HRMAS-NMR food analysis. RSC food analysis monographs
tool in foodstuff characterisation. Magn Reson 3. Santos ADC, Fonseca FA, Lião LM, Alcantara
Chem 49(Suppl 1):S121–S1S5. https://doi. GB, Barison A (2015) High-resolution magic
org/10.1002/mrc.2826 angle spinning nuclear magnetic resonance in
 Olive Fruit Phenolic Profiling Using High Resolution-Magic Angle. . . 283

foodstuff analysis. TrAC Trends Anal Chem 13. Vermathen M, Marzorati M, Diserens G,
73:10–18. https://doi.org/10.1016/j.trac. Baumgartner D, Good C, Gasser F et al
2015.05.003 (2017) Metabolic profiling of apples from dif-
4. Corsaro C, Cicero N, Mallamace D, Vasi S, ferent production systems before and after con-
Naccari C, Salvo A et al (2016) HR-MAS and trolled atmosphere (CA) storage studied by 1H
NMR towards foodomics. Food Res Int 89: high resolution-magic angle spinning
1085–1094. https://doi.org/10.1016/j. (HR-MAS) NMR. Food Chem 233:391–400.
foodres.2016.09.033 https://doi.org/10.1016/j.foodchem.2017.
5. Sundekilde UK, Eggers N, Bertram HC 04.089
(2019) NMR-based metabolomics of 14. Song EH, Kim HJ, Jeong J, Chung HJ, Kim
food. In: Methods in molecular biology. HY, Bang E et al (2016) A 1H HR-MAS
Humana Press, pp 335–344 NMR-based metabolomic study for metabolic
6. Tang F, Vasas M, Hatzakis E, Spyros A (2019) characterization of rice grain from various
Magnetic resonance applications in food Oryza sativa L. Cultivars. J Agric Food Chem
analysis. In: Webb GA (ed) Annual reports on 64(15):3009–3016. https://doi.org/10.
NMR spectroscopy. Academic Press, pp 1021/acs.jafc.5b05667
239–306. https://doi.org/10.1016/bs. 15. Augustijn D, de Groot HJM, Alia A (2021)
arnmr.2019.04.005 HR-MAS NMR applications in plant metabo-
7. Brescia MA, Jambrenghi AC, Di Martino V, lomics. Molecules 26(4). https://doi.org/10.
Sacco D, Giannico F, Vonghia G et al (2002) 3390/molecules26040931
High resolution nuclear magnetic resonance 16. Larsen FH, Blennow A, Engelsen SB (2008)
spectroscopy (NMR) studies on meat compo- Starch granule hydration-a MAS NMR investi-
nents: potentialities and prospects. Ital J Anim gation. Food Biophys 3(1):25–32. https://doi.
Sci 1(2):151–158. https://doi.org/10.4081/ org/10.1007/s11483-007-9045-4
ijas.2002.151 17. Manolopoulou E, Spinella A, Caponetti E,
8. Shintu L, Caldarelli S, Franke BM (2007) Dais P, Spyros A (2013) High resolution-
Pre-selection of potential molecular markers magic angle spinning NMR study of olive
for the geographic origin of dried beef by leaves. In: van Duynhoven J, Belton PS, Webb
HR-MAS NMR spectroscopy. Meat Sci 76(4): GA, van As H (eds) Magnetic resonance in
700–707. https://doi.org/10.1016/j.mea food science: food for thought. The Royal
tsci.2007.02.010 Society of Chemistry, pp 164–170
9. Panthi RR, Sundekilde UK, Kelly AL, 18. Beteinakis S, Papachristodoulou A, Gogou G,
Hennessy D, Kilcawley KN, Mannion DT Katsikis S, Mikros E, Halabalaki M (2020)
et al (2019) Influence of herd diet on the NMR-based metabolic profiling of edible
metabolome of Maasdam cheeses. Food Res olives – determination of quality parameters.
Int 123:722–731. https://doi.org/10.1016/ Molecules 25(15):3339. https://doi.org/10.
j.foodres.2019.05.026 3390/molecules25153339
10. Vermathen M, Marzorati M, Vermathen P 19. Irmak Ş, Dıraman H (2022) Chemometric
(2012) Exploring high-resolution magic angle characterization of raw olives from important
spinning (HR-MAS) NMR spectroscopy for Turkish table olive cultivars Cvs. Using HPLC–
metabonomic analysis of apples. Chimia DAD method based on their biophenolic pro-
66(10):747–751. https://doi.org/10.2533/ files. J Food Sci Technol 59(5):1792–1800.
chimia.2012.747 https://doi.org/10.1007/s13197-021-
11. Corsaro C, Mallamace D, Dugo G (2014) A 05190-6
quantitative 1H HR-MAS study of the meta- 20. Rocha J, Borges N, Pinho O (2020)
bolic profile of the PGI cherry tomato of Table olives and health: a review. J Nutr Sci 9:
pachino. In: Nuclear magnetic resonance e57. https://doi.org/10.1017/jns.2020.50
(NMR): theory, applications and technology. 21. Infante R, Infante M, Pastore D, Pacifici F,
Nova Science, pp 163–178 Chiereghin F, Malatesta G et al (2023) An
12. Cicero N, Corsaro C, Salvo A, Vasi S, Giofré appraisal of the oleocanthal-rich extra virgin
SV, Ferrantelli V et al (2015) The metabolic olive oil (EVOO) and its potential anticancer
profile of lemon juice by proton and neuroprotective properties. Int J Mol Sci
HR-MAS NMR: the case of the PGI Interdo- 2 4 ( 2 4 ) . h t t p s : // d o i . o r g / 1 0 . 3 3 9 0 /
nato lemon of Messina. Nat Prod Res 29(20): ijms242417323
1894–1902. https://doi.org/10.1080/ 22. El Qarnifa S, El Antari A, Hafidi A (2019)
14786419.2015.1012166 Effect of maturity and environmental
284 Efstathia Manolopoulou and Apostolos Spyros

conditions on chemical composition of olive 28. Kazou M, Nikolopoulou E, Panagou EZ

oils of introduced cultivars in Morocco. J (2023) Greek table olives: an overview on the
Food Qual 2019. https://doi.org/10.1155/ impact of processing elaborations on the con-
2019/1854539 tent of biophenols and related nutritional
23. Lukić I, Žanetić M, Jukić Špika M, Lukić M, implications. Eur Food Res Technol 249(12):
Koprivnjak O, Brkić Bubola K (2017) Complex 3151–3164. https://doi.org/10.1007/
interactive effects of ripening degree, malaxa- s00217-023-04356-0
tion duration and temperature on Oblica 29. Mousouri E, Melliou E, Magiatis P (2014)
cv. Virgin olive oil phenols, volatiles and sen- Isolation of megaritolactones and other bioac-
sory quality. Food Chem 232:610–620. tive metabolites from ‘Megaritiki’ table olives
https://doi.org/10.1016/j.foodchem.2017. and debittering water. J Agric Food Chem
04.047 62(3):660–667. https://doi.org/10.1021/
24. Zoidou E, Melliou E, Gikas E, jf404685h
Tsarbopoulos A, Magiatis P, Skaltsounis A-L 30. Bastoni L, Bianco A, Piccioni F, Uccella N
(2010) Identification of Throuba Thassos, a (2001) Biophenolic profile in olives by nuclear
traditional greek table olive variety, as a nutri- magnetic resonance. Food Chem 73(2):
tional rich source of oleuropein. J Agric Food 145–151. https://doi.org/10.1016/S0308-
Chem 58(1):46–50. https://doi.org/10. 8146(00)00250-8
1021/jf903405e 31. Bianchi G (2003) Lipids and phenols in table
25. Brenes-Álvarez M, Romero C, Medina E, olives. Eur J Lipid Sci Technol 105(5):
Brenes M (2023) Phenolic and oleosidic com- 229–242. https://doi.org/10.1002/ejlt.
pounds in different parts of Spanish olive culti- 200390046
vars: influence of maturation and processing. 32. Karkoula E, Skantzari A, Melliou E, Magiatis P
LWT 184. https://doi.org/10.1016/j.lwt. (2012) Direct measurement of oleocanthal and
2023.115098 oleacein levels in olive oil by quantitative 1H
26. FooDB version 1.0. https://foodb.ca. NMR. Establishment of a new index for the
Accessed 1 Apr 2024 characterization of extra virgin olive oils. J
27. Biological Magnetic Resonance Bank version Agric Food Chem 60(47):11696–11703.
3.6. https://bmrb.io/. Accessed 1 Mar 2024 https://doi.org/10.1021/jf3032765
 INDEX

A H
Alkyl chloroformate derivatization............................... 206 HDMSE ...............................................240, 245, 247, 252
Amino acids (AAs) ................................... 6, 11, 160, 167, HILIC-MS/MS ...........................................160, 181–202
182, 183, 198, 199, 205–218
Authentication............................................................... 277 I
Identifications.......................................3, 6, 8, 10, 11, 20,
B
27, 32, 46, 56, 60, 68, 70, 73, 75, 78, 84, 99, 103,
Biological samples .............................................6, 56, 109, 110, 111, 114, 121, 123, 125, 166, 174, 175,
112, 114, 118, 124, 131–149 181, 182, 205, 232, 249, 253, 270, 277, 280
Biomarker discovery...................................................... 154 Ion mobility (IM) ........................................ 4, 8, 26, 110,
113, 114, 119, 222, 224, 225, 227, 232, 235,
C 239–242, 249, 253
Cells .............................................1, 3, 4, 11, 54, 55, 109, Ion pair chromatography..................................... 166, 182
112, 114, 131, 138, 142, 144, 181, 195, 198,
L
210, 226, 235, 241, 244, 245, 257
Chiral analysis ....................................................... 205–207 Lipidomics .......................................................6, 112, 127,
Collision cross section (CCS)............................. 8, 11, 26, 131–149, 221–236
119, 232, 233, 235, 240, 242–244, 247–253 Lipoproteins ................................................ 138, 143, 144
Liquid chromatography (LC).........................1, 110, 134,
D 156, 167–173, 183–188
Liquid chromatography-mass spectrometry
Data qualities....................................31, 61–63, 111, 115,
146, 175, 178, 179, 234 (LC-MS) ....................................... 3, 4, 10, 19, 25,
47, 53–65, 91–106, 109–128, 157, 160,
E 165–179, 194, 195, 198, 240–242, 270, 271
Liquid chromatography-tandem mass spectrometry
Extracellular vesicles............................................. 139, 144 (LC-MS/MS) .......................................... 110, 155,
157, 158, 160, 161, 187, 258, 259, 266
F
Liquid-liquid extraction.............................. 139, 140, 145
Fecal extracts ..............................155, 156, 158, 159, 161
Fecal samples ..................................................39, 154–161 M
Food analysis ................................................................. 277 Mass spectra handling ........................................ 68, 75–77
Formula handling............................................................ 69 Mass spectrometry (MS)..................................... 2, 26, 29,
39, 53, 55, 75–77, 79, 83, 85, 87, 109, 121,
G 131–149, 165, 174, 181, 239
Gas chromatography-mass spectrometry Metabolic fingerprints................................................... 270
(GC-MS)................................ 6, 10, 31, 109, 156, Metabolic phenotyping....................................... 1–4, 6, 8,
157, 159–161, 205–218, 258, 270 11, 109–128
Glycolipids ..................................................................... 135 Metabolic profiling..................................... 2, 4, 8, 15–49,
Gut microbiome................................................... 154, 156 54, 56, 62, 109–112, 119, 122–124, 155, 156,
Gut microbiota.............................................................. 154 166, 249

Olga Deda et al. (eds.), Metabolic Profiling: Methods and Protocols, Methods in Molecular Biology, vol. 2891,
https://doi.org/10.1007/978-1-0716-4334-1,
© The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Science+Business Media, LLC, part of Springer
Nature 2025

285
 METABOLIC PROFILING: METHODS AND PROTOCOLS
286 Index
Metabolite annotation ....................................67, 68, 103, Quality management (QM).................. 15–18, 21, 27, 38
240, 249–251, 253 Quantification .............................................. 3, 12, 16, 20,
Metabolite profiling .................................. 1–11, 181, 218 38, 54, 60, 110, 114, 132, 134, 137, 138, 141,
Metabolites ............................................. 1, 19, 53, 67–87, 143, 147, 166, 181, 182, 187, 196, 202, 208,
94, 109, 142, 154, 165, 181, 214, 240, 257 214, 215, 217, 264, 266, 270–272, 277
Metabolome ......................................................3, 6, 8, 11, Quantitative analysis ................................... 141, 196, 235
109–111, 114, 142, 154, 165, 240, 269–276
Metabolomics ................................................1, 2, 6, 9–12, R
15–49, 53–65, 67–87, 103, 104, 123, 125, 142,
Rats ...................................................................7, 160, 161
154–156, 181–202, 218, 240, 245, 246, 252, Retention time correction .............................93–101, 106
253, 269–272 RforMassSpectrometry ................................................... 68
Metabonomics..........................1–3, 9, 11, 165, 166, 181
Robust method ........................................... 15, 16, 18, 45
Metadata ..................................................... 32, 76, 93–95,
104, 105, 245, 252 S
Metanephrines (MNs) ........................ 257–261, 263–266
Missing values............................................ 31, 35, 99, 252 Sample preparation ...........................................3, 6, 7, 19,
24, 30, 44, 58–60, 64, 111, 114, 116, 117, 126,
N 138–140, 143–145, 154–156, 158–160,
172–174, 181, 182, 194, 195, 198, 207, 211,
NMR spectroscopy................................... 3, 4, 10, 11, 53, 213, 214, 218, 221, 241, 245–246, 252, 258, 279
109, 155–158, 160, 165, 166, 277–281 Secondary AAs ..................................................... 205–218
Nuclear magnetic resonance (NMR) ................. 2–4, 155,
Serum............................................. 1, 6, 7, 31, 55–59, 63,
181, 269–276 109, 112, 114–116, 119, 121, 122, 143, 148,
166, 182, 194, 200, 202, 206–210, 214, 217
O
Solid-state ............................................................. 277–282
Organs ................................... 1, 131, 139, 142, 144, 153 Spectra similarity calculations ...................................78–83
Sphingolipids ........................................................ 132, 145
P Supercritical fluid chromatography (SFC)............. 6, 8, 9,
Parallel accumulation serial fragmentation 131–149
(PASEF) ..........................222, 223, 225, 227, 235
T
Peak picking ......................................... 57, 62, 67, 93–97,
105, 106, 123, 143, 147, 235, 248, 252 Targeted metabolomics .................................54, 111, 182
Peptide hydrolysates...................207, 208, 210, 214, 218 Tissues.....................................................3, 4, 6, 7, 46, 54,
Phenolic profiling................................................. 277–282 55, 109–112, 114, 117–119, 122, 123, 126, 139,
Pheochromocytoma (PHEO) ............................. 257, 258 142, 144, 155, 165–179, 181, 194, 195, 210
Phospholipids .............................................. 141, 145, 149 Trapped ion mobility spectrometry (TIMS) ..... 113, 222,
Plants ......................................6, 109, 269–276, 278, 282 223, 225, 226, 235, 240
Plasma ...............................................1, 3, 6, 7, 30, 55–59, Traveling wave ion mobility ......................................... 240
63, 112, 114, 116, 119, 121, 122, 127, 138, 140, Two-phase extraction.................................. 270, 271, 274
142, 143, 166, 182, 210, 214, 217, 230, 235,
257, 258, 265 U
Polar analytes........................................................ 111, 182 Untargeted metabolic profiling........................... 110, 111
Polar metabolites.............................................6, 111, 114,
Untargeted metabolomics ........................... 9, 16, 20, 44,
148, 166, 182, 205 53–56, 240, 245
Preprocessing........................................31, 61, 65, 67, 92, Urine......................................1, 3, 4, 6, 7, 54, 55, 57–59,
94–103, 105, 106, 122, 123, 252, 270
63, 109, 112, 114, 116, 118–121, 124, 154, 166,
182, 194–200, 202, 207–210, 214, 217, 218,
Q
258, 259, 262, 263, 265, 266
Quality assurance (QA).............................. 15, 16, 18, 19,
21–43, 45, 49, 54, 55 V
Quality checks ............................93, 95, 96, 99, 101, 234 Validation.............................................. 20, 37, 44, 53–65,
Quality control (QC)......................................8, 9, 15, 18,
167, 174, 175, 177, 252, 264
22, 29–31, 35, 36, 44, 46, 48, 53–65, 96, 105, Vitis ................................................................................ 239
111, 114–120, 123–125, 127, 128, 147, 149,
174–176, 178, 179, 195, 214, 215, 217, 218,
229, 245–248, 251–253, 262