The handbook of plant metabolomics

Hill and Ute Roessner 1.2.1.2 Homogenization and Extraction 7 1.2.1.3 Procedure for Polar Extraction of Metabolites 8 1.2.2 Chemical Derivatization: Methoxymation and Silylation 9 1.2.2.

and Günter KahlThe Handbook of PlantMetabolomics

Tai Lieu Chat Luong

Titles of the Series “Molecular Plant Biology Handbook Series”

Kahl, G., Meksem, K (eds.)

The Handbook of Plant Functional Genomics

Concepts and Protocols

2008

ISBN: 978-3-527-31885-8

Meksem, K., Kahl, G (eds.)

The Handbook of Plant Mutation Screening

Mining of Natural and Induced Alleles

2010

ISBN: 978-3-527-32604-4

Meksem, K., Kahl, G (eds.)

The Handbook of Plant Genome Mapping

Genetic and Physical Mapping

2005

ISBN: 978-3-527-31116-3

Related Titles

Harbers, M., Kahl, G (eds.)

Tag-based Next Generation Sequencing

2012

ISBN: 978-3-527-32819-2

Hirt, H (ed.)

Plant Stress Biology

From Genomics to Systems Biology

2010

ISBN: 978-3-527-32290-9

Hayat, S., Mori, M., Pichtel, J., Ahmad, A (eds.)

Nitric Oxide in Plant Physiology

The Handbook of Plant Metabolomics

The cover picture presents some structures of

representative phytochemicals and

biosynthetic pathways and enzymes of

Arabidopsis thaliana, referred to in the chapter

“Integrative analysis of secondary metabolism

and transcript regulation in Arabidopsis

thaliana ” by Fumio Matsuda and Kazuki Saito

(for further details see Chapter 9, Fig 4) The

figure was originally published in “Matsuda,

F., et al (2010) AtMeteEpress development: A

phytochemical atlas of Arabidopsis

development Plant Physiol, 152, 566–578),

www.plantphysiol.org, # American Society of

Plant Biologists The permission of the authors

to partly use their figure in a changed format

is greatly appreciated Foto of Arabidopsis:

# Vasiliy Koval, Fotolia.com

and author have used their best efforts in preparing this book, they make no representations or warranties with respect to the accuracy or completeness of the contents of this book and specifically disclaim any implied warranties of merchantability or fitness for a particular purpose No warranty can be created or extended by sales representatives

or written sales materials The Advice and strategies contained herein may not be suitable for your situation You should consult with a professional where appropriate Neither the publisher nor authors shall be liable for any loss

of profit or any other commercial damages, including but not limited to special, incidental, consequential, or other damages.

Library of Congress Card No.: applied for

British Library Cataloguing-in-Publication Data

A catalogue record for this book is available from the British Library.

Bibliographic information published by the Deutsche Nationalbibliothek

The Deutsche Nationalbibliothek lists this publication in the Deutsche Nationalbibliografie; detailed bibliographic data are available on the Internet at < http://dnb.d-nb.de >.

#2013 Wiley-VCH Verlag GmbH & Co KGaA, Boschstr 12,

69469 Weinheim, Germany Wiley-Blackwell is an imprint of John Wiley & Sons, formed

by the merger of Wiley’s global Scientific, Technical, and Medical business with Blackwell Publishing.

All rights reserved (including those of translation into other languages) No part of this book may be reproduced in any form – by photoprinting, microfilm, or any other means – nor transmitted or translated into a machine language without written permission from the publishers Registered names, trademarks, etc used in this book, even when not specifically marked as such, are not to be considered unprotected by law.

Print ISBN: 978-3-527-32777-5 ePDF ISBN: 978-3-527-66989-9 ePub ISBN: 978-3-527-66990-5 mobi ISBN: 978-3-527-66991-2 oBook ISBN: 978-3-527-66988-2 Cover Design Adam-Design, Weinheim Typesetting Thomson Digital, Noida, India Printing and Binding Markono Print Media Pte Ltd, Singapore

Printed in Singapore Printed on acid-free paper

Ulrich and Hannelore Weckwerth

for their endless sympathy, patience and guidance

Preface XVII

List of Contributors XIX

Camilla B Hill and Ute Roessner

1.2.1.2 Homogenization and Extraction 7

1.2.1.3 Procedure for Polar Extraction of Metabolites 8

1.2.2 Chemical Derivatization: Methoxymation and Silylation 9

1.2.2.1 Procedure for the Chemical Derivatization of Plant Extracts 9

1.2.3.1 Procedure to Acquire GC–MS Data 11

1.2.4.1 Procedure for Postacquisition Data Preprocessing 12

1.2.4.2 Data Analysis and Statistics 14

1.2.4.3 Procedure for Postacquisition Data Analysis 15

References 18

2 Isotopologue Profiling– Toward a Better Understanding

of Metabolic Pathways 25

Wolfgang Eisenreich, Claudia Huber, Erika Kutzner, Nihat Knispel,

and Nicholas Schramek

2.2.2.1 Protein-Bound Amino Acids 36

2.2.2.2 Metabolic Intermediates and Polar Products 37

2.2.4 Protocols for Isotopologue Profiling by NMR 41

2.2.5 Deconvolution of Isotopologue Data 43

2.2.6 Expanding the Metabolic Space by Retrobiosynthetic Analysis 45

4.2.1 Instrumentation 81

4.2.5 Quantitative Analysis of Selected Metabolites 84

References 90

Ale9sSvato9sand Hans-Peter Mock

5.2.3.1 Bruker Ultraflex Instruments 103

References 109

6 Medicago truncatula Root and Shoot Metabolomics: Protocol

for the Investigation of the Primary Carbon and Nitrogen MetabolismBased on GC–MS 111

Vlora Mehmeti, Lena Fragner, and Stefanie Wienkoop

6.2.3 Plant Material and Harvest 113

6.2.7 Metabolite Identification and Quantification: Data Matrix

Processing 116

ContentsjIX

6.2.8 Data Mining 119

References 123

Part II Secondary and Lipid Metabolism 125

7 Study of the Volatile Metabolome in Plant–Insect Interactions 127

Georg J.F Weingart, Nora C Lawo, Astrid Forneck, Rudolf Krska,

and Rainer Schuhmacher

7.1.1 Plant–Insect Interactions 127

7.1.2 Significance of Volatile Plant Metabolites 128

7.1.3 Study of the Plant Volatile Metabolome in Plant–Insect Interactions 1287.1.3.1 Setting Up of Biological Experiments 129

7.1.3.2 Sampling, Quenching, and Sample Preparation 130

7.1.3.3 Headspace Extraction and Measurement by GC–MS 131

7.1.3.5 Biological Interpretation 135

7.2.2 Cultivation of Grapevine Plants and Inoculation with Phylloxera 136

7.2.3 Sampling and Quenching of Plant Tissue (Roots and Leaves) 1387.2.3.1 Sampling and Quenching of Root Tips 138

7.2.3.2 Sampling and Quenching of Grapevine Leaves 139

7.2.4 Milling and Weighing of Plant Tissue (Roots and Leaves) 140

7.2.4.1 Milling and Weighing of Root Samples 140

7.2.4.2 Milling and Weighing of Leaf Samples 141

7.2.6.1 An In-House Reference Library Has to be Established in Advance 1457.2.6.2 Generation of RI Calibration File 146

7.2.6.3 Batch Job Analysis for the Simultaneous Processing of Multiple

7.2.7.1 Univariate Statistics 147

7.2.7.2 Multivariate Statistics 148

7.3 Applications of the Technology 148

References 150

Lie-Fen Shyur, Chiu-Ping Liu, and Shih-Chang Chien

8.2.2.1 Sample Handling for Medicinal Plants 160

8.2.2.2 Sample Preparation for LC–MS Analysis 160

8.2.2.4 HPLC–Photodiode Array (PDA) MS Setup and Analysis 161

8.2.2.6 Plant Extract Preparation for GC–MS Analysis 163

8.2.2.7 GC–MS Parameters and Analysis 164

8.2.2.10 Sample Preparation and LC–SPE–NMR Analysis 167

References 170

9 Integrative Analysis of Secondary Metabolism and Transcript

Regulation in Arabidopsis thaliana 175

Fumio Matsuda and Kazuki Saito

9.2.1 Metabolome Analysis of Plant Secondary Metabolites 177

9.2.1.1 Sample Preparation 177

9.2.1.2 Data Acquisition 178

9.2.1.3 Preparation of Metabolite Accumulation Data from the Raw

9.2.2 Preparation of Combined Data Matrix 180

9.2.2.1 Preparation of Gene Expression Data 180

9.2.2.2 Combination of Data Matrices 180

9.2.3.2 Correlation Analysis 181

9.2.3.3 Principal Component Analysis and Application of Other

Data Mining Techniques 183

ContentsjXI

9.3 Applications of the Technology 183

10.2.2 Plant Cultivation Conditions 208

10.2.3 Preparation of Biological Material with Biotechnological Methods

(Callus, Cell, or Hairy Root Cultures) 208

10.2.4 Extraction of Plant Tissue or Biotechnologically Prepared Material 20810.2.4.1 Extraction Procedure 209

10.2.5 Solid-Phase Extraction of Culture Medium or Apoplastic Fluids 20910.2.6 Preparation of Samples for LC–MS Analyses 210

10.2.7 Chromatographic Protocols for Separation of Flavonoid

Glyconjugates 210

10.2.8 Control of Ionization Parameters During Mass Spectrometric

Analysis and Identification of Compounds During LC–MS

Metabolite Profiling 211

10.3 Applications of the Technology 211

References 212

11 Introduction to Lipid (FAME) Analysis in Algae Using Gas

Chromatography–Mass Spectrometry 215

Takeshi Furuhashi and Wolfram Weckwerth

11.2.8.2 Protocol II 221

12.2.2 Plasmid Construction of Multi-Gene Transformation 233

12.2.3 Preparation ofDual Terminator (DT) Fragment by PCR-Based

Overlap Extension Method 233

12.2.4 Plasmid Construction ofpUHR KS CSPS Thsp 236

12.2.5 Construction ofpHSG299 CSPS 35S-CYP88-DT (Figure 12.2a) 23612.2.6 Construction ofpHSG299 CSPS 35S-CYP72-DT2 (Figure 12.2a) 23712.2.7 Construction ofpHSG299-CYP93(RNAi)-DT (Figure 12.2a) 238

12.2.9 Transformation of Soybean by Particle Bombardment 239

12.2.9.1 Preparation of Embryogenic Suspension Tissue Culture 239

12.2.9.2 Preparation of Plasmid DNA for Particle Bombardment 240

12.2.9.3 Conditions of Particle Bombardment 240

12.2.9.4 Selection and Generation of Transgenic Soybean Plants 240

12.2.10 GC-MS Analysis for Triterpene Glycone 241

Part III Metabolomics and Genomics 245

Alexander Herrmann and Nicolas Schauer

14 Conducting Genome-Wide Association Mapping of Metabolites 255

Susanna Atwell and Daniel J Kliebenstein

14.2.1 Biological Question to Be Addressed 256

14.2.6 Computational Platform to Use for Analysis 261

14.2.6.1 Single Marker Analysis 262

14.2.6.2 Population Structure Modification 262

14.2.6.3 Resulting GWA Plots 262

14.2.6.4 Gene-Based Approaches 263

14.2.6.5 What Should I Use and How Do I Use It? 263

14.2.8 Candidate Gene Validation 266

14.2.8.1 Validate That the Gene Influences the Phenotype? 267

14.2.8.2 Validate That Natural Variation in the Gene Influences the

References 268

Part IV Metabolomics and Bioinformatics 273

15 Metabolite Clustering and Visualization of Mass Spectrometry Data

Using One-Dimensional Self-Organizing Maps 275

Alexander Kaever, Manuel Landesfeind, Kirstin Feussner, Ivo Feussner,and Peter Meinicke

16 Metabolite Identification and Computational Mass Spectrometry 289

Steffen Neumann, Florian Rasche, Sebastian Wolf, and Sebastian B€ocker

16.2 Annotation and Identification of Metabolites 290

16.2.1 Exact Mass Search in Compound Libraries 291

16.2.2 Deriving the Elemental Composition from MS1 292

16.2.3 Elemental Composition from MS2and MSn 293

16.2.4 In Silico Library Search with MetFrag 294

16.2.5 Reference Spectral Library Lookup 299

References 303

Xiaoliang Sun and Wolfram Weckwerth

17.2.1.1 Imputation of Missing Values 308

17.2.1.2 Transformations to Satisfy Prerequisites of Statistical Methods 31017.2.1.3 Adjusting Outliers 310

17.2.1.4 Scaling 310

17.2.1.5 Filtering by Statistical Features 310

17.2.2 Uni- and Bivariate Statistical Methods for Individual

Metabolite-Level Analysis 311

17.2.2.1 ANOVA Compares Single Metabolite Levels 311

17.2.2.2 Correlation Coefficients Interpret the Relationships Between

Pairwise Two Metabolites 311

17.2.2.3 Granger Causality Analysis Identifies the Causation Between

Pairwise Two Metabolites in Time-Series Data 311

17.2.3 Multivariate Statistical Methods for Group-Level Analysis 312

17.2.3.1 PCA Distinguishes Phenotypes and Finds Most Influencing

Metabolites 312

17.2.3.2 Independent Component Analysis Distinguishes Phenotypes

and Finds the Latent Sources of Metabolites in Time-Series Data 31217.2.3.3 Clustering Classifies Data Into Groups 312

Analysis, and Clustering 313

17.2.5.2 On the Variance and Covariance: ANOVA, PCA, and ICA 314

References 320

ContentsjXV

18 Mass Spectral Search and Analysis Using the Golm Metabolome

18.2.2 The Text Search Queries 325

18.2.3 The Mass Spectrum Query Submission and Analysis Options 32518.2.3.1 Mass Spectral Matching 326

18.2.3.2 Decision Tree (DT)-Supported Substructure Prediction 32918.2.4 Interpreting the Mass Spectral Analysis Results 329

18.2.4.1 The Mass Spectral Matching Results 329

18.2.4.2 The Substructure Prediction Results 332

18.2.4.3 Interpreting Decision Trees 333

18.2.5.1 General Considerations 336

References 342

Glossary 345

Whereas the most modern topics of plant research, grouped into the term“omics,”such as genomics, transcriptomics, and proteomics, are comparatively new, thestrategies to isolate, purify, identify, and quantify a multitude of low molecularweight cellular compounds (called metabolites) look back onto a long history Whileinitially only metabolites present in comparatively high concentrations could beisolated and quantified (sometimes only semiquantified), the advent of enzymaticdetection techniques in the 1970s brought a breakthrough in precise metaboliteanalysis Unfortunately, these techniques required the coupling of a metabolite’sdetection to the reduction/oxidation of NADþor NADPþ, and therefore excludedthe majority of metabolites, especially secondary metabolites More recent develop-ments in mass spectrometry (MS), matrix-assisted laser desorption/ionization(MALDI)–MS for metabolite imaging, gas chromatography coupled to mass spec-trometry (GC–MS), liquid chromatography coupled to mass spectrometry (LC–MS),and NMR technology for medium- to high-throughput identification and quantifi-cation of low molecular weight compounds pushed metabolite analysis to today’sadvanced level, where various physico-chemical separation techniques are com-bined to analyze metabolite profiles in considerable detail and accuracy The presentstate of technology for metabolite analysis has been denoted“metabolomics.”

Metabolomics reflects the physiological state of an organism or its organs, tissues,

or cells, and therefore allows a hitherto not possible comprehensive understanding

of the biology of an organism and its response to intrinsic or environmental changes

or influences The various techniques of metabolomics allow the comprehensiveprofiling of cellular metabolites at the systems level, thereby providing a directreadout of biochemical activity that can be correlated with phenotype and used toidentify therapeutic targets This omics discipline then bridges the gap betweengenotype and phenotype The present Handbook of Plant Metabolomics not onlywitnesses the present state-of-the-art metabolomics and its widespread applications,but also portrays up-to-date technical advances in metabolicfingerprinting and the

in silico analysis of the resulting, mostly very complex, metabolite patterns

The Handbook of Plant Metabolomics (Metabolite Profiling and Networking) is thefourth volume in the successful Wiley-VCH series of Handbooks of Plant GenomeAnalysis, and follows the warmly welcomed The Handbook of Plant Genome Mapping(Genetic and Physical Mapping), The Handbook of Plant Functional Genomics

jXVII

(Concepts and Protocols), and The Handbook of Plant Mutation Screening (Mining ofNatural and Induced Alleles) It provides informative introductions to each chapter,detailed descriptions of techniques for metabolite profiling, and robust and ready-to-

go laboratory protocols, in addition to some applications, all written by ally renowned experts in their researchfields Although this volume focuses on plantmetabolomics, the techniques presented are broadly applicable to other biologicalsystems exemplifying the pioneering and original character of metabolomics inplant biology This rapid development of metabolomics to a mature technology iscatalyzing the application of metabolomics in otherfields of research also, such asbiomedicine

internation-The Editors very much appreciate the excellent chapters contributed by all theauthors, and expect that The Handbook of Plant Metabolomics will reproduce theworldwide success of its three progenitors

August 2012

University of California, Davis

Department of Plant Sciences

One Shields Avenue

No 250, Kuo Kuang RoadTaichung 402

TaiwanYoung Hae ChoiLeiden UniversityInstitute of BiologyNatural Products LaboratorySylviusweg 72, 2333 BELeiden

The NetherlandsKatja DettmerUniversity of RegensburgInstitute of Functional GenomicsJosef-Engert-Strasse 9

93053 RegensburgGermany

Wolfgang EisenreichTechnische Universit€at M€unchenLehrstuhl f€ur BiochemieLichtenbergstrasse 4

85435 GarchingGermany

jXIX

University of Natural Resources and

Life Sciences, Vienna

Department of Crop Sciences

Division of Viticulture and Pomology

Althanstrasse 14

1090 ViennaAustriaAlexander HerrmannMetabolomic Discoveries GmbH

Am M€uhlenberg 11

14476 Potsdam-GolmGermany

Camilla B HillThe University of MelbourneSchool of Botany, Building 122Australian Centre for PlantFunctional Genomics (ACPFG)Professors Walk

Parkville, VIC 3052Australia

Claudia HuberTechnische Universit€at M€unchenLehrstuhl f€ur BiochemieLichtenbergstrasse 4

85435 GarchingGermanyJan HummelMax Planck Institute of MolecularPlant Physiology (MPIMP)Bioinformatics Group

Am Muehlenberg 1

14476 Potsdam-GolmGermany

Masao IshimotoNational Institute of AgrobiologicalSciences

2-1-2 KannondaiTsukubaIbaraki 305-8602Japan

University of California, Davis

Department of Plant Sciences

One Shields Avenue

Davis, CA 95616

USA

Nihat Knispel

Technische Universit€at M€unchen

Lehrstuhl f€ur Biochemie

Center for Analytical ChemistryDepartment IFA-Tulln

Konrad-Lorenz-Strasse 20

3430 TullnAustria

Erika KutznerTechnische Universit€at M€unchenLehrstuhl f€ur BiochemieLichtenbergstrasse 4

85435 GarchingGermanyManuel LandesfeindGeorg-August-Universität GöttingenInstitute of Microbiology and GeneticsDepartment of BioinformaticsGoldschmidtstrasse 1

37077 GöttingenGermanyNora C LawoUniversity of Natural Resources andLife Sciences, Vienna

Division of Viticulture and PomologyDepartment of Crop SciencesKonrad-Lorenz-Strasse 24

3430 TullnAustriaandSyngenta Crop ProtectionResearch Stein

Schaffhauserstrasse 101

4332 SteinSwitzerland

List of ContributorsjXXI

RIKEN Plant Science Center

Metabolomic Function Research Group

Graduate School of Information

Science and Technology

Eindhovenweg 20, 2333 ZCLeiden

The NetherlandsHans-Peter MockLeibniz Institute of Plant Geneticsand Crop Plant Research (IPK)Corrensstrasse 3

06466 GaterslebenGermany

Toshiya MuranakaYokohama City UniversityKihara Institute for BiologicalResearch

641-12 Maioka-choTotsuka-kuYokohamaKanagawa 244-0813Japan

andOsaka UniversityDepartment of Biotechnology2-1 Yamadaoka

Suita-shiOsaka 565-0871Japan

Steffen NeumannLeibniz Institute of PlantBiochemistry, IPB HalleDepartment of Stress andDevelopmental BiologyWeinberg 3

06120 Halle (Saale)Germany

Tokyo Institute of Technology

Graduate School of Engineering

The University of Melbourne

School of Botany, Building 122

Australian Centre for Plant

Functional Genomics (ACPFG) and

Suehiro-cho 1-7-22Tsurumi-kuYokohama 230-0045Japan

andGraduate School of PharmaceuticalSciences

Department of Molecular Biology andBiotechnology

Chiba UniversityInohana 1-8-1Chuo-kuChiba 260-8675Japan

Nozomu SakuraiKazusa DNA Research Institute2-6-7 Kazusa-Kamatari

KisarazuChiba 292-0818Japan

Satoru SawaiChiba UniversityGraduate School of PharmaceuticalSciences

1-33 Yayoi-choInage-kuChiba 263-8522Japan

andTokiwa Phytochemical Co., Ltd

58 KinokoSakuraChiba 285-0801Japan

List of ContributorsjXXIII

Technische Universit€at M€unchen

Lehrstuhl f€ur Biochemie

Center for Analytical ChemistryDepartment IFA-Tulln

Konrad-Lorenz-Strasse 20

3430 TullnAustriaHikaru SekiYokohama City UniversityKihara Institute for BiologicalResearch

641-12 Maioka-choTotsuka-kuYokohamaKanagawa 244-0813Japan

andOsaka UniversityDepartment of Biotechnology2-1 Yamadaoka

Suita-shiOsaka 565-0871Japan

Daisuke ShibataKazusa DNA Research Institute2-6-7 Kazusa-Kamatari

KisarazuChiba 292-0818Japan

Lie-Fen ShyurAgricultural Biotechnology ResearchCenter

Academia Sinica

No 128, Sec 2, Academia RoadNankang

Taipei 115Taiwan

Leibniz Institute of Plant

Biochemistry, IPB Halle

Department of Stress and

Althanstrasse 14

1090 ViennaAustriaHideyuki SuzukiKazusa DNA Research Institute2-6-7 Kazusa-Kamatari

KisarazuChiba 292-0818Japan

Ale9s Svato9sMax Planck Institute for ChemicalEcology

Mass Spectrometry Research GroupHans-Knoell-Strasse 8

07745 JenaGermanyEiji TakitaKazusa DNA Research Institute2-6-7 Kazusa-Kamatari

KisarazuChiba 292-0818Japan

andIdemitsu Kosan Co., Ltd

1280 KamiizumiSodegaura-shiChiba 299-0293Japan

Sonia van der SarLeiden UniversityInstitute of BiologyNatural Products LaboratorySylviusweg 72, 2333 BELeiden

The Netherlands

List of ContributorsjXXV

Max Planck Institute of Molecular

Plant Physiology (MPIMP)

Center for Analytical ChemistryDepartment IFA-Tulln

Konrad-Lorenz-Strasse 20

3430 TullnAustriaandFondazione Edmund MachResearch and Innovation CentreFood Quality and NutritionDepartment

Via E Mach 1

38010 San Michele all’Adige (TN)Italy

Stefanie WienkoopUniversity of ViennaDepartment of Molecular SystemsBiology

Althanstrasse 14

1090 ViennaAustriaSebastian WolfLeibniz Institute of PlantBiochemistry, IPB HalleDepartment of Stress andDevelopmental BiologyWeinberg 3

06120 Halle (Saale)Germany

Camilla B Hill and Ute Roessner

1.1

Introduction

For numerous organisms, complete genomes have been sequenced [1–3] andtranscriptome [4–6] and proteome studies [7–9] have been described, but onlyrecently have metabolome analyses using mass spectrometry (MS)-based platformsattracted attention Recent advances in analytical technologies have now allowed theanalysis of complex metabolic structures in an organism

Metabolomics is currently a very powerful tool for characterizing metabolites andmetabolic pathways and aims to provide a“snapshot” of the biochemical state of abiological sample The number of metabolites is expected to be significantly lowerthan the number of genes, proteins, or mRNAs, which reduces the complexity of thesample However, the total number of metabolites in the plant kingdom is estimated

to be between 100 000 and 200 000, which makes cataloging of all metabolites achallenging task [10,11] The metabolic composition of plants is likely to be alteredduring different physiological and environmental conditions and can also reflectdifferent genetic backgrounds Metabolomics aims to provide a comprehensive andunbiased analysis of all metabolites with a low molecular weight present in abiological sample, such as an organism, a specific tissue, or a cell, under certainconditions [12]

Analytical strategies for plant metabolite analysis include metabolic profiling,metabolite target analysis, and metabolicfingerprinting and are chosen according toeither the focus of the research or the research question [12–14] Metabolite profilingaims to detect as many metabolites as possible within a structurally relatedpredefined group, for example, organic acids, amino acids, and carbohydrates.Metabolic profiling does not necessarily aim to determine absolute concentrations

of metabolites but rather their comparative levels In contrast, the aim of targetedmetabolite analysis is to determine pool sizes (e.g., absolute concentrations) ofmetabolites involved in a particular pathway by utilizing specialized extractionprotocols and adapted separation and detection methods A third conceptualapproach in metabolome analysis is metabolicfingerprinting, which generally isnot intended to identify individual metabolites, but rather provides afingerprint ofThe Handbook of Plant Metabolomics, First Edition Edited by Wolfram Weckwerth and Günter Kahl

#

all chemicals measurable for sample comparison and discrimination analysis bynonspecific rapid analysis of crude metabolite mixtures Depending on the analyticalstrategy, a number of different instrumental platforms with different configurationsmay need to be utilized to ensure optimal data acquisition [15].

Because of the diversity of structural classes of metabolites, ranging from primarymetabolites such as carbohydrates, amino acids, and organic acids to very complexsecondary metabolites such as phenolics, alkaloids, and terpenoids, there is nosingle methodology that can measure the complete metabolome in one step It isnecessary to combine different techniques to detect all metabolites in a complexmixture [13] It is possible that two samples, although very different, may show thesame metabolite profile using one strategy Therefore, only by employing a combi-nation of different instrument platforms and techniques can the suite of differences

in the metabolite profiles be revealed

Several extraction methods and instrument platforms have been established toanalyze highly complex mixtures, and each has to be chosen according toparticular interests These include nuclear magnetic resonance (NMR), Fouriertransform ion cyclotron resonance mass spectrometry (FT-ICR-MS or FT-MS),and mass spectrometry (MS) coupled with liquid chromatography (LC) [liquidchromatography–mass spectrometry (LC–MS)] or gas chromatography (GC)[gas chromatography–mass spectrometry (GC–MS)] Section 1.2 focuses on theapplication of GC–MS to plant metabolomics studies; the advantages and dis-advantages of other instrument platforms for metabolomics were discussed inRefs [16–19]

The coupling of GC to electron impact ionization (EI) MS is possibly the oldesthybrid technique in analytical chemistry and is considered to be one of the mostdeveloped, robust, and highly sensitive instrument platforms for metabolite analysis[20–22] GC–MS offers high chromatographic separation power, robust quantifica-tion methods, and the capability to identify metabolites with highfidelity, and istherefore often referred to as the“gold standard” in metabolomics [23] GC–MS-based methodologies were among thefirst to be applied to metabolite profiling andtarget analysis, thus offering established protocols for machine setup, data mining,and interpretation Compared with other instrument platforms, it offers the lowestacquisition, operating, and maintenance expenses [24] Furthermore, both commer-cially and publicly available EI spectral libraries facilitate the use of GC–MS as ametabolomics platform [25]

Historically, the first chromatographic separation techniques were developedbetween 1940 and 1950 by Martin and Synge, who won the 1952 Nobel Prizefor their invention of partition chromatography [26,27] They further contributedsubstantially to the development of GC and high-performance liquid chromatogra-phy (HPLC) During the 1970s, the term“metabolite profiling” was coined and wasfirst applied in studies of steroid and steroid derivatives, amino acids, and drugmetabolites [28,29] in 1971 In the following years, metabolite research developedtoward the utilization of metabolic profiling by GC–MS as a diagnostic technique inmedicine to monitor metabolites present in urine [30] But it was not until the 1990sthat metabolomics found its way into plant research In the late 1990s, Oliver et al

4j1 Metabolic Profiling of Plants by GC–MS

were thefirst to introduce the terms metabolome and metabolomics [31] About adecade ago, one of the first approaches for high-throughput, large-scale, andcomprehensive plant metabolite analysis was conducted by Roessner et al.[21,32,33], who analyzed more than 150 compounds simultaneously within a singlepotato (Solanum tuberosum) tuber sample using GC–MS, and Fiehn et al [20], whoanalyzed 326 distinct compounds from Arabidopsis thaliana leaf extracts of fourgenotypes by GC–MS, and identified 50% of these compounds Several studieshave now implemented this approach, and it has been applied to various plantspecies and tissues, including A thaliana leaf tissue [13], phloem exudates ofbuttercup squash (Cucubita maxima) [34], tomato leaves and fruit (Solanum lyco-persicum) [35,36], and barley leaf and root tissue (Hordeum vulgare) [37] GC–MSapplications include studies that associate certain metabolites with biotic [38] andabiotic stress responses [39–42], define metabolic differences of genetically modifiedplants [32,33,35,43,44], or integrate genetic and metabolite data for plant functionalgenomics [45–49].

GC is the preferred technique for the separation of low molecular weightmetabolites which are either volatile or can be converted into volatile andthermally stable compounds through chemical derivatization before analysis[15] This includes especially primary metabolites, such as amino acids, amines,sugars, organic acids, fatty acids, long-chain alcohols, and sterols, whereas LC–MSanalysis is favored for detecting a broader range of metabolites, includingsecondary metabolites such as alkaloids, terpenes, flavonoids, glucosinolates,and phenylpropanoids [50,51] Derivatization is usually needed to increasevolatility and to reduce the polarity of polar hydroxyl (OH), amine (NH2),carboxyl (COOH), and thiol (SH) groups [25] Exceptions include plantvolatiles [52] and metabolites present in essential oils [53], which can be injecteddirectly into the GC column

The greatest challenge of any metabolomics project is to make sense of the wealth

of data that has been produced during metabolite analysis Targeted metaboliteanalysis employs optimized measurements of preselected metabolites, which arecharacterized by their mass spectrum and retention time/index, and allows the fastand easy construction of the data matrix [25,54] It is a highly quantitative methodwith a very high detection rate for known metabolites, which must be available inpurified form To quantify metabolites, either external calibration (which requirespreparation of standard solutions) or internal calibration (based on the relationbetween the peak area of the compound and that of an internal standard) can beemployed [16]

In contrast, untargeted analysis distinguishes all mass peaks above a certainthreshold by their mass spectrum and retention time/index, with the majority ofthem not being identified, and can be used to detect novel metabolic markers In thiscase, data mining is more complex than in targeted analysis and requires bio-informatics and statistical tools to avoid labor-intensive and time-consumingmanual data handling

In our laboratory, we routinely use GC–MS as a tool to investigate tolerancemechanisms of plants, particularly cereal crops such as wheat, rice, and barley,

under abiotic stress, including drought, cold, salinity, and nutritional de ficiencies ormineral toxicities (www.acpfg.com.au; www.metabolomics.com.au).

Plant metabolite profiling using GC–MS involves the steps depicted in Figure 1.1.The most relevant sections of this experimental workflow are detailed in Section1.2 The chapter then turns to the implementation of GC–MS in plant metabolo-mics, portraying various examples of applications of this technology The finalsection reports new developments in GC–MS technology

Figure 1.1 Workflow showing the general strategy and experimental steps of a GC–MS experiment.

6j1 Metabolic Profiling of Plants by GC–MS

on the accuracy and precision of the information gained from the experiment Inparticular, when studying plant samples, the influences of environmental factorssuch as harvesting time (day/night, season), light conditions, temperature, develop-mental stage of the plant or plant cells, the type of harvested tissue/plant organs, andgenetic factors have to be considered [55].

1.2.1.1 Sampling

Thefirst step in sample preparation for plant metabolite analysis is harvesting ofplant tissue by rapid freezing in liquid nitrogen (196C) and storing at80C, orfreeze-drying for longer storage until used This will stop all enzymatic processesand avoid degradation and modification of metabolites in the sample Moreuncommon ways to quench the metabolism involve the use of cold methanol,perchloric acid, or nitric acid [56] Harvesting should be performed at the same time

of day for all samples to minimize biological variations due to diurnal changes ofmetabolism The number of replicates is dependent on the experimental sources ofvariation, but since the biological variation generally exceeds the analytical variation,

a minimum of three to six biological replicates per line is recommended [57,58].Technical replicates ensure that the effect of instrument variations during theanalytical run are minimized

1.2.1.2 Homogenization and Extraction

Before extraction of metabolites, the plant tissue has to be homogenized to afinepowder to allow the solvent to penetrate the tissue to extract metabolites effectively.This is typically done using one of the following methods: grinding with a mortarand pestle using liquid nitrogen [44,59], milling in a ball-mill with precooled holders[20], or using ULTRA-TURRAX tissue homogenizers [21,35,60]

The next step in sample preparation is the extraction of plant metabolites, whichhas to be optimized to ensure minimal losses of metabolites due to enzymaticconversion or chemical degradation Blank samples containing only extractionsolution and no metabolite extract should be derivatized along with other samplesand analyzed in each analytical run to identify contaminants, which are thenexcluded from further analysis Additionally, pooled samples are prepared by acombination of aliquots from each biological sample as suggested by Sangster et al.[61] These are used to produce a set of replicates, which are analyzed together withthe real samples at the beginning, at the end, and randomly throughout the

analytical run Therefore, all metabolites of the real samples are present in the(pooled) reference samples, which can be used to normalize the metabolite levels inthe real samples Furthermore, using principal component analysis (PCA), thequality of the data set can be inferred from the clustering of the pooled qualitycontrol samples (see Section 1.2.4) Since the quality control samples are replicates

of the same sample, they should have very similar values for their principalcomponents, which ensure that instrument sensitivity and chromatography duringthe analytical run are not changed significantly

Internal standards are compounds that are not present in the biological sample(e.g., stable isotope-labeled compounds) and are included before or during metabo-lite extraction In the case of targeted analysis, stable isotope-labeled internalstandards that have chemical properties identical with those of the target metabolitesare often used

1.2.1.3 Procedure for Polar Extraction of Metabolites

The procedure is outlined in Figure 1.2 Weigh 30 3 mg (the amount depends onthe origin of the sample and needs to be confirmed for each tissue type) of frozensample plant tissue into a 2 ml soft tissue homogenizing tube with 1.4 mm ceramicbeads (Bertin Technologies) (1), and add 0.5 ml of 100% methanol extractionsolution to the plant sample (2) Record exact sample weights Perform homogeni-zation for 1 30 s at 6000 rpm using a high-throughput tissue homogenizer(Precellys 24, Bertin Technologies) Following incubation for 15 min at 70C in athermomixer at 850 rpm (3), centrifuge the sample for 10 min at 14 000 rpm at roomtemperature (RT) (4) Transfer the supernatant into a new 1.5 ml reaction tube (5a)and add 0.5 ml of 50% aqueous methanol solution containing internal standards

Figure 1.2 Experimental procedure for homogenization and polar extraction of plant metabolites for GC–MS profiling.

8j1 Metabolic Profiling of Plants by GC–MS

(20ml per sample from a stock solution containing 0.2 mg/ml [13C]sorbitol and

1 mg/ml [13C]valine in 100% methanol) to the pellet (5b) After a second nization (6) and centrifugation step (7), pool the supernatants and transfer 50mlaliquots (again, the amount needs to be confirmed for each tissue type for optimalanalysis) into glass vial inserts suitable for GC–MS analysis (8) Dry all resultingaliquots in vacuo using a vacuum concentrator (9) For a subsequent GC–MS analysis,derivatize the sample immediately before analysis (see Section 1.2.3) Note: Prepare asufficient amount of backup samples Store the dried sample aliquots in plastic bagsfilled with silica gel beads at RT For long-term storage, sample aliquots should be keptunder argon to avoid oxidation and degradation of metabolites

homoge-1.2.2

Chemical Derivatization: Methoxymation and Silylation

A variety of derivatizing agents with different properties have been developed,including alkylation, silylation, esterification, and acylation reagents [17,62] Trime-thylsilylation is a commonly used method to derivatize a broad range of metabolites,including sugars, sugar alcohols, amines, amino acids, and organic acids, in orderfor them to become volatile and thermally stable [21] A two-step derivatizationmethod involving oximation followed by silylation is commonly applied for GC–MSmetabolite analysis: First, carbonyl groups are converted into the correspondingoximes using hydroxylamine or alkylhydroxylamine reagents (such as O-methyl-hydroxylamine hydrochloride, MeOx) to stabilize sugars in the open-ring conforma-tion [16,17] (Figure 1.3 a) Oximes exist as two (syn and anti) stereoisomers, andtherefore are often present as two peaks per compound in the chromatograms (denotedMx1 and Mx2) This is followed by trimethylsilylation using silylating reagents such asN-methyl-N-(trimethylsilyl)trifluoroacetamide (MSTFA) þ 1% trimethylchlorosilane(TMCS, a catalyst of the reaction), or alternatively N,O-bis(trimethylsilyl)trifluoroa-cetamide (BSTFA)þ 1% TMCS, which replace active hydrogen in polar functionalgroups such asOH, COOH, NH, and SH with a TMS [Si(CH3)3] group(Figure 1.3b) TMS derivatives are sensitive to moisture, which may cleave TMSderivatives In contrast, tri-tert-butyldimethylsilyl (TBDMS) derivatives, which useN-methyl-N-(tert-butyldimethylsilyl)trifluoroacetamide (MTBSTFA) as a derivatizationreagent, are more moisture resistant [63], but show a significant increase in molecularweight, which may lead to only partial derivatization due to steric hindrance [25].Note: To ensure optimal sample stability, derivatization should be performed immedi-ately before sample injection

1.2.2.1 Procedure for the Chemical Derivatization of Plant Extracts

In our laboratory, a Gerstel MPS2XL GC–MS autosampler performs the tion procedure immediately before injection Add the samples and the derivatizationreagents (MeOx and BSTFA) to a glass vial and then place them in the autosamplertray The autosampler mixes the sample with derivatization reagents automaticallyusing the following program for derivatization using TMS Plant extracts werederivatized for 120 min at 37C using 20ml of MeOx solution (30 mg/ml MeOx

derivatiza-dissolved in pyridine) per sample This was followed by trimethylsilylation with

40ml of BSTFA þ 1% TMCS per sample for 30 min at 37C Finally, 2ml ofretention time standard mixture [0.029% v/v n-dodecane, n-pentadecane, n-nonadecane, n-docosane, n-octacosane, n-dotriacontane, and n-hexatriacontanedissolved in pyridine; Sigma) per sample was added before injection into the GCcolumn Note: To prepare the MeOx solution, weigh 30 mg of MeOx in a reactiontube and after addition of 1 ml of pyridine heat the mixture for 5 min at 50C todissolve the MeOx Store the solution at RT for up to 1 month and avoid moisture.Caution: the derivatization reagents are extremely toxic and should be handledunder a fume hood while wearing gloves

1.2.3

GC–MS Analysis

Components are separated on the basis of differential partitioning between a mobilegas phase (typically helium) and a solid stationary phase (typically based on siliconepolymers), which is bound to the inner surface of a fused-silica tube [18,64] In theion source, analytes are ionized by EI, creating distinct fragmentation patterns foreach component GC–MS traces of plant metabolites are commonly acquired using

Derivatization reagent

Trimethylsilyl amine Amine

tert-butyldimethylsilylether

Carboxyl

Trimethylsilyl ether Hydroxyl

O O

10j1 Metabolic Profiling of Plants by GC–MS

a gas chromatograph coupled with either a single-quadrupole (QUAD), flight (TOF), or ion-trap (TRAP) mass analyzer, which separates the fragment ionsaccording to their m/z values [20,21,50] QUADs are comparably simple but versatilemass analyzers that consist of a set of four parallel metal rods that create anoscillating electricfield when radiofrequency (RF) and DC voltages are applied to therods [65] Ions are separated depending on the stability of their trajectories throughthe electricfield between the four rods GC–QUAD-MS provides a large dynamicmass range of 2–4000 Da/e, but with a mass resolution around 1 : 1500 nominalmass accuracy and slow scan speeds compared with GC–TOF-MS [64,66] Onlyrecently have rapid-scanning QUADs been introduced, offering scan speeds of

time-of-10 000 amu/s [67]

In GC–TOF-MS instruments, bundles of ions are accelerated to high kineticenergy by an electricfield and are separated along a flight tube as a result of theirdifferent velocities, depending on their m/z ratio [50,68] GC–TOF-MS offers ahigher m/z accuracy than conventional GC– QUAD-MS, which is important for theidentification of unknowns [17] Furthermore, GC–TOF-MS gives data acquisitionrates with narrow high-resolution chromatographic peak widths (0.5–1 s), andtherefore allows a higher sample throughput with shorter analysis times comparedwith QUAD- and TRAP-MS [50,66] This is combined with a nominal massresolution similar to that of a QUAD-MS [17]

TRAP instruments work by trapping and sequentially ejecting ions of successivemasses [50] Both QUAD and TRAP instruments are limited by low resolution;however, TRAPs are capable of reaction monitoring, which scans masses slowly over

a predefined mass range to perform a second fragmentation step This can facilitatecompound identification and increases the mass resolution [50,65]

1.2.3.1 Procedure to Acquire GC–MS Data

In our laboratory, GC–MS traces are typically acquired using an Agilent 5975C gaschromatograph coupled with an Agilent Triple-Axis QUAD detector, operated byChemstation software (Agilent) Samples are placed in random order on the sampletray and are analyzed along with several blank and pooled reference samples (seeSection 1.2.1.2) Inject 1ml of derivatized sample into the GC column using a hotneedle technique with a 10ml Hamilton syringe Operate the injector in the splitlessmode isothermally at 230C Use helium as the carrier gas with aflow rate of 1 ml/min Perform chromatographic separation on a 30 m VF-5MS column [with a 10 mIntegra guard column of 0.25 mm i.d., 0.25 nmfilm thickness (Varian)] Fix the MStransfer line to the quadrupole at 280C, the EI ion source at 250C, and the MSQUAD at 100C Tune the mass spectrometer according to the manufacturer’sprotocols using tris(perfluorobutyl)amine (CF43)

Perform GC–MS analysis of plant tissue extracts using the following oventemperature program: set the injection temperature at 70C, followed by a 7C/min oven temperature gradient to afinal 325C, and then hold for 3.6 min at 325C.The GC–MS system is then temperature equilibrated for 1 min at 70C beforeinjecting the next sample Ions are generated by a 70 eV electron beam at anionization current of 2.0 mA and spectra are recorded at 2.91 scans per second with

an m/z scanning range of 50–550 amu Retention time locking (RTL) of thechromatographic peak of mannitol before the sample run ensures repeatableretention times across the systems regardless of operator, detector type, and columnmaintenance Note: For optimal sample analysis, GC–MS settings, including theinjection temperature and the oven temperature gradient, have to be optimized andtailored for each type of plant sample and type of targeted metabolite class(es).1.2.4

Data Preprocessing and Export

After the acquisition of mass spectra, the data sets have to be preprocessed, whichincludes the reduction of background noise, adjusting for baseline shifts andmachine drift, peak alignment, peak detection, and mass spectral deconvolution,before they are subjected to searching against compound databases [69] Softwarepackages for effective in silico data preprocessing include the commercial softwarepackages AnalyzerPro (SpectralWorks), Masshunter (Agilent), Xcalibur (Thermo-Fisher Scientific), and the freely available AMDIS (National Institute of Standardsand Technology, Gaithersburg, MD, USA) (NIST) software (Table 1.1) The softwaredetects component peaks in the chromatograms and calculates the relative amount

by integration of the peak area below the peak, usually relative to the unique m/z ofinternal standards (standardization) [70] To make the data suitable for statisticalanalysis (see Section 1.2.4.3), normalization has to adjust the data for experimentalerrors during sample preparation and changes in instrument sensitivity during theanalytical run Furthermore, retention time index (RI) systems based on eitheralkanes [71] or fatty acid methyl esters [72] are used for correct peak assignment,which depends on the relative elution of a compound between two RI standards.Compounds are identified by matching the RI and mass spectra of each compound,

to minimize false peak assignment due to retention time shifts during the analyticalrun [73] Automated calculation of the RI for all compounds and automated massspectral deconvolution are implemented in most current software packages.1.2.4.1 Procedure for Postacquisition Data Preprocessing

In this section, the data processing procedure using the commercially availableAnalyzerPro software package (SpectralWorks, current version: 2.5.1.7) with thefully integrated NIST05 mass spectral search program (NIST) is described.1) Import all datafiles into the AnalyzerPro (.swx) format

2) Create a manual RI ladder by creating a .csv file with alkane specifications(name/RI/RT)

3) Set up qualitative data processing of all datafiles of the pooled reference samplesusing a number of parameters in the“Processing Method” of AnalyzerPro Fortargeted analysis, the use of the default settings is recommended: minimummasses¼ 4; area threshold ¼ 500; height threshold ¼ 1%; signal-to-noise ratio

¼ 3; width threshold ¼ 0.01 min; resolution ¼ very low; scan windows ¼ 3;smoothing¼ 3 The masses of m/z 73 (TMS), 147 (TMS-O-DMS), and 207

12j1 Metabolic Profiling of Plants by GC–MS

(column bleed) appear in nearly every plant chromatogram after derivatizationwith MTSFA and therefore have to be excluded from further analysis Leave thebox for target component searching unchecked, and specify a library and theconfidence threshold for targeted analysis Note: It is important to adjust peakpicking parameters according to the quality of the chromatogram (e.g., peakwidth, signal-to-noise ratio, and resolution) to be able to pick as many compo-nents as possible that are present in the biological sample; furthermore,deconvolution and peak picking parameters have to be optimized to avoid falsepositives (for a review on the quality of peak picking using different softwareprograms, see [79]).

4) Generate a target component library (TCL) of the pooled reference samplefilewith the most deconvoluted components The TCL contains a list of (identified)components of this representative chromatogram and has to be additionallyspecified by the target ion and ion ratios of the second and third most abundantions (fill in manually) Additionally, perform background subtraction by choosing

a blank sample datafile to remove contaminants and components not present inthe biological sample

5) Enable and configure the Matrix Analyzer plug-in Enable the box for targetcomponent searching Match all components found in the other chromatogramsagainst the TCL using the same initial parameter settings

6) After processing of all data files, the Matrix Analyzer plug-in report can beaccessed via the“reports” tab Save the data matrix in one of the specified formats(.csv or .xls) for further data mining (see Section 1.2.4.2)

7) Control the quality of the raw data Ensure that peaks are accurately identifiedand peak areas are correctly integrated

8) Normalize the data by dividing the integrated peak areas of all detected lites by the peak area of the internal standard and by the sample weight (ingrams)

metabo-1.2.4.2 Data Analysis and Statistics

Following data preprocessing and normalization, the data are typically logarithmtransformed to minimize possible effects of outliers [80,19] Subsequently, effectivestatistical discriminant analysis is applied to the data set to extract biologicallyrelevant information This aims tofind patterns or relationships within the data toextract the information needed to generate scientific hypotheses, which have to befurther tested using Student’s t-test and analysis of variance (ANOVA) Metabolitedata can be mined using different pattern recognition methods to separate the datainto classes, either knowing that classes exist, using supervised learning algorithms,

or in the absence of any advanced knowledge, using unsupervised learningalgorithms [81]

Univariate analysis is the simplest statistical method and is carried out with onlyone variable at a time Basic univariate statistical measures are mean, variance,standard deviation, covariance, and correlation [82] Multivariate statistics deal withthe analysis of multiple variables simultaneously, and include unsupervised classi-fication methods such as PCA, hierarchical cluster analysis (HCA), and self-

14j1 Metabolic Profiling of Plants by GC–MS

organization mapping (SOM) and supervised approaches such as partial leastsquares (PLS) to classify metabolites [15] HCA and PCA are most widely usedfor comparison and visualization of similarities and differences between data sets.Additionally, tools displaying data sets on metabolic pathway maps are often used

to visualize metabolic profi les, and can also be combined with gene expressionpro files [83]

1.2.4.3 Procedure for Postacquisition Data Analysis

There is a huge amount of different commercially and freely available softwarepackages to explore data sets statistically Many statistical tests and classificationmethods, including PCA, PLS, and HCA, can be performed using The Unscramblerstatistical software (CAMO) or using scripts integrated in the R software environ-ment (www.r-project.org) Furthermore, normalization using internal standards andsample weight, log transformation, and statistical analyses can be performed usingdesigned R scripts and Excel macros that are well documented and freely available

at Metabolomics Australia (http://code.google.com/p/ma-bioinformatics/) Onlyrecently have web-based metabolomic data tools such as MetaboAnalyst beenmade available; this combines several complex data analysis techniques includingdata processing, normalization, statistics, and pathway mapping and is freely available

on a web server (www.metaboanalyst.ca; [84]) Further information is available inseparate chapters on data analysis and multivariate statistics (Sun and Weckwerth,Chapter 16) and metabolite clustering and visualization (Kaever et al., Chapter 14)

1.3

Applications of the Technology

Numerous applications have been reported in which GC–MS-based metabolomicshas been used to investigate metabolites and pathways that are differentiallyregulated due to genetic or environmental perturbations There have been extensivereports and reviews describing how GC–MS-based metabolite profiling has beenemployed to study plant metabolism in great detail [19,50,85] Here, we mention just

a few examples of research areas where metabolomics has already made acontribution

Metabolite profiles generated by GC–MS can be used as biochemical readouts toclassify organisms according to genetic and environmental stimuli and to identifythe differences and similarities between the different conditions As described inthis chapter, GC–MS can generate hundreds of data points and, regardless ofwhether those data points can be referred to a known metabolite or not, the presenceand relative abundance of those data points can be related to genetic background andenvironmental conditions similarly to a signature DNA sequences are still thestandard used for the identification of genetically different individuals However, it

is known that the biochemical readout of individuals even with similar genomes will

be different with environmental changes Therefore, metabolomics has already beensuccessfully applied to classify genetically similar individuals grown at different

locations (provenance) or under different conditions An example where olomics has been used to determine the geographical origin of samples waspresented by Choi et al [86], who used 1H NMR-based metabolite profiling incombination with multivariate analysis to classify 12 Cannabis sativa cultivars based

metab-on the regimetab-on they were grown (see also Chapter 3 by Choi et al.) Extensive studies

on comparisons of metabolite profiles of plants grown in different conditions havebeen carried out For example, the metabolomes of plants grown in unfavorableconditions such as abiotic and biotic stresses can increase our understanding of howplants respond and adapt to harsh environments Researchers aim to understandhow plants have evolved mechanisms to deal with stress and especially how someplants perform better than others Abiotic stresses including cold, frost, heat,drought, and salinity cause massive losses in crop yields every year An under-standing of stress tolerance mechanisms and the transfer of those mechanisms tocommercial crop varieties will reduce agricultural losses Contributions made bymetabolomics approaches to learning about the physiology and biochemistry ofplants in different stress conditions have been reported, for example, for cold andheat stress in Arabidopsis [87], salinity in rice [88], Lotus japonicus [89], and barley [90],and water deficiency in Arabidopsis [91]

Metabolomics as a tool to characterize a plant chemically is becoming increasinglyimportant for risk assessments of genetic modifications Genetic alterations canhave an impact not only on the visible phenotype but also on the biochemicalcomposition of the cells, potentially leading to effects that are unexpected on thebasis of current genetic or biochemical knowledge [92,93] There have been anumber of reports where the introduction or deletion of a gene has alteredmetabolism and therefore metabolite concentrations compared with wild-typecontrols [32,33,43] It has also been demonstrated that the introduction of thesame gene into different species could result in differential changes of themetabolomes [43] A substantial equivalence concept is a framework for safetyevaluations where existing crops and foods are taken as the baseline considered asbeing safe, and the properties of any new foods and crops are compared with thebaseline Therefore, it is important to monitor the metabolomes (and all other cellproducts) of genetically engineered plants and compare them with the naturalvariation of metabolomes of their wild-type counterparts [93,94]

The last example mentioned here is the application of metabolomics in breedingand quantitative trait loci (QTLs) analysis, which is recognized to have enormouspotential Often agronomic traits are controlled by many genes or QTLs potentiallyresiding on different chromosomes but their expression works together as a networkdetermining that particular phenotype or trait Especially if the trait of interest isbased on a metabolite of interest (e.g., vitamins or essential amino acids), theutilization of metabolomics as a strategy to link phenotypes with QTLs has alreadybeen demonstrated in a number of different species [46,47,95] Now that metab-olomics technologies have become faster and cheaper, it is possible to analyze hugenumbers of compounds simultaneously in a large genetic mapping population Thisnew approach of combining conventional genetic methods such as QTL mappingwith omics technologies such as transcriptomics, proteomics, and metabolomics,

16j1 Metabolic Profiling of Plants by GC–MS