PLACE IN RETURN BOX to remove this checkout from your record.
TO AVOID FINES return on or before date due.
MAY BE RECALLED with earlier due date if requested.

 

DATE DUE DATE DUE DATE DUE

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

5/08 K:lProj/Aoc&Pres/CIRCIDateDue.indd

 

PROFILING OF SPECIALIZED METABOLITES IN GLANDULAR TRICHOMES OF
THE GENUS SOLAN UM USING LIQUID CHROMATOGRAPHY AND MASS
SPECTROMETRY

By

Feng Shi

A DISSERTATION

Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of

DOCTOR OF PHILOSOPHY
Chemistry

2009

ABSTRACT

PROF ILING OF SPECIALIZED METABOLITES IN GLANDULAR
TRICHOMES OF THE GENUS SOLANUM USING LIQUID
CHROMATOGRAPHY AND MASS SPECTROMETRY

By
Feng Shi

The plant kingdom synthesizes thousands of biologically active secondary metabolites,
but our understanding of the fimctions of genes responsible for regulating their
biosynthesis is far from complete. Existing knowledge of the diversity of plant chemistry
is also remarkably limited, but an emerging approach known as metabolomics offers the
potential for explosive growth in this area of research. Accelerating the pace of
discoveries in this realm will require more rapid and powerful analytical tools capable of
identifying and quantifying from hundreds to thousands of secondary metabolites using
nontargeted metabolite profiling. Since the majority of plant metabolites are either polar
or nonvolatile molecules, efforts to push the performance limits of coupling liquid
chromatographic separations with mass spectrometry (LC/MS) are bringing about a new
“Golden Age” of plant biochemistry.

This thesis describes an approach that takes advantage of fast spectrum acquisition
available with time-of-flight mass spectrometers to extend nonselective collision induced
dissociation (CID) by performing quasi-simultaneous acquisition of mass spectra using
multiple collision conditions. This approach, termed multiplexed CID (muxCID),
generates accurate mass measurements of molecular and fragment ion species with the
additional benefit that the dependence of each ion’s abundance upon collision energy is
obtained throughout the entire course of an LC/MS analysis. The additional information

obtained from ion breakdown curves, coupled with accurate mass measurements, aids

 

assignments of ions as molecular, fragment, noncovalent oligomer, or adducts. Coeluting
metabolites are resolved using collision energy-selective fragmentation of individual
metabolite classes.

This multiplexed CID approach has been adapted for use with fast ultraperformance
fused-core LC separations to provide an analytical platform capable of measuring about
100 plant metabolites per minute of instrument time. Rapid profiling of secondary
metabolites was performed for extracts of specialized epidermal structures called
glandular trichomes for numerous relatives of tomato plus plant lines derived from
tomato and a wild relative. This technology has made it possible to screen about 300
samples within 30 hours. Automated data processing strategies yielded measurements of
more than 1500 analytical signals. Multivariate statistical analysis of the metabolite
profiles revealed five specific phenotypes differing in abundances and profiles of
bioactive compounds including acylsugars, glycoalkaloids,and flavonoids, including
several novel metabolites. Mass spectra generated using collision induced dissociation
also provided important information regarding metabolite structures that have opened
windows into the metabolic diversity within the plant genus Solanum. This work was
extended to use metabolite profiles to help guide efforts to breed tomato plants rich in
anti-insect acylsugar metabolites.

The analytical methods and approaches described herein will serve to expand the
potential of LC/MS for metabolite identification in metabolomics and lead the way to

speeding up discoveries of gene functions in plants and other organisms.

ACKNOWLEDGEMENT

Many people deserve thanks and appreciation for this thesis. First and foremost, I own
my sincerest gratitude to my advisor Dr. A. Daniel Jones for his continuous support of
my Ph.D research. As greatest advisor and mentor in my life ever, his patience,
knowledge, motivation and consideration guide and light my journey up to today. I could
not imagine my research without him.

A special thank from me to my outstanding collaborator Dr. Rob Last, also the
principle investigator of the project, for his invaluable advice and critiques, which would
benefit not only the research but my feature career. I also would like to thank my other
collaborators Dr. Gregg Howe, Dr. Martha Mutschler from Cornell University, Dr. Eran
Pichersky from University of Michigan and Dr. David Gang from University of Arizona.
It is an honor to learn from them and work with them. All the open minded discussions
are memorable.

I also would like to acknowledge the postdocts Dr. Jin-ho Kang, Dr Tony Schilmiller
in this project for being good colleagues with me. We have great collaborations to make
this project going smoothly and create the foundation for this thesis.

It is a pleasure to thank my committee: Dr. Gavin Reid, Dr. Merlin Bruening and Dr.
Kevin Walker, for their insightful comments and criticism, and encouragement.

Specially, I would like to express my thanks to my coworkers Mike, Ruth, Siobhan,
Chao, Ramin, Behnaz, Bao, xiaoli from Jone’s group and Jeongwoon, Amanda from
Last’s group for all the assistant and happiness they gave to me. I would also like to thank

Lijun and Bev from mass facility for their help during these years.

Last but never the least, my deeply thank to my family: my husband Hongyang, Li for
always supporting me and sharing the happiness and difficulties with me; my little
daughter Sophia for being such a sweet baby and bringing so much joy into our life; my
parents and my parents in law, for taking care of the baby and helping us with the
housework. I am really grateful to have them around me. Their unconditional love and
support accompany with me through these tough years and make this thesis going so well.

I definitely could not work through my Ph.D. without them.

TABLE OF CONTENTS

List of Tables ............................................................................ x
List of Figures ......................................................................... Xii
List of Abbreviations .............................................................. xxiii
Chapter 1 Introduction ................................................................. 1
1.1 The Emergence of Metabolomics in Biological Research °°°°°°°°°°°°°°°°°°°° 3
1.2 Challenge of Metabolomics ......................................................... 6
13 Techniques for Metabolomics ....................................................... 7
1.4 Multivariate Statistic tools for Metabolomic Analyses °°°°°°°°°°°°°°°°°°°°° 11
1.5 Metabolite Identification ........................................................... 16
1.6 Secondary Metabolites in Solanum Glandular Trichomes """""""""""""""""" 18
17 Summary Of Research Goals ....................................................... 21

Chapter 2 Discovery and deep profiling of metabolites using liquid
chromatography/multiplexed collision-induced dissociation (muxCID)/mass

spectrometry .............................................................................. 23
2.1 Introduction ......................................................................... 24
2.2 Experimental Section ............................................................... 26

2.2.] Materials ........................................................................ 26
2.2.2 Chemicals ...................................................................... 26
2.2.3 Extraction Method .............................................................. 27
2.2.4 Liquid Chromatography ....................................................... 27
2.2.5 Mass Spectrometry ............................................................. 27
2.3 Results and Discussion ............................................................. 28
2.4 Conclusions ......................................................................... 63

Chapter 3 Annotation of Acylsugar metabolites in glandular trichomes from
accessions of the genus Solanum based on LC/TOF MS coupled with

multiplexed CID ......................................................................... 65
3.1 Introduction ......................................................................... 66
3.2 Experimental Section ............................................................... 68

3_2_1 Materials ........................................................................ 69
3.22 Chemicals ...................................................................... 69
3.2.3 Extraction Method .............................................................. 69
3.2.4 High-Performance Liquid Chromatography """""""""""""""""""""""" 69

vi

3.2.5 Mass Spectrometry .............................................................. 69

3.26 Characterization of Metabolites ................................................ 70
3.2.7 Nomenclature ................................................................... 70
3.3 Results and Discussion .............................................................. 71
3.3.1 Structure annotation for glucose trimesters in S. pennellii LA0716 °°°°°°°°°° 71
3.3.2. Structure annotation for glucose trimesters and sucrose trimesters in S.
pennellii LA1522 ................................................................ 90

3.3.3. Structure annotation of sucrose tetraesters in S. habrochaites LA1777°113
3.3.4 Structure annotation of sucrose triesters and tetraesters in S. habrochaites

LAI353 ........................................................................ 127
3.4 Conclusions ........................................................................ 1 3 4

Chapter 4 Annotation of polyphenol, glycoalkaloid, and terpenoid
secondary metabolites in glandular trichomes from tomato, its wild relatives,
and near isogenic lines, based on LC/TOF MS coupled with multiplexed

CID ........................................................................................ 136
41 Introduction ........................................................................ 137
4.2 Methods ............................................................................ 141

421 Materials ....................................................................... 14]
4.2.2 Chemicals ...................................................................... 14]
423 Extraction Method ............................................................. 141
4.2.4 High-Performance Liquid Chromatography °°°°°°°°°°°°°°°°°°°°°°°°°°°°°° 142
425 Mass Spectrometry ............................................................ 142
4.3 Results .............................................................................. 143
4.3.] Structural annotation for polyphenols ......................................... 143
432 Structural annotations for glycoalkaloids ..................................... 163
433 Structural annotation of terpenoids ............................................ 176
4.4 Conclusions ........................................................................ 184

Chapter 5 Fast LC/time-of flight mass spectrometry for screening metabolic

phenotypes for tomato chromosomal substitution lines ----------- 185
5.1 Introduction ........................................................................ 186
5.2 Experimental Method ............... . ............................................... 189

5.2.] Plant Growth Conditions ...................................................... 189
522 Plant Extractions ............................................................... 190
5.2.3 Chromatography ............................................................. 190
5.2.4 Mass Spectrometry ............................................................ 190
5.2.5 Chemometric Data Analysis ................................................... 191
5.3 Results and Discussion ............................................................ 19]
5.3.]. Analysis of Chromosome Substitution Lines Using LC/TOF MS °°°°°°°° 191

vii

Ill. .

‘IIJ

5.3.1.1 ILs 1-3 and 1—4 are missing an acetyl group on major acyl sucrose

metabolites ................................................................. 204
5.3.1.2 Two introgression lines with lower total acylsugars """""""""""""" 207
5.3.1.3 ILs 8-1 and 8-1-1 causes a shift in acyl chain lengths without altering

nurnbers of substitutions ................................................... 210
5.3.1.4 Discovery of differences in accumulation of metabolites of lower

abundance ILs 1-1 and 1-1-3 with novel glycoalkaloids °°°°°°°°°°°°° 217

5.3.1.5 ILs 6-3 and 6-4 with higher ratio of tri- to dimethylmyricetin" ° ' ' ° “"223
5.3.2 F2 plants are screened to determine wether phenotypes dominant or recessive

for recurrent parent M82 ...................................................... 225
5.4. Conclusions ....................................................................... 228

Chapter 6 Qualitative and quantitative profiling of acylsugar metabolic
phenotypes for tomato breeding lines using liquid chromatography/time-of-

flight mass spectrometry .............................................................. 230
6.1 Introduction ........................................................................ 231
62 Experimental Section .............................................................. 233

6.22 Chemicals ...................................................................... 233
6.2.3 Preparation of internal standard (3-decanoyl-glucofuranose) °°°°°°°°°°°°° 233
6..24 Extraction Method ............................................................. 234
6.2.5 High-Performance Liquid Chromatography °°°°°°°°°°°°°°°°°°°°°°°°°°°°° 234
62.6 Mass Spectrometry ............................................................ 235
6.2.7 Gas Chromatography/Mass Spectrometry .................................... 235
6.2.8 Chemometric data analysis .................................................... 236
6.2.9 Extraction and isolation of S 4.17 ............................................. 237
63 Results and Discussion ............................................................ 237
6.3.1 Annotation of acylsugar from conventional tomato Solanum lycopersicum
M82 ............................................................................. 237
6.3.2 Compare total amounts of acylsugars and distributed amounts of acylsugars
using LC /MS .................................................................. 246

6.3.3 Fatty acyl substituents of acylsugars among wild type S. pennellii LA0716,
LA 1522, S. habrochaites LA1777 and cultivated tomato M82 using
GC/MS ......................................................................... 249

6.3.4 Compare total amounts of acylsugars and distributed amounts of glucose
triesters, sucrose triesters and sucrose tetraesters, and fatty acyl substituents of

acylsugars among Cornell breeding lines using LC/MS °°°°°°°°°°°°°°°°°°°°° '25]

6.4 Conclusions ........................................................................ 271

Chapter 7 COflCluding Remarks .................................................... 273

Appendix ............................................................................... 278
viii

Bibliography ........................................................................... 284

Tat

LIST OF TABLES

Table 1-1. MarkerLynx software first detects peaks using specific mass data, aligns the
peak according to retention time, integrates peaks over all samples, and

assembles the results into a table ............................................... 13

Table. 3-1. Fragments and fatty acid constituents for detected glucose triesters from S.
pennellii LAO716 in negative and positive mode (CID potential, 25 V) °°°° 88

Table 3-2. Exact mass measurements of [M+HCOO]' ions of detected glucose triesters
from S. pennellii LAO716 detected using multiplexed CID with collision

potential (10 V) .................................................................. 89

Table. 3-3. Fragment ions and fatty acid constituents of detected glucose triesters in an
extract of S. pennellii LA1522 leaf trichomes using negative and positive

mode electrospray ionization (CID potential, 25 V) °°°°°°°°°°°°°°°°°°°° 94

Table.3-4. Fragments and fatty acid constituents for detected sucrose triesters in S.
pennellii LA1522 using ESI negative (CID potential, 55 V) and positive mode

(CID potential, 40 V) ......................................................... 108

Table 3—5. Exact mass measurement of [M+HCOO]' Ions for detetable sucrose triesters in
accession LA1522 detected using ESI negative mode under CID potential 10

V ................................................................................ 109

Table 3-6. Fragments and fatty acid constituents of detected sucrose tetraesters from S.
habrochaites LA1777 using ESI positive mode (CID potential, 40 V) and

negative (CID potential, 55 V) ............................................... 124

Table 3-7. Exact mass measurement of [M+HCOO]’ ions for detected sucrose tetraesters
in S. habrochaites LA1777 detected using 1381 negative mode under CID

potential 10 V .................................................................. 125

Table 3-8. Fragments and fatty acid constituents of detected acylsugars from S.
habrochaites LA1353 using ESI positive mode (CID potential, 40 V) and

negative (CID potential’ 55 V) ................................................. 129

Table 5-1. The detectable acylsugars for ILs were listed with retention time (min) and
molecular mass for [M+HCOO]' ............................................. 208

Table 6-1. Structure of S 4:17 in Solanum chopersicum M82 characterized by NMR (1H,
13
C, DEPT, gI-IMQC, gHMBC, gCOSY, and TOCSY). Chemical shifts of

Til

l3 _ . .
proton and C resonances are listed Wlth proton-proton coupling
constants ooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooo 242

Table 6-2. Acyl sucrose metabolites detected in trichome extracts from S. lycopersicum
M82 and their adduct ions and fragment ions observed using ESI negative

(CID potential 55 V) and positive mode (40 V) °°°°°°°°°°°°°°°°°°°°°°° 246

Table 6-3. Fatty acyl constituents for detected acylsugars in 8-1X0716 using ESI negative
mOde With CID potential at 55 V .............................................. 259

Table 6-4. Fatty acyl constituents for detected acylsugars in 8-2X0716, 3-2><0716 and 7-
5XO716 using ESI negative mode with CID potential at 55 V """""""""" 260

Table 6-5. Fatty acid constituents of detected sucrose triesters from fixed lines using ESI
negative (CID potential, 55 V) ................................................ 269

Table 6-6. Fatty acid constituents of detected acylsugars from crosses between fixed
lines and S. pennellii LAO716 sing ESI negative (CID potential, 55 V) ”'270

Table.A-l Description of breeding lines from Cornell including introgression lines, fixed
lines, crosses between introgression lines and S. pennellii LAO716, and new

fixed lines ...................................................................... 279

Table.A-2 Description of breeding lines from Cornell including S. pennellii LAO716,
normal tomato lines, fixed lines and crosses between fixed lines and S. penellii

0716 ............................................................................ 280

xi

FL

LIST OF FIGURES

Figure 2-1 Schematic of LC/multiplexed CID method showing (A) application of
multiple collision potentials to the ion transit lens to effect nonselective ion
fragmentation, (B) mass spectra acquired using five different quasi-
simultaneous CID conditions, showing molecular (*), noncovalent
oligomer (O), and fragment ion species ([1) using color-coded symbols to
distinguish ions coeluting metabolites, and (C) breakdown curves showing
common behavior for molecular ion species for two coeluting metabolites, a

noncovalent dimer ion, and a fragment ion ' ' ' '° " '30

Figure 2-2 Negative mode electrospray ionization mass spectra obtained using
multiplexed collision induced dissociation using five aperture 1 voltages
(from bottom to top: 10, 25, 40, 55 and 80 V) for two metabolites extracted

from S pennellii LAO716 by leaf dip ........................................ 33

Figure 2-3 TIC (total ion chromatograph) for S. habrochaites LA1777 (top) and S.
pennellii LAO716 (bottom) using ESI negative with CID potential at 10
V ............................................................................... 35

Figure 2-4 Breakdown curves showing energy dependence of various ion abundances for
metabolites extracted from leaf dips of S. pennellii LAO716 and S.

habrOChaites LA1777 ........................................................ 36

Figure 2-5 A: Breakdown curves derived from metabolites detected during LC/negative
ion ESI MS analyses of a leaf dip extract from S. habrochaites LA1777° ' '39

Figure 2-6 Breakdown curves derived from LC/negative ion ESI MS analyses of a leaf
dip extract from S pennellii LAO716 ......................................... 40

Figure 2-7 Breakdown curves for the flavonoid glycoside rutin obtained using negative
mode electrospray ionization following injections of 10 ul of solutions at

five di fferent concentrations .................................................. 42

Figure 2-8 Multiplexed CID mass spectra (electrospray, negative mode) of coeluting
metabolites from LC/TOF MS analysis of an extract of S. pennelli LAO716,

showing spectra obtained from three of the five Aperture 1 potentials '''' 45

Figure 2-9 Total ion chromatography (TIC) for ILl-l (Bottom) and XIC of m/z 609,
1076 (elute at 2.46 min), 1094 and 593 (elute at 2.60min) under ESI

negative mOde ................................................................. 49

Figure 2-10 Averaged mass spectra across the LC/MS TIC peak eluting at 2.46 min
using three CID voltages (10’ 55 and 80 V) ................................. 50

xii

Figure 2-11 XICs for the ions labeled with o and 0 from Figure 2-10 were generated for
LC/MS analysis Of an extract Of IL 1-1 ...................................... 51

Figure 2-12 Negative mode mass spectra from an extract of IL-1-1 trichomes. Spectra
were generated for three CID voltages (10, 55, and 80 V) by averaging

spectra across the TIC peak eluting at 2.60 min °°°°°°°°°°°°°°°°°°°°°°° 54

Figure 2-13 Breakdown curves showing energy dependence of various ion abundances
for glycoalkaloid metabolites eluting at 2.46, 2.60 and 2.70min extracted
from leaf dips of transgenetic plant. Ion abundances are normalized to the
total ion current for each individual ion as summed for all five collision

potentials ...................................................................... 5 5

Figure 2-14 Breakdown curves showing energy dependence of various ion abundances
for flavonoid metabolites eluting at 2.46 and 2.60min extracted from leaf

dips 0f transgenetic plant ...................................... . .............. 57

Figure 2-15 Concentration dependence of extracted ion chromatogram (XIC) peak areas
for ions derived from rutin at different collision potentials using 10 pl
injections of rutin standard solution and negative mode electrospray

ionization ...................................................................... 60

Figure 2-16 (Top) TIC of S. habrochaites LA1777 and XIC of m/z 233 under highest
CID voltage in ES] negative mode. (Bottom)Breakdown curves derived
from LC/negative ion ESI MS analyses of a leaf dip extract from S.
habrochaites LA1777 corresponding to three isomers A (O), B (I) and C
(A) of deprotonated sesquiterpene acid ([M—H]' ions at m/z 233) showing

isomer-dependence of signal upon collision potential °°°°°°°°°°°°°°°° 62

Figure 3-1 Generalized structure of triacylglucose metabolites from S. pennellii LAO716
as identified by Burke et a]. ................................................... 72

Figure 3-2 Extracted ion chromatograms (XICs) generated from an extract of leaf tissue
from S. pennellii LAO716 of ions of m/z values corresponding to
[M+HCOO]— for glucose triesters with total fatty acid carbon atoms ranging

from 12 to 22 ............................................................ 74

Figure 3-3 Proposed fragmentation pathway of the formate adduct of triacylglucose
S3218 extracted from S. pennellii LAO716 (RT = 21.11 min) using ESI

negative mOde with CID voltage 25 V ....................................... 78

Figure 3-4 Electrospray ionization mass spectra of G 3:18 (RT = 27.11 min) extracted
from S. pennellii LAO716 using an Aperture 1 CID potential of 25 V in

negative (top) and 40 V in positive (bottom) mode °°°°°°°°°°°°°°°°°°°°° 81

xiii

Figure 3-5 Product ion MS/MS spectrum of m/z 519 ([M+HCOO]_) from triacylglucose
G 3:18 from S. pennellii LAO716 using collision energy 20 V """"""""" 82

Figure 3-6 Top: Product ion MS/MS spectrum of m/z 492 ([M+NH4]+) from
triacylglucose G 3:18 from S. pennellii LAO716. Bottom: product ion

MS/MS spectrum for m/z 457 ([M+H-H20]+) for G 3:18; m/z 457 was
generated in source ............................................................ 83

Figure 3-7 (Top) Extracted ion chromatogram of m/z 547 (G 3:20) from LC/MS analysis
of an extract of S. pennellii LAO716 under CID potential 10 V in ESI
negative mode. Chromatographic peaks corresponding to four isomers
detected as m/z 547 were resolved. (Bottom) ESI mass spectra obtained

from these four isomers at 25 V in negative mode (1-4) """"""""""""" 85

Figure 3-8 Fatty acyl groups from total acylsugars in an extract of S. pennellii LAO716
were transesterified to form fatty acid ethyl esters, and analyzed using
GC/MS. Data for each fatty acid were calculated as a percentage of the total
ion current chromatogram peak areas of fatty acids detected through GC/MS.

“‘H

The abbreviations 1 and “ai” refer to iso- and anteiso- branched
isomers OOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOO 90

Figure 3-9 Extracted ion chromatograms (XICs) generated from an extract of leaf tissue
from S. pennellii LA1522 for ions corresponding to [M+HCOO]_ for

glucose triesters with total fatty acid carbon atoms ranging from 12 to 22 '92

Figure 3-10 Extracted ion chromatograms (XICs) generated from an extract of leaf tissue
from S. pennellii LA1522 of ions of m/z values corresponding to
[M+HCOO]' for sucrose triesters with total fatty acid carbon atoms ranging

from 12 to 23 ............................................................ 96

Figure 3-11 Breakdown curves derived from LC/negative ion ESI MS analyses of a leaf
dip extract from S. pennellii LA1522 showing collision energy dependence
of abundances of: (A)'formate ion adducts of three triacylsucrose
metabolites ([M+HCOO]_ at m/z 737 83:22 (A); [M-H]_ at m/z 691 (0);

C12 fatty acid at m/z 199 (I); C5 fatty acid (El) ............................ 99

Figure 3-12 (TOP) ESI spectrum of triacylsucrose 83:22 (RT = 36.50 min) extracted
from S. pennellii LA1522 using CID potential of 55 V. (Bottom) Product
ion MS/MS spectrum for formate adduct of 83:22 (products of m/z 737.39

using collision potential of 30 V) ........................................... 100

Figure 3-13 Positive mode ESI spectrum of triacylsucrose S 3:22 (RT = 36.50 min)
extracted from S. pennellii LA1522 using CID potential of 55 V """"" 103

xiv

Figure 3-14 (Top) Product ion MS/MS spectra of (Top) [M+NH4]+ at m/z 710.43 and
(Bottom) [M+Na]+ at m/z 715.39 for S 3:22 extracted from S. pennellii
LA1522 ....................................................................... 103

Figure 3-15 Proposed fragment pathway of formate adduct of S 3:22 (RT = 36.50 min)
extracted from S. pennellii LA1522 using ESI negative mode with CID
voltage 55 V. Specific substitution positions of individual fatty acids remain

uncertain ...................................................................... 1 04

Figure 3-16 Proposed fragment pathway for sodium adduct of S 3:21 extracted from S.
pennellii LA1522 based on MS/MS product ion spectrum. Specific

substitution positions of individual fatty acids remain uncertain """"""" 106

Figure 3-17 Proposed fragment pathway of ammonium adduct for S 3:22 (RT = 36.50
min) extracted from S. pennellii LA1522 using ESI positive mode with CID
voltage 40 V. Specific substitution positions of individual fatty acids remain

lmcertain ..................................................................... 1 07

Figure 3-18 Fatty acyl groups from total acylsugars in an extract of S. pennellii LA1522
were transesterified to form fatty acid ethyl esters, and analyzed using
GC/MS. Data for each fatty acid were calculated as a percentage of the total
ion current chromatogram peak areas of fatty acids detected through

GC/MS ....................................................................... 111

Figure 3-19 Multiplexed CID mass spectra (electrospray, negative mode) of coeluting
metabolites from LC/T OF MS analysis of an extract of S. pennellii LA1522

showing spectra obtained from three of the five Aperture 1 potentiaIS' ' "113

Figure 3-20 (A) Extracted ion chromatograms (XICs) fi'om negative ion mode LC/MS
analyses of (A) S. pennellii LA1522 and (B) S. habrochaites LA1777.
Selected m/z values correspond to [M+HCOO]— of sucrose triesters
(highlighted with blue dashed lines) and sucrose tetraesters (red dashed

line) ........................................................................... 1 15

Figure 3-21 (Top) Negative mode ESI spectrumof tetraacylsucrose S 4:21 (RT = 31.50
min) under CID potential 55 V and (Bottom) product ion MS/MS spectrum
of formate adduct (m/z 737.35) for S 4:21 from S. habrochaites

LA1777 ...................................................................... 117

Figure 3-22 Positive mode ESI spectrum of tetraacylsucrose S 4:21 (RT = 31.50 min)
from S. habrochaites LA1777 using CID potential 40 V """"""""""" 119

Figure 3-23 (Top) Product ion MS/MS spectrum of [M+NH4]+ at m/z 710.40 and
(Bottom) product ion MS/MS spectrum for m/z 317 ([FRU+acyl]+) of S 4:21

XV

from S. habrochaites LA1777; m/z 317 was generated in source from
tetraesters 84:21 from S. pennellii LA11777 °°°°°°°°°°°°°°°°°°°°°°°° 120

Figure 3-24 Collision induced dissociation of [M+NH4]+ acylsugar ions yield a dominant
acylfructofuranose fragment ion, and a less abundant acylglucopyranose

fragment ion ................................................................. 1 2 1

Figure 3-25 Proposed fragment pathway upon CID of the formate adduct for S 4:21 (RT
= 31.5 min) extracted from S. habrochaites LA1777 using ESI negative

mode With CID voltage 5 5 V ................................................ 122

Figure 3-26 Proposed fragment pathway of ammonium adduct for S 4:21 (RT = 31.5 min)
extracted from S. habrochaites LA1777 using ESI positive mode with CID

voltage 55 V ................................................................. 123

Figure 3-27 Fatty acyl groups from total acylsugars in an extract of S. habrochaites
LA1777 were transesterified to form fatty acid ethyl esters. and analyzed
using GC/MS. Levels for each fatty acid were calculated as a percentage of
the total ion current chromatogram peak areas of fatty acids detected

through GC/MS .............................................................. 1 27

Figure 3-28 (Top) Negative mode ESI spectra of S 5:25 (RT = 31.91 min) and (Bottom)
S 4:26 (RT = 36.04 min) from S. habrochaites LA1353 using CID potential

55 V .......................................................................... 130

Figure 3-29 Putative structures for detected acylsugars among S. pennellii LAO716 and
LA1522 and S. habrochaites LA1777 and LA1353. Specific substitution
positions of individual fatty acids remain uncertain for some

metabolites ................................................................... 1 3 2

Figure 4-1 Generalized structure of flavonoid metabolites known in tomato and its
relatives, where R groups can be hydrogen, hydroxyl, or other substituted

oxygen-containing groups ................................................... 1 38

Figure 4-2 Generalized structure of glycoalkaloid metabolites known in tomato and its
relatives, where R groups are glycosides .................................... 139

Figure 4-3 Putative structures for several flavonoid derivatives from Solanum wild
species oooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooo 145

Figure 4-4 ESI mass spectra for quercetin-3-rutinoside from S. pennellii LAO716 using
positive mode With CID 40 V ............................................... 146

Figure 4-5 (Top) XIC for m/z 609 from S. pennellii LAO716 using ESI negative mode
with CID voltage 10 V. (Bottom) ESI spectrum of quercetin-3-rutinoside

(RT = 11.52 min) using ESI negative mode with CID voltage 55 V ''''' 147

XV 1

Figure 4-6 Proposed fragmentation pathway for quercetin-3-rutinoside from S. pennellii
LAO716 using ESI negative mode with CID potential 55 V °°°°°°°°°°° 148

Figure 4-7 (Top) Two isomeric metabolites appear in the extracted ion chromatogram
m/z 609 for an extract of S. habrochaites LA1777 using CID voltage 10 V.
(Middle) ESI spectrum of first peak at 11.04 min for kaempferol-3-
glucosylglucoside at CID voltage of 55 V. (Bottom) ESI spectrum of second
peak (RT = 11.45 min) annotated as quercetin-3-rutinoside using ESI

negative mode With CID voltage 55 V ..................................... 150

Figure 4-8 Putative tructures of four anthocyanins petunidin-3-O-(p-coumaroyl)-
rutinoside-S-O-glucoside (m/z 933), malvidin-3-O-(p-coumaroyl)-
rutinoside-S-O-glucoside(m/z 947), delphinidin-3-O-(p-coumaroyl)-
rutinoside-S-O-glucoside (m/z 919) and petunidin-3-(caffeoyl)-rutinoside-5-
O-glucoside (m/z 949) ...................................................... 152

Figure 4-9 ESI mass spectra from (top) anthocyanins at m/z 933 (RT = 18.69 min) and
(bottom) m/z 947 (RT = 20.23 min) from L. lycopersicum M82 using

positive mode With CID 40 V ................................................ 154

Figure 4-10 Important steps in the biosynthesis of anthocyanins and flavonoids (adapted
from B. Winkel-Shirley and K. Saito30’3l) ................................. 155

Figure 4-11 LC/MS XICs for a trichome extract from S. habrochaites LA1777 using ESI
positive mode with CID voltage 10 V showing: m/z 347 (dimethylmyricetin,
2M), m/z 361 (trimethylmyricetin, 3M), m/z 375 (tetramethylmyricetin, 4M)
and W2 389 (pentamethylmyricetin, 5M) .................................. 157

Figure 4-12 Putative structures of major methylated flavonoids in Solanum species
LA1777. The exact positions of methyl groups on these flavonoids retain

unknown ..................................................................... 1 59

Figure 4-13 Nomenclature of fragment ions generated from flavonoidS°"“ "-160

Figure 4-14 (Top) Positive mode mass spectrum of trimethylmyricetin from S.
habrochaites LA1777 using CID potential 80 V. (Bottom) Product ion
MS/MS spectrum of [M+H]+, m/z 361 trimethylmyricetin (positive mode)

118ng collision energy of 35 V .............................................. 160

Figure 4-15 Proposed formation of fragment ions of trimethylmyricetin in ESI positive
mode, obtained from extract of S. habrochaites LA1777 °°°°°°°°°°°°°°° 161

Figure 4-16 (Top) Positive mode mass spectrum of tetramethylmyricetin extracted from
S. habrochaites LA1777 with CID potential 80 V. (Bottom) Positive mode
product ion MS/MS spectrum of tetramethylmyricetin at m/z 361 using

collision energy 40 V ........................................................ 162

xvii

Figure 4_l7 Chemical Structure of tomatine .............................................. 163

Figure 4-18 (Top) LC/MS XICs of [M+H]+ for a-tomatine (RT = 14.78 min) and
(Bottom) dehydrotomatine (RT = 14.29 min) from a leaf dip extract of

tomato (S lycopersicum M82) ......................................... 167

Figure 4-19 Mass spectra of a-tomatine (Top) and dehydrotomatine (Bottom) from
LC/MS analysis of an extract from S. lycopersicum M82 using ESI positive

mOde With CID potential 80 V .............................................. 168

Figure 4-20 A proposed fragmentation pathway for a-tomatine using ESI positive mode
based on literature reports .................................................... 169

Figure 4-21 LC/MS XICs of an extract of S. pennellii LAO716 using ESI in positive ion
mode corresponding to protonated molecules of (Top) a-tomatine and

(Bottom) dehydrotomatine ................................................... 1 72

Figure 4-22 (Top) LC/MS XIC of m/z 1032.5 from ILl-l using ESI positive mode and
CID voltage 10 V. ESI spectrum (Middle) of first peak (RT = 2.73 min)
corresponding to dehydrotomatine isomer and ESI spectrum (Bottom) of
second peak (RT = 3.00 min) corresponding to dehydrotomatine (double

bond in 5,6‘pOSiti0n) using CID 80 V ....................................... 173

Figure 4-23 A proposed fragmentation pathway for first eluting dehydrotomatine isomer
peak (RT = 2.73 min) from IL 1-1 using ESI positive mode with CID
voltage 80 V. Fragment ions at m/z 273 and 255 indicate that the position of

the double bond is not on rings A-D ........................................ 174

Figure 4-24 LC/MS XIC of m/z 1050 corresponding to a putative hydroxytomatine
metabolite from ILl-l using ESI positive mode. (Bottom) ESI spectrum for
hydroxyltomatine from ILl-l using ESI positive mode with CID potential of

80 V ........................................................................... 175

Figure 4—25 A proposed fragmentation pathway for hydroxytomatine from IL l-l using
ESI positive mode ........................................................... 176

Figure 4-26 LC/MS XIC of sesquiterpene acid at m/z 233 from S. habrochaites using
ESI negative mOde ........................................................ 177

Figure 4-27 Product ion MS/MS spectrum of sesquiterpene acid at m/z 233 from S.
habrochaites LA1777 using ESI negative mode with collision energy 30

V ..................... _ ......................................................... 178

Figure 4-28 A proposed fragmentation pathway for sestertepene acid at m/z 233 from S.
habrOChaiteS using ESI negative mOde ..................................... 179

xviii

Figure 4-29 (Top) LC/MS XICs of diterpene acid at m/z 319 and (Bottom)
hydroxysesquiterpene acid at m/z 249 from S. habrochaites LA1777 using

ESI negative mOde ........................................................... 180

Figure 4-30 (Top) LC/MS XICs of m/z 649 [M-H]_ of sesterterpene metabolite using
ESI negative mode. (Middle) Possible Elemental formulas based on accurate
mass measurement. (Bottom) ESI spectrum of m/z 649 (RT — 16. 83 min)
from S. habrochaites LA1777 using ESI negative mode with CID potential

55 V... ......... ......... ..183

Figure 4-31 Constant neutral loss (86 Da) spectrum from an extract S. habrochaites
LA1777 using flow injection analysis and ESI negative mode """""" 184

Figure 5-1 S. pennellii introgression lines. (adapted from http://zamir.sgn.comell.edu/
Qtl/il_story.htm ) ............................................................ 188

Figure 5-2 LC/MS total ion chromatogram (TIC) obtained from leaf extraction of
introgression lines (ILs) 8-1-1 with 5 min gradient using Ascentis Express
C18 fused core column, 2.1 X 50 mm; 2.7 pm (Top) compared to 43 min
gradient with Thermo BetaBasic C18 column, 1 X 150 mm, 3 um (Bottom)

in ESI negative mOdC ........................................................ 194

Figure 5-3 Extracted ion chromatograms (XICs) of [M+HCOO]_ for leaf dip extract of
IL 8-1-1 showing peaks corresponding to sucrose triesters or tetraesters
from m/z 639 to m/z 779 (homologs differing by 14 Da in molecular mass

owing to additional methylene group in the acid moieties) °°°°°°°°°°°°°° 196

Figure 5-4. Markerlynx data for part of introgression line samples. The software first
generates extracted ion chromatograms (XICs) for each mass, then detects
and integrates the peak intensity, and aligns the peaks according to retention

time. The metabolite ID was assigned at the first column for each peak '199

Figure 5-5 Principal component analysis (PCA) on 66 ILS plus recurrent parent M82
(total 276 samples) obtained using pareto scaling with mean centering. PCA
score plot; Symbols: ILS 1-3, 1-4 (0); 5-3, 11-3 (0); 8-1, 8-1-1 (A); Squares
(I) represent recurrent parent M82 (n=31); Cross (*) represent other 60 lLs
(“:3 or 4 for eaCh IL) ....................................................... 201

Figure 5-6 (Left) PCA scores plot from LC/TOF MS metabolite data represent three
clusters of [LS separated from recurrent parent M82 using Pareto scaling
with mean centering. The clusters are 1-3, 1-4 (0); 5-3, 11-3 (O); 8-1, 8-1-1
(A) and M82 (I). (Right) PCA loading plot showed corresponding
analytical signals (retention time-mass pairs) contributing to the difference

among clusters in PCA SCOI‘CS plOt ......................................... 202

xix

Figure 5-7 LC/MS total ion LC/MS chromatograms obtained from leaf dip of M82 (Top)
and IL 1-3 (Bottom) using ESI negative mode. The peaks are labeled after

identified using multiplexed CID MS method """"""""""""""""""" 206

Figure 5-8 (Top) Mass spectrum of peak at RT = 3.35 min in IL 1-3 and (Bottom) Mass
spectrum of peak at RT = 3.42 min from S. lycopersicum M82 with CID

potential 55 V ................................................................ 206

Figure 5-9 Total amount of acylsugars for 66 ILS and recurrent parent M82 based on
Quanlynx software in Masslynx 4.1 (waters). Sum of all the peak areas of
acylsugars were normalized to internal standard (Propy1-4-

hydroxybenzoate) and dry leaf dip .......................................... 209

Figure 5—10 A. Total ion chromatograms from LC-TOF MS analysis of acylsugars in leaf
dips of M82 and IL8—1-1. Labeling nomenclature for acylsugars - S 3:22
(5,5,12) is an acylsucrose with 3 acyl chains having a total of 22 carbons
and numbers in parentheses indicate the lengths of the individual acyl
chains. B. Amounts of acylsugars showing differences in abundance
between M82 and IL8-1-1 are shown as integrated peak areas normalized to

the internal standard and the dry weight of the extracted leaflet """"""" 212

Figure 5-11 ESI negative LC/MS XICs of C4 m/z 87.04 and C5 m/z 101.06 fatty acyl
groups cleaved from acylsugars with CID potential 55 V for A: S.

lycopersicum M82 and B: S. pennellii LAO716. C: ILs 8-1 °°°°°°°°°°° 214

Figure 5-12 Side chains on acylsugars from leaf dip samples were transesterified to the
corresponding ethyl esters and analyzed by GC/MS as described in chapter
6. The peak area % values for the base peak XIC for each fatty acid in M82
and lL8-1-1 are shown along with the standard error with n = 2 for IL 8-1-1

andn: 11 for M82 .......................................................... 217

Figure 5-13 A. Extracted ion chromatograms of M82 and ILl-l for m/z 1076.5 show an
earlier eluting peak found only in the IL and not in M82. B. Schematic
representation of chromosome 1 introgressions showing locus controlling
the glycoalkaloid phenotype is located on bin l-A or l-B (IL1-1-2 was not

analyzed in this study). C. Structures for dehydrotomatine isomers °°°°° 220

Figure 5-14 OPLS—DA loading plot showed interclass difference between IL 1-1 (Right)
and recurrent parent M82 (Left) using fused core column with 5 min
gradient. The ions were labeled with retention time and m/z. The upper right
corner and the lower left comer are corresponding to IL 1-1 (n = 4) and M82

(n 2 4), respectively ......................................................... 222

Figure 5-15 Total amount of mono-, di-, trimethylmyricetin for 66 ILs and recurrent
parent M82 were obtained based on Quanlynx software in Masslynx 4.1
(waters). The peak area of mono-, di-, and trimethylmyricetin is normalized

XX

to internal standard (propyl-4-hydroxybenzoate) and dry leaf dip
individually for all the samples ............................................. 224

Figure 5-16 Total amount of sucrose triesters and tetraesters were obtained based on
Quanlynx software in Masslynx 4.1 (Waters) for 18 F2 plants of 1-3 and 22

F2 plants of l_4 and 10 control plants M82 ................................ 226

Figure 6—1 Structure of synthesized internal standard 3-decanoylglucofi1ranose° '° '° ° "234

Figure 6-2 XICs of formate adducts from S 4:15 to S 4:24 (highlighted with red dash line)
and S 3:15, S 3:20 to S 3:22 (highlighted with blue dash line) from S.

lycopersicum M82 ........................................................... 239

Figure 6-3 Electrospray ionization mass spectra of S 4:17 (RT = 3.11 min) extracted
from S. lycopersicum M82 using CID potential of 55 V (top) in negative and

40 V (bottom) in positive mode ............................................ 245

Figure 6-4 Distributed amounts of glucose triesters, sucrose triesters, and sucrose
tetraesters of S. pennellii LAO716, LA1522, S. habrochaites LA1777 and S.
lycopersicum M82 based on sum of total peak areas of formate adducts for
detected acylsugars normalized to internal standard and dry leaf weight for

each acylsugar ............................................................... 248

Figure 65 Distribution of fatty acyl groups for total acylsugars among S. pennellii 0716,
1522, S. habrochaites 1777 and S. chopersicum M82 obtained from

GC/MS ....................................................................... 251

Figure 6-6 Total amounts of acylsugars in Cornell tomato breeding lines. Reported
levels represent the sum of XIC peak areas of acylsugar formate adducts

normalized to internal standard and dry leaf weight """""""""""""""" 254

Figure 6-7. Distribution of glucose triesters, sucrose triesters and sucrose tetraesters
among all the breeding lines ................................................. 255

Figure 6-8 XICs of deprotonated ion [M-H]- for C4 and C5 fatty acids using CID
potential 55 V, corresponding to m/z 87 and 101, from crosslines 8-1 ><0716

and 8_2x07l6 ................................................................ 258

Figure 6-9 Amounts of acylsugars in S. pennellii, normal tomato lines, fixed-lines and
crosslines based on sum of peak areas, then normalized to internal standard
and dry leaf weight (Top). Composition is broken down into glucose
triesters, sucrose triesters and sucrose tetraesters were among all the

breeding lines (Bottom) ..................................................... 263

Figure 6-10 TIC for LAO716 (top), 071026><LA0716 (middle) and 071026 (bottom)
using ESI negative mode with CID potential at 10 V °°°°°°°°°°°°°°°° 266

xxi

(Figure 6-11 ESI negative Xle of C9 m/z 157.13 and C10 m/z 185.16 fatty acyl groups
using CID potential 55 V for fixed line 071026 and crosses between 071026

and S pennellii LAO716 ..................................................... 267

Figure 6-12 Distribution of amounts of individual fatty acyl groups among fixed-lines,
crosslines and S. pennellii LAO716 based on GC/MS profiles. 077044-6,
077048, 077049-6, 077051-6, 077058-6 are fixed lines; 077082-6 to

077089_6 are crosslines ...................................................... 268

xxii

APCI
C E—MS
C ID
DDA
DFI-MS
ES I
F I D
FR'U

F T- I (IR-MS

GC—NS
GLU
GS Ts
HPLC
IL
MS

OPLs-DA

PC A
QTL
RT

$S’E

TIC

LIST OF ABBREVIATIONS

atmospheric pressure chemical ionization
capillary electrophoresis-mass spectrometry
collision induced dissociation
data dependent analysis
direct flow injection-mass spectrometry
electrospray ionization

flame ionization detection

fructofuranose

Fourier-transformation ion cyclotron resonance
mass spectrometer

gas chromatography-mass spectrometry
glucopyranose

glandular secreting trichomes
high-performance liquid chromatography
introgression line

mass spectrometry

orthogonal projection to latent structures-
discriminant analysis

principal component analysis
quantitative trait locus
retention time

solid phase extraction

total ion chromatogram

xxiii

TOF time-of—flight

XIC extracted ion chromatogram
UPLC ultra-performance liquid chromatography
U V—vis ultraviolet-visible spectrometry

xxiv

Chapter 1. Introduction

The rapid growing world population will have an astounding impact on food and
energy resources.1 In order to meet the needs for quick expanding human population,
large amounts of energy are devoted to irrigation and production of fertilizers and
pesticides, to maintain agricultural productivity. The world needs more productive
agriculture to feed a growing population and provide renewable feedstocks for chemical
production and energy. One way to achieve these goals is through metabolic engineering
of plants and microbes. Genetic modifications of plants offers improvements in
production of diverse natural products (secondary metabolites), that can protect plants
against insect attack, pathogen stress and animal herbivores and increase agricultural
productivityz‘3 Many plant metabolites also have important value as pharmaceuticals,
fragrances and food additives.4 In addition, engineering of plant biomass through genetic
modification is an area of active research that aims to develop renewable sources of
energy. However, many genes involved in regulation of important metabolites remain
largely unknown, and those involved in biosynthesis of key metabolites are not
completely understood. Metabolite levels can be viewed as the responses of biological
systems to environmental or genetic manipulation,5 thus, identification and quantification
chemical metabolites among transgenic plants can aid assignment of functions to genes.6

Glandular trichomes were chosen as a targeted model in the Solanum Trichome
Project to investigate gene functions important in regulation of specialized natural
products. Trichomes are highly specialized epidermal cells common to many plants and
are involved in the synthesis, storage, and secretion of secondary metabolites.7 They are
easily removed from plant tissue surfaces since they protrude from the epidermis.8

Therefore, RNA, proteins and small molecules that they contain are easily sampled.8

Information about DNA sequences, metabolite identities and levels, and gene expression
obtained from trichomes have great utility for identifying gene and enzyme functionsQ
when specialized type of trichomes are isolated from other cell types. First, metabolite
profiling can be performed for isolated trichomes, and transcripts (mRNA) are extracted
for construction of cDNA libraries and EST (Expressed Sequence Tag) database. Gang et
al.'0 and Iijima et al.”‘'2 isolated glandular trichomes from three different cultivars,
prepared EST databases from them, and profiled metabolites for mint and basil. Through
annotation of correlations between transcript and metabolite levels or from altered
metabolite profiles among different cultivars, functions of specific genes that regulate
production of methylated phenylpropenes were discovered.9 So far this approach has

13"5 and has accomplished

provided important information about secondary metabolism,
this more efficiently than early studies that explored gene functions through enzyme
assays.’6 Since altered metabolite profiles may be associated to specific genes to support
annotations of gene function, this approach is also used in the research described below
for trichomes that are prolific in synthesizing specialized metabolites. This dissertation
describes metabolite profiling of several wild type species in the genus Solanum and

hundreds of transgenic plants. My role in this research has focused on development of '

analytical methods for metabolite profiling and rapid screening of metabolic phenotypes

screening to guide discoveries of gene functions important for metabolic engineering.

1.1 The Emergence of Metabolomics in Biological Research

Metabolites are often classified as primary or secondary metabolites.'7 Primary

metabolites are important to plant growth, development18 whereas secondary metabolites

are often connected cell signaling, interspecies communication and responses to biotic
and abiotic stress.19

Broad assessment of the entire suite of metabolite produced by a biological system is
named “metabolomics”.5‘20'23 The general idea of ‘metabolomics’ was first brought up
several years ago in the field of microbiology,24 and its importance in plant science was
discovered soon after. The variety of metabolites, as intermediates and end products of
biochemical pathways, is regulated by many structural and regulatory genes in addition to

25 Plants are sources of numbers of metabolites whose

environmental influences.
structures, functions and utility have been largely unknown. At this stage, we have only a
vague idea of how large the complement of metabolites within a particular species is;
estimates from 5000-25000, and are comparable in order of magnitude to the number of
genes.26 These numbers considerably exceed estimates of the number of metabolites
synthesized by prokaryotic microorganisms (~1500) and animals (~2500).27 Although
more than 100,000 plant secondary metabolites have already been identified, in fact these
may represent only 10% of the actual total synthesized by the plant kingdom.28

The metabolomics approach strives for an overview of whole-cell metabolic
patterns.29 Depending on the number of compounds they analyze, the level of structure
information they obtain, and their sensitivity, the analysis of metabolites is divided into
target analysis, metabolic profiling, metabolomics, or metabolic fingerprinting.30 Target
analysis aims at quantitative analysis of known substrate and/or product of targeted
protein.23 The most common approach, metabolite profiling, a term first brought up in the

19705 to refer to the qualitative and quantitative analysis of complex mixtures of

physiological origin.“32 One of the early pioneers in this field, Professor Charles

Sweeley, was one of the first to demonstrate such an approach, using gas chromatography,
GC/MS, and computerized data systems and library searching for profiling of metabolites
from assorted classes.33 Much of this work was conducted at Michigan State University.
At the other extreme, metabolic fingerprinting detects many compounds without

34 This technology focuses on pattern

requiring identification of their structures.
recognition of metabolic phenotype to produce a broad picture,3O rather then cataloguing
specific metabolites.

In order to find out the functions of metabolites and corresponding enzymes, two
basic strategies can be performed, termed forward genetics and reverse genetics.30 In the
forward genetic approach, naturally occurring mutants and mutagenized plants with
interesting phenotypes can be used to guide identification and characterization of genes
correlated with one or more phenotypes. In reverse genetic approaches, analysis begins
with a specific cloned gene, a protein sequence, or a specific genetic mutation and seeks
to define the possible phenotypes derived from a specific gene.30 No matter which
approach is applied, our understanding of phenotype is based on what can be observed
about the characteristic behavior of an organism, and this can often be observed at the
metabolite level.5 Since the metabolites are the end products of gene expression,30
efficient and comprehensive approaches to genomic analysis must include measurements
of chemical constituents—the metabolites—of individual cells or tissues of the organism.
By coupling metabolite analysis with the information provided by genomics, we create a

science of metabolic genomics, which accelerates completely deciphering the complex

inner workings of living systems.

1.2 Challenge of Metabolomics

The major challenge in metabolomics lies in the diversity and complexity of the plant
metabolome and these factors provide the impetus for pursuing cutting-edge
technological deveIOpments.35 Metabolome analyses aim to determine all metabolites in a
plant extract; however, no single routine technology for metabolomics, such as DNA
sequencing for genome analyses or DNA arrays for transcriptome analyses, is available.27
The genome and the transcriptome are constructed from four nucleotides with highly
similar chemical properties, thus facilitating high throughput analytical approaches.23
Unfortunately, metabolites have a much greater variability in the arrangements of atoms
and functional groups compared to the linear 4-letter codes for genes or the linear 20-
letter codes for proteins.5 Most metabolomics methods rely on GC/MS, LC/MS to
identify metabolites through comparison against commercially available standards.23
Structural elucidation of metabolites is a major obstacle since considerable amounts of
high quality purified materials were needed to perform multiple dimensional NMR.
Although tandem mass spectrometry requires smaller amounts of material, complete
structures of unknown metabolites only can be identified at some limited level by it alone.
The alternative approach for metabolite identification is to build a library of standard
profiles, created with purified metabolites.36 Although the libraries for NMR, IR and
Mass Spectrometry have existed for decades, they are still not complete for the nearly
200,000 known natural products.37 In addition, the range of chemical properties of
metabolites covers ionic inorganic species. hydrophilic carbohydrates, hydrophobic lipids.

and complex natural products.38 No single analytical platform yields efficient

identification and quantification of all metabolites, therefore, multiple technologies are
generally employed.39
Dynamic range of quantitative analysis is also a major challenge for metabolomics.
The dynamic range of the most powerful analytical techniques lies between 4 and 6
orders of magnitude for individual components.23 However, the dynamic range of many
techniques is often limited by sample matrix or the presence of other compounds that can
interfere with measurement of specific analytes. For instance, oligosaccharide profiling
by liquid chromatography-mass spectrometry is compromised when amino acids and
peptides are present in the sample because these latter substances are preferentially
ionized and can suppress ionization of oligosaccharides.23
Another difference for metabolomics compared to other “omics” is that it is difficult
or impossible to establish a direct link between individual metabolites and genes as
mENA and proteins.35 The number of metabolites is far beyond the number of genes in
one species involved in the biosynthesis since multiple metabolites were produced from
one gene due to several mechanisms.40 Therefore, metabolome analysis need to cover the

‘dentification and quantification of all intracellular and extracellular metabolites to study

géne function, which creates more challenge for metabolomics.

1.3 Techniques for Metabolomics

Advancing analytical technologies offer great opportunities for viewing global
changes of metabolites in a system caused by perturbation of genes in order to associate
metabolic phenotypes with gene functions of an organism.34 Due to the complexity and

diversity of chemical and physical properties of plant metabolites, multiple analytical

approaches must be combined to achieve this goal, including photodiode array detection
( UV/V is), GC, NMR and MS.

Photodiode array detection (UV/Vis) coupled with various separation techniques has
found only limited application due to the low extinction coefficients of many metabolites
and lack of selectivity in analyte detection.“ It is typically limited to select classes of
metabolites containing strong and distinct chromophores, such as the anthocyanin
pigments in plants.42’43 Therefore, UV/V is detection is primarily used for target analysis
and metabolic profiling,”45 but not for global metabolome characterization. Gas
Chromatographic separation coupled to flame ionization detection (F ID) has found wider
use in metabolite analysis.”47 The chromatographic resolution of GC provides can
SeIDarate hundreds of metabolites in a single analysis and allows for high throughput
£11'l'c‘tlyses, but flame ionization detection provides minimal information regarding the
iclfintities of the analytes beyond chromatographic retention times. For that reason, GC is
211 &0 limited to target analysis and metabolic profiling where the only compounds of

ill-":erest have been identified and are available as standards.
NMR serves as an alternative approach for determining chemical structures of
metabolites and measuring their abundances. Although 'H NMR provides comprehensive
measurements of hydrogen-containing metabolites, it is generally low sensitivity and
suffers from overlapping signals.48 Recently, NMR has been interfaced with HPLC to
yield multidimensional data useful for metabolite structure elucidation.49 The primary
disadvantage of current HPLC/NMR technology lies in its low duty cycle and elevated

expense due to the need for deuterated mobile phase.23

Recent metabolome studies have expanded to a variety of body fluids and tissues and

now employ a range of instruments including F T-lCR-MS, GC-MS, GC-GC-MS, HPLC-

MS, UPLC-MS and capillary electrophoresis-MS.50‘5l FT-ICR-MS offers the highest
mass resolution (100,000-1,000,000) and mass accuracy (0.1-1 mDa), detection limits in

the attomole to femtomole range and MSn capability,52 and these features make it a
powerful tool for metabolite fingerprinting investigation.53 Despite these features, the
application of mass spectrometry methods, including FT-ICR-MS, often cannot
differentiate isomeric metabolites without resorting to an additional dimension of analysis.
such as physical separation or isomer-selective fragmentation. In addition, the high
capital and operating costs of high-performance F T-ICR-MS instrumentation serve as a
barrier to the adoption of this technology as the analytical technique of choice for

metabolome analyses.

Both targeted and nontargeted metabolite analyses are often performed using GC/MS,
which has proven capability for profiling from several hundred to slightly more than a
thousand components.24’54'56 GC/MS can contribute high separation efficiencies with up

to 250,000 theoretical plates.23 Furthermore, GC/MS offers the advantage of having
extensive libraries of 70 eV electron ionization mass spectra that aid metabolite
identification. However, one drawback of using GC/MS derives from the requirement
that metabolites must be volatile and not decompose at elevated temperatures used for
chromatographic separation. Since many metabolites are ionic or polar compounds of low
volatility, chemical derivatization must be performed to convert them to more volatile
forms. Because many metabolites must be derivatized, extracts must often be evaporated

to dryness before derivatization. Numerous metabolites, including secondary metabolites
such as multiply glycosylated conjugates and betaines with permanently charged groups,
still cannot be detected using GC/MS even after derivatization owing to their

decomposition during heating and vaporation. In addition, identification of unknown

ILHI

.1

3’1;

("1

metabolites following derivatization and GC/MS analysis has frequently failed because
libraries of metabolite mass spectra are incomplete. This is illustrated in one of the more
successful metabolomic studies to date. In a collaboration between the laboratories of
Oliver Fiehn and Michigan State University’s Michael Thomashow, the effects of cold
stress on the metabolome of the model plant Arabidopsis were explored using
derivatization and GC/MS analysis,57 In this study, only 26% (30 of 114) of metabolites
exhibiting greater than 5-fold change after cold treatment were identified. Such results
highlight how the incomplete success of metabolite identification in metabolomic
investigations results from inadequacies in existing libraries of metabolite mass spectra
and the challenges of interpreting spectra. Therefore, alternative analytical techniques are
needed to identify and quantify a broader range of specialized secondary metabolites.
The combination of liquid chromatography with mass spectrometry (LC/MS) has
emerged as a rapid, efficient and powerful approach to complement GC/MS for analysis
Of a broad range of metabolites including polar and semi-polar compounds.5860 LC/MS
removes the need for sample derivatization and therefore samples can be analyzed with
minimal sample processing or even directly from the tissue source.“62 Most semi-polar
metabolites (phenolic acids, flavonoids, alkaloids and other glycosylated species) can be
separated using reverse-phase columns. Polar compounds can be measured58 (sugars,
amino sugars, amino acids, vitamins, carboxylic acids and nucleotides) using hydrophilic
and ion-exchange separations. Recent developments termed ultra performance liquid
chromatography (UPLC), provide higher separation efficiency and resolution through use
of sub-2 um particles.“66 Short UPLC provide separations of efficiency equal to longer
conventional HPLC columns, and this allows for analysis times to be shortened relative

to conventional methods. UPLC separations are beginning to find application in

10

63.67.68 66.69.70

nontargeted metabolite analysis or targeted profiling, in view of their
performance in terms of speed, robustness, and high sample throughput.

Technologies that couple liquid separations to mass spectrometry have also improved
in recent years, following the development of the soft atmospheric pressure ionization
methods atmospheric pressure chemical ionization (APCI) and electrospray ionization
(ESI). APCI is a less "soft" ionization technique than ESI,7| i.e. it generates more

fragment ions relative to pseudomolecular ions, and is more frequently applied for
ionization of less polar and less thermally labile compounds. ESI has provided a major
breakthrough for the analysis of biological samples by mass spectrometry because this

technique is applied directly to the analytes in solution,72‘7

3 requires no derivatization,
provides low detection limits, and generates reproducible results. Furthermore, recent
advances in electronics and computing have hastened the development of mass
Spectrometers. Among this new generation, time-of-flight (TOF) is the most prominent
because it requires fast processing of signals, usually at rates faster than 1 GHz. TOF
mass analysis offers high mass accuracy, medium to high mass resolution, adequate
dynamic range and fast spectrum acquisition capability.”75 These features have made
TOF analyses highly valued and broadly applied for targeted and nontargeted metabolite
profilingm‘m'78 Therefore, development of technologies in separation and mass

spectrometers make LC/MS become a powerful high-throughput approach for

metabolomics.

1.4 Multivariate statistical tools for metabolomic analyses

11

Compared to traditional targeted analysis, one of the major advantages for
metabolomics lies in the measurement of changes in metabolism resulting from
environmental, genetic or developmental changes.”80 Instead of tracking a few
metabolites, a single GC-MS profile can measure 300—500 distinct compounds.23
Although a wealth of information can be obtained from this, the challenges lie more in
data processing than data acquisition. Large scale metabolomic studies usually involve

screening of thousands of samples varying in genotype, treatment, or time in a single
study. The large volume of generated data accentuates the need to apply and develop
data processing software and multivariate analysis tools.

Recent developments in processing software or unbiased mass peak extraction and
al i gnment of LC/MS data, such as Waters MarkerLynx, now offer improved workflows
for untargeted metabolomics approaches, which intend to gather information on as many
metabolites as possible for each extract analyzed.“83 MarkerLynx software extracts data

from multiple LC/MS analyses, then detects, integrates, and aligns chromatographic
Peaks. A list of intensities of the peaks detected was generated for the first sample, using
retention time (RT) and m/z data pairs as the identifier of each peak. An arbitrary number
was assigned to each of these RT-m/z pairs in the order of their elution for data
alignment.69 Ions of different samples were considered to be the same ion when they
demonstrated the same RT and m/z within limited mass and retention time windows.
After such alignment, peak area data for each RT-m/z pair are assembled into a table.
(Figure 1-1). If a peak is not detected in the sample, a zero is recorded as peak area in the
data table. This process is repeated for each sample, and data from each run in the batch
sorted so that the correct peak area data for each RT—m/z pair are aligned in a final data

table.84 Integrated peak areas within a sample are then normalized to the sum of the peak

12

intensities for that sample, as this facilitates certain statistical comparisons of sample data.
After alignment and normalization, data from a batch of samples were ready for

appropriate statistic and multivariate analysis tools.

Table 1-1 MarkerLynx software first detects peaks using specific mass data, aligns the
peak according to retention time, integrates peaks over all samples, and assembles the

results into a table.

 

 

Mass Ret. Time Ion intensity Ion intensity Ion intensity Ion intensity
ml: (min) (sample 1) (sample 2) (sample 3) (sample 4)
681.3215 23.4517 38873.9219 18603.3223 40866.7187 23265.209
519.3077 30.6518 0 0 58.5839 41.1124
723.412 33.3373 450.8516 241.5029 386. 5996 172.6767
533. 3221 33.0887 0 0 1 19. 8434 73.4247
709.3905 32.1405 4977.9126 3300.7446 5445.8208 2131.6414
519.3061 31.7391 138.9787 0 164.2645 95.6386
765.4271 34.5681 533.9459 284.1887 430.8036 164.2941
449.2242 19.5197 58.3577 65.1068 114.182 90.7419
435.2096 17.4071 0 0 0 0
547.3348 34.7608 70.6505 25.1308 42.1335 0
575.3652 37.4728 40.1302 0 0 0
435.2106 16.7895 0 0 0 0
737.4205 35.3425 16415.6641 135580674 17265.7578 110480996
477.2637 24.6339 0 0 0 0
561.3495 35.6978 135. 8656 39.7652 50.622 32.1877
853.4323 19.5759 0 0 0 0
993.5841 31.7724 0 0 0 0
561.3501 36.5467 25.7131 0 28.8188 0
519.3071 31.1841 0 0 21.776 0
695.3716 29.9849 0 0 0 0
533.3224 33.577 0 0 49.091 31.9687
463.2445 22.2556 0 0 0 0
547.3354 34.2186 0 25.1308 0 26.4573
519.3069 28.8797 0 0 0 0

 

13

Principal component analysis (PCA) is one of the oldest and most widely used
unsupervised multivariate techniques.23 PCA can reduce the dimensions of a complex
dataset and provides a rapid visualization of similarity and difference among samples.
The similar samples group together into one cluster and different ones are far from each
other based on components they contain. This is a mathematical way to present the
correlation among sample sets. The most important use of PCA is to represent a
multivariate data table as a low-dimensional plane, usually consisting of 2 to 5
dimensions,85 with each dimension representing an orthogonal component of the data
variance. Selection of two dimensions, often chosen to represent a substantial fraction of
data variance, leads to construction of a two-dimensional plane, and each of the sample
points was projected into this plane to visualize all the samples. The coordinates of the
observation in this plane are called scores, and visualization of those scores is termed a
score plot. Figure 1—1 (Left) displays a scores plot, which presents a projection of

metabolite data from three different plant genotypes, with each point representing a

Single sample. The samples belonging to one accession cluster together and different

accessions are far from each other. Examinations of scores plots often reveal groups of
observation, trends or outliers.

Once the relation among the samples is recognized, it is necessary to identify the
variables responsible for the clustering. Such information is provided by a loadings plot,
in which each point represents a specific variable (e.g. peak area for a specific mass-
retention time pair). The loadings plot identifies variables most responsible for clustering
of the data. In the case of metabolome data, this corresponds to signals from specific

metabolites. An important aspect is that directions in the scores plot correspond to

14

directions in the loadings plot. PCA analysis provides a powerful tool for understanding
the underlying patterns in multivariate data,"9 and finds frequent application for
visualization of metabolomics data and discovery of samples that are metabolic
outliers.”88
PCA often fails to reveal contributions to sample discrimination by metabolite of

minor abundance, and is often followed by OPLS-DA (orthogonal projection to latent
structures-discriminant analysis). OPLS-DA is a supervised statistical tool to display
difference in signal intensity between samples and potentially correspond to
bi omarkers.84‘89’9O OPLS-DA is used to take advantage of additional knowledge about
each the class to which each sample belongs (e.g. wild type vs. mutant), and this
information is organized in what is known as the Y matrix.79'°' OPLS is to separate the
systematic variation in X into two groups, one that is linearly correlated to Y and one that
is orthogonal to Y.79‘ 92 Same as PCA, OPLS-DA also has scores plot, which clearly
displays the differences of the two sample groups in the horizontal as well as the
differences within the same sample groups along the vertical axis. In the S-plot, each
point represents a pair with exact mass and retention time. As a result, the largest and
reliable magnitudes variables in the upper and lower outer the S-plot. Thus, OPLS-DA is
applied for modeling two classes of data to increase the class separation and find
potential marker590'93 In fact, for assessment of metabolomics study, PCA is suggested as

a starting point for analyzing multivariate data and rapidly present overview information
hidden in the data.85 Then OPLS-DA was followed to obtain a clear and more
straightforward interpretation, especially, the interclass variation can be obtained from

it.85

15

Minor genetic variations can lead to subtle changes in complex plant chemistry, and
recognition of these metabolic differences from mass spectrometry data poses substantial
challenges. Owing to the potential for several hundred metabolites to be present in plant
tissue extracts, identification of every metabolite is usually impractical and unnecessary
for assigning plant gene functions. A more efficient approach should first aim to
recognize metabolite signals that distinguish genotypes through application of
multivariate statistical analysis of MS data, and the statistical tools described above
provide important functions to aid recognition of metabolites that distinguish samples.

Efforts to assign chemical structures can then be focused on the outlier metabolites.

1.5 Metabolite Identification

Chemical identification of metabolites in metabolomics is fundamental for the
extraction of biological perspective from the data.62 The basic strategy behind metabolite
identification today is similar to the foundations laid by the pioneers of organic mass
spectrometry including John Beynon, Fred McLafferty and Klaus Biemann.84 First,
providing high mass accuracies (<5 ppm) for the ion of interest, the range of possibilities
of molecular formulas is narrowed down, and in the ideal case, leads to a short list of
possible molecular formulas for each ion.94 However, there are still more possible
mathematical combination of elements fit certain molecular mass than a number
molecular formulas that exist chemically, even the instrument can provide the accurate
measurement for molecular mass, especially for mass of ion above 400 amu.62 Several
methods can be used to exclude impossibility, such as nitrogen rules, used for evaluation

of the presence or absence of N atoms in a molecular or ion.94 Another tool for narrowing

16

the number of molecular formulas is to make use of the isotopic pattern for specific
mass.95’96 According to Kind and Fiehn,96 this strategy can remove more than 95% of
false positives.

Following assignment of likely elemental compositions to a molecular ion, the ion is
converted into fragment ions using one of several ion activation methods and tandem or
multistage mass spectrometry. Isolating one ion and performing MS/MS can track
functional groups and connectivity of fragments for elucidating the structures of

metabolites.94 In addition, an in-source collision induced dissociation (CID) approach,

E _
S ’97 99

named M produces accurate mass measurement for fragments, not just molecular

ions. The advantage for accurate mass measurement of fragment ions is association of

fragments with molecular ions to figure out the functional group corresponding to mass

loss between them. For example, the mass loss 44.02 Da (C02) obtained from CID in

negative mode indicates the ion contains one or more mono carboxylic acid groups.

MS/MS spectra can provide an indication of putative structures of metabolites using

database and/or manual interpretation of the fragments'oo’m'

In addition, a term “mass defect” is also very important to classify the metabolites.

. . l 14 . .
Among elements commonly found in metabolites, II and N make posmve

contributions to mass defect (+7.8 and +3.1 mDa per nucleus, respectively), heavier

elements (160, 31P and 32S) make negative contributions to mass defect (-5.1, -26.2 and -

27.9 mDa respectively), and 12C contributes zero to mass defect. Owing to the high

number abundance of hydrogen and low frequency of phosphorus and sulfur in most
metabolites, the mass defect for molecular, dimer, and fragment ions largely reflects the

number of hydrogen atoms. Divided the mass defect by the molecular mass will give a

17

measurement of %H. This value might facilitate mining certain compounds. For

examples, if the ion locates within 200-300 ppm in tomato trichomes, it often belongs to

polyphenols.

Retention time is another parameter which can assess the polarity of metabolites
when separation is coupled with mass spectrometer.94 And the advanced techniques make
retention time reproducible, allowing direct comparison of chromatograms and construct
metabolite databasem‘103 The structure can be determined from exact mass measurement,
isotopic pattern, fragmentation pattern or search the existed libraries or database. A

number of databases give lists of known compounds for particular elemental
compositions or exact mass values,84 including SciFinder Scholar
(http://www.cas.org/products/facad/fsflash.html), Metlin (http://metlin.scripps.
edu/metabosearch.hhp), ChemSpider (http://wvwv.chemspider.com/) and many others.5 '
The final proof for identification of a compound is through comparison of LC retention
time and MS/MS spectrum with that of an authentic standard.62 However, in most cases,
there are few economically available standard compounds for such comparision,
especially in the secondary metabolism field.94 Thus, for identification of novel unknown
structures or non-novel compounds without standards,8 we can use traditionally
acceptable physicochemical properties, such as accurate mass, elemental analysis, UV/IR

spectra, and/or NMR spectra?“94

1.6 Secondary Metabolites in Solanum Glandular Trichomes

Most plants have hairs on their aerial surfaces, called trichomes, that serve a number

of functions ranging from protection against insects to heat and moisture conservation'O4

18

Trichomes range in size, shape, number of cells and morphology, as well as composition
across species and accessions.8"05 There are two general categories of trichomes:
glandular and nonglandular in Solanum species. Glandular trichomes contain or secrete a
mixture of secondary metabolites at their tips, and many have significant commercial
value as drugs, fragrances, food additives, and natural pesticides.4’106 Glandular trichomes
are accessible cell types that are easily sampled, and for this reason they provide a useful
model system for investigation of secondary metabolism.

Several classes of chemical compounds have been studied identified in the glandular
trichomes. Acylsugars are one of them and wildly distributed within the trichomes. These
compounds have glucose and sucrose as center, acylated by short- to medium-chain
length fatty acids (C2-C12) analogs, some of which are branched and derived in part
from amino acids.'07'”3 Acylsugars are secreted from the gland exudates and they cause
the plant surface to become sticky and provides a strong deterrent to insectsg. Except the
possible taste different, no significant variation in biological activity has been attributed
to structural differences among acylsugars.l '4 Therefore, why the trichomes secrete the
diverse and complex acylsugars across species and accessionsm’l '5'” is still unknown.
A more complete understanding of mechanism of acylsugar synthesis and regulation of
these secondary metabolites is necessary for their engineering of plants for optimal

acylsugar accumulation.1 '9
Another metabolite class, the terpenoids, is recognized as the most abundant and
structurally diverse,‘20 consisting of the C10 mono-, C15 sesqui-, and C20 diterpenoids,

derived from the MEP (2-methyl-D-erythritol—4-phosphate) biosynthetic pathway.'2'

Terpenoids are essential to the plant development and growth'22 and those volatile

19

I“

compounds can attract natural enemies of herbivores, either predators or parasotoids.123
These compounds are wildly distributed in the plant kingdom, that includes the mints,
sages, and basils.9

In addition, flavonoids are commonly found in trichomes. Trichomes of Phyllyrea

I24

Iatifolia8 and Betula accumulate flavonoid glycosides. Flavonoids aglycones are found

on the surface of many plant species125 and present in the glands of peppermint leaves.‘26
Alkaloids are another important class among secondary metabolites found in gland.
Toxins nicotine and related alkaloids exist in the roots, leaves and trichomes'27 Alkaloids
are reported to have a variety of pharmacological and nutritional properties in animal and
humans, thus, we need a better understanding of the role of these compounds both in the

127,128 In summary, the metabolic pathways involved in the

plant and in the diet.
biosynthesis of most secondary metabolites are not well understood at this stage. Thus,
my efforts have been part of a larger Solanum Trichome Project which aims to
understand the evolution of glandular trichome chemistry and function.

Tomato and related species in the family Solanaceae (e. g. peppers, eggplant, tobacco,
and petunia) were chosen as target plants since a variety of trichomes are present on the
surfaces of leaves, stems, and reproductive structures of those plantsm‘130 Plus, a rich
history of research on tomato genetics, has led to a strong set of tools based upon
mutagenesis and introgression breeding with wild relatives of tomatom. The ability to
get fertile offspring from crosses of tomato with related wild species provides important
tools for a genetic dissection of glandular trichome development and metabolism since
128.132

accumulation of secondary metabolites is often species- or accession-specific.

Therefore, many functional genomics tools for tomato and related species will assist our

20

research on plant secondary metabolism. For example, after analyzing the introgression
lines for changes in trichome metabolites and comparing back to the recurrent parent (S.
lycopersicum cv. M82), the region of the genome controlling any observed changes can
be narrowed down to a particular locus based on the metabolic phenotype of ILs with

overlapping introgressions in our study.

1.7 Summary of Research Goals

In order to provide comprehensive information and speed up metabolite profiling, a
new approach, described in chapter 2, named multiplexed CID was developed, which can
expand the information yielded from LC/MS metabolite screening. Accurate mass
measurement for fragment ions and molecular ions are obtained in a single analysis, and
this information accelerates structural elucidation. The technique also generates
information about the dependence of ion abundances on an electric potential used to
induce ion-molecule collisions. The shapes of breakdown curves constructed based on
multiple CID potentials provide the possibility to assign unknown ions as oligomers,
fragments and molecular ions, with reduced matrix effects and enhanced quantitative
dynamic range. Hundreds of metabolites are annotated and classified using this approach
in Chapters 3 and 4, such as acylsugars, flavonoids, alkaloids and terpenoids. The
annotation of these metabolites broadens our knowledge of metabolites in Solanum
species, and especially opens the eyes of biochemists to study the biosynthetic pathway
for these secondary metabolites and their function to herbivore resistance. In addition, a
high-throughput approach was developed based on multiplexed CID for phenotype

screening of introgression lines to accelerate gene discovery in Chapter 5. Three hundred

21

samples were screened within two days and five phenotypes were pointed out from 65
introgression lines derived from cultivated tomato M82 and wild type S. pennellii
LAO716, which accelerate the discovery of gene functions. Finally, in Chapter 6, this
approach has been applied to qualitative and quantificative analysis of glucose and
sucrose esters pattern using LC/TOF MS, and compares branched- or straight-chain fatty
acids of different breeding lines by GC/MS for hundreds of Cornell breeding lines. The
results indicate this approach has great potential to support plant breeders in efforts to

develop desirable insect-resistant acylsugar breeding lines of tomato in enconomiclly

important crops.

22

 

Chapter 2. Discovery and deep profiling of metabolites using liquid
chromatography/multiplexed collision-induced dissociation (muxCID)/mass

spectrometry

23

2.1 Introduction

Plants and fungi produce a remarkable diversity of specialized secondary
metabolites, many of which have biological activities important in medicine, industry,
and agriculture. Often these are synthesized in specialized cell types such as glandular
trichomes!“ which produce chemicals with important pharmacological, flavor, and
fragrance properties and in plant defenses against insects and disease. Despite valiant
research efforts, progress in identifying and exploiting the genes responsible for
metabolite biosynthesis has moved slowly owing to a lack of “-omics” resources that
reach beyond standard model organisms. Recent developments in high-throughput DNA
sequencing offer prospects for rapid gene discoveries, but association of these genes with
metabolite biosynthesis will require efficient identification and deep profiling of
metabolites that can keep pace with transcript profiling based on next-generation
sequencingzg’I33

Levels of specialized metabolites are dynamic, dependent on tissue and cell type, and
are sensitive to influences of both genetics and environment.30 Non-targeted and deep
profiling of specialized metabolites presents substantial technical challenges, and modern
mass spectrometry provides a powerful tool for metabolite profiling.I34 More common
approaches for identification and quantification of specialized metabolites require
generation of mass spectra that can provide masses of molecular and fragment ions.
While this can be achieved through real-time data-dependent selection of ions that

l36-l38 this

undergo subsequent fragmentationI35 as is common in proteomic analyses,
approach is best performed using slow gradient elution and often fails to detect low

abundance analytes. DDA results in a loss of data in the MS mode when MS/MS data are

24

being acquired, especially for low abundant metabolites which are below the predefined
threshold, thus making it less than ideal for non-targeted metabolite profilingm’ '40
Furthermore, the data-dependent switching between MS and MS/MS experiments
dictates that the analytical method varies from sample to sample, thus posing potential for
compromising quantitative comparisons of metabolite levels.

The energy dependence of ion fragmentation has been shown to distinguish the
presence of specific functional groups in peptides140 and oligosacchan'des,'4"I42 drug

143"“ and carotenoid metabolites,‘5 but the latter two investigations used

metabolites,
slower scanning quadrupole mass analyzers. One promising option for enhancing
information content of LC/MS analyses involves alternating of spectrum acquisition
between gentle and energetic fragmentation conditions without precursor ion selection, a
process recently termed MSE acquisitionf’n'gg’l45'l5 I This approach yields molecular and
fragment information in one analysis with a single and reproducible method of execution,
often fast time-of-flight mass analyzers capable of recording signals for ions across
typical m/z ranges used for LC/MS analyses on the time frame of <100 us. Algorithms
have been developed to separate matrix interference from analyte signals.15 ' These efforts
have employed two collision conditions, fragmenting and non-fragmenting.'52‘ '53 To our
knowledge few reports have generated spectra using more than two CID conditions in a
single analysis, but the results were typically used to choose two conditions, fragmenting
and non-fragmenting, used for sample analyses.’54 This report also suggested that a single
high collision energy CID condition often did not yield desired fragmentation for all

metabolites, but did not explore use of the energy-dependence of fragmentation beyond

generation of structurally useful fragment ions for structure confirmation.

25

Deep profiling of complex mixtures of specialized metabolites poses the challenge of
characterizing more than 1000 metabolites in a single LC/MS analysis. Further
discoveries of the genes involved in metabolite biosynthesis will require large-scale
profiling efforts of numerous genotypes, tissues, treatments, and time points for each.
The needs of such efforts will drive LC/MS technologies to deliver more chemical
information using shorter analysis times. In this study, we describe the multiplexing of
non-selective CID (muxCID) as a powerful tool for enhancing information from non-
target metabolite profiling in the form of nonselective generation of ion breakdown
curves. We take advantage of the rapid acquisition capabilities of time-of-flight mass
spectrometry to generate energy-resolved fragmentation with accurate mass
measurements by switching among five different CID voltages on the chromatographic
time scale. As the following sections will describe, the muxCID approach offers
additional advantages for annotation of ions, extending quantitative dynamic range, and

resolving matrix effects.

2.2 Experimental Section

2.2.1 Materials. Seeds for plants were obtained from the C. M. Rick Tomato Genetics
Resource Center at the University of Califomia—Davis. Plants used for this study were
grown in a greenhouse at Michigan State University, and leaves were harvested

immediately before metabolite extraction.

2.2.2 Chemicals. Acetonitrile (HPLC grade), 2-propanol (HPLC grade), methanol
(HPLC grade), and formic acid (88% aqueous solution) were purchased from VWR

Scientific; rutin was obtained from (Sigma-Aldrich, St. Louis, MO, USA).

26

2.2.3 Extraction Method. To extract metabolites from Solanum trichomes, ~100 mg
leaves were dipped in 2 ml isopropanol: CH3CN: H20 (3:3:2 v/v/v) for 1 min. Extracts

were centrifuged, and the supematants were analyzed without further processing. Extracts

were stored at ~20 °C until analyzed.

2.2.4 Liquid Chromatography. A Shimadzu (Columbia, MD) LC-20AD HPLC ternary
pump with SIL—5000 autosampler was used. Separations were performed on a Thermo
BetaBasic column (1 X 150 mm, 3 pm particles) held at 30 °C that was interfaced to the
mass spectrometer via electrospray ionization probe. A solvent gradient was executed
based on 0.15% aqueous formic acid (A) and methanol (B) as follows: initial condition
5% B; linear gradient to 50% B (5 min); 95% B (33 min); 100% B (35 min); hold at

100% B until 38 min; return to 5% B (43 min). The flow rate was 0.1 ml-min-l, and the

injection volume was 10 pl.

2.2.5 Mass Spectrometry. Analyses were performed by coupling the HPLC separations
(above) to a Waters LCT Premier mass spectrometer (Milford, MA), operated using
electrospray ionization (ESI) and under control of MassLynx version 4.1 software. The

ESI conditions were as follows: capillary voltage, 3200 V; cone voltage, 10 V; source

temperature 90 °C; desolvation gas flow, 300 L-h-I; desolvation gas temperature, 200 °C;

--1 . . . .
cone gas flow, 20 L-h . Detection was performed 1n the negative ron mode over m/z 50-

1500 using centroided peak acquisition and dynamic range enhancement. Aperture 1
voltages were 10, 25, 40, 55, 80 V for five functions (where the scan time for each mode

was 0.4 s with an interscan delay of 0.1 s).

27

2.3 Results and Discussion

Three hard realities confront investigators designing LC/MS protocols for
comprehensive and nontargeted metabolite analysis. First is the notion that no single set
of ion source conditions will be optimal for all metabolites. Conditions that deposit
enough energy to dissociate noncovalent clusters of one metabolite may cause excessive
fragmentation in others. Second, comprehensive and non-targeted metabolite profiling
using LC/MS presents challenges to annotate each ion species as molecular, fragment,
adduct or noncovalent dimer ions. This can present substantial challenges when
metabolites coelute, as is often the case for large-scale metabolite profiling studies that
require short analysis times to enable analyses of large numbers of biological replicates.
Third, plant tissues produce a diverse suite of metabolites with abundances spanning
several orders of magnitude, and extended dynamic range is an important issue for
metabolite identification and quantification.

Extending nontargeted metabolite analysis to low-level metabolites demands
recognition of ion type and acquisition of molecular and fragment mass information for
as many metabolites as possible. This can be achieved by rapid switching among multiple
CID conditions during a single analysis using a predetermined protocol (Figure 2-1A).
Switching between CID conditions can be achieved on the time scale of about 10 ms, and
spectra for each CID condition are stored in separate data acquisition functions within a
single data file. Such analyses yield quasi-simultaneous acquisition of spectra under

gentle and harsh conditions in addition to several conditions of intermediate severity

28

(Figure 2-lB). For the purposes of this study, five different CID conditions were
employed to allow for sufficient time for each stored spectrum to correspond to
accumulated signal from 8000 transients. Spectra acquired at any given time present a
mixture of molecular, fragment, and noncovalent oligomer ions derived from all
substances eluting at the time. Switching between CID conditions yields different
amounts of decomposition of each ion into fragments, with the yields of various fragment
ions depending on the reaction kinetics and activation energies for each fragmentation
pathway. Further data processing can be performed for each retention time-ion m/z
combination to generate breakdown curves (Figure 2-1C) showing the dependence of ion
abundance on collision potential. The figure depicts behavior of ions from three coeluting
metabolites. The first metabolite yields molecular, noncovalent dimer, and fragment ions,
even at the lowest CID potential. lts low threshold for fragmentation allows it to yield
substantial amounts of fragment ions at lower CID energies than the other two
metabolites. Breakdown curves can be used to recognize noncovalent oligomers, as their
signals decrease sharply with CID potential, and fragment ions, which have abundances
that increase with CID potential until they fragment further. Different classes of
metabolites may be distinguished based on collision energy thresholds and the slopes of

breakdown curves for these ions.

29

 

 

 

 

 

 

’f'
E
on
y
b’

 

r—‘m
is-
:19

 

 

 

 

 

 

 

 

 

 

 

 

e) C

7E7
8 x
E
.9 l

. “5
39 .

Collision potential (V)

 

Figure 2-1. Schematic of LC/multiplexed CID method showing (A) application of
multiple collision potentials to the ion transit lens to effect nonselective ion
fragmentation, (B) mass spectra acquired using five different quasi—simultaneous CID
conditions, showing molecular (r913), noncovalent oligomer (O), and fragment ion species
(1:!) using color-coded symbols to distinguish ions coeluting metabolites, and (C)
breakdown curves showing common behavior for molecular ion species for two coeluting

metabolites, a noncovalent dimer ion, and a fragment ion.

30

To demonstrate proof-of-principle for the multiplexed CID approach, LC/MS
metabolite analyses were performed on solvent dip extracts of leaves of two wild
relatives of tomato, S. pennellii LAO716 and S. habrochaites LA1777. These extracts
each contain more than 300 metabolites detectable by LC/MS, many of them acyl glucose
and acylsucrose esters, flavonol glycosides, glycoalkaloids, and others yet to be
identified. For these analyses, pre-analyzer CID was controlled by adjusting the potential
applied to Aperture 1 in the transit region between the ion source and the TOF analyzer.

Pressures in this region are not controlled or measured on this instrument, but are

. -4 . .
estimated to exceed 10 mbar and are sufficrent to ensure >99% conversron of some

molecular ion species to fragment ions. The instrument control software allows multiple
acquisition functions, each with distinct Aperture 1 settings. For these experiments, five
functions were set up, with Aperture 1 voltages ranging from 10-80 V. The choice of 5
functions was made to enable switching among the functions with a duty cycle
compatible with conventional HPLC separations.

The quasi-simultaneous acquisition of metabolite mass spectra is demonstrated in
Figure 2-2, which presents metabolite mass spectra acquired in negative mode ESI using
five different CID potentials. In each case, the bottom spectrum was generated under the
gentlest conditions (Aperture 1 = 10 V), with the top spectrum acquired under harshest
conditions (80 V). The first set of spectra (Figure 2-2A) demonstrates behavior for the
glycoalkaloid a-tomatine. At CID settings of 10 and 25 V, only the formate adduct (m/z
1078) is observed owing to the lack of acidic groups that favor formation of the

deprotonated ion [M—H]'. With increasing CID potential, [M-H]' at m/z 1032 appears

31

as a result of loss of formic acid, but more extensive fragmentation was only observed at
the highest potential (80 V), primarily involving losses of portions of the oligosaccharide
group. In Figure 2-23, behavior of a triacylglucose (two C4 and one C5 fatty acid esters)
is demonstrated. At the lowest CID energy, noncovalent dimer ions corresponding to
[2M-H]' and [2M+formate]“ were observed at m/z 807 and 853 respectively,
[M+formate]_ at m/z 449 was the most abundant ion, and minor amounts of [M—H]‘ (m/z
403) and a fragment ion at m/z 227 were present but barely detectable. As is often the
case for specialized metabolites, the acylsugar lacks groups with moderate acidity, so
formate adducts were the most abundant signals at lowest CID potential. As this potential
was increased, [M-H]— grew in abundance along with several fragment ions such as m/z
315 and 227, which form via successive losses of neutral C4 fatty acids. At the highest
CID energy, the C4 fatty acid anion (m/z 87) dominates the spectrum, with a smaller
amount of the C5 fatty acid anion observed at m/z 101. These two fragment ion masses
provide evidence for the fatty acid groups. We speculate that differences in yields of the
two fatty acid anions may be attributable to differences in the substitution positions of
these groups, and a more detailed analysis of CID spectra of Solanum metabolites will be

described in Chapter 3 and 4.

 

 

 

 

 

 

 

 

 

 

 

 

OH
O O
O
0H 0.. ”m
4
O O R o OH
HO O 3 /<
o 0 OH 0 R2 0
OH
HO O 0 0H 03:13
Ho OH OH R2,R3,R4 = C4, C4. C5
OH
OH Tomatine
100 576 87090;032 10° 87
167273811 1 l 1““
87
100° 1032 1000 227
A /o 101 I 315403
0 0 l l 1
too 1078 100
%I 1032 J 087112127315403
0 l ll101l1l l
100 1078 100 one-I710 22731‘259344M9
% 1068i 313.1 I ll1 1—
o
100 1078 100 449 853
o 0 807
AiL 0,0 l 1 L

 

Figure 2-2. Negative mode electrospray ionization mass spectra obtained using

multiplexed collision induced dissociation using five aperture 1 voltages (from bottom to

top: 10, 25, 40, 55, and 80 V) for two metabolites extracted from S. pennellii LAO716 by

leaf dip. Panel A: a-tomatine at retention time of 15.25 min; panel B: triacylglucose G

3:13 with two iso-C4 and one branched C5 fatty acid esters with retention time 19.99

min.

33

Annotation of detected ions as molecular, noncovalent oligomer, adduct, or fragment
ions serves as the foundation for interpreting nontarget metabolite mass spectra.
Metabolites often coelute, and this makes such assignments more challenging. However,
multiplexing of CID conditions makes possible the construction of breakdown curves for
every m/z value at every retention time during a chromatographic separation. Figure 2-3
displayed TIC (total ion chromatogram) for S. habrochaites LA1777 and S. pennellii
LAO716 under lowest CID voltage and the metabolites used in Figure 2-4 and 2-5 were
labeled in TIC. To illustrate the practice of this feature, we constructed breakdown curves
(Figure 2-4) for dimer, adduct, and fragment ions and deprotonated molecules derived
from four metabolites extracted from the tomato relatives S. pennellii LAO716 and S.
habrochaites LA1777. The structures for those four metabolites were labeled on the top
of Figure 2-4. The absolute abundances of all dimer ions and formate adducts exhibited a
sharp drop with increasing CID potential (Figure 2-4A, 4B), but most chloride adducts
were more persistent owing to their higher thresholds for decomposition (data not
shown). Various fragment ions appeared only at higher CID potentials (Figure 2-4D) and
the abundances increased with elevated CID voltages. Such behavior allows fragment
ions to be distinguished from molecular ion species, and targets these signals for use in
structure assignments. Abundances of [M-H]" ions display various dependence on CID
potential, often reaching maxima at intermediate voltages (Figure 2-4C) owing to their
formation from decomposition of dimers and some adduct ions. The most important
aspect of this finding suggests that shapes of breakdown curves can guide assignment of

those ions as molecular, noncovalent oligomer, or fragment species.

34

tetraacylsucrose

 

 

 

 

tetraacylsucroseS 4:20 tetrasacylsucrose
4:21
100- tetraacylsucrose S 4:16 /
l S 4115 l l triacylsucrose
. \ ‘\ 1 s 3:22
°\°-« R tin 1 ll ll ll l i/ l
« ” l l 11 .1 11,1 1 11.11.11 1
4 ' ’1 l‘ 1‘ 1‘ ‘5' l . 1| ,’ Lkl ‘l 1' ll . 1 1
0+ f Efl-WVA' [NU/”u‘fltl LL] Wu} \‘k/‘v‘ ‘l\ f\\\v‘/\.\v,/ J u (i if" 1&1ka law“,
10.00 ' 20.00 30.00 40.00
Triacylglucose
1001 Tnagylglpecose G 3318 l Triacylglucose
5 1 G 3:19
°\°‘ tomatine fl {1H ‘ ,1. Triacylglucose
\ (I l / \ fl 1 G 3:20
A1 A l"- ‘h‘trhv' I1 [l
1 ’ . 1' l 1 ii} i“ ‘l‘ l
17* r T 1 r IE- T...... “‘r‘fi—T—"“"“'_"_r‘" a —r——1— 7 *fi' 1 ' 7 ' ' %:M lme
10.00 20.00 30.00 40.00

Figure 2-3. Total ion chromatogram (TIC) for S. habrochaites LA1777 (top) and S.
pennellii LAO716 (bottom) using ESl negative with CID potential at 10 V. The labeled

metabolites were used as examples for Figure 2-4, 2-5 and 2-6.

35

Figure 2-4. Breakdown curves showing energy dependence of various ion abundances
for metabolites extracted fi'om leaf dips of S. pennellii LAO716 and S. habrochaites
LA1777. Ion abundances are normalized to the total ion current for each individual ion
as sumed for all five collision potentials. Panel A: Breakdown curves of non-covalent
dimer ions for the flavonoid glycoside rutin with [2M—H]" (O) and [2M+HCOO]_ (O)
and two acylsugar metabolites, tetraacylsucrose with 21 carbon atoms in acyl groups
detected as [2M-H]' (A) and [2M+HCOO]" (A); dimer ions of triacylglucose with 19
carbon atoms in acyl groups detected as [2M—H]_ (I) and [2M+HCOO]- (El); Panel B:
breakdown curves for formate adduct ions [M+HCOO]- for the glycoalkaloid tomatine
(0), rutin (:1), tetraacylsucrose (C1), and triacylglucose (0); Panel C: breakdown curves for
monomeric [M-H]- ions for tomatine (*), rutin (X), tetraacylsucrose (Cl), and
triacylglucose (6); Panel D: breakdown curves for major fragment ions for tomatine at
m/z 870 (I), rutin at m/z 300 (A), C5 (X) and C4 (C1) fatty acid anion fragments for
tetraacylsucrose, and C5 (0) and C10 (A) fatty acid anion fragments for triacylglucose.

The proposed structures of four metabolites are on the top of breakdown curves.

36

My 1;...

Rutin

0H
R2
G 3:19
1123334 = C4, cs, C10

0H 0H

@000

HO OH O OH

OH
“ Tomatine
Rb) mo:
R3t<0 oj/0 O>/

S4zR321

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

R3,R4 = C4, C5, R31: C10
100 100
80 E 807
9.’
60 3 6d
C
40 -9 40
"6
20 o\° 20
0 , 0
10 25 55 80 10 25 40 55 80
Aperture 1 Voltage Aperture 1 Voltage
A B
50 100
E 530i
~40 “-3
8 '5 60
I: 0
.9. 1:
“520 340
,_.,\° o
o\0 20
c ‘ _
10 25 40 55 80 U 25 4o 55
Aperture 1 Voltage Aperture 1 Voltage
C D

37

4.

L42

{‘9

1111-1?

Shapes of breakdown curves are determined by the ease with which compounds form
various ions. At low CID potentials, the more acidic metabolites were more likely to form
[M-H]_ than adducts, and this was reflected in the dominant abundances of [M—H]_ and
dimer ions in spectra of rutin at low CID potentials. Abundances of formate adducts of
the weakly acidic metabolite rutin had the sharpest drop among adducts of these four
metabolites. The relative abundances of formate adduct ions to deprotonated molecular
species therefore depend on metabolite formate ion affinities, metabolite acidities (e.g.
ease of conversion of [M+formate]' to [M—H]_) and energy barriers toward fragment ion
formation. Collision energy dependences of ions derived from six acylsugars (one
triacylsucrose, two tetraacylsucroses, and three triacylglucoses) were compared
respectively (Figure 2-5 and 2-6). The results highlight the consistency of shapes of
breakdown curves generated by multiplexed CID within each chemical class. Application
of our findings to several hundred metabolite breakdown curves generated by
multiplexed CID suggested that comparisons of breakdown curve shapes have value for
recognition of chemical classes of metabolites. This feature is important for plant
metabolomics since it was common there were no report in database or literature for
many secondary metabolites, but at least they can be assigned to chemical classes under

the help of breakdown curves.

38

 

 

 

 

 

 

 

60 80
E 8 60
E40 s
U U
.5 .5 4°
*5 20 "(Z 20
o\° °\
\ 0‘
05 25 4'5 6'5 85 10 25 40 55 80
Aperture 1 voltage Aperture 1 voltage
A B

Figure 2-5 A: Breakdown curves derived from metabolites detected during LC/negative
ion ESI MS analyses of a leaf dip extract from S. habrochaites LA1777, showing
collision energy dependence of abundances of (A) formate ion adducts of three
acylsucrose metabolites ([M+HCOO] at m/z 737 83:22 (O); [M+HCOO] at m/z 653
S4115 (Cl); [M+HCOO]’ at m/z 723 S4220 (A) and (B): fatty acid fragment ions of
various carbon lengths (C5 fatty acid at m/z 101 (l) and C12 fatty acid at m/z 199 (El)
for 83:22), (C4 fatty acid at m/z 87 (A) and C5 fatty acid at ni/z 101 (A) for 84:16), (C4
fatty acid at m/z 87 (O) and C10 fatty acid at 171 (O) for $420). The acylsucroses are
designed as 83:22, S4216 and S4:20, with the letter S representing the sucrose core, the
first number represents the number of fatty acyl groups, and the second number

corresponds to the total number of carbon atoms in the fatty acyl groups.

39

 

 

on
.0

 

 

 

 

 

 

 

 

‘6 :5.
260 e
3 3
0 O
.540 .5
“5 “5
$201 °\°
0 - a _ r _. 0,. e . .
5 25 45 65 85 5 25 45 65 85
Aperture 1 voltage Aperture 1 voltage
A B

Figure 2-6 Breakdown curves derived from LC/negative ion ESI MS analyses of a leaf
dip extract from S. pennellii LAO716 showing collision energy dependence of
abundances of: (A) formate ion adducts of three triacylglucose metabolites ([M+HCOO]“
at m/z 547 G3:20 (O); [M+HCOO]_ at m/z 519 G3:18 (Cl); [M+HCOO]— at m/z 449
G3:13 (A) and (B): fatty acid anion fragments of various carbon lengths. (C5 fatty acid
at m/z 101 (O) and C10 fatty acid at m/z 171 (El) for G 3:20); (C4 fatty acid at m/z 87
(A) and C10 fatty acid at m/z 171 (X) for G3:18); (C4 fatty acid at m/z 87 (*) and C5
fatty acid at m/z 101 (O) for G3:13). The triacylglucoses are designated as 63:13, G3:18,
and G320 with the letter G representing the glucose core, the number 3 refers to three
fatty acyl groups, and the last number corresponds to the total number of carbon atoms in

the fatty acyl groups.

The discussion above has emphasized behavior of negative ions, but similar results

were obtained using positive ion mode. For acylsugar metabolites, the adduct ion
[M+NH4]+ for all tested acylsugars underwent fragmentation at lower CID threshold
energies than [M+Na]+, and MS/MS product ion spectra showed that most fragment ions
arose from [M+NH4]+ and not [M+Na]+.

We considered the possibility that analyte concentration could influence shapes of
breakdown curves owing to formation of dimer and other oligomer ions at higher analyte

concentrations. Breakdown curves were generated using LC/multiplexed CID for rutin

40

standard ranging from 20 to 800 pmol injected (2-80 uM solutions). Breakdown curves
displayed in Figure 2-7 showed minimal concentration-dependence for fragment ions.
Earlier studies have shown that influences of solvent composition and pH on breakdown

curves were negligible.‘55

4]

Figure 2-7 Breakdown curves for the flavonoid glycoside rutin obtained using negative
mode electrospray ionization following injections of 10 uL of solutions at five different
concentrations (2 (O), 5 (Cl), 10 (A), 40 (X), and 80 (0) uM). A: non-covalent dimer
ion [2M—H]" B: formate adduct ion [M+HCOO]_ C: monomer deprotonated ion
[M—H]_ D: aglycone radical anion fragment with m/z 300 and E: aglycone fragment ion
with m/z 301.

42

 

 

\___

 

100

 

WOOnwO

8 6 4 2
029 228910 o\o

 

 

5
8

.5
6

4'5
Aperture 1 voltage

25

 

 

0 0 0

4. 2
E930 :2 E a.

Aperture 1 voltage

B

A

 

 

 

 

 

0
6

0 nw
4 2

Emtso co_ *0 fix.

 

 

O
3

0 0
2 1
E250 :2 *0 .x.

 

0

85

45
Aperture 1 voltage

Aperture 1 voltage

 

 

65 85

Aperture 1 voltage

 

0
6

0 0
4 2
E950 co_ E .x.

 

43

Two great advantages of the multiplexed CID approach lie in its ability to generate
fragment mass information for low abundance metabolites and to discriminate coeluting
metabolites based on different collision energy thresholds for fragmentation. To illustrate
this point, we present an example using two coeluting metabolites derived from a leaf dip
extract of S. pennellii LAO716 (Figure 2-8). Two metabolites of minor abundance coelute
at 22.5 min, a triacylglucose (with two C4 and one C5 fatty acid ester groups) that gives a
formate ion adduct at m/z 463, and trimethquuercetin that yields [M—H]_ at m/z 343 that
is barely evident in the low voltage spectrum. Triacylglucoses undergo facile
fragmentation and lack acidic functionality. Even at this lowest collision potential, the
spectrum displays multiple triacylglucose peaks including deprotonated dimer (m/z 835),
formate-bound dimer (m/z 881), the deprotonated triacylglucose (m/z 417), chloride
adduct (m/z 453 and 455) and a fragment ion (m/z 375, corresponding to loss of a neutral
C4 fatty acid from the formate adduct) are observed at higher abundance than the
deprotonated molecule of trimethquuercetin. The mass spectrum generated at 25 V CID
potential shows fragment ions derived from the triacylglucose that correspond to neutral
losses of the fatty acid groups (m/z 329, loss of C4 fatty acid from [M—H]_ and m/z 241
and 227 arise from subsequent losses of C4 or C5 fatty acid respectively). Fragments
corresponding to the anions of C4 and C5 fatty acids are also evident at m/z 87 and 101.
In contrast, the spectrum at 80 V is characterized by fragment ions derived from
trimethquuercetin, which requires greater collision energy than the triacylglucose to
yield abundant fragment ions. At the higher collision potential, the triacylglucose is
largely converted to low mass fragment ions, which no longer obscure the

trimethquuercetin fragments.

44

&

 

 

 

100‘ 313

‘ 8

; a 299 &

2 1 1

. 87 14 343 417
0 1 ....... [,1 ........ 191,421,4114-9- ..... l "1...: ..................................................................................................................
100.0(10 u

i 0
t

0/ * O
. o: * * 227 343 * A
: * 101143 375 453 o Rz/Q 835

a: * *
4 37 197 241 315

 

 

 

 

 

R3
0 3 '
'k
463
1007 I
. l 10 V
l 8‘ 375 453 835 881
l
l 343. l l l J, ”V7
o ..................................................................... , .................................. , .......................................................................

Figure 2-8. Multiplexed CID mass spectra (electrospray, negative mode) of coeluting
metabolites from LC/T OF MS analysis of an extract of S. pennelli LAO716, showing
spectra obtained from three of the five Aperture 1 potentials. The lowest collision
potential (10 V) yields multiple ions from a triacylglucose (83:14) including formate and
chloride adducts (m/z 463 and 453), deprotonated (m/z 417), fragment (m/z 375), and
dimer ions (m/z 881 and 835). Of markedly lower abundance is [M-H]" of
trimethquuercetin (m/z 343). Slighly higher CID potential (25 V, middle panel) shows
extensive fragmentation of the triacylglucose (ions marked with *); spreading the signal
among numerous fragment ions allows more ready observation of the methylated
flavonoid ion (marked with &). Fatty acyl anions are observed at m/z 87 and 101
corresponding to C4 and C5 fatty acyl groups. In the top panel, the highest collision
voltage of 80 V yields fragmentation of the methylated flavonoid through losses of
methyl groups.

45

Multiplexed CID-TOF mass spectra generate information about molecular and
fragment ion species including accurate mass measurements and elemental composition
based upon abundances of isotopomer ions. With proper instrument calibration,
multiplexed CID-TOF spectra yielded ion masses accurate to within 0.005 Da of true
values for all molecular and fragment ions that have adequate ion statistics for mass
measurement. The coupling of muxCID with TOF mass measurement allows ions of
overlapping nominal mass to be distinguished based upon mass defects (the fraction of

ion mass following the decimal point). Among elements commonly found in metabolites,

1 14 . . . .
H and N make pOSitive contributions to mass defect (+7.8 and +3.1 mDa per nucleus,
. . 16 31 32 . . .
respectively), heav1er elements ( O, P and S) make negative contributions to mass

defect (-5.1, -26.2 and -27.9 mDa respectiveIY), and 12C contributes zero to mass defect.

Owing to the high number abundance of hydrogen and low frequency of phosphorus and
sulfur in most metabolites, the mass defect for molecular, dimer, and fragment ions
largely reflects the number of hydrogen atoms.

High mass accuracy provides information essential for metabolite discovery, but
even technologies capable of the highest accuracy mass measurements fail to yield

'56 The number of elemental formulas that

conclusive molecular formula assignments.
yield agreement with a measured mass grows exponentially with mass, and even mass
measurement accuracy of 1 part-per-million can result in hundreds of matching formulas
for metabolites with molecular masses greater than 500 Da. Combining accurate mass
measurements with quantitative assessments of abundances of ions with heavy isotopes

can limit the range of matching elemental formulas. Extending this approach to include

accurate fragment mass measurements and fragment isotopomer abundances adds

46

additional discriminating information about metabolite formulas and structures.157
Linking of fragment and molecular ion masses through their common chromatographic
elution profiles enjoys common practice in GC/MS analyses, and has been implemented
in such context using software packages such as AMDIS.‘58 We suggest that rapid
multiplexing of CID with accurate measurements of mass and isotopomer abundances for
molecular and fragment ions adds the additional dimension of measuring energy-
dependence of ion abundances to accelerate discrimination of metabolites through several
features: (1) suggestion of elemental formulas for molecular and fragment ions, (2)
recognition of functional group types via assignments of group-specific fragment ions or
neutral mass losses, and (3) discrimination of molecular structure and ion types according
to ion appearance thresholds from breakdown curve analysis. Multiplexing of CID in a
single analysis allows this approach to reach low-abundance metabolites providing that
ion signals are abundant enough to allow for precise measurements of these signals, and
does not require data-dependent ion selection for MS/MS scans. In essence, this approach
provides the benefits of MSE analysis99 with the additional dimension of energy-
resolution of ion fragmentation.

When chromatographic separations of metabolites are incomplete, multiple
metabolites will arrive at the mass spectrometer at the same time. In such cases, extracted
ion chromatogram profiles and mass defects are useful aids for assigning fragment ions
with appropriate molecular ion species. This is illustrated in the case of coeluting
glycoalkaloid and flavonoid metabolites present in extracts of leaves from an
introgression line (IL) 1-1 derived from S. lycopersicum M82 and S. pennellii LAO716.

Based on extracted ion chromatograms generated at low CID potential (10 V), the

47

chromatographic profiles of two metabolites detected at m/z 609 and 1076 overlap, with
retention times of 2.46 and 2.50 min. respectively (Figure 2-9). Second pair of
metabolites, detected at m/z 593.15 and 1094.54 are not resolved at all, with a retention
time of 2.60 min for both.

For the first pair of metabolites which overlap, the TIC shows a skewed peak that
does not resolve the two metabolites. A composite spectrum was generated for the 2.46
min peak by averaging spectra across the peak (Figure 2-9 A and B XICs). This was
performed for all five CID potentials, and spectra for three of the five acquired functions
are displayed in Figure 2-10. The lowest energy mass spectrum shows three prominent
peaks at m/z 609.15, 655.15, and 1076.52, and additional ions are observed at higher CID
potentials. Extracted ion chromatograms (XICs) were generated for all ions observed at
the highest (80 V) CID potential, and these are displayed in Figure 2-11. Fragment ions
commonly inherit the elution profiles from molecular ions, and two distinct sets of
chromatographic profiles were observed, with one set of ions eluting with a peak apex at
2.46 min, and the other set eluting with an apex at 2.50 min. The earlier eluting peak is
annotated as quercetin-3-O-rutinoside, and the second peak corresponds to an isomer of
dehydrotomatine that is described in more detail in Chapter 4. In this case, resolution of
the ultraperforrnance LC column provided distinct chromatographic elution profiles that

allowed fragment ions to be associated with specific metabolites without MS/MS spectra.

48

 

 

 

 

100:1 D ' m/z 593
°o JL
0 ' .7 l I i I I AMT # AA 7A A g?
1.0 2.0 3.0 4.0
*
10 C 2'62 m/z 1095
% A ,
c I A I I A I I I A I I L I
1.0 2.0 3.0 40
2.72
10 B 2.50 m/z 1077
% 3.
c I I I I I I I I I
1.0 2.0 3.0 40
2.46
10 A m/z 609
% &
C T I I I I I I I I
1.0 2.0 3.0 4,0
10 &
% W
M *
G I I I I I I I

 

1.0 2.0

Figure 2-9 TIC for ILl-l (Bottom) and XIC of m/z 609, 1076 (elute at 2.46 min), 1094
and 593 (elute at 2.60min) under ESI negative mode. The peaks from XICs of m/z 609
and m/z 1077 labeled with & were corresponding to coeluting peak in TIC at RT = 2.46
min; the peaks from XICs of m/z 593 and m/z 1095 labeled with * were corresponding to

peak in TIC at RT = 2.62 min.

49

 

 

 

 

100 609
, i
% 300 1 80V
0 E O o 883 0 10311 O
0 255271 574 I 670 736 X 8?8 1 967
o
100 609
96 5 55v
0 : 0 o
r 1031
C 30.1 3r ‘ l 1957
o
100 599
I 0 10V 0
0 tr _ . ,

 

 

Figure 2-10 Averaged mass spectra across the LC/MS TIC peak eluting at 2.46 min

using three CID voltages (10, 55 and 80 V). The metabolites are from an extract of

trichomes from IL 1-1. The ions labeled with o belong to one dehydrotomatine isomer

and the ion peaks labeled with O are corresponding to quercetin-3-O-rutinoside (rutin).

50

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

OH OH OH OH 0 o
o 0
HO I O O 0 OH
H OH OH HO OH (334311
OH
rutin 0“ dehyrotomatine isomer
Apex=2.46min Apex=2.50min
100 g 100 I
% m/z 655 j‘y % "V2 1076 I
0 ‘5’ 0 ' ‘5
s 100 '
1(3/2 L m/Z 609 i, % m/z 1030 , I\ \
3 I
3 I
100 |__
120 m/z 300 \ % m/z 898 ,1 a
a 1
= I
100 E 100
% m/z 271 :\\ % m/z 868 1’ i“ \\
0’ gr 0 - .
5 I
100 i 100 m/z 736 L.
% [Tl/Z 255 -._\ F 0/0 1': ‘
0 ‘ 3: “ “ ‘““ o ’77 "“ I 7-
g I
"3° m/z 179 100 m/z 574 .-,
/° ‘ % . I
O ~M___/ «v «w...— 0 , / ' k-‘.__

 

 

 

 

Figure 2-11 XICs for the ions labeled with o and 0 from Figure 2-10 were generated for

LC/MS analysis of an extract of IL 1-1. The chromatograms on the left are assigned as

being derived from quercetin-3-O-rutinoside (rutin) and the chromatograms on the right

are related to an isomer of dehydrotomatine.

51

It is almost unavoidable that some metabolites will have indistinguishable
chromatographic elution profiles, and this was the case for metabolites eluting at 2.60
min. (Figure 2-9, C and D XICs). The spectrum derived from averaging across the TIC
peak showed major ions at m/z 593.15, 639.15, and 1094.54 using the lowest CID
potential. As before, spectra were also generated at raised CID potentials (Figure 2-12).
Breakdown curves were calculated for the three ions above (Figures 2-13 and 2-14). The
peaks at m/z 639.15 and 1094.54 had disappeared when the CID potential was raised to
55 V. Such behavior is consistent with both being formate adduct ions, and this
assignment is supported by the appearance of ions 46 Da lower (corresponding to
[M—H]_) for both cases. The abundance of the ion of m/z 639.15 drops off more quickly
with increasing CID potential than is the case for m/z 1094.54. This behavior suggests
greater acidity or proton mobility in the former, because a proton must migrate to allow
for neutral loss of formic acid. Also, at 55 V CID potential, two fragment ions appear at
m/z 285.04 and 284.04 (unlabeled), and at the highest CID potential (80 V), additional
fragment ions appear at m/z 255.03, 227.02, 592.39, 754.44, 886.48, 916.50. Support for
annotation of these ions as fragment ions comes from slopes of their breakdown curves,
which increase with increasing CID potential. It is worthy to note that fragment ions at
m/z 592.39, 754.44, 886.48 and 916.50 all have even masses and positive mass defects
above 390 mDa, which are traits common to the ions at m/z 1094.54 [M+HCOO]" and
1030.52 [M-H]'. In addition, the shapes of breakdown curves for these molecular ion
and fragment ions were similar with corresponding ions from dehydrotomatine (Figure 2-
13). These similarities in mass defects and breakdown curves suggested this metabolite

has similarities with other glycoalkaloids. Based on the combined information from

52

accurate mass measurements and breakdown curves, this metabolite was assigned as
hydroxytomatine. Ions at m/z 593.15, 639.15, 284.04, and 285.04 did not fit behavior or
mass defects of ions from glycoalkaloids, and were attributed to a coeluting metabolite
with structural units and elemental compositions distinct from the glycoalkaloid family.
Fragment ions at m/z 284.04 and m/z 285.04 gave calculated elemental formulas
characteristic of fragments corresponding to the aglycone kaempferol, and breakdown
curves were similar to the flavonoid glycoside rutin (Figure 2-14). Based on the sum of
information from accurate mass measurements and breakdown curves, this metabolite
was assigned as kaempferol-3-O-rutinoside. Thus, with the help of breakdown curves and
mass defect, even coeluting metabolites can be distinguished and annotated. The

annotation of these and other metabolites is described in more detail in Chapter 4.

53

A
100 593

 

 

 

 

 

 

80V
% 234
A A A , .
227 255i 285 427 592 658 7'54 8.35 9'18 105191085
0 ‘ g 1 \ \A L /
A
100 593
96 55\/ .
A r 1049 .
285 J 1085
0
A
100 593 A
10 v '
9b 639 1095

 

 

 

Figure 2-12. Negative mode mass spectra from an extract of IL-l-l trichomes. Spectra
were generated for three CID voltages (10, 55, and 80 V) by averaging spectra across the
TIC peak eluting at 2.60 min. The peaks labeled with A were attributed to kaempferol-3-

O-rutinoside and the peaks labeled with o are attributed to hydroxytomatine.

54

Figure 2-13 Breakdown curves showing energy dependence of various ion abundances
for glycoalkaloid metabolites eluting at 2.46, 2.60 and 2.70 min extracted from trichomes
of IL 1-1. Ion abundances are normalized to the total ion current for each individual ion
as sumed for all five collision potentials. Breakdown curves of A: formate adduct ion
[M+HCOO]_ for dehydrotomatine isomer (9), hydroxytomatine (A), dehydrotomatine
(Cl); B: monomeric [M-H]_ ions for isomer of dehydrotomatine (O), hydroxytomatine
(A), dehydrotomatine (D); C: major fragment ions for dehydrotomatine isomer at m/z
898 (O) and at m/z 574 (1:1); for dehydrotomatine at m/z 898 (A) and fragment ion at m/z
574 ( - ); for hydroxytomatine at m/z 916 (4‘) and at m/z 592 (o).

55

 

 

 

 

 

HN N
o o
H H
Hog H H HOQO IP“
H H OH HO OH OH
OH . OH . .
0” dehydrotomatine OH dehydrotomatine isomer
HN
o s
OH
H
OH 0 0
HO 0 El
OH
HOQ or—PH
HO OH OH
H OH hydroxytomatine
60 ——A»—~VA77~1 100 fi fi— 7 v —;
l 80 l
. k
40 ¥;:.;;ig\x I 60 . ' ‘
. ‘ A l
20' ‘ ‘9. ‘ 4° . .4 l
\L.\\ I ZO‘I , '4 ... j i
0 * brim} I 0L“, lg" , . J
5 25 45 65 85 5 25 45 65 85
Aperture 1 voltage Aperture 1 voltage
A B
100 ,1
80
60 / l
40 y/ l
.1/ l
20 .,
0 ,H'.;.;;£;(.,....; .. '/ _.,~i ,7 7 31‘
5 25 45 65 85

Aperture 1 voltage

C

56

Figure 2-14 Breakdown curves showing energy dependence of various ion abundances
for flavonoid metabolites eluting at 2.46 and 2.60 min extracted from trichomes of IL 1-
1. Ion abundances are normalized to the total ion current for each individual ion as
sumed for all five collision potentials. Breakdown curves of A: formate adduct ion
[M+HCOO]_ for rutin (Cl) and kaempferol—3-O-rutinoside (O), B: monomeric [M—H]_
ions for rutin (D), kaempferol-3-O-rutinoside (O), C: major fragment ions for rutin at m/z

301 (El), for kaempferol-3-O-rutinoside at m/z 285 (O).

57

 

 

kaempferol-3-rutinoside

rutin

 

 

 

 

 

 

 

0 0 n. 0 0
N no go 4 2 0
®D_N> mus-Owfl< ho o\o
3;
i u
0 0 _ 0 D... .
w 8 m A» 2 O

0205 05.03.13 5\5

65 85
Aperture 1 voltage

45

25

45 65 85

Aperture 1 voltage

25

B

85

 

 

0
0
1

45
Aperture 1 voltage

 

i-——1—¢ ' -
25

00000
8642

02m> 0u20mn< 50 0\0

58

Quantitative profiling is an essential aspect of correlating metabolite levels with gene
expression. The initial impetus for this work was driven by the limited dynamic range of
TOF mass spectrometers, coupled with the idea that multiplexed CID could extend this
range by attenuating strong molecular ion signals via inducing their fragmentation. In
large-scale metabolite screening that can involve many thousands of LC/MS analyses,
repeating each analysis using a more dilute sample is often not practical. In Figure 2-15,
integrated ion currents for ions from the flavonoid glycoside rutin demonstrate how
multiplexed CID extends the dynamic range of LC/TOF MS. At low collision potentials.
the response of [M-H]_ shows a steep slope but deviates from linearity at concentrations
in the low pM range owing to formation of noncovalent dimer ions at higher rutin
concentrations. As expected, a collision potential of 80 V yielded a less steep slope but
not just from a proportional reduction in ion current. The high collision potentials
induced dissociation of adducts and noncovalent complexes and yielded linear response

up to the highest concentration (50 uM) tested.

59

 

8000

6000 .

4000 7

Peak Area

2000 1

 

 

A
v

 

 

0 I A j
0 10 20 30 40 50

Concentration (uM)

Figure 2-15. Concentration dependence of extracted ion chromatogram (XIC) peak areas
for ions derived from rutin at different collision potentials using 10 uL injections of rutin
standard solution and negative mode electrospray ionization. Curve A (A): the
deprotonated molecular ion [M—HI using the lowest CID potential (10 V); B (O): the
deprotonated molecular ion [M-HT‘ using the highest CID potential (80 V); C (0):
fragment ion at m/z 300 from rutin using the highest CID potential (80 V); D (l): the
rutin noncovalent dimer ion [2M-H]_ using the lowest CID potential (10 V); E (O): the
rutin noncovalent dimer ion [2M-H]' using the highest CID potential (80 V). N=3

separate plants, error bars represent standard error of the mean.

Matrix suppression can be revealed and minimized through use of multiplexed CID.
Hydrophobic components that coelute with specific analytes are notorious contributors to

'59 These components preferentially partition to surfaces

matrix suppression of ionization.
of electrosprayed droplets and compete for ionization. In some cases, they form
heterodimeric ions with other analytes, reducing molecular ion signals of the latter.

Matrix suppression can be reduced through use of more aggressive CID conditions that

drive dissociation of heterodimeric ions toward molecular ion forms. The behaviors of

60

three major isomers of sesquiterpene acid metabolites in extracts of the tomato relative S.
habrochaites LA1777 are presented as examples of this effect. These metabolites elute in
the midst of numerous acylsugar metabolites that are more than lO-fold more abundant,
often yielding ions corresponding to noncovalent heterodimers between acylsugar and
sesquiterpene acid. Figure 2-16 shows breakdown curves for three isomeric deprotonated
sesquiterpene acids ([M—H]: m/z 233). The breakdown curve for the second isomer
shows behavior typical of deprotonated molecules in the absence of coeluting
interferences, with abundance decreasing as CID potential is increased. In contrast,
abundances of corresponding ions for the other two isomers increase with CID voltage.
This behavior is attributed to increased formation of [M*H]‘ at elevated collision
energies via dissociation of heterodimeric ions. Substantial signals were observed for
noncovalent heterodimer ions (not shown) formed by associations of sesquiterpene
isomer 1 with a triacylsucrose, and sesquiterpene isomer 3 formed heterodimer ions with
a tetraacylsucrose. These findings highlight the diagnoses of effects of coeluting

interferences through examination of breakdown curves for molecular ion species.

61

 

 

 

100
% A C
o ............................ f- . . , - . , .
10 20 30 40
100
%
0 :‘v vvvvvvvvvvvvvvvvvvvvvvvv fl .......... Time
10 20 30 40

 

3000 -

N
O
O
0

Peak Area

1000 ~

 

 

 

 

 

5 25 45 65 85
Aperture 1 voltage

Figure 2-16 (Top) TIC of S. habrochaites LA1777 and XIC of m/z 233 using ESI
negative mode with CID potential at 80 V. (Bottom) Breakdown curves derived from
LC/negative ion ESI MS analyses of a leaf dip extract from S. habrochaites LA1777
corresponding to three isomers A (O), B (I) and C (A) of deprotonated sesquiterpene
acid ([M-H]_ ions at m/z 233) showing isomer-dependence of signal upon collision
potential. Ion yields for isomers A and C exhibit an unusual positive slope, suggesting
contributions toward these ion abundances from fragmentation of heterodimers that result

from coelution with abundant acylsugar metabolite.

62

2.4 Conclusions

Acquisition of LC/TOF MS data using multiplexed nonselective CID with LC/TOF
MS provides accurate ion masses and ion breakdown curves on the ultra-performance
chromatographic time scale. The approach employs a suite of instrumental conditions that
are consistent across analyses as needed for comparisons of large numbers of diverse
samples in metabolomic discovery research. The approach enables deep coverage of the
metabolome, yielding accurate measurements of molecular and fragment masses and ion
behavior in a collision energy-resolved context. Multiplexed CID has particular value
when limited analyte quantities or time constraints preclude repeat analyses.

One might ask whether more than two CID conditions are necessary, or whether there
is an optimal number of CID conditions that ought to be employed. Multiplexing of CID
conditions exacts minimal costs in terms of LC/MS method performance, providing the
duty cycle is compatible with adequate sampling across chromatographic peaks. Our
findings, using either fewer or more CID conditions, provided minimal additional
information relative to five collision potentials. Use of five CID potentials provides a
minimal definition of shapes of breakdown curves for all detected ions including
appearance thresholds for fragment ions, and can direct assignment of those ions as
molecular, noncovalent oligomer, or fragment species. The combination of this
information offers the prospect for recognition of chemical classes of metabolites. These
improvements in analytical throughput promise to accelerate discoveries in emerging
areas including profiling of temporal and spatial dependence of metabolomes or analyses

of other high complexity samples.

63

Multiplexing of collision conditions also provides a rich source of information for
quantitative comparisons of metabolite profiles. Collision induced fragmentation can
attenuate ion signals and extend dynamic range for abundant metabolites. As is the case
with traditional GC/MS with electron ionization, metabolites can be quantified based on
various ions including molecular and fragment species, and virtually guarantees that both
kinds of ions will be detected for all metabolites. In addition, elevated collision potentials
can drive dissociation of heterodimeric ions formed by coeluting metabolites, and
decrease matrix suppression of molecular ion formation.

One of the most promising applications of LC/TOF MS with multiplexed CID lies in
large-scale screening of metabolic phenotypes. During the past year, our laboratory has
used a five-minute LC/TOF MS protocol that employs multiplexed CID to screen plant
genotypes for associations with metabolic traits. This approach has generated analyses of
300 samples over a 30h period and yielded more than 1500 analytical signals after
automated peak alignment and integration. The additional dimensionality of the MS data
(CID energy) aids metabolite annotation even when chromatographic profiles overlap.
More detailed results of this rapid method for discoveries of genes that influence

metabolite levels will be described in chapter 5.

64

Chapter 3. Annotation of acylsugar metabolites in glandular trichomes
from accessions of the genus Solarium based on LC/TOF MS coupled with
multiplexed CID

65

3.1 Introduction

Secretory and glandular trichomes (SGTs) of wild tomato relatives within the genus
Solanum exude a cocktail of secondary metabolites that play critical roles in protection
against biotic and abiotic stresses including herbivory, pathogen infection, extreme
temperature and excessive light.‘06‘I60 Many of these specialized metabolites were
reported to have important potential value as commercial products such as drugs, natural
pesticides, consumer products, or food additives.'06"61'163 SGTs are specialized plant cells
that are prolific chemical factories, but our understanding of the functions that confer
such biochemical productivity remains limited. Identification of the genes involved in
biosynthetic pathways and regulation of biosynthesis of these secondary metabolites will
facilitate manipulation of metabolite production in plants, allowing for improved insect
resistance in tomato and other crops. Furthermore, expression of these genes in other
organisms such as bacteria and yeasts offers potential for using microorganisms as
biological catalysts for specific chemical transformations. However, discoveries of genes
involved in metabolite biosynthesis require profiling and identification of metabolites in
an assortment of different cell types in numerous plant genotypes. Only then can gene
expression be correlated with plant chemistry.

Assignment of structures to SGT metabolites has been slow to develop. Much of our
knowledge has been derived for a limited range of plants, primarily S. pennellii
LAO716108’I32’164 and S. habrochaites LA1777,”5 and has been focused on a limited
number of acylsugar metabolites that were purified from plant extracts. Most knowledge
of metabolite structures was obtained from NMRH8’M’5'167 or GC/MS'68"70 analysis.

However, NMR requires previous purification of each compound and is not practical for

66

large numbers of different plants. Many metabolites lack volatility needed for GC/MS
analysis. As a result, such compounds must be derivatized before analysis, and these
procedures are often time-consuming, and structures are inferred from chemical
derivatives that differ from the original metabolite. Such barriers might be overcome in
part through use of LC/MS, which avoids derivatization steps, and LC/MS approaches
are finding increasing use in global characterization of metabolites in mixturesmo’lm’nl'
174 Mass spectrometry is most powerful, when used for metabolite identification, when
accurate molecular mass measurements allow assignment of an elemental formula to each
metabolite. In recent years, coupling of LC separations to mass analyzers such as
TOF'O3’I75 or FTICRW"78 have facilitated assignments of formulas to previously
unidentified metabolites. Yet more information can be generated by selecting metabolite
molecular ion species and generating MS/MS fragmentation patterns that aid structure
assignments.'°3"72"74"79

While coupling of HPLC separation with tandem mass spectrometry provides
valuable information for profiling and identification of secondary metabolites, it suffers
from important limitations. The standard approach employs a data dependent strategy,1 '9
in which the data system performs real-time selection of molecular ions followed by
acquisition of MS/MS product ion spectra for the selected ions. The nature of these
analyses dictates that the analysis is different for each analysis, and this complicates
quantitative comparisons of metabolite levels in different sample extracts. Furthermore,
many of the mass analyzers are used for such analyses offer limited mass measurement

accuracy for fragment ions. The approach described below for metabolite identification

has relied on methods that fragment all ions in a data-independent fashion, yielding

67

accurate measurements of fragment ion masses. This multiplexed CID approach,
described in Chapter 2, provides important information about metabolite structures and
quantities with high throughput needed for screening large numbers of samples.

The multiplexed CID method provides for quasisimultaneous generation of molecular
and fragment ions by switching the CID potential from low to high in this approach.
Accurate mass measurements on molecular and fragment ions in a single analysis
simplify spectrum interpretation and support elemental formula assignment. Use of
multiple collision conditions in a single analysis yields information about the collision
energy-dependence of all ion abundances, providing additional evidence for assignment
of ions to chemical classes of metabolites.

To test the utility of LC/multiplexed CID MS for metabolite profiling and identification,
this approach was first applied for identifying metabolites extracted from two wild
tomato relatives, Solanum pennellii LAO716 and S. habrochaites LA1777. A limited
number of metabolites from these accessions had been characterized in earlier studies.
Demonstration of the value of LC/multiplexed CID MS was then achieved through
analyses of a different accession for each species, S. pennellii LA1522 and S.
habrochaites LA1353. Acylsugars, alkaloids, phenolic acids, flavonoids and terpenoids
are also annotated for these and additional Solanum accessions. Results from multiplexed
CID are compared to results from product ion MS/MS spectra for many metabolites in

order to support interpretation of fragment ion formation.

3.2 Experimental Section

68

3.2.1 Materials. Plants were grown in a greenhouse in Michigan State University and the
seeds were obtained from C. M. Rick Tomato Genomics Resource Center at the

Univeresity of California-Davis.

3.2.2 Chemicals. Acetonitrile (HPLC grade), 2-propanol (HPLC grade), methanol
(HPLC grade), and formic acid 88% were obtained from VWR Scientific and Fisher

Scientific.

3.2.3 Extraction Method. To extract metabolites from Solanum trichomes, ~100 mg of

individual leaflets were dipped in 2 m1 isopropanol: CH3CN: H20 (323:2 v/v/v) for 1 min.

A 1.0 ml aliquot of each extract was transferred to a 1.5 ml polypropylene
microcentrifuge tube, and the tubes were centrifuged (5000 rpm) for 2 min at 25°C, and

the supematants were analyzed without further processing. Extracts were stored at -20°C.

3.2.4 High-Performance Liquid Chromatography. A Shimadzu (Columbia, MD) LC-
20AD HPLC ternary pump was used. The separation was performed on a Therrno 1 X
150 mm BetaBasic C13 column (100 X 1 mm, 3 pm). Gradient elution was executed based
on 0.15% aqueous formic acid (solvent A) and methanol (solvent B) as follows: initial
conditions 5% B; linear gradient to 50% B (5 min), 95% B (33 min), and 100% B (35

min); hold at 100% B until 38 min, and return to 5% B (43 min). The flow rate was 0.1

. -l . . . . . .
ml-min , and the injection volume was 10 ii]. The column was maintained at 30 °C in a

column oven.
3.2.5 Mass Spectrometry. Analyses were performed using a model LCT Premier mass

spectrometer (Waters, Milford, MA). Operating parameters were set as follows: capillary

voltage: 3.2 kV in both positive and negative mode; desolvation gas flow: 300 L-h-I;

69

-l
desolvation gas temperature: 200 °C; source temperature: 90 °C; cone gas flow: 20 L-h .

The instrument was operated in both negative and positive ion electrospray modes using
separate injections of extract for each ionization mode. Analyses were performed using
rapid switching of Aperture 1 voltage in the ion transit region of the mass spectrometer,
proving quasisimultaneous generation of spectra under fragmenting and nonfragmenting
conditions. Mass spectra were acquired over m/z 50 to 1500 with a scan time of 0.4 s for
each fimction. Spectra were acquired in centroid format using ‘dynamic range
enhancement’ (DRE) enabled, which acquires separate mass spectra by switching a beam
attenuator on and off, followed by post-acquisition stitching together of spectra. The
instrument mass scale was calibrated daily in both ionization modes over m/z 50—1500
using 0.1% aqueous phosphoric acid as reference. Most analyses were performed using
the V-mode ion pathway in the mass analyzer, which generated a resolving power of

5000 (FWHM). For some analyses requiring more exact mass measurements, 0.1%

H3PO4 was infused through the reference sprayer of the lock-spray source as a reference

lock mass, and the lockspray attachment switched between sample and reference streams.

3.2.6 Characterization of Metabolites. Measured ion masses were used to generate
possible elemental formulas using the Masslynx elemental composition calculator
(Waters), using a mass tolerance of 10 ppm. In addition, isotopic pattern filter or the mass
defect can assist to determine the natural of elemental composition.

3.2.7 Nomenclature. To simplify the reporting of acylsugars, they are described using a
single letter corresponding to the base sugar (G = glucose; S = sucrose) followed by the

designation of number of fatty acyl groups and total number of acyl carbon atoms. For

70

example, a glucose triester acylated with three C5 fatty acids would be designated as G

3:15.

3.3 Results and Discussion

3.3.1 Structure annotation for glucose trimesters in S. pennellii LA0716

In earlier studies, acylsugars were identified as major components in SGTs from
several species in the plant genus Solanum.”5‘ '64 Known acylsugars are based upon
either glucose or sucrose as a core, and most of these sugars are esterified with 3-4 fatty
acyl groups consisting of either straight or branched chains ranging in carbon numbers
from 2 to 12.10“” Acylglucoses are the major metabolites exudates from glandular
trichomes of the wild tomato relative S. pennellii'm'l'l' 164’ '80. Burke and coworkers'64
discovered nine 2,3,4-tri-O-acylglucose esters in S. pennellii using magnetic NMR,
GC/MS and desorption chemical ionization (DCI) MS (Figure 3-1). Only three
acylsugars were purified as needed for complete structure elucidation by NMR owing to
the limited separation of metabolites by preparative chromatography. In all cases, the
positions of fatty acyl groups were assigned at 2-, 3-, and 4- positions using NMR. This
suite of acylglucoses consisted of variants containing iso-C4, iso- and anteiso-CS, and
straight chain and iso-ClO fatty acid groups, with a total of nine different combinations of

substitution.

7]

/H

.21.
R31?) 0:?

R2

OH R2 ,R3 ,R4 = various combinations of branched
and linear C4, C5, and C10 fatty acids

Figure 3-1. Generalized structure of triacylglucose metabolites from S. pennellii

LA0716 as identified by Burke et at.I64

To simplify discussion of acylsugars, we propose the following nomenclature to
describe them. Acylsugars are assigned a letter (G or S) corresponding to glucose or
sucrose core, followed by two numbers. The first number indicates the number of fatty
acyl groups attached to the core, and the second number describes the total number of
carbon atoms in the fatty acyl chains. In some cases, the numbers of carbons in the
individual fatty acids are included in parentheses. The report of Burke’s laboratory
described glucose triesters of four molecular masses that we abbreviate as follows: G
3:12 (4,4,4), G 3:13 (4,4,5), G 3:18 (4,4,10) and G 3:19 (4,5,10).

Multiplexed CID coupled with LC/TOF MS is applied for structure annotation of
acylsugars in S. pennellii LA0716 to compare LC/MS results with Burke’s original

findings. Acylsugars lack readily ionized basic or acidic functional groups. As a

consequence, various adducts with cations ([M+Na]+, [M+NH4]+) or anions ([M+Cl]_,
[M+HCOO]'), are formed in ESI under gentle CID conditions. The dominant peaks in
the total ion chromatogram of an extract of S. pennellii leaf tissue displayed

[M+formate]— ions for a series of glucose esters ranging from G 3:13 to G 3:23. Extracted

ion chromatograms for these m/z values are presented in Figure 3-2. The acylglucoses

72

were resolved at different retention times using a reversed phase C13 column, with

retention time increasing with the number of fatty acyl carbon atoms. Findings from the
current work document 23 chromatographic peaks corresponding to formate adduct of
triacylglucoses, and these ranges across 11 different molecular masses. This LC/MS
approach has more than doubled the known diversity of acylglucoses in S. pennellii

LA0716.

73

Figure 3-2. Extracted ion chromatograms (XICs) generated from an extract of leaf tissue
from S. pennellii LA0716 of ions of m/z values corresponding to [M+HCOO]' for
glucose triesters with total fatty acid carbon atoms ranging from 12 to 22. A: G 3:12 (m/z
435). B: G 3:13 (m/z 449). C: G 3:14 (m/z 463). D: G 3:15 (m/z 477). E: G 3:16 (m/z 491).
F: G 3:17 (m/z 505). G: G 3:18 (m/z 519). H: G 3:19 (m/z 533). I: G 3:20 (m/z 547). J: G
3:21 (m/z 561). K: G 3:22 (m/z 575). A-K (from bottom to top).

74

33.18

 

 

 

 

 

 

 

 

 

 

 

10
on 32.77 K
0 1'0 2‘0 30 ' i ' 40
10 32.28
% 31.79 J
0 . . . 3 . .
10 20 30 40
10 31.16
29.83
10 20 30 40
10 28.50 H
%
0 . . . .
10 20 30 40
27.11
10 26.62 G
%
0 , . . .
1o 20 30 40
10 25.08 F
% 25.64
0 , , , ,
10 20 3o 40
10 22.91 2340 E
% ll
0 . 4 . . . . .
1o 30 40
100i
%
O , J‘ , . .
1o 20 30 40
17.46
10 C
% ll
10 20 30 40
10 15.09
% ll
0 , . . . ,
1o 20 30 40
13.14
10
% ll
0 . . . , . - 4 .
10 20 30 40

75

. Time

C03

"4171'

j i

Glucose triester G 3:18(4,4,10) from S. pennellii LA0716 was used to illustrate how
multiplexed CID spectra were used for metabolite annotation and identification.
Assignment of ions as molecular ions (both adduct and deprotonated forms), non-
covalent dimers, and fragment ions is aided by evaluating slopes of breakdown curves
generated under multiplexed CID (Figure 2-6. G 3:18). At the lowest CID potential (10
V), [M+HCOO]- at m/z 519.29 is the most abundant ion observed, with minor amounts
of [M—H]_ at m/z 473.28 and non-covalent dimer ions corresponding to [2M+HCOO]_
and [2M—H]_ at m/z 993.59 and 947.59 respectively. As CID potential was increased, the
abundance of formate adduct ions decreased, and relative abundances of [M-Hj' and
several fragment ions increased. At the highest CID potential of 80 V, the fatty acid
anions corresponding to deprotonated C10 (m/z 171.14) and C4 (m/z 87.04) fatty acids
dominated the spectrum. Between these extremes of CID potential (25 V), fragment ions
at m/z 347.13, 301.13 and 213.08 reached maximum relative abundance. These fragment
ions were assigned as products formed by elimination of neutral C10, C10 plus formic,
and C10 plus formic plus C4 acids, respectively, from [M+HCOO]" of G 3:18 (Figure 3-
4 top). It is worthy of note that the formate group remains attached after fatty acid
elimination in the fragment at m/z 347. A fragment ion at m/z 143.04 was assigned as
corresponding to losses of two acyl groups as fatty acids and elimination of the third as a
fatty ketene. This fragment (m/z 143.04) was common to all observed triacylglucoses,
and provides a characteristic marker for recognition of triacylglucoses.

High accuracy mass measurements of all observed ions accelerate spectrum

interpretation and help confirm associations of fragments with molecular ions. For

example, the fragment ion at m/z 143.04 (dehydrated glucose fragment C6H7O4') is

76

readily distinguished by the TOF mass analyzer from the C8 fatty acid anion (C3H15Of;

theoretical m/z 143.11). Therefore, the masses of fatty acid constituents of each acylsugar
can be determined by multiplexed CID in negative mode based on accurate mass
measurements of deprotonated fatty acid fragment ions.

In order to confirm assignments of fragments from multiplexed CID, product ion
MS/MS spectra were generated for the formate adduct (m/z 519) of triacylglucose G 3: l 8
using a QTof mass spectrometer (Figure 3-5). Product ions corresponding to neutral
losses of C10, C10 plus formic acid, and C10, formic acid plus C4 fatty acids were
observed at m/z 347.14, m/z 301.14, and m/z 213.09 respectively, remnants of the glucose
core appeared at m/z 143.04 and m/z 125.03, and C10 and C4 fatty anions were observed
at m/z 171.14 and m/z 87.04. All of these fragment ions were similar as observed in
multiplexed CID spectra using 25 V as described above. These findings demonstrate that
multiplexed CID experiments generate information in fragment ion masses similar to
those generated during MS/MS experiments.

Formation of fatty acid anion fragments from formate adducts of triacylglucoses is
consistent with formation of the fatty acid anion fragments (RCOO_) occurring via
nucleophilic displacement of the fatty acid anion by formate at the carbon where the fatty
acid ester is attached on the carbohydrate backbonem‘182 Additional evidence in support
of this mechanism comes from the appearance of [M+formate—RCOOH]_ at m/z 347 in
the case of loss of the neutral C10 fatty acid. Additional fragment ions corresponding to
losses of neutral fatty acids from [M—H—RCOOHT or [M+HCOO_—RCOOH]_ are
consistent with cis elimination reactions.'8"'82 A proposed fragmentation for the formate

adduct of G 3:18 is presented in Figure 3-3.

77

Figure 3-3. Proposed fragmentation pathway of the formate adduct of triacylglucose
S3:18 extracted from S. pennellii LA0716 (RT = 21.11 min) using ESI negative mode
with CID voltage 25 V. Specific substitution positions of individual fatty acids remain
uncertain.

78

m/ZI171

_O\

O
k
A
A0
m/z 301 13

 

6
o
/<
m/z 473 28

oz't/O‘C‘H
0
j:
519 29

0
W2:

OH

HO\

 

m/z: 143.08

m/z:213.08

79

The structure assigned for triacylglucose G 3:18 from negative ion mass spectra was
further confirmed by generating multiplexed CID mass spectra in positive mode. The

peaks observed, using a CID potential of 10 V, at m/z 492.23 and m/z 497.19 correspond

to [M+NH4]+ and [M+Na]+. Figure 3-4 displays ESI spectra from a triacylglucose G 3:18

(RT = 27.11 min) under CID potential 25 V in negative (top) and 40 V in positive
(bottom) mode. In both positive and negative ion modes, fragment ions are observed that
correspond to neutral losses of fatty acid groups and to fragments characteristic of the
fatty acid moieties. In positive mode, acylium ions corresponding to C4 and C10 fatty
acids are observed at m/z 71 and 155 respectively using collision potential of 40 V, and

their relative abundance increased when the potential was raised to 80 V.

80

100

 

 

 

 

 

 

 

 

 

171.14
%4 213.08
1 473 28519.29
1 87.04 143.04 301.13 347.13 I l
0 .4 . 1-4+ W1 .fiwnmmq—Lwrewwww .- 4. 7.4+. 4. - r -, . . were
100 285.12 457.21
1 127.13
%1 71-03 5L”? 07 ‘
b10913 197-09 492.23
- 55.13 369.19 497 19
0‘ ' lvl lyil ll .1 Y l VVVVV TI -+~~vr~+v , l, . , , -, A, ll, .4- , . r . x m/z
1 00 200 300 400 500 600

Figure 3-4. Electrospray ionization mass spectra of G 3:18 (RT = 27.11 min) extracted
from S. pennellii LA0716 using an Aperture 1 CID potential of 25 V in negative (top)
and 40 V in positive (bottom) mode.

81

171

 

 

 

 

100. 473
,
g; 213
4 474 519
301 ,
172 l 347 495 520
87 143 ’ 214’ 259 l l l
Ofi.T.r..Irfi.fi . YT.‘ ........ 1.4%.;T744‘1l...14.wl7 ...... .L ...... v vvvvvvvvvvvv I‘%"T'm/Z

100 "150 200 250 300 350 400 450 500

Figure 3-5 Product ion MS/MS spectrum of m/z 519 ([M+HCOO]') from triacylglucose
G 3:18 from S. pennellii LA0716 using collision energy 30 eV.

Most fragment ions were derived from [M+NH4]+ instead of [M+Na]+, which was

demonstrated by generating product ion MS/MS spectra of [M+NH4]+ generated using

the QTOF instrument. The ammonium adduct of acylglucose G 3:18 undergoes facile

loss of NH3 and H20 to form [M+H-HZO]+ (m/z 457.21) at collision potential of 10 V.
Product ion spectra for m/z 492 [M+NH4]+ and m/z 457 [M+H-H20]+ were performed

using the QTOF (Figure 3-6). The [M+H-HzO]+ fragment was generated in source

without changing ion source parameters, and a product ion spectrum was generated. The
major product ions using 30 V collision energy in the QTOF were the same as those

generated using multiplexed CID at 40 V. All of the major product ions observed from all

three precursors ([M+NH4]+, and [M+H-HZO]+) corresponded to losses of the various

neutral fatty acids. None of the product ions could be attributed to cross-ring
fragmentation reactions that would have been useful for establishing positions of

substitution of the fatty acid groups.

82

1 285.12
[M+H-H20-C10—2C4]+ .
197,09 \ [M+H-H20]
%, 1 457.21

1
109.12 155.13 I
‘ .

 

 

l
l l [M+H-H20-C4]+
‘ 369.16 ‘
1 $127.13 I , l ‘ ‘
05 1 i l . l l
1 285.12
1 155.13 197.09 ‘ MS/MS457 ESl+
1
1 127.13 I 1 1 36916
137.1 I ' ‘
$110212! (7 7 1 l i
l 1
l 1 J j 1 l ‘ 1 457.21
All I l l l l ‘
O .
200 250 400 450 ”V2

Figure 3-6 Top: Product ion MS/MS spectrum of m/z 492 ([M+NH4]+) from
triacylglucose G 3:18 from S. pennellii LAO716 (collision energyz30 eV). Bottom:
product ion MS/MS spectrum for m/z 457 ([M+H—H20]+) for G 3:18; m/z 457 was

generated in source (Collision energy235 eV).

Isomeric acylsugar metabolites can be distinguished based on fragment ion masses
produced using more aggressive CID voltage. Triacylglucose G 3:20 from S. pennellii
LAO716 provides a good example, yielding at least four resolved chromatographic peaks
yielding abundant formate adduct ions (Figure 3-7). Fatty acyl constituents could be
determined from the anions of fatty acids cleaved from acylglucoses at CID potentials of
25 V and greater. The extracted ion chromatogram for the formate adduct of G 3:20
revealed four major isomers. The first eluting peak (peak 1) shows fragments for C5 and
C10 fatty anions, peak 2 shows C4, C5, and C11 fatty anions, and peaks 3 and 4 show C4

and C12 fatty anion peaks. In order to match the observed masses, peak 1 isomer must

83

contain two C5 groups, and peaks 3 and 4 must contain two C4 groups. Given the limited
number of observed peaks, we concluded that variation in substitution patterns for
specific fatty acids was limited. For peaks 3 and 4, additional GC/MS evidence suggested
the presence of both linear and branched C12 fatty acid groups. Therefore, we attribute
the two isomeric forms (C4, C4, C12) to arise from different C12 fatty acid groups, and
not to the presence of multiple sites of C12 substitution. All the ions were labeled in the

mass spectrum of each triacylglucose isomer in Figure 3-5.

84

Figure 3-7. (Top) Extracted ion chromatogram of m/z 547 (G 3:20) from LC/MS analysis
of an extract of S. pennellii LAO716 under CID potential 10 V in ES] negative mode.
Chromatographic peaks corresponding to four isomers detected as m/z 547 were resolved.

(Bottom) ESI mass spectra obtained from these four isomers at 40 V in negative mode.

85

1 00" ’l

 

 

 

o\°“1 ,‘1 1‘ . 1
1 {l \1 l *1 l 1‘
1 ‘1V1‘ i 1; l l
. 1 1‘ l 1 l
‘ I ll 1 l
. l “I 1v.“ l
044—4444444444. rrvfifi 144—444 - 1 . 1 1 1 1 .“‘r: :v~44ee.4.4.44444_.4_. Time
30.00 32.00 34.00 36.00 38.00
C12 fatty anion [M—H]
199.17 _ -
100104 [fattyHalnion [M- H— 0121' [M+HCOO'CIZI 4 501. 31 IM+HC°OI
% 143. 04 21308 30113 347_13 1 54731
04443101....444444444Wt ..444 ..444 .4444 . %H.—444444445 414.. W411, .444
C12fatty anion
19917 [M— H]
133104140 anion 1213081” H 0121 [M+HCOO- C12] 3 50131IM+H9001
o 87 04 301 13 547. 31
0144444414444 15543499440441. 444,444.44 ...—444....-.341134444444433. 4..-,532Ll...4,44444444 -...
c11 fa ' [Ni—H]—
100, “y amo" [M+HCOO—C11]’ 2 501-31 [M+HCOO]"
1 C5fattyanion185.16 M~ H- -C11 1 1
° 227.09 I 1 547.3
61870410106143.04l 31.514 351-14 1
0’“ Wm 4444.444 4L4.» 4L“- 4444444 .4. L. 44444444444 44. .4 44.44. . 4.. (44444244st-.4 “3%.
C10 fatty anion “II—Hg-
100 171.14 [M- H- -C10 0] 15 _ 31 _
'1 C5 fatty anion l 227 09 IM+HCOO C10] [M+HCOO]
0.1 101 061 1 .' 329.15 375.15 1 547.31
014 . . 430%. . _ .L 4. 1 4 . 1 . . . . l . 1.33-.. m
100 200 300 400 500

86

All detectable acylglucoses from S. pennellii LA0716 were annotated from
multiplexed CID mass spectra and product ion MS/MS scans. Major fragment ions and
fatty acyl constituents determined using multiplexed CID for each detectable acylglucose
of LA0716 (S. pennellii) are presented in Table 3-1. Accurate masses for formate adducts
for each acylglucose were determined by flow injection analysis, and are listed in Table
3-2. One original finding of this work is discovery of C8, C9, C11, and C12 fatty ester
groups on triacylglucoses from this plant genotype. The C8 and C9 fatty acids were not
observed in extracts from several other accessions within the genus Solanum (S.
habrochaites LA1777 and LA1353), and we conclude that their occurrence in accession
LA0716 may help guide future discoveries of genes involved in branched fatty acid

elongation.

87

Table. 3-1. Fragments and fatty acid constituents for detected glucose triesters from S.

pennellii LA0716 in negative and positive mode (CID potential, 25 V)

 

 

 

 

 

Acylsugar metabolite Fatty acid [M+formate|' lM—HI' Fragment ions
annotation ester groups (m/z) (m/z) (m/z)
G 3:12 C4, C4, C4 435 389 347, 301,213,143, 125, 87
03:13 C4,C4,C5 449 403 361,315, 227, 143, 125, 101,87
G 3:14 C4, C5, C5 463 417 375, 329, 227, 143, 125, 101, 87
G 3:15 C5, C5, C5 477 431 375, 329, 227, 143, 125, 101
G 3:16 C4, C4, C8 491 445 347, 301, 213, 125, 143, 87
3:88
G3:18 C4,C4,C10 519 473 347,301,213, 143, 125, 171,87
G3:19 C4,C5,C10 533 487 361,315, 227, 143, 171, 101,87
C5,C5,C10 347,301,213,143,125,171,101
G3:20 C4,C4,Cll 547 501 361,315,213, 143,125, 185,101,87
C4, C5, C12 375, 329, 227, 143,125, 199, 87
G3:21 C4,C5,C12 561 515 361,315, 227, 143, 199, 101,87
G 3:22 C5, C5, C12 575 529 375, 329, 227, 143, 199, 101, 101
Acylsugar Fatty acid [M+NH4|+ |M+Na|+ Fragments
metabolite annotation constituents (m/z) (m/z) (m/z)
G 3:12 C4, C4, C4 408 413 373, 285, 197, 127, 109
G 3:13 C4, C4, C5 422 427 387, 299, 197, 127, 109
G3:14 C4, C5, C5 436 441 401,299,211, 127
G3:15 C5, C5, C5 450 455 415,313,211, 127
G 3: 16 C4, C4, C8 464 469 429, 285, I97, 127
G 3:17 C4, C5, C8 478 483 443, 299, 197, 127
G 3:18 C4, C4, C10 492 497 457, 285, 197, 127
G3:19 C4,C5,ClO 506 511 471,299,211, 127
C5,C5,C10 485,313,211,127
G 3:20 C4, C4, C12 520 525 485, 285, 211, 127
C4, C5, C11 485, 299, 211, 109
G 3:21 C4, C5, C12 534 539 499, 299, 211, 127
G322 C5,C5,C12 548 553 513,313,211, 127

 

88

Table 3-2. Exact mass measurements of [M+HCOO]' ions of detected glucose triesters
from S. pennellii LA0716 detected using ESI negative mode with CID potential (10 V).

 

Acyl sugar

 

2:32:22: Formuilzgchprmate M2232?“ Calculated (m/z) ppm (error)
G 3:12 C19H31011 435.1866 435.1865 -0.2
G 3:13 CZOH33011 449.2022 449.2030 -0.2
G 3:14 C21H35011 463.2179 463.2169 -2.2
G 3: 1 5 022H37011 477.2336 477.2327 -1.9
G 3:16 CZ3H39011 491.2492 491.2515 4.7
G 3:17 0241141011 505.2615 505.2649 -6.7
G 3:18 025H43O11 519.2805 519.2830 4.8
G 3:19 C26H45011 533.2962 533.2964 0.4
G 3:20 C27H47O11 547.3118 547.3122 0.7
G 3:21 028H49011 561.3275 561.3262 -2.3
G 3:22 029H51011 575.3431 575.3434 0.5

 

Despite the successes in annotating acylsugar metabolites based on the number of
acyl chains and the number of carbons in the fatty acid groups, the CID spectra did not
distinguish branched from straight chain of fatty acyl groups or their positions of
attachment. This shortcoming arises because fragment ions that would distinguish fatty
acid isomers were not observed. GC/MS was performed to distinguish isomeric fatty
acids after transesterification from all the acylsugars in these wild type species. The
GC/MS results for the composite composition of fatty acyl groups in leaf extracts of S.
pennellii LA0716 are summarized in Figure 3-8. All fatty acid chain lengths observed in
LC/MS results were confirmed by GC/MS data, and isomeric fatty acids were
distinguished by their retention times and electron ionization mass spectra. The proposed
structures for branched or straight chain fatty acyl groups were based on fatty acid esters
identified from GC/MS results by comparing retention times and mass spectra with

authentic standards or the NIST 05 mass spectrum library and support from literatures.l '2

89

 

40
35 ~
30
25
20

1

4 l_ 1

104

% to the total
31

 

   

|C4
aICS
IC5
aiC6
1C8
C
1C9
1010
C12
C12

nC1O
iC11
nC11

Fatty acids

 

 

 

Figure 3-8 Fatty acyl groups from total acylsugars in an extract of S. pennellii LA0716
were transesterified to form fatty acid ethyl esters, and analyzed using GC/MS. Data for
each fatty acid were calculated as a percentage of the total ion current chromatogram
peak areas of fatty acids detected through GC/MS. The abbreviations “i" and “ai” refer to

130- and anteiso- branched isomers. Error bars indicate standard error with n = 3.

3.3.2. Structure annotation for glucose triesters and sucrose triesters in S. pennellii
LA1522.

After the performance of multiplexed CID was verified by detecting known
triacylglucoses and previously undetected acylsugars in extracts of S. pennellii LA0716,
this approach was applied to profile acylsugar metabolites in a different accession of the
same species, LA1522. No previous reports have documented acylsugars in LA1522
accession. The LC/MS results revealed triacylglucoses ranging from G 3:12 to G 3:24 in

leaf trichome extracts from this accession. All the XICs for formate adducts of the

90

acylglucoses are displayed in Figure 3-9. Findings from the current work document 21
chromatographic peaks corresponding to formate adduct of triacylglucoses in LA1522,

and these are distributed across 12 different molecular masses.

91

Figure 3-9. Extracted ion chromatograms (XICs) generated from an extract of leaf tissue
from S. pennellii LA1522 for ions corresponding to [M+HCOO]_ for glucose triesters
with total fatty acid carbon atoms ranging from 12 to 22. A: G 3:12 (m/z 435). B: G 3:13
(m/z 449). C: G 3:14 (m/z 463). D: G 3:15 (m/z 477). E: G 3:16 (m/z 491). F: G 3:17 (m/z
505). G: G 3:18 (m/z 519). H: G 3:19 (m/z 533). I: G 3:20 (m/z 547). J: G 3:21(m/z 561).
K: G 3:22 (m/z 575). L: G 3:23 (m/z 589).

92

 

 

 

 

 

 

 

 

 

 

 

 

 

G 10 20 30 40
,0 37.87 K
95
G 10 T 20 if 30 IF 7* ' 40
36.27
‘° 119
0
{3’ 2 . .4 . ..4, , . ,
10 20 30 40
35.15
10 I
95 A
0 F10 i 20 5‘ Fl 30 ' ’ '5 40
33.96
10 p.
95
0‘ 10 20 - 30 ' ' 40
10 32.07G
95
c . 4 . .
10 20 30 40
10 29.00 0-12F
95
G 10 20 T ' 30 ";7° '7 40
10 27.11 E
<5
0 4 . - . . 4 .
10 30 40
10 25.16 D
1
o L . 4 . 4 . .
10 20 30 40
,0 22.71 C
95
o . 4 . . b . 4 .
10 20 30 40
10 20.19
1 B
0 4 . . . . . ,
10 20 30 40
10 17.89
95 .l [K
o . . . . . ,
10 20 30 40

93

Although the patterns of acylsugars were similar with LA0716, the relative abundances
of fatty acyl groups differed between accessions, as did fatty acyl substitution patterns.
For example, triacylglucose G 3:19 (m/z 533) in LA0716 was substituted with C4, C5,
and C10 fatty acyl groups in two isomers believed to differ only in substitution of a
straight chain C10 ester by an iso-ClO ester. In contrast, for LA1522, the fatty acyl
constituents were C5, C5, and C9 in all three observed chromatographic peaks. It is
postulated that the isomers in LA1522 may arise from different isomeric C5 fatty acids
(both iso-CS and anteiso-CS fatty acyl groups). The glucose triesters were assigned based
on accurate mass measurement and fragments, as was performed before for LA0716. All

the detected glucose triesters are described in table 3-3.

Table 3-3. Fragment'ions and fatty acid constituents of detected glucose triesters in an
extract of S. pennellii LA1522 leaf trichomes using negative and positive mode

electrospray ionization (CID potential, 25 V)

 

 

Acylsugar . _ -
metabolite :13”:de IM+(}:§())OI “:31“ Fragment ions (m/z)
annotation g up Z z
G 3:12 C4, C4, C4 435 339 347, 301, 213,143, 125,87
G 3:13 C4, C4, C5 449 403 361, 315,227, 143, 125, 101, 87
G 3:14 C4, C5, C5 463 417 375,329,227, 143, 125, 101, 87
G 3:15 C5, C5, C5 477 431 375,329,227, 143, 125, 101
G 3:16 C5, C5, C6 491 445 389, 343, 241, 227,143, 115, 101
, C4,C5, C8 361, 315,227, 125, 143, 101, 37
G 3'” C5, C6, C6 505 459 339, 343,241, 227, 143,115, 101
, C4,C4, ClO 347,301,213, 143, 125, 171,87
0 3'18 C5, C5, C3 5'9 473 375,329,227, 143. 125, 101
63:19 C5, C5, C9 533 487 361,315,227, 143,171, 101,87
, C5,C5,C10 347,301,213, 143,125,171, 101
G 3'20 C4, C5, C12 547 50' 375,329,227, 143,125, 199,37
G 3:21 C5, C6, C10 561 515 361,315, 227, 143, 199, 101, 87
G 3:22 C5, C5, C12 575 529 375,329,227, 143, 199, 101, 101
G 3:23 C5, C6, C12 589 543 389, 343, 241, 227, 199, 143. 115, 101

 

94

In addition to glucose triesters, sucrose triesters were also detected in LA1522 at
levels similar to triacylglucoses. There were 15 chromatographic peaks corresponding to
formate adduct of triacylsucroses in LA1522, and these range across 10 different
molecular masses. All XICs of these sucrose triesters are shown in Figure 3-10. One
striking difference between the triacylglucose and triacylsucrose chromatograms is
evident. Specifically, fewer isomeric peaks are observed in the triacylsucroses than
triacylglucoses. Two potential explanations come to mind. Either the LC separation fails
to resolve some isomeric triacylsucroses, or the substitution patterns differ between
acylglucoses and acylsucroses. This question remains unresolved, pending purification
and NMR characterization of more acylsugar metabolites by other researchers in the

Jones laboratory.

95

Figure 3-10. Extracted ion chromatograms (XICs) generated from an extract of leaf
tissue from S. pennellii LA1522 of ions of m/z values corresponding to [M+HCOO]‘ for
sucrose triesters with total fatty acid carbon atoms ranging from 12 to 23. A: S 3:15 (m/z
639). B: S 3:16 (m/z 653). C: S 3:17 (m/z 667). D: S 3:18 (m/z 681). E: S 3:19 (m/z 695).
F: S 3:20 (m/z 709). G: S 3:21 (m/z 723). H: S 3:22 (m/z 737). I: S 3:23 (m/z 751). J: S
3:24 (m/z 765).

96

 

40

3'0

20

10

 

4'0

3'0

2'0

1'0

 

40

3'0

20

1'0

 

40

3'0

2'0

1'0

 

40

3'0

2'0

1'0

10

 

O

40

3'0

2'0

1'0

10

30

2'0

1'0

 

021
100

0

40

3'0

2'0

1'0

 

 

%.

 

 

100

%‘1

4'0

3'0

2'0

1'0

100

 

%

40

3'0

2'0

1'0

97

The process of annotating acylsucrose metabolites is illustrated using S 3:22 from
LA1522 as an example. This metabolite was chosen because only one isomer was
apparent from LC/MS results. Based on the slope of breakdown curves generated using
multiplexed CID, the ions were assigned as adducts and fragments (Figure 3-11). At the
lowest CID potential (10 V), [M+HCOO]‘ was the most abundant ion observed,
accompanied by minor amounts of [M+Cl]—. At a setting of 25 V, [M-H]' at m/z 691.39
appeared, but its abundance decreased at higher CID voltages. Yet higher CID potential
yielded more extensive fragmentation, most notably the fatty acid anions at m/z 101.06
(C5) and 199.17 (C12). Other major fragments at m/z 607.33 (loss of C5), m/z 509.22
(loss of C12), m/z 425.17 (losses of C5 and C12) and m/z 341.11 (losses of two C5 and
C12) are attributed to neutral losses of each of the three fatty acid groups as fatty ketenes.
It is worthy of emphasis that eliminations of fatty ketenes from triacylsucroses stand as
distinctly different from eliminations of neutral fatty acids from triacylglucoses. This

point will be discussed in more detail below.

98

80

 

604

/.

40~

% of ion current
I

20 ~ » ~
/I/ /‘

,«/

 

 

 

0 E J I t I t

5 25 45 65 85
Aperture 1 voltage

 

 

Figure 3-11. Breakdown curves derived from LC/negative ion ESI MS analyses of a leaf
dip extract from S. pennellii LA1522 showing collision energy dependence of
abundances of: (A) formate ion adducts of three triacylsucrose metabolites ([M+HCOO]"
at m/z 737 83:22 (A); [M-H]— at m/z 691 (0); C12 fatty acid at m/z 199 (I); C5 fatty

acid anion (1]).

To establish relationships between precursor and fragment ions, a product ion MS/MS
spectrum was performed for the formate adduct of triacylsucrose S 3:22 from accession
LA1522. Figure 3-12 compares ESI spectrum of S 3:22 obtained using LC/MS and
multiplexed CID (top) to the product ion MS/MS spectrum (bottom) for [M+HCOO]-.
The two spectra share remarkable similarity, and the results suggest that all major

fragments could be explained by decomposition of formate adducts. These findings allow

99

Losses of fatty ketene are characteristic of negative mode CID spectra of
triacylsucroses. Their formation could be attributed to either a “charge-directed” process,
which involves attack of anion charge site on the hydrogen of the fatty acid constituents,
or a “charge-remote” mechanism occurring through cis-elimination reaction, as observed
in CID experiments performed on acyl glycerophospholipids.18" '82 Loss of neutral fatty
acid might arise from “charge-remote” mechanisms through cis- elimination reactions,
but those fragment ions only were observed to have relatively low abundance in this
work. The proposed fragmentation pathway for the formate adduct of S 3:22 is presented

in Figure 3-15.

C12 fa acid anion

   
 

 

 

 

 

 

 

 

 

99. 7
100*
1C5 fatty acid
amon - [M'l'Cll-
%: 10106 [M-H-C12-C5-C5]’ [MES-P12] 737%.» H
1 [M-H-C12-C5] [M—H-Cs] 13938
425.17 607. 3
\l/ 40716 - -1
1 l 323.10 lM-Hll
0 +4Y1Y1Y Y l+Y Y Y fYJY Y YllYY Y YJ-LY YYYYYY
425.16
100; TOF MS/MS 737 ESI-
. 509.22
%4
407.15 607.32
199.17
3231034111 l 691.39
100 200 300 400 500 600 700

Figure 3-12 (TOP) ESI spectrum of triacylsucrose S 3:22 (RT = 36.50 min) extracted
from S. pennellii LA1522 using CID potential of 55 V. (Bottom) Product ion MS/MS
spectrum for formate adduct of S 3:22 (products of m/z 737.39 using collision potential of
35 eV).

100

Positions of fatty acid substitution on acylsugars were not evident from negative
mode CID spectra, and structural information was limited to information about the
number of carbons in fatty acyl moieties. CID spectra generated using positive mode
provide complementary information about positions of fatty acyl groups. No [M+H]+ ions

were observed in positive mode ESI spectra of acylsugars, and the low CID potential

spectra were dominated by adducts [M+NPI4]+ and [M+Na]+. Figure 3-13 displays a

positive mode ESI spectrum of triacylsucrose S 3:22 from LA1522 leaf extracts using
CID potential of 40 V.
Product ion MS/MS spectra were generated for ammonium and sodium adducts of S

3:22 (Figure 3-14) in order to establish the source(s) of fragment ions observed using

multiplexed CID. The results suggested that most fragments under multiplexed CID in

positive mode were derived from [M+NH4]+ because the predominant products of

[M+Na]Jr were sodium-containing fragment ions not observed in the multiplexed CID
spectra. Collision induced dissociation of [M+Na]+ gave results similar to a report on
acylsucroses from oriental tobacco reported by Xu and coworkers,183 and a proposed
fiagmentation pathway of S 3:22 [M+Na]+ is displayed in Figure 3-16. These findings are
consistent with greater activation barriers to fragmentation of [M+Na]+ owing to limited
proton mobility.

Information from positive mode CID spectra of triacylsucroses complements

assignments of fatty acyl composition derived from negative ion CID spectra. The most

obvious example of this comes from the dominant loss of NH3+1 80 (fructofuranose unit)

through “charge-directed” cleavage of the glycosidic bond to produce a fragment ion at

m/z 513.34. This fragment consists of the anhydroglucose core ion with all three fatty

101

acyl groups attached. The importance of this fragmentation reaction is that the fragment
mass serves to indicate which sugar groups are substituted by specific fatty ester groups.
This information is not evident in negative mode CID spectra. For all of the

triacylsucroses from S. pennellii LA1522, all fatty acid groups are attached on the

glucopyranose ring. Nearly all remaining products from [M+NH4]+ can be attributed to

neutral losses of fatty acids or fatty ketenes. Proposed fragment pattern of S 3:22
[M+NH4]+ is shown in Figure 3-17.

A summary of the information from positive and negative mode CID spectra of
triacylsucrose S 3:22 from LA1522 suggest C5,. C5 and C12 fatty acyl groups, all of
which are attached on the glucopyranose ring. As described in subsequent chapters, this
substitution pattern is less common among Solanum accessions than substitution divided
between the two rings. A compilation of fragment ions and fatty acyl constituents for
acylsucroses in accession LA1522 was prepared from multiplexed CID spectra, and is
assembled in Table 3-4. Accurate masses for sucrose esters are displayed in Table 3-5. In
contrast to its close relative LA0716, the accession LA1522 contained both
triacylglucoses and triacylsucroses, with the latter having all three fatty acyl groups on
the glucopyranose ring. This discovery highlights substantial biochemical differences
between two accessions of the same plant species, and points to remarkable chemical
diversity within the genus Solanum. Of equal importance, the observed acylsugars exhibit
some specificity in substitution patterns, and are not random combinations of attachments
of fatty acids to sugars. The selectivity of substitution reflects enzyme selectivity in
biosynthesis. The chemical diversity of trichome metabolites is explored in more detail

across a broader range of genetic variants in subsequent chapters.

102

[M+H-FRU]+

 

 

 

 

. 513.34

( [M+H-FRU-C12]"

l +
, [M+H-FRU-C12-CS]+ 710.43
/0‘

1 211.10 +

4 C10fa ac Iium [M+H'FRU'C5I ,

tty183,16 299.22 411-28 [MtNH4'FRU1 715.39
ol 1.27125YJYL. . .Lmfi. is . xi. .il§8:37.lY Y Y ..m/z
100 200 300 400 500 600 700

Figure 3-13 Positive mode ESI spectrum of triacylsucrose S 3:22 (RT = 36.50 min)
extracted from S. pennellii LA1522 using CID potential of 55 V. The abbreviation
‘FRU’ refers to a fragment derived from the fructofuranose ring.

[M+H-FRU-C12]+ TOF MSIMS 710 ESI+
313

100 . ,
, [M+H-FRU]
[M+H-FRU-2C5] 513
1

 

 

 

 

 

 

 

 

[M+H-FRU-C12-C5]+
%] ‘V M+H FRU C5+
' C10fatty acylium 183 211 I 411 l
0 . YYfi. . eigehfi Y Y: WY Y YL. Y h
TOF msms 715 ESI-I-
553
10° 1 [M+Na-FRU]’
[M+H-FRU-C12-ZC5]* [M+Na-FRU-012]"
%j \fM+Na-FRU-C12-C5]+ 3‘53
. 1A [M+Na-FRU—CS]+
L 1E7 1671?5 229 251 i 451
04 Y Y Y Y W Jrfi #42 Yl .1 Y 1 Y Y: Y Y Y I Y Y . Y Y Y m/Z
100 200 300 400 500 600

Figure 3-14 (Top) Product ion MS/MS spectra of (Top) [M+NH4]+ at m/z 710.43

(collision energy: 45 eV) and (Bottom) [M+Na]+ at m/z 715.39 (collision energy: 55 eV)
for S 3:22 extracted from S. pennellii LA1522. The abbreviation ‘FRU’ refers to a

fragment derived from the fructofuranose ring.

103

Figure 3-15 Proposed fragment pathway of formate adduct of S 3:22 (RT = 36.50 min)
extracted from S. pennellii LA1522 using ESI negative mode with CID voltage 55 V.

Specific substitution positions of individual fatty acids remain uncertain.

104

b 14an OH
:H_OC QA HC§

O
9 1.
:m/zg7; 40 m/ZZH:07.331\ Hobos-(Edi

/—HCOOH b9 0”
€330 0“ “film/z: 407.16
6 —-0 EHKOM)
Haxg .

H OH
O H;C'O O C? 0 OH
OH \\C,403 O
o‘zjdoo: , ,_ . OH

6 H.0H QO‘OH OHO
6a OH

C=O
Lu H<Siam/z: 425.17
HO
m/z: 691.39 la M: 50922 CO $011
0
_ OH OH

\ -

b0 / O . OH H23: 3:300” sz323.10
(5 OH Ho-
(3&0

9 ‘OH OH
m/z: 341 .11

W2: 589.32

H’ ' ‘ '
K< 0
W2: 737.40 ('3

m/z:199.17
OH OH
9,9 @0451:
030 (30 O

m/z:101.06

105

b
W23715-39 m/z:533.34 \ ”V1451 27

OH +
0 Na
06’0éé/L0”
OH 1
OH OH
0 ”8+ \ / S\
\

<— O
/ OH Vb=0

b’ H4
I \ a ’

Ho

m/z:167.03 m/z:251.09 fir/21353.16

Figure 3—16 Proposed fragment pathway for sodium adduct of S 3:21 extracted from S.
pennellii LA1522 based on MS/MS product ion spectrum. Specific substitution positions

of individual fatty acids remain uncertain

106

m/z: 710.43 m/z: 513.34

H OH 1)
0‘ 3) \
b’ __ 81H 32H
_a__, 9 </ > < >
C/ko \ _\ / \
1 \55 9511(‘0. O
Woj: L2
H / L
m/z:411.28 </ C1>
_ m/z:211.10 m/z:313.17

Ho'
m/z:127.04

NH3 H H
OH H<CH °H+ o‘\ \
OH (to
O ( O 0‘ 0-6 CNN H %H ‘
°6'°' o 6' 6 H ‘0
OH ' ‘0 I \ a ‘C,O
1 1 O £C\O
O (I) OH | =0¢=0j§i
‘C=OC=O S\\\<O
m/z: 693. 41

Figure 3-17 Proposed fragment pathway of ammonium adduct for S 3:22 (RT = 36.50
min) extracted from S. pennellii LA1522 using ESI positive mode with CID voltage 40 V.

Specific substitution positions of individual fatty acids remain uncertain.

107

Table.3-4. Fragments and fatty acid constituents for detected sucrose triesters in S.
pennellii LA1522 using ESI negative (CID potential, 55 V) and positive mode (CID
potential, 40 V)

 

 

 

 

 

Acylsugar metabolite |M+HCOO]' IM+35CW Fragment ions
and Fatty acid (m/z) (In/z)
constituents ("i/Z)
S 3: l 5 C5,C5,C5 639 629 593,425, 407, 341, 323, 179,101
S 3: l6 C5,C5,C6 653 643 607, 523, 425, 341, 323, 115, 101
S 3:18 C5,C5,C8 681 671 635, 551. 425, 407, 341, 179, 143, 101
S 3:19 C5,C5,C9 695 685 649, 565, 509, 425, 341, 323, 157, 101
S 3:20 C5,C5,C10 709 699 663, 579, 509, 425, 407, 341, 323, 171, 101
S 3:21 C5,C6,C10 723 713 677, 593, 579,509, 425, 407, 341, 323.171, 115, 101
S 3:22 C5,C5,C12 737 727 691, 607, 509, 425, 407, 341, 323, 199, 179, 101
S 3:23 C5,C6,C12 751 741 705, 621, 607, 523, 439, 421, 341, 323, 199, 115,101
S 3:24 C6,C6,C12 765 755 719, 537, 439, 341, 323, 199, 115
Acylsugar metabolite and lM+NH4I+ |M+Na|+ Fragment ions
Fatty acid constituents (In/z) (m/z) (m/z)
S3:15 C5,C5,C5 612 617 455,415,313,21|,127
S 3:16 C5,C5,C6 626 631 469, 429, 313, 211, 127
S 3:18 C5,C5, C8 654 659 497, 457, 355, 313, 211, 127
83:19 C5,C5,C9 668 673 471,313,211, 127
S 3:20 C5,C5,C10 682 687 485, 313, 211, 127
S 3:21 C5,C6,C10 696 701 499, 327, 225, 127
83:22 C5,C5,C12 710 715 513,411,313, 309,211, 127
S 3:23 C5,C6,C12 724 729 527, 327, 211, 127
S 3:24 C6,C6,C12 738 743 541,341, 225, 127

 

108

Table 3-5. Exact mass measurement of [M+HCOO]_ Ions for detetable sucrose triesters

in accession LA1522 detected using ESI negative mode under CID potential 10 V.

 

 

1.3.3.2333. 31111283739999 ... (......)
s 3:15 CZBH47O16 639.2885 639.2864 3.3
s 3:16 029H49016 653.3036 653.3021 2.3
s 3:17 0301151016 667.3150 667.3177 40
s 3:18 6311453016 681.3334 681.3336 0.3
s 3:19 (3321155016 695.3505 695.3490 2.2
s 3:20 c33H55016 709.3619 709.3647 -3.9
s 3:21 0341157016 723.3303 723.3322 2.6
s 3:22 0351459016 737.3960 737.3972 1.6
s 3:23 C36H61O16 751.4116 751.4102 -1.9

 

As was the case for S. pennellii LA0716, the branched or straight chain fatty acyl
groups for acylsugars from accession LA1522 could not be distinguished from CID
spectra. Therefore, GC/MS was performed to identify fatty acids after transesterification
from all the acylsugars in these wild type species. The GC/MS results for the composite
composition of fatty acyl groups in leaf extracts of S. pennellii LA1522 are summarized
in Figure 3-18. All fatty acid chain lengths observed in LC/MS results were supported by
GC/MS data. The identities of fatty acyl groups were based comparisons of EI mass
spectra from GC/MS analyses with retention times and mass spectra of authentic
standards. When standards were not available, the fatty acid ethyl esters were identified
using comparisons to the NIST 05 mass spectrum library. GC/MS results revealed that in
LA1522, the iC5 was most abundant (~40% of peak area), with about 15% each of aiC5,
nC10, and nC12. In constrast, iC5 made up less than 5% of total TIC area for LA0716,
for which iC4 accounted for nearly 40% of the peak area. The aiC5 and iC10 esters made

up about 15% and 20% respectively, and nC10 was the next most abundant chain (~

109

10%). It is also worthy to note the inverse ratio of anteiso to iso-CS isomers in LA0716
relative to LA1522. The iso- and anteiso-CS are believed to be derived from different
amino acid precursors, leucine and isoleucine respectively. Whether the shift in C5
isomer abundances results from different relative sizes of the leucine and isoleucine pools
or from different pathway efficiencies cannot be determined from available evidence.
Likewise, the iso-C4 fatty acid is believed to be derived from valine, but reasons for the
high abundance of iC4 esters in LA0716 cannot yet be established with certainty. It is
recommended that future work address these questions, perhaps through use of metabolic

flux measurements using stable isotope labeled tracers.

110

 

% to the total

 

 

Fatty acids

 

Figure 3-18 Fatty acyl groups from total acylsugars in an extract of S. pennellii LA1522
were transesterified to form fatty acid ethyl esters, and analde using GC/MS. Data for
each fatty acid were calculated as a percentage of the total ion current chromatogram
peak areas of fatty acids detected through GC/MS. The abbreviations “i" and “a1" refer to

160- and anteiso- branched isomers. Error bars indicate standard error with n = 3.

Generation of fragment ions from triacylsucroses required higher collision potentials
than for analogous triacylglucoses, as was evident from inspection of plots of various ion
abundances as a function of collision potential (breakdown curves) described in Chapter
2 (Figures 2-5 and 2-6). For acylglucoses, [M+formate]’ abundances drop sharply as
Collision potential is increased above 10 V, while for acylsucroses, the drop occurs at

collision potentials greater than 25 V. While this behavior can be attributed in part to

111

 

lower center-of-mass collision energies for the more massive triacylsucroses, a
comparison of the behavior of the formate adduct of the lowest mass triacylsucrose
formate (m/z 625) shows striking differences relative to the most massive triacylglucose
(m/z 589), with the latter forming greater yields of CID products at lower collision
potentials. Based on these observations, we judge the activation energy for fragmentation
of triacylglucoses to be lower than triacylsucroses based on the threshold CID potential
needed to observe fragment ions.

Differences in energy thresholds for fragmentation can be exploited using
multiplexed CID to distinguish acylglucoses and acylsucroses when they coelute, as is
often the case when analyzing plant extracts with diverse composition of acylsugar
metabolites. Resolution of fragments from coeluting acylsugars is illustrated in Figure 3-
19, using coeluting triacylglucose G 3:14 (4,5,5) and triacylsucrose S 3:15 (5,5,5). At the
lowest collision potential (10 V) the spectrum displays multiple ions from. a
triacylsucrose (83:15) including formate (m/z 639) and chloride adducts (m/z 629 and
627), and a formate-bound dimer ion (m/z 1223). Of lower abundance is [M-H]- of
triacylglucose G 3:14 (m/z 463). Raising the collision potential to 25 V (middle panel)
leads to extensive fragmentation of the triacylglucose (ions marked with *), but no
fragments (e.g. [M-HT) derived from CID of the triacylsucrose are observed at this
potential. Fatty acyl anions are observed at m/z 87 and 101 corresponding to C4 and C5
fatty acyl groups from G 3:14. In the bottom panel, the highest collision voltage of 80 V
yields fragmentation of the triacylsucrose as evident from appearance of [M-H]_ from
83:15 (m/z 593) and neutral losses of fatty ketenes or fatty acids. This feature aided

profiling of metabolites of accession LA1522 which produces complex mixtures of

112

triesters of both glucose and sucrose. This multiplexed CID approach offers great

potential to resolve coeluting compounds based on different fragmentation energy

 

 

thresholds.
&
100 629
8‘ &
&425 &
o 3 , 509
/° 1801 3234071 l 593
0_ L L111 ...li i t Ll U.
8
100« .1 639
% * 227* * 4:7 * &
170121313153611453 1233
0 . ...r 11 r. l. r .. l 1 1
&
100 63.9
% * 1:33
0 463 ,

 

Figure 3-19. Multiplexed CID mass spectra (electrospray, negative mode) of coeluting
metabolites from LC/TOF MS analysis of an extract of S. pennellii LA1522 showing
spectra obtained from three of the five Aperture 1 potentials. Top panel: 10 V; middle
panel: 25 V; bottom panel: 80V. m/z of formate adduct ions at m/z 639 is sucrose triester

S 3:15 (5,5,5) and m/z 463 is glucose triester G 3:14 (4,5,5) at RT = 22.8 min.

3.3.3. Structure annatation of sucrose tetraesters in S. habrochaites LA1777.
Although the acylsucroses described in S. pennellii LA1522 were all triesters, other

Solanum accessions Exhibited metabolites of the same nominal mass but earlier

113

chromatographic elution times. Closer examination of the m/z values of these metabolites
suggested elemental formulas more consistent with sucrose tetraesters. Sucrose tetraesters
have the same nominal mass as sucrose triesters but with one less carbon atom in the fatty
acyl chains. Accurate mass measurements allowed us to distinguish them by mass
difference, as the tetraesters are lower in mass than the isobaric triester by 0.036 Da. In
addition, sucrose tetraesters can be distinguished from triesters based on different
fragmentation behavior with ESI positive mode, which is described in detail in following
example for structure annotation for S 4:21 from LA1777.

Based on differences in exact masses and fragmentation behavior, most acylsugars in
extracts of S. habrochaites LA1777 were assigned as sucrose tetraesters. Figure 3-20
displays extracted ion chromatograms for m/z values corresponding to formate adduct of
acylsucroses for LA1522 and LA1777 accessions. Observed peaks in these
chromatograms separate into two distinct retention time groups. Peaks corresponding to
tetraesters of sucrose are highlighted with a blue dashed line, and those corresponding to
triacylsucroses are highlighted with a red dashed line. The tetraesters were easily
distinguished by chromatography, as all eluted about 5 minutes earlier than the triesters
of the same nominal mass. Five sucrose tetraesters were reported by King1 '3 as
constituents of S. habrochaites that we abbreviate as follows: 8 4:14 (2,4,4,4), 8 4:17
(2,5,5,5), S 4:20 (2,4,4,10) and 8 4:21 (2,4,5,10). All these acylsucroses were tetraesters
with one acetate group on glucopyranose. The LC/MS profiling recognized 18
chromatographic peaks corresponding to formate adducts of tetraacylsucroses from
accession LA1777, and these span 11 different molecular masses ranging from 14 to 24

total fatty acyl carbons.

114

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

. 84:23 '
l [Ti/Z765 2 L 2 i: l fl
3 34:22 :
m/ 751 83:23 .
1 z r 1 ,1 -
m/z737 83:22 .‘ 84:21 .‘
I 1 1 _ ...
m/z 723 S 3:21 : s 4:20 .
1 r 1 . ..
l m/z709 53:20 i l 84:19 : ' 53:20
' - AL;
| m/z695 53:19 :l I 343134 3 ;' s3:29
l m/2681 3311391 l 84:17 - L 83:18
1 - 13L 1 L m
m/2653 $3316 .' 84:15 .'
I 11- 1 ft -
53:15 -' .-
° 4:14 .
l m/z639 l: l s l- .4
0 10 20 ° 30 40 0 10 20 30 40
min min

Figure 3-20. (A) Extracted ion chromatograms (XICs) from negative ion mode LC/MS
analyses of (A) S. pennellii LA1522 and (B) S. habrochaites LA1777. Selected m/z
values correspond to [M+HCOO]_ of sucrose triesters (highlighted with dashed lines) and

sucrose tetraesters (solid line).

The tetraacylsucrose S 4:21 from S. habrochaites LA1777 is used to illustrate how
sucrose tetraesters were annotated. Figure 3-21 (top) shows the mass spectrum obtained
using CID potential of 55 V for the most abundant isomer of S 4:21. Fragments of m/z
87.04, 101.06 and 171.12 were assigned as C4, C5 and C10 fatty acid anions,

correspondingly. This, and all other tetraacylsucroses from this accession showed

115

87.04, 101.06 and 171.12 were assigned as C4, C5 and C10 fatty acid anions,
correspondingly. This, and all other tetraacylsucroses from this accession showed

formation of an ion corresponding to loss of formic acid and ketene from the formate

adduct ([M-H-C2H20]—; m/z 649.34 for S 4:21) that was not present for any of the

sucrose triesters. This fragment serves as evidence of an acetyl ester. Additional
fragments at m/z 537.22, 495.21, 425.17, and 341.11 correspond to further losses of the
longer chain neutral fatty ketenes. Product ion MS/MS spectrum were generated for
formate adducts of S 4:21 in order to establish the source(s) of fragment ions observed
using multiplexed CID, as we did for S 3:22 from LA1522 (Figure 3-21 bottom). The two
spectra share remarkable similarity, and the results suggest that all major fragments could
be explained by decomposition of formate adducts. These findings allow annotation of S
4:21 from LA1777 as a sucrose tetraester with one C2, one C4, one C5 and one C10 fatty
acyl groups. For sucrose tetraesters in S. habrochaites LA1777, the fragmentation
pathway in negative mode was similar with sucrose trimesters in S. penellii LA1522.
Losses of fatty ketene are characteristic of negative mode CID spectra of triacylsucroses.
Their formation could be attributed to either a “charge-directed” process, which involves
attack of anion charge site on the hydrogen of the fatty acid constituents, or a “charge-

remote” mechanism occurring through cis-elimination reaction.

116

[M+HCOO]"
100. 737.36

« C5 fa anion _ M+C| '
1 fly C10 fatty amon [M-H—Cz-C10-C4]' [ l

0 .1 _ _ _ ‘ _ -
/°. ”1'06 171.14 [M-H—cz—C1o—2C41‘ [M H CZ C10] [M_H]727-35
67.04 34111 49521 [M-H-C2l 691.36

323.10 \L4o7_15425.17 [513722 64193411
- r r'LrwJ. Y #44va .firhy rfivrvw .‘. J

 

 

 

 

 

 

0 1711H41.1Lé~4441
100 495.21 TOF msms 737 ESI-
%.
407.15 53721 737.35
1010617114 25-16 54 -33
0879i - 1' 323-193’41'1111L LL ‘1 - ,1. L 591-354 /
100 200 300 400 500 600 700 ”'2

Figure 3-21. (Top) Negative mode ESI spectrum of tetraacylsucrose S 4:21 (RT = 31.50
min) under CID potential 55 V and (Bottom) product ion MS/MS spectrum of formate
adduct (m/z 737.35) for S 4:21 from S. habrochaites LA1777.

More information about which rings bear individual fatty acids position was yielded
from positive mode CID spectra. Figure 3-22 displays a positive mode ESI spectrum of
triacylsucrose S 4:21 from S. habrochaites LA1777 leaf extracts using a CID potential of
40 V. Product ion MS/MS spectra were also generated for ammonium and sodium
adducts of S 4:21 in order to establish the source(s) of fragment ions observed using

multiplexed CID (Figure 3-23). The results suggested that most fragments observed in

positive mode using multiplexed CID were derived from [M+NH4]+, in similar fashion as

the triesters from LA1522 discussed above. [M+NH4]+ for sucrose tetraesters from

LA1777 was cleaved at the glycosidic bond that connects the two rings, forming two

fragments corresponding to acyl substituted glucopyranose and fructofuranose ions.

117

. . . . . . . . 119
Slmllar 1011 behav10r was observed for acylsugars 1n Nicotzana benthamzana.

Dissociation of [M+NH4]+ preferentially cleaved the glycosidic bond and formed cations

stabilized by the adjacent oxygen. This selectivity in yields of fragment ions is attributed
formation of a more stable tertiary carbocation from the fructofuranose ring relative to a

secondary carbocation from the glucopyranose ring (Figure 3-24). For S 4:21, CID of

[M+NH4]+ generated fragment ions at m/z 317.20 and m/z 359.18, with the former being

more abundant under a wide range of collision conditions. In the case of S 4:21, the mass

of the fructofuranose fragment ion (m/z 371.20) is consistent with substitution by one
C10 fatty acyl group. A pseudo MS3 spectrum of product ions from m/z 371 was

generated after forming this ion in the source by raising the source cone voltage (Figure
3-23, bottom). The prominent product ion at m/z 155.12 was assigned as the C10 acylium
ion. The remaining C2, C4, and C5 acyl groups were on the glucopyranose unit for S
4:21 as supported by the fragment ion at m/z 359.17. Fragmentation reactions that

generate structurally useful fragment ions for acylsugars from CID of [M+formate]_ and

[M+NH4]Jr are summarized in Figures 3-25 and 3-26.

Results from CID in positive ion mode allow the fatty acid groups to be assigned to
either the glucopyranose or fructofuranose rings. Sucrose tetraesters in S. habrochaites
LA1777 were substituted with combinations of C2, C4, C5, C10, and C12 fatty ester
groups, but these groups were not randomly distributed. Fragments derived from
cleavage of the glycosidic bonds showed that the glucopyranose ring was substituted with
three variations of a single fatty ester. All C2 esters were on the glucopyranose ring, with
the variations consisting of either two C4, one each of C4 and C5, or two C5 groups.

These are reflected in the minor abundance fragments at m/z 345, 359, and 373 (Table 3-

118

6). Substitution on the fructofuranose ring was limited to a single acyl group, varying
among C4, C5, C10, and C12 esters. Note that the longer C10 and C12 esters are never
observed on the glucopyranose ring. This behavior is the opposite of tomato (S.
lycopersicum M82), where the longer fatty acyl groups were always on the
glucopyranose ring. Accurate mass measurements for major acylsucroses from accession
LA1777 are presented in Table 3-7, and these measurements are consistent with the
assignments described above. Findings from this work indicate that tomato and its
relatives exert tight control over which fatty acyl groups are attached at each position in
acylsugars, yet there is dramatic diversity in acylgroup substitution among related species
within the genus Solanum. Mass spectra that exploit CID behavior offer a powerful
window into the diversity of acylsugar metabolite structures that should accelerate

discoveries of the genetic basis for this selectivity.

 

 

 

 

 

[GLU+nacyI]
100 317.20
%.
[M+Na]+
127.04 251.16 [FRumacy'] [M+NH4]+ 71536
137.11 -
910A] “153.13 271.14 359.18 710.40LA
,l1llLfi1 ,fij , , , l 1 , ,W .-fi . . . . - . . . . 'L'Lm/z
100 200 300 400 500 600 700

Figure 3-22. Positive mode ESI spectrum of tetraacylsucrose S 4:21 (RT = 31.50 min)
from S. habrochaites LA1777 using CID potential 40 V. The abbreviation ‘FRU’ refers
to a fragment derived from the fructofuranose ring and ‘GLU’ refers to a fragment

derived from glucopyranose ring.

119

TOF MS/MS 710 ESI+

 

 

 

 

 

 

100 317.21
1
o/o‘
2 1.17
5 359.19 710.40
0, 11 1 fl L1 1 . . kl. .
100 155.14 TOF MS/MS 317 ES|+
1 137.12
%1
127.04
109.02
J 317.20
0'5160’H200'H300" '4OOH'Y5007'V600'1'700T'm/z

Figure 3-23. (Top) Product ion MS/MS spectrum of [M+NH4]+ at m/z 710.40 and
(Bottom) product ion MS/MS spectrum for m/z 317 ([FRU+acyl]+) of S 4:21 from S.

habrochaites LA1777; m/z 317 was generated in source from tetraesters 84:21 from S.

pennellii LA11777. The abbreviation ‘FRU’ refers to a fragment derived from the

fructofuranose ring.

120

O HO H‘P‘L‘“ 1M+NH41+
0 H
R‘t/ILO ,0 0
O 0
R3—( CY OH
O R 0 OH
2
.>=0
R3
1‘ H
OH /N\ 111
J 0
R4 0 H O R4—LO O OH+ 0
R3\(O «0 HO OH R O 0
R2 0 OH 3\\/ R f O OH OH
O O ' 0
R3. R3

Figure 3-24. Collision induced dissociation of [M+NH4]+ acylsugar ions yield a

dominant acylfructofuranose fragment ion, and a less abundant acylglucopyranose

fragment ion.

121

 

0" ‘OH
C§HO
o l
0&6 l
‘5 m/z:495.21 m/Z'425-17
m/z:.537 22 {£20011 011 6'0 3:343” OH 011
OC\\ 0 Sij: H0<<:C>O»HO do?
C~o
m/z: 3930 15 . 2 0.10
‘5 m/z:407.16 m 3 3

Figure 3-25 Proposed fragment pathway upon CID of the formate adduct for S 4:21 (RT
= 31.5 min) extracted from S. habrochaites LA1777 using ESI negative mode with CID
voltage 55 V.

122

H— H
0” 0H3 0” boH OH
OH 03H 0
O O O + CH
\C C OH 0 0*
/< o o 0" __. 0 0 O ..-..a_, “'0' \
CQ '»0 ’0 /c C» C=O C . .
I O§C (3.40
7'
mz:35917
OH
,0

b . H OH
\ c8 OH

m/z:710.40 \0

m/z:317.20
OH
3
O OH
\\ ‘— .
9W Iago OH

m/z:155.14

Figure 3-26 Proposed fragment pathway of ammonium adduct for S 4:21 (RT = 31.5 min)
extracted from S. habrochaites LA1777 using ESI positive mode with CID voltage 55 V

123

Table 3-6. Fragments and fatty acid constituents of detected sucrose tetraesters from S.
habrochaites LA1777 using ESI positive mode (CID potential, 40 V) and negative (CID
potential, 55 V).

 

Major Fragments in

 

 

 

 

AFiltilyg:clid“ce:::t(iiilt:nat:d “VH'NHM+ INSLZT+ order of decreasing
("l/Z) abundance

S 4:14 C2,C4,C4,C4 612 617 233, 345

S 4:15 C2,C4,C4,C5 626 631 247, 345

S 4:16 C2,C4,C5,C5 640 645 247, 359

S 4:17 C2,C5,C5,C5 654 659 247, 373

S 4:20 C2,C4,C4,C10 696 701 317, 345

S 4:21 C2,C4,C5,C10 710 715 317, 359

S 4:22 C2,C4,C4,C12 724 729 345, 345

S 4:23 C2,C4,C5,Cl2 738 743 345, 359

S 4:24 C2,C5,C5,C12 752 757 345, 373
Acy:::gF;tTye;:ti’:me [NI-5350015 1M+35C'l— Fragment ions (m/z)

constituents z) ("l/Z)

S 4:14 C2,C4,C4,C4 639 629 593, 551, 87
S 4215 C2,C4,C4,C5 653 643 607, 565, 101, 87
S 4:16 C2,C4,C5,C5 667 657 621, 579, 537, 495, 393, 341, 323, 101, 87
S 4:17 C2,C5,C5,C5 681 671 635, 551, 509, 425, 407, 341, 323, 101
S 4:20 C2,C4,C4,C10 723 713 677, 635, 481, 411, 393, 341, 323, 101, 87
S 4:21 C2,C4,C5,C10 737 727 691, 649, 495, 393, 341, 323, 171, 101, 87
S 4222 C2,C4,C4,C12 751 741 709, 663, 481, 393, 341, 323, 305, 199, 87
S 4:23 C2,C4,C5,C12 765 755 719, 677, 495, 411, 341, 323, 199, 101, 87
S 4:24 C2,C5,C5,Cl2 779 769 733, 691, 509, 425, 407, 341, 323, 199, 101

 

124

Table 3-7. Exact mass measurements of [M+HCOO]— ions for detected sucrose

tetraesters in S. habrochaites LA1777 detected using ESI negative mode with CID

potential 10 V.

 

Formula of

 

AcyISEIgzztzlgnbollte 5.25:3: MES/gm Calculated m/z ppm (error)
S 4:14 C27H43017 639.2500 639.2520 3.1
S 4:15 C28H45O17 653.2657 653.2686 4.4
S 4:16 029H47O17 667.2813 667.2815 0.3
S 4: 1 7 C30H49017 681.2970 681.2992 3.2
S 4: 19 (3321153017 709.3283 709.3290 2.5
S 4:20 C33H55017 723.3439 723.3432 -1.0
S 4:21 C34H57017 737.3596 737.3577 -2.6
S 4:22 C35H59017 751.3752 751.3737 2.0
S 4:23 CB6H61017 765.3909 765.3875 -4.4
S 4:24 C37H63017 779.4065 779.4094 3.7

 

In order to provide additional information about structures of fatty acyl groups in
acylsugars from S. habrochaites LA1777 detected using LC/MS and multiplexed CID,
GC/MS analyses were performed for the composite mixture of acylsugar fatty acids in
this accession after transesterification to ethyl esters (Figure 3-27). The result confirmed
the LC/MS findings that C4, C5, C10, and C12 were the dominant fatty acid groups. For
this accession, the ratio of iso-CS to anteiso-CS was about 2:1, a value similar to the S.
pennellii accession (LA1522), but the inverse of the ratio found in S. pennellii LA0716.
The iso-C4 fatty acyl group was dominant in LA1777 and LA0716 (about 35% of total
chains), but C4 was much less abundant (about 2% of total) in LA1522. For C10 and C 12
fatty acyl groups, ratios of branched (iso) to linear fatty acyl chains showed intriguing
variation. For S. pennellii, formation of the branched isomer was favored in LA0716 by

about 2:1, whereas accession LA1522 favored the linear isomer (1 :8 branched/linear). In

125

contrast, S. habrochaites LA1777 favored the branched C10 chain by about 11:].
Corresponding ratios for the C12 isomers followed a different pattern. For S. pennellii,
accession LA0716 favored the linear isomer (1:4 branched/linear ratio), and only the
linear C12 isomer was observed in LA1522. For S. habrochaites LA1777, the branched
C12 isomer was favored over the linear chain by 2:1. These findings suggest that all three
accessions synthesize and attach the same fatty acyl groups onto sugars, with the
exception that LA1522 does not make acylsugars with branched C12 fatty acyl
attachments. However, the relative amounts of the different fatty acid groups show
marked diversity among these three accessions. It is tempting to attribute the variability
in fatty acyl substitution to differences in substrate selectivity during either fatty acid
biosynthesis, attachment of fatty acyl groups to sugars, or a combination of both. The
major conclusion from this exploration of acylsugar structure points to substantial
chemical diversity in acylsugar composition. These findings will be explored in further

detail in Chapter 6.

126

 

iC10

 

 

 

 

 

 

9 11'
a 1 i
5 £ 31
a
.\° -
1C12 nC12
.__,.__,1.~...1.,.i.1_,_,fl,.

Fatty acids

 

 

 

Figure 3-27 Fatty acyl groups from total acylsugars in an extract of S. habrochaites
LA1777 were transesterified to form fatty acid ethyl esters, and analyzed using GC/MS.
Levels for each fatty acid were calculated as a percentage of the total ion current
chromatogram peak areas of fatty acids detected through GC/MS. The abbreviations “i”
and “ai” refer to iso- and anteiso- branched isomers, respectively. Fatty acids with the
same number of carbon atoms are listed in order of increasing retention time from left to

right. Error bars indicate standard error with n = 3.

3.3.4 Structure elucidation of sucrose triesters and tetraesters in S. habrochaites
LA1353.

Another S. habrochaites accession, LA1353, was extracted, and metabolite profiles
were generated to broaden our knowledge about acylsugar diversity, especially since
there were no published reports about acylsugars for this accession. Analyses using
LC/MS suggested that comparable amounts of sucrose tri- and tetraesters are
accumulated by LA1353, but acylglucoses were not detected. Table 3-8 displays all the

detected acylsugars from S. habrochaites LA1353. Fatty acid ester constituents were C4,

127

C5, C10, C11, and C12, but no acetate ester were present and only one chain of C10 or
longer was present in any of the tri- or tetraesters. Accurate mass measurements and a
relatively short chromatographic retention time identified one acylsugar as a pentaester,
with five C5 chains. No other pentaesters were observed in substantial abundance. For
this accession, GC/MS analyses were not performed, so the composition of linear and
branched isomeric fatty acid groups remains undetermined.

Positive ion CID spectra indicated that all triester fatty acid groups were substituted
on the glucopyranose ring. All tetraesters contained a single C5 fatty acid on the
fructofuranose ring (evident as m/z 247 fragment), with all three other ester groups,
including all C10, C11, and C12 groups, were attached on the glucopyranose moiety. The
tetraesters correspond to observed sucrose triesters modified by addition of one C5 fatty
acyl group on the fructofuranose ring. For the pentaester (S 5:25), two fatty acyl groups
reside on the fructofuranose ring (evidence as m/z 313 fragment), while three
substitutions are on the glucopyranose ring, based on the fragment at m/z 415. Specific
positions of substitutions of individual ester groups remain unknown, pending metabolite
purification and NMR characterization.

Fatty acyl substitution patterns in S. habrochaites LA1353 are distinct from all other
accessions described above. For sucrose triesters, all fatty acyl groups are on
glucopyranose. This pattern differs from LA1777 where one group (including the longer
chains) is on the fructofuranose ring, but is similar to S. pennellii LA1522. For sucrose
tetraesters, accession LA1353 always has the longer chain esters on the glucopyranose

ring, a pattern differing from LA1777. Accession LA1522 did not produce tetraesters.

128

Substitution of two fatty ester groups, both C5, on the fructofuranose ring is a trait unique
to LA1353.

Table 3-8. Fragments and fatty acid constituents of detected acylsugars from S.
habrochaites LA1353 using ESI positive mode (CID potential, 40 V) and negative (CID
potential, 55 V).

 

 

 

 

 

Acylsugar metabolite Fatty acid |M+NH 41+ |M+Nal+ Fragment ions
annotation constituents (m/z) (m/z) (m/z)
S 3:19 C4,C5,C10 668 673 511,339,251
8 3:20 C5,C5,C10 682 687 525,485,353,313,229
S 3:21 C5,C5,C1 l 696 701 539,499,397,313,211
S 3:22 C5,C5,C12 710 715 553,513,353,313,211
S 4:19 C4,C5,C5,C5 682 687 247,441
S 4:20 C5,C5,C5,C5 696 701 247,455
S 4:24 C4,C5,C5,C10 752 757 247,511
S 4:25 C5,C5,C5,C10 766 771 247,525
S 4:26 C5,C5,C5,C] l 780 785 247,539
S 4:27 C5,C5,C5,C 12 794 799 247,553
S 5:25 C5,C5,C5,C5,C5 780 785 331,415
Acylsugar metabolite Fatty acid |M+HCOO]' ““3501" Fragment ions
annotation constituents . (m/z) (m/z) (m/z)
S 3:19 C4,C5,C10 695 685 565,495,425,407,341,171,101
S 3:20 C5,C5,C10 709 699 579,495,425,34l,l7l,101
S 3:21 C5,C5,Cl l 723 713 593,509,425,407,341,323,]85,1OI
S 3:22 C5,C5,C12 737 727 607,509,425,407,34l,323,]99,101
S 4:19 C4,C5,C5,C5 709 699 663,579,495,425,34l
S 4:20 C5,C5,C5,C5 723 713 677,593,509,425,341,101
S 4:24 C4,C5,C5,C10 779 769 677,593,509,425,407,341,l71,lOl
S 4:25 C5,C5,C5,C10 793 783 663,579,425,34l, 1 71,101
S 4:26 C5,C5,C5,C] l 807 797 76l,677,593,509,425,185,341,101
S 4:27 C5,C5,C5,C12 821 81 1 775,691 ,607,509,425,34l 323,199,101
S 5:25 C5,C5,C5,C5,C5 807 797 761,677,593,509,425,407,34l,323,]01

 

129

Accurate mass measurements played a key role in the identification of the sucrose
tetraester $525 from accession LA1353. Two sucrose esters with the same formate
adduct nominal mass (m/z 807) were observed, but subtle mass differences allowed them
to be assigned as S 4:26 (5,5,5,11) and S 5:25 (5,5,5,5,5) (Figure 3-28). Additional
support for these assignments came from differences in fragment masses (36 mDa
theoretical difference between isobaric fragments from tetra- and penta-esters). These
small mass differences facilitate data mining for acylsugars with different numbers of
ester groups via generation of extracted ion chromatograms using a narrow mass window,

such as 10 mDa.

 

 

 

 

 

 

[M-H-205]- [M-H-CS]- [M+CI]-
100 797"mm-41000].
593.32 577-38 .
1807.44
0 05 fatty acid _ - - '
/° anion [M-H-505]- [M-H-4C5]-[M H 3C5] [M-Hl-
761.43
101.06 3:11 '11 425.17 \50926 1 L
0 J ’ n 1 - . _. A 1 - n n1 1 1 I
[M-H-CS]—
100 677.41 [M+CI]-
. [M-H-4C5]- [M'H'ZC51' 797°46[M+HCOO]-
C11 fatty ac1d [M-H-3C5]-
aniO" lM-H-3C5-C11l- 595.35 '
0 . ' - -
/" 135.16 509.30 1 [M H] ’ 807-47
C5 fatty ac1d 341.11 ‘
anion Ll 42517 L 761.46 .
0 .101'.9§.1_ hmnm L 3:1 41'. -_ -‘L , -5 11 .311. 1L .. _ A

 

 

 

 

Figure 3—28 (Top) Negative mode ESI spectra of S 5:25 (RT = 31.91 min) and (Bottom)
S 4:26 (RT = 36.04 min) from S. habrochaites LA1353 using CID potential 55 V.

l

30

This survey of two acccesions from each of two plant species (S. pennellii and S.
habrochaites) showed dramatic differences in acylsugar fatty acyl chains and positions of
fatty ester substitutions. To facilitate comparisons of these genotypes, the structures of

acylsugars are summarized in Figure 3-29.

131

Figure 3-29 Putative structures for detected acylsugars among S. pennellii LA0716 and
LA1522 and S. habrochaites LA1777 and LA1353. Specific substitution positions of

individual fatty acids remain uncertain for some metabolites.

132

O H O H O OH
R410&§\0H R4106 0‘ H
S pennellii /< o O H (11

LA0716 R330 02: R3 0 712
Glucose triesters Sucrose triesters
R2,R3,R4=C4,C5,08,CQ,C10,C11,C12 R2,R3,R4=C4,CS,C1O,C11
Only one chain C3 0' longer Only one chain C10 or C11

0 H

0 OH
0
S. enne/Iii R450§§ lat/Low O
p 0 o H o 0
LA1522 R3»\{ 0 R3 0 j H H 3
0021 2

Sucrose triesters

R2, R3, R4: C5, C6 C8, 09, C10, C12
R2 R3 R4: C4 C5 06: 03 09 C10 912 Only one chain CB or longer, no R3’ substitution

Only one chain C8 or longer
R4R3 £&{ 0%.?

S. habrochaites Sucrose tetraeszters

LA1777 R2=CH3; R3,R4,R3-=C:C5,C1O,C12
Only one chain C10 or longer

110 H 0 ”o
Rizfléjjwfi) 5

H H
.>=°
Sucrose triesters R3
R3,R4,R3.=C4,C5,C10,C12

Glucose triesters

Only one ochain C10 or C12
OH
Fill/Clio 0% 0H0
S. habrochaites R340 0 O OH R534: Oio
LA1353 :12 H H

Sucrose tetraestezrs Rg'>=o
R3=CS; R2 R3 R4: C4, C C10, C11, C12

R2 R3 R4= C4 C5 C10 C11 012 Only one chain C10 or longer
Only one chain C10 or longer 0 OH

Sucrose triesters

OH
Rfio 0 o
0 o 0 ._
2 [ho

Sucrose pentaResters
R2=R3= R4=R3' =R4=05

133

3.4 Conclusions

This research has presented application of LC/TOF with multiplexed CID for
comprehensive profiling of acylsugar metabolites extracted from trichomes of four wild
tomato relatives. This approach has generated accurate masses for both 3 adduct and
fragment ions, yielding information content similar to MS/MS spectra but without the
loss of information that can occur during LC/MS/MS analyses. These analyses led to
detection of about 50 different kinds of acylsugars, not including isomers within specific
accessions. The analyses were performed using a relatively inexpensive model of mass
spectrometer, and the successes of this work suggest that laboratories that do not enjoy
access to more expensive instruments can make substantial contributions to metabolite
discovery.

However, the exact positions of fatty acyl chains on acylsugars and assignment of
fatty acyl groups as branched or straight chains remain to be determined. Efforts to
produce position-specific fragment ions through metal cationization yielded
unsatisfactory results owing to facile losses of fatty acid groups during collision induced
dissociation. Purification and characterization using NMR are underway in the Jones
laboratory. Even though complete structural details for acylsugar metabolites have not yet
been achieved, these findings raise awareness of diversity of secondary metabolites in the
genus Solanum. More extensive profiling of other accessions, not described in this
dissertation, has shown that every accession displays a profile of secondary metabolites
that distinguish it from all others. Experimental determination of metabolic diversity
provides researchers with an opening to discover how genes and environment combine to

produce different biochemical outcomes, and rapid screening for differences in plant

134

'84 Future research should exploit

chemistry can guide efforts toward gene discovery.
information about the secondary metabolome to establish a scientific foundation for

understanding of biological roles of genes and metabolites.

135

Chapter 4. Annotation of polyphenol, glycoalkaloid, and terpenoid
secondary metabolites in glandular trichomes from tomato, its wild
relatives, and near isogenic lines, based on LC/TOF MS coupled with

multiplexed CID

136

4.1 Introduction

Glandular trichomes of wild relatives of tomato produce secondary metabolites in
abundance. Chapter 3 of this dissertation described annotations of a diverse set of
acylsugar metabolites, and in this chapter, additional metabolites from other compound
classes are the focus.

Three classes of secondary metabolites, the polyphenols, glycoalkaloids and
terpenoids, are the most recognized chemical defenses used by plants as protection

s
'23"8“'86 These compounds also have been touted as

against insects and pathogens.
providing benefits, in the case for of polyphpenolic flavonoid antioxidants, and harm, as
observed with teratogenic behavior of glycoalkaloids, to human diet and health.'22 The
biological potency of these compounds lends impetus to efforts aimed at discovering
genes responsible for their biosynthesis in plants. All three classes of metabolites present
potential scaffolds for development of novel pharmaceutical and nutraceutical agents.
Complete genetic maps of the enzymes involved in their biosynthesis are not available
for any of these classes of metabolites. Glandular trichomes are prolific chemical
factories known to synthesize chemical defense compounds in plants, and are therefore
attractive targets for investigating relationships between gene expression and chemical
composition.

Flavonoids are a prominent group of polyphenols that consist of two aromatic rings
with six carbon atoms (A and B ring) and linked through a heterocyclic six-membered

ring (C ring)'87

containing an ether oxygen (Figure 4-1). In tomato and its relatives,
structural diversity among flavonoids is due to differences in positions of hydroxyl

substitution, and the presence or absence of methylated, acylated and glycosylated

137

hydroxyl groups.'88 Owing to their importance in tomato flavor (they impart bitterness),
flavonoid content has primarily been investigated in fruits of tomato plants, where

flavonoid glycosides are abundant and have been catalogued in a tomato metabolite

103

database. This database reported that observed glycosylation primarily occurred at the

'26" 89 are important

3-position. Flavonoids and their structurally related derivatives
secondary metabolites in trichomes of Solanum species that contribute color to plant
organs, protect against biotic and abiotic stress, and regulate normal plant

'90 However, closely related plants often differ in flavonoid profiles, and

development.
more advances are needed to identify the genetic basis for differences in flavonoid

composition in specific plants and cell types.

Figure 4-1. Generalized structure of flavonoid metabolites known in tomato and its
relatives, where R groups can be hydrogen, hydroxyl, or other substituted oxygen-

containing groups.

Glycoalkaloids are secondary metabolites common in the family Solanaceae that
contain a basic nitrogen in a heterocyclic ring and are conjugated to an assortment of
carbohydrates. These compounds are produced by a variety of Solanaceous plants,

including potato, tomato and eggplant,'9 and largely differ in the structure of the

138

oligosaccharide and in the number of double bonds (Figure 4-2). Glycoalkaloids are
bioactive secondary metabolites with toxicity to bacteria, fungi, viruses, and insects,
which play an important role in protection of disease in plants.192 In tomato,
glycoalkaloids include a—tomatine and dehydrotomatine, which contain a spiro-ring
geometry somewhat unusual in plant metabolites, and exhibit an assortment of subtle
structural differences relative to chaconine or solanine and leptine in potato or

192

solamargine and solasonine in eggplant. Alkaloids are reported to have a variety of

pharmacological and nutritional properties in animal and humans, thus, we need a better
understanding of the role of these compounds both in the plant and in the diet.'27"28
Although much effort has been put into study of biosynthesis of glycoalkaloids,I93 many

biosynthetic steps remain unknown, especially the formation of the nitrogen heterocyclic

F ring of tomatine.

 

Figure 4-2. Generalized structure of glycoalkaloid metabolites known in tomato and its

relatives, where R groups are glycosides.

Terpenes, the most abundant and structurally diverse group of plant secondary

metabolites,120 consist in tomato largely of the C10 mono-, and C15 sesquiterpenes.8"94

139

Mono- and sesqui-terpenes are abundant and volatile metabolites that accumulate in
tomato glandular trichomes.195 Aside from these volatiles, relatively few oxidized
terpenoid metabolites have been observed in tomato and its relatives, with sesquiterpene
carboxylic acids from S. habrochaites LA1777 being an exception'w"I97 The potential
diversity of terpenoids arises from different precursors, multiple structures that can arise
during cyclization, and variations in substitution patterns of metabolic oxidation.
Metabolite identifications offer potential for accelerating discoveries of gene functions,
and findings of this kind help integrate metabolite information into system-level studies
of how plants synthesize defense chemicals.'98

One long standing approach used for discovery of gene functions is called forward
genetics, which involves searching for genes associated with specific traits. For the
purpose of this study, those traits are levels of individual secondary metabolites. In order
for this approach to be efficient, it is necessary to perform deep and rapid screening of
genetic variants for differences in phenotype, which in this case is plant chemistry. Once
variations in chemistry are recognized in specific plants, the metabolite profiles serve to
guide discovery of genes responsible for plant chemistry. These investigations often
explore differences between naturally occurring plant species and between offspring
produced by genetic crosses.

One of the most successful examples of the use of metabolite profiling for discovery
of plant gene functions was reported by the group of Kazuki Saito of the RIKEN Plant
Science Center.'99 In this study, metabolite profiles generated using combinations of

HPLC and mass spectrometry were employed to establish functions for two Arabidopsis

glycosyltransferases in the biosynthesis of flavonoid and anthocyanin glycosides.

140

In the studies described below, LC/TOF MS with nonselective and multiplexed
collision induced dissociation (multiplexed CID) was developed and applied to compare
metabolite profiles for tomato and several genetic variants. These include several wild
tomato relatives and lines derived from crossing tomato (S. lycopersicon M82) with the
wild tomato S. pennellii (LAO716), and generating genetic variants of tomato that contain
small segments of genomic DNA from the wild tomato line. These new genotypes,
termed introgression lines, were generated in the laboratory of Dani Zamir (REF), and
they are useful for establishing roles of specific DNA regions on the chemical
composition in specific plant tissues. Comparisons of metabolite profiles of conventional
tomato with these introgression lines and wild tomato relatives help target subsequent

gene mapping efforts to narrow regions of specific tomato chromosomes.

4.2 METHODS

4.2.1 Materials. Plants were grown in a greenhouse in Michigan State University from
seeds obtained from C. M. Rick Tomato Genomics Resource Center at the University of

California-Davis.

4.2.2 Chemicals. Acetonitrile (HPLC grade), 2-propanol (HPLC grade), methanol
(HPLC grade), and formic acid 88% were obtained from VWR Scientific and Fisher

Scientific.

4.2.3 Extraction Method. To extract metabolites from Solanum trichomes, ~100 mg of

individual leaflets were dipped in 2 mL isopropanol: CH3CN: H20 (3:322 v/v/v) for 1 min.

A 1.0 mL aliquot of each extract was transferred to a 1.5 mL polypropylene

141

microcentrifuge tube, and the tubes were centrifuged (5000 rpm) for 2 min at 25°C, and

the supematants were analyzed without further processing. Extracts were stored at -20°C.

4.2.4 High-Performance Liquid Chromatography. A Shimadzu (Columbia, MD) LC-

20AD HPLC ternary pump was used. The separation was performed on a Thermo 1 X

150 mm BetaBasic C13 column (100 X 1 mm, 311m). Gradient elution was executed based

on 0.15% aqueous formic acid (solvent A) and methanol (solvent B) as follows: initial
conditions 5% B; linear gradient to 50% B (5 min), 95% B (33 min), and 100% B (35

min); hold at 100% B until 38 min, and return to 5% B (43 min). The flow rate was 0.1

. -l . . . . . .
ml- mm , and the Injection volume was 10 1,11. The column was maintained at 30 °C in a

column oven.

4.2.5 Mass Spectrometry. Analyses were performed using a model LCT Premier mass

spectrometer (Waters, Milford, MA). Operating parameters were set as follows: capillary

, -1
voltage: 3.2 kV in both positive and negative mode; desolvatlon gas flow: 300 L-h ;

. -1
desolvation gas temperature: 200 °C; source temperature: 90 °C; cone gas flow: 20 Lb .

The instrument was operated in both negative and positive ion electrospray modes using
separate injections of extract for each ionization mode. Analyses were performed using
rapid switching of Aperture 1 voltage in the ion transit region of the mass spectrometer,
proving quasisimultaneous generation of spectra under fragmenting and nonfragmenting
conditions. Mass spectra were acquired over m/z 50 to 1500 with a scan time of 0.4 s for
each function. Spectra were acquired in centroid format using ‘dynamic range
enhancement’ (DRE) enabled, which acquires separate mass spectra by switching a beam
attenuator on and off, followed by post-acquisition stitching together of spectra. The

instrument mass scale was calibrated daily in both ionization modes over m/z 50-1500

142

using 0.1% aqueous phosphoric acid as reference. Most analyses were performed using
the V-mode ion pathway in the mass analyzer, which generated a resolving power of

5000 (FWHM). For some analyses requiring more exact mass measurements, 0.1%

H3PO4 was infused through the reference sprayer of the lock-spray source as a reference

lock mass, and the lockspray attachment switched between sample and reference streams.

4.3 RESULTS

4.3.] Structural annotation for polyphenols

Flavonoid metabolites were detected in extracts of all of the plants described above,
and their annotations were based on chromatographic retention times, accurate
measurements of molecular and fragment ion masses, proposed elemental formulas
derived from these measurements, and knowledge of previously-identified metabolites
from the tomato fruit metabolite database (MoToDB) .200 Flavonoid metabolites from
three Solanum accessions consist of several flavonoid glycosides and methylated
flavonoid aglycones as described below.

Putative structures for detected flavonoid glycosides are displayed in Figure 4-3.
Rutin (quercetin-3-rutinoside) is used to illustrate how LC/T OF MS data were used to
drive metabolite annotation. First, candidate polyphenol metabolite ions were
distinguished in negative ion mode based on having lower mass defects (the value
following the decimal place) than other metabolite ions of major abundance. Positive
mass defects in the range of 100-200 mDa were considered likely candidates. Second,
candidate flavonoid glycosides eluted earlier than acylsugar metabolites, making them

easy to distinguish. Third, shapes of breakdown curves generated using multiplexed C1D

143

(Figure 2-4 for rutin), were used to assign ions of major abundance as fragments, adducts
or molecular ions for each candidate flavonoid glycoside. In one specific case, an ion at
m/z 609.15 was most abundant at the lowest CID potential of 10 V, and gradually
decreased in abundance with increasing CID potential, suggesting a pseudomolecular ion
([M-H]'). Calculation of possible elemental formulas was performed using Masslynx

software, using a mass window of 10 ppm, and the assumption that no nitrogen, sulfur, or

phosphorus atoms were present. Only one elemental formula, C27H29016‘ was consistent

with the measured m/z. The calculated value for the double bond equivalents (DBE,
corresponding to the number of C atoms minus half the number of H atoms, plus 1) for
this formula was 13.5, which indicated it had numerous rings plus double bonds, as
would be consistent with an aromatic compound. Additional evidence about glycoside
structures came from masses of fragments in both negative and positive modes. In the
latter, fragments at m/z 465 corresponding to loss anhydrorhamnose (146 Da) and m/z
303 (loss of anhydrorhamnosylglucose) were suggestive of glycosylation pattern
consisting of a rhamnosylglucose attachment (Figure 4-4). In negative mode, both sugar
groups were eliminated together, showing fragments for this metabolite at m/z 301 and
300 (Figure 4-5). These ions are annotated as quercetin aglycone [M-H-
anhydrorharnnosylglucose]' and a corresponding aglycone radical anion, respectively.
Aglycone refers non-sugar compounds after replacement of glycosyl group from a
glycoside by a hydrogen atom. Such fragment ions are consistent with earlier work from
the laboratory of Magda Claeys.2°' The facile formation of the radical anion (m/z 300) by
homolytic cleavage between sugar and aglycone is common for flavonoids glycosylated

in the 3-position, as many of these flavonoids form stable radical anion fragments. Based

144

on these results, this metabolite was annotated as quercetin-3-rutinoside (rutin). Figure 4-
6 displays proposed fragmentation pathway for rutin (quercetin-3-rutinoside) using

negative mode.

 

Quercetin-di glucose

OH

Figure 4-3 Putative structures for several flavonoid derivatives from Solanum wild

species.

145

303

 

 

 

 

 

 

 

 

100~ [M+Na]+
1 633
0/0_
[ll/”HI+
611
304
‘ 465 Rha
(NC
0. 4.4. ,- ..1. . . .1. . . . . W4 1. . . . . .ITI/Z
100 200 300 400 500 600 700

Figure 4-4 ESI mass spectra for quercetin-3-rutinoside from S. pennellii LA0716 using
positive mode with CID 40 V.

146

100-

0/0"

 

 

 

 

 

 

 

/“——’/ '\\
0 .4 4 4 -.:— ‘;‘ lirfi—fi—wWQ3x W <~— . 4 . . . min
10 1 12 13 14
100 300.03
1
°/°‘ , 301.04
‘ 271.02
1 255.02 /237.03 607115
1070115101 243033 1 l l
O 1 .‘ 1 -51 r r1 r . 5" 15L v» - rrrrr . 1 T '7 Y - 1 1 m/z
100 200 300 400 500 600

Figure 4-5 (Top) XIC for m/z 609 from S. pennellii LA0716 using ESI negative mode

with CID voltage 10 V. (Bottom) ESI spectrum of quercetin-3-rutinoside (RT = 11.52
min) using ESI negative mode with CID voltage 55 V.

147

  

OH 0 OH OHOH
OH

H
m/z: 609.15 \

 

301. 04

Om/z? 030:)00:/OO/ OHm/z:

OHO

 

OH O
m/z: 179.00 / m/z: 151,0]
6
O —9 ° 7'
OH OH OH
m/z: 243.03 mm: 107.01

Figure 4-6 Proposed fiagmentation pathway for quercetin-3 -rutinoside from S. pennellii
LA0716 using ESI negative mode with CID potential 55 V.

148

Two chromatographic peaks exhibiting the same exact mass at m/z 609.15 in negative
ion mode (Figure 4-7) were observed in an extract of S. habrochaites LA1777. Based on
exact mass measurement and fragment ion masses, the second peak was assigned as
quercetin-3-rutinoside due to characteristic fragment ions at m/z 300.03 and 301.04 for
this metabolite. In contrast, the later eluting metabolite yielded fragment ions at m/z
284.03 and 285.04 consistent with the aglycone of kaempferol, 16 Da less than aglycone
of quercetin. These findings together suggested the major and earlier eluting metabolite
was kaempferol-3-glucosylglucoside, although the mass spectra cannot distinguish the
various hexose isomers. Additional support for this annotation is based on positive mode
CID results including fragments at m/z 449.03 (-162, loss of one hexose) and 287.04 (-

162-162, loss of two hexoses).

149

1001

 

 

 

 

 

11.45!
1 ll
(1
1
1
l1
1:
M
1 11
1 1' ‘1
/ 'U/I'" .
.,-3.fi-,--, -..“--. , vvvvvvvvv <fivv *va rmkrfi‘fi Tfi'u ' ' ‘ T'mln
0 4 6 8 10 12 14 16 13
1004 284.03
%j 285.04
1 255.02
4 27.021 609.15
Oji—r—rwrfi—fi—fi—r—H—fi—T—JZg—flowo—Ffi—rrlw—f fifiJl—fi affirvv ., ,nnivfivrn f 4 fififi v 4- »Y~i——r
.1
100— 609 5
%.‘ 300.03

 

 

 

041013.11 WWI 'J‘Ahlai,4LJL—I+h ' HWJ».11LA.LL++JILL'1 4W1 . l'MWLvW-WWLHLJWHWWVLWf' "*7 m/z
100 200 300 400 500 600

Figure 4-7 (Top) Two isomeric metabolites appear in the extracted ion chromatogram
m/z 609 for an extract of S. habrochaites LA1777 using CID voltage 10 V. (Middle) ESI
spectrum of first peak at 11.04 min for kaempferol-3-glucosylglucoside at CID voltage of
55 V. (Bottom) ESI spectrum of second peak (RT = 11.45 min) annotated as quercetin-3-
rutinoside using ESI negative mode with CID voltage 55 V.

150

In addition to flavonoid glycosides, anthocyanins are also observed in leaf tissue from
Solanum species. Anthocyanins possess a permanent positive charge, and are best
characterized using positive mode ESI. Structures of anthocyanins from extracts of
glandular trichomes of tomato and its relatives were annotated based on comparisons of
exact masses and fragmentation patterns with metabolites in the MoTo tomato fruit
metabolite database.202 Putative structures of four anthocyanins were detected and
displayed in Figure 4-8. They were assigned as coumaroyl esters of glycosides of
petunidin, malvidin, and delphinidin and a caffeoyl ester of petunidin glycosides based on
ions at m/z 933.27, 947.28, 919.25 and 946.26 respectively. In all cases, the masses and
CID spectra were consistent with glucose conjugation at the 5-position and conjugation
by the disaccharide rhamnosylglucose at the 3-position. The phenolic acid esters, p-
coumaroyl and caffeoyl, are attached on the disaccharide. Anthocyanin aglycones were
annotated based on fragment ions at m/z 303, 317 and 331 corresponding to delphinidin,
petunidin and malvidin respectively. For example, for pettmidin-3-O-(p-coumaroyl)- and
Malvidin—3-O-(p-coumaroyl)-rutinoside-5-O-glucoside, fragments ions at m/z 317.06 and
331.07 (Figure 4-9) were due to loss of glucoside (-162 Da), rutinoside (-162) and p-
coumaroyl (-292) from m/z 933.27 and 947.28, so the aglycones for those two

anthocyanins are determined as petunidin and malvidin, respectively.'()3

151

Figure 4-8 Putative tructures of four anthocyanins petunidin-3-O-(p-coumaroyl)-
rutinoside-S-O-glucoside (m/z 933), malvidin-3-O-(p-coumaroyl)-rutinoside-5-O-
glucoside(m/z 947), delphinidin-3-O-(p-coumaroyl)-rutinoside-5-O-glucoside (m/z 919)
and petunidin-3-(caffeoyl)-rutinoside-.5-O-glucoside (m/z 949).

152

  

O
HO
HO
HO
OH
m/z 949 26
OCH3
OH
OCH3
g
0

OH

 

HO

HO

HO
HO
m/z: 919.25

OH

 

 

m/z: 947.28

m/z: 933.27

153

317.06

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

109. 771.11

1 479“ 933.27

°/0.: 4

0' ......... 14,4. .1.7_. -4 . MAJ—fl mwafigflwwLfiF—fimH—W . 444 r
1 493-12 947.28

%1 < <

O 4T.-,9r--.fid#fil ...... ..lr --.%.+45,4. -T.-..,...-,J- ,fi-4T-.. H4. 4—4—er/z
200 300 400 500 600 700 800 900

Figure 4-9 ESI mass spectra from (top) anthocyanins at m/z 933 (RT = 10.97 min) and

(bottom) m/z 947 (RT = 11.25 min) from S. lycopersicum M82 using positive mode with
CID 40 V.

Nontargeted analysis and annotation of polyphenolic metabolites is critical to
establish whether they represent end products such as flavonoid glycosides, products of
competing side reactions such as anthocyanins, or intermediates that accumulate owing to
inefficient catalysis of biosynthetic steps. Although flavonoid biosynthesis has been
widely studied,203'205 identification of all genes involved of the flavonoid pathway,
including metabolic modification of flavonoids, has yet to be achieved in specific plants
and cell types.43 A generalized depiction of the biosynthetic steps in synthesis of

flavonoids and anthocyanins is presented in Figure 4-10.

154

101:3“

HOOC CoASOC pc—oumarHoyl CoA

01
p-coumaric :18“

HO
1 Naringenin chalcone
0”

OHO

  

Naringenin
Homem' “DWI: Kaempferol glycoside
OH 0 OH 0
Dihyrokaempferol1O kaempferolR OCH 3
l + OH
/ OCH3
OGlc O-Glc-o.Rha
OH OH . . -
leucocyanidins 3-OH-anthocyan1nd1ns anthocyanins

Figure 4-10 Important steps in the biosynthesis of anthocyanins and flavonoids (adapted
from B. Winkel-Shirley and K. Saito43’206).

Parallel investigations were conducted in collaboration with the laboratory of
Professor Gregg Howe of Michigan State University to explore the role of trichome
chemistry in susceptibility of plants to herbivory of insects using wild type tomato and
mutant af (anthocyanin free). The mutant line, which was recognized by its lack of red
pigmentation, was more susceptible to herbivory by the insect Manduca sexta, and the
biological findings are described in more detail elsewhere (Kang et a1., in preparation).

Metabolites extracted from leaf trichomes by solvent dip were analyzed using LC/T OF

155

MS and multiplexed CID. An assortment of glycosides of the flavonols quercetin and
kaempferol were detected in both wild type and mutant plants, with quercetin-3-
rharnnoside the most abundant. Levels of all flavonoid glycosides, were about 6-fold
lower in the af mutant, and anthocyanins were below detectable levels in the mutant.
Metabolites with masses consistent with aglycones and glycoconjugates of the flavonoid
precursors naringenin and/or naringenin chalcone were detected in wild type but not in
the af mutant. These results suggest some steps in flavonoid pathway are inefficient in the
af mutant, but this mutation does not completely eliminate flavonoid biosynthesis.
Further investigations of differences in secondary metabolite profiles between wild type
and af mutant are continuing, with one goal being the discovery of tomato genes that
might contribute greater insect resistance in tomato plants.

F lavonoids are often present as flavonoid glycosides in tomato plant flowers, leaves.

20' No flavonoid aglycones have been reported in tomato tissues, however,

stems or roots.
based on our results, various methylated aglycones of flavonoids were found in extracts
of trichomes from S. habrochaites LA1777 and S. lycopersicum M82 (commercial
tomato). Figures 4-11 displays extracted ion chromatograms for pseudomolecular ions of

methylated myricetins in an extract of S. habrochaites LA1777, with masses

corresponding to 2-5 methyl groups.

156

100
96

 

A 5M

 

 

 

 

0 10 Y ' ’ 20 " Y 7 30 i ' ' 40
100 4'“
%1
o . . . . 7}"1 . . - - . . . . 4 .
10 20 30 40
100 1 3|“
1 1
O . . 1* +74 'h—itrfi 4 , 4 . . Y . v T
10 20 30 40
100 2'“
96] J
O -SW<<-4w¢nmanmxfihal¢:£fiaxmifllLfigmwaammwdumLficxfltKSLAQflwmflawnu:wnun
10 20 3O 40

Figure 4-11 LC/MS XICs for a trichome extract from S. habrochaites LA1777 using ESI
positive mode with CID voltage 10 V showing: m/z 347 (dimethylmyricetin, 2M), m/z
361 (trimethylmyricetin, 3M), m/z 375 (tetramethylmyricetin, 4M) and m/z 389
(pentamethylmyricetin, 5M).

Annotation of methylated flavonoids is based on masses of pseudomolecular and
fragment ions generated using ESI in both positive mode. Proposed structures from
mono- to pentamethylmyricetins are shown in Figure 4-12. Methylated flavonoid
candidates have positive mass defects in the range of 40-100 mDa. These metabolites
elute later than flavonoid glycosides owing to the lack of the polar carbohydrate groups,

but earlier than acylsugars. Fragment ions generated using elevated CID energies were

consistent with multiple losses of —CH3 (A15 Da) groups in both positive and negative

modes. Additional fragmentation involves cross-ring cleavages, and common positions of

ring cleavage of flavonoids are displayed in Figure 4-13.207 Fragmentation pathways of

157

tri- and tetrarnethylmyricetins are used as examples to show how these structures were
annotated. Trimethylmyricetin from S. habrochaites LA1777 has its three methyl groups
distributed among the A, B, and C rings based on fragment ion masses produced using
ESI positive mode. Determination of the positions of methylation proved challenging
owing to the facile losses of methyl radicals as discussed in more detail below. Although
evidence for one methyl group on the A ring for this metabolite was suggested by the
fragment ion at m/z 167.03, attributed to 1,3A+ cleavage using the nomenclature
developed by Mabry and Markham33 However, the same fragment mass also suggests
localization of a methyl group on the B ring as a 1,ZB+ fragment ion. Extensive
homolytic cleavage of the bond between the oxygen and methyl group was observed in
multiplexed CID and in MS/MS product ion spectra. The dominant fragments for
trimethylmyricetin included those arising from losses of one and two methyl radicals
from [M+H]+. Figure 4-14 displays ESI spectrum of trimethylmyricetin using ESI
positive mode and product ion MS/MS spectrum for trimethylmyricetin. More detailed

characterization of methylated flavonoid aglycones is being conducted by graduate

student Chao Li who is using MS/MS and M83, and these results will be presented

separately. Proposed fragmentation pathways for trimethyl myricetin using positive mode

are shown in Figure 4-15.

158

 

penta-methylmyricetin

Figure 4-12 Putative structures of major methylated flavonoids in Solanum species

LA1777. The exact positions of methyl groups on these flavonoids retain unknown.

159

0,2A+ 0,3A+ R2

 

Flavonol
Kaempferol: R1= H; R2 = H
Quercetin: Rl= OH; R2 = H
Myricetin: R1: OH; R2 = H

Figure 4-13 Nomenclature of fragment ions generated from flavonoids.207

 

 

 

1001 . [M+H-15]+ 3
1 346.07 1M+H1
1 361.09
1 [M+H—30]+1 1
o , + ‘
/°l 167.03 [M+H—30—28] 331‘~06‘ 1
1 303.05 1 1 1
‘1 275.05 1 1 1 ‘
1
0 1 ' 1 1 ‘ 11,1 11111 'l 1' 1‘ 11 ‘l ‘ 1,1 11'11' '
[M+H—15]+
100 346
l .1M+H—301*?
l [M+H—30-28] 1
. 303 331 .
% 167 275 l 1 1 [M+H]*
1 .
1 l l 1 317 ' 1; 361.09
0' ' > ‘ 1 1 . . ' . “ " i 1', " , 11 ‘1 '1' 1 11 1 m/z
30 120 160 200 240 280 320 360

Figure 4-14 (Top) Positive mode mass spectrum of trimethylmyricetin from S.
habrochaites LA1777 using CID potential 80 V. (Bottom) Product ion MS/MS spectrum
of [M+H]+, m/z 361 trimethylmyricetin (positive mode) using collision energy of 60 eV.

160

   

OH 9H2
m/z; 361.09 m/z:346.07 m/z:275.08

l \
OCH3
H3CO 6H
1?; we“

OH 0 m/z:167.03
m/z:167.03

Figure 4-15 Proposed formation of fragment ions of trimethylmyricetin in ESI positive
mode, obtained from extract of S. habrochaites LA1777

Positive mode mass spectra and MS/MS product ion spectra were generated for

tetramethylmyricetin from S. habrochaites LA1777 (Figure 4-16) Fragment ions for

suggested two methyl groups on B ring and another two on A and C ring respectively.

Two methyl groups attached on B ring were suggested by fragment ion at m/z 181.05, but

this fragment might also be explained by methyl substitutions on the 5 and 7 positions on

the A ring. Therefore, positions of methyl group substitution were not established from

these experiments.

16]

[M+H-30]+

 

 

 

 

 

 

100 34507
%~ [M+H-15r
1 314.05 [M+H]”
289.07 330 06 3600937510
167.03 181.05 271-07 1 1' 1 ‘ U l
0 fr v. 7W r r [W Jer Y_.Lfi ng fir LY - 1L" 1 ltlr if; ,1 ‘erfivlvlrv'v'fiil Y-" yl vfifigfi
314
1
100 360
%
167 271 345
J 181 243 257 1 #139 1! 330 1 1
OJ. . ... , ..L .4-m-aawim-u-di,_ta.1afiu4lwattwliwwlmama.m/z
140 180 220 260 300 340 380

Figure 4-16. (Top) Positive mode mass spectrum of tetramethylmyricetin extracted from
S. habrochaites LA1777 with CID potential 80 V. (Bottom) Positive mode product ion
MS/MS spectrum of tetramethylmyricetin at m/z 361 using collision energy 40 V.

Although the exact positions of methyl groups on flavonoids could not be definitively
assigned without standards, annotation of these metabolites is important for metabolic
phenotype screening and association of methylation patterns with investigations of
specific methyltransferase enzymes. Methylation patterns differed in various tomato lines.
For example, S. lycopersicum M82 contain methylmyricetins ranging from 1-3 methyl
groups, whereas in S. pennellii LAO716, the core aglycones differed, spanning
methylated kaempferol, and quercetin, but barely detectable amounts of methylmyricetins.
Kaempferol aglycones contained 1-2 methyl groups and quercetin had 1-3 methyl groups
in this genotype. When metabolic profiles of flavonoids were surveyed among

introgressions of LAO716 >< M82, introgression lines (ILs) IL6-3 and IL6-4 displayed

higher ratios of tri- to dimethylmyricetin, This ratio was inverted relative to other ILs and
recurrent parent M82 (detailed in Chapter 5), implying that methyltransferases
originating from chromosome number 6 of S. pennellii somehow alter the amount of
trimethylmyricetins among these two ILs. These findings serve as the basis for
continuing efforts to characterize O-methytransferases involved in myricetin methylation
in S. habrochaites LA1777 in the laboratories of Jones and Eran Pichersky (University of

Michigan).

4.3.2 Structural annotations for glycoalkaloids

Glycoalkaloids make up another important class of specialized metabolites in tomato
plants.200 These compounds act as toxic defense compounds and exist in numerous plant
tissues, including the fruit and root.127 The name refers to two characteristics — the term
“glyco” implies that these metabolites are conjugated to carbohydrates. Alkaloids are
metabolites containing basic nitrogen functional groups. In tomato, the best-known

glycoalkaloids are variations of tomatine (Figure 4-17).

 

HO HO
HOW
HO HO
HO
H
HO

Figure 4-17 Chemical structure of tomatine

163

Glycoalkaloids are easily protonated in positive mode owing to the basic secondary
amine group, and they are easily detected using electrospray ionization in positive mode.
They lack acidic functional groups, and are more likely to be observed as adduct ions (e. g.
formate adducts) in negative mode. The polar carbohydrate groups and the positively-
charged protonated amine confer sufficient interaction with polar solvent for these
compounds to elute earlier than acylsugars on reversed phase separations used in this
work. Glycoalkaloids are recognized based on their large positive mass defects in the
range of 400-600 mDa, owing to the substantial hydrogen content in the steroid
aglycones.

Mass spectrometry has been a useful tool for characterization of glycoalkaloids.
Positive mode MS/MS product ion scans generated using an ion trap mass
spectrometer208 were generated for steroid glycoalkaloids extracted from tomato leaves to
elucidate fragmentation mechanisms. The findings can be summarized as follows:
protonated glycoalkaloids undergo ready loss of water accompanied by losses of
carbohydrate moieties, with relatively low levels of fragments useful for investigating the
structure of the aglycone portion (minus carbohydrates).

Based on characteristic fragments summarized in a 2005 report,208 the glycoalkaloid
metabolites were identified in trichome extracts using LC/T OF MS with multiplexed CID.
Manual inspection of negative ion LC/MS results for an extract from tomato S.
lycopersicum M82 trichomes led to recognition of two abundant peaks with even nominal
masses, indicative of an odd number of nitrogen atoms present. Extracted ion
chromatograms (XICs) for masses of formate adducts of glycoalkaloids a-tomatine and

dehydrotomatine, which were previously known tomato metabolites. are shown in Figure

164

4-18. Identifications of these metabolites were supported by chromatographic retention
time, accurate mass measurements in both positive and negative ion modes, fragment ion
masses, proposed elemental formulas derived from these measurements, and knowledge
of previously—identified metabolites from the tomato fruit metabolite database (MoToDB)
and Kazusa OMICS (KOMICS) database.198

The known tomato glycoalkaloid a-tomatine is used to illustrate how mass
spectrometry supports glycoalkaloid structure elucidation. LC/MS and CID spectra are
particularly usefirl because these are large molecules with complex NMR spectra in the
aliphatic region, and NMR-based structure elucidation requires prior purification of
milligram quantities and extensive spectrum interpretation. First, the accurate mass
measurement of [M+H]+ for the strongest glycoalkaloid signal was an even nominal mass
at m/z 1034.55, suggesting this molecule contained an odd number of nitrogen atoms.
Fragment ions generated at higher CID potentials appeared at m/z 902.50, 740.45, and
578.41, are consistent with consecutive neutral losses of anhydro forms of xylose (132
Da), xylose-glucose unit (132+162 Da) and xylose-glucose-glucose (132+l62+162 Da),
in agreement with the documented sequence of the oligosaccharide conjugate. Neutral
loss of another glucose (162 Da) would account for the fragment ion at m/z 416.36, which
corresponds to the characteristic fragment of the protonated aglycone ([tomatidine+H]+).

In addition to fragments mentioned above, a doubly-charged ion at m/z 528.78 is assigned

as [M+H+Na]2+ and other fragment ions observed at m/z 884.52, 722.43, 560.38 and

398.33 using the highest C1D potential (80 V) are in agreement with consecutive loss

glycosides plus H20 through a rearrangement process in the E ring,208 as described in

Figure 4-19. Lower mass fragment ions at m/z 273.22 and 255.21 are consistent with

165

cleavage of E-ring from the protonated aglycone [tomatidine+H]+ at m/z 416.36 with CID
voltage 55 V, which was consistent with earlier studies by Cataldi and colleagues.208 The
ion at m/z 161.13 can be assigned as the result of water loss and multiple cleavages in the
C-ring that can be explained by a charge driven mechanism, involving the generation of
resonance stabilized allylic carbocations209 (Figure 4-20).

Comparisons of mass spectra of other putative glycoalkaloids assist their structure
annotation. Figure 4-21 displays ESI mass spectra of a-tomatine and dehydrotomatine
using ESI positive mode with CID voltage 80 V. Another metabolite, assigned as
dehydrotomatine, gives a protonated molecule 2 Da lower (m/z 1032 [M+H]+) than
tomatine. All major fragment ions are also 2 Da lower in mass than the corresponding
fragment ions from a-tomatine, indicative that the unsaturation lies in the aglycone
portion and not in the carbohydrate portion. These results are consistent with the
difference of one unsaturation relative to tomatidine,208 and with a double bond at

position 5, 6 in the steroidal aglycone, as reported for tomato.‘98

166

100] 14.78

 

 

 

 

 

1?
"/0“ I41
‘ 1
1 11
0 . ~44 . 1 . - .gw. .
1O 20 3O 40
1001 14.29
‘ 1
. 11
0/0“ i1
. 11
1 11
. J.
0 4 . .14 4 1531 A 1 1 . ~ 4 . .
10 20 30 40 min

Figure 4-18 (Top) LC/MS Xle of [M+H]+ for a-tomatine m/z 1034.5 (RT = 14.78 min)
and (Bottom) dehydrotomatine m/z 1032.5 (RT = 14.29 min) from a leaf dip extract of
tomato (S. lycopersicum M82).

167

1032.54

 

 

 

 

 

 

1004
%1‘ 253.20 5516.38 882 9
4 .4
4 159 12 I 396. 32153734761576 397 8870.48 900.50
12.7121
1
0 4--2,1M1.LLJ11_1__~l1'_ILL__LHl,L 2.5L .IL ..730 {fifi 4.443444 .i,g~l~z._1__~-1 ._._-1-.--4 44 4~~
1001 1034.55
255-21 560.40
%1 161.13
‘ 1 416.35 57341 88450
398134 528-78 672 50 13¢ 902 51
1 1 1 1 2173.22 11 722. 44 740 45 :1 /
11,,11Hdfififizfizrm .1 4444 44444m2
200 400 600 800 1000 1200 1400

Figure 4-19 Mass spectra of a-tomatine (Top) and dehydrotomatine (Bottom) from
LC/MS analysis of an extract from S. chopersicum M82 using ESI positive mode with
CID potential 80 V.

168

Figure 4-20. A proposed fragmentation pathway for a-tomatine using ESI positive mode

based on literature reportszog’ 209

169

   

m/z:416

  
 

  

HO
H30;

CH

 

  
 
  

   
    

 

HO
H39: _
CH3‘\?
Po
H3C
HO
1 644.
H3O
m/z:255
H39: ; "‘\\\CH3 CH : if:
CH 4 N
N -H20 9H2
H3C 0 H2 —’ H3O
m/z:416
HO
HO
H3C
m/z:161

170

In contrast to S. lycopersicum M82, two chromatographic peaks suggestive of
additional isomers for both a-tomatine and dehydrotomatine are present in extracts of S.
pennellii LA0716 (Fig.4-21). These two isomeric peaks produce same fragment ions
using ESI positive mode, suggesting they are either stereoisomers or regioisomers.
Metabolic phenotypes of glycoalkaloids were also surveyed among more than 50
introgression lines of LA0716 >< M82. Chromatographic peaks corresponding to novel
glycoalkaloids were discovered in extracts of IL1-1 and 1-1-3 that were not present in
other ILs or M82, suggesting that a region at the top of chromosome 1 is responsible for
the metabolic difference. Two peaks with same exact mass as protonated
dehydrotomatine, at m/z 1032.56 were detected for IL1-1 and lLl-l-3 (Figure 4-22 top).
Based on fragment ions generated using a CID voltage 55 V, both yield losses of neutral
carbohydrate fragments at m/z 900.50, 738.44 and 576.39, and give the same mass for the
steroid aglycone at m/z 414.34. However, lower mass fragment ions attributed to
fragments of the aglycone are different after E-ring is cleaved. The early eluting isomer
(RT = 2.73 min) generates fragment ions (Figure 4-22 middle) at m/z 271.21 and 253.20,
but the later eluting isomer (RT = 3.00 min) is consistent with the known
dehydrotomatine metabolite, yielding fragment ions (Figure 4-22 bottom) 2 Da mass
greater at m/z 273.22, 255.21. The fragments from the early eluting isomer are identical
with the corresponding peaks from a-tomatine, which does not have any double bonds in
rings A-D. These observations suggest the early eluting isomer corresponds to tomatine,
but with an additional unsaturation, probably a double bond, on the F ring which contains
the nitrogen atom. A proposed pathway for formation of the fragments at m/z 273 and

255 is displayed in Figure 4-23. This is a novel metabolite with no previous literature

171

reported. Purification of this metabolite is underway by another student in Jones’ lab to

confirm the structure by NMR. If the metabolite is indeed the imine, it may reveal a view

toward the key biosynthetic steps in formation of the heterocyclic F ring, and the

association of this trait with chromosome 1 should steer research toward the genes in this

region.

100-

0/0 _

 

100—

 

vvvvvv

‘10 ” ""20

vavavava

' '30

min

Figure 4-21 LC/MS XICs of an extract of S. pennellii LAO716 using ESI in positive ion
mode corresponding to protonated molecules of (Top) (it-tomatine at m/z 1034.5 and

(Bottom) dehydrotomatine at m/z 1032.5.

172

100

 

 

 

.1

2.73

 

 

 

 

 

 

 

 

%1 3.00
11
1 1
\1/ \\ min
0 11111111111111111 "2' ' r'fifi'é ****** 41 ' 's“
576.29 1032.54
100
%
. 255.21 414 34 1015 53
161.13 '
0 -21 211.121 558 391 . ..... 3.791.493??? 1. z .2. - -2.
1001 576.29
%1 527.76
1 414.34 852. 481015 115332 54
. 2 .
04159.12 1 271.2111 1 1/ 720 43738 441 900. 5011 /
- ,-.~--,-4-L.,141--4.,444,341. -41- 4, ~41-~~~4 4mz
200 400 600 800 1000 1200 1400

Figure 4-22 (Top) LC/MS XIC of m/z 1032.5 from ILl-l using ESI positive mode and
CID voltage 10 V. ESI spectrum (Middle) of first peak (RT = 2.73 min) corresponding to
dehydrotomatine isomer and ESI spectrum (Bottom) of second peak (RT = 3.00 min)
corresponding to dehydrotomatine (double bond in 5,6-position) using CID 80 V.

173

 

H3C
1 H30. flee“
H3C \c'jHN/
H3C
m/z:273 6:33:20

m/z2255
Figure 4-23 A proposed fragmentation pathway for first eluting dehydrotomatine isomer

peak (RT = 2.73 min) from IL 1-1 using ESI positive mode with CID voltage 80 V.
Fragment ions at m/z 273 and 255 indicate that the position of the double bond is not on
rings A-D.

In addition, another putative glycoalkaloid at m/z 1050.55 ([M+H]+) was detected in
extracts of ILl—l and 1-1-3. For this metabolite, the aglycone fragment is observed at m/z
432.35, suggesting one hydroxyl group based on the 16 Da increase in mass relative to
the corresponding fragment from (at-tomatine. Spectrum generated with higher CID
potentials revealed associated fragment ions at m/z 273.23 and 255.22, as are found also
for a—tomatine, attributed to cleavage of E-ring. These fragment ions suggest that the
hydroxyl group is on the heterocyclic F ring of the metabolite, though an alternative
explanation would involve a tautomeric form of a steroidal amino aldehyde. Figure 4-24
displays XIC of m/z 1050.56 ([M+H]+) and ESI spectrum for this metabolite using ESI

positive mode with CID potential 80 V. A fragmentation pathway that accounts for

174

formation of the fragments at m/z 273 and 255 from the protonated aglycone fragment is

proposed in Figure 4-25.

 

 

 

 

 

 

100 ~ 2.8611l
O/o _.
1
. A
0 I ' I ' f r WT , ' f '/ 'V* F r I * w I r I min
1 2 3 4
100- 576.49
1032.54
%1
4 414.34 594.40
255.21 32.35 1050.55
4161.13 273 22 527.76 7 2.45
l L - k I “738.44 900.50 918.51
r-,f-.‘_fr-,l—. 1&1? -4L $14.11;.[i-el.;LkJ,Affi.1.§L...A,- 511,..-vaivv-T ...,..-v m/z
200 400 600 800 1000 1200 1400

Figure 4-24 LC/MS XIC of m/z 1050 corresponding to a putative hydroxytomatine

metabolite from ILl-l using ESI positive mode. (Bottom) ESI spectrum for

hydroxyltomatine from IL] -I using ESI positive mode with CID potential of 80 V.

175

 

m/z:273
m/z:255

Figure 4-25 A proposed fragmentation pathway for hydroxytomatine from IL l-l using
ESI positive mode.

To summarize the results from above, characteristics of fragmentation for
glycoalkaloids208 were used to assign this class of metabolites using multiplexed CID.
The carbohydrate side chain is easily cleaved to leave fragments corresponding to the
protonated aglycone. Lower mass fragment ions caused by cleavage of E-ring appear and

are useful for diagnosing structural modifications that occur on the F-ring.

4.3.3 Structural annotation of terpenoid metabolites
Although volatile monoterpenes (10 carbon atoms) and sesquiterpenes ( 15 carbon
atoms) impart much of the characteristic “tomato” smell to S lycopersicum leaves.

characterization of less volatile terpenoids has been less explored in the genus Solarium/(r.

176

Oxidized and more polar terpenoids are less common, and are usually not observed in
GC/MS analyses of tomato or its relatives.
One exception to this rule is the wild tomato relative, known now as S. habrochaites

LA1777. This plant produces three isomeric sesquiterpene acids: a-santalenoic acid and
two isomers of bergamoten-lZ-oic acid, all of which have the formula C15H2202 and

molecular mass of 234 Da. Analyses of this accession using LC/MS revealed three major
isomers of sesquiterpene acid at m/z 233.15 (Figure 4-26) in extracts of S. habrochaites
LA 1777 and their assignment was based on exact mass, fragments and prior literature

'97 One unusual feature was observed in the product ion MS/MS spectrum for

report.
deprotonated ion of a-santalenoic acid and is displayed in Figure 4-27. Products derived
from [M-H]' included an abundant fragment ion at m/z 98 that is a radical anion. Such
odd electron fragments are relatively unusual products from CID of aliphatic

carboxylates, but stable radical due to the near double bond. A proposed fragmentation

pathway is shown in Figure 4-28.

100— m
1

1/

 

\
\
\1 .
i _ O
,1 b ~._ ‘\-/—\_
f . , e if.-..-.~.r.._.-- ,ffi- ,-+fi.,.-. , --- |

 

 

Figure 4-26. LC/MS XIC of sesquiterpene acid at m/z 233 from S. habrochaites using

ESI negative mode.

177

 

 

 

 

 

99
%
‘ 98
‘ 67
55 L
0 i .2 r. . & W2
10 50 100 150 200 250

Figure 4-27 Product ion MS/MS spectrum of sesquiterpene acid at m/z 233 from S.
habrochaites LA1777 using ESI negative mode with collision energy 45eV using flow

injection analysis.

178

god; gt.

m/z: 233.15 m/z:98.04
o
m/z: 99.05 m/z: 55.06

Figure 4-28 A proposed fragmentation pathway for a-santalenoic acid from S.

habrochaites using ESI negative mode.

Several additional terpenoid acids from S. habrochaites LA1777 were also observed
and annotated based on accurate mass measurements. These include three peaks
annotated as oxidized sesquiterene acids and a diterpene acid. (Figure 4-29). For the
former, the mass difference between these and the sesquiterpene acids (m/z 249.15 and
m/z 233.15) was A 16.00 Da, suggestative of one oxygen atom, so m/z 249.15 was
tentatively assigned as hydroxylated variants of the sesquiterpene acids. The three
chromatographic peaks may correspond to the three isomeric sesquiterpene acids, with

each modified by addition of an oxygen atom. For the diterpene acid at m/z 319.22, the

mass difference relative to m/z 233.15 was A8602 Da, corresponding to C5H100.

Though the negative mode CID spectra did not yield sufficient information to determine

179

more structural details, the difference in formula would be consistent with addition of an

isoprene unit to sesquiterpene acid plus H20.

 

 

1001
01
°\..
.-. I ' Y 10:001‘777720700‘7777303001 I V 10:00
1003.
1 .1
I °\°f 111 1*
‘ 1.511 1 11
, 15.1161 A A "'11
OWKII ' 'FMYVJWIJV t ' ”W Tlme
10.00 20.00 30.00 40.00

Figure 4-29 (Top) LC/MS XICs of diterpene acid at m/z 319 and (Bottom)
hydroxysesquiterpene acid at m/z 249 from S. habrochaites LA1777 using ESI negative
mode with CID potential of 10 V.

In addition to sesquiterpene acids and diterpene acids, several metabolite ions were
were discovered in extracts of S. habrochaites LA1777, with the analyses relying heavily
on accurate mass measurements. Sesterterpenoids are a group of metabolites containing
25 carbon atoms (5 isoprene units) in core structures that are rarely found in plants,
though several have been isolated from marine sponges. The few known examples
exhibit a number of interesting pharmaceutical propertiesm‘z” One prominent signal

detected in extracts of accession LA1777 showed a putative [M-H]° peak at m/z 649.32

180

was used as an example to show characterization of these metabolites. First, three
possible elemental formula were obtained based on accurate mass measurement. Then,
fragment ions are generated for one isomer of m/z 649.32 (RT = 16.83 min) with CID
voltage 55 V (Figure 4-30). A product ion MS/MS spectrum was generated for this
metabolite to confirm that the major fragments observed using multiplexed CID were

also observed in products of m/z 649. For both kinds of spectra, major fragments were
observed at m/z 605.32, corresponding to neutral loss of C02. This neutral loss occurred
at low CID potentials as would be consistent of a carboxylic acid group that undergoes
decarboxylation easily such as a B-keto acid. Mass difference between m/z 605.32 and

m/z 563.31 is A4202, consistent with neutral loss of ketene (CH2=C=O). These findings

are consistent with a malonic ester group that undergoes successive losses of C02 and

ketene. The next prominent fragment is observed at m/z 401.26, corresponding to

subsequent neutral loss of anhydrohexose (Al62.05). Exact mass measurements

suggested an elemental formula for m/z 401.26 to be C25H37O4'. No further

fragmentation of this ion was observed, therefore this is assigned as the core of this

terpenoid metabolite. Several mass peaks at m/z 811, 729, 691, 567, 581, 499 and 487

were observed to consecutively loss A44.01 (C00) and then A42.02 (CH2CO)

corresponding to malonic ester in addition to m/z 649. Neutral loss scan of A 86 was
generated for S. habrochaites LA1777 to support this result (Figure 4-31). Based on exact
mass measurements and their fragmentation pattern, they were all assigned as
sesterterpene acid derivatives. For example, m/z 811.38, mass difference with m/z 649.32

is A 162.0650, corresponding to hexose. The retention time of mass peak at m/z 811.38

181

also elute earlier (1.05 min) than mass peak at m/z 649.32 since the polarity increased due
to one more hexose. Elucidation of those metabolite structures is important for exploring
the metabolite pathway of terpene acids. Therefore, much effort was put into the
purification of these metabolites by preparative HPLC in our lab now. Although exact
structures can not determined so far, the classes of metabolites at least can be assigned

based on characteristic fragments and exact mass with the aid of multiplexed CID.

182

 

 

 

 

 

 

 

 

 

 

 

100 - 16.83
% 4 1
1
. . 11
O .2- -“QDL ...... r....,%..,.......min
1O 20 3O 40
5.32
100~ 60
°/ ~ 401.26
0 563.31
161.04 443.26
0 WA, .1 ,L ”h- .2. - , ,. , , ,. 1.1.1 ..+- 1.44994. m/z
200 300 400 500 600

Figure 4-30 (Top) LC/MS XICs of m/z 649 [M-H]' of sesterterpene metabolite using ESI
negative mode. (Bottom) ESI spectrum of m/z 649 (RT = 16.83 min) from S.
habrochaites LA1777 using ESI negative mode with CID potential 55 V.

183

567

 

 

 

 

 

 

 

100— - 649
96- 499 581
1
497 811
487 729
501 521 582 650 691
o. .. .-1 ..1, “A , .... -. ...;1692 1 - .139? ..m/z
500 600 700 800

Figure 4—31 Constant neutral loss (86 Da) spectrum from an extract S. habrochaites

LA1777 using flow injection analysis and ESI negative mode.

4.4 Conclusions

These efforts have annotated several novel secondary metabolites from trichomes
within the genus Solanum using LC/TOF with multiplexed CID and LC/MS/MS
performed on a QTof instrument. These discoveries have included findings of methylated
variants of the flavonoid myricetin, novel glycoalkaloids with F-ring modifications that
point to potential biosynthetic intermediates, and new terpenoid acids including malonate
esters of sesterterpene glycosides, which constitute a family of specialized metabolites
rarely described within the plant kingdom. The observation of new metabolites in several
genotypes of tomato and its relatives suggest that these plants may represent a rich

resource for mining of genes useful as biological catalysts.

184

Chapter 5. Fast LC/time-of flight mass spectrometry for screening

metabolic phenotypes for tomato chromosomal substitution lines

185

5.1 Introduction

Glandular secreting trichomes (GSTs) are epidermal protuberances that are found on
a wide variety of plant speciesm‘213 These structures typically manufacture relatively
large amounts of specialized (secondary) metabolites, and either store these molecules in
an extracellular space or deposit them on the epidermal surface, sometimes in amounts
sufficient to cause stickiness to the touch. Trichome compounds such as those in basil,”4
mint”5 and hops216 are responsible for their smell and taste. Other plants manufacture
compounds of medicinal importance such as the anti-nausea agent tetrahydrocannabinol
from Cannabis and antimalarial artemisinin from wonnwood. The combination of active
biosynthesis, species-specificity of metabolite accumulation and ease of purification of
trichomes make them an excellent system for specialized metabolite pathway elucidation
and identification of genes encoding the pathway enzymesm‘213 The combination of
cDNA sequencing and metabolite analysis of SGTs to discover candidates for in vitro
enzyme analysis has been very effective in elucidation of specialized metabolic
pathways.2”’56‘2'8 However, this molecular biochemistry approach can be very
challenging or impossible when seeking enzymes that have structures unrelated to known
activities or are members of multigene families. Forward genetics (screening for an
altered phenotype in a collection of genetically diverse individuals) is a complementary
approach to discovery of enzymes, transporters or regulators of poorly defined
biosynthetic pathways or primary sequences not recognizable by homology. One of the
advantages to this approach is that, by directly screening genetic variants for changes in
metabolite levels, it is possible to identify genes encoding completely novel products.2l9

Therefore, we applied forward genetics into our research. A series of nearly isogenic

186

lines (ILs) (Figure 5-1) are chosen as targets for study widespread natural diversity.
Among those lines, the individual chromosomal segments from the wild species S.
pennellii (accession LAO716) were introduced into cultivated tomato (S. lycopersicum)

0 These lines were used for

M82 by recurrent backcrossing and genotyping.22
identification or verification of genes affecting a variety of phenotypes, including
differences in metabolites.‘7"7"'223 The phenotypic and genetic study of ILs will help with
the identification of diverse genetic variants and understanding the gene function. Tomato
trichome research can take advantage of the available genetic resources to characterize
specialized metabolites and to identify loci/genes resulting in the production of the
compounds. Exploiting the genetic basis of the chemical phenotype in trichomes will
enable further understanding of what chemicals are produced, what genes are involved in
the production of the compounds and how they are regulated. In addition, it will also
provide important knowledge for plant metabolic engineering to improve plant defense
and human nutrition by regulating the production of specific metabolites. Therefore,

successful analysis of the complex network of plant metabolism requires analytical

method able to record information on as many metabolites as possible.

187

IL bin mapping

Donor parent

 

 

Recurrent parent

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

lL1-1rhl‘I I I I 1 I I I
IL1-2 :1 fri i i i i i
lL1-3 1 imititgiigi i i j i
lL1-4 f i i iffi.;“'"ii‘i:i‘:'.ifi'i’.i i i
lL1-5 . i i i i i’..j.._.:.i......_._:i
IL1-6 f z z ; ; 2 :

BIN

1A: 18 i1Ci1Di1E i1Fi1Gi1Hi 1| 1J

Figure 5-1. Incorporation of genomic DNA from S. pennellii donor parent (red) and
tomato recurrent parent (black) into DNA for introgression lines. (adapted from

http://zamir.sgn.cornell.edu/ Qtl/il_story.htm )

Currently, however, no single technique can provide detection of all metabolites due
to the complexity of chemical compounds.224 Over the past decade, several methods
suitable for large-scale analysis and comparison of metabolites in plant extracts have

been established,225 including gas chromatography coupled to mass spectrometry (GC-

224.227.228

MS),5’24’56‘226 direct flow injection—mass spectrometry (DFI-MS), liquid

103229-233

chromatography-mass spectrometry (LC-MS), capillary electrophoresis-mass

spectrometry (CE-MSW“35 and NMR technologies.48‘235 LC-MS-based approaches are

expected to be of particular importance in plants, owing to the highly rich biochemistry

188

of plants, which covers many semi-polar compounds, which can be separated and
detected by LC-MS approaches.236 Here, in our study, a fast LC/time of flight mass
spectrometry with multiplexed collision CID method is developed for discovery of plant
phenotypes. The multiplexed CID system switches voltages from low to high and
generates molecular and fragments in a single analysis. Exact mass measurements for all
of ions using this approach assist to limit elemental formula and associate fragments with
molecular ions. Another advantage of the method is that it is fast and reliable. For
instance, ~300 plants can be screened within 30 h with reproducible and accurate results.
Also, pre-existing sofiware— Quanlynx and Markerlynx- in Waters Masslynx 4.1 is
helpful for data processing including peak integration and statistical tests such as
Principal Component Analysis (PCA). Five specific metabolic phenotypes are discovered
among 65 ILs using this approach. In summary, this high-throughput approach offers a

great opportunity for large-scale chemical screening.
5.2 Experimental Method

5.2.1 Plant Growth Conditions. Tomato seed, S. lycopersicum cv. M82 and S. pennellii
LAO716, was obtained from the Tomato Genetic Resource Center (http:

220

//tgrc/ucdavis.edu). S. pennellii ILs were obtained from Dr. Dani Zamir (Hebrew

University Faculty of Agriculture, Rehovot, Israel). Plant seedlings were grown in Jiffy

peat pots (Hummert International) for 3 weeks in a growth chamber maintained for 16 h

at 28°C in the light (300 uE/mz/s, mixed cool white and incandescent light bulbs) and 8 h

at 20°C in the dark.

189

5.2.2 Plant Extractions For analysis of total trichome metabolite, leaflets from the

second leaf after the newly emerging leaf of 3-week-old plants were dipped with gentle
rocking for 1 min in 750 pl of isopropanol : CH3CN : H20 (3:3:2 v/v/v) containing 10

11M of internal standard propyl-4-hydroxybenzoate.

5.2.3 Chromatography Reverse-phase analysis was performed on a Shimadzu
(Columbia, MD) LC-20AD HPLC system using a fused core Ascentis Express C18
column, 2.1 X 50 mm with 2.7 pm particles. The column was maintained at 30 °C. A
steep gradient was executed for rapid metabolite screening based on solvent 0.15%
formic acid in MilliQ water (A) and Methanol (B) with analysis time of 5 min/sample.
Gradient profile consisted of: initial conditions 10%B; linear gradient to 60% B (2 min);

100% B (3 min); hold at 100% B until 4 min.; return to 10% B (5 min). Flow rate was 0.4
. -l
ml°m1n .

5.2.4 Mass Spectrometry A Waters LCT Premier mass spectrometer (Milford, MA)
coupled to a Shimadzu LC-20AD HPLC system and SIL-SOOO auto sampler were used.
Ionization mode: electrospray negative mode; capillary voltage: 3.00 kV; Source

temperature: 90 °C; desolvation temperature: 300 °C; nebulizer nitrogen flow rate: 20

L'h—l; desolvation nitrogen gas flow rate: 300 L-h_l; mass range: 50-1500 Da. Analyses

were performed using rapid switching of the Aperture 1 voltage in the ion transit region
of the mass spectrometer, providing quasisimultaneous generation of spectra under
fragmenting and nonfragmenting conditions. Aperture 1 voltage were 10, 25, 40, 55, 80
V for five functions. Data were acquired in centroid format with ‘dynamic range
enhancement’ (DRE) enabled for a wide dynamic range for quantification. Peak areas of

extracted ion chromatograms for characteristic masses of each metabolite and the internal

190

standard were integrated using Waters QuanLynx software. Measured levels of secondary

metabolites were normalized to the dried weight of the tissue used for each extraction.

5.2.5 Chemometric Data Analysis The ESI' raw data were analyzed by the Markerlynx
applications manager version 4.1 (Waters, Manchester, UK). The parameter were RT
range 0-5 min, mass range 100-1500 Da, mass tolerance 20 ppm, internal standard
detection parameters were deselected for peak retention time alignment, isotopic peaks
were excluded for analysis, noise elimination level was set at 50, minimal intensity was
set to 1% of base peak intensity, maximum mass per RT was set at 50 and, finally, RT
tolerance was set at 0.2 min. No specific mass or adduct ions were excluded. Ions of
different samples were considerate to be the same ion when they demonstrated the same
RT and m/z value within tolerance above. If the peak was not detected in the sample, the
ion intensity was documented as zero in the final data table. A list of the intensity of the
peaks detected was generated for all the samples, using RT and m/z data pairs as the
identifier of each peak. The resulting three-data comprising of peak number (RT-m/z
pair), sample name, and ion intensity were analyzed by unsupervised method principal
component analysis (PCA) within the Markerlynx software. Following that, the
supervised method orthogonal projection to latent structures-discriminant analysis (OPLS)

analysis is carried out using SIMCA-P.

5.3 Results and Discussion

5.3.1. Analysis of Chromosome Substitution Lines Using LC/T OF MS 1

191

Initial evaluation of rapid methods for screening metabolic phenotypes focused on
determining whether physical separation of metabolites upstream of the mass

837' 228’ 2374'“ offers the most rapid option

spectrometer was desirable. Direct injection M
for mass spectrometric analyses, but suffers from suppression of ionization,37 limited
ability to distinguish of chemical isomeric metabolites, and complications that arise from
in-source formation of adduct, oligomer, and fragment ions. Chromatographic separation
offers a mechanism to resolve these features, but usually involves long analysis times (30
min or more) per sample. Recent improvements in LC column technologies provide
enhanced chromatographic resolution through use of smaller particle sizes and
superficially porous particles?”243 Thus, a fast LC separation employing a 5 min/sample
analysis time coupled with exact mass measurements from time-of-flight mass
spectrometry was developed for nontargeted metabolite profiling.

Fast S-minute LC/TOF MS method was compared with a slower 43-min method on a
longer column using a leaf dip extract of metabolites from of IL 8-1 (Figure 5-2). The
major observed metabolites, including malic acid, quinic acid, chlorogenic acid,
alkaloidglycosides and several acylsugars were resolved using both methods. The slower
separation yielded chromatographic peak widths about 4-fold broader (20-30 s vs. 6 s)
than the fast separation. Extracted ion chromatograms (XICs) for some specific
metabolites were also compared between these two techniques to assess separation of
isomeric metabolites. We generated 11 different XICs corresponding to formate adduct
ions of acylsugars differing in molecular mass owing to different fatty acyl groups.

(Figure 5—3). Fast HPLC separation resolved as many isobaric acylsugar metabolite peaks

as the slower method. For example, three peaks at m/z 695 were resolved by fast HPLC

192

and the slower HPLC separations. One peak is assigned as a tetraacylsucrose with 18
fatty acyl carbon atoms, and the other two are triacylsucroses with 19 fatty acyl carbons.
These assignments could be made owing to the high mass accuracy measurements which
distinguish the tetraester from the triester from the 36.4 milliDalton difference in

molecular masses.

193

Figure 5-2. Total ion chromatograms obtained from leaf dip extraction of M82 (A) and
IL8-1-1 (B) with a 5 min gradient using an Ascentis Express C18 fused core column, 2.1
x 50 mm; 2.7 pm (Top of A and B) compared to a 43 min gradient with a Thermo
BetaBasic C18 column, 1 x 150 mm, 3 pm (Bottom of A and B) in ESI negative mode. A:

Malic acid and quinic acid; B: Chlorogenic acid; C: Rutin; D: Tomatine; E-L: Acylsugars.

I94

Acvlsuqars

    
 

glycoalkaloid
orqanic acids flavonoid E
C

 

 

100 200 3.00 400 500
100} F
:11 E G H
A c °
0%

 

 

 

5.00 10.00 15.00 20.00 25.00 30.00 35.00 40.00

195

 

1 S 4:24 I. 1 m/z 779 6 4:24 1

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

l S 4323 1 1 m/z 765 6 4:23 _

L . . 6 4:22 -1. 4 m/z 751 6 4:22 I

1 6 4:21 1 6 3:22 m/z 737 6 4:21 1 1 6 3:22

I 1 6 3:21 m/z 723 6 3:21

L 1 6 3:20 1 m/z 709 .' 6 3:20

1 64:16 .- L 11 63:19 1 m/z695 64:16 1 63:19

1 S 4:17 .1. 1 m/z 681 S 4:17 1

1 S 4316 1 1 m/z 667 6 4:16 1

1 6 4:15 1 . 1 m/z 653 6 4:15

1 84:14 1 - 1 1 ””639 84:14 I

0 43 0 5 .
mm mm

Figure 5-3 Extracted ion chromatograms (XICs) of [M+HCOO]‘ for leaf dip extract of
IL 8-1-1 showing peaks corresponding to sucrose triesters or tetraesters from m/z 639 to
m/z 779 (homologs differing by 14 D6 in molecular mass owing to additional methylene
group in the acid moieties). Left: 43 min gradient in ESI negative mode. Right: 5 min

gradient in ESI negative mode.

196

We have frequently used the fast LC/MS method for screening several hundred
metabolite extracts from a single sample set. For one set of experiments, 296 samples
including blanks were screened over one weekend using this fast HPLC/MS method.
Automated peak detection, integration, and alignment was performed, generating an
extensive data matrix of analytical signals (annotated as retention time-mass tags), with
each represented as the value for the integrated peak area for the specific mass and
chromatographic retention time (Figure 5-4). In a typical analysis, this data matrix
compiles information from more than 1500 analytical signals after automated peak
alignment and integration with Markerlynx software. This was not equivalent to 1500
individual analytes as it was confirmed some of the RT-m/z pairs were adducts or minor
fragment ions. Based on our analysis of mass spectra of a single peak and considering the
number of adducts and fragments ions associated with each compound (approximately 5-
7 ions), it was predicted that the total number of metabolites close to 200 or more, which
means one metabolite/1.5 second by this approach.

Data analysis and visualization was performed on normalized, mean-centered and
Pareto-scaled peak areas, using unsupervised principal component analysis (PCA). The
results were displayed as loadings plots for the first two principal components, in which
the loadings of each signal on the latent variables (principal components) are displayed.
The sum of the loadings for each sample extract is then calculated, and these values are
displayed as 3 scores plot, where each symbol represents a distinct sample. Clustering of
samples based on metabolite profiles may be evident from the scores plot. PCA scores
plot generated from LC/MS metabolite data of recurrent parent M82 and 65 [Ls is

displayed in Figure 5-5. Two distinct sample groups emerge, with lLs 1-3, 1-4, 8-1, 8-1-

197

1, 11-3, and 5-3 separated from all other ILs and M82. Further examination suggested the
outlying samples could be divided into four clusters: ILs 8-1 and 8-1-1; IL “-3, 5-3; l-3,
1-4 and M82 (Figure 5-6, Left). The corresponding loadings plot (Figure 5-6, Right)
aided identification of the analytical signals contributing most to the measured variance,
and steered efforts toward metabolite identification of these metabolites. Once
metabolites are recognized that distinguish specific genotypes, these compounds must be

identified to aid attribution of functions to the responsible genes.

198

Figure 5-4. Metabolite data extracted using MarkerLynx software from LC/MS analyses
showing a subset of 5 introgression line samples. The software first generates extracted
ion chromatograms (XICs) for each mass, then detects and integrates the peak intensity,
and aligns the peaks according to retention time. The metabolite identifications are
assigned in the first column. Values in the various columns correspond to extracted ion

peak areas for the retention times and masses listed in the third and fourth columns.

199

 

-33E£+§

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

8%. 3 :35. 333m 3333 33.2 333.83 2 :2 3; _.3 gm fishnvvmqvm
- FE
SE33 $32 333M: 333” 88.8 52.83 3:3 .8333. 3883 amoéaé
88.3. 333 3:33 838 33.33 35 33.3 $-23 53m
.358?ng
35.33 333 332.33 $5.8 3333 23.3 a: 33.3 5 :.3v 3%
:33 833.33 M38. : o 3:; 8338 3933 3238032 3.3.6 2%
-Houmnborzg
3232 3332 3332 $3..on $32 33:: £83 3 3:53 Bumeoceafia
S 2.2% 3:38 @3333 33.38 $03.23 3. av :83 -.oséoté 32 32
”$33. 8338 803.5. 33.33 82.83 3228 33 -.oafiorE 3.3.6 8%
$3.83 ”$5.33 3323 8333 28.33 3:38 «3:3 338635 3.3V 3%
£338 383% 3:133 3838 $8. :3 8233 $9.3 Agassiz. 23m
$3.5” 395.8 0333 33.3 E32: 333$ 33 -Eé:8+§ 313V 333m
-Eafioié
:33: $2.62 8.52 come: 3.6: $33.33 $33 _ Eéfiu 3825
:83» 308: $3.83 $8.3 £38 3922 333 -waéotzu 3.3.9 2%
3%. SS 3%: SSS 23.3% $3.39: 393: @333 fiascoté 3.2.3.3 33m
w 33% RE. _ a $3. :3 3.8.9...” 3. z 3% 33 z .38 83.3 -235 55m
3332 59%: 35.33.: 33:83 33.8: $32: 33.3 33.88.45 2:363
$33.33 $3.88 3:133 3.33 8332 3328 33% 33.83% 3.3.3 :3
fig
ES oEmum—musowom 538058 8208302
.-m 3 73. 3 E. 3 d
383 883 $83 $83 883 3 295m

 

 

 

 

 

 

 

 

 

200

ZOd

Figure 5-5 Principal component analysis (PCA) on 66 ILS plus recurrent parent M82
(total 276 samples) obtained using Pareto scaling with mean centering. PCA scores plot;
Symbols: ILS 1-3, 1-4 (0); 5-3, 11-3 (A); 8-1, 8-1-1 (I); Squares (9) represent
recurrent parent M82 (n=31); Cross (61>) represent other 60 ILs (n = 3 or 4 for each IL).

 

 

 

 

 

 

60
o O
40 4 ' .2 0 '3“?
° .- ' % 9;.9'; «“p
20 - J 3 J; 0.0 :°
0 'Ih °° 2" ° ‘9')" '
o o i
0 -1 . | : .0 ‘g. 0:... o. 5 O
. ° .°-."
-20 d ‘ I °2?‘°; . o o
A O 0 1 o .
40- e - . ' ° ‘° °
_ - lL1-3/1-4 . °° T“
'50 - lL8-1/8-1-1 ° .
‘ IL5-3/11-3 °
'80 ‘ . M82
° h
-100 0't er IFS I I I I 1 I I T
-140 -120 -100 -80 -60 40 -2o 0 20 4o 60 80

PC1

201

Figure 5-6 (Left) PCA scores plot from LC/TOF MS metabolite data represent three
clusters of ILs separated from recurrent parent M82 using Pareto scaling with mean
centering. The clusters are 1-3, 1-4 (0); 5-3, 11-3 (V); 8-1, 8-1-1 (0) and M82 (A). (Right)
PCA loadings plot showed corresponding analytical signals (retention time-mass pairs)

contributing to the difference among clusters in PCA scores plot.

202

 

 

 

 

 

 

 

 

 
        
 

 

 

 

80
60~ .
O
40- ’
0
20- v AAmfiAA A
v V
N 0— V, AAQQAM
Q A
‘1 -20- v
-40-
-50— o o |L1-3/1-4
o 00 0 IL 8-1/8-1-1
-80- %g v IL 5-3/11-3
A M82
'100 I I l I I l
-80 -60 -40 -20 0 20 40 60 80
PC1
0.4
. 3.34_639. 1 . 33433142
0.3a
0.24
0.1~ 2.47_655.15, ' .3_42_1317.6o
' . 7 .4 .
8 0.0- 2.47 609.15 ._.- 392—7 9 3 .
IL . . . ° 3.43_681.32
-0.1- ’
2.58_1078.55-
-0.2-
3.81_765.41-
'0-3‘ 3.25_653.30°
- 3.34_667.31
'0.4 l I I I
-0.4 -0.2 0.0 0.2 0.4 0.6
P01

203

5.3.1.1 ILs 1-3 and 1-4 are missing an acetyl group on major acyl sucrose
metabolites

Two metabolites, of m/z 639 and 737, contributed most to distinguishing ILS 1-3 and
1-4 as judged by their positions near the top of the loadings plot. These contributions are
clearly observed in the LC/MS chromatograms for M82 and IL1-3 shown in Figure 5-7.
Metabolites at m/z 639.28 (RT 3.35 min) and 737.40 (RT 3.84 min) are more abundant in
IL1-3, but were largely less from M82, which displayed abundant metabolites at m/z
681.30 (RT 3.42 min) and m/z 779.41 (RT 3.92 min) that were largely absent from lLl—3.
ILs 1-3 and 1-4 stands out as having substantially greater abundance of this metabolite
than the parent line M82 and all other ILs. Multiplexed CID approach was performed and
yielded ion masses and mass accuracy within 10 ppm of true values for all molecular and
fragment ions that have adequate ion statistics for mass measurement. Multiplexed CID
provided spectra with minimal fragmentation at the lowest CID voltage, and
progressively more extensive fragmentation at higher potentials. Processing of data
generated under different quasi-simultaneous instrument settings aided recognition of
substructures based on structure-specific fragment ions and assignments of putative
metabolite structures. The great aspect of using multiplexed CID lies in it minimized
need for doing MS/MS, which makes this fast LC/MS even more compelling.

S 3:15 from IL1-3 and S 4:17 from S. lycopersicum M82 were used to illustrate how
to annotate metabolite structures. The annotation of metabolite structures were described
in detail in chapter 3. Mass spectra of those two metabolites with CID voltage 55 V are
described in Figure 5-8. For mass spectra of S 4:17 from S. lycopersicum M82, raising

CID potential to 55 V generated numerous ions observed at coeluting retention time with

204

593.28 (losses of C2), 509.22 (loss of C5 plus C2), 425.17 (loss of two C5 plus C2),
341.11 (loss of three C5 plus C2) due to loss of fatty acyl groups as neutral fatty ketene.
Fatty anion at m/z 101.06 corresponding to fatty anion of C5 fatty acid supported S 4:17
had C5 fatty acyl groups. Thus, acylsugar with formate adduct ion at m/z 681.29 was
determined as a sucrose tetraester (S 4:17) with one C2 and three C5 fatty acyl groups.
The structure of this metabolite is supported by NMR data (see detail, Table 6-1).
Compared to formate adduct ion at m/z 681.30 for S 4:17, ion at m/z 639.28 for S 3:15
has fragments at 509.22 (loss of C5), 425.17 (loss of two C5), 341.11 (loss of three C5)
corresponding to consecutive loss of neutral fatty ketenes and ion at m/z 101.06 (fatty
anion of C5). Therefore, acylsugar with formate adduct ion at m/z 639.28 is sucrose
tetraester (S 5:15), with one less acetate group compared to ion at m/z 681.30. Similar
with these, acylsugars with ions at m/z 737.40 and m/z 779.41 were assigned as S 3:22
(5,5,12) and S 4:24 (2,5,5,12). The result indicated IL1-3 and 1-4 were two specific traits
with much higher abundance of sucrose triesters, with one less acyl group compared to
tetraesters. Therefore, this is consistent with the hypothesis that this introgression region

is missing a gene necessary for normal acetyltransferase activity.

205

S 4:17 (2,555)

i
\ s 4:24 (255,12)

U J 10%

1 r;—1——r—Y——rr1v77 1W fij’fivffirTY

2 00 3. 00 4.00
S 3:15 (5,5,5)
S 3:22 (5,5,12)

o‘ \
°\“* ‘ ‘ [L
l J\/ .
j /\\\\ ’,\m_/\J\H/ \L/qkfl- \_ f fw\\

_.,A\

100

 

 

 

 

\‘ _ . __

 

 

OL—L‘L—LLLLL L—H ..LmnnTL LT vamTime

1 00 2 00 3.00 V 34.00

Figure 5-7 Total ion LC/MS chromatograms obtained from leaf dip of M82 (Top) and IL
1-3 (Bottom) using ESI negative mode. The peaks for most variance between two lines

are labeled.

[M-H-csr [M+HCOO]'

 

 

 

 

 

 

 

100 509 539
[M'Hl'
. OH OH _
%~ - [M-H-205] 593
[M- --H -305]
C51fatty acid anion 425 p 629 ..
\ 323V 407 '
0 305 341 I , 49} 555 . 1
Mo [M+Hcool’
10"} igfiL %OQK 681
l M- --H 02-05"
. OH OH [ l ..
05 [M- -H- -cz-205] [M'H‘Czl
CS fatty acid anion [M- H- CZ— -305] ¢ 509 [M-H]'671 '
510551 593
0 101 323.31” 407425 457 l I 635 I
100' ' ’200' ' '300'_' '400' ' ’500' ' '600' ' 700””

Figure 5-8 (Top) Mass spectrum of peak at RT = 3.35 min in IL 1-3 and (Bottom) Mass
spectrum of peak at RT = 3.42 min from S. lycopersicum M82 with CID potential 55 V.

206

5.3.1.2 Two introgression lines with lower total acylsugars

Results from screening the ILs illustrate that the rapid LC-TOF MS method can
reveal quantitative metabolite changes. In contrast to the effect of the S. pennellii
chromosomal regions 1-3 and 1-4 and 8-1-1 on acylsugar substitution, two other ILs were
found to cause an unexpected change in total acylsugar levels. ILs 5-3 and 11-3 were
found to have changes in the proportions of acylsugars relative to tomatine (m/z 1078)
and rutin (m/z 609) from PCA data. Quanlynx Software in Masslynx 4.1 was performed to
quantify the total amounts of detectable acylsugars among all [LS and M82. The nominal
mass, retention time of corresponding acylsugars were edited into the. quantification
method (Table 5-1) and use chemical compound propyl-4-hydroxybenzoate (synthesized)
as internal standard. Total peak areas of observable acylsugars were normalized to
internal standard and dry leaf weight. ILs 5-3 and 11-3 showed the lowest amounts of
acylsugars (Figure 5-9) compared to other ILs and M82, which supported the result from
PCA. Decreased acylsugar levels caused by introgression of chromosomal segments from
S. pennellii LA0716 into M82 tomato is counterintuitive because this wild tomato species
accumulates much higher amounts of acylsugars than the cultivated tomato.244 Two
general hypotheses would account for this unexpected behavior: either there are alleles of
M82 important for normal acylsugar levels or alleles from LA0716, which when removed
from the LA0716 genomic context, negatively influence acylsugar accumulation
compared to the genes from M82 tomato. This high-throughput with PCA statistic tool
greatly shortens the time to discover specific traits with phenotype among hundreds of

samples.

207

Table 5-1. The detectable acylsugars for ILs were listed with retention time (min) and

molecular mass for [M+HCOO]'.

 

Molecular mass

 

Acylsugars Retention time [M +H COO]-
S 4:14 3.14 639.25
S 4:15 3.25 653.27
S 4:16 3.35 667.28
S 4:17 3.43 681.30
S 4:18 3.49 695.33
S 4:20 3.65 723.34
S 4:21 3.69 737.36
S 4222(1) 3.74 751.37
S 4:22(2) 3.79 751.37
S 4:23 3.80 765.39
S 4:24 3.92 779.41
S 3:14 3.25 625.27
S 3:15 3.35 639.29
83:16 3.43 653.31
S3:18 3.58 681.33
S 3219(1) 3.65 695.35
S 3:19(2) 3.75 695.35
S 3:20 3.70 709.36
S 3:21 3.76 723.38
S 3:22 3.84 737.40

 

208

.>

 

peak area/lS/gdw

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

B T '2225.A ‘_ "'
In _ _5-B ;-'_ 11-A
-.--

3 SC N __11-B
_ ‘L __11-C
--5-D ‘-

m 11-D
u.)- "5-E <1; _—

Y -- :E : 11-E

“It-.. 1: :11:

u) 1- I _-

«5 "-5" " Y[-_11-H

11

Figure 5-9 (A) Total amount of acylsugars for 66 ILS and recurrent parent M82 based on
Quanlynx sofiware in Masslynx 4.1 (waters). Sum of all the peak areas of acylsugars
were normalized to internal standard (Propyl-4-hydroxybenzoate) and dry leaf dip.
Acylsugars are detected under ESI negative mode. The retention time and m/z of
[M+HCOO]_ corresponding to acylsugars are edit into method in Quanlynx. N=3, or 4
were tested for each IL that was grown except for IL3-2 (n = l) and IL2-6 (n = 2). (B)

Map of chromosomal introgressions for chromosomes 5 and 11.

209

5.3.1.3 ILs 8-1 and 8-1-1 causes a shift in acyl chain lengths without altering

numbers of substitutions

ILS 8-1 and 8-1-1 also stand out from other ILs and M82 based on PCA scores plot
and the signals responsible for this cluster in PCA loadings plot are m/z 667 (S 4:16), m/z
653 (S 4:15) and m/z 765 (S 4:23). TICs for IL8-1 and IL8-1-1 with changes in major
acylsugar peaks compared to M82 also supported PCA result (Figure 5-10A). The major
compound was 14 daltons lower in molecular mass than S 4:17 for IL 8-1-1.
F ragrnentation and accurate mass measurements indicated that this peak is a S 4:16
acylsucrose with a C4 fatty acid substituted for one of the C5 moieties (Figure 5-10 A
and B). Note that this is a quantitative change and M82 still accumulates S 4: 16 at ~ 50%
of the level seen in IL8-l and 8-1-1. Further analysis of the acylsugars in IL8-1-1 showed
an increase in the abundance of acylsugars having one or more C4 acyl chains compared
to M82 (Figure 5-10 A and B). Acylsugars lacking a C4 acyl chain however were
decreased in IL8-l-1 compared to M82 (Figure 5-10 B).

This feature was also supported through comparing specific functional groups using
multiplexed CID with aggressive CID condition. XICs of functional groups of C4 and C5
fatty acid from S. lycopersicum M82, S. pennellii LA0716 and IL 8-1 using CID potential
55 V are shown in Figure 5-11. Through comparison of the sum of integrated peak areas
of fatty anions for C5 to C4 among those species, C5 shifts to C4 in IL 8-1 compared to
M82, but similar with LA0716. Therefore, a great aspect for multiplexed CID is,
observation of characteristic functional groups for specific class of metabolites produced
under higher CID potential using multiplexed CID can assist profiling metabolic

phenotypes.

210

The acyl chains present on the acylsugars are proposed to come from precursors of
branched chain amino acid biosynthesis?” 246 Because the C5 acyl chains can come from
precursors of either leucine or isoleucine, two types of C5 acyl chains can be present. 3-
methyl-butyrate (iso-C5 or iC5) comes from the leucine biosynthetic pathway and 2-
methyl-butyrate (anteiso-CS or aiC5) comes from the isoleucine pathway. The higher
proportion of C4 acyl chains and the corresponding decrease in acylsugars with C5 acyl
chains prompted us to further characterize the types of acyl chains present on acylsugars
in M82 and IL8-1-l. Acyl chains on the acylsugars in leaf dips were trans—esterified to
the corresponding ethyl esters for analysis by GC-MS. This type of analysis allows for
determining whether acyl chains are branched or straight chain, and more specifically, the
position of the branch for the C5 branched chain. Following ethyl esterification, the leaf
clips from IL8-1-l showed an increase in isobutyrate ethyl ester compared to M82 (Figure
5-2.) This increase in the C4 chain was associated with a decrease in 3-methylbutyrate
ethyl ester (iC5) and not in the 2-methylbutyrate ethyl ester (aiC5). One possible reason
for this phenotype could be that since the iC5 precursors are derived from the C4
precursor, a specific decrease in only the iC5 side chains in IL8-l-l suggests that

elongation of the C4 precursor to iC5 is perturbed in this line.

211

Figure 5-10 A. Total ion chromatograms from LC-TOF MS analysis of acylsugars in leaf
dips of M82 and IL8-1-1. Labeling nomenclature for acylsugars - S 3:22 (5,5,12) is an
acylsucrose with 3 acyl chains having a total of 22 carbons and numbers in parentheses
indicate the lengths of the individual acyl chains. B. Amounts of acylsugars showing
differences in abundance between M82 and IL8-1-1 are shown as integrated peak areas
normalized to the internal standard and the dry weight of the extracted leaflet. Error bars
indicate standard deviation with n = 4. C. Schematic representation of chromosome 8
introgressions showing locus controlling the acylsugar phenotype located on bin 8-B.
Structure of S 4:17 (2,5,5,5) is based on LC-MS and NMR data. Structure of S 4:14
(2,4,4,4), an acylsugar detected only in IL 8-1-1 and not in M82, is inferred based on
negative and positive mode LC-MS.

212

 

 

 

 

 

 

 

 

 

 

100-
1%.
O‘T"‘1""1""l W11 ’1" '1””1"Iv:"":""i""1"”1'”-1'v'rtrflrt'rrrirfi'r'“:W"
1 2 3 4
84:16
100‘ 34:15 (2.4.5.5)
IL81-1 (2,445) 34:22
(2,4412)

%1. . S4223
. (24.5.12)
0 l 1 If‘TI l l”"l“"l I l I l l l

1 2 3 4
3000
2500, l
/ 6
12504 1 4
3 1000
750
5001
250
0‘ A ' A A A A A A
fi ‘0. 9.‘ N N ‘1'. ‘0. ‘0. 9.‘ N
696:“;‘1‘963
e’ :2- ~°Z v“ s; 2.- 23.- : v: v:
2‘) i: at. 9‘. 9! 3 53 ‘3 N c:
if -- "3 °° if £15 ‘4‘ N 9‘.
(I) V a) v
a) a) ‘0 co 0) a) a) (7)
C.
i | '1 H
obi “_
<1 OHOH
: co
"" ' S4:17(2,,55,5) —<.\

 

8-3
9°
.n

0::

°° 8'5 <02 9H
v.- ---3-0
(‘0

-. _3-H S4:,,,14(2444)

_ ”--

 

 

 

213

Figure 5-11 ESI negative LC/MS XICs of C4 m/z 87.04 and C5 m/z 101.06 fatty acyl
groups cleaved from acylsugars with CID potential 55 V for A: S. lycopersicum M82 and
B: S. pennellii LA0716. C: ILs 8-1. C5 fatty acyl group was predominated in M82, while
comparable amount of C4 and C5 fatty acyl groups were in LA0716 and C5 fatty acyl
groups shifted to C4 from IL8-1 based on the integration of peak areas for each fatty acyl
group. D: Mass spectra of peak A for S 3:20 from IL8-1 using ESI negative mode with
CID potential 55 V.

214

M82

 

C5 fatty acid

 

 

 

100;
. 101.06
%~ 113
1 \ J
04 v .. r r ,. Hi .-./1... -, --tg’?\5, ,VLfianfik/Lvn
10 20 30
100‘
‘ C4 fatty acid
i 87.04
i 51
0A4

 

0 that} - 1AA} A} A! Ml Ar nllmnur 11.1mm ”0“qu figlélfihu‘il' 15‘2me 1 i!”

 

 

 

 

 

 

 

 

 

1 0
0716 C5 fatty acid
1 00‘
1 01 .06
1 13
% 1 A x
J
1 . ' {l f H \A
0 w ...“ij -4; \kxij ~ V I LL x/ '1 R] ‘ V, Afkfl‘
10 20 30 40
100
. C4 fatty acud
4 l x 87.04
%- l s ‘ 51
1 1 .
‘ / /l R11 ‘10 1A ll
4 / K I U lvl ‘ NJ \r
‘ \ j l p. ,6 L. IR l x 1 Win/V L AW
0 g -T 4L». , , \ .--~-~a;4:L. .- , - ..... 5““ . vLTime
1 0 20 30 40

215

CS fatty acid

 

C4 fatty acid 87.04
51

 

 

 

 

 

[M'HT
10 C12 fatty acid 325
0' 199
‘ D
481
% _
253 I [M-Hl
[M-Hr ' 393 663 6199709
C4 fatty acid ‘ y 411 593 | 1
87 . 1 I 575
ll] 1'] l ll Ill l1 'l mu ['1 ii I‘I Infill“ 11:! ll lilllhll I I 11.1111 1 l i l . lll‘l m/Z
00 2 300 400 500 600 700

Figure 5-11 (Con’t)

216

50

 

Figure 5-12 Side chains on acylsugars from leaf dip samples were transesterified to the
corresponding ethyl esters and analyzed by GC/MS as described in chapter 6. The peak
area % values for the base peak XIC for each fatty acid in M82 and IL8-1—1 are shown
along with the standard error with n = 3 for IL 8-1-1 and n=5 for M82. For these data C4
ethylester levels were below limits of quantification for M82, which is presumed to arise

from the high volatility of this product.

5.3.1.4 Discovery of differences in accumulation of metabolites of lower abundance

ILs 1-1 and 1-1-3 with novel glycoalkaloids

Abundant metabolites that distinguish NILs from a reference genotype can be
recognized by visual inspection of LC/MS chromatograms, but discovery of
accumulation of less abundant metabolites requires alternative approaches. Low

abundance metabolites often are obscured in PCA plots, and rapid separations increase

217

the likelihood of metabolite coelution. These challenges can be overcome through
automated generation of extracted ion chromatograms (XICs) for specific molecular or
fragment masses of anticipated metabolites or metabolite fragment ions. In the current
study, XICs for the glycoalkaloid dehydrotomatine (m/z 1076.5 for the formate adduct
ion, negative ion mode) were generated and integrated, as this dehydrogenated form of
tomatine had been reported in earlier studies247 (Figure 5-13). While the quantitative
measurements showed modest differences in dehydrotomatine levels across the series of
all NILs, the chromatograms revealed an earlier-eluting isomer of dehydrotomatine that
was present only in the overlapping IL 1-1 and 1-1-3, and not observable in recurrent
parent M82 and about 50 folds in another parent LA0716. Fragmentation patterns
generated using positive ion mode suggested the position of the double bond in the early
eluting isomer to be in the F -ring,'98 probably in the form of an amine group. Once NILs
exhibiting a metabolic difference are identified, deeper probing of differences in
metabolite accumulation is achieved using a supervised multivariate statistical analysis
termed Orthogonal Projection to Latent Structures-Discriminant Analysis (OPLS-DA).92
This algorithm employs multiple linear regression and orthogonal filtering to aid
recognition of metabolites that discriminate sample classes, in this case the two classes
were M82 parent (class 1) and IL1-1-3 (class 2) (Figure 5-14) and the LC/MS data were
automatically extracted using MarkerLynx software to yield an array of aligned and
integrated peak areas. These processing steps generate lists of peak areas annotated with
specific LC retention times and masses, and the statistical analysis assigns a p(corr) score
suggestive of how well they discriminate classes, and this score is independent of the

metabolite abundance. Analysis by OPLS-DA suggested three metabolites abundant in

218

ILs 1-1 and 1-1-3, and these were manually annotated as didehydrotomatine,
hydroxytomatine, and the early eluting dehydrotomatine isomer based on fragmentation
patterns and accurate mass measurements. All three were undetected in every one of the
other 63 NILs. LC/MS signals for didehydrotomatine, hydroxytomatine metabolites

ranged from 0.8-6 % of the signal for the abundant S4: 1 7 tetraacylsucrose.

219

Figure 5-13. A. Extracted ion chromatograms of M82 and ILl-l for m/z 1076.5 show an
earlier eluting peak found only in the IL and not in M82. B. Schematic representation of
chromosome 1 introgressions showing locus controlling the glycoalkaloid phenotype is
located on bin l-A or l-B (ILl-l-2 was not analyzed in this study). C. Structures for
dehydrotomatine isomers. Compound with RT = 2.81 min having the double bond in the
B-ring has been described before.247 Structure of compound with RT = 2.52 min showing
the double bond in the F -ring is inferred based on fragmentation from positive ion mode

MS.

220

100 M82
('0
A % ($4--
‘7 1 __1-A
a LL15

 

 

. ""1‘C
0 . . - 1-D
1 2 3 4 ______ HE
100 " ::::1_F
lL1-1 . ____1-G
°/ ______ 1-H
° gig-"44
: M...
0 . 1 . . . . .. .,
1 2 3 4

 

(l) RT=2.52

221

_x
O

 

.0
CD
I

.0
C)
1

 

.0
it

.0
N
1

P(corr)[1]P

.0
o

 

-0.2-1

-0.44

 

 

 

-0.6 I

 

0.4 0-5

Pi1lP

 

Retention time_m/z p[1]P p(corr)[1]P

 

A 2.51_1066.5019 0.114 0.889
B 2.47_1074.4997 0.088 0.860
c 2.49_1076.5256 0.453 0.887
D 2.50_1094.5345 0.145 0.847

 

Figure 5-14. OPLS-DA loading plot showed interclass difference between IL 1-1 (Right)
and recurrent parent M82 (Left) using fused core coltunn with 5 min gradient. The ions
were labeled with retention time and m/z. The upper right corner and the lower lefi comer
are corresponding to IL 1-3 (n = 3) and M82 (n = 4), respectively. A: [M+Cl]_ of isomer
of dehydrotomatine at m/z 1066; B: [M+HCOO]_ of didehydrotomatine at m/z 1074; C:
[M+HCOO]- of isomer of dehydrotomatine at m/z 1076; D: [M+HCOO]_ of
hydroxytomatine at m/z 1094.

222

5.3.1.5 ILs 6—3 and 6-4 with higher ratio of tri- to dimethylmyricetin

Mono-, di- and trimethylmyricetins were observed in S. lycopersicum M82 and
structure annotation was described in chapter 4. Quanlynx software was used to quantify
these three methylmyricetins among all ILs and S. lycopersicum M82 (Fig. 5-15) based
on extracted ion chromatograms. IL6-3 and IL6-4 were discovered to have more
percentage of trimethylmyricetin compared to mon- and di-methylmyricetin, but
dimethylmyricetin was most abundance one among other [LS and recurrent parent M82,
implying that methyltransferases originating from S. pennellii somehow alter the amount
of trimethylmyricetin among these two ILS. These findings serve as the basis for
continuing efforts to characterize O-methytransferases involved in myricetin methylation
in S. habrochaites LA1777 in the laboratories of Jones and Eran Pichersky (University of

Michigan).

223

 

Fmgmefiiylrnyricetin lidirnethylinyricetin [3 trirhethylfnJfioeLtini

 

 

 

 

__ _ W ‘1' J TWA

2-4 6-3 6-4 M82

   

Figure 5-15. Total amount of mono-, di-, trimethylmyricetin for 66 ILs and recurrent
parent M82 were obtained based on Quanlynx software in Masslynx 4.1 (waters). The
peak area of mono-, di-, and trimethylmyricetin is normalized to internal standard
(propyl-4-hydroxybenzoate) and dry leaf dip individually for all the samples. Myricetin is
detected under ESI negative mode. The retention time and m/z of [M+HCOO]‘
corresponding to mono-, di- and trimethylmyricetin are edited into method in Quanlynx.
N = 3 for ILs and n = 30 for M82. ILs 6-3 and 6-4 showed the‘different phenotypes
compared to other ILs and samples. ILs 6-3 and 6-4 has more abundance

trimethylmyricetin, while for M82 and other ILs, dimethylmyricetin is predominant.

224

Combination of PCA, OPLS-DA and fast LC/MS, five specific phenotypes were
discovered from 65 ILs within minimal analysis time. Then genotypes mapping for F2
plants were performed to determine the phenotypes dominant or recessive compared to
their recurrent parent S. lycopersicum M82. The result of fast screening of F2 plants using
fast LC/MS was also used to guild genotype mapping. F2 plants were obtained by self-

fertilization of F1 plants.

5.3.2 F2 plants are screened whether the metabolic phenotypes dominant or

recessive for recurrent parent M82

The phenotype of [Ll-4 was sucrose triesters predominant compared to tetraesters.
PCA scores plot for 22 F2 plants of IL1-4 and 10 control plants recurrent parent
S.lycopersicum M82. 17 out of 22 (~80%) displayed the same phenotype with S.
lycopersium M82, which suggested phenotype of ILl-4 was S.lyc0persicum M82
predominant. Quanlynx software was also performed for acylsugars and compared the
ratio of sucrose trimesters to tetraesters. Figure 5-18 displayed the ratio of sucrose
trimesters to tetraesters for all those F2 plants of 1-4 and control plants M82. The result
showed 17 F2 IL1-4 out of 22 had predominant sucrose tetraesters instead of trimesters,

same as M82, which supported PCA result.

225

Figure 5-18. Total amount of sucrose triesters and tetraesters were obtained based on
Quanlynx software in Masslynx 4.1 (Waters) for 22 F2 plants of 1-4 and 12 control plants
M82. 17 out of 22 showed the similar ratio of sucrose triesters to tetraesters as

S. lycopersicum M82, which supported PCA scores plot for F2 plants of 1-4.

226

 

OH
4

R

Ratio of triesters to tetraesters

I
O

ZBW

’v-L lel
V'L lel
7'1 lel
V'L lel
7'1 Zl'll
’7'l 24'”
7'1 21'“
7'1 lel
V'L ZHI
V'l ZHI
V‘l lel
TL 21‘“
7'1 Zl'll
V'l Zl'll
V'l Zl'll
V‘L lel
TL 31'”
V'L Zl'll
V'L Zl'll
V'l Zl'll
V'l 31'"
T1 le|

 

 

60«

 

'7 1 1

O

50
40 —
0
0
10

227

Sample 10

 

Around 2000 F2 plants were screened using this fast LC/MS approach (lining the last
two years in our lab to guide genotype mapping, demonstrating the suitability of this
approach for large-scale phenotype analysis. Exploiting these capabilities continues to
assist our efforts to locate genes to specific chromosomal regions and accelerate the

discovery of gene functions relevant to plant chemistry.

5.4. Conclusions

Fast LC/TOF MS with multiplexed CID has been demonstrated as a powerful tool to
profile metabolic phenotypes in plant metabolomics within minimal analysis time. Three
advances provided by this fast LC/T OF MS approach for metabolite profiling of complex
Solanum extracts are fast detection of many metabolites (roughly one metabolite per
second of instrument time), high chromatographic resolution, and accurate mass
measurements for molecular and fragment ions for structure elucidation. This approach
has generated analyses of 300 samples over a 30 h period and yielded more than 1500
analytical signals after automated peak alignment and integration. Five specific
phenotypes including 10 ILs were discovered among 65 ILs within one week under
assistance by PCA and OPLS-DA. The phenotypes including low abundance acylsugars
in ILS-3, 11-3; fatty acyl groups shift from C5 to C4 in IL8-1, 8-1-1; acyl group missing
in IL1-3, 1-4; methylflavonoid ratio shift in IL6-3, 6-4 and isomers of dihydrotomatine in
ILl-l, 1-1-3. Significant novel markers responsible for those phenotypes were identified
using LC/MS with multiplexed CID.

Around 4000 plant extracts were screened during last two years in our lab for gene

mapping using this fast LC/MS approach. The genes regulating metabolic phenotype for

228

low abundance acylsugars were limited to a small chromosome region. Once gene
sequence of tomato become public, we can rapidly limit the gene location for those

phenotypes with the help of fast LC/MS.

229

Chapter 6. Qualitative and quantitative profiling of acylsugar metabolic
phenotypes for tomato breeding lines using liquid chromatography/time-of-

flight mass spectrometry

230

6.1 Introduction

Glandular trichomes of Solanum species secrete acylsugars enhance resistance of the
plants to several species of insects.]28‘248'25' These highly viscous lipids constitute a
significant proportion of leaf biomass in the Solanaceae and are produced in particularly
large amounts in the wild tomato species Solanum pennellii accession LA0716 (~ 20%
leaf dry weight).252 Certain wild tobacco (Nicot'z'ana) species produce acylsugars at up to
15% leaf dry weight, and these metabolites are also produced, but at lower levels, in the
tobacco species Nicotiana benthamiana.25 3 '108 Increasing acyl sugar production has long
been a target of tomato and potato (Solanum tuberosum) breeding programsm‘ 254'256
because of the potent antifeedant, antioviposition, and in some cases toxic properties of
these compounds toward insect pests. Many acyl sugars exhibit insecticidal activity

'6' Trichome-derived

exceeding the efficiency of a comparable standard, insecticidal soap.
acyl sugars have stimulated great interest owing to their functions as alternatives to
chemically synthesized sucrose ester insecticides, as antibiotics, and for consumer
products such as cosmetics.'63’257 Therefore, development of cultivars with increased
levels of acylsugar metabolites would provide an important alternative that could to
reduce applications of chemical insecticides, allowing for diminished environmental
impact. Knowledge of the genes involved in acylsugar biosynthesis will provide
information needed to manipulate production of acylsugars and other metabolites. While
production of genetically modified transgenic plants would stimulate controversy owing
to the potential that a modified genome will cause changes in the patterns or

concentrations of metabolites,40 traditional breeding approaches that involve genetic

crosses of two naturally occurring plants is considered more acceptable by many.

231

Therefore, understanding the difference of metabolic phenotypes between “normal”
plants and “modified” plants through classical breeding, mutation or manipulation can
facilitate generation of “natural” variants that incorporate a metabolic trait from a wild
species into a commercially viable plant such as tomato.

The research described in this chapter has involved collaboration with an expert in
tomato breeding for acylsugar content, Professor Martha Mutschler of Cornell University,
to refine a series of unique acylsugar-producing tomato breeding lines. The long-term
goal is to improve acylsugar production in tomato and develop insect-resistant tomato
lines. The lines investigated in this study were derived from crosses between the
conventional tomato S. lycopersicum and its wild relative S. pennellii LA0716. The latter
accumulates large amounts of glucose-based acylsugars in trichomes, but before this
study, conventional tomato was considered to not produce any acylsugars. These efforts
involved both identification and quantitative comparisons of acylsugar profiles in various
plant genotypes. Profiling of acylsugars was accomplished using LC/T OF MS, and
multiplexed C1D was applied to enhance metabolite identification and for quantitative
comparisons of metabolite levels across breeding lines. The basics underlying annotation
of acylsugar metabolite structures using LC/T OF mass spectrometry with multiplexed
CID was discussed earlier in chapter 3. This chapter will focus on quantitative
comparisons of glucose and sucrose esters and branched- or straight-chain fatty acid
esters among different breeding lines. One of the aims of this study has been to
demonstrate the performance of metabolite profiling using LC/T OF MS coupled with

multiplexed CID for accelerating development of crops with desired metabolic traits.

232

6.2 Experimental Section

6.2.1 Materials. Plant breeding lines were grown at Cornell University, and extracts of
plant leaves were prepared from freshly harvested leaves. These extracts were shipped to
Michigan State University on dry ice. Leaf extracts from wild type species LA0716,

LA1522, LA1777 were grown in green house of Michigan state university.

6.2.2 Chemicals. Acetonitrile (HPLC grade), 2-propanol (HPLC grade), methanol

(HPLC grade), formic acid (88%, CHC13 (HPLC grade) were from Fisher Scientific of

VWR Scientific. Thionyl chloride (analytical grade), 1,2,5,6-di-O-isopropylidene-u-D-

glucofuranose (98%), and decanoic acid were purchased from Sigma-Aldrich.

6.2.3 Preparation of internal standard (3-decanoyl-glucofuranose). 570 mg of
decanoic acid was dissolved in 1 ml thionyl chloride and refluxed in a 25 ml round

bottom flask for 4 h. After cooling to room temperature, volatiles were evaporated under

aspirator vacuum to dryness, and 5 ml CHC13 was added to the flask, followed by 100 mg

1,2,5,6-di-O-isopropylidene-u-D-glucofuranose. The mixture was refluxed for another 5
h, then cooled to room temperature. The reaction mixture was evaporated to dryness
under vacuum using a rotary evaporator. Raw product was dissolved in 2 ml methanol
and hydrolyzed under 1 ml 0.1 M NaOH, and then purified by C18 solid phase extraction
column, eluting with methanol/water. Yield was 31.3 %, and consisted of a mixture of
four isomers as determined by LC/MS. It is expected that the isomers represent products
of epimerization at the l-position and contributions of two epimers of 3-

decanoylglucopyranose formed by ring opening and closing to the pyranose form. The

233

structure of one isomer of the 3-decanoylglucofi1ranose product used as an internal

standard for acylsugar analyses is shown in Figure 6-1.

HO
HO

o OH
0 OH

Figure 6-1 Structure of synthesized internal standard 3-decanoylglucofuranose.

6.2.4 Extraction Method. To extract metabolites from Solanum trichomes, ~10 mg

leaves were dipped in 2 ml isopropanol : CH3CN : H20 (3:3:2 v/v/v) for 1 min. Extracts

were centrifuged, and the supematants were analyzed without further processing. Extracts

were stored at -20 °C until they were thawed for LC/MS analysis.

6.2.5 High-Performance Liquid Chromatography. A Shimadzu (Columbia, MD) LC-

20AD HPLC ternary pump was used. The separation was performed on a Thermo 1 X

150 mm BetaBasic C13 column with 3 pm particles. A gradient was executed based on

0.15% aqueous formic acid (A) and methanol (B). Gradient profile consisted of: initial
conditions 5% B; linear gradient to 50% B (5 min); 95% B (33 min); 100% B (35 min);
hold at 100% B until 38 min; return to 5% B (43 min). The flow rate was 0.1 ml-min_',

and the injection volume was 10 ul. The column was maintained at 30 °C.

234

6.2.6 Mass Spectrometry. Analyses were performed using a Waters LCT Premier mass
spectrometry (Milford, MA). Capillary voltage of 3.2 kV was used in both positive and

negative modes, using the following experimental parameters: desolvation gas flow, 300

L-h 1; desolvation gas temperature, 200 °C; source temperature, 90 °C; cone gas flow, 20

L-h . Mass spectra were acqu1red into five dlfl‘erent acqu1s1t10n fimctlons usmg rapid

switching of Aperture 1 voltage in the ion transit region of the mass spectrometer,
providing quasisimultanoeous generation of spectra under fragmenting and
nonfragmenting conditions. Mass spectra were acquired over m/z 50 to 1500 with a scan
time of 0.43 per function. Spectra were acquired in centroid format with ‘dynamic range
enhancement’ (DRE) enabled. Prior to performing all experiments, the instrument was
calibrated in both ionization modes over m/z 50-1500 range. This instrument was

equipped with a 4.0-GHz time-to-digital converter and a reflectron, and typically yielded

a resolving power of 5000 (fwhm). 0.1% H3PO4 was infused through the reference

sprayer of the lock-spray source as a lock mass.

6.2.7 Gas Chromatography/Mass Spectrometry

The mixed solvent (acetonitrile: isopropanol: water (3:3:2, v/v/v) acidified with 0.1%
formic acid) were evaporated using speed-vac to dryness. Then, add 20 pl 21% sodium
ethoxide (in denatured alcohol) to 1.5 ml microcentrifuge tubes with dry trichome
extraction residue and the tanseterification allowed reaction to precede for 30 min at
room temperature. The reaction mixture was washed with a saturated sodium chloride
solution (50 pl) to remove 50 pl of the top, hexane layer and transfer to autosampler vials
with low volume inserts. The sample was analyzed by capillary gas chromatography-—

mass spectrometry (GC-MS on a 10 m, 0.1 mm (id) fused-silica column with a 0.34 pm

235

thick stationary phase (DB-5; Agilent, Santa Clara, CA, USA). The GC program was as
follows: injector temperature 280°C; initial column temperature 36°C held for 4 min;
ramped at 10°C/min to 150°C, 20°C/min to 220°C; and finally at 40°C/min to 300°C,
held for 2 min. The helium carrier gas flow was set to 0.4 ml/min. All compounds were
analyzed using an Agilent 5975B quadrupole mass spectrometer (Santa Clara, CA, USA)
operated in 70 eV electron ionization mode using an Agilent 6890N GC system. This

approach is modified from Walters and coworkers’ paper. ' ‘2

6.2.8 Chemometric data analysis

All data obtained from ESI-MS of the trichome samples were extracted using
MarkerLynx 4.1 (Waters, Manchester, UK.) Mass spectral data were processed over the
entire retention time and m/z ranges using automated peak detection, alignment, and
integration. The parameters used included mass tolerance of 20 ppm, heavy isotopic
peaks were excluded for analysis, noise elimination level was set at 10.00, minimum
intensity was set to 10 % of base peak intensity, maximum masses per retention time was
set at 20 and finally, chromatographic retention time tolerance was set at 0.5 min. A list
of the integrated areas of the peaks detected was generated for all the samples, using RT
and m/z data pairs as the identifier of each analytical signal. These integrated areas were
normalized to the sum of all integrated signals for each sample. In separate analysis,
Quanlynx software was used to integrate peak areas of XICs for formate adducts of

targeted acylsugar metabolites.

236

6.2.9 Extraction and isolation of S 4:17

Sixty grams of freshly harvested leaves from 3-week old S. lycopersicum cv. M82
were extracted by dipping leaf material into 1.0 liter of isopropanol: acetonitrile: water

(3:3:2, v/v/v) containing 0.1% formic acid for 5 min. The resulting extract was partitioned
into two phases by addition of 500 ml CHCI3. The lower chloroform layer was collected

and evaporated to dryness on a rotary evaporator. The residue was dissolved in 10 ml
Methanol: H20 (1:1, v/v) for fiirther purification using a SupelcleanTM LC-C 18 solid

phase extraction (SPE) (10 ml column). A 0.2 ml aliquot was loaded onto the SPE

colmnn for each step of the process, and the process was repeated 25 times to accumulate

sufficient quantity for NMR analysis since we used two columns each run. Material was
eluted using mixtures of methanol/water, using 0.1% formic acid in the aqueous

component for the following methanol/aqueous compositions: Fraction 1 = 20 ml of

Methanol:H20 (9:1 1, v/v); Fraction 2 = 30 ml of MethanoleZO (11:9, v/v); Fraction 3 =

30 ml of MethanolegO (13:7, v/v), and Fraction 4 = 50 ml of MethanoleZO (7:3, v/v).

A wash with 20 ml 100% methanol was used to clean the column after each elution. After
the presence of the desired component was confirmed by flow injection analysis-
electrospray ionization MS in negative ion mode, the solvent was evaporated and all of

the material collected for Fraction 4 was pooled to obtain S 4: l 7 for NMR analysis.

6.3 Results and Discussion

6.3.1 Annotation of acylsugar from conventional tomato Solanum lycopersicum M82.

One parent for some of Professor Mutschler’s breeding lines is the conventional tomato,

line M82 of S. lycopersicum. Therefore, understanding the metabolic pattern of

237

acylsugars from this species is important for breeders to track the genetic basis of
phenotypes for progeny of this line. However, acylsugars had not been detected from S.
lycopersicum M82 prior to this study, presumably due to the limited sensitivity of those

. . . 9
on ginal analytical procedures. ' 32‘ l 6

This report describes, for the first time, the
accumulation of acylsugars in cultivated M82, with twelve distinct peaks in the LC/MS

results suggestive of tri- and tetraacylsucroses. Figure 6-2 shows the extracted ion

chromatograms for formate adducts of acylsugars for an extract of S. lycopersicum M82.

238

Figure 6-2 Xle of formate adducts from S 4:15 to S 4:24 (highlighted with red dash line)

and S 3:15, S 3:20 to S 3:22 (highlighted with blue dash line) from S. lycopersicum M82.

239

S 4:24

 

20

 

40

35.15

30

2'0

10

S 4:22

 

2'0

1'0

 

2'0

10

AA

5 3:21

8 4:20
20

 

1'0

4'0

2'0

1'0

 

S 4:18

 

 

 

0 0 O
.4 .4 44
L
0
w w r3
C
6
o1 x
5000
ZALo:
1 Jo... -
2
2
2 2
7
1
4
6
s ...
4
s
1 1 .1.
m % 0 0m % 0
1

LA

4-

A
T

4*
V Y

 

40

30

10

17.89

time

40

2'0

1‘0

 

240

Four major acylsugars annotated as S 3:15, S 4:17, S 3:22 and S 4:24 were detected in
accession M82. First, the most abundant component S 4: 17 was purified and the structure

was characterized using NMR. Isolation of other three is underway by undergraduate

student Thomas Westbrook. Structural information for S 4: 17 was generated from IH,
l3C, DEPT, gHMQC, gHMBC, gCOSY, and TOCSY NMR experiments. NMR spectra

are in Appendix.

24]

Table 6-1. Structure of S 4:17 in Solanum lycopersicum M82 characterized by NMR (1H,
l3C, DEPT, gHMQC, gHMBC, gCOSY, and TOCSY). Chemical shifls of proton and

I3 . . .
C resonances are listed With proton-proton coupling constants.

242

 

 

l3

 

Carbon # (multiplicity) 1H mult. (J, Hz) C
1(CH) 5.61 d(3.7) 89.3
2(CH) 4.85 dd(10.6, 3.7) 70.5
2-0—
-1 (CO) 170.2
-2 (CH3) 1.98 s 20.6
3 (CH) 5.44 dd (10.6, 10.0) 68.9
3-0-
-1 (CO) 175.4
-2 (CH) 2.29 m 40.9
-2 (-CH3) 1.03 d (7.0) 16.4
-3 (CH2) 1.56, 1.37 sex! (7.0) 26.5
-4 (CH3) 0.81 t (7.4) 11.4
4 (CH) 4.92 dd (10.2, 10.0) 68.5
4—0-
-1 (CO) 172.2
-2 (CH2) 2.1 - 2.4 mb 42.9
-3 (CH) 2.15 m 25.8
-4 (CH3) 0.9 d (7)a 22.3
-5 (CH3) 1.0 d (7)a 22.3
5 (CH) 4.09 m 71.5
6 (CH2) 3.56, 3.60 m 61.4
1' (CH2) 3.47, 3.58 d(12.3) 64.2
2' (C) 103.9
3' (CH) 5.21 d(7.9) 78.7
3 '-o-
-1 (CO) 174.0
-2 (CH2) 2.3 - 2.4 mb 43.0
-3 (CH) 2.03 m 25.3
—4 (CH3) 0.91 d(7)“ 22.3
-5 (CH3) 1.01 d (7)8 22.3
4' (CH) 4.45 dd (8.0, 8.0) 71.5
5' (CH) 3.91 m 82.5
6' (CH2) 3.70, 3.85 m 60.3

 

243

Twelve signals between 8 59 and 103 in the 13C NMR spectrum, which were

assigned (by DEPT) to seven oxymethyns, three oxymethylenes, one anomeric methyne
(8 89.6), and one nonprotonated anomeric carbon (8 102.2), were indicative of a
disaccharide. The 1H NMR signals and the connectivities observed in the gCOSY
spectrum revealed both glucose and fiiranose rings in the disaccharide. The HMBC
correlation from H-l of glucose (8 5.61) to C-2’ of fructose (8 102.2) completes the
identification of the disaccharide as sucrosezsg. The low-field shifts of the H-3, H-3’, H-4,

H-2 signals and the presence of carbonyl signals at 8 170.2, 175.4, 172.2, 174.0 were

assigned as 2, 3, 4, and 3’ carbons of sucrose with ester attachments. 1H, 13C, COSY,

HMQC and HMBC spectra allowed identification of the acylgroups as one 2-
methylbutanoyl, two 3-methylbutanoyls and one acetyl. The positions of these groups
were established by the HMBC correlations from H3 to C-1 (8 175.4) for 2-
methylbutanoyl, H-4 to C-1 (8 172.2) for 3-methylbutanoyl, H-2 to C-1 (8 170.2) for
acetyl, H-3’ to C-1 (8 174.0) for another 3-methylbutanoyl. Therefore, acylsugar S 4:17 is
substituted with an acetate ester at the 2-position, iso-CS esters at the 4- and 3’- positions,
and an anteiso-CS ester at the 3-position of sucrose.

Knowledge of the structure of S 4:17 made it possible to better interpret multiplexed
CID spectra generated using ESI and both negative and positive ion modes (Figure 6-3).
The anion at m/z 101.06 corresponds to deprotonated C5 fatty acid, and several fragment
ions correspond to losses of fatty ketenes from [M—H]_ as follows: m/z 593.27 (loss of
C2), 551.24 (loss of C5), 509.23 (loss of C2 plus C5), 425.16 (loss of C2 plus two C5),

341.11 (loss of C2 plus three C5, corresponds to [M-H]' of sucrose). In positive mode,

[M+NH4]+ ions undergo cleavage of the glycosidic bond to form two prominent fragment

244

ions with the more abundant at m/z 247.12 corresponding to the residue from the

fructofuranose ring with one C5 fatty acid ester, and the other at m/z 373.22

corresponding to the glucopyranose residue with two C5 and one C2 ester groups.

The acylsucrose metabolites observed in trichome extracts of leaflest of S.

lycopersicum M82 are descrobed in Table 6-2, which includes major fragment ions

generated for both positive and negative mode ESI mass spectra. It is worthy to note that

all the detected sucrose esters produce the fiagment ion at 247.12 in positive ion mode,

which suggested all have one C5 fatty acyl group on the fructofuranose ring. This result,

combined with the NMR data, suggest remarkable selectivity of substitution positions of

fatty ester groups in tomato acylsucroses.

100‘

0/o ‘

1001

 

[FRU +acyl]+

 

 

 

 

%‘

 

 

 

 

 

247.12
+ [M+Nal’
181.09 , [M+NH4] 559.30
[GLU + nacyl] 654 33
970412705 271.17 373.22 ' I
. 1. .1 . . . l- . . . . 1, . T
CS fatty anion -
101 06 [M+HCOO]
' 681.30
[M-H-02-2051' [1263;32'051- [M+Cl]'
, [M-H-cz—acsr 407-15 ' [M‘H’§21 571.27
A . 141304 341.11 425.16 1 5511-24 593.27 N L
. 'Hmfiul JILL LL . . 4L. .LL. 1 .. 4 . 4 1. - . .1...
100 200 300 400 500 600 m”

Figure 6-3 Electrospray ionization mass spectra of S 4:17 (RT = 3.11 min) extracted

from S. lycopersicum M82 using CID potential of 55 V (top) in negative and 40 V

(bottom) in positive mode.

245

Table 6-2. Acyl sucrose metabolites detected in trichome extracts from S. lycopersicum
M82 and their adduct ions and fragment ions observed using ESI negative (CID potential
55 V) and positive mode (40 V).

 

 

 

 

 

Acylsugar Fatty acid IM+NH4I+ |M+Na|+ Fragments
metabolite annotation constituents (m/z) (tn/z) (m/z)

S3:15 C5, C5, C5 612 617 247,331

S 4:17 C2, C5, C5, C5 654 659 247,373

S 3:20 C5, C5, C10 682 687 247,401

8 4:22 C2,C5, C5, C10 724 729 247,443

S 3:22 C5, C5, C12 710 715 247,429

S 4:24 C2, C5, C5,C12 752 757 247,471

Acylsugar metabolite Fatty acid |M+formate]' lM+Cl|‘ Fragment ions
annotation ester groups (m/z) (m/z) (m/z)

S 3:15 C5, C5, C5 639 629 593,509,425,407,34l,323,]01
S 4:17 C2, C5, C5, C5 681 671 635,593,509,425,407,34l,305,101
S 3:20 C5, C5, C10 709 699 663,579,425,407,34l,323,]71,101
S 4:22 C2,C5, C5, C10 75 'l 741 705, 621,579,425,407,341,323,]71,101
S 3:22 C5, C5, C12 737 727 691,607,509,425,407,34l,323,]99,lOl
S 4:24 C2,C5,C5,C l 2 779 769 733,691,607,425,407,341,323,]99,lOl

 

6.3.2 Compare total amounts of acylsugars and distributed amounts of acylsugars
from S. pennellii LA0716, LA1522, S. habrochaites LA1777 and S. lycopersicum M82

using LC/MS

Compared to cultivated S. lycopersicum M82, many wild type Solanum species
accumulate large amounts of acylsugars, which is the reason why plant breeders cross the
cultivated tomato with wild species to produce lines of cultivated tomato that have

inherited the genetic capacity to accumulate substantial amounts of acylsugars. This study

246

has aimed to make quantitative comparisons of acylsugar levels in tomato and several
lines of wild tomato previously known to accumulate acylsugars. To this end, analytical
responses for glucose triesters, sucrose triesters and sucrose tetraesters were quantified
for S. pennellii LA0716, L1522, S. habrochaites LA1777 and S. lycopersicum M82 by
adding all the LC/MS extracted ion chromatogram peak areas of formate adducts for
detected acylsugars normalized to the internal standard and dry leaf weight (Figure 6-4).
Unfortunately, no authentic standards are available for determination of instrument

response factors for the various acylsugar metabolites.

247

lGlucosetriesters ISucrosetriesters II Sucrose tetraesters 1 1

30000 - I

25000. 7 x
20000
15000. l
10000. i
5000. p77;
0., , J___JJ ,JLJ; .

LA0716 LA1522 LA1777 M82 l
Sample ID

 

Peak areas/ISlDry leaf weight(g)

 

 

Figure 6-4 Distributed amounts of glucose triesters, sucrose triesters, and sucrose
tetraesters of S. pennellii LA0716, LA1522, S. habrochaites LA1777 and S. lycopersicum
M82 based on sum of total peak areas of formate adducts for detected acylsugars
normalized to internal standard and dry leaf weight for each acylsugar. N=3, SE. Plants

were grow in MSU green house.

Based on the LC/MS results, the S. pennellii accession LA0716 produced the greatest
total amount of acylsugars, accumulating about 5-fold more than LA1522 and LA1777,
and about 36-fold more than tomato (M82). Glucose triesters predominated in both
accessions of S. pennellii, whereas sucrose esters were the dominant form in S.
habrochaites and S. lycopersicum. Glucose triester levels from S. pennellii LA0716 were
about 6-fold greater than from LA1522 (same species). Levels of glucose triesters from S.

habrochaites LA1777 were much lower, around 0.6% of the LA0716 levels. Tomato

248

(M82) did not yield detectable glucose triesters. The two S. pennellii lines (LA0716 and
1522) produced substantial amounts of sucrose triesters, but virtually no sucrose
tetraesters. The reverse was observed for S. habrochaites LA1777 and S. lycopersicum

M82, where sucrose tetraesters were the dominant forms.

6.3.3 Fatty acyl substituents of acylsugars among wild type S. pennellii LA0716, LA

1522, S. habrochaites LA1777 and S. lycopersicum M82 using GC/MS

In addition to differences among the four lines in amounts and ratios of acylsugars,
the distribution of fatty acyl substituents for acylsugars was also different, as established
by performing GC/MS analyses of fatty acid ethyl esters after transesterification.
Acylsugars from S. pennellii LA0716 contained the following branched chain fatty acids:
2-methylpropanoic acid (iC4), 2-methylbutanoic acid and 3-methylbutanoic acid (aiC5
and iC5), and 8-methylnonanoic acid (iC10), as well as some straight-chain fatty acids
(nC10 and nC12). These findings were consistent with a literature report that employed
GC/l\/IS.”O"64‘I69 For same species LA1522, the ratio of aiC5 to iC5 was about 2:1, and
was similar to the ratios observed for S. habrochaites LA1777 and S. lycopersicum M82.
However, in the close relative LA0716 gave the inverse ratio (about 1:2 aiC5:iC5).

Quantitative differences in the amounts of the individual fatty acid ester groups were
also determined using GC/MS (Figure 6-5). The iso-C4 fatty acyl group was dominant in
LA1777 and LA0716 (about 35% of total chains), but C4 was much less abundant (about
2% of total) in LA1522 and not detectable in extracts from S. lycopersicum M82 used in
this study. However, in separate analyses of M82 extracts from different sources obtained

at different times, low levels of iC4 fatty acid groups were detected.

249

Accumulation of large amounts of metabolites containing short or medium-length
branched chain fatty acid esters is not common in plants.169 Walters and Steffens”2
proposed the short branched chain fatty acids, such as 2-methylpropanoic acid (iC4), 2-
methylbutanoic acid and 3-methylbutanoic acid (aiC5 and iC5), arise from branched
chain amino acids such as valine, leucine and isoleucine. However, the mechanisms of
fatty acid elongation and the genetic control of carbon flux from amino acid to fatty acid
pathways in these systems are still uncertain.259 Combining information about identities
and levels of fatty acyl esters across numerous species with information about gene
expression may provide critical knowledge that will guide metabolic engineering of

plants and microbes capable of biosynthesis of these and other metabolites.

250

(fjiJL-aiGSLIRJS i505 iiCB .08 1109 31010 in010
501 971011 5011 11012 7.50127

 

 

 

% to the total
_I N (A) -h 01
O Q - _c.’ E’— .9 0.4
l
.1

1 LA1777 M82 LA1522 LA0716
l Accessions

Figure 6-5 Distribution of fatty acyl groups for total acylsugars among S. pennellii 0716.
1522, S. habrochaites 1777 and S. lycopersicum M82 obtained from GC/MS. N=3

separate plants, error bars represent standard error of the mean.

6.3.4 Compare total amounts of acylsugars and distributed amounts of glucose
triesters, sucrose triesters and sucrose tetraesters, and fatty acyl substituents of
acylsugars among Cornell breeding lines using LC/MS.

To determine how well a large scale metabolite profiling effort could guide breeding
of tomato lines for acylsugar production, extracts from leaflets from 90 different plants
samples were detected including the parent lines S. lycopersicum M82, S. pennellii
LA0716, introgression lines derived from crosses of LA0716 x M82, Second- to fifth

generation progeny from crosses of introgression lines with S. pennellii LA0716, and

251

crosses between a separate high acylsugar producing line (termed “97 fixed-lines”,
usually derived from several introgressions of LA0716 genes into the tomato parent) with
introgression lines, and several new lines developed at Cornell. The description of these
breeding lines is shown in Appendix Table A-1. Crosslines were detected for acylsugar
accumulation and subjected to QTL (Quantitative Trait Locus) analysis to determine the
genomic regions associated with the accumulation of acylglucoses, acylsucroses, and
total acylsugars. Total amounts of acylsugars and distribution of acylglucoses,
acylsucroses were surveyed to determine variance among those breeding lines. This
information is then used to identify breeding lines that carry desirable metabolic traits.
Several quantitative measures were used in this study: (1) total amounts of acylsugars
determined by LC/MS, (2) the ratio of acylglucoses to acylsucroses (LC/MS), (3) relative
amounts of aiC5, iC5, and iC4 fatty acid groups (determined by GC/MS). These
measurements help plant breeders identify desirable traits in these breeding lines, and
select lines for field testing of insect resistance.

Figure 6-6 displays the total amounts of acylsugars for crosslines and fixed acylsugar
lines. Distributions of glucose triesters, sucrose triesters and sucrose tetraesters are shown
in Figure 6-7. The parent line S. pennellii LA0716 accumulated ~3-folds more total
acylsugars than the Cornell fixed line 071402, which ranked second most abundance
acylsugars. Fixed lines and crosslines contained at least 10-fold more total acylsugars
than the parent tomato M82 and assorted introgression lines. All fixed lines produced
similar relative amounts of individual of sucrose triesters and tetraesters, but differed in
total amounts of acylsucroses. Therefore any difference in the susceptibility of the lines

to insect damage was attributed to variance in the level of total acylsugars but not in

252

relative levels of different acylsugars. These findings led to recommendations that
breeding efforts should focus on enhancing total acylsugar yields rather than engineering
a specific acylsugar composition.

Distribution of acylsugars was different among introgression lines, which are tomato
lines that contain small amounts of genomic DNA from an acylsugar-producing wild
parent. In comparing parent tomato M82 with introgression lines, sucrose tetraesters were
~10-folds of sucrose triesters except for IL1-3 and 1-4, for which the triesters were
dominant as described in chapter 5 (Figure 5-6). Several introgression lines were selected
based upon changes in either acylsugar amounts or profiles, and were crossed with the
high acylsugar producer S. pennellii LA0716 and with Cornell fixed lines. The purpose
of this experiment was to determine whether introducing genes that shift fatty acid
composition would also influence total acylsugar yields. Results from LC/MS analyses
were generated for crosses between introgression lines and S. pennellii (8-1 X 0716, 8-2 X
0716, 3-2 X 0716 and 7-5 X 0716). Three of these genotypes (8-2 X 0716, 3-2 X 0716 and
7-5 X 0716) produced similar levels of acylglucoses and acylsucroses, but for one
genotype (8-1 X 0716) acylglucoses were only about 30% of acylsucrose levels.,

In a related set of experiments, the Cornell acylsugar fixed line (97FL) was crossed
with tomato M82 and several introgression lines. Three genotypes (M82 X 97FL, 3-2 X
97FL, 7-4 X 97F L), did not accumulate detectable levels of acylglucoses, as expected

. because none of the parent lines accumulated acylglucoses.

253

 

.erra
PLVLH
OLVFN
movra

.I. 89.:

I BE
Li
I 69.:

r

J85:
Ill 35

T
11 88:

.r

lJNovK
.Ewmammi: I 53L.

7
_

Jukmsmw—Z Jmom:
.EE..YB_ I mom:

$5.93. 1|... 6.5:
$5.98. Ilsa:
$511.58
23.95.111.88:
98.92.11] 65:
I BE

ornodméd 0:04—39: loom:

0 0 0 0 0 mo 0 0
0 0 0 0 0 0 0
4 2 O 8 6 4 2

r .. 26.5.; am». am a 25.285 558.5 d 586 5.55....

New fixed lines

 

 

“H500

1 for IL8-2 X 0716. Error

254

2 or 3 for those lines except N

 

Figure 6-6 Total amounts of acylsugars in Cornell tomato breeding lines. Reported levels
represent the sum of XIC peak areas of acylsugar formate adducts normalized to internal

standard and dry leaf weight. N
bars represent standard error of the mean

l IGIucosetriesters ISucrosetriesters ;; Sucrose tetraestersj

   
  
 

 

  

L l

‘P_

 
 
 

 

:52

«1400 ’fii i’*

.C

D)

g1200l ’

.. i r; 1

3 ' 3 N r dl'
E1000, d 5 d ( ew1xe Ines

a ‘ ‘— to T Y N '0‘, g

.5 st 5 .
t; 600 E ‘7' to l 2 2 l
_ loco I _

~ = l l r l

E — . 1

3 i I 1 H l
.s ‘ I

F:

(U

2

N

X

(I

0

fl.

... 5
<3 .8. 3
E”: M-
71315 Elm-25716

(Dix Vwixoacoms-Nm Ow-N
oo Nrwu—oooooggggggvu—r
mm mommmvavvvvq’vv-Vv"
F‘— F‘OV-VFVY‘Ft—errFFFr-F
“N N‘DNNNNNNNNNNNNFNNI‘

.

 

Figure 6-7. Distribution of glucose triesters, sucrose triesters and sucrose tetraesters
among all the breeding lines. N=2 or 3 for those lines except N=l for IL8-2 X 0716. Error

bars represent standard error of the mean.

In addition to variations in total amounts of acylsugars and in distributions of
acylglucoses versus acylsucroses among those breeding lines, another feature of great
interest was a shift in fatty acyl substitutions from C5 to C4 fatty acids observed for 8-1 X
0716 compared to other crosslines and fixed lines. Such a shift in metabolite abundances
is not readily apparent from LC/MS data generated using gentle conditions, but the
presence of specific fatty acid groups is apparent in high CID energy spectra where fatty
acid anions are abundant. Screening for this shift in fatty acyl composition can be
assessed by generating XICs for specific fatty acid anions (e.g. m/z 87 and 101 for C4 and

C5 fatty acids respectively) using high CID potentials.

255

Figure 6-8 shows the XICs of C4 and C5 fatty acid anion fi'agments in 8-1 X 0716 and
8-2 X 0716 using 55 V CID potential. Both lines accumulated acylglucoses and
acylsucroses. Several differences in acylsugar composition are evident from these
chromatograms. First, the acylsugars of line 8-1 X0716 display more peaks with C4 fatty
acid fragments than 8-2X0716 (approximately 18 notable peaks compared to 4 peaks),
suggesting a more diverse population of C4-containing acylsugars in crossline 8-1 X 0716.
In addition, integration of chromatographic peak areas led to the conclusion that 8-1 X
0716 accumulated comparable amounts of C5 and C4 fatty acyl groups, but for 8-2 X
0716, acylsugars with C5 groups were ~ 5-folds more abundant that acylsugars with C4
chains. The latter distribution was similar to results fi'om other crosses between
introgression lines and wild type S. pennellii LA0716.

Acylsugar metabolites were identified with the aid of multiplexed CID for crosses
between introgression lines and wild type S. pennellii LAO716 (Table 6-3 and 6-4). For in
8-1 X 0716, most sucrose triesters and tetraesters have at least one C4 fatty acyl group,
but almost no C4 on acylsucroses was observed for other cross lines. This result indicated
that genes related to C4 fatty acid substitution are located at the top of chromosome 8,
which is in agreement with GC data from Martha Mutschler’s group.169 The remarkable
diversity of acylsugar metabolites from this line presents substantial challenges for
metabolite isolation for more detailed structure analysis.

The LC/MS results were also employed to generate quantitative information about
levels of individual acylsugars in the lines described above. These profiles were derived
by manually constructing a table of retention times and m/z values for formate adduct

ions, and using QuanLynx software to integrate all observed peaks corresponding to

256

acylsugar formate adducts. The results demonstrated that line 8-1 X 0716 had lower
accumulation of glucose triesters and lower total acylsugars compared to lines 8-2 X 0716,
3-2 X 0716 and 7-5 X 0716. Compared to all other lines of tomato, wild relatives, and
genetic crosses, the suite of acylsugars accumulated by the cross of IL 8-1 with S.
pennellii LA0716 accumulates about twice as many distinguishable acylsugars than any
other plants investigated in this research. This investigation therefore documents two
important findings. First, the LC/multiplexed CID TOF MS approach generates sufficient
information to describe acylsugar type (tri- vs. tetra-esters, glucose vs. sucrose esters) to
be useful for phenotypic screens. Second, this tool has demonstrated its utility in
development of “gain-of-function” lines that produce metabolites not synthesized by
either parent line. The combination of these approaches lends support to metabolic

engineering of plants for desired biochemical traits.

257

m m/2101.06

 

 

 

 

 

 

 

 

100 152
D (:35 fatty acid
% l. Ill ,1 v if) 1 M
0 ' J“ L. . . *rwwif‘LE/LJYW iii—i . L ”I: Jaw?“ WAN/l, L ”Cm/1M 1 . . 1
10 20 40
8-1 x 0716 m/z 87.04
100 . 111
o/ C fir C4 fatty acrd
o
0 I. . . . "’"f’v- ............... fifty-(~63?) ,‘L - .I ‘ \wfljfirfi. T'fiLJB‘Tfl . 3g. . >0 q‘CV‘S-r—L ...... . .......
10 20 40

m/z10106
......

100
,4 3 Wk C5 fatty acid

 

 

1’0 30 40""
W2 87.04
100 A 331
% C4 fatty acid
. ... . . Y a-.. .. 1.5/”5 - 3.8.5-..- ..... .. -
O ‘ 10 20 30 40 T'me

Figure 6-8. ESI negative XICs of C4 m/z 87.04 and C5 m/z 101.06 fatty acyl groups
using CID potential 55 V for crosses 8-2 X 0716 (A and B) and 8-1 X 0716 (C and D).
The number labeled below mass to charge ratio on the right of chromatogram indicate the

base peak intensity,

258

Table 6-3. Fatty acyl constituents for detected acylsucroses in 8-1 X 0716 using ESI
negative mode with CID potential at 55 V.

 

 

 

 

Acylsugar metabolite Formula of Measured m/z . .
annotation formate adduct [M+HCOO]" Fatty acnds constituents

S 3.12 C26H43O16 597.20 C4,C4,“

S 3.13 C26H43O16 611.23 C4,C4,05

S 3.15 C28H47O16 639.28 C5,CS,C5
s 3:18 C31H53016 681.33 gj'gg'ggo
C4,C4,C12
CS,C5,C1O

Aey\sugar metabolite Formula of Measured m/z

Fatty acids constituents

 

annotation formate adduct [M+HCOO]—
S 4:14 C27H43017 639.25 C2,C4,C4,C4
S 4: 16 C29H47O17 667.28 C2,C4,C5,C5
. C2,C4,C5,C10
S 4.21 C34H57017 737.36 C2,C4,C4,C11
C2,C4,C4,C12
C2,C5,C5,C10
. C2,C4,C5,C12
S 4.23 C36H61017 765.39 C2,C5,C5,C11
S 4.24 C37H63O17 779.41 C2,C5,C§,C12

 

259

Table 6-4. Fatty acyl constituents for detected acylsucroses in 8-2 X 0716, 3-2 X 0716
and 7-5 X 0716 using ESI negative mode with CID potential at 55 V.

 

 

Acylsugar metabolite Formula of formate Measures m/z Fatty acids
annotation. adduct [M+HCOO]" constituents
S 3: 15 'C28H47016 639.28 05.05.05
8 3: 19 C32H55016 695.35 05,05,09
s 3:20 C33H55016 709.36 gigggl‘:
S 3:21 C34H57016 723.38 05,05,011
S 3:22 C35H59016 737.39 05,05,012
S 3:23 C36H61016 751.41 05,06,012

 

 

 

Acylsugar metabolite Formula of formate Measures m/z Fatty acids
annotation adduct [M+HCOO]" constituents
S 4:16 C29H47017 667.28 CZ,C4,C5,C5
S 4:17 (3301149017 681.29 02.05.05.05
8 4:21 C34H57017 737.36 C2,CS,C§,CQ
S 4:22 0351159017 751.38 02,05,05,C10
S 4:23 C36H61O17 765.39 02,05,05,C11
S 4:24 CB7H63017 779.41 02,05,05,C12
S 4:25 C38H65017 793.43 02,05,06,012

 

Although the metabolite profiling results for 90 breeding lines (described above)
identified lines that accumulate high levels of acylsugars and display significant
reduction of insect infestation in field trials, these lines exhibit undesirable traits from the
parent S. pennellii LAO716 including small fruit, and poor gemination and growth. In
order to eliminate those negative traits, the Cornell group has combined data from these
acylsugar profiles with DNA marker information to select plants with high acylsugar
production, but reduced contributions of S. pennellii DNA in regions not associated with

acylsugar production. These plants have undergone further field trials, and new

260

commercial tomato lines with high trichome acylsugar accumulation may be introduced
to the commercial market within the next few years.

Continued efforts to improve acylsugar accumulation in tomato trichomes has
continued with investigation of 13 additional lines not described above. For this study,
LC/MS metabolite profiles were generated for a total of 60 plants including S. pennellii,
normal tomato lines, five fixed-lines and eight crosses between fixed lines and S.
pennellii LAO716. The description of these breeding lines was displayed in Appendix
Table A-2.

Annotation of metabolites as acylsugars was performed for the set of 60 plants
mentioned above, using LC/TOF MS and multiple CID conditions. From this information,
a list of target acylsugar metabolites was constructed, and XIC for the formate adduct of
each was integrated using QuanLynx software. For each extract, the sum of all integrated
peak areas was calculated (Figure 6-9 Top). S. pennellii LAO716 accmnulates 3 to 7-fold
more total acylsugars compared to all fixed lines, and the conventional tomato line
accumulated minimal amounts relative to all otheres. Acylsugar totals in crosses between
fixed lines and S. pennellii LA0716 roughly doubled compared to fixed-lines. These
findings were expected, since inter-specific crosses between a species with high levels of
any of these compounds and a species with low levels usually results in plants with

259260. Substantial diversity in acylsugar

intermediate levels of the compounds
accumulation was observed across this set of plants. For S. pennellii LA0716, ~90% of

acylsugar signal was attributed to triacylglucoses. There were no detectable acylglucoses

in fixed lines and normal tomato lines. Glucose triesters account for 30-50 % of the total

261

acylsugars (Figure 6-9 bottom) among crosses, which suggests a biochemical inheritance

from their parent S. pennellii.

262

Figure 6-9 Amounts of acylsugars in S. pennellii, normal tomato lines, fixed-lines and
crosslines based on sum of peak areas, then normalized to internal standard and dry leaf
weight (Top). Composition is broken down into glucose triesters, sucrose triesters and
sucrose tetraesters were among all the breeding lines (Bottom). N=4. Error bars represent

standard error of the mean.

263

 

 

 

   

Crosslines

if Fixed lines and normal tomato lines

 

 

6
1
7
o
l . 1
A0
000000000 ,0
00000000 0
00000000
00000000
87654321

.2220; 08. 0.20.0020 000..

Sample ID

 

 

 

 

.I... @0028

                     

I 0.0023
. 0.2023

I 00023
I @0028

Sucrose tetraesters
Crosslines

I 0.328

  

I Qmwonno

.| 0.30520

I @0028

0.028
I 0.0.23
I 0.928

Fixed lines and
normal tomato lines

I 0.28

IIGIucose triesters I Sucrose triesters

I 0.3.28

 

 

 

 

 

6
1
7 I......-....III......OIIOCIIUOIDO .-
0
0-: 2.30
0 0 0 0 0 0 0 0 mu
0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0 0
9 8 7 6 5 4 3 2 1

.2 2225 03. En \ 23:5» .3305 2 «9:0 :09...

Samples ID

 

 

264

One of the more interesting aspects about these breeding lines was the increased
diversity of acylsugars in plants derived by crossing fixed lines with wild type S.
pennellii LAO716. Figure 6-10 compares LAO716, fixed line 071026 and crosses 071026
X LAO716. Many acylsugars that were not present in either parent appeared after crossing.
Fatty acyl substituents C11 and C9 were prominent in the cross, but had low abundance
in both parents. (Figure 6—11). This new variation in metabolite composition may derive
from two sources: (I) changed abundances of specific fatty acid precursors, and (2)
additional variation in positions of fatty acid substitution. One may ask how new
substitution patterns arise after crossing. One hypothesis is that the cross leads to
expression of enzymes in the cross that was not expressed in either parent. .Future

investigations will be needed to test this hypothesis.

265

100, LA0716

1 f V
«1.. ll W1 1r
3 n~_ __Mflum,,\/10/kJJM\ V ofidfliggfaa,fA¢AmeNLfU %kN\J\0__¥~_m___

 

 

 

 

 

 

O f I 16 """""""" id T *3‘0' """"""""" '4'0'
100: 071026XLA0716
%: jl UL A lily/V10“ fl
1 A W V l 1“
M ...... 1 VI . .. if , f- 00
10 20 30 40
mo
1”. 071026
1..
% {1‘ ()1 ll I“ 1;
0 wwkflw. ’dW.HU%M#MmAflMMfi0 T
“Mfr—74*» I FF—F F T ' ' firfififi ................. me
10 20 30 4o '

Figure 6-10. TIC from wild type S. pennellii LAO716 (top), cross of 071026 X LAO716
(middle) and 071026 (bottom) using ESI negative mode with CID potential at 10 V.

266

m/z 157.13
100] 071026xLA0716 27593

 

 

 

 

 

0/o I A
0 v 1 Y Y , . ha T . 7— ,J‘fi—Jis—J L/‘LJJE—eufi Y 7 V , ,
1 0 20 30 40
1 00 W2 1 85. 16
071026XLA0716 A 10764
%.
1 MRJLJ’A‘VFKJ [in
O ‘1'o"' ’ 2b ’ ' ‘ 3b 40
W2 1 57.13
100i 071026 l 11
% I It: ' ll ll b
0 LL—‘éy—AA;LT"IE“~1AAM1A¥LAArwflk';'afi‘ 4313.“.1Piul‘h3 ‘ filll'fiifl‘:¢pfikiiilew’lqrjk{A $64“Lw‘llLuij' 'U VVI ”1'! lhfldLyfiAld my}.
1 O 20 30 4O
m/z 1 85. 16
100] 071026 ‘1 924
%l 1'1
1’1 ,1 I
A: ' ' r A‘ 1 ' Y I Y T r ‘ Tr‘ ‘T ‘4: WE r k ‘ V T ”W‘an-zn; "me '
0 1o 20 30 4o T'me

Figure 6-11 ESI negative XICs of C9 m/z 157.13 and C10 m/z 185.16 fatty acyl groups
using CID potential 55 V for fixed line 071026 and crosses between 071026 and S.
pennellii LAO716. The numbers below m/z on the right top of chromatograms indicate the
base ion peak intensity.

The results from LC/MS were supported by using GC/MS for fatty acyl group
analysis. Fatty acyl groups among fixed lines, crosses between fixed lines and S. pennellii
LA0716 were analyzed using GC/MS after transesterification in Figure 6-12. For fixed
lines, they contain predominantly: 2-methylbutanoic acid and 3-methylbutanoic acid
(aiC5 and iC5), and straight-chain fatty acid nC12, the same as their parent normal
tomato line. But in plants generated by crossing fixed lines with S. pennellii LAO716, the

distribution of fatty acids changed. iC4, aiC6, iC9 and iC11 fatty acids appear that were

267

barely detectable in the fixed lines. For example, the branched chain iC11 fatty acyl
group became the second most abundant fatty acyl group (about 40% to the total fatty
acyl) after iC5 among those crosslines, but iC11 was not only 1% of fatty acyl groups in

LA0716 and was not detected in the fixed lines.

 

 

 

 

i’IiC4 Iaicsliics liCiDiCéEl K3195 glow-{ca} ninmz'

(O

O

0
VJ

800 l

0105\1
GOO
000

N0)
00
OO

 

Peak area/lsldry leaf weight (mg)
A
O
O

 

 

 

 

Figure 6-12 Distribution of amounts of individual fatty acyl groups among fixed-lines,
crosslines and S. pennellii LAO716 based on GC/MS profiles. 077044-6, 077048,
077049-6, 077051-6, 077058-6 are fixed lines; 077082-6 to 077089-6 are crosses from
fixed lines and S. pennellii LA0716; N=4, Error bars represent standard error of the mean.

Annotation of acylsugars were based on LC/TOF MS data for retention time and
accurate masses of formate adducts and fragment ions, following the same approach
described in Chapter 3. Fatty acyl chains for acylsugars were determined based on fatty

acid anions in negative mode. Fatty acid compositions for observed glucose triesters for

268

fixed lines and crosslines between fixed lines and S. pennellii LAO716 are displayed in
the Tables.6-5 and 6-6. The suite of acylsugars accumulated by the cross of fixed lines
with S. pennellii LA0716 accumulates about three folds as many distinguishable

acylsugars (about 22) than fixed lines themselves (about 7) investigated in this research.

Table 6-5. Fatty acid constituents of detected acylsucroses from fixed lines using ESI

negative (CID potential, 55 V).

 

 

 

 

 

Acylsugar metabolite Formula of formate Measures m/z Fatty acids
annotation adduct [M+HCOO]- constituents
S 3: 1 5 C28H47O16 639.28 C5,C5,Cfi
S 3:20 033H55016 709.36 CS,CS,C1O
S 3:21 034H57015 723.37 04,05,012
S 3:22 035H59016 737.39 C5,C5,C12
Acylsugar metabolite Formula of Measures m/z Fatty acids
annotation formate adduct [M+HCOO]- constituents
S 4:17 C30H49017 681.29 CZ,C§.CS,C5
S 4:22 C35H59017 751.38 C2,C5.C5,C10
S 4:24 C37H63017 779 41 C2,C5,C5,C12

 

269

Table 6—6. Fatty acid constituents of detected acylsugars from crosses between fixed

lines and S. pennellii LAO716 using ESI negative (CID potential, 55 V).

 

 

 

 

 

 

 

 

Acyls:Ig:;tl:tei:lbolite Formuala:l ((1):; cfprmate llVlMeislt'iEW Fatty acids constituents
G 3:14 020H33011 463.21 04.05.05
G 3:15 021H35O11 477.23 05.05.05
G 3:16 C22H37011 491.24 05.05.06
G 3:18 023H39011 519.28 04.05.09
G 3:19 025H43011 533.29 05.05.09
G 3120 CZ6H45011 547-3’1 (033.838?
G 3:21 0271447011 561.32 05,05,011
G 3:22 C28H49011 575.34 05,05,012
G 3:23 029H51O11 589.36 05,06,012
Acyls:lg:;tl:tei:‘bollte F ormu:i :1): cfprmate miles-22:31? Fatty acids constituents
S 3:15 C28H47O16 639.28 05.05.05
S 3:19 C32H55016 695.35 05.05.09
3 3:20 C33H55016 709.36 giggg}?
S 3:21 C34H57016 723.38 05,05,011
S 3:22 C35H59016 737.39 05,05,012
S 3:23 0361461016 751.41 05,06,012
Acylsugar metabolite Formula of formate Measures m/z Fatty acids
annotation adduct [M+HCOO]" constituents
S 4:16 029H47017 667.28 02.04.05.05
S 4:17 C30H49017 681.29 02.05.05.05
S 4:21 C34|~|57017 737.36 02.05.05.09
S 4:22 C35H59017 751.38 02.05.05.010
S 4:23 (3361461017 765.39 02.05.05.011
S 4:24 (3371163017 779.41 02.05.05.012
S 4:25 C38H65017 793.43 02.05.06.012

 

270

6.4 Conclusion

This chapter has presented results from profiling of acylsugar metabolites in (about

100) different genotypes of tomato breeding lines. Several aspects of this research

represent original scientific contributions that demonstrate how nontargeted metabolomic

profiling can guide plant breeding efforts to develop desirable traits.

1.

Contrary to prior belief, conventional tomato (line M82) does accumulate
acylsugars, but at low levels compared to other lines. The LC/MS approach
developed for this study proved capable of detecting acylsugar production in all
lines studied. This advance allows investigation of factors that regulate acylsugar
production, as distinct from measuring the presence or absence of acylsugars.
Recognition that all tomato lines produce measureable quantities of acylsugars
offers the prospect for a paradigm shift in understanding factors that regulate
acylsugar accumulation.

Before this work, acylsugar analyses were almost entirely performed by
performing base hydrolysis of organic plant extracts, followed by colorimetric
measurement of glucose. Measurements of sucrose esters required enzymatic
conversion to glucose (using invertase), that could also be detected by colorimetry.
The LC/MS method developed in this work has led to more than IOO-fold
lowering of detection limits, and has allowed for comprehensive profiling of 50 or
more different acylsugars in a single plant extract.

Tomato breeding lines derived by crossing S. pennellii LA0716 with other fixed

lines of accumulated acylsugar mixtures with greater diversity than combinations

271

of the two parents might suggest. These new crosslines demonstrate that a
bottleneck to metabolic engineering has been overcome through traditional
breeding, without need for recombinant DNA techniques. The LC/MS
methodology developed here offers a powerful tool for screening plants for

biochemical composition, and should aid future plant breeding efforts.

272

Chapter 7. Concluding remarks

273

For most of the modern era, discoveries of genes have been driven by a reductionist
approach in which organisms were reduced to individual genes that were studied one-at-
a-time and in isolation. Although these efforts generated a great deal of knowledge,
complex networks control most biological processes, and these are not amenable to yield
a comprehensive understanding when they are studied in isolated model systems. For
several reasons, engineering of desirable secondary metabolite levels in plants has found
limited success. Fortunately, new high-throughput technologies that can link gene
expression levels with comprehensive chemical profiles offer the prospect for
overcoming many technical obstacles to metabolic engineering of plants and other
organisms. These technologies have brought about an era of systems biology, which is a
more holistic way to view biological functions. To move such efforts forward, this thesis
has presented development and application of a new metabolomic methodology using
LC/MS for discovery and profiling of specialized metabolites in glandular trichomes of
numerous plants in the genus Solanum. One central goal of my efforts has been to push
the limits of a relatively low-cost mass spectrometer to provide rich information
regarding metabolite identities and quantities that could be accessible to laboratories
without the resources to purchase expensive and more powerful mass spectrometers.

Technologies for metabolite profiling have lagged behind other genomic and
proteomic methods owing to the chemical diversity of metabolites. Improved capabilities
in analytical throughput for both metabolite identification and quantification are needed
to generate comprehensive metabolite databases that will be useful for gene discoveries.
To this end, Chapter 2 describes development of a new LC/TOF MS approach for

metabolite profiling that uses nonselective and multiplexed CID. This analytical

274

methodology enables deep coverage of the metabolome using a consistent protocol for
identification and quantification, yielding accurate measurements of molecular and
fragment masses and ion behavior in a collision energy-resolved context. The
combination of this information offers the prospect for annotating ions as molecular,
oligomer, or fragment ions and for improved recognition of chemical classes of
metabolites based on mass and ease of fragmentation, even when multiple metabolites
coelute. These improvements in analytical throughput allow faster metabolic profiling
analyses to be performed to identify major and minor metabolites in high complexity
biological extracts. In Chapters 3 and 4, the LC/multiplexed CID MS approach was
applied for nontargeted profiling of metabolites extracted from trichomes of genus
Solanum. These analyses led to detection and annotation of many secondary metabolites,
covering acylsugars, flavonoids, alkaloids and terpenoids, many of which were
previously unknown in specific species. Experimental determination of metabolic
diversity provides researchers with an opening to discover how genes and environment
combine to produce different biochemical outcomes, and rapid screening for differences
in plant chemistry can guide efforts toward gene discovery.

Chapter 5 takes multiplexed CID into the realm of high-throughput screening, and
describes a new fast LC/MS approach used to screen trichome metabolic phenotypes for
thousands of transgenic plants. The additional dimensionality of the MS data (CID
energy) aids metabolite annotation even when chromatographic profiles overlap, and
enables measurements of approximately 100 metabolites per minute of instrument time,
and analyses of several thousand samples per year on a single mass spectrometer.

Results from screening of tomato introgression lines have identified several chromosome

275

 

regions that influence trichome chemical composition and amounts of trichome
metabolites and these findings are guiding gene mapping efforts. Finally, Chapter 6
applies LC/T OF MS with GC/MS for qualitative and quantitative profiling of anti-insect
acylsugar metabolites for Cornell breeding lines of tomato. These results demonstrated
nontargeted metabolite profiling using LC/MS reveals remarkable chemical diversity
among plant genotypes, and can guide plant breeding efforts to develop desirable traits in
economically important crops.

The fast LC/TOF MS-multiplexed CID method offers great value for deep profiling
of metabolomes but requiring minimal instrument time per sample. It allows metabolic
phenotype screening for thousands of samples to become practical, and generates
qualitative and quantitative analysis results for unidentified metabolites to assist
discovery of their biosynthetic pathways. Additional applications of the LC/T OF and
multiplexed CID approach, not included in this dissertation, have generated metabolite
profiles as parts of two collaborative studies, one involving plant resistance to insect
herbivory with the laboratory of Professor Gregg Howe, and the other a survey of other
plants within the family Solanaceae with Professor Comelius Barry’s group. In addition,
this approach is widely performed by other members of the Jones laboratory for profiling
of plant metabolic responses to pathogen stress, responses of cancer cells to
chemotherapeutic agents, discovery of novel bioactive oxylipins in plants and algae, and
profiling of soluble byproducts of biomass processing for generation of cellulosic
ethanol.

Perhaps the most striking conclusion that can be drawn from the thousands of LC/MS

analyses performed during these studies is the. remarkable chemical diversity observed

276

within a single plant genus. Even closely related lines within the same species showed
distinct metabolite profiles. These are largely attributed to small changes in DNA
sequence resulting in substantial changes to plant chemistry. Improved understanding of
the biochemical basis for these differences will come from ongoing and future research,
and offers hope that the current century will see improved use of biological sources of

chemicals.

277

Appendix

278

.II
A

Table A-1 Description of breeding lines from Cornell including introgression lines,

fixed lines, crosses between introgression lines and S. pennellii LA0716, and new fixed

lines.

Ml sample #

I—I—I———I—_I___
\OOOQONUIADJN_O\OOO\IO\UI-th—s

WWWNNNNNNNNNN
N—OOOOQONMALQN—O

Samples for GH dips
66018
71302
71304
71305
71306
71307
71308
71309
71311
71312
71313
71314
71315
71316
71317
71318
71319
71320
71321
71322
71401
71402
71403
71404
71405
71406
71407
71408
71409
71410
71411
71412

 

ped [qree

 

97FL

 

M82

 

M82

 

lL lO-l

 

lL8-l X LA716

 

lL8-2 X LA7l6

 

M82 X 97FL

 

IL3'5 X 97FL

 

[L 3-5

 

IL 8-l

 

lL 8-l-l

 

[L 8-2

 

lL3-2

 

lL3-2 X LA7l6

 

lL3-2 X 97FL

 

IL7-4

 

IL7-4 X 97FL

 

IL 7-5

 

lL7-5 X LA7l6

 

IL- 1 0-1 / control

 

[LP cyto X 01 lOl4-DO7] X (prog sel from 01 lOl4-BO3)F2

 

(short 3 X 99FL F2 selection 01-1014-B03)F6

 

[LP cyto X 01 lOl4-DO7] X (prog sel from 01 lOl4—BO3)F3

 

(short 3 X 99FL F2 selection 01-1014-BO3) F5

 

[LP cyto X 01 lOl4-DO7] X (prog sel from 01 lOl4-BO3)F3

 

[LP cyto X 0|]014-DO7] X (prog sel from 011014-803)F2

 

(short 3 X 99FL F2 selection 01-1014-BO3)F5

 

[LP cyto X 011014-DO7] X (prog sel from 01 1014-BO3)F3

 

[LP cyto X 01 lOl4-DO7] X (prog sel fi'om 01 1014-BO3)F3

 

[LP cyto X 01 1014-DO7] X (prog sel from 01 lOl4-BOB)F3

 

(short 3 X 99FL F2 selection Ol-lOl4-D07)F4

 

 

[LP cyto x 01 10144307] x (prog sel from 011014-803)F3

 

279

 

Table A-2 Description of breeding lines from Cornell including S. pennellii LAO716,

normal tomato lines, fixed lines and crosses between fixed lines and S. penellii 0716.

 

Ml

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

samples for GH ID and Types .
sarptple dips of crosses pedigree
1 055117-9 LA716 S. pennellii LA716 long internode parent
2 055117-9 LA716 S. pennellii LA716 long internode parent
3 055117-9 LA716 S. pennellii LA716 long internode parent
4 055117-9 LA716 S. pennellii LA716 longinternode parent
_ ([LPC X 011014-007] X (prog sel from 011014-
5 0770446 066076 7 803)F2)F4 '071025 (0660764)
_ ([LPC X 011014-DO7] X (prog sel from 011014-
6 077°“ 6 066076 7 BO3)F2)F4 '071025 (0660764)
_ ([LPC X 011014-DO7] X (prog sel from 011014-
7 0770446 066076 7 803)F2)F4 '071025 (066076-4)
_ ([LPC X 011014-DO7] X (prog sel from 011014-
8 0770446 065076 7 803)F2)F4 '071025 (066076-4)
([LPC X 011014-DO7] X (prog sel from 011014-
9 077048 071026 BO3)F2)F4
([LPC X 011014—DO7] X (prog sel from 011014-
10 077048 071026 BO3)F2)F4
([LPC X 011014-007] X (prog sel from 011014-
1 1 077048 071026 BO3)F2)F4
([LPC X 011014-DO7] X (prog sel from 011014-
12 077048 071026 3033me
_ ([LPC X 011014—DO7] X (prog sel from 011014-
13 077049-6 66060 5 BO3)F2)F4
_ ([LPC X 011014-DO7] X (prog sel from 011014-
14 077049-6 66060 5 BO3)F2)F4
_ ([LPC X 011014—DO7] X (prog sel from 011014-
15 077049-6 66060 5 303)F2)F4
_ ([LPC X 011014-DO7] X (prog sel from 011014-
16 0770496 66060 5 BO3)F2)F4
077051-6 0665‘” ([LPC X 0110144307] X (prog sel from 011014-
17 (also 071024 BO3)F2)F4 '071024
077051—6 066254-7 ([LPC X 011014-D07] X (prog sel from 011014-
18 (also 071024 BO3)F2)F4 '071024
0770515 066254'7 (ll-PC X 011014-DO71X (prog sel from 011014-
19 (also 071024 BO3)F2)F4 '071024

 

 

 

 

280

 

Table A-2 (continued)

I?

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

077051-6 066254-7 (also ([LPC X 011014-007] X (prog sel from 011014-
20 071024 BO3)F2)F4 '071024
21 0770526 066083 N033EB-1 (PH-2+PH-3)
22 0770526 066083 NC33EB-1 (PH-2+PH—3)
23 0770526 066083 N033EB-1 (PH-2+PH-3)
24 0770526 066083 NC33EB-1 (PH-2+PH-3)
077058-6 066112-8 also ([LPC X 011014-DO7] X (prog sel from 011014-
25 071021 803)F2)F3 '071021
077058 -6 066112-8 also ([LPC X O11014—DO7] X (prog sel from 011014-
26 071021 BO3)F2)F3 '071021
077058 -6 066112-8 also ([LPC X 011014-DO7] X (prog sel from 011014-
27 071021 803)F2)F3 '071021
0770586 066112-8 also ([LPC x 011014-0071x (prog sel from 011014-
28 071021 BO3)F2)F3 '071021
0770826 07102” x 071021-3 x LA716 long internode
0551176
29
0770826 07102” x 071021-3 x LA716 long internode
0551176
30
0770826 07102” x 071021-3 x LA716 long internode
31 0551176
0770826 071°21'3 x 071021-3 x LA716 long internode
32 0551176
0770836 0710263 x 071026-3 x LA716 long internode
33 055117-9
077083-6 0710233 x 071026-3 x LA716 long internode
34 0551176
0770836 07102647 x 071026-3 x LA716 long internode
35 0551176
0770836 0710263 x 071026-3 x LA716 long internode
36 055117-9

 

 

 

 

 

281

Table A-2 (continued)

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

1' 7- 077084-6 06606075 x 066060-5 x LA716 long internode
0551176
37
077084—6 0660606 x 066060—5 x LA716 long internode
0551176
38
077084-6 0660605 x 066060-5 x LA716 long internode
055117-9
39
' ' "077084-65 0660605 x 066060-5 x LA716 long internode
, 0551176
40 .
‘ 0770856 0710244 x 071024-1 x LA716 long internode
41 055117—9
- 70770856 0710244 x 071024—1 x LA716 long internode
. 0551176
42 . .
0770856 0710244 x 071024-1 x LA716 long internode
43 055117-9
077085-6 071024-1 x 071024-1 x LA716 long internode
0551176
44
- 0770866 0660764 x 066076-4 x LA716 long internode
45 0551176
0770866 0660764 x 0660764 x LA716 long internode
46 0551176
0770866 0660764 x 0660764 x LA716 long internode
47 055117-9
0770866 0660764 X 0660764 x LA716 long internode
48 055117-9
077087-6 0551134 X 0551134 x LA716 long internode
49 0551176
0770876 0551134 x 0551 131 x LA716 long internode
50 0551176
0770876 0551 13'1 X 055113-1 x LA716 long internode
51 055117-9

 

 

 

 

282

 

 

Table A-2 (continued)

 

 

 

 

 

 

 

 

 

0770876 0551 13'1 X 055113—1 x LA716 long internode
0551176
52
0770886 06112123 X NC33EB-1 x long internode LA716
0551176
53
0770886 06112143 x NC33EB—1 x long internode LA716
0551176
54
0770886 06112123 x NC33EB—1 x long internode LA716
55 0551176
0770886 06112143 x N033EB-1 x long internode LA716
0551176
56
061120-23 x
0770896 055177_9 LIL x LA716 long
57
061120-23 x
077089-6 055177_9 LIL x LA716 long
58
061120-23 x
077089-6 055177_9 LIL x LA716 long
59
061120-23 x
60 0770896 0551776 LIL x LA716 long

 

 

 

 

283

 

 

Bibliography

284

Bibliography

(1)

(2)
(3)

(4)

(5)

(6)
(7)
(8)
(9)

(10)

(11)

(12)

(13)

(14)

(15)

Pimentel, D.; Pimentel, M. American Journal of Clinical Nutrition 2003, 78.
660s-663s.

Wittstock, U.; Gershenzon, J. Current Opinion in Plant Biology 2002, 5, 300-307.

Theis, N.; Lerdau, M. International Journal of Plant Sciences 2003, 164, S93-
S102.

Duke, S. 0.; Canel, C.; Rimando, A. M.; Tellez, M. R.; Duke, M. V.; Paul, R. N.
Advances in Botanical Research Incorporating Advances in Plant Pathology, Vol
31 20002000, 31,121-151.

Fiehn, 0.; Kopka, J .; Dormann, P.; Altmann, T.; Trethewey, R. N.; Willmitzer, L.
Nature Biotechnology2000, 18, 1157-1161.

Schauer, N.; Femie, A. R. Trends in Plant Science 2006, I I , 508-516.
Xie, Z.; Kapteyn, J .; Gang, D. R. Plant Journal 2008, 54, 349-361.
Schilmiller, A. L.; Last, R. L.; Pichersky, E. Plant Journal 2008, 54, 702-711.

Fridman, B; Wang, J. H.; Iijima, Y.; Froehlich, J. E.; Gang, D. R.; Ohlrogge, J.;
Pichersky, E. Plant Cell 2005, I 7, 1252-1267.

Gang, D. R.; Wang, J. H.; Dudareva, N.; Nam, K. 11.; Simon, J. E.; Lewinsohn, E.;
Pichersky, E. Plant Physiology 2001, 125, 539-555.

Iijima, Y.; Gang, D. R.; Fridman, E.; Lewinsohn, E.; Pichersky, E. Plant
Physiology 2004, 1 3 4, 370-379.

Iijima, Y.; Davidovich-Rikanati, R.; Fridman, E.; Gang, D. R.; Bar, B;
Lewinsohn, E.; Pichersky, E. Plant Physiology 2004, 136, 3724-3736.

Suzuki, H.; Reddy, M. S. S.; Naoumkina, M.; Aziz, N.; May, G. D.; Huhman, D.
V.; Sumner, L. W.; Blount, J. W.; Mendes, P.; Dixon, R. A. Planta 2005, 220,
696-707.

Goossens, A.; Hakkinen, S. T.; Laakso, I.; Seppanen-Laakso, T.; Biondi, S.; De
Sutter, V.; Lammertyn, F.; Nuutila, A. M.; Soderlund, H.; Zabeau, M.; Inze, D.;
Oksman-Caldentey, K. M. Proceedings of the National Academy of Sciences of
the United States of America 2003, 100, 8595-8600.

Morikawa, T.; Mizutani, M.; Aoki, N.; Watanabe, B.; Saga, H.; Saito, S.; Oikawa,

A.; Suzuki, H.; Sakurai, N.; Shibata, D.; Wadano, A.; Sakata, K.; Ohta, D. Plant
Cell 2006, 18, 1008-1022.

285

(16)

(17)

(18)
(19)
(20)

(21)
(22)
(23)

(24)

(25)

(26)

(27)

(28)

(29)

(30)

(31)
(32)

(33)
(34)

Martzen, M. R.; McCraith, S. M.; Spinelli, S. L.; Torres, F. M.; Fields, S.;
Grayhack, E. J .; Phizicky, E. M. Science 1999, 286, 1153-1155.

Fiehn, 0.; Kopka, J .; Dorrnann, P.; Altmann, T.; Trethewey, R. N.; Willmitzer, L.
Nature Biotechnology 2001, 19, 173-173.

Koch, K. Current Opinion in Plant Biology 2004, 7, 23 5-246.
Mitchell-Olds, T.; Pedersen, D. Genetics 1998, 149, 739-747.

Trethewey, R. N.; Krotzky, A. J.; Willmitzer, L. Current Opinion in Plant
Biology 1999, 2, 83-85.

Trethewey, R. N. Current Opinion in Biotechnology 2001, 12, 135-138.
Oliver, D. J .; Nikolau, B.; Wurtele, E. S. Metabolic Engineering 2002, 4, 98-106.
Sumner, L. W.; Mendes, P.; Dixon, R. A. Phytochemistry 2003, 62, 817-836.

Roessner, U.; Luedemann, A.; Brust, D.; Fiehn, 0.; Linke, T.; Willmitzer, L.;
Femie, A. R. Plant Cell 2001, 13, 11-29.

Harrigan, G. G.; Stork, L. G.; Riordan, S. G.; Reynolds, T. L.; Ridley, W. P.;
Masucci, J. D.; Maclsaac, S.; Halls, S. C.; Orth, R.; Smith, R. G.; Wen, L.; Brown.

W. E.; Welsch, M.; Riley, R.; Mcfarland, D.; Pandravada, A.; Glenn, K. C.
Journal of A gricultural and Food Chemistry 2007, 55 , 6177-6185.

Trethewey, R. N. Current Opinion in Plant Biology 2004, 7, 196-201.

Oksman-Caldentey, K. M.; Saito, K. Current Opinion in Biotechnology 2005, 16.
174-179.

Wink, M. Theoretical and Applied Genetics 1988, 75, 225-233.

Last, R. L.; Jones, A. D.; Shachar-Hill, Y. Nature Reviews Molecular Cell
Biology 2007, 8, 167-174.

Fiehn, 0. Plant Molecular Biology 2002, 48, 155-171.
Devaux, P. G.; Homing, M. G.; Homing, E. C. Analytical Letters 1971, 4, 151-&.

Homing, E. C.; Homing, M. G. Journal of Chromatographic Science 197], 9,
129-&.

Sweeley, C. 0.; Homing, E. C. Nature 1960, 18 7, 144-145.

Oldiges, M.; Lutz, S.; Pflug, S.; Schroer, K.; Stein, N.; Wiendahl, C. Applied
Microbiology and Biotechnology 2007, 76, 495-51 1.

286

(35)

(36)

(3 7)

(38)

(39)

(40)
(41)

(42)

(43)

(44)

(45)

(46)

(47)

(48)
(49)

(50)

(51)

Bino, R. J .; Hall, R. D.; Fiehn, 0.; Kopka, J .; Saito, K.; Draper, J .; Nikolau, B. J .;
Mendes, P.; Roessner-Tunali, U.; Beale, M. H.; Trethewey, R. N.; Lange, B. M.;
Wurtele, E. S.; Sumner, L. W. Trends in Plant Science 2004, 9, 418-425.

Wagner, 0; Sefkow, M.; Kopka, J. Phytochemistry 2003, 62, 887-900.

Dettmer, K.; Aronov, P. A.; Hammock, B. D. Mass Spectrometry Reviews 2007,
26, 51-78.

Villas-Boas, S. G.; Bruheim, P. Omics-a Journal of Integrative Biology 2007, 11,
305-313.

Villas-Boas, S. G.; Mas, S.; Akesson, M.; Smedsgaard, J.; Nielsen, J. Mass
Spectrometry Reviews 2005, 24, 613-646.

Schwab, W. Phytochemistry 2003, 62, 837-849.

Wyss, R. Journal of Chromatography B-Biomedical Applications 1995, 671, 381-
425.

Besseau, S.; Hoffmann, L.; Geoffroy, P.; Lapierre, C.; Pollet, B.; Legrand, M.
Plant Cell 2007, 19, 148-162.

Yonekura-Sakakibara, K.; Tohge, T.; Matsuda, F .; Nakabayashi, R.; Takayama,
H.; Niida, R.; Watanabe-Takahashi, A.; Inoue, E.; Saito, K. Plant Cell 2008, 20,
2160-2176.

Cuesta-Rubio, 0.; Piccinelli, A. L.; Fernandez, M. C.; Hernandez, I. M.; Rosado,
A.; Rastrelli, L. Journal of Agricultural and Food Chemistry 2007, 55, 7502-7509.

Rieger, G.; Muller, M.; Guttenberger, H.; Bucar, F. Journal of Agricultural and
Food Chemistry 2008, 56, 9080-9086.

Yeh, T. F.; Morris, C. R.; Goldfarb, B.; Chang, H. M.; Kadla, J. F. Tree
Physiology 2006, 26, 1497-1503.

Wang, G.; Zhang, R.; Sun, Y.; Xie, K.; Ma, C. Chromatographia 2007, 65, 363-
366.

Ward, J. L.; Harris, C.; Lewis, J .; Beale, M. H. Phytochemistry 2003, 62, 949-957.

Bailey, N. J. 0; Stanley, P. D.; Hadfield, S. T.; Lindon, J. C.; Nicholson, J. K.
Rapid Communications in Mass Spectrometry 2000, 14, 679-684.

Lenz, E. M.; Wilson, I. D. Journal of Proteome Research 2007, 6, 443-458.

Want, E. J.; Nordstrom, A.; Morita, H.; Siuzdak, G. Journal of Proteome
Research 2007, 6, 459-468.

287

(52)

(53)

(54)

(55)

(56)

(57)

(58)
(59)

(60)

(61)

(62)

(63)

(64)

(65)

(66)

(67)

Marshall, A. G.; Hendrickson, C. L.; Jackson, G. S. Mass Spectrometry Reviews
1998, 17, 1-35.

Brown, S. C.; Kruppa, G.; Dasseux, J. L. Mass Spectrometry Reviews 2005, 24,
223-231.

Birkemeyer, C.; Kolasa, A.; Kopka, J. Journal of Chromatography A 2003, 993,
89—102.

Broeckling, C. D.; Huhman, D. V.; Farag, M. A.; Smith, J. T.; May, G. D.;
Mendes, P.; Dixon, R. A.; Sumner, L. W. Journal of Experimental Botany 2005,
56, 323-336.

Schauer, N.; Steinhauser, D.; Strelkov, S.; Schomburg, D.; Allison. G.; Moritz, T.;
Lundgren, K.; Roessner-Tunali, U.; Forbes, M. G.; Willmitzer, L.; Femie, A. R.;
Kopka, J. Febs Letters 2005, 5 79, 1332-1337.

Cook, D.; Fowler, 8.; F iehn, 0.; Thomashow. M. F. Proceedings of the National
Academy of Sciences of the United States of A merica 2004, 101, 15243-15248.

Tolstikov, V. V.; Fiehn, 0. Analytical Biochemistry 2002, 301, 298-307.
Huhman, D. V.; Sumner, L. W. Phytochemistry 2002, 59, 347-360.

Femie, A. R.; Trethewey, R. N.; Krotzky, A. J.; Willmitzer, L. Nature Reviews
Molecular Cell Biology 2004, 5, 763-769.

Takats, Z.; Wiseman, J. M.; Gologan, B.; Cooks. R. G. Science 2004, 306, 471-
473.

Bedair, M.; Sumner, L. W. T rac-Trends in Analytical Chemistry 2008, 27, 238-
250.

Pongsuwan, W.; Bamba, T.; Harada, K.; Yonetani, T.; Kobayashi, A.; Fukusaki,
B. Journal of Agricultural and Food Chemistry 2008, 56, 10705-10708.

Shen, Y. F.; Zhang, R.; Moore, R. J .; Kim, J .; Metz, T. 0.; Hixson, K. K.; Zhao,
R.; Livesay, E. A.; Udseth, H. R.; Smith, R. D. Analytical Chemistry 2005, 77,
3090-3100.

Wilson, I. D.; Nicholson, J. K.; Castro-Perez, J.; Granger, J. H.; Johnson, K. A.;
Smith, B. W.; Plumb, R. S. Journal of Proteome Research 2005, 4, 591-598.

Nordstrom, A.; O'Maille, G.; Qin, C.; Siuzdak, G. Analytical Chemistry 2006, 78,
3289-3295.

Hanhineva, K.; Rogachev, I.; Kokko, H.; Mintz-Oron, S.; Venger, 1.; Karenlampi,
S.; Aharoni, A. Phytochemistry 2008, 69, 2463-2481.

288

(68)

(69)

(70)

(71)

(72)

(73)

(74)

(75)

(76)

(77)

(78)

(79)

(30)

(81)

(82)

(83)

(84)

Goodacre, R.; Vaidyanathan, S.; Dunn, W. B.; Harrigan, G. G.; Kell, D. B. Trends
in Biotechnology 2004, 22, 245-252.

Chan, E. C. Y.; Yap, S. L.; Lau, A. J.; Leow, P. C.; Toh, D. F.; Koh, H. L. Rapid
Communications in Mass Spectrometry 2007, 21, 519-528.

Xie, G. X.; Plumb, R.; Su, M. M.; Xu, Z. H.; Zhao, A. H.; Qiu, M. F.; Long, X. B.;
Liu, Z.; Jia, W. Journal of Separation Science 2008, 31, 1015-1026.

Zaikin, V. G.; Halket, J. M. European Journal of Mass Spectrometry 2006, 12,
79-115.

Fenn, J. B.; Mann, M.; Meng, C. K.; Wong, S. F.; Whitehouse, C. M. Science
1989, 246, 64-71.

Wilm, M.; Mann, M. Analatical Chemistry 1996, 68, 1-8.

Femandez-Alba, A. R.; Garcia-Reyes, J. F. True-Trends in Analytical Chemistry
2008, 27, 973-990.

Ferrer, I.; Thurman, E. M. Trac-Trends in Analytical Chemistry 2003, 22, 750-
756.

Mihaleva, V. V.; Vorst, 0.; Maliepaard, C.; Verhoeven, H. A.; de Vos, R. C. H.;
Hall, R. D.; van Ham, R. C. H. J. Metabolomics 2008, 4, 171-182.

Xie, G. X.; Ni, Y.; Su, M. M.; Zhang, Y. Y.; Zhao, A. H.; Gao, X. F.; Liu, Z.;
Xiao, P. G.; Jia, W. Metabolomics 2008, 4, 248-260.

Senyuva, H. Z.; Gilbert, J .; Ozturkoglu, S. Analytica Chimica Acta 2008, 617, 97-
106.

Trygg, J. Journal of Chemometrics 2002, 16, 283-293.
Trygg, J .; Wold, S. Journal of Chemometrics 2002, 16, 119-128.

Idborg, H.; Zamani, L.; Edlund, P. 0.; Schuppe-Koistinen, I.; Jacobsson, S. P.
Journal of Chromatography B—Analytical Technologies in the Biomedical and
Life Sciences 2005, 828, 14-20.

Lenz, E. M.; Bright, J.; Knight, R.; Westwood, F. R.; Davies, D.; Major, 11.;
Wilson, I. D. Biomarkers 2005, 10, 173-187.

Whitfield, P. D.; Noble, P. J. M.; Major, H.; Beynon, R. J .; Burrow, R.; Freeman,
A. I.; German, A. J. Metabolomics 2005, 1, 215-225.

Griffiths, W. J .; Wang, Y. Q. Chemical Society Reviews 2009, 38, 1882-1896.

289

(85)

(86)

(87)

(88)

(89)

(90)

(91)

(92)

(93)

(94)

(95)

(96)

(97)

(98)

Trygg, J .; Holmes, E.; Lundstedt, T. Journal of Proteome Research 2007, 6, 469—
479.

Wang, Y. L.; Bollard, M. E.; Keun, H.; Antti, H.; Beckonert, 0.; Ebbels, T. M.;
Lindon, J. 0.; Holmes, E.; Tang, H. R.; Nicholson, J. K. Analytical Biochemistry
2003, 323, 26-32.

Plumb, R. S.; Stumpf, C. L.; Gorenstein, M. V.; Castro-Perez, J. M.; Dear, G. J .;
Anthony, M.; Sweatman, B. 0; Connor, S. C.; Haselden, J. N. Rapid
Communications in Mass Spectrometry 2002. 16, 1991-1996.

Kim, S. W.; Ban, S. H.; Ahn, C. Y.; Oh, H. M.; Chung, H.; Cho, S. 11.; Park, Y.
M.; Liu, J. R. Journal of Plant Biology 2006, 49, 271-275.

Xiao, C. N.; Dai, H.; Liu, H. B.; Wang, Y. L.; Tang, H. R. Journal of Agricultural
and Food Chemistry 2008, 56, 10142—10153.

Consonni, R.; Cagliani, L. R.; Stocchero, M.; Porretta. S. Journal of Agricultural
and Food Chemistry 2009, 5 7, 4506-4513.

Wold, S.; Ruhe, A.; Wold, H.; Dunn, W. J. Siam Journal on Scientific and
Statistical Computing 1984, 5, 735-743.

Rantalainen, M.; Cloarec, 0.; Beckonert, 0.; Wilson, I. D.; Jackson, D.; Tonge,
R.; Rowlinson, R.; Rayner, S.; Nickson, J .; Wilkinson, R. W.; Mills, J. D.; Trygg,
J.; Nicholson, J. K.; Holmes, B. Journal of Proteome Research 2006, 5, 2642-
2655.

Stella, C.; Beckwith-Hall, B.; Cloarec, 0.; Holmes, E.; Lindon, J. C.; Powell, J.;
van der Ouderaa, F.; Bingharn, 8.; Cross, A. J.; Nicholson, J. K. Journal of
Proteome Research 2006, 5, 2780-2788.

Moco, S.; Bino, R. J.; De Vos, R. C. H.; Vervoort, J. Trac-Trends in Analytical
Chemistry 2007, 26, 855-866.

Grange, A. H.; Winnik, W.; Ferguson, P. L.; Sovocool, G. W. Rapid
Communications in Mass Spectrometry 2005, 19, 2699-2715.

Kind, T.; Fiehn, O. Bmc Bioinformatics 2006, 7, 234.

Bateman, R. H.; Carruthers, R.; Hoyes, J. B.; Jones, C.; Langridge, J. 1.; Millar,
A.; Vissers, J. P. C. Journal of the American Society for Mass Spectrometry 2002,
13, 792-803.

Kosaka, T.; Yoneyama-Takazawa, T.; Kubota, K.; Matsuoka, T.; Sato, 1.; Sakaki,
T.; Tanaka, Y. Journal of Mass Spectrometry 2003, 38, 1281-1287.

290

(99)

(100)

(101)

(102)

(103)

(104)

(105)

(106)
(107)

(108)
(109)

(110)

(111)
(112)
(113)

(114)

(115)

Plumb, R. S.; Johnson, K. A.; Rainville, P.; Smith, B. W.; Wilson, I. D.; Castro-
Perez, J. M.; Nicholson, J. K. Rapid Communications in Mass Spectrometry 2006,
20, 1989-1994.

Bottcher, 0; von Roepenack-Lahaye, E.; Schmidt, J .; Schmotz, C.; Neumann. S.;
Scheel, D.; Clemens, S. Plant Physiology 2008, 147, 2107-2120.

Rochfort, S. J .; Trenerry, V. C.; Imsic, M.; Panozzo, J .; Jones, R. Phytochemistry
2008, 69, 1671-1679.

Schauer, N.; Semel, Y.; Roessner, U.; Gur, A.; Balbo, I.; Carrari, F.; Pleban, T.;
Perez-Melis, A.; Bruedigam, C.; Kopka, J .; Wilhnitzer, L.; Zamir, D.; Femie, A.
R. Nature Biotechnology 2006, 24, 447-454.

Moco, S.; Bino, R. J .; Vorst, 0.; Verhoeven, H. A.; de Groot, J .; van Beck, T. A.;
Vervoort, J .; de Vos, C. H. R. Plant Physiology 2006, 141. 1205-1218.

Dayan, F. E.; Watson, S. B.; Nanayakkara, N. P. D. Journal of Experimental
Botany 2007, 58, 3263-3272.

Werker, E. Advances in Botanical Research Incorporating Advances in Plant
Pathology, V0131 2000 2000, 31, 1-35.

Wagner, G. J. Plant Physiology 1991, 96, 675-679.

Kandra, L.; Severson, R.; Wagner, G. J. European Journal of Biochemistry 1990,
188, 385-391.

Kroumova, A. B.; Wagner, G. J. Planta 2003, 216, 1013-1021.

Kroumova, A. B.; Xie, Z. Y.; Wagner, G. J. Proceedings of the National Academy
of Sciences of the United States of America 1994, 91, 11437-11441.

Li, A. X.; Eannetta, N.; Ghangas, G. s.; Steffens, J. C. Plant Physiology 1999,
121, 453-460.

van der Hoeven, R. S.; Steffens, J. C. Plant Physiology 2000, 122, 275-282.
Walters, D. S.; Steffens, J. C. Plant Physiology 1990, 93, 1544-1551.

King, R. R.; Calhoun, L. A.; Singh, R. P.; Boucher, A. Journal of Agricultural
and Food Chemistry 1993, 41, 469-473.

Severson, R. F.; Jackson, D. M.; Johnson, A. W.; Sisson, V. A.; Stephenson, M.
G. Acs Symposium Series 1991, 449, 264-277.

King, R. R.; Calhoun, L. A.; Singh, R. P.; Boucher, A. Phytochemistry 1990, 29,
2115-2118.

291

(116)
(117)

(118)

(119)

(120)

(121)

(122)

(123)

(124)

(125)

(126)
(127)
(128)

(129)

(130)

(131)
(132)

(133)

King, R. R.; Calhoun, L. A. Phytochemistry 1988, 27, 3761-3763.
King, R. R.; Calhoun, L. A.; Singh, R. P. Phytochemistry 1988, 27, 3765-3 768.

King, R. R.; Singh, R. P.; Calhoun, L. A. Carbohydrate Research 1987, I66, 113-
121.

Slocombe, S. P.; Schauvinhold, 1.; McQuinn, R. P.; Besser, K.; Welsby, N. A.;
Harper, A.; Aziz, N.; Li, Y.; Larson, T. R.; Giovannoni, J .; Dixon, R. A.; Broun, P.
Plant Physiology 2008, 148, 1830-1846.

Yu, F. N. A.; Utsumi, R. Cellular and Molecular Life Sciences 2009, 66, 3043-
3052.

Dudareva, N.; Andersson, S.; Orlova, 1.; Gatto, N.; Reichelt, M.; Rhodes, D.;
Boland, W.; Gershenzon, J. Proceedings of the National Academy of Sciences of
the United States of America 2005, 102, 933-938.

Chappell, J. Annual Review of Plant Physiology and Plant Molecular Biology
1995, 46, 521-547.

Dicke, M.; van Loon, J. J. A. Entomologia Experimentalis Et Applicata 2000, 97,
237—249.

Valkama, E.; Salminen, J. P.; Koricheva. J .; Pihlaja, K. Annals of Botany 2004, 94,
233-242.

Wollenweber, E.; Mann, K. Zeitschrift Fur Naturforschung C-a Journal of
Biosciences 1984, 39, 303-306.

Voirin, B.; Bayet, C.; Colson, M. Phytochemistry 1993, 34, 85-87.
Friedman, M. Journal of Agricultural and Food Chemistry 2002, 50, 5751-5780.
Kennedy, G. G. Annual Review of Entomology 2003, 48, 51-72.

Carter, C. D.; Gianfagna, T. J.; Sacalis, J. N. Journal of Agricultural and Food
Chemistry 1989, 37, 1425-1428.

Li, L.; Zhao, Y. F.; McCaig, B. C.; Wingerd, B. A.; Wang, J. H.; Whalon, M. E.;
Pichersky, E.; Howe, G. A. Plant Cell 2004, 16, 783-783.

Eshed, Y.; Zamir, D. Genetics 1995. 141, 1147-1162.

Goffreda, J. C.; Steffens, J. C.; Mutschler, M. A. Journal of the American Society
for Horticultural Science 1990, 115, 161-165.

DellaPenna, D.; Last, R. L. Science 2008, 320. 479-481.

292

(134)
(135)

(136)

(137)

(138)

(139)

(140)

(141)

(142)

(143)

(144)

(145)

(146)

(147)

(148)

Want, E. J .; Cravatt, B. F.; Siuzdak, G. Chembiochem 2005, 6, 1941-1951.

Tiller, P. R.; Land, A. P.; Jardine. 1.; Murphy, D. M.; Sozio, R.; Ayrton, A.;
Schaefer, W. H. Journal of ChromatographyA 1998, 794, 15-25.

Blackler, A. R.; Klammer, A. A.; MacCoss, M. J.; Wu, C. C. Analytical
Chemistry 2006, 78, 1337-1344.

Spahr, C. S.; Davis, M. T.; McGinley, M. D.; Robinson, J. H.; Bures, E. J.;
Beierle, J .; Mort, J .; Courchesne, P. L.; Chen, K.; Wahl, R. 0.; Yu, W.; Luethy, R.;
Patterson, S. D. Proteomics 2001, 1, 93-107.

Williams, J. D.; Flanagan, M.; Lopez, L.; Fischer, 8.; Miller, L. A. D. Journal of
ChromatographyA 2003, 1020, 11-26.

Chakraborty, A. B.; Berger, S. J.; Gebler, J. C. Rapid Communications in Mass
Spectrometry 2007, 21, 730-744.

Wang, Y. S.; Vivekananda, 8.; Men, L. J .; Zhang, Q. B. Journal of the American
Society for Mass Spectrometry 2004, 15, 697-702.

Seymour, J. L.; Costello, C. E.; Zaia, J. Journal of the American Societyfor Mass
Spectrometry 2006, 17, 844-854.

Zaia, J.; Miller, M. J. C.; Seymour, J. L.; Costello, C. E. Journal ofthe American
Societyfor Mass Spectrometry 2007, 18, 952-960.

Lips, A. G. A. M.; Lameijer, W.; Fokkens, R. H.; Nibbering, N. M. M. Journal of
Chromatography B 2001, 759, 191-207.

Tian, Q. G.; Duncan, C. J. G.; Schwartz, S. J. Journal of Mass Spectrometry 2003,
38, 990-995.

Bateman, K. P.; Castro-Perez, J .; Wrona, M.; Shockcor, J. P.; Yu, K.; Oballa, R.;
Nicoll-Griffith, D. A. Rapid Communications in Mass Spectrometry 2007, 21,
1485-1496.

Rainville, P. D.; Stumpf, C. L.; Shockcor, J. P.; Plumb, R. S.; Nicholson, J. K.
Journal of Proteome Research 2007, 6, 552-558.

Silva, J. C.; Denny, R.; Dorschel, C. A.; Gorenstein, M.; Kass, I. J.; Li, G. 2.;
McKenna, T.; Nold, M. J .; Richardson, K.; Young, R; Geromanos, S. Analytical
Chemistry 2005, 77, 2187-2200.

Ramos, A. A.; Yang, H.; Rosen, L. E.; Yao, X. D. Analytical Chemistry 2006, 78,
6391-6397. 7

293

(149)

(150)

(151)

(152)

(153)

(154)

(155)

(156)
(157)
(158)

(159)

(160)
(161)

(162)

(163)
(164)

(165)

(166)

Wrona, M.; Mauriala, T.; Bateman, K. P.; Mortishire-Smith, R. J.; O'Connor, D.
Rapid Communications in Mass Spectrometry 2005, 19, 2597-2602.

Cheng, F. Y.; Blackburn, K.; Lin, Y. M.; Goshe, M. B.; Williamson, J. D. Journal
of Proteome Research 2009, 8, 82-93.

Zhang, H. Y.; Grubb, M.; Wu, W.; Josephs, J.; Humphreys, W. G. Analytical
Chemistry 2009, 81, 2695-2700.

Grata, E.; Boccard, J.; Guillarme, D.; Glauser, G.; Carrupt, P. A.; Farmer, E. E.;
Wolfender, J. L.; Rudaz, S. Journal of Chromatography B-Analytical
Technologies in the Biomedical and Life Sciences 2008, 871, 261-270.

Kaufmann, A.; Butcher, P.; Maden, K.; Widmer, M. Analytica Chimica A cta 2007,
586, 13-21.

Glauser, G.; Grata, E.; Rudaz, S.; Wolfender, J. L. Rapid Communications in
Mass Spectrometry 2008, 22, 3154-3160.

Weinmann, W.; Wiedemann, A.; Eppinger. B.; Renz, M.; Svoboda, M. Journal of
American Society for Mass Spectrometry 1999, 10, 1028-1 03 7.

Kind, T.; F iehn, O. Bmc Bioinformatics 2006, 7.
Kind, T.; F iehn, O. BMC Bioinformatics 2007, 8, 105.

Stein, S. E. Journal of the American Society for Mass Spectrometry 1999, 10,
770-781.

Mei, H.; Hsieh, Y. S.; Nardo, C.; Xu, X. Y.; Wang, S. Y.; Ng, K.; Korfmacher, W.
A. Rapid Communications in Mass Spectrometry 2003, 17, 97-103.

Wagner, G. J .; Wang, R.; Shepherd, R. W. Annals of Botany 2004, 93, 3-11.

Puterka, G. J.; Farone, W.; Palmer, T.; Barrington, A. Journal of Economic
Entomology 2003, 96, 636-644.

Xia, Y. L.; Johnson, A. W.; Chortyk, O. T. Journal of Economic Entomology
1997, 90,1015-1021.

Hill, K.; Rhode, O. Fett-Lipid 1999, 101, 25-33.
Burke, B. A.; Goldsby, G.; Mudd, J. B. Phytochemistry 1987, 26, 2567-2571.

King, R. R.; Singh, R. P.; Boucher, A. American Potato Journal 1987, 64, 529-
534.

Spring, 0.; Heil, N.; Vogler. B. Phytochemistry 1997, 46, 1369-1373.

294

(167)

(168)

(169)

(170)

(171)

(172)

(173)

(174)

(175)

(176)

(177)

(178)

(179)

(180)

(181)

(182)

Wollenweber, E.; Stevens, J. F.; Ivanic, M.; Deinzer, M. L. Phytochemistry 1998,
48, 931-939.

Asai, T.; Hara, N.; Kobayashi, S.; Kohshima, S.; Fujimoto, Y. Helvetica Chimica
Acta 2009, 92, 1473-1494.

Shapiro, J. A.; Steffens, J. C.; Mutschler, M. A. Biochemical Systematics and
Ecology 1994, 22, 545-561.

SerratoValenti, G.; Bisio, A.; Comara, L.; Ciarallo, G. Annals of Botany 1997, 79,
329-336.

Juo, C. G.; Chiu, D. T. Y.; Shiao, M. S. Biofactors 2008, 34, 159-169.

Yang, S.; Sadilek, M.; Synovec, R. E.; Lidstrom, M. E. Journal of
Chromatography A 2009, 1216, 3280-3289.

Griffiths, W. J .; Karua, K.; Homshaw, M.; Woffendin. G.; Wanga, Y. European
Journal of Mass Spectrometry 2007, 13, 45-50.

Stobiecki, M.; Skirycz, A.; Kerhoas, L.; Kachlicki, P.; Muth, D.; Einhom, J.;
Mueller-Roeber, B. Metabolomics 2006, 2, 197-219.

Farag, M. A.; Huhman, D. V.; Lei, Z. T.; Sumner, L. W. Phytochemistry 2007, 68,
342-354.

Kong, H.; Wang, M.; Venema, K.; Maathuis, A.; van der Heijden, R.; van der
Greef, J.; Xu, G.; Hankemeier, T. Journal of Chromatography A 2009, 1216,
2195-2203.

Nakamura, Y.; Kanaya, S.; Sakurai, N.; Iijima, Y.; Aoki, K.; Okazaki, K.; Suzuki,
H.; Kitayama, M.; Shibata, D. Plant Biotechnology 2008, 25, 377-380.

Holman, S. W.; Wright, P.; Langley, G. J. Rapid Communications in Mass
Spectrometry 2008, 22, 2355-2365.

Kachlicki, P.; Marczak, L.; Kerhoas, L.; Einhom, J.; Stobiecki, M. Journal of
Mass Spectrometry 2005, 40, 1088-1 103.

Li, A. X.; Steffens, J. C. Proceedings of the National Academy of Sciences of the
United States of America 2000, 97, 6902-6907.

Hsu, F. F .; Turk, J. Journal of the American Society for Mass Spectrometry 2000,
11, 892-899.

Hsu, F. F .; Turk, J. Journal of the American Society for Mass Spectrometry 2000,
11, 797-803.

295

(183)

(184)

(185)
(186)
(187)

(188)
(189)
(190)

(191)

(192)

(193)

(194)

(195)

(196)

(197)

(198)

(199)

Ding, L.; Xie, F. W.; Zhao, M. Y.; Xie, J. P.; Xu, G. W. Rapid Communications
in Mass Spectrometry 2006, 20, 2816-2822.

Matsuda, F.; Yonekura-Sakakibara, K.; Niida, R.; Kuromori, T.; Shinozaki, K.;
Saito. K. Plant Journal 2009, 5 7, 555-577.

Friedman, M. Journal of Agricultural and Food Chemistry 1997, 45, 1523-1540.
Barbour,.J. D.; Kennedy, G. G. Journal of Chemical Ecology 1991, 1 7, 989-1005.

Schijlen, E. G. W.; de Vos, C. H. R.; van Tunen, A. J.; Bovy, A. G.
Phytochemistry 2004, 65, 2631-2648.

Bovy, A.; Schijlen, E.; Hall, R. D. Metabolomics 2007, 3, 399-412.
Horper, W.; Marner, F. J. Phytochemistry 1996, 41, 451-456.

Lazar, G.; Goodman, H. M. Proceedings of the National Academy of Sciences of
the United States of America 2006, 103, 472-476.

McCue, K. F.; Shepherd, L. V. T.; Rockhold, D. R.; Allen, P. V.; Davies, H. V.;
Belknap, W. R. Abstracts of Papers of the American Chemical Society 2005, 229,
U83-U83.

Friedman, M. Journal of Agricultural and Food Chemistry 2006, 54, 8655-8681.

McCue, K. F.; Shepherd, L. V. T.; Allen, P. V.; Maccree, M. M.; Rockhold, D. R.;
Corsini, D. L.; Davies, H. V.; Belknap, W. R. Plant Science 2005, 168, 267-273.

Colby, S. M.; Crock, J .; Dowdle-Rizzo, B.; Lemaux, P. G.; Croteau, R.
Proceedings of the National Academy of Sciences of the United States of America
1998, 95, 2216-2221.

van Schie, C. C. N.; Haring, M. A.; Schuurink, R. C. Plant Molecular Biology
2007, 64, 251-263.

Coates, R. M.; Ho, J. 2.; Klobus, M.; Zhu, L. J. Journal of Organic Chemistry
1998, 63, 9166-9176.

Besser, K.; Harper, A.; Welsby, N.; Schauvinhold, 1.; Slocombe, S.; Li, Y.; Dixon,
R. A.; Broun, P. Plant Physiology 2009, 149, 499-514.

Iijima, Y.; Nakamura, Y.; Ogata, Y.; Tanaka, K.; Sakurai, N.; Suda, K.; Suzuki,
T.; Suzuki, H.; Okazaki, K.; Kitayama, M.; Kanaya, S.; Aoki, K.; Shibata, D.
Plant Journal 2008, 54, 949-962.

Tohge, T.; Nishiyama, Y.; Hirai, M. Y.; Yano, M.; Nakajima, J.; Awazuhara, M.;

Inoue, B.; Takahashi, H.; Goodenowe, D. B.; Kitayama, M.; Noji, M.; Yamazaki,
M.; Saito, K. Plant Journal 2005, 42, 218-235.

296

(200)

(201)

(202)

(203)

(204)

(205)

(206)

(207)

(208)

(209)

(210)

(211)
(212)
(213)

(214)

(215)

Moco, S.; Forshed, J .; De Vos, R. C. H.; Bino, R. J .; Vervoort, J. Metabolomics
2008, 4, 202-215.

Cuyckens, F.; Claeys, M. Journal of Mass Spectrometry 2004, 39, 1-15.

Moco, S.; Capanoglu, B.; Tikunov, Y.; Bino, R. J.; Boyacioglu, D.; Hall, R. D.;
Vervoort, J .; De Vos, R. C. H. Journal of Experimental Botany 2007, 58, 4131-
4146.

Tohge, T.; Matsui, K.; Ohme-Takagi, M.; Yamazaki, M.; Saito, K. Biotechnology
Letters 2005, 27, 297-303.

Tohge, T.; Yonekura—Sakakibara, K.; Niida, R.; Watanabe-Takahashi, A.; Saito,
K. Pure and Applied Chemistry 2007 , 79, 811-823.

Lepiniec, L.; Debeaujon, 1.; Routaboul, J. M.; Baudry, A.; Pourcel, L.; Nesi, N.;
Caboche, M. Annual Review of Plant Biology 2006, 5 7, 405-430.

Winkel-Shirley, B. Plant Physiology 2001, 126, 485-493.

Ma, Y. L.; Li, Q. M.; VandenHeuvel, H.; Claeys, M. Rapid Communications in
Mass Spectrometry 1997, 11, 1357-1364.

Cataldi, T. R. I.; Lelario, F.; Bufo, S. A. Rapid Communications in Mass
Spectrometry 2005, 19, 3 103 -31 10.

Claeys, M.; VandenHeuvel, H.; Chen, S.; Derrick, P. J .; Mellon, F. A.; Price, K. R.
Journal of the American Society for Mass Spectrometry 1996, 7, 173-181.

Dal Piaz, F.; Imparato, S.; Lepore, L.; Bader, A.; De Tommasi, N. Journal of
Pharmaceutical and Biomedical Analysis 2010, 51, 70-77.

Ungur, ‘N.; Kulcitki, V. Tetrahedron 2009, 65, 3815-3828.
Schilmiller, A. L.; Last, R. L.; Pichersky, E. Plant Journal 2008, 54, 702-711.
Wagner, G. J. Plant Physiol 1991, 96, 675-679.

Iijima, Y.; Davidovich-Rikanati, R.; Fridman, B.; Gang, D. R.; Bar, B.;
Lewinsohn, E.; Pichersky, E. Plant Physi012004, 136, 3724-3736.

Croteau, R. B.; Davis, E. M.; Ringer, K. L.; Wildung, M. R. Naturwissenschaften
2005, 92, 562-577.

(216) Nagel, J.; Culley, L. K.; Lu, Y.; Lin, B.; Matthews, P. D.; Stevens, J. F .; Page, J. E.

(217)

Plant Cell 2008, 20, 186-200.

Gang, D. R.; Wang, J.; Dudareva, N.; Nam, K. H.; Simon. J. B.; Lewinsohn, E.;
Pichersky, E. Plant Physiol 2001, 125, 539-555.

297

(218)

(219)
(220)

(221)

(222)

(223)

(224)

(225)
(226)

(227)

(228)

(229)

(230)

(231)

(232)

Schilmiller, A. L.; Schauvinhold, 1.; Larson, M.; Xu, R.; Charbonneau, A. L.;
Schmidt, A.; Wilkerson, C.; Last, R. L.; Pichersky, E. Proceedings of the
National Academy of Sciences of the United States of America 2009, 106, 10865-
10870.

Benning, C. Analytical Biochemistry 2004, 332, 1.
Eshed, Y.; Zamir, D. Genetics 1995, 141, 1147-1162.

Liu, Y. S.; Gur, A.; Ronen, G.; Causse, M.; Damidaux, R.; Buret, M.; Hirschberg,
J .; Zamir, D. Plant Biotechnol J 2003, 1, 195-207.

Ronen, G.; Carmel-Goren, L.; Zamir, D.; Hirschberg, J. Proceedings of the
National Academy of Sciences of the United States of America 2000, 97, 1 1102-
11107.

Schilmiller, A. L.; Schauvinhold, 1.; Larson, M.; Xu, R.; Charbonneau, A. L.;
Schmidt, A.; Wilkerson, C.; Last, R. L.; Pichersky, E. Proceedings of the
National Academy of Sciences of the United States of America 2009, 106, 10865-
10870.

Overy, S. A.; Walker, H. J .; Malone, 8.; Howard, T. P.; Baxter, C. J .; Sweetlove,
L. J .; Hill, S. A.; Quick, W. P. Journal of Experimental Botany 2005, 56, 287-296.

Hall, R. D. New Phytologist 2006, 169, 453-468.

Roessner, U.; Willmitzer, L.; Femie, A. R. Plant Cell Reports 2002, 21, 189-196.
Hirai, M. Y.; Klein, M.; Fujikawa, Y.; Yano, M.; Goodenowe, D. B.; Yamazaki,
Y.; Kanaya, S.; Nakamura, Y.; Kitayama, M.; Suzuki, H.; Sakurai, N.; Shibata, D.;
Tokuhisa, J .; Reichelt, M.; Gershenzon, J .; Papenbrock, J .; Saito, K. Journal of
Biological Chemistry 2005, 280, 25590-25595.

Goodacre, R.; York, E. V.; Heald, J. K.; Scott, I. M. Phytochemistry 2003, 62,
859-863.

Jander, G.; Norris, S. R.; Joshi, V.; Fraga, M.; Rugg, A.; Yu, S. X.; Li, L. L.; Last,
R. L. Plant Journal 2004, 39, 465-475.

Tolstikov, V. V.; Lommen, A.; Nakanishi, K.; Tanaka, N.; Fiehn, 0. Analytical
Chemistry 2003, 75, 6737-6740.

von Roepenack-Lahaye, B.; Degenkolb, T.; Zerjeski, M.; Franz, M.; Roth, U.;
Wessjohann, L.; Schmidt, J .; Scheel, D.; Clemens, S. Plant Physiology 2004, 134,
548-559.

Vorst, 0.; de Vos, C. H. R.; Lommen, A.; Staps, R. V.; Visser, R. G. F.; Bino, R.
J .; Hall, R. D. Metabolomics 2005, 1, 169-180.

298

(233)

(234)

(235)

(236)

(237)
(238)

(239)

(240)

(241)

(242)

(243)
(244)

(245)

(246)

(247)

(248)

(249)

Rischer, H.; Oresic, M.; Seppanen-Laakso, T.; Katajamaa, M.; Lammertyn, F.;
Ardiles-Diaz, W.; Van Montagu, M. C. E.; Inze, D.; Oksman-Caldentey, K. M.;
Goossens, A. Proceedings of the National Academy of Sciences of the United
States of America 2006, 103, 5614-5619.

Sato, S.; Soga, T.; Nishioka, T.; Tomita, M. Plant Journal 2004, 40, 151-163.

Le Gall, G.; Colquhoun, I. J.; Davis, A. L.; Collins, G. J.; Verhoeyen, M. E.
Journal of Agricultural and Food Chemistry 2003, 51, 2447-2456.

De Vos, R. C. H.; Moco, S.; Lommen, A.; Keurentjes, J. J. B.; Bino, R. J .; Hall, R.
D. Nature Protocols 2007, 2, 778-791.

Smedsgaard, J.; Frisvad, J. C. Journal of Microbiological Methods 1996, 25, 5-17.
Smedsgaard, J. Journal of Chromatography A 1997. 760, 264-270.

Allen, J .; Davey, H. M.; Broadhurst, D.; Heald, J. K.; Rowland, J. J .; Oliver, S. G.;
Kell, D. B. Nature Biotechnology 2003, 21 , 692-696.

Castrillo, J. 1.; Hayes, A.; Mohammed, S.; Gaskell, S. J.; Oliver, S. G.
Phytochemistry 2003, 62, 929-937.

Kirkland, J. J .; Langlois, T. J .; DeStefano, J. J. American Laboratory 2007, 39,
18-21.

Cunliffe, J. M.; Adams-Hall, S. B.; Maloney, T. D. Journal of Separation Science
2007, 30, 1214-1223.

Salisbury, J. J. Journal of Chromatographic Science 2008, 46, 883-886.
Fobes, J. F.; Mudd, J. B.; Marsden, M. P. Plant Physiol 1985, 77, 567-570.

Kandra, G.; Severson, R.; Wagner, G. J. European Journal of Biochemistry 1990,
188, 385—391.

van der Hoeven, R. S.; StefTens, J. C. Plant Physiol 2000, 122, 275-282.

Friedman, M.; Kozukue, N.; Harden, L. A. Journal of Agricultural and Food
Chemistry 1997, 45, 1541-1547.

Lawson, D. M.; Lunde, C. F .; Mutschler, M. A. Molecular Breeding 1997, 3, 307-
317.

Liedl, B. B.; Lawson, D. M.; White, K. K.; Shapiro, J. A.; Cohen, D. B.; Carson,
W. G.; Trumble, J. T.; Mutschler, M. A. Journal of Economic Entomology 1995,
88, 742-748.

299

(250)

(251)

(252)

(253)

(254)

(255)

(256)

(257)

(258)

(259)

(260)

Hawthome, D. J .; Shapiro, J. A.; Tingey, W. M.; Mutschler, M. A. Entomologia
Experimentalis et Applicata 1992, 65, 65-73.

Nonomura, T.; Xu, L.; Wada, M.; Kawamura, S.; Miyajima, T.; Nishitomi, A.;
Kakutani, K.; Takikawa, Y.; Matsuda, Y.; Toyoda, H. Plant Science 2009, 176,
31-37.

Fobes, J. F .; Mudd, J. B.; Marsden, M. P. F. Plant Physiology 1985, 77, 567-570.

Shinozaki, Y.; Matsuzaki, T.; Suhara, S.; Tobita, T.; Shigematsu, H.; Koiwai, A.
Agricultural and Biological Chemistry 1991, 55, 751-756.

Bonierbale. M. W.; Plaisted, R. L.; Pineda. 0.; Tanksley, S. D. Theoretical and
Applied Genetics 1994, 87, 973-987.

Chortyk, O. T.; Kays, S. J .; Teng, Q. Journal of A gricultural and Food Chemistry
1997, 45, 270-275.

McKenzie, C. L.; Weathersbee, A. A.; Puterka, G. J. Journal of Economic
Entomology 2005, 98, 1242-1247.

Chortyk, O. T.; Severson, R. F.; Cutler, H. C.; Sisson, V. A. Bioscience
Biotechnology and Biochemistry 1993, 5 7, 1355-1356.

Maldonado, B.; Torres, F. R.; Martinez, M.; Perez-Castorena, A. L. Journal of
Natural Products 2006, 69, 1511-1513.

Blauth, S. L.; Steffens, J. C.; Churchill, G. A.; Mutschler, M. A. Theoretical and
Applied Genetics 1999, 99, 373-381.

Blauth, S. L.; Churchill, G. A.; Mutschler, M. A. Theoretical and Applied
Genetics 1998, 96. 458-467.

300