Dissertation Diana Sirbu

Comprehensive Analysis of
Cocoa Lipidomics – unraveling the unknown

chemistry of cocoa butter
by
Diana Sirbu
a Thesis submitted in partial fulfillment

of the requirements for the degree of
Doctor of Philosophy
in Chemistry
Approved Dissertation Committee
Prof. Dr. Nikolai Kuhnert, Jacobs University

Bremen, Germany
Prof. Dr. Matthias Ullrich, Jacobs University

Bremen, Germany
Dr. Frédéric Depypere, Barry Callebaut

Lebbeke-Wieze, Belgium
Date of Defense: June 1st, 2018
Department of Life Sciences and Chemistry

Sirbu Diana
20331305
Ph. D. Thesis
07.09.2018
I dedicated this thesis to my family
for their encouragement, lifelong support and inconditional love.
Table of contents
Preface ...............................................................................................................................................................I
Abstract ........................................................................................................................................................... III
List of abbreviations........................................................................................................................................ VI
List of figures ............................................................................................................................................... VIII
List of tables .................................................................................................................................................. XV
Chapter 1: .......................................................................................................................................................... 1
1 Introduction .................................................................................................................................................... 1
1.1 Lipidomics ............................................................................................................................................. 1
1.1.1 General concept of Lipidomics........................................................................................................... 1
1.1.2 Lipid classification and functions ....................................................................................................... 1
1.1.3 Experimental approaches in lipidomics research................................................................................ 5
1.1.4 Plant lipidomics .................................................................................................................................. 7
1.1.5 Lipidomics and food science .............................................................................................................. 8
1.2 Theobroma cacao – from bean to bar .................................................................................................... 9
1.2.1 History, origin, and cocoa varieties .................................................................................................... 9
1.2.2 Agriculture ....................................................................................................................................... 10
1.2.3 Post-harvesting treatments ................................................................................................................ 11
1.2.4 Factory production process............................................................................................................... 12
1.2.5 Economic impact and health benefits ............................................................................................... 15
Chapter 2: ........................................................................................................................................................ 21
2 Review on cocoa lipidomics – state of knowledge and future needs ........................................................... 21
Abstract ........................................................................................................................................................... 22
2.1 Introduction ......................................................................................................................................... 23
2.2 From cocoa tree to cocoa butter ........................................................................................................... 24
2.3 Quality criteria of cocoa butter ............................................................................................................ 29
2.4 Chemical composition background and significance ........................................................................... 30
2.4.1 Cocoa fat content .............................................................................................................................. 30
2.4.2 Fatty acids composition .................................................................................................................... 31
2.4.3 Triacylglycerol composition ............................................................................................................. 34
2.4.4 Minor components ............................................................................................................................ 40
2.5 Correlation of chemical composition with the physical properties of cocoa butter ............................. 42
2.5.1 Cocoa butter polymorphism ............................................................................................................. 43
2.5.2 Crystallization and melting behavior ................................................................................................ 45
2.5.3 Influence of the origin – softness assessment ................................................................................... 47
2.6 Cocoa butter alternatives ..................................................................................................................... 48
2.7 Conclusions ......................................................................................................................................... 48
2.8 Aim of the thesis .................................................................................................................................. 49
Chapter 3: ........................................................................................................................................................ 61
3 Exploring cocoa lipid classes ....................................................................................................................... 61
Abstract ........................................................................................................................................................... 61
3.1 Introduction ......................................................................................................................................... 62
3.2 Material and Methods .......................................................................................................................... 64
3.2.1 Chemicals and reagents .................................................................................................................... 64
3.2.2 Preparation of the seed sample ......................................................................................................... 64
3.2.3 Thin layer chromatography (TLC) ................................................................................................... 65
3.2.4 Solid phase extraction (SPE) ............................................................................................................ 65
3.2.5 Silica gel column chromatography ................................................................................................... 66
3.2.6 HPLC chromatographic conditions .................................................................................................. 66
3.2.7 Mass spectrometry detection ............................................................................................................ 66
3.3 Results and discussion ......................................................................................................................... 67
3.3.1 Extraction optimization .................................................................................................................... 67
3.3.2 TLC lipid profile .............................................................................................................................. 70
3.3.3 Separation of lipid classes ................................................................................................................ 73
3.4 Conclusions ......................................................................................................................................... 76
Chapter 4: ....................................................................................................................................................... 79
4 Characterization of triacylglycerols in unfermented cocoa beans by HPLC-ESI mass spectrometry.......... 79
Abstract ........................................................................................................................................................... 80
4.1 Introduction ......................................................................................................................................... 81
4.2 Material and methods .......................................................................................................................... 83
4.2.1 Chemicals and reagents .................................................................................................................... 83
4.2.2 Sample preparation .......................................................................................................................... 83
4.2.3 HPLC chromatographic conditions .................................................................................................. 83
4.2.4 High-resolution Mass spectrometry conditions ................................................................................ 84
4.2.5 Tandem Mass spectrometry conditions ............................................................................................ 84
4.2.6 Enrichment of polar TAG fraction ................................................................................................... 84
4.2.7 Derivatization of hydroxylated fatty acids ....................................................................................... 85
4.2.8 NMR spectroscopy ........................................................................................................................... 85
4.3 Results and discussion ......................................................................................................................... 85
4.3.1 High resolution LC-ESI-MS and Molecular Formulae assignment ................................................. 87
4.3.2 Tandem mass spectrometry .............................................................................................................. 90
4.3.3 Characterization of TAGs with hydroxylated fatty acids substituents ............................................. 96
4.4 Conclusions ......................................................................................................................................... 99
Chapter 5: ..................................................................................................................................................... 104
5 Variation of triacylglycerol profiles in unfermented and dried fermented beans ....................................... 104
Abstract ......................................................................................................................................................... 105
5.1 Introduction ....................................................................................................................................... 106
5.2 Material and methods ........................................................................................................................ 108
5.2.1 Chemicals and reagents .................................................................................................................. 108
5.2.2 Sample preparation ........................................................................................................................ 108
5.2.3 Moisture analysis ........................................................................................................................... 108
5.2.4 TLC separation ............................................................................................................................... 109
5.2.5 HPLC chromatographic conditions ................................................................................................ 109
5.2.6 High-resolution Mass spectrometry conditions .............................................................................. 109
5.2.7 Statistical analysis .......................................................................................................................... 110
5.3 Results and discussion ....................................................................................................................... 111
5.3.1 Total lipid content ........................................................................................................................... 111
5.3.2 Triacylglycerol profiles .................................................................................................................. 112
5.3.3 Comparison of Triacylglycerol profiles of different origins ........................................................... 115
5.3.4 Correlation of triacylglycerol profiles with cocoa butter hardness ................................................. 117
5.3.5 Comparison of triacylglycerol format in unfermented and fermented dried beans ........................ 120
5.4 Conclusions ....................................................................................................................................... 122
Chapter 6: ...................................................................................................................................................... 126
6 Fatty acids profiles in unfermented and fermented dried cocoa beans of different origins and correlation with
the triacylglycerol profile ................................................................................................................................... 126
Abstract ......................................................................................................................................................... 126
6.1 Introduction ....................................................................................................................................... 127
6.2.2 Sample preparation ......................................................................................................................... 128
6.2.3 FAME synthesis ............................................................................................................................. 128
6.2.4 GC-FID method .............................................................................................................................. 129
6.3.1 Fatty acids profile ........................................................................................................................... 130
6.3.2 Fatty acids distribution ................................................................................................................... 132
6.4 Conclusions ....................................................................................................................................... 139
Chapter 7: ...................................................................................................................................................... 142
7 Monitoring lipids of Theobroma cacao during the fermentation ............................................................... 142
Abstract ......................................................................................................................................................... 142
7.1 Introduction ....................................................................................................................................... 143
7.2.2 Sample preparation ......................................................................................................................... 145
7.2.3 Moisture analysis ............................................................................................................................ 145
7.2.4 Solid phase extraction (SPE) .......................................................................................................... 145
7.2.7 FAME synthesis ............................................................................................................................. 147
7.2.8 GC-FID method .............................................................................................................................. 147
7.2.9 Spectrophotometric analysis ........................................................................................................... 148
7.2.10 Lipid extraction from the pulp ....................................................................................................... 148
7.3 Results ............................................................................................................................................... 149
7.3.1 Lipid amount .................................................................................................................................. 149
7.3.2 TAGs profile during the fermentation ............................................................................................ 149
7.3.3 Statistical analysis on TAGs ........................................................................................................... 154
7.3.4 Fatty acid profile during the fermentation ...................................................................................... 155
7.3.5 Colour of the lipid extract during the fermentation ........................................................................ 156
7.3.6 Minor lipid classes during the fermentation ................................................................................... 157
7.4 Discussion ......................................................................................................................................... 160
7.5 Conclusions ....................................................................................................................................... 164
Chapter 8: ..................................................................................................................................................... 168
8 Evolution of cocoa lipids across factory production process ..................................................................... 168
8.1 Introduction ....................................................................................................................................... 168
8.2.5 DSC analysis .................................................................................................................................. 171
8.3.1 Lipid content in dried beans and cocoa liquor ............................................................................... 172
8.3.2 TAGs profile of dried beans versus roasted beans in form of cocoa liquor .................................... 172
8.3.3 DAGs profile of dried beans versus roasted beans in form of cocoa liquor ................................... 175
8.3.4 Correlations studies on TAG and DAG profiles ............................................................................. 176
8.3.5 Elucidation of the fatty acids amides as markers of alkalization ................................................... 177
8.3.6 Influence of the fatty acid amides on the crystallization behavior of cocoa butter ........................ 178
8.4 Conclusions ....................................................................................................................................... 180
Chapter 9: ..................................................................................................................................................... 183
9 Minor lipid compounds in cocoa ............................................................................................................... 183
9.1 Introduction ....................................................................................................................................... 183
9.2.3 Solid phase extraction (SPE) .......................................................................................................... 185
9.2.6 Tandem Mass spectrometry conditions .......................................................................................... 186
9.3.1 Molecular species fraction 1 .......................................................................................................... 187
9.3.2 Molecular species fraction 3 ........................................................................................................... 188
9.4 Conclusions ....................................................................................................................................... 196
Chapter 10: .................................................................................................................................................... 197
10 Comparative lipidomics of Cylindrotheca closterium and Scenedesmus sp. reveals differences in lipid
production under nitrogen starvation .................................................................................................................. 197
Abstract ......................................................................................................................................................... 198
10.1 Introduction ....................................................................................................................................... 199
10.2.1 Algae and culture condition ............................................................................................................ 201
10.2.3 Total lipids extraction ..................................................................................................................... 202
10.2.4 FAME preparation and GC analysis ............................................................................................... 202
10.2.5 High-performance liquid chromatography ..................................................................................... 203
10.2.6 High-resolution mass spectrometry ................................................................................................ 203
10.2.7 Tandem mass spectrometry ............................................................................................................ 203
10.3.1 Total lipid content and growth kinetics under nitrogen starvation.................................................. 204
10.3.2 Fatty acids profile under nitrogen starvation. ................................................................................. 206
10.3.3 Relative quantitative analysis of glycerolipids and pigments ......................................................... 208
10.3.4 Lipid and pigment molecules alteration ......................................................................................... 211
10.4 Conclusions ....................................................................................................................................... 218
General Conclusions ..................................................................................................................................... 223
Preface
“It’s not enough to do your best, you must know what to do and then do your best.”
W. Edwards Deming
The years as a Ph.D. student at Jacobs University have been full of memorable experiences,
challenges, and opportunities. Coming to the end of this learning path, I would like to express my
gratitude to everyone who contributed to it accomplishment.
I would like to acknowledge and thank my supervisor Prof. Dr. Nikolai Kuhnert for his guidance,
encouragements, and for allowing me the opportunity to do research in my way while always pushing
me indirectly to overcome my limits. I have learned a lot from him, from chemistry to human social
interactions. Appreciations also go to Prof. Dr. Matthias Ullrich for his continuous engagement and
advisement, as well for his ‘keep going, you are doing well!’ smile whenever we met in the university
corridors. I would like to thank Dr. Frédéric Depypere for joining my dissertation committee and
reviewing this thesis. I am grateful for the support and help from my former and current colleagues,
group and lab members. I am especially grateful to Dr. Maria Alexandra Patras for always being there
for me for whatever issues. I am thankful to Anja Mueller for her technical support and to Britta Behrens
for helping me to sort out the enormous number of cocoa lipid samples. Having great colleagues made
my tough days better. I would like to thank CoMeta team for their support, commitment, and valuable
advises.
Moreover, I would like to thank Anja Muller, Britta Behrens, Nina Boettcher and Dr. Hussein Ali El
Damrany Hussein for their technical support, for helping out with sample preparation and in sample
logistics. Many thanks to Dr. Maria Alexandra Patras, Dr. Abhinandan Shrestha, Dr. Sergio Grimbs, Dr.
Anne Grimbs, Dr. Roy D’Souza and Dr. Foad Tehrani Najafian for their valuable guidance on
performing the data analysis, excel platforms and advises. Dr. Gino Vrancken for support and many
productive discussions. Enis Bardhi for helping with the nuts lipid extractions and TLC. Audrey’s
Bergounhou for helping to perform analytical measurments on fatty acids amide.
My immeasurable gratitude is to my mother, Ioana Sirbu, whose support; love and patience have no
equal. This achievement I have to share with her. Further, I would like to thank the rest part of my
family for always being proud of me and having the patience for a forever student in the family. Besides,
I want to show my love and appreciation for my partner and best friend who found the best way to
support me during the most stressful period of my Ph. D. Another thank you must go to my
physiotherapist Ariane von der Kaus for helping me to relax my back after long hours in front of
computer.
Additionally, I would like to thank all my friends outside Kuhnert’s group for their support,
friendship, and encouragement during the ups and downs of the last few years. I would like to express
I
my gratitude to all of my co-authors for their openness, collaboration, and patience. I wish to
acknowledge all my students whom I have got to supervise during my Ph. D. studies. Many thanks to
the people of other research groups on Campus. It felt encouraging when I could meet so many people
ready to help. I am thankful to Jacobs’ community for teaching me to be more patient and to understand
and accept diversity.
Finally, I would like to thank Barry Callebaut Company for financial support and people for the
constant exchanges.
Without the support of all these people, it would not have been possible to complete this thesis.
Thank you
II
Abstract
Abstract
Theobroma cacao is a tree that produces pods containing high valuable seeds, referred to as
cocoa beans. About 50 million people around the world depend on cocoa as a source of income.
From tree to bar, cocoa beans undergo an intricate processing chain leading to various chemical
changes within the seed. Lipids constitute half of the dry weight of a cocoa bean and the most
significant part of the entire cocoa market value taking the name of cocoa butter. Cocoa butter
is composed of about 95% triacylglycerols (TAGs) and 5% of other minor components. The
physical and chemical characteristics of this fat have been at the forefront of many research
studies in the past decade. However, there are several knowledge gaps regarding the chemical
composition and modifications that cocoa butter undergoes during the production process. This
research aimed at investigating the chemistry of lipids in cocoa samples from different
production steps and geographical regions, thus enriching the present knowledge of cocoa
chemistry.
In the first place, analytical methods for the extraction, separation, and detection of cocoa
lipids were developed and optimized. The overall lipid profile of unfermented, dried and
roasted cocoa beans was investigated, and a solid phase extraction method, separating cocoa
lipid extract in five fractions was established to aid the evaluation of the all cocoa lipid
constituents.
A non-aqueous reversed-phase liquid chromatographic method using ethanol as the mobile
phase, coupled to electrospray ionization (ESI) tandem mass spectrometry was developed for
TAGs analysis, which allowed the characterization of 83 different TAGs in unfermented cocoa
beans, the largest number of TAGs identified in cocoa so far. From these, 58 TAGs were not
previously reported in cocoa and 31 TAGs were characterized by the presence of one to three
hydroxyl groups on the unsaturated fatty acid chain. Applying the same analytical approach,
relative quantification of the TAGs in lipid extracts of unfermented raw and fermented dried
beans from different origins was performed. Results revealed that the presence of hydroxy-
allyl fatty acid derived TAGs (proposed name cacaoic acid) is unique to unfermented beans
and they may serve as biomarkers for the fermentation grade of the beans. A simple analytical
test, based on thin layer chromatography, testing for the presence of these biomarkers has been
introduced.
Multivariate statistical analysis was carried out to compare triacylglycerol profiles across
the different origins. Only small variations in individual compound concentrations across the
different origins were observed. Nonetheless, TAGs profile of each geographical origin
III
Abstract
reflected the influence of the climate conditions. Therefore, a new softness/hardness indicator
was proposed, which takes into account only four di-unsaturated TAGs relative quantities. In
addition, the fatty acids profile showed the impact of both fermentation status and origin of the
beans. 18 fatty acids were successfully identified and many others detected. The primary
outcome was the significantly increased quantity of palmitic and linoleic acid in dried
fermented beans when compared to unfermented wet beans in all geographical locations.
Moreover, the analysis of variance showed significant differences in the amount of the short
chain fatty acids for South American cocoa beans, whereas Malaysian samples contained a
significantly higher amount of long chain fatty acids.
Concerning the chemical changes occurring during the fermentation process, 72 and 96
hours (h) of fermentation showed to be crucial time points for both triacylglycerol and fatty
acid content. Hydroxy-allyl fatty acids containing TAGs decreased almost entirely after 96 h
of fermentation. The lipid extracts obtained from different time points have shown a distinct
change of the color from pale yellow to brown yellow. Absorbance at 400 nm revealed a
maximum between 96 and 120 h. Moreover, the lipid profile of minor components pointed out
towards the influence of the microorganisms involved in the fermentation process. These
results suggest that lipids could also be involved in defense and stress response and in the
cellular metabolism of cocoa.
Furthermore, the evolution of cocoa lipids during the factory production process was
studied. While the total lipid amount decreased from bean to liquor, considerably random
amount dissimilarities were observed in the entire TAGs profile, which suggests that other
chemical reactions of an unknown nature might take place during the roasting and grinding
processes. Moreover, the impact of the alkalization treatment was explored. HPLC-MS data
revealed the presence of novel compounds, identified as fatty acids amides, namely
palmitamide, oleamide, and stearamide, which were proposed as markers of the alkalization
process.
Separation of cocoa lipid extract in 5 fractions allowed structure elucidation of 62 minor
lipid components which are part of the six different lipid classes including sterol esters,
diacylglycerols, waxes, free fatty acids, sterol esters glucosides and phospholipids.
The lipidomics approach using the HPLC-MS method developed for cocoa was applied for
the analysis of the lipid extracts of 2 algae species, Scenedesmus sp. (Chlorophyceae) and
Cylindrotheca closterium (Bacillariophyceae). Nine different lipid classes were identified and
monitored in response to nitrogen stress during 96 h. Although the two algae exhibited different
IV
Abstract
lipid metabolic responses with respect to nitrogen starvation, both of them showed a significant
increase of triacylglycerols, which is valuable for biofuel production.
A large part of the research was carried out aiming at product valorization and exploration
of new possible industrial applications performing multiple analysis on different samples
provided by the company. The outcomes give insights into the potential of cocoa butter and
open new horizons for future studies and industrial applications.
V
List of abbreviations
A arachidic acid
ANOVA one-way analysis of variance
APCI atmospheric pressure chemical ionization
API atmospheric pressure ionization
CB cocoa butter
CHCl3 chloroform
CN carbon number
DAG diacylglycerol
DBs double bonds
DCM dichloromethane
DSC differential scanning calorimetry
ECN equivalent carbon number
EFSA European Food Safety Authority
EIC extracted ion chromatogram
ELSD evaporative light scattering detector
ESI electrospray ionization
F fermented
FA fatty acid
FAA fatty acid amide
FAME fatty acid methyl esters
FID flame-ionization detector
FTICR Fourier-transform ion cyclotron resonance
GC gas chromatography
GLC gas liquid chromatography
HCA hierarchical cluster analysis
HPLC high performance liquid chromatography
L linoleic acid
LC liquid chromatography
Ln linolenic acid
LOQ limit of quantification
LOX Lipoxygenase
M miristic acid
Ma margaric acid
MANOVA multivariate analysis of variance
MeOH methanol
MOSH mineral oil saturated hydrocarbons
VI
MOAH mineral oil aromatic hydrocarbons

MS mass spectrometry
NARP non-aqueous reversed-phase
NMR nuclear magnetic resonance spectroscopy
O oleic acid
OC19:0 1-oleoyl-2-nonadecanoyl-sn-glycerol
OO 1,2-dioleoyl-sn-glycerol
P palmitic acid
PCA principal component analyses
PLP 1,3-dipalmitoyl-2-linoleoyl-sn-glycerol
PLS 1-palmitoyl-2-linoleoyl-3-stearoyl-sn-glycerol
POA 1-palmitoyl-2-oleoyl-3-arachidoyl-sn-glycerol
POO 1-palmitoyl-2,3-dioleoyl-sn-glycerol
POP 1,3-dipalmitoyl-2-oleoyl-sn-glycerol
POS 1-palmitoyl-2-oleoyl-3-stearoyl-sn-glycerol
Rf retention factor
Rt retention time
S stearic acid
SFC solid fat content
SLR 1-stearoyl-2-linoleoyl-3-ricinoleoyl-sn-glycerol
SLS 1,3-distearoyl-2-linoleoyl-sn-glycerol
sn stereospecific number
SOA 1-stearoyl-2-oleoyl-3-arachidoyl-sn-glycerol
SOO 1-stearoyl-2,3-dioleoyl-sn-glycerol
SOS 1,3-distearoyl-2-oleoyl-sn-glycerol
sp. species
SP 1-stearoyl-2-palmytoyl-glycerol/2-palmytoyl-3-stearoyl-sn-glycerol
SPE solid phase extraction
SPS 1-stearoyl-2-palmytoyl-3-stearoyl-sn-glycerol
SS 1,3-distearoyl-sn-glycerol
S'U'S' 1,3-disaturated-2-unsaturated-sn-glycerol
S'U'U'/U'U'S' 1(3),2-diunsaturated-3(1)-saturated-sn-glycerol
TAG triacylglycerol
TLC thin layer chromatography
TOF time of flight
2D map two dimensional map
2D NMR two dimensional nuclear magnetic resonance spectroscopy
U unfermented
VII
List of figures
List of figures
Figure 1.1 Lipid classes and prominent examples.

Figure 2.1 Representation of the main cocoa process practices and the steps in cocoa butter
production.
Figure 2.2 Chemical structure of the most predominant fatty acids in cocoa.
Figure 2.3 Box plots* illustrating the distribution of major fatty acids** in different origin countries.
*Normalization to 100 was performed to reduce the variability due to analytical discrepancies by
different authors.
** Fatty acids data from thetable wereused to build the box plots.
Figure 2.4 Triacylglycerol structure of a 1,3-disaturated-2-unsaturated-sn-glycerol represented by
POS
Figure 2.5 Box plots* illustrating the distribution of major TAGs** in different origin countries.
*Normalization to 100 was performed to reduce the variability due to analytical discrepancies by
different authors.
**Triacylglycerol data from the table were used to build the box plots.
Figure 2.6 Thin-layer chromatographic separation of phospholipids from seed crystals (SCPL 3),
phospholipid standard mix (STDPL mix) and Ivory Coast cocoa butter sample (CBPL2). SL–simple
lipids, DPG–diphosphatidylglycerol, GL–glycolipid, PG–phosphatidylglycerol,
PA–phosphatidylethanolamine, PC–phosphatidylcholine, PI+LPE–phosphatidylinositol and
lysophosphatidylethanolamine, PS–phosphatidylserine, PA–phosphatidic acid (adapted from Arruda
& Dimick, 1991).
Figure 2.7 (a) Chain-length packing structures and (b) the subcell structures of the 3 most common
polymorphs in TAGs (Himawan et al., 2006; Sato, 1999, 2001).
Figure 2.8 (a) Phase transition scheme of cocoa butter as proposed by Van Malssen et al., (1999) (b)
polymorphic behavior of POP and SOS as shown by Yano et al., (1993).
Figure 2.9 Time-temperature state diagram for the polymorphism of statically crystallized cocoa
butter. Symbol (*) represents the polymorphic forms that have been determined by XRD (Marangoni
& McGauley, 2003).
Figure 2.10 Ideal solid fat content curve of chocolate as function of time (Klagge & K. Sen Gupta,
1990).
Figure 3.1 (a) TLC plate image showing the lipid profile of fermented dried beans extracted with
Soxhlet apparatus: 1 – petroleum ether; 2 – CHCl3:MeOH (1:1); 3 – heptane; 4 –DCM; 5 – Folch
extraction CHCl3:MeOH:0.05 N potassium chloride (8:4:3); and 6 – diethyl ether. Iodine staining and
UV light at 254 nm visualized the spots. (b) and (c) show non-lipid compounds extracted with
CHCl3:MeOH (2:1, v/v), and (d) TLC plate treated with ferric chloride for the phenols presence (F –
Folch; DCM, DCM:MeOH, CHCl3:MeOH).
VIII
List of figures
Figure 3.2 Optimization of lipid extraction in cocoa by Soxhlet method using DCM as extraction
solvent. 3 g of fermented dried powdered cocoa beans and 150 mL solvent.
Figure 3.3 Percentage of lipid extracted form 6 seeds and 1 dried leaves samples by Soxhlet method
using three different extraction solvents. 3 g of powdered plant material and 150 mL solvent.
Figure 3.4 (a) Visualization under UV light at 254 nm and (b) Hanessian's stained TLC plate image
showing the lipid profile of different treated beans extracted with Soxhlet apparatus using P –
petrolium ether (2, 5, 8), H – heptane (3, 6, 9) and DCM – dichloromethane (4, 7, 10) by subsequent
extraction compared with Folch extraction of the dried cocoa beans.
Figure 3.5 TLC plate* image showing the lipid profile of white tea leaves, cocoa beans, coffee beans
and sunflower seeds extracted with Soxhlet apparatus using P – petrolium; H – heptane; and DCM –
dichloromethane (3 g sample and 150 mL solvent). *Hanessian's stain was used.
Figure 3.6 TLC plate* image showing the lipid profile of white hazelnuts and almonds in comparison
to cocoa beans extracted with Soxhlet apparatus using P – petrolium; H – heptane; and DCM –
dichloromethane (3 g sample and 150 mL solvent). *Hanessian's stain was used.
Figure 3.7 TLC plate* image showing the separation of 5 lipid fractions corresponding to: cholesterol
esters; tryacylglicerols, diacylglycerols, free fatty acids and polar lipids classes asserted by MS
fragmentation. On the right side, the proposed structures of identified compounds in cocoa lipid
extract. *Hanessian's stain was used.
Figure 3.8 (a) TLC plate image showing the separation of a silica column fractions of lipidic
compounds stained with Hannessian stain and (b) under UV light. (c) 2D maps build by HPLC-MS
data of the main fractions illustrating the complexity of cocoa lipid. Fraction 3 contains the well
known TAGs in cocoa.
Figure 4.1 (a) HPLC-MS profile of Ivory Coast raw unfermented cocoa TAGs in positive ionization
mode; (b) Two-dimensional (2D) map of TAGs from raw cocoa beans, Ivory Coast origin, by
reversed-phase high resolution HPLC/ESI-MicroOTOF MS analysis. Colour rapresents the intensity
of the peaks with red being the highest and blue the lowest: (c) 2D HPLC-MS feature map displaying
TAGs building blocks of homologous series.
Figure 4.2 Positive ion mode tandem mass spectrum showing the fragmentation of triacylglycerols
(a) SPS (m/z 881.9), (b) POP (m/z 851.9), (c) PLP (m/z 848.8) and (d) PLnP (m/z 829.7) ammonium
adduct ions. P, palmitic; O, oleic; S, stearic; L, linoleic; Ln, linolenic and neutral loss chromatograms
of (e) 273 Da, (f) 301 Da, (g) 299 Da, (h) 297 Da and (i) 295 Da (from up to down) corresponding to
ammonia adducts of palmitic acid, stearic acid, oleic acid, linoleic acid and linolenic acid in Ivory
Coast cocoa lipid extract.
(a) PC18:2-OHP (m/z 865.0), (c) PC18:1-OHP (m/z 867.0) ammonium adduct ions and (b) PC18:2-
OHP (m/z 870.0) and (d) PC18:1-OHP (m/z 872.0) sodium adduct ions. (e) and (f) proposed possible
structures of the unusual TAGs in raw cocoa beans.
IX
List of figures
Figure 5.1 Box plots of total lipid content determined gravimetrically following Soxhlet extraction
for (a) unfermented raw and (b) fermented dried cocoa beans from different countries of origin.
Figure 5.2 Representative chromatography results for fermented dried and unfermented raw cocoa
lipid extract of Ivory Coast origin. (a) TLC plate image and the parallel corresponding (b) HPLC-MS
chromatograms in positive ion mode (P – palmitic acid, O - oleic acid, S – stearic acid, L – linoleic
acid, Ln – linolenic acid, Ma – margaric acid, A – arachidic acid). (c) Two-dimensional (2D) map of
reversed-phase high-resolution MS of unfermented raw and (d) fermented dried cocoa bean samples
from Ivory Coast. The colormap of the 2D map represents the peak intensity with red being the
highest and blue the lowest.
Figure 5.3 Relative abundance of the key 12 TAGs for unfermented and fermented cocoa beans from
different countries of origin. The sample labels denote the countries of origin, namely, B – Brazil, E –
Ecuador, I – Indonesia, M – Malaysia, T – Tanzania, and V – Ivory Coast, as well as the fermentation
status, F – fermented and U – unfermented. The gray painted bars illustrate the remaining detected
TAGs.
Figure 5.4 PCA of the 12 most relevant TAGs for unfermented and fermented cocoa beans from
different countries of origin. The sample labeling and coloring in (a) the scores plot as well as the
coloring of the two-standard deviational ellipses illustrate the different countries of origin, namely, B
– Brazil, E – Ecuador, I – Indonesia, M – Malaysia, T – Tanzania, and V – Ivory Coast. In addition,
the sample label as well as the shape indicate the fermentation status, F – fermented and U –
unfermented. The labelling in (b) the loadings plot denotes the major TAGs.
Figure 5.5 Box plots of the 12 TAGs for unfermented and fermented cocoa beans from different
countries of origin. The subplot label denotes the significance value (p) of the respective ANOVA.
The heat map on the lower right illustrates the pairwise T-tests with a name for each TAG, which is
significantly different for the respective country pair. The brightness of the heat map additionally
indicates the number of significantly different TAGs per country couple.
Figure 5.6 (a) Extract ion chromatogram of m/z 876.8015 corresponding to POO (left peak) and PLS
(right peak); (b) Scatter plot on the SLS:SOO ratio and the PLS:POO (of all the samples) and (c)
Extract ion chromatogram of m/z 904.8328 corresponding to SOO (left peak) and SLS (right peak) of
2 Malaysian samples, 2 Ivorian and 2 Ecuadorian samples.
Figure 5.7 (a) Box plots of softness indicators for unfermented and fermented cocoa beans from
different countries of origin. As softness indicator serve the established (a) ratio of polyunsaturated
and monounsaturated plus saturated TAGs and (b) the newly conceived cumulative values of POO,
PLS, SOO and SLS.
Figure 5.8 PCA of the TAGs profiles from LC-MS data, excluding the major 12 TAGs for
unfermented and fermented cocoa beans from different countries of origin. The sample labeling and
coloring illustrate the different countries of origin, namely, B – Brazil, E – Ecuador, I – Indonesia, M
– Malaysia, T – Tanzania, and V – Ivory Coast. In addition, the sample label, as well as the shape and
X
List of figures
the coloring of the two-standard deviational ellipses indicate the fermentation status, F – fermented
and U – unfermented.
Figure 5.9 Heat map of hierarchical clustering of the peak profiles from LC-MS data, excluding the
major 12 TAGs for unfermented (U) and fermented (F) cocoa beans from different countries of
origin. The detected peaks are given in ascending retention time order.
Figure 5.10 Proposed structure of a triacylglycerol containing a hydroxyl-allyl fatty acid substituent,
namely cacaoic acid.
Figure 6.1 Representative GC-FID chromatogram of fatty acids methyl esters of cocoa lipid extract,
Ecuador unfermented beans.
Figure 6.2 Representative GC-FID chromatogram of FAME extract of (a) and (c) unfermented wet
beans and (b) and (d) fermented dried beans from Ecuador and Indonesia respectively.
Figure 6.3 Most abundant fatty acid distribution according to the origin.
Figure 6.4 (a) Bar plot of the most abundant 3 fatty acids and (b) of the other 3 fatty acids in
unfermented wet and fermented dried cocoa beans of different origins.
Figure 6.5 Bar plot illustrating the total fatty acids* content in unfermented wet and fermented dried
cocoa beans of different origins and their corresponding ratio unsaturated/saturated fatty acids*.
*Only the data of the positively identified peaks were taken into account.
Figure 6.6 Bar plot illustrating the hydroxyl fatty acids content in unfermented wet and fermented
dried cocoa beans of different origins. *the amount was calculated as mg of stearic acid/g of cocoa lipid extract.
Figure 6.7 (a) Scatter plots illustrating the variation of oleic acid and main TAGs in unfermented wet
cocoa beans and (b) in fermented dried cocoa beans and of different origins E – Ecuador, M –
Malaysia, I– Indonesia, B – Brazil, V – Ivory Coast.
Figure 7.1 Lipid amount as the percentage in dry mass of six fermentation trials from 6 different
origins.
Figure 7.2 Monounsaturated TAGs relative amount in 3 fermentation trials from 3 different origins.
Figure 7.3 Di- und triunsaturated TAGs relative amount in 3 fermentation trials from 3 different
origins.
Figure 7.4 Saturated TAGs relative amount in 3 fermentation trials from 3 different origins.
Figure 7.5 Evolution of TAGs profiles in CCN51 hybrid during the fermentation.
Figure 7.6 Unusual TAGs during the fermentation process.
Figure 7.7 Unusual TAGs during the fermentation in (a) 2 batched of the same hybrid and (b) in 2
batched of two hybrids of the same origin.
Figure 7.8 Unusual TAGs during the fermentation in samples from Malaysia.
Figure 7.9 Means of the distribution of 4 homologues series TAGs in Malaysian samples.
Figure 7.11 Total fatty acids and main fatty acids profile of a fermentation series from Ivory Coast .
Figure 7.12 Total fatty acids and main fatty acids profile of a spontaneous and controlled
fermentation series from Ivory Coast.
XI
List of figures
Figure 7.13 Evolution of the color during the fermentation of cocoa beans.
Figure 7.14 2D map of the LC-MS chromatogram in positive ion mode of minor components of the
lipid extract of a fermentation trial from Ivory Coast. Blue circle indicates the unusual hydroxyl TAGs
and red circle the phospholipid class occurrence.
Figure 7.15 2D map of the LC-MS chromatogram in positive ion mode of minor components of the
lipid extract of a fermentation trial from Ecuador. Blue circle indicates the unusual hydroxyl TAGs
and red circle the phospholipid class occurrence.
Figure 7.16 2D map of the LC-MS chromatogram in positive ion mode of the lipid extract from pulp
at 24 and 48h from 3 different hybrids from 2 origins: Ivory Coast and Malaysia. At the upper part,
same 2D map* left and right side of a reference chromatogram from Ivory Coast. *these 2D maps are inserted
for comparison purposes.
Figure 8.1 Alkalized ingredients used in alkalization of cocoa adapted from (“Published Articles |
Blommer Chocolate Company,” ).
Figure 8.2 Lipid amount in cocoa beans and cocoa liquors from different origin.
Figure 8.3 Profile amount of monounsaturated TAGs in cocoa beans and cocoa liquors from different
origin.
Figure 8.4 Profile amount of di- and triunsaturated TAGs in cocoa beans and cocoa liquors from
different origin.
Figure 8.5 Ratios between di- and triunsaturated TAGs and ratios between saturated and
monounsaturated TAGs in cocoa beans and cocoa liquors from different origin.
Figure 8.6 Profile amount of hydroxyl-TAGs in cocoa beans and cocoa liquors from different origin.
Figure 8.7 Profile amount of saturated TAGs in cocoa beans and cocoa liquors from different origin.
Figure 8.8 Profile amount of DAGs in cocoa beans and cocoa liquors from different origin.
Figure 8.9 Positive ion mode HPLC-MS chromatogram of normal and alkalized lipid profile.
Figure 8.10 Positive ion mode HPLC-MS chromatogram (zoomed) of alkalized butter and mass
spectra of the new identified compounds.
Figure 8.11 Tandem MS spectra of oleamide.
Figure 8.12 DSC crystallization thermograms recorded on cooling of natural CB, natural cocoa butter
(CB), natural CB plus 0.1% fatty acid amides (FAA) mixture, natural CB plus 1% FAA, and natural
CB plus 2% FAA mixture.
Figure 8.13 DSC crystallization melting curves of natural CB, of natural cocoa butter (CB), natural
CB plus 0.1% fatty acid amides (FAA) mixture, natural CB plus 1% FAA, and natural CB plus 2%
FAA mixture.
Figure 8.14 Solid fat content (in %) as a function of temperature (°C) of natural cocoa butter (CB),
natural CB plus 0.1% fatty acid amides (FAA) mixture, and natural CB plus 1% FAA mixture.
Figure 9.1 2D map of the LC-MS chromatogram in positive ion mode of fraction 1 minor
components of the SPE lipid extract from fermented dried beans. Red circle indicates the sterol esters
class.
XII
List of figures
Figure 9.2: Positive ion mode tandem mass spectrum of sterol esters (a) 18:2 Campesteryl ester (m/z
680.7) ammonia adducts, (b) 18:2 Sitosteryl ester (m/z 694.6) ammonia adducts, (c) cholesterol
standard (m/z 369.2) corresponding to [M-OH]+ molecular ion.
components of the SPE lipid extract from (a) unfermented wet cocoa beans and (b) fermented dried
cocoa beans. Blue circle indicates the hydroxyl-TAGs, green circle indicates the diacylglycerol class,
and magenta circle indicates waxes class.
Figure 9.4 (a) HPLC-MS extract ion chromatogram of m/z 577.5 in positive ion mode corresponding
to palmitoyl-oleoyl-sn-glycerol and (b) tandem mass spectrum displaying the fragmentation of 2
regioisomers and tandem MS spectrum of the standard 1-palmitoyl-2-oleoyl-sn-glycerol.
Figure 9.5 MS spectrum illustrating the adduct ions of 1-palmitoyl-2-oleoyl-sn-glycerol, 1-palmitoyl-
3-oleoyl-sn-glycerol from cocoa lipid extract, and of 1-palmitoyl-2-oleoyl-sn-glycerol standard
compound.
Figure 9.6 HPLC-MS extract ion chromatogram in positive mode of diacylglycerols identified in
cocoa lipid extract.
Figure 9.7 Tandem mass spectrum heptadecyl-palmitate, m/z 512.5, as representative of wax class.
Figure 9.8 HPLC-MS extract ion chromatogram in negative mode of free fatty identified in cocoa
lipid extract.
cocoa beans. Blue circle indicates other unusual TAGs, light brown circle indicates other unknown
lipid classes.
cocoa beans. Red circle indicated sterol esters glucosides class and light blue circle indicates
phospholipid class in cocoa.
Figure 9.11 Positive ion mode tandem mass spectrum of sterol esters glucosides (a) 16:0-Glc-
Sitosterol (m/z 832.6) ammonia adducts and (b) 18:0-Glc-Stigmasterol (m/z 858.7) ammonia adducts.
Figure 9.12 Positive ion mode tandem mass spectrum of phospholipids (a) 1-palmitoyl-2-oleoyl-sn-
glycero-3-phosphocholine, (b) 1,2-dioleoyl-sn-phosphatidylethanolamine, (c) and (d) their
corresponding chemical structures.
Figure 10.1 (a) Total lipids and (b) growth kinetics of Scenedesmus sp. and C. closterium under
nitrogen starvation. Results were shown as meansSD, n=3.
Figure 10.2 Fatty acids composition of Scenedesmus sp. and C. closterium under nitrogen starvation.
(a) Major fatty acids in Scenedesmus sp.; (b) Minor fatty acids in Scenedesmus sp.; (c) Major fatty
acids in C. closterium; (d) Minor fatty acids in C. closterium. Results were shown as meansSD, n=3.
Compounds could be assigned by standards were named by the retention time.
XIII
List of figures
Figure 10.3 Changes of glycerolipids and pigments in Scenedesmus sp. and C. closterium under
nitrogen starvation: (a) and (b) Glycerolipids and pigments in Scenedesmus sp.; (c) and (d)
Glycerolipids and pigments in C. closterium. Results were shown as meansSD, n=3.
Figure 10.4 Principal component analysis of Scenedesmus sp. and C. closterium under nitrogen
starvation: (a) Samples from Scenedesmus sp.; (b) Samples from C. closterium. Three biological
replicates were performed at each time point.
Figure 10.5 Heat map of glycerolipid and pigment molecules in Scenedesmus sp. and C. closterium
under nitrogen starvation: (a) glycerolipid and pigment molecules from C. closterium; (b)
glycerolipid and pigment molecules from Scenedesmus sp. All the identified molecules were primarily
clustered in 4 groups: molecules in group a' and c' displayed a decreasing trend during the time
course; molecules in group b' decreased at the beginning, then increased till 96 h; group d' exhibited
an increasing trend. compounds were tentatively identified due to the insufficient fragment information.
Figure 10.6 Chemical structures of the most abundant lipid compounds in Scenedesmus sp.
Figure 10.7 Chemical structures of the most abundant lipid compounds in C. closterium.
XIV
List of tables
List of tables
Table 1.1 The number of possible TAGs.

Table 2.1 Definitions and characteristics of cocoa butter products according to EU Directive
2000/36/EC and Codex Standards.
Table 2.2 Fatty acid content (%) in cocoa beans from different origins.
Table 2.3 (a) and (b) Triacylglycerol content (%) in cocoa beans from different origins.
Table 2.4 Cocoa butter polymorphs and their melting point (°C) according to selected authors.
Table 3.1 Common methods for lipid extraction in literature and their characteristics
Table 3.2 Amount of cocoa lipids extracted from 3 g of dried beans by different extraction methods
and different solvents.
Table 4.1 Origin data and total lipid amount of raw cocoa samples used in this study.
Table 4.2 Abbreviations of fatty acids with the characteristic fragment ion acyls and neutral losses.
Table 4.3 TAG species in raw cocoa beans lipid extract subdivided into homologous series.
Table 5.1 Triacylglycerol composition of cocoa lipid extract of different origins*.
Table 6.1 Significance value (p) of the respective ANOVA of the five samples compared to Q1-
Tanzania origin as reference point for the origins and Q2-U (unfermented) as reference point for
unfermented wet and fermented dried comparison.
Table 6.2 Correlation matrixes of fatty acid and TAGs in unfermented wet and dried fermented cocoa
beans.
Table 7.1 Analysis of variance p-value for POS in two Malaysian fermentation trials.
Table 8.1 Correlation values of the TAG lipid classes, DAG, and total lipid amount.
Table 9.1 Diacylglycerol molecular species in cocoa beans lipid extract.
Table 10.1 The content of TAGs (% dry biomass) with C16 or C18 FA in the sn-2 position and ratio
of TAG with C16 and C18 in the sn-2 position.
XV
Chapter 1
Chapter 1:
1 Introduction
1.1 Lipidomics
1.1.1 General concept of Lipidomics
Lipidomics is an emerging field of metabolomics centered on the analysis of all lipids

present in a biological system known as a lipidome. A lipidome is the comprehensive and
quantitative description of a set of lipid species present in an organism (Navas-Iglesias,
Carrasco-Pancorbo, & Cuadros-Rodríguez, 2009; Wenk, 2005). Lipidomics covers not only
the study of all lipids, identifying the structures of cellular lipid species including the number
of atoms, the number and location of double bonds, the core structures and head groups,
individual fatty acyl chains, and the regiochemistry of each isomer, but also the molecules with
which they interact, and their function within the cell. A more comprehensive definition
describes lipidomics as: the full characterization of lipid molecular species and their biological
functions, with respect to expression of proteins involved in lipid metabolism and function,
including gene regulation (Spener, Lagarde, Géloên, & Record, 2003). This branch of
metabolomics uses multiple techniques to quantify hundreds of chemically different lipids and
define pathways of the molecular mechanisms through which they enable cellular function
(Meer, 2005; Wenk, 2010).
1.1.2 Lipid classification and functions
Although there is not a unified definition of lipids, most of the authors accept the definition
provided by Christie, which restricts the use of ‘lipids’ to “fatty acids, their naturally-occurring
derivatives (esters or amides), and substances related biosynthetically or functionally to these
compounds” (Christie & Han, 2010). Lipids are various compounds that are insoluble in water,
soluble in organic solvents and chemically and structurally diverse. The number of distinct
lipid chemical species in a living organism is estimated to be between 104–105. Recently, The
LipidMaps consortium has developed the first internationally accepted lipid classification
1
Chapter 1
system (Brügger, 2014; LIPID MAPS Lipidomics Gateway), as well as nomenclature and
structural representation system. This database contains over 40,000 unique lipid structures of
biological relevant lipids from human, mouse, plants, bacteria, fungi, algae and marine
organisms, and other computationally generated structures. There are eight lipid categories,
each complete with its classification and sub-classification hierarchy (Fig. 1.1). Each lipid is
chemically different and thus set with different functional properties.
Fatty acyls
Fatty acids and conjugates

Octadecanoids oleic acid
Eicosanoids
Docosanoids
Fatty alcohols
3-oxo-2-(2-pentenyl)-cyclopentane-1-octanoic acid
Wax esters (cis-(+)-OPDA)
stearyl palmitate
Glycerolipids
Triacylglycerol
Diacylglycerol
Monoacylglycerol
1-palmitoyl-2-oleoyl-3-stearoyl-sn-glycerol
Glycerophospholipids
Phosphatidic acids
Phosphatidylcholines
Phosphatidylserines
Phosphatidylglycerols
Phosphatidylethanolamines
Phosphatidylinositols and phosphoinositides
Cardiolipins
1-oleoyl-2-linoleoyl-sn-glycero-3-phosphocholine
Figure 1.1 Lipid classes and prominent examples (continues on next page).
2
Chapter 1
Sterol lipids
Sterols and derivatives
sitosterol
Prenol lipids
Isoprenoids
α tocopherol – vitamin E
Sphingolipids
Sphingoid bases
Ceramides
Phosphosphingolipids
Neutral, acidic and basic glycosphingolipids
N-(octadecanoyl)-sphing-4-enine-1-phosphocholine
Saccharolipids
2,3-di-o-hexanoyl-glucopyranose
Polyketides
Erytromycin
Figure 1.1 (continued) Lipid classes and prominent examples.
Fatty acids are considered the simplest lipid compounds. The common fatty acid chain
contains an even number of carbon atoms in a straight chain, which can vary from 2 to 36
3
Chapter 1
carbon atoms in animal and plant tissues. Fatty acids from plant tissues can contain a wide
variety of functional groups including double bonds, hydroxyl, keto and other groups. Fatty
acids are the building blocks of more complex lipids such as acylglycerols, sterol-esters and
sphingolipids, and precursors of biologically active lipids such as the eicosanoids,
prostaglandins, and leukotrienes. Combinations of different fatty acids, typically linked by a
backbone, and functional head groups lead to a massive structural diversity in lipids (Christie
& Han, 2010).
A typical complex structure formed by one molecule of glycerol esterified by three fatty
acid molecules is a triacylglycerol (TAG). The three fatty acids may be identical, completely
different, or two of them the same. Because they are biosynthesized from acetyl-CoA, fatty
acids in plants and animals are in general composed of only even-numbered chain lengths.
Bacteria, on the other hand, can synthesize acids with odd chain lengths and branched chains.
Therefore, ruminant fat also contains acids with an odd number of carbon atoms, for example,
15 or 17, because of the bacteria in the rumen.
A special nomenclature is used for the stereochemistry of TAGs instead of the usual D/L or
R/S nomenclature, where the primary hydroxyl groups are often termed α- and α' and the
secondary β, or the "stereospecific numbering" (sn) system was recommended by the IUPAC-
IUB commission.
The number of theoretically possible molecular species increases with the growing number
of fatty acids forming a mixture of TAGs, see Table 1.1. However, many of these theoretically
possible species do not exist in nature or exist in very low abundance.
Table 1.1 The number of possible TAGs.

Description Number of Possible TAGs Number of Possible TAGs for y = 10
without isomers x = (y3 + 3y2 + 2y)/6 220
without enantiomers x = (y2 + y3)/2 550
all isomers x = y3 1000
(where x is the number of TAGs and y is the number of FAs in TAGs)
Lipid compounds have diverse physical and chemical properties and occur in vastly
different concentration ranges. Therefore, lipids exhibit a wide variety of functions within cells.
Their primary function is to form the matrix of our cell membranes, including cellular and sub-
cellular partitioning, where they support a variety of tasks essential for life. Moreover, lipids
also store energy; they serve as hormones and signaling molecules; participate in the
4
Chapter 1
maintenance of electrochemical gradients; and a recent described function as cofactors in

modulating protein activity (Coskun, Grzybek, Drechsel, & Simons, 2011).
1.1.3 Experimental approaches in lipidomics research
The workflow of lipidomics uses a set of tools defined as novel analytical approaches
mainly based on liquid chromatography (LC) and mass spectrometry (MS) (Han, 2016b; Wenk,
2010). A typical analytical procedure for the characterization of individual lipid compounds
includes four steps: sample preparation and lipid extraction; chromatographic separation;
identification and quantification; and data analysis and interpretation. Despite this, due to the
vast chemical diversity of lipid species, it is nearly impossible to analyze all lipid classes with
a standard method for extraction, chromatography, and detection. Therefore, numerous
methods are available for lipid analysis. It is the task of the analyst to choose the most
appropriate tool for each application. Many lipid analysis techniques are using advances in
electronics and computer science (especially the areas of data acquisition, data analysis, and
interpretation), but some “low tech” analytical methods, such as solid phase extraction (SPE)
and thin layer chromatography (TLC), can also provide valuable information at modest costs
(Christie & Han, 2010).
Lately, the global analysis of lipids, namely shotgun lipidomics has acquired interest. The
workflow of shotgun lipidomics is based on direct infusion of the sample into the mass
spectrometer. Because there is no chromatographical separation involved, several MS scan can
be performed for both identification and quantitation (Ryan & Reid, 2016). The disadvantage
of the shotgun lipidomics workflow is that many isobaric species may overlap. Therefore, high
mass resolution and/or multiple stages of MSn are required to separate and distinguish the many
isobaric and isomeric species found within biological lipid extracts. Additionally, ion
suppression and enhancement in a direct infusion hamper quantification. Nevertheless, MS is
the most preferred method of detection, enabling structure elucidation and quantification
studies at a previously unattainable level of complexity.
Lipid research has experienced overgrown concerning both developments and applications
since the development of numerous different LC systems and sensitive mass spectrometry
technologies for the quantitative characterization of the lipidome on the market (Holčapek,
Jirásko, & Lísa, 2012). Separation of lipid classes is generally achieved by adsorption
chromatography. Due to the complexity of the lipid extract, quite often, it is not possible to
analyze different lipid classes with a single analytical procedure. Therefore, it is required to
5
Chapter 1
isolate distinct lipid fractions prior to analysis. Typically, systems employing silica gel
columns, TLC, and solid phase extraction cartridges are the most applied. However, the
separation of individual lipid molecules entails the use of HPLC, mostly centered on reversed-
phase columns. Many other different types of column packing materials (bonded phases) such
as diol, nitrile, cyanopropyl, or polyvinyl alcohols are also used. Reversed-phase HPLC
separates the lipid molecules according to the combined chain-lengths of the fatty acid and the
double bonds in the fatty acids, which reduces the retention time of the lipid molecule by
approximately two carbon atoms. Silver ion chromatography is an auxiliary technique, which
separates the lipid molecules according to their degree of unsaturation. Simplifying the lipid
molecules by synthesis and extraction procedures of some of their building blocks, fatty acids
methyl esters, individual fatty acids can be determined. Fatty acids are usually analyzed by gas
chromatography (GC), following derivatization, which allows the separation of volatile
molecules. Fused silica gel column with a polyester phase, with different polarity, enable high
resolution and accurate quantitative analysis. With the non-polar phases, unsaturated fatty acids
are eluted before the saturated compounds of the same chain-length, while the opposite is true
for the polar phases (Christie & Han, 2010). Current lipidomics encompasses the combination
of different methods, technics and columns. However, the utmost effort is in the separation and
analysis of individual molecular species in its native form, to better assess its function and
physical properties in a living organism or material. This task is mostly achievable by HPLC.
The coupling of an HPLC with the high vacuum required for ion analysis in a mass
spectrometer obliges rapid evaporation of high amounts of liquid mobile phase to release the
molecules into the gas phase, along with ionization of those molecules so that the detector can
detect them. Nitrogen gas is usually the gas employed to spray the effluent stream, in the
process called nebulization. Nebulization and ionization regularly occur at atmospheric
pressure, and thus the formation of ions under these conditions is called “atmospheric pressure
ionization” (API). The two most commonly used API techniques are atmospheric pressure
chemical ionization (APCI), and electrospray ionization (ESI). In APCI the nebulized
solvent/analyte mixture is sprayed down a heated vaporizer tube to aid in desolvation, and the
vaporized mist passes by either a corona discharge needle held at high voltage (~2-5 kV).
Therefore, the nature of the solvents used in the LC mobile phase has a substantial effect on
the ionization of the analyte molecules (William C. Byrdwell, 2017).
Electrospray ionization uses a concentric nebulization gas to assist desolvation by charging
of the liquid to accomplish nebulization. It also includes the capability to provide heat to aid in
6
Chapter 1
desolvation, which allows higher flow rates. Evaporation of the solvent caused a decrease in
droplet size and simultaneously caused an increase in surface-charge density, finally leading to
desorption of molecular ions into the gas phase (Coulomb explosion) at atmospheric pressure.
ESI almost always produces an intact pseudomolecular ion, [M+H]+ (not to be confused with
a ‘molecular ion’, which is an odd-electron species formed by loss or addition of an electron
only), or adduct ion, [M+X]+ (X = Li, Na, NH4, etc.), and just a small abundances of fragments,
unless tandem MS is applied. While phospholipids are naturally amenable to ESI,
triacylglycerols frequently require the addition of an electrolyte, such as ammonium formate
or ammonium acetate, to produce adduct ions, [M+NH4]+ (i.e. ammonium ions in this example)
with small amount of Li and Na adduct in addition (William Craig Byrdwell, 2003).
APCI produces diacylglycerol-like fragment ions ([DAG]+), in case of TAGs, and their the
ratios are indicative of the composition of regioisomers. However, with this technique,
polyunsaturated TAGs produce a [M+H]+ base peak, while saturated TAGs often yield little or
no [M+H]+ ion, whereas with ESI-MS intact molecule adduct ions are obtained, but the ratios
of the regioisomers are less clear. Consequently, for these reasons, acquisition of both ESI-MS
and APCI-MS is desirable(William C. Byrdwell, 2017).
Electrospray ionization is the most popular ionization method as it provides minimal in-
source fragmentation and an outstanding sensitivity (femtomole range). This ionization
technique can comprehensively analyze thousands of lipid molecular species in an extract in a
reasonably short time (Christie & Han, 2010).
Although high-resolution mass spectrometry approaches were shown to be successful in
lipid profiling, and some structural elucidation, other methods such as chemical derivatization
or 2D NMR may be needed for the unequivocal assignment of structures such as double bond
configurations (Iijima et al., 2008).
1.1.4 Plant lipidomics
Both direct infusion-based shotgun lipidomics and LC-MS based approaches are applicable
in plant lipidomics (Han, 2016a). Due to the remarkable increase of the vegetable oil market,
both edible oils and sustainable substitutes for petrochemicals, plant lipid biochemistry is
facing massive growth. Moreover, there is a significant interest in widening the array of
applications and considering new techniques of oil-formation into non-traditional tissues. To
achieve this scope, it is necessary to fully understand the complex pathways at the core of plant
lipid synthesis, at both genetic and biochemical level. At this point in time, the overall "picture”
7
Chapter 1
of plant lipid metabolism can be grasped, yet there is still much left to learn. Lipidomics offers
the possibilities to couple functional genomics and systems biology with cutting-edge lipid
analytical techniques, bridging a vital gap between gene function and in vivo physiological
outcomes (Li-Beisson et al., 2013).
Plants accumulate lipids mostly in their seeds. The lipid content comprises mainly of
triacylglycerols and other various kinds of lipidic compounds including fatty acids,
tocopherols, phospholipids, sphingolipids, and sterols. The variation in the composition of
these compounds in plant seeds is very high and partially characteristic of individual plant
families. Databases containing a vast collection of information about the fatty acid composition
of wild plant seeds, of particular interest to food scientists, biochemists, chemists and botanists,
managers of food and oleochemical industry, is nowadays available online (Matthäus, 2012).
1.1.5 Lipidomics and food science
Oils and fats are a primary source of energy; provide essential fatty acids; and other minor
components as fat-soluble vitamins. In foods, lipids contribute to the texture and the flavor.
Human’s brain integrates the texture, taste, and flavor in the orbitofrontal cortex to give the
overall perception of a particular food. Sensorial perception of foods is affected by the
microstructural composition often determined by the emulsion properties associated with the
continuous phase. This fact denotes the complexity of the effect of lipids on the overall flavor
of a product. Nevertheless, the presence of lipid sensors in taste is still under discussion.
A significant source of oils are oilseeds and tropical fruits whereas fats are usually of animal
origin. TAGs account for over 95% of the components of a fat or oil. The remaining minor
constituents comprise phospholipids, glycolipids, tocopherols, sterols and other compounds
referred to as non-TAG components. During the processing, both TAG and minor constituents
undergo chemical and physical changes, which play a strong impact on the quality of the oils,
and hence the importance of post-extraction active processing, as, deodorization, degumming
or refining (Akoh, 2017; Shahidi & Weenan, 2006).
Lipidomics can be efficiently used to understand the effect of the processing on edible oils
and fats. Consequently, it can be used to streamline food handling and to enhance the
physicochemical properties of a desirable product. Nonetheless, it aids understanding the
dietary value, and provides indicators for the evaluation of food-related health effects and
toxicity, through monitoring the lipids changes (Sebedio & Brennan, 2014).
8
Chapter 1
1.2 Theobroma cacao – from bean to bar
Theobroma cacao is a small evergreen perennial crop tree in the family Sterculiaceae. The
scientific name was given for the first time, in 1753, by the Swedish botanist Carl Linnaeus in
his book Species Plantarum. Theobroma means "food of the gods" and cacao is derived from
the Nahuatl (Aztec language) word xocolatl, from xococ (bitter) and atl (water) (Coe & Coe,
2007).
1.2.1 History, origin, and cocoa varieties
The story of cocoa begins around 4,000 years ago when the first cocoa plantations were
established by Olmecs in the south of Yucatan (Mexico) and Central America (Coe & Coe,
2007). The Maya and the Aztecs of Mexico after them, as well as the Incas from Perù were
also growing cocoa trees. Thus, cocoa beans have been at the center of Mesoamerican
mythology for thousands of years (Coe & Coe, 2007; Dreiss & Greenhill, 2008). Centuries
before, a sophisticated process was necessary to transform the bitter seeds into what is known
today as chocolate, it was consumed as unsweetened liquid form and used as currency by the
Maya and the Aztecs. The chocolate was introduced to Europe in the 1520s, after the Spanish
conquest of Central America. At first, it was consumed as a drink by the kings and aristocrats
of Spain, France, United Kingdom, and Italy and then it was spread in coffeehouses. Later, in
the nineteenth and twentieth centuries, industrialization made chocolate available for all, and
now, it has turned out to be once again a more than just a bar – a luxury product (Beckett, 2008;
Coe & Coe, 2007).
There are four different species of cocoa tree: Criollo – produces pods with a very thin peel
and beans with white cotyledons. This variety produces small harvests, and it is very fragile,
but its cocoa has a unique, refined aroma; Forastero – is a more vigorous type of tree and it is
easier to cultivate and produces larger harvests. The fruit has a thicker peel and a stronger
aroma. Cocoa from the Forastero beans is often called bulk cocoa because it gives chocolate a
typical chocolate aroma; Trinitario – is a cross between Criollo and Forastero and is not found
in the wild. Its characteristics are a strong but relatively refined aroma moreover and it is easier
to cultivate, and National – the fourth type of cocoa tree is grown only in Ecuador and probably
originates from the Amazonian area of Ecuador. Cocoa from National beans has a full cocoa
flavor with additional floral and spicy flavors (Beckett, 2008; Coe & Coe, 2007a; Young,
2007).
9
Chapter 1
International Cocoa Germplasm Database (“ICGD Homepage”) stores records on about

30,000 reported cocoa clone names but only slightly more than 1,000 clones could be grouped
into ten cocoa varieties (genetic groups). For example, the clone CCN51 is a hybrid of tree
varieties called Amelonado, Iquitos and Criollo. Motamayor et al. proposed a new
classification based on the analyses of 1241 genotyped individuals of wild and cultivated cacao
covering a large geographical area with 106 microsatellite markers that resulted in 10 genetic
clusters named according to the geographical location: Amelonado, Contamana, Criollo,
Curaray, Guiana, Iquitos, Maranon, Nacional, Nanay, and Purus. In this way, Forastero
definition disappears because it includes various genetic clusters, while Criollo definition
remains (Motamayor et al., 2008).
1.2.2 Agriculture
Cocoa trees naturally grow within 20° latitude of the equator because they need about 2000
mm of rainfall a year and temperatures in the range of 21 to 32 °C. The humidity should be
very close to 100% during the day and between 70-80% during the night. Very often cocoa
crops are cultivated in mixed crops among tall standing plants like bananas or palms, which
offer shadow to cocoa plants protecting them from the sun. The soil needs to be permeable and
nutrients rich to a depth of around 1.5 m, this will allow the roots to grow. The pH of the soil
should be between 5 and 7.5 as lower, or higher pH would be harmful to the plant. The sowing
density affects the yield; thus, most of the farmers use a density between 1000 and 1200 trees
per hectare (Young, 2007). The tree starts bearing fruits in the 5-6th year of its life when peak
production levels occur, and they can continue at this level for ten years. The average life of a
cocoa tree is 25 years, and after this time, the trees must be replaced to provide the same yield
from year to year. The fruit is called cacao pod and it is ovoid, 15-30 cm long, and 8-10 cm
wide, colored yellow to red-orange and weighs about 500 g when ripe. The cocoa tree flowers
in 2 cycles of 6 months the year around and about 40 flowers develop into cocoa pods. The pod
contains up to 60 seeds, usually called “beans”, implanted in a white pulp (Cuatrecasas, 1964;
Young, 2007). Around 80% to 90% of cocoa comes from small, family-run farms, with
approximately five to six million cocoa farmers worldwide. In Africa and Asia, the typical farm
covers two to four hectares. Each hectare produces 300 to 400 kilograms of cocoa beans in
Africa and about 500 kilograms in Asia. Cocoa farms in the Americas tend to be a little larger
and produce 500 to 600 kilograms of cocoa beans per hectare. More than 75% of the global
production in the last year originated from African countries, while the leading cocoa bean
10
Chapter 1
processing countries are European countries grinding 38.7% of the 4.6 million tons of cocoa
harvested in 2016. Indonesia is third in world exports with around 290 thousand tons produced
last year (“The International Cocoa Organization (ICCO) | Cocoa Producing and Cocoa
Consuming Countries,”). Other counties as Ecuador, Cameroon, Brazil, Madagascar, Nigeria, Sri
Lanka, and Venezuela export significant amounts of beans. Cocoa is also cultivated for export in
Columbia, Tanzania, Congo/Zaire, Costa Rica, Cuba, Dominican Republic, Fiji, Gabon, Grenada,
Haiti, Jamaica, Malaysia, South Central Mexico, Panama, Papua New Guinea, Peru, Philippines,
Sao Tome, Sierra Leone, Togo, Trinidad and Western Samoa (Update, 2014).
Cacao trees produce pods throughout the year, but large harvests occur twice annually. The
main crop is harvested from October to March and is usually of a higher quality than midcrop,
harvested from May to August. There are no machines for harvesting cacao so the process may
take weeks to accomplish. Cacao trees produce flowers and fruit year-round in this way
machines could harm the clusters of flowers or pods that grow from the trunk. Workers must
harvest the pods by hand, using short, hooked blades mounted on long poles to reach the highest
fruit. The pods are collected into baskets and then carried to the edge of the field.
1.2.3 Post-harvesting treatments
Collected pods are opened one by one; the pulp-covered cacao seeds are scooped out and
quickly placed into boxes, or heaped into piles, and covered with banana leaves. For the next
three to seven days, the intense bitter seeds ferment, allowing natural yeasts and bacteria to
grow and produce physicochemical transformations into the beans. Fermentation is the first
essential step that creates the familiar chocolate flavor (Afoakwa, Paterson, Fowler, & Ryan,
2008). This process is divided into 2 phases: a first anaerobic and a second aerobic phase. The
anaerobic phase takes place in the first 24-48 h in which the beans are exposed to numerous
various microorganisms. The major species include Hanseniaspora sp., Pichia sp.,
Saccharomyces cerevisiae for the yeasts and Acetobacter sp., Lactobacillus sp. and
Leuconostoc sp. for the bacteria (Ho, Zhao, & Fleet, 2014). During this stage, the initial pH is
bellow 4 and yeasts produce ethanol from the sugar present in the pulp. Moreover, some yeasts
degrade the pectin of the pulp cell walls, which favor aeration (Aprotosoaie, Luca, & Miron,
2016). After 48 to 96, the aerobic phase takes place in which the yeasts are inhibited by
aeration, alcohol concentration and increased pH, giving space to the growth of lactic acid
bacteria. At the end of this phase, lactic acid bacteria are replaced by acetic acid bacteria which
are responsible for the oxidation of ethanol to acetic acid in an exothermic reaction which leads
11
Chapter 1
to a temperature increase up to 45-52 °C. During the fermentation process ethanol, lactic acid,
acetic acid, carbon dioxide are produces and the pH values increases. These compounds and
pH conditions play a key role in the formation of flavor precursor. Acetic acid diffuses into the
bean causing the death of the embryo and the pH values of 3.8 and 5.8 offer the optimal
conditions for the activity of numerous enzymes as, for instance, proteases. Hence, complex
biochemical reactions occur generating flavor precursor as reducing sugars, peptides and
amino acids, oxidized polyphenols and other compounds. Besides, the color of the beans
changes developing from purple or light color to brown color (Afoakwa et al., 2008;
Aprotosoaie et al., 2016; Giacometti, Jolić, & Josić, 2015). If the seeds are over-fermented,
unwanted molds may grow which produce the hammy off-flavor compounds. Moreover, the
pH value continues to increase causing a consequent darkening of the beans (Biehl & Ziegleder,
2003; Schwan & Wheals, 2004). Most of the beans around the world are fermented under
spontaneous conditions. However, controlled conditions using specific culture strains are
implemented as well (De Vuyst L. & Weckx S., 2016).
After fermentation, the beans have to be dried. Drying process lasts between 6 to 14 days
and consists of spreading out the beans in the sun or under solar dryers. The beans are regularly
turned so that they retain just a fraction of their moisture content (less than 8%). Drying is
crucial, both for stopping the fermentation process and for storage; hence, to avoid
microorganisms growth. Sometimes farmers are compelled to use equipment like hot pipes or
air dryers to speed the drying process and this can generate changes in the cocoa flavor.
Once the cacao seeds are dry, the cocoa farmers pack them into jute sacks immediately.
They inspect the crops to make sure that they have properly fermented. After weighing and
packing them into bales of 50-60 kg jute sacks or in bulk or containers, cocoa beans are shipped
to their final destinations – chocolate and cocoa-making companies (Beckett, 2008).
1.2.4 Factory production process
Once the shipments of seeds have reached the factories, they quickly begin the process that
transforms the beans into chocolate or other cocoa products. The conversion is a complex and
time-consuming process. An individual chocolate bar production can take anywhere from two
to four days, or more. Initially, the seeds are sorted according to type and country of origin.
They are passed through a cleaning machine that removes bits of remaining pulp and debris.
Each manufacturer uses special formulations to blend the seeds to create different types of
chocolates. African beans, for instance, are known for the basic aroma that they give to
12
Chapter 1
chocolate. American and Asian beans give the chocolate a more refined aroma. Then large,
rotating ovens at temperatures of 120 °C or more are used to roast the seeds in order to release
their rich aromas and delicious taste. Roasting is the second most important step in the cocoa
processing. This process takes from 15 minutes to two hours and it depends upon the variety
of the seeds. The beans lose much of their moisture during the roasting. Eventually, they turn
into a deep brown color, similar to coffee beans. A sterilization step is normally applied in order
to destroy the heat-resistant bacteria and spores even if most of the microorganisms are
destroyed in the early stages of roasting under moist conditions. The procedure consists of an
injection, over a period of 20 sec, of a fine water spray into the roasting drum at the end of the
roasting period. This procedure guarantees acceptable values in terms of bacteria and spores
for a good quality cocoa. The roasting processes makes the shells of the cacao rather brittle and
after the beans are cooled, the giant winnowing machine open and remove the thin shells. Giant
fans then blow away these empty crusts. The remaining broken seed bits called nibs are then
passed through a series of sieves, which strain and sort the nibs according to their size, a process
called “winnowing”. Sometimes this process along with the roasting occurs together.
In the next step, the nibs are milled – crushed by heavy steel discs, hammer mills, ball mills,
three-roll refiners or a combination of the four. During this process, sufficient friction and heat
are generated which liquefies the nibs into a thick paste, called chocolate liquor. The heat
emanated by this process ensures that the cocoa butter present in the liquor melts, making the
cocoa liquor into liquid. The cocoa liquor is used as an ingredient of chocolate, or it can be
further processed into cocoa powder and cocoa butter. The nibs themselves are made up of 53-
55% cocoa butter and 45-47% pure cocoa solids. The separation of these two substances is a
laborious process. Huge hydraulic presses are used to squeeze out the cocoa butter from the
chocolate liquor. Once cocoa butter is obtained, the remaining solid cocoa is pulverized into
cocoa powder – both are essential for several different applications as beverages, cooking,
chocolate, baking, and in cosmetics.
About 50% of the cocoa liquor that is used to produce cocoa powder is alkalized. Solutions
of alkali, typically potassium, ammonium or sodium carbonate are commonly employed.
Usually, the solutions are added to the cocoa nibs before roasting, even though the cocoa liquor
or the cocoa powder can also be treated. The initial reason for doing this procedure was to
make the powder less predisposed to agglomerate or sink to the bottom, when mixed with milk
or water-based drink. Now the alkalization is mainly used to produce changes in the color and
the flavor.
13
Chapter 1
For chocolate production, manufacturers mix cocoa liquor with milk powder, sugar, and
extra cocoa butter. The extra cocoa butter keeps the chocolate solid at room temperature and
that is why it melts in the warmth of the mouth. The raw mixture of milk powder, liquor, sugar,
and cocoa butter is churned until it becomes a coarse, brown powder called “crumb”. The
ingredients are first weighed according to the recipe for the particular type of chocolate, mixed,
and then blended into homogeneous chocolate mass. The chocolate crumb mixture is then
pressed between rollers to form a fine powder. The rollers break down the tiny particles of milk
powder, cocoa, and sugar within the crumb and give a smooth texture and a homogenous flavor
to the finished chocolate. The average size of the particles in this chocolate powder should be
smaller than the distance between the taste buds on our tongues – so that we can never
physically feel them, this guarantees a smooth chocolate. There is a right proportion in the
mixing of this mixture – not enough crushing will give a coarse and grainy chocolate and too
much blending will give a pasty and gummy chocolate. In general, Swiss and German
chocolates are refined for a longer period. This makes them smoother and finer than American
or English candy, which is a matter of national taste preferences. Subsequently, the refined
chocolate paste is transferred into a vat in which a large heavy roller kneads, blends, and grinds
it for hours until the aromas have fully developed. These machines are called conches and they
are large tanks with a powerful stirring apparatus inside that slowly kneads the mixture. Due
to the friction caused by the stirring, heat develops. This heat melts the powder into a
homogeneous paste and makes the unpleasant, acid aromas evaporate. Moreover, agitating and
aerating this paste smoothes out the sugar grains and give a silky texture to the chocolate, which
creates a mellower, more well-rounded flavor. This process, called “conching” can take up to
six days to complete.
The refined chocolate is cooled and warmed repeatedly in a process called “tempering”.
This gives to the chocolate its glossy sheen, and ensures that it will melt properly. Tempering
ensures the formation of the right cocoa butter crystals so that the chocolate will harden into
shiny, hard and solid shapes. Finally, only after it has been tempered, the chocolate can be
poured into molds or processed further into blocks, drops or other solid shapes. In a classic
assembly-line fashion, the machines squirt tempered chocolate into several hundred molds per
minute. Some devices dispense chocolate over flavored centers, in a process called “enrobing”.
Others create chocolate shapes that will be filled with liquid before their bottoms are sealed.
Eventually, wrapping-and-packaging machines box the chocolates at speeds unattainable by
human hands (Beckett, 2008; Minifie, 1989).
14
Chapter 1
1.2.5 Economic impact and health benefits
It has been discovered that cocoa was already playing an important economic and social
role in the communities living in the Ulúa valley in Honduras about some thousand years ago.
Nowadays, people consume more than 4.5 million tons of cocoa beans annually in a multitude
of different forms. Each country has unique preferences and distinctive assortments of candy
and desserts (Young, 2007). More than half of the top 20 consuming countries are European
with people in the Northern nations in average consuming more than people in the South.
Cocoa is a cash crop and a critical export commodity for producing countries and is also a key
import for consuming countries, which typically do not have suitable climates for cocoa
production. When traveling a global supply chain, cocoa beans go through a complicated
production process that includes farmers, buyers, shipping organizations, processors,
manufacturers, chocolatiers, and distributors. Many people are involved in cocoa and chocolate
industry all around the world, and they are crucial users of other agricultural commodities such
as sugar, dairy, nuts, and fruits. According to the World Cocoa Foundation, some 50 million
people around the world depend on cocoa as a source of livelihood. Progressively increasing
demand from worldwide consumers encourages some global efforts and funds committed to
support and improve cocoa farm sustainability (“World Cocoa Foundation,”). The world leader
in business news, CNBC, reports that the overall chocolate market between 2010 and 2015
rose to hit 101 billion US $, according to a market research firm. Cocoa prices had increased
over the last four months reaching the price of € 2,259.33 per metric ton on April 26th, 2018
(“ICCO Daily Prices of Cocoa Beans,”).
Evidence for the use of cacao or chocolate for medicinal purposes can be traced to ancient
Aztecs sources. Moreover initially, Spanish and Italian empire entered cocoa in their
pharmacopoeia (Coe & Coe, 2007). Cocoa is particularly rich in polyphenols. The total
polyphenol content of the bean arises 6% to 8% by weight of the dry bean (Rusconi & Conti,
2010; Zumbé, 1998) with monomeric, oligomeric, and polymeric procyanidins being the main
constituents. They are well-known antioxidant agents, which reduce the risk of cardiovascular
diseases. The European Food Safety Authority (EFSA) has recently recognized that cocoa is a
precious source of compounds with specific antioxidant properties (flavanols). Daily
consumption of cocoa powder (2.5 g) or dark chocolate (10 g) can help protect the walls of the
arteries, stimulate blood circulation, and regulate blood pressure; thus, safeguarding the heart
(EFSA Panel on Dietetic Products, 2014). A long-term feeding study to examine the relative
15
Chapter 1
activities of cocoa constituents on diet-induced obesity and insulin resistance was

accomplished by Dorenkott, M. R. et al. (2014). The oligomer-rich fraction showed to be most
effective in preventing weight gain, fat mass, impaired glucose tolerance, and insulin resistance
in mice model (Dorenkott et al., 2014). Other authors’ results claim that stearic acid is a unique
saturated FA in cocoa beans, having unique effects on plasma total cholesterol levels (Kris-
Etherton et al., 1993). Although the mechanism of its hypocholesterolemic action is still not
clear, a short-term cocoa consumption significantly reduced blood cholesterol, but the changes
were dependent on the dose of cocoa consumption and the health status of participants (Jia et
al., 2010). Cocoa butter has a high digestibility similar to that of corn oil (Shahkhalili, Duruz,
& Acheson, 2000), and its absorption can be reduced by calcium supplementation (Shahkhalili
et al., 2000).
Cocoa butter is used on a tiny scale in the pharmaceutical industry for the preparation of
suppositories and in the cosmetics industry for the manufacturing of lipstick and body lotions.
Some decennia ago cocoa butter was used intensively for the production of cosmetics, but now
it has mostly been replaced by synthetic products (Pontillon, 1998), which are preferred
because of their price despite their high content of mineral oil hydrocarbons (MOSH and
MOAH). Niederer reported that in 38 paraffin oils contained in lip-gloss products, MOAH
range is between 0.05-4.5% (Niederer, 2014), whereas in cocoa beans, MOSH and MOAH
represent contaminants absorbed during the transportation and storage, and their concentration
is much lower. EFSA has expressed concern and issued regulation since some mineral oil
aromatic hydrocarbons (MOAH) may be mutagenic and carcinogenic (“Scientific Opinion on
Mineral Oil Hydrocarbons in Food,” 2012). Nevertheless, the most important use of cocoa
butter is, of course, the preparation of chocolate.
References
Afoakwa, E. O., Paterson, A., Fowler, M., & Ryan, A. (2008). Flavor Formation and Character in Cocoa and
Chocolate: A Critical Review. Critical Reviews in Food Science and Nutrition, 48(9), 840–857.
https://doi.org/10.1080/10408390701719272
Akoh, C. C. (2017). Food Lipids: Chemistry, Nutrition, and Biotechnology, Fourth Edition. CRC Press.
Aprotosoaie, A. C., Luca, S. V., & Miron, A. (2016). Flavor Chemistry of Cocoa and Cocoa Products—An
Overview. Comprehensive Reviews in Food Science and Food Safety, 15(1), 73–91. https://doi.org/10.1111/1541-
4337.12180
16
Chapter 1
Beckett, S. T. (2008). The Science of Chocolate. Royal Society of Chemistry.
Biehl, B., & Ziegleder, G. (2003). COCOA | Chemistry of Processing. In B. Caballero (Ed.), Encyclopedia of
Food Sciences and Nutrition (Second Edition) (pp. 1436–1448). Oxford: Academic Press.
https://doi.org/10.1016/B0-12-227055-X/00261-3
Brügger, B. (2014). Lipidomics: Analysis of the Lipid Composition of Cells and Subcellular Organelles by
Electrospray Ionization Mass Spectrometry. Annual Review of Biochemistry, 83(1), 79–98.
https://doi.org/10.1146/annurev-biochem-060713-035324
Byrdwell, William C. (2017). Chapter 10 - Multiple Parallel Mass Spectrometry for Liquid Chromatography. In
Handbook of Advanced Chromatography/Mass Spectrometry Techniques (pp. 365–405). AOCS Press.
https://doi.org/10.1016/B978-0-12-811732-3.00010-8
Byrdwell, William Craig. (2003). Dual Parallel Liquid Chromatography/Dual Mass Spectrometry (LC2/MS2) of
Bovine Brain Total Lipid Extract. Journal of Liquid Chromatography & Related Technologies, 26(19), 3147–
3181. https://doi.org/10.1081/JLC-120025516
Christie, W. W., & Han, X. (2010). Lipid Analysis, Fourth Edition: Isolation, Separation, Identification and
Lipidomic Analysis (4 edition). Bridgwater, England: Oily Press.
Coe, S. D., & Coe, M. D. (2007). The True History of Chocolate (Second Edition edition). New York: Thames &
Hudson.
Coskun, Ü., Grzybek, M., Drechsel, D., & Simons, K. (2011). Regulation of human EGF receptor by lipids.
Proceedings of the National Academy of Sciences, 108(22), 9044–9048. https://doi.org/10.1073/pnas.1105666108
Cuatrecasas, J. (1964). Cacao and its allies a taxonomic revision of the genus Theobroma. Washington:
Smithsonian Institution.
De Vuyst L., & Weckx S. (2016). The cocoa bean fermentation process: from ecosystem analysis to starter culture
development. Journal of Applied Microbiology, 121(1), 5–17. https://doi.org/10.1111/jam.13045
Dorenkott, M. R., Griffin, L. E., Goodrich, K. M., Thompson-Witrick, K. A., Fundaro, G., Ye, L., … Neilson, A.
P. (2014). Oligomeric Cocoa Procyanidins Possess Enhanced Bioactivity Compared to Monomeric and Polymeric
Cocoa Procyanidins for Preventing the Development of Obesity, Insulin Resistance, and Impaired Glucose
Tolerance during High-Fat Feeding. Journal of Agricultural and Food Chemistry, 62(10), 2216–2227.
https://doi.org/10.1021/jf500333y
Dreiss, M. L., & Greenhill, S. E. (2008). Chocolate: Pathway to the Gods (Har/DVD edition). Tucson: University
of Arizona Press.
17
Chapter 1
EFSA Panel on Dietetic Products, N. and A. (NDA). (2014). Scientific Opinion on the modification of the
authorisation of a health claim related to cocoa flavanols and maintenance of normal endothelium‐dependent
vasodilation pursuant to Article 13 (5) of Regulation (EC) No 1924/2006 following a request in accordance with
Article 19 of Regulation (EC) No 1924/2006. EFSA Journal, 12(5), 3654.
Giacometti, J., Jolić, S. M., & Josić, D. (2015). Chapter 73 - Cocoa Processing and Impact on Composition. In V.
Preedy (Ed.), Processing and Impact on Active Components in Food (pp. 605–612). San Diego: Academic Press.
https://doi.org/10.1016/B978-0-12-404699-3.00073-1
Han, X. (2016a). Lipidomics: Comprehensive Mass Spectrometry of Lipids. John Wiley & Sons.
Han, X. (2016b). Lipids and Lipidomics. In Lipidomics (pp. 1–20). John Wiley & Sons, Inc.
https://doi.org/10.1002/9781119085263.ch1
Ho, V. T. T., Zhao, J., & Fleet, G. (2014). Yeasts are essential for cocoa bean fermentation. International Journal
of Food Microbiology, 174, 72–87. https://doi.org/10.1016/j.ijfoodmicro.2013.12.014
Holčapek, M., Jirásko, R., & Lísa, M. (2012). Recent developments in liquid chromatography-mass spectrometry
and related techniques. Journal of Chromatography A, 1259, 3–15. https://doi.org/10.1016/j.chroma.2012.08.072
ICCO Daily Prices of Cocoa Beans. (n.d.). Retrieved April 27, 2018, from https://www.icco.org/statistics/cocoa-
prices/daily-prices.html
ICGD Homepage. (n.d.). Retrieved April 26, 2018, from http://www.icgd.reading.ac.uk/
Iijima, Y., Nakamura, Y., Ogata, Y., Tanaka, K., Sakurai, N., Suda, K., … Shibata, D. (2008). Metabolite
annotations based on the integration of mass spectral information. The Plant Journal, 54(5), 949–962.
https://doi.org/10.1111/j.1365-313X.2008.03434.x
Jia, L., Liu, X., Bai, Y. Y., Li, S. H., Sun, K., He, C., & Hui, R. (2010). Short-term effect of cocoa product
consumption on lipid profile: a meta-analysis of randomized controlled trials. The American Journal of Clinical
Nutrition, 92(1), 218–225. https://doi.org/10.3945/ajcn.2009.28202
Kris-Etherton, P. M., Derr, J., Mitchell, D. C., Mustad, V. A., Russell, M. E., McDonnell, E. T., … Pearson, T. A.
(1993). The role of fatty acid saturation on plasma lipids, lipoproteins, and apolipoproteins: I. Effects of whole
food diets high in cocoa butter, olive oil, soybean oil, dairy butter, and milk chocolate on the plasma lipids of
young men. Metabolism: Clinical and Experimental, 42(1), 121–129.
Li-Beisson, Y., Shorrosh, B., Beisson, F., Andersson, M. X., Arondel, V., Bates, P. D., … Ohlrogge, J. (2013).
Acyl-Lipid Metabolism. The Arabidopsis Book, e0161. https://doi.org/10.1199/tab.0161
LIPID MAPS Lipidomics Gateway : Home. (n.d.). Retrieved January 22, 2018, from http://www.lipidmaps.org/
18
Chapter 1
Matthäus, B. (2012). The new database Seed Oil Fatty Acids (SOFA). Lipid Technology, 24(10), 230–234.
https://doi.org/10.1002/lite.201200227
Meer, G. van. (2005). Cellular lipidomics. The EMBO Journal, 24(18), 3159–3165.
https://doi.org/10.1038/sj.emboj.7600798
Minifie, B. W. (1989). Chocolate, Cocoa and Confectionery: Science and Technology. Springer Science &
Business Media.
Motamayor, J. C., Lachenaud, P., Mota, J. W. da S. e, Loor, R., Kuhn, D. N., Brown, J. S., & Schnell, R. J. (2008).
Geographic and Genetic Population Differentiation of the Amazonian Chocolate Tree (Theobroma cacao L).
PLOS ONE, 3(10), e3311. https://doi.org/10.1371/journal.pone.0003311
Navas-Iglesias, N., Carrasco-Pancorbo, A., & Cuadros-Rodríguez, L. (2009). From lipids analysis towards
lipidomics, a new challenge for the analytical chemistry of the 21st century. Part II: Analytical lipidomics. TrAC
Trends in Analytical Chemistry, 28(4), 393–403. https://doi.org/10.1016/j.trac.2008.12.004
Niederer, M. (2014). 1.1. 1 Lippenpflegeprodukte/Mineralparaffine (MOSH/MOAH) und allergene Duftstoffe.

Food and Chemical Toxicology, 72, 312–321.
Pontillon, J. (1998). Cocoa and chocolate: production, utilization, characteristics. Cocoa and chocolate:
production, utilization, characteristics. Retrieved from
https://www.cabdirect.org/cabdirect/abstract/20026788577
Published Articles | Blommer Chocolate Company. (n.d.). Retrieved March 23, 2018, from
http://www.blommer.com/articles.php
Rusconi, M., & Conti, A. (2010). Theobroma cacao L., the Food of the Gods: A scientific approach beyond myths
and claims. Pharmacological Research, 61(1), 5–13. https://doi.org/10.1016/j.phrs.2009.08.008
Schwan, R. F., & Wheals, A. E. (2004). The Microbiology of Cocoa Fermentation and its Role in Chocolate
Quality. Critical Reviews in Food Science and Nutrition, 44(4), 205–221.
https://doi.org/10.1080/10408690490464104
Scientific Opinion on Mineral Oil Hydrocarbons in Food. (2012). EFSA Journal, 10(6).
https://doi.org/10.2903/j.efsa.2012.2704
Sebedio, J.-L., & Brennan, L. (2014). Metabolomics as a Tool in Nutrition Research. Elsevier.
Shahidi, F., & Weenan, H. (Eds.). (2006). Food Lipids: Chemistry, Flavor, and Texture (1 edition). Washington,
DC : New York: American Chemical Society.
19
Chapter 1
Spener, F., Lagarde, M., Géloên, A., & Record, M. (2003). Editorial: What is lipidomics? European Journal of
Lipid Science and Technology, 105(9), 481–482. https://doi.org/10.1002/ejlt.200390101
The International Cocoa Organization (ICCO) | Cocoa Producing and Cocoa Consuming Countries. Retrieved
April 26, 2018, from https://www.icco.org/
Update, C. M. (2014). World Cocoa Foundation. Retrieved March.
Wenk, M. R. (2005). The emerging field of lipidomics. Nature Reviews Drug Discovery, 4(7), 594.
https://doi.org/10.1038/nrd1776
Wenk, M. R. (2010). Lipidomics: New Tools and Applications. Cell, 143(6), 888–895.
https://doi.org/10.1016/j.cell.2010.11.033
World Cocoa Foundation. Retrieved February 26, 2018, from http://www.worldcocoafoundation.org/
Young, A. M. (2007). The Chocolate Tree: A Natural History of Cacao (Enlarged/Expanded edition). Gainesville:
University Press of Florida.
Zumbé, A. (1998). Polyphenols in cocoa: are there health benefits? Nutrition Bulletin, 23(1), 94–102.
https://doi.org/10.1111/j.1467-3010.1998.tb01088.x
20
Chapter 2
Chapter 2:
2 Review on cocoa lipidomics – state of knowledge

and future needs
Diana Sirbua and Nikolai Kuhnerta*
a
Department of Life Sciences and Chemistry, Jacobs University, Campus Ring 1, D-28759 Bremen, Germany
*Correspondence to:
Prof. Dr. Nikolai Kuhnert,
Analytical Chemistry Laboratory,
Jacobs University, Campus Ring 1,
D-28759 Bremen, Germany.
Phone: +49 421 200 3120,
Fax: +49 421 200 3102.
E-mail: [email protected]
21
Chapter 2
Abstract
Cocoa butter is an essential ingredient in many commercial products. The chemical

composition of cocoa butter can vary depending on the growing conditions and on the
production process of the cocoa on the one hand, and on the refining treatments of cocoa butter
on the other hand. These slight differences in the chemical composition also lead to small
differences in the physical properties, such as the crystallization kinetics.
This review provides a state of the art information on chemical compositions of cocoa butter,
taking into account the paramount interest on the aspects essential for a good crystallization
performance. Besides some links are drawn helping the understanding of physical properties
of cocoa butter based on its chemistry. Understanding better the chemical composition and
consequently, its correlation with the physical properties of the cocoa butter will be of excellent
use for setting adequate, cost-effective parameters for the tempering of chocolate and the
development of products with tailored melting characteristics.
Keywords: Cacao butter, chemistry, fatty acids, triacylglycerols, minor components, physical
properties, polymorphism.
22
Chapter 2
2.1 Introduction
Cocoa butter (CB) is an ivory colored fat extracted from cocoa beans and constitutes the
primary part of the final industrial product of cocoa. The physical and chemical characteristics
of CB have made this fat of high demand in food, cosmetic and pharmaceutical industries
(Liendo, Padilla, & Quintana, 1997). Identified as the primary transport and suspension
medium (continuous phase) for the cocoa powder, chocolate liquor, sugar and several other
ingredients used in chocolate production, CB gives chocolate its substance, consistency,
plasticity, natural diffusion of taste and flavor, and the typical melt-in-the-mouth characteristics
(Chaiseri & Dimick, 1989; Lipp & Anklam, 1998; Ribeiro et al., 2013). Since cocoa butter has
a cocoa flavor and aroma, it is used for making white chocolate as the only cocoa ingredient
(Beckett, 2011).
Cocoa Butter is considered to be a precious commodity because not only cocoa trees are
difficult to cultivate, and the extraction of this fat is a very complex and costly process, but
also because the income derived from the sale of CB constitutes the most significant part of
the entire cocoa market value (Badrie et al., 2015). Global cocoa production is increasing every
year and in 2016/2017 cocoa year (Oct-Sep) has reached 4,748 thousand tons with more than
90% produced by the smallholder farmers. Cocoa butter prices have seen fluctuations over the
years. The year of 2014 has seen cocoa butter prices increases by 28%, to as high as $ 8,200 a
ton, because the global grinders cut processing to reduce inventory (“As cocoa price soars,
chocolate makers devour substitutes,” Reuters, 2014). However, it has declined progressively
over the last years. Since the start of the 2017/2018 crop year, cocoa butter prices followed a
near upward trend and stood at over $ 5,000 per ton, reflecting the growing demand for cocoa
registered in recent months, as reported by the leading regional cocoa associations for the
fourth quarter of 2017. Worldwide the annual trade of cocoa butter exceeds 800,000 tons (“The
International Cocoa Organization (ICCO)| Cocoa Producing and Cocoa Consuming
Countries,”).
The quality of commercial cocoa butter is defined by a sum of properties including
appearance, snap, hardness, elasticity, and smoothness, which regulates its use in
confectionery, cosmetic, and pharmaceutical applications. Nevertheless, there is still a
significant knowledge gap on how the production process affects the character of the final
product. Despite many investigators have directed their attention towards the physical
23
Chapter 2
properties of cocoa butter and its significance in the confectionary industry, striving to connect
them to the chemical properties of this fat has so far not been possible.
While extensive work has addressed the understanding of the chemical changes occurring
during cocoa bean processing within constituent proteins, carbohydrates and polyphenols,
considered as the prime reagents of the Maillard reaction and flavor precursors, little research
has been done so far on the chemical changes occurring within cocoa lipids, which account for
50% or more of the dry mass of cocoa beans. Recently, Garti & Widlak wrote a book covering
mostly the progress on the physical characteristics of the past ten years’ studies on cocoa butter
(Garti & Widlak, 2015).
Therefore, the motivation behind writing this review is primarily to give researchers and
industry a valuable concise overview of the state of art of cocoa butter research covering the
progress of many years of studies on cocoa butter chemistry. Our principal focus is to describe
the cocoa butter chemical composition in detail and to emphasize the importance of
understanding how this chemical composition may reflect the physical properties of the final
product (like crystallization behavior, etc.). Moreover, the data obtained by various authors
were compared to individualize the possible missing pieces of the puzzle thus, to relate it to
the lipid composition and hardness/softness characteristics of cocoa butter. Additionally, the
review attempts to correlate numerous studies conducted worldwide to investigate the
crystallization properties of cocoa butter concerning the various constituents of this fat.
2.2 From cocoa tree to cocoa butter
Cocoa beans are the seeds of the Theobroma cacao tree growing on small farms in tropical
environments. Three main species rule over the global cocoa production, namely: Forastero,
Criollo, and Trinitario. About 95% of the world’s cacao is now derived from Forastero trees,
and in the trade, this cacao is termed bulk cacao. Criollo has white cotyledons and is the
originally cultivated type. The origin of the Trinitarios is usually asserted as the result of
hybridization between Forastero and Criollo trees. The Trinitario and Criollo varieties produce
mainly the fine flavor cocoa. Plant characteristics, genetic diversity and agronomy approaches
have been reviewed and acknowledged by Lima et al. and recently by Badrie et al. (Badrie,
Bekele, Sikora, & Sikora, 2015; Lima, Almeida, Nout, & Zwietering, 2011).
The cocoa fruits called pods, sprout directly from the trunk and the thicker branches of the
cocoa tree. Being a delicate and sensitive crop, cocoa trees must be protected from the wind
24
Chapter 2
and sun, as well as from pests and disease attacks and soil fertilization is required. With proper
care, most cocoa trees begin to yield pods at peak production levels by the fifth year of life,
which can
continue for another ten years. Cocoa pods grow throughout the year starting from thousands
of small flowers, but normally there are two harvests periods per year, namely main crop,
harvested from October to March, and mid-crop, harvested from May to August. Usually,
during the main-crop trees produce higher quantity and quality of pods than the mid-crop. Each
pod contains 30 to 40 seeds implanted in a mucilaginous pulp, which accounts for
approximately 40% of the fresh beans mass (Schwan & Wheals, 2004). The pulp is
characterized mainly by high water content, 80-86% (w/w), and around 9 to 13% (w/w) sugar.
A seed consists of two cotyledons (the nib) and a small embryo plant, all enclosed in a skin
(the shell). In the cotyledons fat is the storage substance, which constitutes the food for the
developing plant and represents half of the weight of the dried beans (Fowler M. S., 2009;
Hanneman, 2000; Lima et al., 2011). Fat is stored in vesicles called lipid droplets with a
diameter of 12 nm, homogeneously distributed in the cotyledons (K. D. Chapman, Dyer, &
Mullen, 2012). The quality and the amount of each constituent in the bean depends on
agricultural factors such as cocoa variety, origin, climate conditions, soil conditions, ripening,
time of harvesting. A schematic representation of the main cocoa process practices is illustrated
in Figure 2.1.
Figure 2.1 Representation of the main cocoa process practices and the steps in cocoa butter
production.
25
Chapter 2
Once collected, the seeds undergo fermentation – a process causing many chemical changes
in both the pulp surrounding the seeds and within the beans themselves. These changes lead to
the development of chocolate flavor precursors and produce color change in the seeds. The
process of the fermentation has been studied extensively. Afoakwa et al., Lima et al., Schwan
& Wheals, and Aprotosoaie et al. cover most of the scientific work achieved on understanding
the chemical transformation occurring during the fermentation process (Afoakwa, Paterson,
Fowler, & Ryan, 2008; Aprotosoaie, Luca, & Miron, 2016; Lima et al., 2011; Schwan &
Wheals, 2004). The method for cocoa beans fermentation starts with piling a certain amount
of beans or transferring it to heaps, boxes, or baskets allowing naturally present
microorganisms to reproduce in the pulp (Thompson, Miller, & Lopez, 2007). The process of
fermentation takes between 2 to 3 days for the beans of Criollo species and from 5 to 7 days
for the beans of Forastero species. Moreover, other factors as such as the country of origin and
the farmers’ practice influence the duration of the fermentation process (Afoakwa et al., 2008;
Fowler M. S., 2009; Lima et al., 2011). In the first stage (24-36 h), part of the pulp drains off
and the anaerobic conditions favors yeasts to convert the sugars in the remaining pulp to
ethanol. Then, bacteria start to oxidize the ethanol to acetic acid and further on to carbon
dioxide and water, producing heat and therefore raising the temperature to 40-55 °C. In the
second stage of the fermentation (48-96 h), namely the aerobic phase, the beans are mixed and
the bacterial activity continues as the rest of the pulp drains away allowing air to diffuse
between the beans. At this stage, ethanol and acetic acid inhibit germination causing bean death
and enzymes and substrates lose compartmentalization moving and being active at the level of
the cotyledons. As the amount of bacteria further increases, lactic acid bacteria become
dominant and they convert sugar and certain organic acids into lactic acid. The formation of
too much lactic acid is not desirable since it enters the cotyledons increasing the acidity of the
beans, but it is important in later reactions. After approximatively 72 h, lactic acid is produced
and the acetic acid bacteria more actively oxidize ethanol to acetic acid, which penetrates into
the beans leading pH drop to 4.5 and causing biochemical transformations. Good fermentation
process is characterized by the dominancy of the acetic acid bacteria (Afoakwa et al., 2008;
Fowler M. S., 2009; Schwan & Wheals, 2004; Thompson et al., 2007). Due to membrane break
down, various beans’ constituents interact with each other(de Brito Edy S et al., 2000). As
results, during the fermentation, precursors of cocoa taste and flavor are produced from the
protein degradation into peptide and free aminoacids and reducing sugars from the enzymatic
degradation of sucrose (Afoakwa, Quao, Takrama, Budu, & Saalia, 2013; Kumari et al., 2016).
26
Chapter 2
Another reaction of particular importance is the breakdown and oxidation of the polyphenols,
which results in reduced bitterness and astringency of the fermented beans. The colored
polyphenols, the anthocyanins, are hydrolysed in cyanidins and sugars, causing the lightning
of the purple color of Forastero cocoa beans. Other oxidative and browning reactions take place
leading to darkening of the beans tissue (D'Souza et al., 2017; Kadow, Bohlmann, Phillips, &
Lieberei, 2013; Lima et al., 2011).
After the fermentation, the seeds are dried to reduce their still high moisture content. The
drying procedure is conducted either by sun drying, when the weather is sunny and dry, or it
can be done artificially using mechanical dryers. It takes between 7 to 15 days to dry the beans
to the desired <8% moisture content, necessary to prevent mold infestation during the storage
and transport (Wood, 2008). Drying rate must not be too high as it can result in acidic beans as
the volatile organic acids tend to be trapped by the hard shells and this is detrimental for the
flavor. Moreover, artificial drying might make use of a source of fire, thus producing an
unpleasant smoky taste which cannot be removed by further processing. On the other hand
slow or insufficient drying would results in beans having low acidity, poorer color, major
presence of molds, and high moisture contents in the bean (Kongor et al., 2016). This can lead
to degradation of the triacylglycerols, causing high free fatty acid and diacylglycerol contents
degrading both flavor and physical characteristics of cocoa butter (Guehi et al., 2008; Hancock
& Fowler, 1994). Chemical changes continue taking place in the beans during the drying
process involving polyphenol oxidation, inducing the formation the brown color (Biehl &
Ziegleder, 2003). This results in bitterness, astringency, and sourness reduction (Giacometti,
Jolić, & Josić, 2015; Kyi Tin Mar et al., 2005).
After the drying process is complete, the beans are shipped to the factories and industrial
processing of cocoa begins with cleaning the beans from foreign objects. Thermal treatment is
applied to facilitate the shell removal and the beans are broken into nibs (Kleinert-Zollinger,
1994). Then the nibs proceed to the roasting. The roasting process is accomplished at
temperatures between 120 and 160 °C during a period of 5 to 120 min (Afoakwa et al., 2008;
Aprotosoaie et al., 2016). Most of the roasting recipes use 100-130 °C for 50 to 70 min.
Roasting is a crucial step of the bean processing for the cocoa aroma enhancement and for
further development of flavor compounds from the precursors formed during the fermentation
and drying, as well as for sterilization. During roasting the main chemical reactions are
associated with non-enzymatic Maillard reactions producing carbonyl derivatives, which react
with free aminoacids during Strecker degradation (Afoakwa et al., 2008). Compounds such as
27
Chapter 2
pyrazines, esters, aldehydes, ketones, furans, imines, ethers, and alcohols are formed during
the roasting process (Counet, Ouwerx, Rosoux, & Collin, 2004; Frauendorfer & Schieberle,
2008; Kongor et al., 2016). Subsequently, the nibs are ground into a liquid mass called cocoa
mass or cocoa liquor. After or during roasting, an alkalizing process, consisting of treating the
nibs with an alkaline solution, using K2CO3 or (NH4)2CO3, can take place to alter flavor and
color, and to improve the dispensability of cocoa powder. This determines the color and taste
of the cocoa liquor, which, as an intermediate or semi-finished product, is supplied to the
chocolate industry or is further processed into cocoa powder and cocoa butter (Beckett, 2008;
Giacometti et al., 2015; Meursing, 1994).
The Dutchman Casparus J.Van Houten extracted cocoa butter for the first time in 1828, who
patented the first cocoa press. Nowadays, the best quality cocoa butter is extracted from the
cocoa liquor made from good quality beans through industrial hydraulic presses and gains the
name of pure pressed cocoa butter. These machines operate at pressures of 100 MPa and
temperatures between 90 and 110 °C. The butter is afterwards filtered, to remove the particles
of cocoa solids. Continuous expeller presses obtain cocoa butter of inferior quality from the
undeshelled beans, nibs, mass, and pressed cake. Expeller cake and other cocoa waste materials
are then defatted by solvent extraction. The use of organic solvents poses a potential health
threat, thus supercritical fluid extraction using carbon dioxide and ethane have been proposed
(Asep et al., 2008; Mohamed, Saldaña, Mazzafera, Zetzl, & Brunner, 2002; Rossi, Arnoldi,
Salvioni, & Schiraldi, 1989; Saldaña, Mohamed, & Mazzafera, 2002). Press, expeller and
solvent extracted cocoa butter undergo refining by treating the fat with an alkali solution and
silica adsorbents (Kamphuis, 2009). Silica pretreatment has been shown to decrease
phosphorus and iron content by 85%, theobromine by 50%, and removing small amount of
tocopherols (De Clercq et al., 2012). Due to the industry demand for different quality CB, other
refining processing as filtering, steam refining or deodorization, bleaching, and degumming
are the most applied. Deodorization is the most often used refining process after the pressing.
The role of the deodorization is to minimize the variability of the flavor components, since the
chocolate industry’s demand is usually for cocoa butter with a neutral taste. This process
consists in removing part of flavor volatiles, free fatty acids, phospholipids, diacylglycerols
and tocopherols means of high temperatures, high vacuum and steam injection. Thus, obtaining
a cocoa butter with a “smooth” flavor (De Clercq et al., 2012; Hanneman, 2000). Another
specific refining step of CB is the removal of the alkaloids: theobromine and caffeine mostly
for technological reasons. Bleaching is used for white chocolate production. Cocoa butter
28
Chapter 2
obtained from alkalized cocoa liquor can have a higher soap content, which is removed by
washing (Beckett, Fowler, & Ziegler, 2017).
2.3 Quality criteria of cocoa butter
European low clearly defines the quality criteria of the different types of butter. Pure press
cocoa butter must not contain more than 1.75% free fatty acids and not more than 0.35%
unsaponifiable matter (Table 2.1) measured by standardized AOAC official methods of
analysis (“AOAC Official Methods of Analysis,”; “Directive 2000/36/EC of the European
Parliament and of the Council relating to cocoa and chocolate products intended for human
consumption.,”). In expeller and refined cocoa butter, the value for free fatty acids is the same
as for the pure pressed cocoa butter but the value of unsaponifiable matter is increased to
0.50%, due to the higher amount of the expeller butter obtained from the shell (Rossi et al.,
1989).
Table 2.1 Definitions and characteristics of cocoa butter products according to EU Directive 2000/36/EC
and Codex Standards.
Pressed cocoa butter
Free fatty acids (expressed as oleic acid) ≤ 1.75%
Unsaponifiable matter (determined using petroleum ether) ≤ 0.35%
Expeller and refined cocoa butter
Free fatty acids (expressed as oleic acid) ≤ 1.75%
Unsaponifiable matter (determined using petroleum ether) ≤ 0.5%
The process may affect drastically the quality of cocoa butter. Thus, high quality beans are
needed to produce the best quality cocoa butter.
Pressed cocoa butter is most commonly used in the chocolate industry, which is only filtered
and often deodorized. The shelf life in solid form is up to one year when stored under suitable
conditions. For refined cocoa butter, the shelf life is reduced dramatically to a maximum of 6
months. The shelf life of liquid cocoa butter is up to one month. Thus, for longer periods
storage, the use of nitrogen to prevent oxidation by air is recommended (Beckett, 2011). The
major factor contributing to the lipid oxidation are: light, temperature, the presence of oxygen,
and metal ions. Low amount of tocopherols were shown to inhibit considerably the oxidation
of the refined CB (Hashim, Hudiyono, & Chaveron, 1997).
29
Chapter 2
2.4 Chemical composition background and significance
2.4.1 Cocoa fat content
Similar to all vegetable oils and fats, CB is mainly composed of a mixture of TAGs with a
small quantity of other minor components. The quality of CB and its properties such as melting
point and hardness depend on the variety of cocoa, age of the plant, geographical origin,
environmental growing circumstances, and nonetheless the processing conditions.
Fat content or yield is an important quality parameter for cocoa grinders during purchasing
of dried fermented cocoa beans. Usually, the fat content of cocoa beans is measured by
extraction using petroleum ether or hexane solvent with subsequent gravimetric assessment,
AOAC methods (“AOAC Official Methods of Analysis,”). Fat accumulates in the cocoa seeds
until the ripened pods are harvested, when in fact, values near to that of matured seed are
reached after 150 days post-pollination (Lehrian & Keeney, 1980). Brazilian cocoa beans had
been considered to have lower fat content in comparison to West African or Malaysian ones
(Wood & Lass, 2001). Thus, seeds from Amazonian hybrids present a range between 58 to 60%
fat in fermented beans and Criollo is reported to contain 53%. In West Africa, fermented and
dried cocoa beans have shown a fat content range between 56 and 58% and most Forastero
hybrids revealed a range of 55 and 59% (Afoakwa et al., 2013; Rohan, 1963). Luis et al.
reported that fat content from beans from genotypes collected in the Upper Amazonian region
showed higher fat levels, while the Trinitario-Criollo and Bahian genotypes tended to have
lower fat content (Pires, Cascardo, Lambert, & Figueira, 1998). Moreover, authors report that
there is a significant negative correlation between dry seed yield per plant and the fat content.
Despite the trend, authors have identified 40 genotypes showing high-fat content and high dry
seed yield. Variations in the bean sizes could also explain the difference in the lipid amount. A
positive correlation between seed weight and fat content was observed in cocoa beans from
West Africa (Toxopeus & Wessel, 1970). Authors described the relationship between the fat
content and the environmental conditions; thus, smaller seeds originated from pods developed
during the dry season have shown low-fat content. Afoakwa et al. reported that the lipid content
decreases with the increasing pod storage time. A fat reduction by 3% after 21 days of pod
storage in unfermented and fermented beans was observed (Afoakwa et al., 2013).
30
Chapter 2
2.4.2 Fatty acids composition
In terms of its fatty acid composition, authors have labelled cocoa butter as a relatively
simple fat because palmitic, stearic, and oleic acid account for over 95% of cocoa fatty acids.
Table 2.2 illustrates the fatty acid composition of cocoa butter according to several authors.
Thus, cocoa butter accounts for about 25% palmitic acid, 34% oleic acid, 35% stearic acid, as
main fatty acids and other minor fatty acids where linoleic acid and arachidic acid are in a
proportion of 3 and 1% respectively. Figure 2.2 illustrates the chemical structure of these fatty
acids. Only Marangoni et al. reported values above the average for linoleic and arachidic acid
in all the origins, despite the total amount of around 98% for the five most abundant fatty acids
(Marangoni & McGauley, 2003). Fatty acid chemical analysis is performed regularly by
methylation of the fatty acids after alkaline hydrolysis of the triacylglycerols and other lipid
components containing fatty acids according to AOAC methods, followed by GC-FID
separation and detection (Foubert, Vanrolleghem, Thas, & Dewettinck, 2004; Torres-Moreno,
Torrescasana, Salas-Salvadó, & Blanch, 2015).
Figure 2.2 Chemical structure of the most predominant fatty acids in cocoa.
The content of fatty acids, which are the building blocks of energy storage molecules, rise
rapidly during pod maturation. Studies report that stearic acid showed an increase during the
last 75-80 days of pod maturation, 100-180 days post-pollination, whereas palmitic acid and
linoleic acid decreased. Oleic acid has shown a notable enhancement after the ovules
solidification, 110-130 days post pollination; then, it showed a decrease by 3 to 4% in the
following 10 days, but remaining subsequently relatively constant until pod harvesting
31
Chapter 2
(Lehrian & Keeney, 1980). Thus, pod ripeness degree applies to a role in the final fatty acids
composition.
Authors suggest that Brazilian cocoa contains a higher amount of unsaturated fatty acids,
compared to other countries of origin, with the most representative being oleic and linoleic
acid ranging from 36.4-37.8% and 3.5-4.1% respectively (Foubert et al., 2004; Klagge & K.
Sen Gupta, 1990; Spangenberg & Dionisi, 2001; Torres-Moreno et al., 2015). In Brazil, cocoa
trees are cultivated between 5 °N and 24 °S latitude, but most of the commercial Brazilian
cocoa beans are coming from Bahia’s region. Vieira et al. and Esteves et al. and have shown
high variability among Brazilian cocoa beans (Esteves, Barreraarellano, Nunes, Galvão, &
Antoniassi, 1994; Vieira, Efraim, Walle, Clercq, & Dewettinck, 2015). Only the fatty acids
values of beans from Bahia region were in the range described by most the other authors. Vieira
et al. and Esteves et al. report values for cultivars grown in warmers regions of Brazil near to
those found in Malaysian cocoa butters (A.h.g & Zainuddin, 1986), thus showing the prominent
influence of the environmental temperature, observed by other authors (Lehrian, Keeney, &
Butler, 1980). The same variability was observed in Criollo cultivars from Venezuela (Liendo
et al., 1997). Notwithstanding, other authors found no difference among the beans from 3
different regions of Brazil (A. P. B. Ribeiro et al., 2012) although they ascertained significant
variations among the triacylglycerols. In consequence, data suggest that geographical origin
and climate are the primary factors influencing the fatty acid content in cocoa beans.
Herein, a graphical representation of normalized values (to 100) for the 5 major fatty acids
is presented for better visualization of the differences across the origins (Fig. 2.3). Only data
from dried cocoa beans and cocoa butter samples are available in literature. What emerges from
the graph, and so from the table, is the very distinctive fatty acid profile of the South and
Central American sample, where the ratio between oleic/stearic acid contents is close to 1 and
in some Brazilian and Peruvian cocoa beans, it is even higher. This possibly indicates the
hybrid effect on the fatty acid profile.
From the graph and the values collected in the table, we could observe a good negative
correlation between stearic and oleic acid and the same for stearic and linoleic acid. By
contrast, oleic and linoleic acid show a very good positive correlation. This indicates that these
fatty acids are interchangeable, however in in-depth studies are needed to elucidate this
hypothesis. A high correlation value was obtained when fewer data from the literature were
taken into account. However, it is quite difficult to compare the data collected from different
authors since the relative quantification was provided for a variable number of fatty acids as it
32
Chapter 2
33
Chapter 2
can be seen from the table. A more accurate quantification would allow a better understanding
of the correlation between these three fatty acids. In view of that, absolute quantification is
needed for equal comparison, since nowadays standards are available for all the major fatty
acids identified in cocoa butter.
As regarding the production process, no data are describing the fatty acids profile during
the fermentation process, while no significant effect was observed for the factory processing
conditions in Ecuadorian and Ghanaian fatty acids (Torres-Moreno et al., 2015). Consequently,
efforts towards determining other factors influencing the final fatty acid profile of cocoa are
necessary.
Figure 2.3 Box plots* illustrating the distribution of major fatty acids** in different origin countries.
*Normalization to 100 was performed to reduce the variability due to analytical discrepancies by different authors.
**Fatty acids data from the table were used to build the box plots.
2.4.3 Triacylglycerol composition
The distribution of the fatty acids (FAs) between the different positions on the glycerol
backbone creating triacylglycerols affects the physical and nutritional properties of CB. Like
in most vegetable oils, in cocoa butter unsaturated FAs usually occupy the sn-2 position,
whereas saturated FAs are mostly located in the sn-1 and sn-3 positions (Fig. 2.4). There are
two main classes of triglycerides in cocoa; the major compounds are of the type 1,3-
disaturated-2-unsaturated-sn-glycerol (S'U'S'), and the minor ones are of the type 1(3),2-
diunsaturated-3(1)-saturated-sn-glycerol (S'U'U' and U'U'S'). The S'U'S' group is principally
composed of 1,3-dipalmitoyl-2-oleoyl-sn-glycerol (POP, 100% regioisomeric purity), 1,3-
distearoyl-2-oleoyl-sn-glycerol (SOS, 100% regioisomeric purity), rac-palmitoyl-stearoyl-2-
34
Chapter 2
oleoyl-sn-glycerol (rac-POS, >90% regioisomeric purity) and, to a lesser extent, 1,3-

dipalmitoyl-2-linoleoyl-sn-glycerol (PLP, 100% regioisomeric purity). The other
triacylglycerol class is minor (rac-S'U'U') and it is composed of 1(3),2-dioleoyl-3(1)-stearoyl-
sn-glycerol (SOO, 90% regioisomeric purity) and 1(3),2-dioleoyl-3(1)-palmitoyl-sn-glycerol
(POO,100% regioisomeric purity). The observed fatty acid distribution confirmed that
saturated acyl groups occur almost exclusively in the 1- and 3-positions of glycerol (Fauconnot
et al., 2004). Glycerol 3-phosphate acyltransferases pathway catalyses cocoa triacylglycerol
biosynthesis in the cytoplasm of cocoa cotyledon cells (Fritz, Kauffman, Robertson, & Wilson,
1986).
Figure 2.4 Triacylglycerol structure of a 1,3-disaturated-2-unsaturated-sn-glycerol represented by

POS.
Theobroma cacao genome contains 84 orthologues genes potentially involved in lipid

biosynthesis (Argout et al., 2011). Griffiths & Harwood have previously appraised the
distribution of the saturated fatty acids in the 1- and 3-positions of glycerol (Griffiths and
Harwood, 1991). Authors have observed that diacylglycerol and phosphatidate, the two
product-steps of triacylglycerols synthesis in plants as in animals arise in the early stages of
maturation, were being relatively rich in both palmitate (33% and 36%, respectively) and oleate
(54% and 46%, respectively). Whereas the proportion of stearate was significantly lower with
values of 13% and 18%, respectively. In these lipids, the saturated fatty acids were almost
exclusively confined to position sn-1 and the oleate to the sn-2 position. By contrast, the
triacylglycerols were relatively rich in stearate (28%). Their results affirm that at position sn-
1, palmitate is predominant over stearate and oleate is largely excluded, whereas position sn-2
is acylated exclusively with oleate. By contrast, position sn-3 is almost equimolar concerning
both of the saturated fatty acids. Later, Tagaki and Ando (Takagi and Ando, 1995) confirmed
the fatty acids distribution by chiral-phase HPLC, but only in one sample of cocoa butter
obtained from unroasted beans. Initially, the three main TAGs were separated into three
35
Chapter 2
fractions using reversed-phase HPLC. Then, the authors analyzed each fraction by chiral-phase
HPLC concluding that the regioisomeric purity of the 3 main peaks, observed in reversed-
phase HPLC, was as follows: POP (100%), POS (51%), SOP (49%), SOS (96.2%), and POA
(3.8). Thus, these results and the results obtained by Fauconnot et al. that were described above,
indicate that cocoa lipid extract showed a regioisomeric mixture. However, no attempts at
studying the racemic content in different hybrids and cultivars from different origins were
reported. Moreover, it is worth noting that the correct assignment of the TAGs was recorded
only after Podlaha et al. (Podlaha, Toregard, & Puschl, 1984).
Studies on the lipid biosynthesis during maturation of cocoa pods have shown changes in
the quantitative composition of TAGs during ripening where oleic fatty acid and linoleic fatty
acid decreased as maturation proceeds and the typical high levels of stearate rise in the storage
fat (Lehrian & Keeney, 1980). Similar behaviour was observed during the ripening of Sea
Buckthorn seeds (Tsydendambaev & Vereshchagin, 2003). Moreover, a linear increase rate of
TAGs between 120 and 170 days post-pollination was observed over 2 cropping seasons,
though the fat accumulation started earlier in pods developed in the May-June period (Lehrian
& Keeney, 1980). Same authors have studied the influence of the temperature during the latter
stages of growth and ripening finding that fruits matured under elevated temperatures showed
an increase in monounsaturated TAGs (S'U'S') with a concomitant decrease of diunsaturated
TAGs (S'U'U') (Lehrian et al., 1980).
In literature, the analysis of TAGs in cocoa butter was performed by capillary GC (Lipp et
al., 2001; A. P. B. Ribeiro et al., 2012; Rossi et al., 1989; Vieira et al., 2015) or reversed-phase
HPLC coupled with evaporative light scattering detector (ELSD) (De Clercq et al., 2012b;
Foubert et al., 2004), and rarely by reversed-phase HPLC coupled with mass spectrometry
(Lísa & Holcapek, 2008; Segall, Artz, Raslan, Ferraz, & Takahashi, 2005). Table 2.3 (a) and
2.3 (b) illustrates the relative amount of the TAGs found in cocoa by several authors. It appears
that amount of the main TAGs in cocoa butter is as follows: POP is between 12.6% and 22.6%;
POS is between 34.6% and 46.3%; and SOS is between 21.7% and 31%. There was a wide
range of variation among the samples from the same country, which could be caused by
temperature differences throughout the whole year and the peculiarities of cocoa trees (Chaiseri
& Dimick, 1989). Moreover, Chaiseri et al. found that factory process affects the quantity of
the major TAGs, by decreasing significantly the amount of POS and SOS (Chaiseri & Dimick,
1989). Not much is known about the impact of the fermentation on CB chemistry. Only our
recent paper gave some insides on CB chemistry of row unfermented beans illustrating the
36
Chapter 2
presence of some TAGs ever described before (Sirbu, Corno, Ullrich, & Kuhnert, 2018). This
indicates that the chemical composition of the bean changes through the fermentation process;
nevertheless, the type of beans, the quality of the fermentation and drying and the subsequent
bean processing will all affect the composition and properties of the final product. It is
recognised that these slight differences in the chemical composition also lead to small
differences in the physical properties, such as the crystallization kinetics (Campos, Ollivon, &
Marangoni, 2010; Chaiseri & Dimick, 1989, 1995; Lehrian et al., 1980; A. P. B. Ribeiro et al.,
2012; Vieira et al., 2015). Furthermore, we were able to describe the presence of a larger
number of TAGs than previously reported, confirming and adding on top of the so far most
comprehensive list given by Lisa et al. (Lísa & Holcapek, 2008).
Concerning origins, the assumption that the farther from the Equator is the origin, the more
polyunsaturated TAGs in the composition of CB, associated with the climate conditions has
been accepted by several authors (Chaiseri & Dimick, 1989; Foubert et al., 2004; Klagge & K.
Sen Gupta, 1990; Lehrian et al., 1980; Liendo et al., 1997; Marty & Marangoni, 2009; Shukla,
1995). Box plots graphical representation of normalized values (to 100) for the 7 major TAGs
gives an overview of the triacylglycerol variation across the countries (Fig. 2.5). It can be
observed that although Dominican Republic is farther from the Equator than Bahia, it has a
higher average temperature, thus supporting the temperature influence considerations. The
contribution of Vieira et al. and Esteves et al. gave the further jutting piece showing that there
is big variability within different Brazilian cocoa samples with respect to the TAGs
composition (Esteves et al., 1994; Vieira et al., 2015).
Figure 2.5 Box plots* illustrating the distribution of major TAGs** in different origin countries.
*Normalization to 100 was performed to reduce the variability due to analytical discrepancies by different authors.
**Triacylglycerol data from the table were used to build the box plots.
37
Chapter 2
38
Chapter 2
39
Chapter 2
Only Bahian cocoa followed the global trend of categorizing Brazilian cocoa butter as a
butter rich in polyunsaturated TAGs. The average temperature in West Africa are rather
constant with a minimum temperature of 20 °C and a maximum 28 °C. Asian islands face the
maximum average temperature of slightly above 30 °C. In contrast, in South American
countries, in certain period of the year, the average temperature goes below 20 °C. Thus, Peru,
Colombia and Bolivia showed a higher amount of diunsaturated TAGs. In regards to major
TAGs, not significant differences could be found, due to high internal variability.
2.4.4 Minor components
As triacylglycerols account for 97-98% of cocoa butter, the remaining part is constituted by
minor components. These compounds include free fatty acids, mono- and diacylglycerols,
phospholipids, glycolipids and unsaponifiable matter. The role of these compounds is
distributed in part on the synthesis triacylglycerols and in part as structural elements of the cell
and compartmentalization of signalling proteins (Griffiths & Harwood, 1991; Lehrian &
Keeney, 1980). There are two potential sources of the minor components in cocoa: these
compounds may be intrinsic to cocoa beans as part of the original cocoa constituents, or they
can be produced during the production process upon degradation of major cocoa butter
constituents. Lehrian et al. observed that the amount of monoacylglycerols, sterols, free fatty
acids, and glycolipids increase during the growth and ripening of cocoa pods. Moreover,
phospholipids and diacylglycerols were present in a higher amount than those above
mentioned, though showing a significant decrease during the last 20 days of pod development.
Thus, a value of a little more than 0.4% of free fatty acids is the intrinsic values of the beans
(Lehrian & Keeney, 1980). Chaiseri et al. reported values of 0.79, 1.25, 1.27, 1.44% for Central
American, South American, African, and Asian cocoa butter respectively (Chaiseri & Dimick,
1989). Foubert et al. have shown a range between 1.16 and 2.77% in 20 sample of different
origin (Foubert et al., 2004). Thus, 8 of these butters presented values above the EU legal
requirements. Authors did not observe any correlation between fatty acids and the origin;
therefore, high amounts of fatty acids were attributed to the inadequate post harvesting
practices such as prolonged storage of fresh beans, extended fermentation or insufficient drying
(Chaiseri & Dimick, 1989; Pontillon, 1998).
40
Chapter 2
Diacylglycerols, as the free fatty acids, have shown no correlation between the amount and
the origin. Studies showed values between 0.59 and 2.8% (Foubert et al., 2004; Pontillon, 1998;
Shukla, 1995). Whereas monoacylglycerols were present only in traces (Pontillon, 1998).
Regarding phospholipids the amount varies and depends on the sources cited: 0.28 and
0.45% in raw cocoa beans (Parsons, Keeney, & Patton, 1969); 0.37% (T. R. Davis & Dimick,
1989), 0.34% (Arruda & Dimick, 1991), 0.1-0.2% (Chaiseri & Dimick, 1995), 0.8-0.9%
(Savage & Dimick, 1995), 0.72-0.94% (Shukla, 1995), and 0.1-0.2% (Foubert et al., 2004) in
cocoa butter. Reported values for glycolipids were of 0.3% (Pontillon, 1998) and 0.3-0.8%
(Chaiseri & Dimick, 1995). Figure 2.6 illustrates a TLC plate illustrating the separation of
phospholipids and glycolipids from seed crystals of cocoa butter (adapted from Arruda &
Dimick, 1991).
Figure 2.6 Thin-layer chromatographic separation of phospholipids from seed crystals (SCPL 3),
phospholipid standard mix (STDPL mix) and Ivory Coast cocoa butter sample (CBPL2). SL–simple
lipids, DPG–diphosphatidylglycerol, GL–glycolipid, PG–phosphatidylglycerol,
PA–phosphatidylethanolamine, PC–phosphatidylcholine, PI+LPE–phosphatidylinositol and
lysophosphatidylethanolamine, PS–phosphatidylserine, PA–phosphatidic acid (adapted from Arruda
& Dimick, 1991).
The amount of unsaponifiable matter ranges from 0.3 to 1.5% (Itoh, Tamura, & Matsumoto,
1973; Schlichter-Aronhime & Garti, 1988). Sterol lipids in cocoa butter are present as mainly
free sterols, with the major being sitosterol (123.3 mg/100 g fat) and stigmasterol (60.1 mg/100
41
Chapter 2
g fat). Besides, esterified sterols amount were found to be 11.5% of the total sterols and
glucosidic sterols 16% (Lipp & Anklam, 1998; Staphylakis & Gegiou, 1985). Other authors
had reported similar sterol composition, 56.2 and 29.3% for sitosterol and stigmasterol
respectively (Homberg & Bielefeld, 1982). El-Saied et al. reported values of 51.48% sitosterol,
39.45% stigmasterol, 5.77% campersterol and 3.29% cholesterol (El-Saied, Morsi, & Amer,
1981). However, no recent studies are available regarding the detailed composition of minor
components in cocoa from beans of different geographical location and coming from various
production stages.
Row cocoa beans contain between 200 and 250 mg/kg fat of tocopherols, and there are four
types of tocopherols in cocoa: α-tocopherol, β-tocopherol, γ-tocopherol and δ-tocopherol, with
γ-tocopherol and αα-tocopherol constituting the majority: approximately 85% and 10%
respectively (De Clercq, 2011). The amount of tocopherols drops dramatically during the
roasting, the parameters of which greatly influence the tocopherol content in cocoa nibs
(Żyżelewicz, Krysiak, Budryn, Oracz, & Nebesny, 2014).
Expeller cocoa butter is obtained mostly from the cocoa bean shells. The lipid content of
the shells was found to vary between 2 and 20% (El-Saied et al., 1981; Okiyama, Navarro, &
Rodrigues, 2017; Rossi et al., 1989). Shell lipid extract showed a higher amount of linoleic
acid and palmitic acid (El-Saied et al., 1981; Ruiz, Antequera, Andres, Petron, & Muriel, 2004),
and different sterol composition as compared with cocoa butter (El-Saied et al., 1981).
2.5 Correlation of chemical composition with the physical

properties of cocoa butter
The degree of unsaturation of the length of the hydrocarbon chain determine the physical
properties of a fatty acid. In the solid state, fatty acids exist in a very regularly ordered
structures called crystals. There are only a few basic crystalline structures that are stable under
a given set of thermodynamic conditions. Some molecules have the ability to group together
in multiple ways giving rise to several distinct crystal structures. This phenomenon is called
polymorphism. The word polymorphism derives from Greek words poly and morphos meaning
multiple forms. The polymorphism of CB triglycerides is the most important advancement in
the understanding of the intricate structure of fat crystal networks, and it has been known since
1820 (Narine & Marangoni, 1999). Fat crystallization affects several properties of cocoa butter
42
Chapter 2
such as the consistency, the mouthfeel, flavor release, and the emulsion stability (Hanneman,
2000; Sato, 2001). Thus, the physical properties of CB affect the status of the final product,
chocolate. Consequently, controlling the crystallization behavior of this fat is of crucial
importance (Chaiseri & Dimick, 1995; Ziegleder G., 2006). This topic was intensively studied,
and it will not be addressed with much of details here, but it will give necessary information to
understand how the chemistry of cocoa butter influences its physical properties in the solid
state. For detailed information, we refer readers to (De Clercq, 2011; Foubert, 2003; Himawan,
Starov, & Stapley, 2006; Langevelde, 2000; Sato, 2001; Stewart, 2017).
2.5.1 Cocoa butter polymorphism
TAGs are “three-legged” molecules that can pack together in two different structures, a
double-chain packing and a triple chain packing. A double-chain structure is usually formed
when the chemical properties of the three fatty acids moieties are the same or very similar. In
contrast, when the chemical properties of one or two of the three fatty acid chain moieties are
different from the other, a triple chain length structure is formed because of chain sorting and
steric hindrance (Sato, 2001) (Fig. 2.7 (a)). The angle of tilt at which the double or the triple
chain pack together determines the stability of the polymorph, and it defines the crystal plan.
The differences between polymorphs are the most apparent from a top view of the planes,
which showed the subcell structure identified by XRD patterns, giving the information
regarding the spacing between the crystal plans and the inter-chain distance. Thus, three
polymorphic forms can be defined: α, β', and β, in this order of increasing stability and melting
point (Fig. 2.7 (b)). In α polymorph, the chains are arranged in hexagonal structure (H) with
no angle. β' polymorph is characterized by the orthorhombic and perpendicular chain packing
(O⊥) with an angle of tilt between 50° and 70°. Whereas in β polymorph, the chain pack in a
triclinic chain packing (T//), in which adjacent chains are in step (parallel), and thus pack tightly
together having an angle of tilt between 50° and 70° (Himawan et al., 2006).
Due to its chemical properties, cocoa butter is an extremely polymorphous fat. The
polymorphic form is determined by the melting temperature of the individual TAGs in CB
(Sato, 2001). The number and the type of polymorphs have been subject for debate for many
years (Table 2.4), but nowadays it is accepted by many scientists that CB can crystallize into 6
different polymorphic forms. Considering a variety of nomenclatures that has been used for
the different polymorphs, herein, a mixed nomenclature recurrently used in the literature based
on 6 roman numerals and 5 Greek letters was adopted.
43
Chapter 2
Figure 2.7 (a) Chain-length packing structures and (b) the subcell structures of the 3 most
common polymorphs in TAGs (Himawan et al., 2006; Sato, 1999, 2001).
Table 2.4 Cocoa butter polymorphs and their melting point (°C) according to selected authors.
(Lovegren,
(Wille & (Chapman, Gray, & (Davis &
(Vaeck, (Duck, Lutton, Akehurst, & Feuge, Dimick, (Van Malssen,
1960) 1964) 1966) Wright, 1971) 1976) 1986) et al. 1999)
γ 17 γ 18 I 17.3 I melting VI 13 I 13.1 γ -5 to +5
α 21-24 α 23.5 II 23.3 II point not V 20 II 17.7 α 17-22
III 25.5 III provided IV 23 III 22.4 β2'
β'' 28 22-27
β' 28 β' 33 IV 27.5 IV 25.6 III 25 IV 26.4 β1'
V 33.8 V 30.8 II 30 V 30.7 β2
β 34-35 β 34-35 29-34
VI 36.3 VI 32.2 I 33.6 VI 33.8 β1
The possible phase transitions in cocoa butter, γ, α and β' polymorphic phases can crystallize
directly from the melt (Fig. 2.8 (a)). The transformation from α to β' is much quicker than if
directly from the melt. The polymorphs βV and βVI are only obtained via a phase transformation
from the β' phase (Van Malssen et al., 1999). Figure 2.8 (b) shows an example of phase
transition in a saturated-O-saturated system.
Figure 2.8 (a) Phase transition scheme of cocoa butter as proposed by Van Malssen et al., (1999)
(b) polymorphic behavior of POP and SOS as shown by Yano et al., (1993).
44
Chapter 2
2.5.2 Crystallization and melting behavior
The crystallization behavior of lipid molecules has essential repercussions in the industrial
processing of food products whose physical characteristics depend mainly on fat
crystallization, thus playing a critical role in determining the sensorial properties and stability
of many fat containing food products (Domingues et al., 2015). The primary crystallization
starts with nucleation and crystal growth after the system reached the supersaturation zone,
then it is followed by aggregation into clusters and network formation (Chaiseri & Dimick,
1995; Himawan et al., 2006; Rousset, 2002; Sato, 2001). The crystallization of fat is often
described under isothermal conditions (van Malssen et al., 1999; Marangoni & McGauley,
2003). A representative time-temperature state diagram for the polymorphism of statically
crystallized cocoa butter based on DSC and XRD data is illustrated in Figure 2.9. Different
polymorphs exhibit various morphologies: the α-form produces an amorphous mass of very
tiny crystals, the β'-form is usually a bulky shape or spherulitic, while the β-form presents a
needle shape (Sato, 2001).
Figure 2.9 Time-temperature state diagram for the polymorphism of statically crystallized cocoa
butter. Symbol ( * ) represents the polymorphic forms that have been determined by XRD (adapted
from Marangoni & McGauley, 2003).
Tempering is a precrystallization technique used to control cocoa butter crystallization. This

technique consists in subjecting the chocolate to carefully controlled shear-temperature-time
parameter combinations to induce crystallisation of CB in a thermodynamically stable βV
polymorphic form. The tempering process produces seed crystals of the form βV of CB crystal,
which enables this polymorph to be the dominant form, the most desirable as it gives to
chocolate a glossy appearance, good snap, contraction, and enhanced shelf life characteristics
45
Chapter 2
(Beckett, 2008; Ziegleder & Schwingshandl, 1998). Poor tempering and storage conditions and
other still unknown factors lead to fat bloom, a physical condition characterized by migration
of high melting TAGs which recrystallize at the surface giving to the chocolate whitish colour
(Beckett, 2008). Studies aiming at overcoming this problem showed that the low-temperature
storage retards TAG migration (Depypere, De Clercq, Segers, Lewille, & Dewettinck, 2009),
whereas diacyglycerols did not have any influence on fat blooming (De Clercq et al., 2014).
Moreover, studies on molecular composition and physical dynamics showed that small
amounts of trisaturated and triunsaturated TAGs influence the crystallization structure, times
and temperature (Campos et al., 2010). Other authors have shown the impact of other additives
on the chocolate crystallization such as sorbitan esters and lecithin (Sonwai, Podchong, &
Rousseau, 2017; Svanberg, Ahrné, Lorén, & Windhab, 2011).
The melting point of triacylglycerol species depends on the fatty acid composition and the
distribution of the fatty acids between the different stereospecific positions of the glycerol
backbone. β' polymorphs of POP melts at 33.5 °C, PPO has a melting point of 35.2 °C, whereas
POP/PPO (50/50) has a melting point of 28 °C (Sato, 2001). The melting points of many TAG
have been measured (Cebula & Smith, 1991), yet a large number of the natural molecular
species remains uncharacterized. Maruzeni et al. have developed a mathematical model for
correlating the value of the melting point of a TAG with the values of the melting points of the
three fatty acids (Maruzeni, 2009). Approximately 20% of the triacylglycerols are liquid at
room temperature (Shukla, 1995).
Determined by its typical chemical composition, cocoa butter has a unique melting
temperature range from 15 to 26 °C, depending on the polymorphic forms present. To describe
the physical properties of CB authors mostly use the solid fat content (SFC) as a function of
temperature (Fig 2.10). The solid fat content below 25 °C is an indication of the hardness of
the butter. Besides, at 25 °C, the solid content should be at least 75%, this contributes to the
acceptable chocolate snap when it is broken. The behavior between 25 and 30 °C designates
the melting resistance of the chocolate. Values lower than 48%, will suggest that the chocolate
will quickly melt in summer. The steepness of the curve (between 25 and 35 °C) shows a rapid
melting of CB, thus releasing intense flavors and a pleasant mouthfeel. Above 37 °C, the
absence of solid content means that cocoa butter is completely melted at body temperature and
there is there is no waxy sensation after eating (Hanneman, 2000; Klagge & K. Sen Gupta,
1990).
46
Chapter 2
Figure 2.10 Ideal solid fat content curve of chocolate as function of time (adapted from Klagge &
K. Sen Gupta, 1990).
2.5.3 Influence of the origin – softness assessment
The physical properties of cocoa butter vary according to the chemical composition. The
softness and the hardness of cocoa butter are tightly correlated with the triacylglycerol
composition. Thus, authors classify cocoa butters from Asian countries as hard butters, whereas
butter from South American countries, especially from Bahia (Brazil) as soft butters, African
butters being in the middle (Chaiseri & Dimick, 1989; Ana Paula Badan Ribeiro et al., 2013;
Schlichter, Sarig, & Garti, 1985; Shukla, 1995; Vieira et al., 2015). The softness of butters was
correlated to some parameters such as high iodine value and low solid fat content explained by
increased POO and SOO content, which lead to undesirable crystallization characteristics
(Chaiseri & Dimick, 1995; A. P. B. Ribeiro et al., 2012). Chaiseri & Dimick revealed that
under the same crystallization conditions cocoa butters from Malaysia and Ivory Coast showed
high nucleation rates, whereas cocoa butters from the Dominican Republic and Bahia (Brazil)
had uniquely lower nucleation rates whereas Ghanaian and Ecuadorian samples have
demonstrated average nucleation rates. Moreover, variation in the triacylglycerol composition
was best correlated with the different crystallization induction times of cocoa butter. Same
authors further showed that cocoa butters with high nucleation rates generally have high crystal
growth rates and vice versa. Only the crystallization characteristics of the Ecuadorian sample
deviate from this general rule in having a slightly high nucleation rate but a low growth rate
47
Chapter 2
(Chaiseri & Dimick, 1995). Besides, Foubert et al. have shown the influence of other minor
components on the crystallization of cocoa butter (Foubert et al., 2004).
2.6 Cocoa butter alternatives
Chocolate typically contains about 20-30% cocoa butter. As cocoa butter is an expensive
product, confectioners in Asia and Europe are looking to save money and replace about a
quarter of the cocoa butter with something substitutes designed to be used as alternatives. Such
non-cocoa fats in chocolate products are usually referred to as compounds. The United States
law requires in chocolate production that 100% cocoa butter must be used. The European
Directive 2000/36/EC allows to mix cocoa butter with not more than 5% by weight of other
vegetable fats in the end product, and it must be clearly indicated on the label. According to
this law, six vegetable fats can be used: palm oil, illipe fat, sal fat, shea butter, coconut butter,
and mango kernel. If this percentage is above 5%, the product should not be called chocolate.
Much of work has been addressed towards studying the chemical composition and physical
properties of cocoa alternatives (Jahurul et al., 2013; Lipp et al., 2001; Lipp & Anklam, 1998;
Spangenberg & Dionisi, 2001; Ulberth & Buchgraber, 2003). Besides, Buchgraber et al.
described a validated method for detection and quantification of cocoa butter equivalents in
chocolate (Buchgraber, Ulberth, & Anklam, 2004).
2.7 Conclusions
The quality of cocoa butter relies on the combination of multiple factors along the entire
production chain. Therefore, it is essential to understand the critical reactions taking place,
which can lead to significant transformation into the lipid content of cocoa. Due to the
complexity of this fat, a lot of research has been conducted, and there is still much more to be
investigated. During the years, authors have taken into consideration a higher number of TAGs.
SOA, for example, although it seems to be present in high amount in cocoa butter, very few
authors report this TAG, and no authors account it for the crystallization behavior despite
studies have shown that minor components influence the crystallization behavior. Therefore,
following questions arises: ‘what about the influence of the remaining TAGs not yet taken into
account by the majority of the authors?’; ‘how does the pull of the entire TAG changes during
the production process?’
48
Chapter 2
On the other hand, driven by flavor peculiarities, nowadays cocoa manufactures tend to
make the classification according to the origins, which is particularly appealing to the
consumers. Data showed that CB could as well be classified according to the source. However,
chocolate producers must consider the hardness and the softness of cocoa butter for the future
of such applications.
Due to rising demand for cocoa butter substituents, a consequent adulteration issue of cocoa
butter might advance. If food industry takes into account only the TAGs described in the
literature, very soon we might come across fats having the same TAG composition but of an
artificial source.
We believe that is time to appraise as much as it is possible on the chemical composition of
cocoa butter and we would like to emphasize our still limited knowledge of the variability of
lipids in cocoa. Furthermore, still many questions require determination of lipid analyte
absolute quantification.
2.8 Aim of the thesis
The rise in cocoa butter demand has determined the industry to invest more in the research
in the past decade. Due to its intricate physical properties, cocoa butter requires deeper insights
into the chemistry of this fat. The current literature does not provide a clear link between the
chemical profile and the physical properties, but rather it offers assumptions often taking into
account only a small number of lipid constituents. Nevertheless, there is very little evidence
regarding the chemical modifications that cocoa butter undergoes during the production
process. To advance the research on understanding the physical behavior of cocoa fat, several
important research gaps need to be addressed.
The work of this thesis is a part of the project between Jacobs University and Barry
Callebaut cooperation named CoMeta – cocoa metabolomics. The main objectives of this
project are to recognize and map the precursor molecules during the production process and
link them with the flavor and color of the final product. Within this framework, variations in
the lipidomics of cocoa from tree to bar have been evaluated. To achieve the general purpose
a series of sub-goals have been intended:
Appraisal of the current state of the art in lipid chemical analysis and cocoa lipidomics
research (literature review);
49
Chapter 2
Analysis of proposed methodologies for the full analysis of plant lipids with particular
attention to oil seeds (literature review);
Optimization of the extraction method to obtain a fast and reliable methodology suitable
for the analysis of a large number of samples;
Optimization of the separation of different lipid classes and characterization of each
lipid classes by chromatographic techniques, along with the development of a fast and reliable
methodology suitable for the analysis of a large number of samples;
Characterization and identification of the largest possible number of lipid molecules
present in cocoa by mass spectrometry techniques;
Qualitative and quantitative assessment of the lipid profile of cocoa concerning different
origins and qualities of cocoa beans;
Study of the influence of selected process parameters in the cocoa and chocolate
manufacturing procedure on the profile and content of the lipids present to identify critical
process conditions responsible for lipid degradation and modification.
Identification, selection, and characterization of biomarkers in each process step and
development of a suitable tool for their assessment.
While the research focuses on cocoa butter chemistry from bean to bar important for the
science of cocoa and chocolate, the overarching aim of this thesis is to derive associations and
practical industrial applications for relevant eventual findings and understandings.
References
https://doi.org/10.1080/10408390701719272
Afoakwa, E. O., Quao, J., Takrama, J., Budu, A. S., & Saalia, F. K. (2013). Chemical composition and physical
quality characteristics of Ghanaian cocoa beans as affected by pulp pre-conditioning and fermentation. Journal
of Food Science and Technology, 50(6), 1097–1105. https://doi.org/10.1007/s13197-011-0446-5
A.h.g, C., & Zainuddin, N. (1986). Characteristics of Malaysian cocoa butter. Presented at the International
Conference on Cocoa and Coconuts: Progress and Outlook. Kuala Lumpur (Malaysia). 15-17 Oct 1984. Retrieved
from http://agris.fao.org/agris-search/search.do?recordID=MY19870059372
AOAC Official Methods of Analysis. (n.d.). Retrieved April 27, 2018, from
http://www.aoac.org/aoac_prod_imis/AOAC/Publications/Official_Methods_of_Analysis/AOAC_Member/Pubs
/OMA/AOAC_Official_Methods_of_Analysis.aspx
50
Chapter 2
4337.12180
Argout, X., Salse, J., Aury, J.-M., Guiltinan, M. J., Droc, G., Gouzy, J., … Lanaud, C. (2011). The genome of
Theobroma cacao. Nature Genetics, 43(2), 101–108. https://doi.org/10.1038/ng.736
Arruda, D. H., & Dimick, P. S. (1991). Phospholipid composition of lipid seed crystal isolates from ivory coast
cocoa butter. Journal of the American Oil Chemists’ Society, 68(6), 385–390. https://doi.org/10.1007/BF02663754
As cocoa price soars, chocolate makers devour substitutes. (2014, September 5). Reuters. Retrieved from
https://www.reuters.com/article/asia-cocoa-substitute/as-cocoa-price-soars-chocolate-makers-devour-
substitutes-idUSL4N0QR2XV20140905
Asep, E. K., Jinap, S., Tan, T. J., Russly, A. R., Harcharan, S., & Nazimah, S. A. H. (2008). The effects of particle
size, fermentation and roasting of cocoa nibs on supercritical fluid extraction of cocoa butter. Journal of Food
Engineering, 85(3), 450–458. https://doi.org/10.1016/j.jfoodeng.2007.08.008
Badrie, N., Bekele, F., Sikora, E., & Sikora, M. (2015). Cocoa Agronomy, Quality, Nutritional, and Health
Aspects. Critical Reviews in Food Science and Nutrition, 55(5), 620–659.
https://doi.org/10.1080/10408398.2012.669428
Beckett, S. T. (2008). The Science of Chocolate. Royal Society of Chemistry.
Beckett, Steve T. (2011). Industrial Chocolate Manufacture and Use. John Wiley & Sons.
Beckett, Steve T., Fowler, M. S., & Ziegler, G. R. (2017). Beckett’s Industrial Chocolate Manufacture and Use.
John Wiley & Sons.
Biehl, B., & Ziegleder, G. (2003). COCOA | Chemistry of Processing. In B. Caballero (Ed.), Encyclopedia of
Food Sciences and Nutrition (Second Edition) (pp. 1436–1448). Oxford: Academic Press.
https://doi.org/10.1016/B0-12-227055-X/00261-3
Buchgraber, M., Ulberth, F., & Anklam, E. (2004). Method validation for detection and quantification of cocoa
butter equivalents in cocoa butter and plain chocolate. Journal of AOAC International, 87(5), 1164–1172.
Campos, R., Ollivon, M., & Marangoni, A. G. (2010). Molecular Composition Dynamics and Structure of Cocoa
Butter. Crystal Growth & Design, 10(1), 205–217. https://doi.org/10.1021/cg900853e
Cebula, D. J., & Smith, K. W. (1991). Differential scanning calorimetry of confectionery fats. Pure triglycerides:
Effects of cooling and heating rate variation. Journal of the American Oil Chemists Society, 68(8), 591–595.
https://doi.org/10.1007/BF02660159
51
Chapter 2
Chaiseri, S., & Dimick, P. S. (1989). Lipid and hardness characteristics of cocoa butters from different geographic
regions. Journal of the American Oil Chemists’ Society, 66(12), 1771–1776. https://doi.org/10.1007/BF02660745
Chaiseri, S., & Dimick, P. S. (1995). Dynamic crystallization of cocoa butter. II. Morphological, thermal, and
chemical characteristics during crystal growth. Journal of the American Oil Chemists’ Society, 72(12), 1497–
1504. https://doi.org/10.1007/BF02577843
Chapman, G. M., Akehurst, E. E., & Wright, W. B. (1971). Cocoa butter and confectionery fats. Studies using
programmed temperature X-ray diffraction and differential scanning calorimetry. Journal of the American Oil
Chemists Society, 48(12), 824–830. https://doi.org/10.1007/BF02609292
Chapman, K. D., Dyer, J. M., & Mullen, R. T. (2012). Biogenesis and functions of lipid droplets in plants Thematic
Review Series: Lipid Droplet Synthesis and Metabolism: from Yeast to Man. Journal of Lipid Research, 53(2),
215–226. https://doi.org/10.1194/jlr.R021436
Clercq, N. D., Depypere, F., Delbaere, C., Nopens, I., Bernaert, H., & Dewettinck, K. (2014). Influence of cocoa
butter diacylglycerols on migration induced fat bloom in filled chocolates. European Journal of Lipid Science and
Technology, n/a-n/a. https://doi.org/10.1002/ejlt.201300476
Counet, C., Ouwerx, C., Rosoux, D., & Collin, S. (2004). Relationship between Procyanidin and Flavor Contents
of Cocoa Liquors from Different Origins. Journal of Agricultural and Food Chemistry, 52(20), 6243–6249.
https://doi.org/10.1021/jf040105b
Davis, T., & Dimick, P. (1986). Solidification of cocoa butter (Vol. 40, pp. 104–108). Presented at the Proc. PMCA
Prod. Conf.
Davis, T. R., & Dimick, P. S. (1989). Lipid composition of high-melting seed crystals formed during cocoa butter
solidification. Journal of the American Oil Chemists Society, 66(10), 1494–1498.
https://doi.org/10.1007/BF02661979
de Brito Edy S, García Nelson H Pezoa, Gallão M I, Cortelazzo Angelo L, Fevereiro Pedro S, & Braga Márcia R.
(2000). Structural and chemical changes in cocoa (Theobroma cacao L) during fermentation, drying and roasting.
Journal of the Science of Food and Agriculture, 81(2), 281–288. https://doi.org/10.1002/1097-
0010(20010115)81:2<281::AID-JSFA808>3.0.CO;2-B
De Clercq, N. (2011). Changing the functionality of cocoa butter (dissertation). Ghent University. Retrieved from
http://hdl.handle.net/1854/LU-1909786
De Clercq, N., Moens, K., Depypere, F., Vila Ayala, J., Calliauw, G., De Greyt, W., & Dewettinck, K. (2012).
Influence of cocoa butter refining on the quality of milk chocolate. Journal of Food Engineering, 111(2), 412–
419. https://doi.org/10.1016/j.jfoodeng.2012.01.033
52
Chapter 2
Depypere, F., De Clercq, N., Segers, M., Lewille, B., & Dewettinck, K. (2009). Triacylglycerol migration and
bloom in filled chocolates: Effects of low‐temperature storage. European Journal of Lipid Science and
Technology, 111(3), 280–289.
Directive 2000/36/EC of the European Parliament and of the Council relating to cocoa and chocolate products
intended for human consumption. (n.d.). Retrieved April 27, 2018, from
https://www.ecolex.org/details/legislation/directive-200036ec-of-the-european-parliament-and-of-the-council-
relating-to-cocoa-and-chocolate-products-intended-for-human-consumption-lex-faoc034904/
Domingues, M. A. F., Ribeiro, A. P. B., Kieckbusch, T. G., Gioielli, L. A., Grimaldi, R., Cardoso, L. P., &
Gonçalves, L. A. G. (2015). Advances in Lipids Crystallization Technology. https://doi.org/10.5772/59767
D'Souza, R., Grimbs, S., Behrends, B., Bernaert, H., Ullrich, M., & Kuhnert, N. (2017). Origin-based
polyphenolic fingerprinting of Theobroma cacao in unfermented and fermented beans. Food Research
International. https://doi.org/10.1016/j.foodres.2017.06.007
Duck, W. (1964). The measurement of unstable fat in finished chocolate. Manufacturing Confectioner, 35(6), 67–
72.
El-Saied, H. M., Morsi, M. K., & Amer, M. M. A. (1981). Composition of cocoa shell fat as related to cocoa
butter. Zeitschrift Für Ernährungswissenschaft, 20(2), 145–151. https://doi.org/10.1007/BF02021260
Esteves, W., Barreraarellano, D., Nunes, M., Galvão, M., & Antoniassi, R. (1994). Composição de ácidos graxos
e triglicerídeos de quatro cultivares de cacau. Cienc. Tecnol. Aliment, 14(2), 247–252.
Foubert, I. (2003). Modelling isothermal cocoa butter crystallization : influence of temperature and chemical
composition /.
Foubert, I., Vanrolleghem, P., Thas, O., & Dewettinck, K. (2004). Influence of chemical composition on the
isothermal cocoa butter crystallization. JOURNAL OF FOOD SCIENCE, 69(9), E478–E487. Retrieved from
http://hdl.handle.net/1854/LU-307446
Fowler M. S. (2009). Cocoa Beans: From Tree to Factory. Industrial Chocolate Manufacture and Use.
https://doi.org/10.1002/9781444301588.ch2
Frauendorfer, F., & Schieberle, P. (2008). Changes in Key Aroma Compounds of Criollo Cocoa Beans During
Roasting. Journal of Agricultural and Food Chemistry, 56(21), 10244–10251. https://doi.org/10.1021/jf802098f
Fritz, P. J., Kauffman, J. M., Robertson, C. A., & Wilson, M. R. (1986). Cocoa butter biosynthesis. Purification
and characterization of a soluble sn-glycerol-3-phosphate acyltransferase from cocoa seeds. The Journal of
Biological Chemistry, 261(1), 194–199.
Garti, N., & Widlak, N. R. (2015). Cocoa Butter and Related Compounds. Elsevier.
53
Chapter 2
Giacometti, J., Jolić, S. M., & Josić, D. (2015). Chapter 73 - Cocoa Processing and Impact on Composition. In V.
Preedy (Ed.), Processing and Impact on Active Components in Food (pp. 605–612). San Diego: Academic Press.
https://doi.org/10.1016/B978-0-12-404699-3.00073-1
Griffiths, G., & Harwood, J. L. (1991). The regulation of triacylglycerol biosynthesis in cocoa Theobroma cacao
L. Planta, 184(2), 279–284. https://doi.org/10.1007/BF01102429
Guehi, T. S., Dingkuhn, M., Cros, E., Fourny, G., Ratomahenina, R., Moulin, G., & Clement, A. (2008). Impact
of cocoa processing technologies in free fatty acids formation in stored raw cocoa beans. African Journal of
Agricultural Research, 3(3), 174–179. Retrieved from http://publications.cirad.fr/une_notice.php?dk=543904
Hancock, B. L., & Fowler, M. S. (1994). Cocoa bean production and transport. In Industrial Chocolate
Manufacture and Use (pp. 8–24). Springer, Boston, MA. https://doi.org/10.1007/978-1-4615-2111-2_2
Hanneman, E. (2000). The Complex World of Cocoa. MANUFACTURING CONFECTIONER, 80, 107–112.
Hashim, L., Hudiyono, S., & Chaveron, H. (1997). Volatile compounds of oxidized cocoa butter. Food Research
International, 30(3), 163–169. https://doi.org/10.1016/S0963-9969(97)00039-2
Himawan, C., Starov, V. M., & Stapley, A. G. F. (2006). Thermodynamic and kinetic aspects of fat crystallization.
Advances in Colloid and Interface Science, 122(1), 3–33. https://doi.org/10.1016/j.cis.2006.06.016
Homberg, E., & Bielefeld, B. (1982). Sterine und Methylsterine in Kakaobutter und Kakaobutter-Ersatzfetten.
Dtsch. Lebensm. Rundsch, 78, 73.
Itoh, T., Tamura, T., & Matsumoto, T. (1973). Sterol composition of 19 vegetable oils. Journal of the American
Oil Chemists Society, 50(4), 122–125. https://doi.org/10.1007/BF02633564
Jahurul, M. H. A., Zaidul, I. S. M., Norulaini, N. A. N., Sahena, F., Jinap, S., Azmir, J., Omar, A. K. M. (2013).
Cocoa butter fats and possibilities of substitution in food products concerning cocoa varieties, alternative sources,
extraction methods, composition, and characteristics. Journal of Food Engineering, 117(4), 467–476.
https://doi.org/10.1016/j.jfoodeng.2012.09.024
Kadow, D., Bohlmann, J., Phillips, W., & Lieberei, R. (2013). Identification of main fine or flavour components
in two genotypes of the cocoa tree (Theobroma cacao L.). Journal of Applied Botany and Food Quality, 86(1),
90–98. https://doi.org/10.5073/JABFQ.2013.086.013
54
Chapter 2
Kamphuis, H. J. (2009). Production and Quality Standards of Cocoa Mass, Cocoa Butter and Cocoa Powder. In
Industrial Chocolate Manufacture and Use (pp. 121–141). Wiley-Blackwell.
https://doi.org/10.1002/9781444301588.ch6
Klagge, P., & K. Sen Gupta, A. (1990). Kakaobutter und ihre Alternativen fir Schokoladenprodukte (Vol. 92).
https://doi.org/10.1002/lipi.19900921208
Kleinert-Zollinger, J. (Lindt und S. (1994). (Cocoa butter provenance and attributes of process technics - Decisive
for product quality). Zucker- und Suesswarenwirtschaft (Germany). Retrieved from http://agris.fao.org/agris-
search/search.do?recordID=DE94Y0220
Kongor, J. E., Hinneh, M., de Walle, D. V., Afoakwa, E. O., Boeckx, P., & Dewettinck, K. (2016). Factors
influencing quality variation in cocoa (Theobroma cacao) bean flavour profile — A review. Food Research
International, 82, 44–52. https://doi.org/10.1016/j.foodres.2016.01.012
Kumari, N., Kofi, K. J., Grimbs, S., D’Souza, R. N., Kuhnert, N., Vrancken, G., & Ullrich, M. S. (2016).
Biochemical fate of vicilin storage protein during fermentation and drying of cocoa beans. Food Research
International (Ottawa, Ont.), 90, 53–65. https://doi.org/10.1016/j.foodres.2016.10.033
Kyi Tin Mar, Daud Wan Ramli Wan, Mohammad Abu Bakar, Wahid Samsudin Mohd., Kadhum Abdul Amir
Hassan, & Talib Meor Zainal Meor. (2005). The kinetics of polyphenol degradation during the drying of
Malaysian cocoa beans. International Journal of Food Science & Technology, 40(3), 323–331.
https://doi.org/10.1111/j.1365-2621.2005.00959.x
Langevelde, A. J. van. (2000). Triacylglycerol Structures and Cocoa-butter Crystallization. Universiteit van
Amsterdam [Host].
Lehrian, D. W., & Keeney, P. G. (1980). Changes in lipid components of seeds during growth and ripening of
cacao fruit. Journal of the American Oil Chemists’ Society, 57(2), 61–65. https://doi.org/10.1007/BF02674361
Lehrian, D. W., Keeney, P. G., & Butler, D. R. (1980). Triglyceride characteristics of cocoa butter from cacao fruit
matured in a microclimate of elevated temperature1. Journal of the American Oil Chemists’ Society, 57(2), 66–
69. https://doi.org/10.1007/BF02674362
Liendo, R., Padilla, F. C., & Quintana, A. (1997). Characterization of cocoa butter extracted from Criollo cultivars
of Theobroma cacao L. Food Research International, 30(9), 727–731. https://doi.org/10.1016/S0963-
9969(98)00025-8
55
Chapter 2
Lima, L. J. R., Almeida, M. H., Nout, M. J. R., & Zwietering, M. H. (2011). Theobroma cacao L., “The Food of
the Gods”: Quality Determinants of Commercial Cocoa Beans, with Particular Reference to the Impact of
Fermentation. Critical Reviews in Food Science and Nutrition, 51(8), 731–761.
https://doi.org/10.1080/10408391003799913
Lipp, M., & Anklam, E. (1998). Review of cocoa butter and alternative fats for use in chocolate—Part A.
Compositional data. Food Chemistry, 62(1), 73–97. https://doi.org/10.1016/S0308-8146(97)00160-X
Lipp, M., Simoneau, C., Ulberth, F., Anklam, E., Crews, C., Brereton, P., … Wiedmaier, C. (2001). Composition
of Genuine Cocoa Butter and Cocoa Butter Equivalents. Journal of Food Composition and Analysis, 14(4), 399–
408. https://doi.org/10.1006/jfca.2000.0984
Lísa, M., & Holcapek, M. (2008). Triacylglycerols profiling in plant oils important in food industry, dietetics and
cosmetics using high-performance liquid chromatography-atmospheric pressure chemical ionization mass
spectrometry. Journal of Chromatography. A, 1198–1199, 115–130.
https://doi.org/10.1016/j.chroma.2008.05.037
Lovegren, N. V., Gray, M. S., & Feuge, R. O. (1976). Effect of liquid fat on melting point and polymorphic
behavior of cocoa butter and a cocoa butter fraction. Journal of the American Oil Chemists Society, 53(3), 108–
112. https://doi.org/10.1007/BF02635960
Marangoni, A. G., & McGauley, S. E. (2003). Relationship between Crystallization Behavior and Structure in
Cocoa Butter. Crystal Growth & Design, 3(1), 95–108. https://doi.org/10.1021/cg025580l
Marty, S., & Marangoni, A. G. (2009). Effects of Cocoa Butter Origin, Tempering Procedure, and Structure on
Oil Migration Kinetics. Crystal Growth & Design, 9(10), 4415–4423. https://doi.org/10.1021/cg9004505
Maruzeni, S. (2009). A mathematical technique for estimating the melting points of triacylglycerols from the
component fatty acid melting points. European Journal of Lipid Science and Technology, 111(12), 1240–1248.
Meursing, E. H. (1994). Cocoa mass, cocoa butter, cocoa powder. In Industrial Chocolate Manufacture and Use
(pp. 70–82). Springer, Boston, MA. https://doi.org/10.1007/978-1-4615-2111-2_6
Mohamed, R. S., Saldaña, M. D. A., Mazzafera, P., Zetzl, C., & Brunner, G. (2002). Extraction of Caffeine,
Theobromine, and Cocoa Butter from Brazilian Cocoa Beans Using Supercritical CO2 and Ethane. Industrial &
Engineering Chemistry Research, 41(26), 6751–6758. https://doi.org/10.1021/ie0203936
Narine, S. S., & Marangoni, A. G. (1999). Relating structure of fat crystal networks to mechanical properties: a
review. Food Research International, 32(4), 227–248. https://doi.org/10.1016/S0963-9969(99)00078-2
Okiyama, D. C. G., Navarro, S. L. B., & Rodrigues, C. E. C. (2017). Cocoa shell and its compounds: Applications
in the food industry. Trends in Food Science & Technology, 63, 103–112.
https://doi.org/10.1016/j.tifs.2017.03.007
56
Chapter 2
Parsons, J. G., Keeney, P. G., & Patton, S. (1969). Identification and Quantitative Analysis of Phospholipids in
Cocoa Beans. Journal of Food Science, 34(6), 497–499. https://doi.org/10.1111/j.1365-2621.1969.tb12069.x
Pires, J. L., Cascardo, J. C. M., Lambert, S. V., & Figueira, A. (1998). Increasing cocoa butter yield through
genetic improvement of Theobroma cacao L.: Seed fat content variability, inheritance, and association with seed
yield. Euphytica, 103(1), 115–121. https://doi.org/10.1023/A:1018327411530
Podlaha, O., Toregard, B., & Puschl, B. (1984). TG-type composition of 28 cocoa butters and correlation between
some TG-type components. Lebensmittel - Wissenschaft + Technologie = Food Science + Technology. Retrieved
from http://agris.fao.org/agris-search/search.do?recordID=US201302561682
Pontillon, J. (1998). Cocoa and chocolate: production, utilization, characteristics. Cocoa and chocolate:
production, utilization, characteristics. Retrieved from
https://www.cabdirect.org/cabdirect/abstract/20026788577
Ribeiro, A. P. B., Silva, R. C. da, Gioielli, L. A., Gonçalves, M. I. de A., Grimaldi, R., Gonçalves, L. a. G., &
Kieckbusch, T. G. (2012). Physico-chemical properties of Brazilian cocoa butter and industrial blends. Part I
Chemical composition, solid fat content and consistency. Grasas y Aceites, 63(1), 79–88.
https://doi.org/10.3989/gya.069011
Ribeiro, Ana Paula Badan, Basso, R. C., dos Santos, A. O., Andrade, G. C., Cardoso, L. P., & Kieckbusch, T. G.
(2013). Hardfats as crystallization modifiers of cocoa butter. European Journal of Lipid Science and Technology,
115(12), 1462–1473. https://doi.org/10.1002/ejlt.201300052
Rohan, T. A. (1963). Processing of raw cocoa for the market. Rome : FAO. Retrieved from
https://trove.nla.gov.au/version/22982824
Rossi, M., Arnoldi, C., Salvioni, G., & Schiraldi, A. (1989). Characterization of cocoa extracts obtained with
supercritical carbon dioxide. ITALIAN JOURNAL OF FOOD SCIENCE, 1(3), 41–50. Retrieved from
https://air.unimi.it/handle/2434/176331#.WsvEki5uapo
Rousset, P. (2002). Physical properties of lipids. Physical Properties of Lipids. Marcell Dekker Inc., New York, 1–
36.
Ruiz, J., Antequera, T., Andres, A. I., Petron, M. J., & Muriel, E. (2004). Improvement of a solid phase extraction
method for analysis of lipid fractions in muscle foods. Analytica Chimica Acta, 520(1–2), 201–205.
https://doi.org/10.1016/j.aca.2004.04.059
Saldaña, M. D. A., Mohamed, R. S., & Mazzafera, P. (2002). Extraction of cocoa butter from Brazilian cocoa
beans using supercritical CO2 and ethane. Fluid Phase Equilibria, 194–197, 885–894.
https://doi.org/10.1016/S0378-3812(01)00719-1
57
Chapter 2
Sato, K. (1999). Solidification and phase transformation behaviour of food fats—a review. Lipid/Fett, 101(12),
467–474.
Sato, K. (2001). Crystallization behaviour of fats and lipids — a review. Chemical Engineering Science, 56(7),
2255–2265. https://doi.org/10.1016/S0009-2509(00)00458-9
Savage, C. M., & Dimick, P. S. (1995). Influence of phospholipids during crystallization of hard and soft cocoa
butters. The Manufacturing Confectioner, 75(5), 127–132.
Schlichter, J., Sarig, S., & Garti, N. (1985). Polymorphic transformations of cocoa butter in the presence of
emulsifier, studied by the DSC. Thermochimica Acta, 85, 517–520. https://doi.org/10.1016/0040-6031(85)85634-
3
Schlichter-Aronhime, J., & Garti, N. (1988). Solidification and polymorphism in cocoa butter and the blooming
problems. Crystallization and Polymorphism of Fats and Fatty Acids, 31, 363–393.
https://doi.org/10.1080/10408690490464104
Shahkhalili, Y., Duruz, E., & Acheson, K. (2000). Digestibility of cocoa butter from chocolate in humans: a
comparison with corn-oil. European Journal of Clinical Nutrition, 54(2), 120–125.
https://doi.org/10.1038/sj.ejcn.1600905
Shukla, V. K. S. (International F. S. C. A. (1995). Cocoa butter properties and quality. Lipid Technology (United
Kingdom). Retrieved from http://agris.fao.org/agris-search/search.do?recordID=GB9609755
Sirbu, D., Corno, M., Ullrich, M. S., & Kuhnert, N. (2018). Characterization of triacylglycerols in unfermented
cocoa beans by HPLC-ESI mass spectrometry. Food Chemistry, 254, 232–240.
https://doi.org/10.1016/j.foodchem.2018.01.194
Sonwai, S., Podchong, P., & Rousseau, D. (2017). Crystallization kinetics of cocoa butter in the presence of
sorbitan esters. Food Chemistry, 214, 497–506. https://doi.org/10.1016/j.foodchem.2016.07.092
Spangenberg, J. E., & Dionisi, F. (2001). Characterization of Cocoa Butter and Cocoa Butter Equivalents by Bulk
and Molecular Carbon Isotope Analyses:  Implications for Vegetable Fat Quantification in Chocolate. Journal of
Agricultural and Food Chemistry, 49(9), 4271–4277. https://doi.org/10.1021/jf001509g
Staphylakis, K., & Gegiou, D. (1985). Free, esterified and glucosidic sterols in cocoa butter. Lipids, 20(11), 723–
728. https://doi.org/10.1007/BF02534394
Stewart, D. I. (2017). The effect of low melting oils on the crystallisation of confectionery fats (Thesis). © David
Stewart. Retrieved from https://dspace.lboro.ac.uk/dspace-jspui/handle/2134/24420
58
Chapter 2
Svanberg, L., Ahrné, L., Lorén, N., & Windhab, E. (2011). Effect of sugar, cocoa particles and lecithin on cocoa
butter crystallisation in seeded and non-seeded chocolate model systems. Journal of Food Engineering, 104(1),
70–80. https://doi.org/10.1016/j.jfoodeng.2010.09.023
The International Cocoa Organization (ICCO) | Cocoa Producing and Cocoa Consuming Countries. (n.d.).
Retrieved March 13, 2018, from https://www.icco.org/
Thompson, S. S., Miller, K. B., & Lopez, A. S. (2007). Cocoa and Coffee. Food Microbiology: Fundamentals and
Frontiers, Third Edition, 837–850. https://doi.org/10.1128/9781555815912.ch39
Torres-Moreno, M., Torrescasana, E., Salas-Salvadó, J., & Blanch, C. (2015). Nutritional composition and fatty
acids profile in cocoa beans and chocolates with different geographical origin and processing conditions. Food
Chemistry, 166(Supplement C), 125–132. https://doi.org/10.1016/j.foodchem.2014.05.141
Toxopeus, H., & Wessel, M. (1970). Studies on pod and bean values of Theobroma cacao L. in Nigeria. I.
Environmental effects on West African Amelonado with particular attention to annual rainfall distribution. Neth J
Agr Sci. Retrieved from http://agris.fao.org/agris-search/search.do?recordID=US201301201159
Tsydendambaev, V. D., & Vereshchagin, A. G. (2003). Changes in Triacylglycerol Composition during Ripening
of Sea Buckthorn (Hippophaë rhamnoides L.) Seeds. Journal of Agricultural and Food Chemistry, 51(5), 1278–
1283. https://doi.org/10.1021/jf020593s
Ulberth, F., & Buchgraber, M. (2003). Analytical platforms to assess the authenticity of cocoa butter. European
Journal of Lipid Science and Technology, 105(1), 32–42. https://doi.org/10.1002/ejlt.200390003
Vaeck, S. (1960). Cocoa butter and fat bloom. Manuf Confect, 35–75.
Van Malssen, K. van, Langevelde, A. van, Peschar, R., & Schenk, H. (1999). Phase behavior and extended phase
scheme of static cocoa butter investigated with real-time X-ray powder diffraction. Journal of the American Oil
Chemists’ Society, 76(6), 669–676. https://doi.org/10.1007/s11746-999-0158-4
Vieira, L. R., Efraim, P., Walle, D. V. de, Clercq, N. D., & Dewettinck, K. (2015). Influence of Brazilian
Geographic Region and Organic Agriculture on the Composition and Crystallization Properties of Cocoa Butter.
Journal of the American Oil Chemists’ Society, 92(11–12), 1579–1592. https://doi.org/10.1007/s11746-015-2728-
y
Wille, R., & Lutton, E. (1966). Polymorphism of cocoa butter. Journal of the American Oil Chemists Society,
43(8), 491–496.
Wood, G. a. R. (2008). From Harvest to Store. In Cocoa (pp. 444–504). Wiley-Blackwell.

https://doi.org/10.1002/9780470698983.ch13
59
Chapter 2
Wood, G. A. R., & Lass, R. A. (2001). Cocoa (4th ed). Oxford : Blackwell Science. Retrieved from
https://trove.nla.gov.au/version/46558938
Yano, J., Ueno, S., Sato, K., Arishima, T., Sagi, N., Kaneko, F., & Kobayashi, M. (1993). FT-IR study of
polymorphic transformations in SOS, POP, and POS. The Journal of Physical Chemistry, 97(49), 12967–12973.
Ziegleder G. (2006). DSC‐Thermoanalyse und Kinetik der Kristallisation von Kakaobutter. Lipid / Fett, 92(12),
481–485. https://doi.org/10.1002/lipi.19900921207
Ziegleder, G., & Schwingshandl, I. (1998). Kinetics of fat migration within chocolate products. Part III: fat bloom.
Lipid/Fett, 100(9), 411–415.
Żyżelewicz, D., Krysiak, W., Budryn, G., Oracz, J., & Nebesny, E. (2014). Tocopherols in cocoa butter obtained
from cocoa bean roasted in different forms and under various process parameters. Food Research International,
63(Part C), 390–399. https://doi.org/10.1016/j.foodres.2014.03.027
60
Chapter 3
Chapter 3:
3 Exploring cocoa lipid classes
Abstract
Cocoa beans store their energy in the form of lipids, which constitute up to 55% of the seed.
It is known that neutral lipids account for more than 95% of the cocoa lipid extract and it mainly
consists of triacylglycerols (TAGs). In order evaluate the remaining part, which includes more
polar lipids (as glycolipids and phospholipids) and other components like sterols, tri-terpenes,
and tocopherols, it is essential to develop efficient extraction and separation methods. Method
development includes all those steps involved in the selection of an adequate solvent system
for chromatographic separation using strength and selectivity as key features. Herein, a simple
procedure for the extraction of lipids from seed material has been developed. Various extraction
conditions, including the solvent, time of extraction, different type of seeds, have been
investigated to obtain the maximum yield of the extract.
Moreover, this chapter examines the overall chemical composition of the cocoa lipids by
fractionating the lipid extract. Five fractions of cocoa lipids were obtained by employing solid
phase extraction (SPE) cartridges, and all the fractions were analysed by thin layer
chromatography (TLC) and by mass spectrometry with electrospray ionization (ESI) on direct
injection analysis and LC-MS. Main TAGs and diacylglycerol (DAG) profiles have not showed
substantial differences during the production process. However, significant differences were
observed in the polar fractions of cocoa lipids, where a variety of more polar lipid compounds
were transformed or degraded during the fermentation process.
Keywords: extraction; dichlomethane (DCM); Soxhlet; cocoa; seeds; solid phase extraction (SPE);
thin layer chromatography (TLC); fraction.
Supplementary material is available on request. E-mail: [email protected]
61
Chapter 3
3.1 Introduction
Comprehensive lipidomics requires a complete extraction of the lipid from any type of
sample. There are many factors responsible for the solubility of lipids in a particular solvent,
and it occurs very frequently that a particular lipid from a specified source is soluble in one
solvent but poorly soluble in another. Suitable extraction solvents have to have the capacity to
overcome the strong forces of association between the lipids and cellular constituents, such as
proteins and polysaccharides. To release all lipids from their association with cell constituents,
the solvent (or solvent mixture) must have a suitable polarity. However, the given polarity must
not permit any chemical reaction with the lipids nor that triacylglycerols and other non-polar
simple lipids do not dissolve and are left adhering to the tissues. Besides, the optimal extracting
solvent may have a function in preventing any enzymatic hydrolysis or other side reactions.
Nevertheless, it is difficult to predict the behavior of a certain solvent as a lipid extractant
(Christie & Han, 2010).
The two main structural features of lipids controlling their solubility in organic solvents are
the hydrophobic hydrocarbon chains of the fatty acid or other aliphatic moieties and any polar
functional groups, such as phosphate or sugar residues, which are markedly hydrophilic.
Simple lipids including fatty acids, triacylglycerols, sterols, sterol and wax ester, are often a
large part of storage tissues, from which they are relatively easy to extract. Complex lipids in
contrast, including glycerophospholipids, glyceroglycolipids, and sphingolipids, are usually
constituents of membranes, being in a close association with compounds like proteins and
polysaccharides, with which they interact; thus, the extraction of these lipids is more laborious.
Lipids with functional groups of low polarity only, such as triacylglycerols or cholesterol
esters, are very soluble in hydrocarbon solvents like hexane, cyclohexane or toluene, and in
solvents of somewhat higher polarity, such as chloroform or ethers. They tend to be rather
insoluble in polar solvents such as alcohols, and methanol especially; solubility in polar
solvents arises as the chain-length of the fatty acid moieties in these lipids decrease or as the
chain-length of the solvent alcohol increases. The polar complex lipids tend to be only
sparingly soluble in hydrocarbon solvents, though the presence of other lipids can aid
dissolution, they do dissolve readily in more polar solvents such as chloroform, methanol, and
ethanol (Christie & Han, 2010; Romanik, Gilgenast, Przyjazny, & Kamiński, 2007; Zahler &
Niggli, 1977).
62
Chapter 3
There are a large number of methods for lipid extraction, which vary in their lipid extraction
efficiency (Christie, 1997; Romanik et al., 2007; Self, 2007). A peculiar method presents its
advantages and disadvantages as, for example, the technique developed by Bligh and Dyer it
is known to be suitable for substrates having less than 10% of lipid content; otherwise the error
will be more prominent (Bligh & Dyer, 1959). As for Soxhlet extraction, it has the advantages
of always applying fresh solvent into contact with the solid matrix and relatively simple and
inexpensive equipment that is easy to operate but requires long extraction time (Luque de
Castro & Garcı́a-Ayuso, 1998). Some of the most common methods for lipid extraction (Asep
et al., 2008; Folch, Lees, & Sloane Stanley, 1957; Hara & Radin, 1978; Matyash, Liebisch,
Kurzchalia, Shevchenko, & Schwudke, 2008) are summarized in Table 3.1.
Table 3.1 Common methods for lipid extraction in literature and their characteristics.
Types of
Extraction
Methods Lipid Observations
Solvent
(Extracted)
Soxhlet-related Hexane or crude extract, extremely time
Non-polar
methods petroleum ether consuming
SFE (supercritical
CO₂ Non-polar polar lipids are not extracted
fluid extraction)
Chloroform, Non-polar
Folch good extraction method
methanol, H₂O and polar
Chloroform, Non-polar samples with a high % of
Bligh and Dyer
methanol, H₂O and polar endogeneus H2O
Hexane, Non-polar
Hara and Radin Isopropanol, glycolipids are not extracted
and polar
H₂O Non-polar
Butanol elimination (SPE)
and polar
Non-polar similar recoveries as those of
Matyash MTBE chloroform but longer incubation
and polar
time
Regarding the separation of the lipid compounds, solid phase extraction (SPE) is known to
be an efficient approach for sample preparation in the chromatographic analysis of edible fats.
This technique requires a small amount of the sample, low volumes of organic solvents, and it
is accomplished in a relatively shorter time if compared to other procedures (Panagiotopoulou
& Tsimidou, 2002; Ruiz-Gutiérrez & Pérez-Camino, 2000). Aminopropyl modified silica
phase it is a moderately polar phase used in alternative to silica offering better recoveries of
the neutral and acidic phospholipids than silica (Alvarez & Touchstone, 1992; Pernet, Pelletier,
& Milley, 2006). Moreover, aminopropyl columns are commercially available in the pre-
63
Chapter 3
packed form and different sizes, constituting a convenient and reproducible tool for lipid
separation.
Considering the absence of a standard extraction method and that about 98% of cocoa lipid
extract are neutral lipids, mainly constituted by triacylglycerols, the main propose of this work
was to achieve the complete extraction of the lipid constituents in cocoa and to analyse the
potential impact of different lipid extraction methods on cocoa lipidomics. Moreover,
separation of the other lipid classes from the TAG lipid class on cocoa beans will allow the
analysis of minor lipid components. For this purpose, selected well-established analytical
methods for lipid extraction and separation have been applied in comparison for a
comprehensive, focused lipidomics in seed plant material. Therefore, this chapter emphasizes
the importance of separation of the lipids in different classes for the effective identification and
further quantification of lipid compounds.
3.2 Material and Methods
3.2.1 Chemicals and reagents
Ethanol gradient grade was purchased from Merck (Darmstadt, Germany), isopropanol
(Rotisolv® HPLC grade), acetonitrile (Rotisolv® HPLC ultra gradient grade), chloroform
(Rotisolv® HPLC grade) and Tetra-dodecylammonium bromide, petroleum ether, boiling
range 35–60 °C, heptane, methanol, diethyl ether, ethyl acetate, was purchased from Carl Roth
(Karlsruhe, Germany), dichloromethane 99,8% stabilized with amylene for synthesis was
purchased from Panreac AppliChem (Darmstadt, Germany), ammonium formate LC-MS Ultra
and formic acid (puriss., ≥ 98% (T) for mass spectrometry), ferric chloride, perchloric acid,
ninhydrin reagent, dinitrophenylhydrazine, Dragendorff reagent (ready-to-use), 4-
methoxybenzaldehyde, and acetic acid were purchased from Sigma-Aldrich Chemie
(Steinheim, Germany). Ethanol was subjected to distillation prior use.
3.2.2 Preparation of the seed sample
Cocoa bean samples, received in several sets from Barry Callebaut and four types of nuts
(almonds, hazelnuts, peanuts, walnuts), coffee beans, one type of oily seed (sunflower seeds),
and a sample of tea (white tea leaves), purchased from the local supermarket (all, except the
tea, are produced by the same company) were deshelled and ground using a grinder Retsch
64
Chapter 3
Grindomix GM200 knife mill (Haan, Germany) to make a fine an homogenous powder.
Henceforth, several extraction methods (Folch et al., 1957; Hara & Radin, 1978; Hirsch &
Ahrens, 1958; Li et al., 2014) have been used to perform the final extraction. Finally, an
overnight Soxhlet (Buchi Extraction System B811 instrument, Flawil, Switzerland) using
dichloromethane as extraction solvent, 150 mL for 5 g of powdered seed material, was used.
All the samples were subjected to each extraction method at least 3 times. Extracted lipids were
quantified gravimetrically after evaporation to dryness in a rotavapor. The dry residue was then
stored at -20 °C until further analysis. Moisture content was determined by heating around 2.5
g of cocoa powder in the oven at 105 °C for 24, cooled in the dry atmosphere of a desiccator,
and then reweighed. Total lipid content was normalized according to:
Lipid extract (g) × 100
= .
Cocoa Powder (g) − Weight Loss (g)
3.2.3 Thin layer chromatography (TLC)
The lipid extract (100 mg) was dissolved in 1 mL dichloromethane/methanol (2/1, v/v) and
8 µL were spotted on the TLC Silica gel 60 F254 HPTLC Aluminium plate 20 cm x 20 cm
(Macherey-Nagel, München, Germany). The dried lipid spots were developed with
hexane/diethyl ether/glacial acetic acid (80:20:1, by vol.). Plates were allowed to dry and run
another time, for half of the plate, with chloroform/methanol/glacial acetic acid (97:3:1, by
vol.). After first drying time spots having chromophores were visualized under UV light at 254
nm. Then, after a second complete drying time, the plate was sprayed with a general purpose
stain, Cerium Molybdate stain (Hanessian's stain), to visualize all the separated spots of the
lipid extract. Others stain solutions were used to support compounds identification: iodine
vapor (high affinity for both unsaturated and aromatic compounds), ferric chloride (ideal for
phenols), ninhydrin (excellent for amino acids), dinitrophenylhydrazine (suitable for aldehydes
and ketones), p-anisaldehyde (multipurpose stain, especially for functional groups which are
strongly and weakly nucleophilic), and Dragendorff’s reagent (optimal for choline and to detect
alkaloids).
3.2.4 Solid phase extraction (SPE)
Chromabond aminopropyl modified silica gel NH2 (Chromabond® LV-NH2, Macherey-

Nagel GmbH & Co. KG, Düren, Germany) column cartridges (3 mL/500 mg) were used for
65
Chapter 3
the SPE fractionation. The procedure was developed by reviewing the available SPE separation
methods, starting with the original method Kaluzny et al. (Kaluzny, Duncan, Merritt, & Epps,
1985) and observing the modifications applied by other authors (Alvarez & Touchstone, 1992;
Laffargue, de Kochko, & Dussert, 2007; Pernet et al., 2006). After few trials and modifications,
the following method has been achieved. The cartridges were first activated with 20 mL
heptane/ethylacetate (80/20, v/v), then, 20 ml heptane. Then, 10 mg of the lipid extract
dissolved in 1 mL of heptane were applied, after which the fractions were eluted with 5
different solvent mixtures: 1) 20 mL heptane, first fraction collected ; 2) 20 mL heptane/ethyl
acetate (97/3, v/v), for the second fraction; 3) 20 mL heptane/ethyl acetate (85/15, v/v), third
fraction collected; 4) 20 mL heptane/ethyl acetate (85/15, v/v) plus 1% acetic acid, for the forth
fraction; and 5) 20 mL methanol. Extracted lipids were evaporated to dryness in a rotavapor
and stored at -20 °C until further analysis.
3.2.5 Silica gel column chromatography
A glass column packed with silica gel 60 (230 mesh) was prepared in petroleum ether and
used. 1 g of raw cocoa lipid extract was passed through the column (15 cm × 3 cm i.d.), using
a solvent gradient from 100% heptane (0% ethylacetate) until 50% ethylacetate in heptane. 15
fractions were collected, evaporated to dryness in a rotavapor, and stored at -20 °C until further
analysis.
3.2.6 HPLC chromatographic conditions
TAGs molecular species were separated using an HPLC equipment (Agilent 1100 series,
Waldbronn, Germany). The column used in this study was a Pursuit XRs C18 (250 mm × 3
mm i.d., 5 μm particles). The temperature of the column oven was set to 35 °C. 3 µL of sample
were injected. Solvent A consisted of acetonitrile with 0.01% formic acid and solvent B
consisted of ethanol with 5 mmol/L ammonium formate and 0.01% formic acid. The mobile
phase was pumped through the column at a flow rate of 0.6 mL/min. The gradient elution
program consisted of holding solvent steady A/B (40/60) for 3 min; followed by a linear
gradient to solvent B (100) for 33 min, and ending with isocratic elution at solvent B (100) for
10 min. The column was equilibrated at 40/60 solvent A/B for 5 min before reuse.
3.2.7 Mass spectrometry detection
The samples (10 μg/mL lipid) were mixed with the corresponding HCCA (α-Cyano-4-
66
Chapter 3
hydroxycinnamic acid) matrix solution (1:1, v/v) and 1 μL of this mixture was transferred onto
the MALDI target. The samples were allowed to crystallize at room temperature under ambient
conditions. All mass spectra were acquired on a Bruker Autoflex MS device (Bruker Daltonics,
Bremen, Germany). One thousand single laser shots were averaged for each mass spectrum.
High-resolution masses were acquired using a time of flight MicrOTOF Focus mass
spectrometer (Bruker Daltonics, Bremen, Germany) fitted with an ESI source used as the
detector with the following parameter settings: capillary voltage of 4.5 kV; nebulizing gas
pressure of 2 Ba; drying gas flow rate of 10 L/min; drying gas temperature of 220 °C. ESI mass
spectra were measured in the range of m/z 200–1200 in the positive ion mode. Internal
calibration was achieved with 10 mL of 0.1 M sodium formate solution injected through a six-
port valve prior to each chromatographic run. Calibration was carried out using the enhanced
quadratic mode.
LC-tandem MS was carried out using an Ion-Trap detector in positive ion mode equipped
with an ESI source (Bruker Daltonics UHT Ultra, Bremen, Germany). The full scan mass
spectra were recorded in the range m/z 200-1200 operating in positive ion mode. Capillary
temperature was set to 350 °C, drying gas flow rate of 10 L/min and nebulizer pressure of 10
psi. Tandem mass spectra were acquired in Auto MSn (smart fragmentation) using a ramping
of the collision energy.
3.3 Results and discussion
3.3.1 Extraction optimization
To analyze the entire content of cocoa, first, it is necessary to optimize the extraction
method. Much of work has been conducted on observing the efficiency of different extraction
procedures for extracting lipids from plant matrices. Among different kinds of extraction
procedures, Soxhlet extraction proved to be the most exhaustive method for the samples.
Moreover, it offers better reproducibility compared to the other methods. Table 3.2 displays the
list of methods used for the lipid extraction in cocoa beans. The extracted total lipid percentage
is highly depending on the extraction solvent used (Table 3.2). It is known that a more polar
solvent will extract a greater amount of lipids due to its releasing properties of the lipids from
their protein-lipid complexes (Christie & Han, 2010; Ryckebosch, Muylaert, & Foubert, 2012).
Nevertheless, our results showed that other non-lipid compounds are extracted each time
67
Chapter 3
mixtures containing methanol were applied (Fig. 3.1). These compounds change their color
passing through red-pink color in acidic environment to a purple then yellow color in basic
environment (Fig. 3.1 (b) and (c)). These compounds are anthocyanins (Pina, 2012) well
known compounds present in cocoa (D'Souza et al., 2017; Patras, Milev, Vrancken, & Kuhnert,
2014). By treating the TLC plate with different stain solutions, we can obtain information about
the nature of the compounds analyzed. Figure 3.1 (d), illustrates the TLC stained with ferric
chloride, which indicates that the extracting mixture DCM/MeOH or CHCl3/MeOH extracts
also some phenolic compounds. From the amount of the lipid extracted (Table 3.1) and the
lipid profile image (Fig. 3.1 (a)) offered by the TLC, it could be asserted that DCM was the
most suitable solvent to extract the highest amount of lipids from qualitative and quantitative
perspective. Cequier-Sanchez et al. have studied the efficiency of DCM with respect to the
well-known Folch method (Cequier-Sánchez, Rodríguez, Ravelo, & Zárate, 2008). Their
results indicate that, for lipid extraction and fatty acid assessment, dichloromethane/methanol
can successfully replace the commonly employed chloroform/methanol, giving the same or
even better yields. DCM is a less hazardous and less toxic solvent and some times more
efficient than CHCl3. Optimization of the extraction (Fig 3.2) has shown that already after 6 h
extraction time, more than 55% of lipid is extracted. DCM has proven to be a good extraction
solvent not only for cocoa but also for other seeds including a tea sample (Fig. 3.3).
Table 3.2 Amount of cocoa lipids extracted from 3 g of dried beans by different extraction methods and
different solvents.
LIPID WEIGHT LOSS IN THE
SOLVENTS
METHODS EXTRACT EXTRACTION THIMBLE
(150 ML)
(GRAMS) (GRAMS)
Soxhlet Heptane 1.64 (±0.13) 1.51 (±0.03)
(continuous Petroleum ether 1.41 (±0.07) 1.44 (±0.03)
mode) Dichloromethane (DCM) 1.61 (±0.06) 1.63 (±0.03)
Methyl tert-butyl ether 1.65 (±0.1) 1.46 (±0.03)
(MTBE):Methanol (MEOH) (10:3)
Chloroform (CHCl3):MeOH (2:1) 1.65 (±0.01) 1.87 (±0.04)
DCM:MeOH (2:1) 1.6 (±0.01) 1.8 (±0.13)
Ultrasonic Bath Petroleum ether 1.23 1.4
Heptane 1.16 1.58
CHCl3:MeOH (2:1) 1.31 1.34
Homogenization DCM 1.58 NOT PRESENT
DCM:MeOH (2:1) 1.69 NOT PRESENT
CHCl3:MeOH (2:1) 1.66 NOT PRESENT
Petroleum ether 1.27 NOT PRESENT
68
Chapter 3
Figure 3.1 (a) TLC plate image showing the lipid profile of fermented dried beans extracted with
Soxhlet apparatus: 1 – petroleum ether; 2 – CHCl3:MeOH (1:1); 3 – heptane; 4 –DCM; 5 – Folch
extraction CHCl3:MeOH:0.05 N potassium chloride (8:4:3); and 6 – diethyl ether. Iodine staining and
UV light at 254 nm visualized the spots. (b) and (c) show non-lipid compounds extracted with
CHCl3:MeOH (2:1, v/v), and (d) TLC plate treated with ferric chloride for the phenols presence (F –
Folch; DCM, DCM:MeOH, CHCl3:MeOH).
Only in coffee sample, DCM has shown to extract less lipid amount than heptane. It might
be because coffee seeds contain less than 10% lipid amount compared to the other oils seed
under study. Moreover, this indicates that the optimization of an extraction has to take into
account the lipid amount in a certain sample. The reasons of choosing a tea leafs sample were
two: tea is a beverages, next to coffee and cocoa (in some applications), and we wanted to test
the extraction power of the three most used organic solvents (petroleum ether, heptane, and
DCM) in extracting pigments. For the last point, DCM revealed good extraction efficiency.
Figure 3.2 Optimization of lipid extraction in cocoa by Soxhlet method using DCM as extraction
solvent. 3 g of fermented dried powdered cocoa beans and 150 mL solvent.
69
Chapter 3
Figure 3.3 Percentage of lipid extracted form 6 seeds and 1 dried leaves samples by Soxhlet
method using three different extraction solvents. 3 g of powdered plant material and 150 mL solvent.
3.3.2 TLC lipid profile
TLC is an established technique of separation of lipids. Although, TLC analyses on cocoa

lipids have been used for the separation of cocoa phospholipids in the past (Levanon, Rossetini,
Raskin, & Mesquita, 1967; Parsons, Keeney, & Patton, 1969), it has proved to be reliable tool
for current lipidomics. In this study, TLC analyses were performed with the aim to assess the
cocoa lipid content in terms of quality and quantity of extracted cocoa lipids. As observed in
Figure 3.4 (a) and (b) TLC separation could provide an important piece of information. In this
example, a continuous extraction employing first petroleum ether for 3 h, then heptane for
another 3 h on the same sample. And ending with another 3 h extraction time with DCM on
the same sample. As result, we could assert that DCM is a more exhaustive solvent compared
to heptane which in turn is more efficient than petroleum ether. Cocoa bean processing implies
many treatments, with the main being: fermentation, drying, and roasting. TLC image (Fig.
3.4) show that processing modifies the lipid profile of the beans as it occurs in proteins (Amin
I, Jinap S, & Jamilah B, 1999; Kumari et al., 2016) and flavor related compounds (Hoskin &
Dimick, 1994). Moreover, Figure 3.4 (a) illustrates the possible presence of aromatic
compounds or heterocyclic compounds which have electron donating or electron withdrawing
functional groups that quench fluorescence and can be seen under UV light. Asep et al. have
70
Chapter 3
shown that the particle size and the extraction time have in impact on the yield but not on the
lipid profile of the fat (Asep et al., 2008). This analysis approach gives relevant evidence on
how the production process affects the lipid extract of cocoa. Figure 3.5 and Figure 3.6
illustrate the lipids profile of the coffee sample, tea leafs, sunflower seeds, hazelnuts and
almonds in comparison to cocoa lipid profile when extracted with different organic solvents.
Images show the efficiency of different solvent and how diverse are the lipid profile of the
samples under study. Although solvent extraction has shown no significant differences in terms
of lipid amount extracted (Fig. 3.3), DCM extracted the spot at the bases (Fig. 3.5 and Fig 3.6,
spot at the origin), known as polar lipids, in higher quantity. As illustrated in the Figures, TLC
image offers a comparison element for assessing lipid profile similarities in different seed
samples. Thus, we can perceive how cocoa lipid profile differs, concerning lipids classes, from
other seed plant material. Along with qualitative results, TLC can also provide a
chromatographic measurement known as an Rf value, which is the retention or the ratio-to-
front factor. This value is determined as the distance traveled by the compound divided by the
distance traveled by the solvent, expressed as a decimal fraction. Compounds with the higher
Rf are less polar as they interact less with the polar adsorbent on the TLC plate.
Accordingly, the polarity of the lipid compounds could be easily inferred. Thus, cocoa lipid
extract showed four prominent spots on the TLC plate with Rf values of 0.95, 0.64, 0.18, and 0
in order of increasing polarity. The spot at Rf of 0.64 are compounds present in the highest
amount in the lipid extract – triacylglycerols.
Figure 3.4 (a) Visualization under UV light at 254 nm and (b) Hanessian's stained TLC plate
image showing the lipid profile of different treated beans extracted with Soxhlet apparatus using P –
petrolium ether (2, 5, 8), H – heptane (3, 6, 9) and DCM – dichloromethane (4, 7, 10) by subsequent
extraction compared with Folch extraction of the dried cocoa beans.
71
Chapter 3
Figure 3.5 TLC plate* image showing the lipid profile of white tea leaves, cocoa beans, coffee
beans and sunflower seeds extracted with Soxhlet apparatus using P – petrolium; H – heptane; and
DCM – dichloromethane (3 g sample and 150 mL solvent). *Hanessian's stain was used.
Figure 3.6 TLC plate* image showing the lipid profile of white hazelnuts and almonds in
comparison to cocoa beans extracted with Soxhlet apparatus using P – petrolium; H – heptane; and
DCM – dichloromethane (3 g sample and 150 mL solvent). *Hanessian's stain was used.
72
Chapter 3
3.3.3 Separation of lipid classes
Due to the high amount of triacylglycerols in cocoa lipid extract, the SPE methods described
in the literature for the separation of lipids (Alvarez & Touchstone, 1992; Kaluzny et al., 1985;
Laffargue et al., 2007; Pernet et al., 2006) seemed not very efficient. Nonetheless, these results
served as a starting point for the optimization of a SPE method for cocoa lipids. Thus, the
optimized method, using heptane/ethyl acetate gradient solvent system for the first four
fractions and methanol for the last elution, can efficiently separate the lipid compounds of
cocoa in five fractions, corresponding to at least one class of lipid compounds. The method has
a relatively rapid speed, simple process, and low cost, providing a practicable analytical
method for basic research.
In oilseeds, the lipid matrix is generally composed by 80 to 90% triacylglycerols, 1 to 5%
diacyl- and monoacylgycerols, 1 to 2% sterols and sterol esters, 0.1-1% polar lipids and ˂1%
free fatty acids (Laffargue et al., 2007). TLC staining with different staining reagents (Fig. A3)
and compound assignment based on mass spectrometry data have allowed the identification
and assignment of the main classes in cocoa (Fig. 3.7), namely sterol esters, triacylglycerols,
diacylglycerols, free fatty acids, and polar lipids. Fraction 5 contains, besides sterols esters
glucosides, identified by tandem MS fragmentation, other polar lipids containing amine and
amide groups as this fraction reacts positively to the ninhydrin and Dragendorff’s reagent (Jork,
1990). Most probably, these compounds belong to phospholipids class.
The other separation method was a simple separation of cocoa lipids using a preparative
silica gel column (Fig. 3.8). The results have been shown to be satisfactory for the separation
of the components on a large scale. With minor changes, this method can be improved to
achieve the separation of each spot seen on the TLC plate. Notwithstanding, an SPE method
would be more appropriate and certainly the information provided by the two methods should
be integrated. The fractions obtained by silica gel column separation were further analyzed by
HPLC-MS. The main purpose of this separation was to illustrate the overall profile of cocoa
fat. In addition, this analysis confirmed that there are more than one class of lipid compounds
when considering separation by SPE. The two dimensional (2D map) maps of each fraction
illustrates the complexity of cocoa lipid extract (Fig. 3.8 (b)). The 2D map of the fraction 3
showed all triacylglycerol known in cocoa, while the rest of the fractions (1, 4-16) represent
other hundreds of compounds not yet fully described in literature.
73
Chapter 3
74
Chapter 3
75
Chapter 3
3.4 Conclusions
Although cocoa butter is known to be a relatively simple fat, in which the predominant
compounds are the triacylglycerols, with an appropriate extraction and separation method the
complexity of this fat can be thorough fully appreciated. There is no doubt that this approach
in combination with proper identification and quantification assessment can give more insight
into the chemistry of cocoa butter and thus help to explain the intricate nature of its physical
properties. Relative changes in lipid amounts with varying conditions may point out to novel
targets.
References
Alvarez, J. G., & Touchstone, J. C. (1992). Separation of acidic and neutral lipids by aminopropyl-bonded silica
gel column chromatography. Journal of Chromatography B: Biomedical Sciences and Applications, 577(1), 142–
145. https://doi.org/10.1016/0378-4347(92)80609-T
Bligh, E. G., & Dyer, W. J. (1959). A Rapid Method of Total Lipid Extraction and Purification. Canadian Journal
of Biochemistry and Physiology, 37(8), 911–917. https://doi.org/10.1139/o59-099
Cequier-Sánchez, E., Rodríguez, C., Ravelo, Á. G., & Zárate, R. (2008). Dichloromethane as a Solvent for Lipid
Extraction and Assessment of Lipid Classes and Fatty Acids from Samples of Different Natures. Journal of
Agricultural and Food Chemistry, 56(12), 4297–4303. https://doi.org/10.1021/jf073471e
Christie, W. W. (1997). Advances in Lipid Methodology. Elsevier.
D'Souza, R., Grimbs, S., Behrends, B., Bernaert, H., Ullrich, M., & Kuhnert, N. (2017). Origin-based
polyphenolic fingerprinting of Theobroma cacao in unfermented and fermented beans. Food Research
International. https://doi.org/10.1016/j.foodres.2017.06.007
Folch, J., Lees, M., & Sloane Stanley, G. H. (1957). A simple method for the isolation and purification of total
lipides from animal tissues. The Journal of Biological Chemistry, 226(1), 497–509.
76
Chapter 3
Hara, A., & Radin, N. S. (1978). Lipid extraction of tissues with a low-toxicity solvent. Analytical Biochemistry,
90(1), 420–426.
Hirsch, J., & Ahrens, E. H. (1958). The Separation of Complex Lipide Mixtures by the Use of Silicic Acid
Chromatography. Journal of Biological Chemistry, 233(2), 311–320. Retrieved from
http://www.jbc.org/content/233/2/311
Jork, H. (1990). Thin-layer chromatography: reagents and detection methods. VCH.
Kaluzny, M. A., Duncan, L. A., Merritt, M. V., & Epps, D. E. (1985). Rapid separation of lipid classes in high
yield and purity using bonded phase columns. Journal of Lipid Research, 26(1), 135–140.
Laffargue, A., de Kochko, A., & Dussert, S. (2007). Development of solid-phase extraction and methylation
procedures to analyse free fatty acids in lipid-rich seeds. Plant Physiology and Biochemistry, 45(3–4), 250–257.
https://doi.org/10.1016/j.plaphy.2007.01.012
Levanon, Y., Rossetini, S. M. O., Raskin, M., & Mesquita, M. T. P. (1967). Thin‐Layer Chromatographic Study
on the Lipid Components of Cocoa Beans and Cocoa Butter. Journal of Food Science, 32(6), 609–610.
https://doi.org/10.1111/j.1365-2621.1967.tb00845.x
Li, Y., Ghasemi Naghdi, F., Garg, S., Adarme-Vega, T. C., Thurecht, K. J., Ghafor, W. A., … Schenk, P. M. (2014).
A comparative study: the impact of different lipid extraction methods on current microalgal lipid research.
Microbial Cell Factories, 13, 14. https://doi.org/10.1186/1475-2859-13-14
Luque de Castro, M. D., & Garcı́a-Ayuso, L. E. (1998). Soxhlet extraction of solid materials: an outdated
technique with a promising innovative future. Analytica Chimica Acta, 369(1–2), 1–10.
https://doi.org/10.1016/S0003-2670(98)00233-5
Matyash, V., Liebisch, G., Kurzchalia, T. V., Shevchenko, A., & Schwudke, D. (2008). Lipid extraction by methyl-
tert-butyl ether for high-throughput lipidomics. Journal of Lipid Research, 49(5), 1137–1146.
https://doi.org/10.1194/jlr.D700041-JLR200
Panagiotopoulou, P. M., & Tsimidou, M. (2002). Solid phase extraction: Applications to the chromatographic
analysis of vegetable oils and fats. Grasas y Aceites, 53(1). https://doi.org/10.3989/gya.2002.v53.i1.292
Parsons, J. G., Keeney, P. G., & Patton, S. (1969). Identification and Quantitative Analysis of Phospholipids in
Cocoa Beans. Journal of Food Science, 34(6), 497–499. https://doi.org/10.1111/j.1365-2621.1969.tb12069.x
Patras, M. A., Milev, B. P., Vrancken, G., & Kuhnert, N. (2014). Identification of novel cocoa flavonoids from
raw fermented cocoa beans by HPLC–MSn. Food Research International, 63(Part C), 353–359.
https://doi.org/10.1016/j.foodres.2014.05.031
77
Chapter 3
Pernet, F., Pelletier, C. J., & Milley, J. (2006). Comparison of three solid-phase extraction methods for fatty acid
analysis of lipid fractions in tissues of marine bivalves. Journal of Chromatography A, 1137(2), 127–137.
Pina, F. (2012). Chemistry and applications of flavylium compounds: a handful of colours. Chemical Society
Reviews, 41, 869–908. https://doi.org/10.1039/c1cs15126f
Romanik, G., Gilgenast, E., Przyjazny, A., & Kamiński, M. (2007). Techniques of preparing plant material for
chromatographic separation and analysis. Journal of Biochemical and Biophysical Methods, 70(2), 253–261.
https://doi.org/10.1016/j.jbbm.2006.09.012
Ruiz-Gutiérrez, V., & Pérez-Camino, M. C. (2000). Update on solid-phase extraction for the analysis of lipid
classes and related compounds. Journal of Chromatography A, 885(1–2), 321–341.
https://doi.org/10.1016/S0021-9673(00)00181-3
Ryckebosch, E., Muylaert, K., & Foubert, I. (2012). Optimization of an Analytical Procedure for Extraction of
Lipids from Microalgae. Journal of the American Oil Chemists’ Society, 89(2), 189–198.
https://doi.org/10.1007/s11746-011-1903-z
Self, R. (2007). Extraction of Organic Analytes from Foods: A Manual of Methods. Royal Society of Chemistry.
Zahler, P., & Niggli, V. (1977). The Use of Organic Solvents in Membrane Research. In Methods in Membrane
Biology (pp. 1–50). Springer, Boston, MA. https://doi.org/10.1007/978-1-4684-2910-7_1
78
Chapter 4
Chapter 4:
4 Characterization of triacylglycerols in
unfermented cocoa beans by HPLC-ESI mass
spectrometry
Diana Sirbua, Marcello Cornob, Matthias S. Ullrich a, and Nikolai Kuhnerta*
a
Department of Life Sciences and Chemistry, Jacobs University, Campus Ring 1, D-28759 Bremen,
Germany
b
Barry Callebaut, Aalstersestraat 122, B-9280 Lebbeke-Wieze, Belgium
*Correspondence to:
Phone: +49 421 200 3120,
Fax: +49 421 200 3102.
Published in Food Chemistry: doi.org/10.1016/j.foodchem.2018.01.194
Supplementary material available with the published article.
79
Chapter 4
Abstract
The nutritional value of cocoa butter is mainly determined by the composition of

triacylglycerols (TAGs). In this paper we have developed a non-aqueous reversed-phase liquid
chromatographic method, using ethanol as the mobile phase, coupled to electrospray ionization
(ESI) tandem mass spectrometry to identify TAGs in raw cocoa beans from six different
origins. Tandem mass spectrometry was adopted to facilitate the identification of TAGs using
unique diacylglycerol product ions and neutral losses. Additionally, two-dimensional m/z
retention time maps aided the identification of entire homologous series of TAGs. A total of 83
different TAGs were identified in unfermented cocoa beans, 58 of which were not previously
reported in cocoa. 31 of these compounds represent a new class of TAGs characterized by the
presence of one to three hydroxyl groups on the unsaturated fatty acid chain. To date, this
represents the largest number of TAGs identified in cocoa.
Keywords: triacylglycerol; high-performance liquid chromatography; tandem mass spectrometry; raw

cocoa beans; homologous series; 2D map; neutral loss.
80
Chapter 4
4.1 Introduction
Theobroma cacao is a small evergreen tree in the family Sterculiaceae, native to Central
and South America. Cacao trees need rich, well-drained soils; they naturally grow within 20°
latitude of the equator as they need about 2000 mm of rainfall a year, and temperatures in the
range of 21 to 32 °C (Young, 2007). The seeds of Theobroma cacao, cocoa beans contain next
to proteins, polyphenols, carbohydrates, alkaloids and minerals, lipids as their main chemical
constituents accounting for 45-60% of their dry mass (Visioli et al., 2009). About 98% of cocoa
lipid mass is represented by triacylglycerols (TAGs), which are hydrophobic molecular species
formed by esterification of three fatty acids (FAs) with a glycerol backbone. Their FAs
composition varies depending on the geographical origins, varieties of Theobroma cacao,
growing seasons, and cultivation method (Lipp and Anklam, 1998a). Two saturated FAs with
oleic acid in the middle characterize a typical pattern of distribution of FAs on the TAG body.
The main TAGs are 1,3-distearoyl-2-oleoyl-sn-glycerol (SOS), 1-palmitoyl-2-oleoyl-3-
stearoyl-sn-glycerol (POS) and 1,3-dipalmitoyl-2-oleoyl-sn-glycerol (POP) that are known to
be major components in cocoa butter (CB) and discriminatory marker compounds allowing
identification of adulterations. Furthermore, this distribution of the FAs within the glycerol
backbone has an essential nutritional and functional value highly appreciated in food, cosmetic
and pharmaceutical industries (Ulberth and Buchgraber, 2003). TAGs are relative simple lipid
substances, but they play an indispensable role in energy production and storage in a living
organism (Carrasco-Pancorbo et al., 2009). TAGs also have an important value in nutrition and
health being the main suppliers of essential fatty acids involved in maintaining life activities
(Visioli et al., 2009). In seed plants, carbohydrates are accumulated as storage energy, and then
transformed in TAGs during the growth stages. TAGs are then reconverted back into
carbohydrates during germination with the incorporation of a high amount of water. One gram
of oil is converted into approximately 2 g of carbohydrates (Murphy, 1994).
The analytical procedure for the characterization of individual TAG compounds involves
three elementary steps: extraction, chromatographic separation, and identification. TAG
profiling and analysis is performed by gas liquid chromatography (GLC) (Ulberth and
Buchgraber, 2003) and high performance liquid chromatography (HPLC) using normal phase
HPLC and non-aqueous reversed-phase (NARP) HPLC (Christie and Han, 2010; Jakab et al.,
2002; Lísa and Holcapek, 2008). NARP-HPLC has been widely applied for the separation of
81
Chapter 4
the intricate acylglycerol class because it offers, under optimized conditions, better resolution
of the peaks and separation of TAGs according to the acyl chain lengths and the number of
double bonds (DBs). The retention is ruled by equivalent carbon number (ECN) defined as
ECN=CN-2DB, where CN is the number of carbon atoms (Holčapek et al., 2003; Lísa and
Holcapek, 2008). Due to lack of chromophores characterization of TAGs is carried out mainly
using mass spectrometry (MS) using either atmospheric pressure chemical ionization (APCI)
or electrospray ionization (ESI) techniques. TAGs show a unique fragmentation pattern in
tandem MS, since the most abundant fragment ions are formed by loss of one acyl moiety with
consequent formation of diaclyglycerols product ions. In spite of this fact, they provide a
challenge for mass spectrometric analysis because of the intricacy of naturally occurring TAGs,
formation of multiple adduct ions and co-elution in chromatographic separation. Combinations
of different fatty acyl substituents in TAGs, different stereochemical positions sn-1, 2 or 3 on
the glycerol backbone (regioisomers) or R/S optical configuration of TAGs esterified in sn-1
and sn-3 positions by two different FAs (optical isomers) results in a large number of molecular
species having the identical elemental composition. Ammonium adducts of TAGs, [M+NH4]+,
generated by electrospray ionization, enable the molecular weight of each TAG molecular
species to be determined. This allowed for tandem mass spectroscopy to be used for the
elucidation of TAG (Cvacka et al., 2006; Holčapek et al., 2003; Hsu and Turk, 2010; McAnoy
et al., 2005; Murphy et al., 2007). Kuhnert et al. have recently shown that fermented cocoa
beans constitute an extremely complex food matrix containing more than 20 000 chemical
entities by using Fourier-transform ion cyclotron resonance (FTICR) mass spectrometry
technique (Kuhnert et al. 2013; Milev et al., 2014). It is, therefore, a challenge to detect and
differentiate different types of TAGs in cocoa beans and various tools have to be applied.
Despite the many studies of the chemical composition of TAGs in cocoa butter previously
published, there is no full characterization of the TAGs in unfermented cocoa beans which are
also known as raw or wet cocoa beans (Lipp and Anklam, 1998b; Lipp et al., 2001; Segall et
al., 2005; Simoneau et al., 1999; Ulberth and Buchgraber, 2003). This contribution presents a
new approach to the analysis of TAGs in different origin raw cocoa beans using a NARP HPLC
with ethanol as mobile phase, ESI tandem mass spectrometry and two dimensional (2D) map
analysis revealing the true complexity of cocoa butter and thus allowing structure elucidation
the of the most abundant lipid molecular species in cocoa.
82
Chapter 4
4.2 Material and methods
(Rotisolv® HPLC grade), acetonitrile (Rotisolv® HPLC ultra gradient grade) and chloroform
(Rotisolv® HPLC grade) was purchased from Carl Roth (Karlsruhe, Germany), dichloromethane
99,8% stabilized with amylene for synthesis was purchased from Panreac AppliChem
(Darmstadt, Germany), ammonium formate LC-MS Ultra and formic acid (puriss., ≥ 98% (T)
for mass spectrometry) were purchased from Sigma-Aldrich Chemie (Steinheim, Germany).
Ethanol was subjected to distillation prior use. Standard TAG samples used in this study were:
TAG (16:0/18:1/16:0), TAG (16:0/16:0/16:0), TAG (18:0/18:0/18:0), and TAG (18:1/18:1/18:1)
all at ≥ 99% were purchased from Sigma-Aldrich Chemie (Steinheim, Germany).
4.2.2 Sample preparation
Frozen cocoa bean samples from different origins were received in several sets from Barry
Callebaut (Lebbeke-Wieze, Belgium). A total of 12 samples coming from Ivory Coast,
Indonesia, Ecuador, Tanzania, Malaysia, and Brazil were analyzed. Firstly, the seed material
was crushed using a grinder Retsch Grindomix GM200 knife mill (Haan, Germany) with the
purpose of making homogenous powder. Henceforth, an overnight Soxhlet (Buchi Extraction
System B811 instrument) method using dichloromethane as an extraction solvent, 150 mL of
5 g of crushed beans, was used. Extracted lipids were quantified gravimetrically after
evaporation to dryness in a rotavapor. The dry residue was then stored at -20 °C until further
analysis. For HPLC analysis, a concentration of 0.045 mg/mL in chloroform/ethanol (50/50)
of cocoa lipid extract was prepared.
83
Chapter 4
consisted of ethanol with 5 mmol/L ammonium formate and 0.01% formic acid. The mobile
4.2.4 High-resolution Mass spectrometry conditions
High-resolution masses were acquired using a time of flight (TOF) MicrOTOF Focus mass
quadratic mode. Molecular formulae suggestions were accepted if the mass error was below 5
ppm.
4.2.5 Tandem Mass spectrometry conditions
4.2.6 Enrichment of polar TAG fraction
1 g of raw cocoa lipid extract was passed through a preparative silica gel column (15 cm ×
3 cm i.d.), using a solvent gradient from 0% ethylacetate in heptane until 50% ethylacetate in
heptane, in order to prepare an enriched fraction of the TAGs eluting at lower retention times
84
Chapter 4
(from 4 to 25 min) in HPLC-MS. This fraction shown a more polar character and was used
for further analysis by nuclear magnetic resonance spectroscopy (NMR).
4.2.7 Derivatization of hydroxylated fatty acids
Derivatization of the hydroxyl group was carried out on 100 mg of the enriched fraction
obtained from the preparative silica gel column. One 0.5 mL benzoyl chloride with 200 µL of
0.1M DMAP (4-(N,N-dimethylamino)pyridine) in pyridine. The reaction was performed in
dichloromethane and stirred at room temperature for 1 h. The benzoate obtained were diluted
in chloroform/ethanol (50/50) and 3 µL were injected in HPLC-MS.
4.2.8 NMR spectroscopy
1
H NMR and 13C NMR spectra were acquired on a JEOL ECX-400 spectrometer (Eching,
Germany) operating at 400 MHz for 1H NMR and 100 MHz for 13C NMR at room temperature
in CDCl3, using a 5 mm probe. The chemical shifts (δ) are reported in parts per million and
were referenced to the residual solvent peak. Enriched cocoa lipid extract fraction was
prepared by dissolving 0.1g of sample in 2.0mL of deuterated chloroform. The 1H-NMR
and 13C NMR shifts were determined from a single analysis of the enriched fraction. To
verify the 1H-NMR and 13
C NMR assignments, 1H1H correlation spectroscopy (COSY) and
1
H13C correlation spectroscopy (HMBC and HMQC) experiments were conducted.
Triacylglycerols of cocoa butter (CB) have been analyzed on several occasions (Podlaha et
al., 1984; Řezanka and Mareš, 1991) and reviewed by Lipp and Anklam (Lipp and Anklam,
1998a, 1998b; Lipp et al., 2001). At this stage, a total of 23 different TAGs were reported in
cocoa butter obtained from fermented cocoa beans. However, no attention has been paid to the
correct assignment of the fatty acid regiochemistry concerning the glycerol backbone. The first
correct assignment was accomplished by a chiral phase HPLC method by Takagi and Ando
(Takagi and Ando, 1995). Fauconnot et al. shown that TAG regioisomers elute at the same
retention time on a C18 column. They have described the major regioisomers in cocoa: 1,3-
85
Chapter 4
dipalmitoyl-2-oleoyl-sn-glycerol (POP, 100% regioisomeric purity), 1,3-distearolyl-2-oleoyl-

sn-glycerol (SOS, 100% regioisomeric purity), rac-palmitoylstearoyl-2-oleoyl-sn-glycerol
(rac-POS, >90% regioisomeric purity), 1,3-dipalmitoyl-2-linoleoyl-sn-glycerol (PLP, 100%
regioisomeric purity), 1(3),2-dioleoyl-3(1)-stearoyl-sn-glycerol (SOO, 90% regioisomeric
purity), and 1(3),2-dioleoyl-3(1)-palmitoyl-sn-glycerol (POO, 100% regioisomeric purity)
identified on the basis of diaclyglycerol fragment abundance in comparison with authentic
standards (Fauconnot et al., 2004). Other analyses of CB were carried out via GLC by Ulberth
and Buchgraber, and Segall et al. using HPLC coupled to APCI-MS (Ulberth and Buchgraber,
2003; Segall et al., 2005). There ten already known TAGs were described in cocoa butter. Lísa
and Holcapek extended the list of the TAGs by identifying TAGs which possess longer chain
fatty acid in their structure, such as lignoceric and behenic acid to a total of 25 compounds
(Lísa and Holcapek, 2008). However, no reports have been published to date on TAGs in
unfermented cocoa beans.
We selected raw cocoa beans from six different origins, representative of beans of different
botanical origin, including beans from Ivory Coast, Tanzania, Ecuador, Brazil, Malaysia and
Indonesia. The total lipids were extracted using an optimized Soxhlet extraction with
dichloromethane as a solvent. It should be noted that most cocoa lipid extraction methods use
hexane or similar resulting in an incomplete extraction, omitting polar lipids. Gravimetric data
of total lipids for each sample are given in Table 4.1. For TAG analysis, we developed an
optimized reversed-phase HPLC method using ESI-mass spectrometry for detection. Ethanol
proved to be an efficient eluent for the good separation of lipids from a reversed C18 stationary
phase. TAG ionization was facilitated by the addition of ammonium formate. A gradient
comprising ethanol and acetonitrile was shown to elute TAGs at earlier retention times if
compared to methanol (at 100%) having a positive effect on the separation efficiency, which
has been described by other authors (Ribeiro et al., 2004). A representative chromatogram is
shown in Figure 4.1 (a). TAGs were eluting at retention times from 4.5 min to 48 min. Further
minor polar lipid species were observed at earlier retention times, which were neglected in this
contribution. For TAG nomenclature we use the characteristic shorthand defining the three
fatty acid substituents by a single letter/abbreviation of three capital letters code (for example
POS for 1-palmytoyl-2-oleoyl-3-stearoyl-glycerol) for the most common fatty acids or a
universal fatty acid code for the less abundant fatty acids in nature (Table 4.2) (Lísa &
Holcapek, 2008).
86
Chapter 4
Table 4.1 Origin data and total lipid amount of raw cocoa samples used in this study.
No. Hybrid name Geographical origin Total lipid amounta* (%)
1 C308XC5 Ivory Coast 41.09
2 C308XC5b* Ivory Coast 40.03
3 G3UTA409XC1 Ivory Coast 37.34
4 C38XC5 Ivory Coast 37.26
5 UPA405XC412 Ivory Coast 37.35
6 ICS60 Indonesia 31.58
7 ICCRI 04 Indonesia 25.61
8 BL2936 Indonesia 30.68
9 PA191 Indonesia 35.15
10 PBS123 Indonesia 30.54
11 ICCRI 03 Indonesia 26.38
12 ICCRI 06H Indonesia 30.79
13 EET544 Ecuador 32.62
14 EET103 Ecuador 35.47
15 CCN51 Ecuador 33.27
16 CCN51b* Ecuador 36.13
17 PBC140 Malaysia 34.10
20 Comun Brazil 33.55
21 unknown Tanzania 38.51
Hybrid name was assigned according to the terminology used by local suppliers.
a*calculated as % of mass ± 1% from raw wet cocoa beans. The moisture content of these samples diverges between 30
and 43%.
b* collected from a different location.
4.3.1 High resolution LC-ESI-MS and Molecular Formulae assignment
Compound identification was carried out using a high-resolution ESI-TOF mass

spectrometer operating in the positive ion mode leading to assignment of molecular formulae.
Molecular formulae were accepted if the mass error was found to be below 5 ppm. Using our
LC-ESI-MS method for TAG identification we observed that for most TAGs two or three
pseudomolecular ions were present, including [M+H]+, [M+NH4]+, and [M+Na]+ with
[M+NH4]+, as the most common base peak. However, not all of TAGs show [M+NH4]+ as main
adduct ion. TAGs at lower retention time have shown to have an intense [M+Na]+ adduct and
87
Chapter 4
the pseudomolecular [M+H]+ in a higher abundance if compared to [M+NH4]+. Byrdawell et

al. (Byrdwell and Neff, 2002) have described this phenomenon with both ESI and APCI
ionization. The relative ratio of these adduct ions varied depending on the structure and the
abundance of the TAGs.
Table 4.2 Abbreviations of fatty acids with the characteristic fragment ion acyls and neutral losses.
Common name Mass Symbol CN:DB [RCO]+ [RCOONH4]+
Myristic acid 228.2089 M C14:0 211.2067 245.2338

Pentadecylic acid 242.2246 C15:0 C15:0 225.2224 259.2495
Palmitic acid 256.2402 P C16:0 239.2380 273.2651
Palmitoleic acid 254.2246 Po C16:1 237.2224 271.2495
Margaric acid 270.2559 Ma C17:0 253.2537 287.2808
Stearic acid 284.2715 S C18:0 267.2693 301.2964
Oleic acid 282.2559 O C18:1 265.2537 299.2808
Linoleic acid 280.2402 L C18:2 263.2380 297.2651
Linolenic acid 278.2246 Ln C18:3 261.2224 295.2495
Unusual 1 fatty acid 298.2351 C18:1-OH C18:1 281.2481 315.2773
Unusual 2 fatty acid 296.2351 C18:2-OH C18:2 279.2324 313.2616
Unusual 3 fatty acid 312.2301 C18:2-(OH)2 C18:2 295.2274 329.2566
Nonadecylic acid 298.2872 C19:0 C19:0 281.2845 315.3138
Arachidic acid 312.3028 A C20:0 295.3006 329.3277
Heneicosanoic acid 309.3163 C21:0 C21:0 325.3112 343.3434
Behenic acid 340.3341 B C22:0 323.3319 357.3590
Tricosanoic acid 337.3476 C23:0 C23:0 353.3425 371.3747
Lignoceric acid 368.3654 Lg C24:0 351.3632 385.3903
88
Chapter 4
Therefore, within the mass spectra corresponding to the individual chromatographic peaks
molecular formulae for two to three ions needed to be evaluated complicating the overall
analysis of the TAG profile. Additionally, database searches using the lipidmaps database (Sud
et al., 2012) were used to compare experimental m/z values with database entries for further
confirmation.
Figure 4.1 (a) HPLC-MS profile of Ivory Coast raw unfermented cocoa TAGs in positive ionization
mode; (b) Two-dimensional (2D) map of TAGs from raw cocoa beans, Ivory Coast origin, by
reversed-phase high resolution HPLC/ESI-MicroOTOF MS analysis. Colour rapresents the intensity
of the peaks with red being the highest and blue the lowest: (c) 2D HPLC-MS feature map displaying
TAGs building blocks of homologous series.
89
Chapter 4
LC-MS data for CB containing signals originating from TAGs can be displayed as a 2D map
with retention times on the x-axis and m/z values on the y-axis (Fig. 4.1(b)). This representation
provides an improved overview revealing all molecular species detected in a complex sample,
in particular showing minor components co-eluting with major components. These plots have
been previously used for phospholipid assignment (Taguchi et al., 2000; Yamada et al., 2013).
Here we used it the first time for TAG identification in plant material. On the plot each TAG
species appear as two or three peaks (peak size in the third dimension corresponds to peak
intensity) with mass differences of 18 Da, corresponding to [M+H]+ and [M+NH4]+ and 23 Da
corresponding to [M+Na]+ at identical retention times. Series of TAGs lie within the plot on
straight lines with a positive slope within an almost equidistant period as indicated in Figure
4.1(c). These series generate TAG homologues series. Homologous series are series of
compounds, which differ in a defined mass increment corresponding to addition or subtraction
of a fixed moiety and hence m/z increment, eg. –H2 for unsaturation of FAs or +C2H4 for chain
elongation of fatty acids. Therefore the information obtained allows identification of
homologous series of fatty acids within the data set similar to the Kendrick formalism
frequently used in petroleum analysis or processed food analysis (Hughey et al., 2001).
The 2D map allows for the assignment of two types of homologous series (Fig. 4.1(c)), the
most common increments of 28 Da corresponding to +C2H4 and a second increment of 26 Da
corresponding +C2H2. The retention time of TAGs is governed by ECN. Retention time
typically increases with ECN number, whereas an increase in unsaturation (measured by double
bond equivalents DB) leads to decrease in retention time. This trend can be clearly seen in the
two dimensional plot. The separation of most TAGs within one ECN group was feasible and
readily predicted by the elution regularity also described by Holčapek et al. (Holčapek et al.,
2003). The molecular formula for a total of 83 TAGs were assigned, 58 of which were not
previously reported in the literature. Table 4.3 shows high resolution MS data and molecular
formulae assignment of all cocoa TAGs identified in Ivory Coast raw unfermented cocoa beans,
the sample showing the highest diversity.
4.3.2 Tandem mass spectrometry
Assignment of individual TAG structures was based on ESI-tandem MS (MSn)

measurements. For tandem mass spectrometry an Ion-Trap mass spectrometer was used,
90
Chapter 4
91
Chapter 4
92
Chapter 4
93
Chapter 4
operating in the positive ion mode using an auto MSn routine. The mechanism of TAG
fragmentation of ammonia adducts of TAGs was first suggested by McAnoy et al. (McAnoy et
al., 2005). Collision induced dissociation of [M+NH4]+ precursor ions resulted in the neutral
loss of NH3 and an acyl side-chain (lost as a carboxylic acid) to generate a diacyl product ion
(Scheme B.1 Appendix). Loss of the first carboxylic acid moiety can occur from the sn-1/3
position or the sn-2 position of the glycerol backbone. The mechanism of fragmentation would
involve a 1,2 acyl migration to produce a five membered oxonium ion (Hsu and Turk, 2010).
In tandem MS the first acyl substituent is believed to be lost from the sn-2 position, therefore
allowing unambiguous assignment of TAG regiochemistry. This assignment requires in our
view a note of caution. It is applied throughout the literature, however based on intuition rather
than on a sufficient set of authentic standards. Therefore, we like to add a further argument.
Migration from the sn-2 to the sn-1/3 position would in terms of organic chemistry be defined
as a “down-hill migration” from a more substituted carbon to a less substituted carbon, whereas
loss of any of the sn-2 acyl group would result in “up-hill migration” towards the sn-1/3 carbon
atoms (Eames et al., 1999).
MS/MS data were used to identify each acyl group present for a given [M+NH4]+ precursor
ion, and this information could be combined with high resolution MS data to identify possible
TAG molecular species present in cocoa bean extracts. Subsequent MS3 experiments on the
resultant diacyl product ions, which gave rise to acylium (RCO+) and related ions, enabled
unambiguous TAG molecular assignments. MS3 experiments using the diacyl product ion as
precursor ion allowed further assignment of the two remaining TAG fatty acid substituents.
Assignment of TAG structures with tandem MS data is given in the supplementary
information. In this section, two representative examples are discussed in more detail. A
compound with a pseudomolecular ion of m/z 881 (C55H110NO6, [M+NH4]+) as precursor ion
yielded fragment ions at m/z 607.6 (C39H74O6, [M–P–NH4]+), 41.4% relative intensity, and
579.6 (C37H70O6, [M–S–NH4]+), 100% relative intensity, corresponding to 1,3-distearoyl-sn-
glycerol (SS) and 1-stearoyl-2-palmytoyl-glycerol/2-palmytoyl-3-stearoyl-sn-glycerol (SP)
diacyl product ions. SP m/z 579.6 was further fragmented and acylium ions of stearic
[C17H35CO]+ m/z 276.2 and palmitic acid [C15H31CO]+ m/z 239.1 could be observed. Hence,
on the basis of MS2 and MS3 fragments and their abundance, the overall structure was assigned
as 1-stearoyl-2-palmytoyl-3-stearoyl-glycerol SPS, TG(18:0/16:0/18:0). A second example
comprises the pseudomolecular ion at m/z 919 (C58H112NO6, [M+NH4]+) as precursor ion,
which yielded two diacyl product fragment ions at m/z ions at m/z 619.6 (C40H74O6, [M–O–
94
Chapter 4
NH4]+), 100% relative intensity, and 605.7 (C39H70O6, [M–C19:0–NH4]+), 41.7% relative
intensity, corresponding to 1-oleoyl-2-nonadecanoyl-sn-glycerol (OC19:0) and 1,2-dioleoyl-
sn-glycerol (OO) diacyl product ions. Further fragmentation led to MS3 acylium ions of oleic
[C17H33CO]+ m/z 265.1 and nonadecylic acid [C19H39CO]+. Therefore, the structure was
assigned as C19:0OO, TAG (19:0/18:1/18:1). Figure 4.2 (a, b, c, d) illustrates the typical tandem
mass spectra of SPS and of 3 other different TAGs of increasing unsaturation fatty acid chain
in position 2 of the glycerol backbone. Interestingly we observed TAGs with fatty acid
substituents carrying an odd number of carbon atoms from pentadecylic acid C 15H30O2 to
pentacosylic acid C25H50O2. Such derivatives have been reported in other plants (Řezanka and
Sigler, 2009), however, they are relatively rare.
As an additional tool for facilitating compound assignment, we used neutral loss analysis.
The first fragmentation step in TAG fragmentation involves loss of the sn-1/3 acyl group
followed by loss of the sn-2 acyl group. Hence, neutral loss chromatograms with defined
neutral losses corresponding to fatty acids could be created. Such neutral loss chromatograms
show all compound displaying loss of a defined acyl group. Figure 4.2 (e, f, g, h, i) shows
exemplary neutral loss chromatograms with neutral loss of 273 Da (-C16H37O2N), 301 Da (-
C18H41O2N), 299 Da (-C18H39O2N), 297 Da (-C18H37O2N) and 295 Da (-C18H35O2N)
corresponding to ammonia adducts of palmitic acid, stearic acid, oleic acid, linoleic acid and
linolenic acid (Table 4.1) (Murphy et al., 2007). Because one of TAG was efficiently identified,
another TAG is more easily recognized from the molecular related mass values with the help
of the assembled information from the retention time point, MS2 fragments, and neutral loss
scanning of the corresponding fatty acid moiety. Combined LC-tandem MS and neutral loss
scanning have been applied to the assignment of other lipid classes so far (Byrdwell and Neff,
2002; Taguchi et al., 2005). Furthermore, the fragmentation pattern of some available standards
was cautiously studied and compared with cocoa TAGs. Their spectra are displayed in
supplementary material. We consider that the above-described approach for the identification
of TAGs will be a useful tool for the identification of TAG in any plant material, especially
seeds lipids.
95
Chapter 4
(a) SPS (m/z 881.9), (b) POP (m/z 851.9), (c) PLP (m/z 848.8) and (d) PLnP (m/z 829.7) ammonium
adduct ions. P, palmitic; O, oleic; S, stearic; L, linoleic; Ln, linolenic and neutral loss chromatograms
of (e) 273 Da, (f) 301 Da, (g) 299 Da, (h) 297 Da and (i) 295 Da (from up to down) corresponding to
ammonia adducts of palmitic acid, stearic acid, oleic acid, linoleic acid and linolenic acid in Ivory
Coast cocoa lipid extract.
4.3.3 Characterization of TAGs with hydroxylated fatty acids substituents
Next, to the TAGs containing saturated and unsaturated fatty acids, a second class of unusual
TAGs was detected in unfermented beans. These compounds eluted at earlier retention times
(from 4 to 25 min) indicating a more polar character. Assignment of molecular formulae from
high resolution mass data indicated the presence of one, two or three additional oxygen atoms
in the TAGs: C55H112NO7, C55H110NO8 and C57H108NO8. The compounds appeared as three sets
of homologues series, each with three to four members, with mass increments of C 2H4
corresponding to +28 Da. Taking into account the TAG abundancy, three main subclasses of
these unusual TAGs could be observed (Fig. 4.3(a)-(f)).
The observed molecular formulae suggest compounds with either an epoxide functionality,
as reported as oxidation products of oleic acid, or an unsaturated fatty acid carrying an
additional alcoholic moiety. The tandem MS data of these polar lipids revealed in MS2 a loss
96
Chapter 4
of 18 Da and 36 Da corresponding to the loss of respectively one or two molecule of H 2O

arguing in favour of OH-derivatives. In a second step both in MS2 and MS3 a neutral loss -276
Da, -278 Da and -280 Da was observed corresponding to a C18 fatty acid with respectively four,
three and two double bonds such as linolenic acid C18:2 (-C18H32O2). This finding suggests that
hydroxylation occurs at an unsaturated fatty acid producing a more polyunsaturated fatty acid
following fragmentation. For example, the compound assigned as PC18:1-OHPS,
pseudomolecular ion of m/z 895 (C55H108NO7, [M+NH4]+) as precursor ion, in which neutral
losses of -273 Da, -301 Da and -280 Da corresponding to stearic acid, palmitic acid, and
linoleic acid were observed defining all three fatty side chains of this polar TAG. The MS 2
spectra of [M+NH4]+ and [M+Na]+ of one representative compound of each class is presented
in Figure 4.3(a)-(f). [M+Na]+ were found to be extremely useful for a better understanding of
the hydroxyl fatty acids moieties for the reason that during MS2 fragmentation these fatty acid
do not lose their hydroxyl group and the value of the neutral loss matches to the absolute m/z
of the corresponding fatty acid. In case of Figure 4.3(b), (d) and (f), it corresponds to palmitic
acid, m/z of 256.
The elemental composition suggests the presence of oxygen within one of the fatty acid side
chains. To answer the question of the nature of oxygen functionality an extract enriched in this
TAG fraction was prepared using preparative column chromatography on silica gel, and NMR
spectra were recorded as well as a derivatization reaction was carried out using benzoyl
chloride.
The 1H-NMR spectrum shows four olefinic signals, two of them at 5.33 and 5.25 ppm
corresponding to a cis-olefinic moiety, probably in (N-9) oleic acid like environment. Two
further signals at 6.46 and 5.63 ppm show cross peaks in the 1H1H-COSY spectrum and an
additional cross peak to a triplet at 5.96 ppm revealing the presence of an allylic alcohol
functionality. They must be assigned to olefinic protons with a trans double bond geometry
due to a large 3J coupling of 16.0 Hz. In the 1H13C-HSQC spectrum, the allylic alcohol CHOH
carbon can be assigned to a signal at 73.1 ppm. The 1H13C-HMBC spectrum reveals from the
CHOH hydrogen signal a 2JHCC interaction to the olefinic carbon at 129.1 ppm and two further
interaction to a CH2 carbon at 30.3 ppm as well as a 3JHCCC interaction to the sn-2 carbonyl
carbon at 173.2 ppm. These interactions clearly place the OH substituent in the β-position of
the carbonyl. A characteristic pair of highfield-shifted carbons as expected for an epoxide are
clearly absent (Xia et al., 2016). Taking into account the tandem Ms data placing the hydroxy-
97
Chapter 4
fatty acid substituent into the sn-2 position we propose the structures shown in Figure 4.3. All
NMR data and key 2D NMR correlations are given in the supplementary material.
(a) PC18:2-OHP (m/z 865.0), (c) PC18:1-OHP (m/z 867.0) ammonium adduct ions and (b) PC18:2-
OHP (m/z 870.0) and (d) PC18:1-OHP (m/z 872.0) sodium adduct ions. (e) and (f) proposed possible
structures of the unusual TAGs in raw cocoa beans.
Following derivatization with benzoyl chloride the intensity of LC-MS signals of all polar
TAGs were significantly reduced, indicating high yielding derivatization. Additional peaks
were observed. In particular, a compound with MS data reminiscent of PLnP was observed with
a fatty acid with three double bonds in the sn-2 position, probably obtained by facile
elimination from the allylic alcohol functionality by β-elimination. EICs signals corresponding
to compounds with m/z values of 971.7, 999.7 or 1027.8 could be observed. Tandem MS data
reveal neutral losses corresponding to loss of 139 Da (-C7H6O2 + NH4+) from the precursor
ions. These findings indicate the presence of benzoyl derivatives, but further tandem MS data
are inconclusive. Chromatographic, MS data and 2D map for the derivatives are provided in
the supplementary material.
In nature hydroxylated fatty acids occur prominently in castor oil with the TAG containing
ricinoleic and dihydroxystearic acid (Lin et al., 2003). We have used castor oil as surrogate
standard (Clifford and Madala, 2017) in a direct comparison with raw cocoa bean lipid extract.
LC-MS data showed that none of the compounds in castor oil were identified in cocoa.
However, contrary to Lin et al. (Lin et al., 2003) assignment as SLR (1-stearoyl-2-linoleoyl-3-
ricinoleoyl-glycerol) TAG, we could suggest from tandem MS data (fragmentation of [M+Na+]
98
Chapter 4
spectra shown in supplementary material) of molecular species with m/z 924 a sn-2 position
for the ricinoleic acid therefore revising the initially suggested structures. We could affirm that
castor oil contain SRL and not SLR TAG moiety, eluting at 0.6 min earlier than SC 18:1(-OH)O
(C57H110NO7) TAG moiety, same m/z ratio, in cocoa but having a different fragmentation
pattern.
Oxygenated FAs, containing hydroxyl, epoxy, or oxo groups are an uncommon component
of seed oils, but have widespread distribution among higher plant families (Smith et al., 2014).
Studies confirm that hydroxy fatty acids are plant self-defense substances (Hou and III, 2000).
In addition, hydroxylated fatty acids have been suggested as precursors in jasmonic acid
biosynthesis, a plant hormone involved in defence reaction that has been observed in fermented
cocoa beans in high abundance (Patras et al., 2014).
When comparing different origins of beans, the large majority of the TAGs was found in all
samples. Indeed, unusual TAGs were found in all origin samples analyzed. However, there
were subtle differences in relative concentrations when comparing samples within the raw
unfermented beans. A detailed analysis of such differences is outside the scope of this
contribution, and accurate quantification will be reported elsewhere.
4.4 Conclusions
A new method for TAG analysis using reversed-phase HPLC coupled to ESI mass
spectrometry was developed, which allowed for the identification not only of the most
abundant TAGs but also of the low abundant TAGs showing weak signals buried into the noise.
Lipid extracts from raw cocoa beans from six different origins were profiled for the very first
time for their TAG content. A total of 83 TAGs could be characterized in unfermented cocoa
beans, 77 of which were positively identified. 58 of these compounds are reported for the first
time in cocoa. Among the new identified compounds, TAGs carrying a hydroxyl side chain
were characterized. The reason why these compounds were not reported before may lay on the
fact that these compounds possibly undergo transformation or degradation during cocoa
production process. These unusual constituents may offer the promise to blend cocoa butter
with unfermented cocoa lipids to produce CB with special physical and chemical properties in
new dietary or cosmetic products. However, further research studies needed to be carried out
to validate this possibility and to elucidate their function.
99
Chapter 4
References
Byrdwell, W.C., and Neff, W.E. (2002). Dual parallel electrospray ionization and atmospheric pressure chemical
ionization mass spectrometry (MS), MS/MS and MS/MS/MS for the analysis of triacylglycerols and
triacylglycerol oxidation products. Rapid Commun. Mass Spectrom. RCM 16, 300–319.
Carrasco-Pancorbo, A., Navas-Iglesias, N., and Cuadros-Rodríguez, L. (2009). From lipid analysis towards
lipidomics, a new challenge for the analytical chemistry of the 21st century. Part I: Modern lipid analysis. TrAC
Trends Anal. Chem. 28, 263–278.
Christie, W.W., and Han, X. (2010). Lipid Analysis, Fourth Edition: Isolation, Separation, Identification and
Lipidomic Analysis (Bridgwater, England: Oily Press).
Clifford, M.N., and Madala, N.E. (2017). Surrogate Standards: A Cost-Effective Strategy for Identification of
Phytochemicals. J. Agric. Food Chem. 65, 3589–3590.
Cvacka, J., Krafková, E., Jiros, P., and Valterová, I. (2006). Computer-assisted interpretation of atmospheric
pressure chemical ionization mass spectra of triacylglycerols. Rapid Commun. Mass Spectrom. RCM 20, 3586–
3594.
Eames, J., Kuhnert, N., Sansbury, F.H., and Warren, S. (1999). Kinetic and Thermodynamic Control in the
Synthesis of Tetrahydro-Pyrans and -Furans from 1,4-Diols by Stereospecific Phenylsulfanyl (PhS) Migration:
Competition Between exo and endo Transition States and between [1,2] and [1,4]Sulfanyl Participation. Synlett
1999, 1211–1214.
Fauconnot, L., Hau, J., Aeschlimann, J.-M., Fay, L.-B., and Dionisi, F. (2004). Quantitative analysis of
triacylglycerol regioisomers in fats and oils using reversed-phase high-performance liquid chromatography and
atmospheric pressure chemical ionization mass spectrometry. Rapid Commun. Mass Spectrom. RCM 18, 218–
224.
Holčapek, M., Jandera, P., Zderadička, P., and Hrubá, L. (2003). Characterization of triacylglycerol and
diacylglycerol composition of plant oils using high-performance liquid chromatography–atmospheric pressure
chemical ionization mass spectrometry. J. Chromatogr. A 1010, 195–215.
Hou, C.T., and III, R.J.F. (2000). Growth inhibition of plant pathogenic fungi by hydroxy fatty acids. J. Ind.
Microbiol. Biotechnol. 24, 275–276.
Hsu, F.-F., and Turk, J. (2010). Electrospray ionization multiple-stage linear ion-trap mass spectrometry for
structural elucidation of triacylglycerols: assignment of fatty acyl groups on the glycerol backbone and location
of double bonds. J. Am. Soc. Mass Spectrom. 21, 657–669.
100
Chapter 4
Hughey, C.A., Hendrickson, C.L., Rodgers, R.P., Marshall, A.G., and Qian, K. (2001). Kendrick Mass Defect
Spectrum:  A Compact Visual Analysis for Ultrahigh-Resolution Broadband Mass Spectra. Anal. Chem. 73, 4676–
4681.
Jakab, A., Héberger, K., and Forgács, E. (2002). Comparative analysis of different plant oils by high-performance
liquid chromatography–atmospheric pressure chemical ionization mass spectrometry. J. Chromatogr. A 976, 255–
263.
Kuhnert, N., Dairpoosh, F., Yassin, G., Golon, A., and Jaiswal, R. (2013). What is under the hump? Mass
spectrometry based analysis of complex mixtures in processed food--lessons from the characterisation of black
tea thearubigins, coffee melanoidines and caramel. Food Funct. 4, 1130–1147.
Lin, J.-T., Turner, C., Liao, L.P., and McKeon, T.A. (2003). Identification and Quantification of the Molecular
Species of Acylglycerols in Castor Oil by HPLC Using ELSD. J. Liq. Chromatogr. Relat. Technol. 26, 773–780.
Lipp, M., and Anklam, E. (1998a). Review of cocoa butter and alternative fats for use in chocolate—Part A.
Compositional data. Food Chem. 62, 73–97.
Lipp, M., and Anklam, E. (1998b). Review of cocoa butter and alternative fats for use in chocolate—Part B.
Analytical approaches for identification and determination. Food Chem. 62, 99–108.
Lipp, M., Simoneau, C., Ulberth, F., Anklam, E., Crews, C., Brereton, P., de Greyt, W., Schwack, W., and
Wiedmaier, C. (2001). Composition of Genuine Cocoa Butter and Cocoa Butter Equivalents. J. Food Compos.
Anal. 14, 399–408.
Lísa, M., and Holcapek, M. (2008). Triacylglycerols profiling in plant oils important in food industry,
dietetics and cosmetics using high-performance liquid chromatography-atmospheric pressure chemical
ionization mass spectrometry. J. Chromatogr. A 1198–1199, 115–130.
McAnoy, A.M., Wu, C.C., and Murphy, R.C. (2005). Direct qualitative analysis of triacylglycerols by electrospray
mass spectrometry using a linear ion trap. J. Am. Soc. Mass Spectrom. 16, 1498–1509.
Milev, B.P., Patras, M.A., Dittmar, T., Vrancken, G., and Kuhnert, N. (2014). Fourier transform ion cyclotron
resonance mass spectrometrical analysis of raw fermented cocoa beans of Cameroon and Ivory Coast origin. Food
Res. Int. 64, 958–961.
Murphy, D.J. (1994). Biogenesis, function and biotechnology of plant storage lipids. Prog. Lipid Res. 33, 71–85.
Murphy, R.C., James, P.F., McAnoy, A.M., Krank, J., Duchoslav, E., and Barkley, R.M. (2007). Detection of the
abundance of diacylglycerol and triacylglycerol molecular species in cells using neutral loss mass spectrometry.
Anal. Biochem. 366, 59–70.
Patras, M.A., Milev, B.P., Vrancken, G., and Kuhnert, N. (2014). Identification of novel cocoa flavonoids from
raw fermented cocoa beans by HPLC–MSn. Food Res. Int. 63, 353–359.
101
Chapter 4
Podlaha, O., Toregard, B., and Puschl, B. (1984). TG-type composition of 28 cocoa butters and correlation
between some TG-type components. Lebensm. - Wiss. Technol. Food Sci. Technol.
Řezanka, T., and Mareš, P. (1991). Determination of plant triacylglycerols using capillary gas chromatography,
high-performance liquid chromatography and mass spectrometry. J. Chromatogr. A 542, 145–159.
Řezanka, T., and Sigler, K. (2009). Odd-numbered very-long-chain fatty acids from the microbial, animal and
plant kingdoms. Prog. Lipid Res. 48, 206–238.
Ribeiro, R.L.V., Bottoli, C.B.G., Collins, K.E., and Collins, C.H. (2004). Reevaluation of ethanol as organic
modifier for use in HPLS-RP mobile phases. J. Braz. Chem. Soc. 15, 300–306.
Segall, S.D., Artz, W.E., Raslan, D.S., Ferraz, V.P., and Takahashi, J.A. (2005). Analysis of triacylglycerol isomers
in Malaysian cocoa butter using HPLC–mass spectrometry. Food Res. Int. 38, 167–174.
Simoneau, C., Hannaert, P., and Anklam, E. (1999). Detection and quantification of cocoa butter equivalents in
chocolate model systems: analysis of triglyceride profiles by high resolution GC. Food Chem. 65, 111–116.
Smith, M.A., Zhang, H., and Purves, R.W. (2014). Identification and Distribution of Oxygenated Fatty Acids in
Plantago Seed Lipids. J. Am. Oil Chem. Soc. 91, 1313–1322.
Sud, M., Fahy, E., and Subramaniam, S. (2012). Template-based combinatorial enumeration of virtual compound
libraries for lipids. J. Cheminformatics 4, 23.
Taguchi, R., Hayakawa, J., Takeuchi, Y., and Ishida, M. (2000). Two-dimensional analysis of phospholipids by
capillary liquid chromatography/electrospray ionization mass spectrometry. J. Mass Spectrom. 35, 953–966.
Taguchi, R., Houjou, T., Nakanishi, H., Yamazaki, T., Ishida, M., Imagawa, M., and Shimizu, T. (2005). Focused
lipidomics by tandem mass spectrometry. J. Chromatogr. B Analyt. Technol. Biomed. Life. Sci. 823, 26–36.
Takagi, T., and Ando, Y. (1995). Stereospecific analysis of monounsaturated triacylglycerols in cocoa butter. J.
Am. Oil Chem. Soc. 72, 1203–1206.
Ulberth, F., and Buchgraber, M. (2003). Analytical platforms to assess the authenticity of cocoa butter. Eur. J.
Lipid Sci. Technol. 105, 32–42.
Velíšek, J., and Cejpek, K. (2006). Biosynthesis of food constituents: Lipids. 1. Fatty acids and derived compounds
- A review. Czech J. Food Sci. 24, 193–216.
Visioli, F., Bernaert, H., Corti, R., Ferri, C., Heptinstall, S., Molinari, E., Poli, A., Serafini, M., Smit, H.J., Vinson,
J.A., et al. (2009). Chocolate, Lifestyle, and Health. Crit. Rev. Food Sci. Nutr. 49, 299–312.
102
Chapter 4
Xia, W., Budge, S.M., and Lumsden, M.D. (2016). 1H-NMR Characterization of Epoxides Derived from
Polyunsaturated Fatty Acids. J. Am. Oil Chem. Soc. 93, 467–478.
Yamada, T., Uchikata, T., Sakamoto, S., Yokoi, Y., Fukusaki, E., and Bamba, T. (2013). Development of a lipid
profiling system using reversed-phase liquid chromatography coupled to high-resolution mass spectrometry with
rapid polarity switching and an automated lipid identification software. J. Chromatogr. A 1292, 211–218.
Young, A.M. (2007). The Chocolate Tree: A Natural History of Cacao (Gainesville: University Press of Florida).
103
Chapter 5
Chapter 5:
5 Variation of triacylglycerol profiles in unfermented

and dried fermented beans
Diana Sirbua, Anne Grimbsa, Marcello Cornob, Matthias S. Ullrich a, and Nikolai Kuhnerta*
a
Department of Life Sciences and Chemistry, Jacobs University, Campus Ring 1, D-28759 Bremen,
Germany
b
Barry Callebaut, Aalstersestraat 122, B-9280 Lebbeke-Wieze, Belgium
*Correspondence to:
Phone: +49 421 200 3120,
Fax: +49 421 200 3102.
Submitted to Food Research International Journal. Status: accepted.
Supplementary material available with the published article.
104
Chapter 5
Abstract
Fermentation and drying are the two crucial processing steps required to produce cocoa
beans with desired properties, especially taste and flavor. To understand their impact on the
lipid profile of cocoa, the lipid composition of unfermented raw and fermented dried beans
from six different origins was investigated using high-performance liquid chromatography-
mass spectrometry methods. While the comparison of triacylglycerol profiles across the
different origins showed only small variations in individual compound concentrations, the
comparison along the fermentation status showed major differences regarding the occurrence
of polar lipids. These compounds may serve as biomarkers for the fermentation status of the
beans and a simple analytical method suitable for field trials is proposed. Finally, a hypothesis
identifying key unsaturated triacylglycerols contributing to the hardness and softness of cocoa
butter is presented.
Keywords: fermentation; drying; cocoa; origin; triacylglycerol; soft butter; lipid; profile.
105
Chapter 5
5.1 Introduction
The production and marketing of origin chocolate constitute one of the latest trends in the
cocoa industry, applying winemakers’ concept of terroir to chocolate. Therefore chocolate
manufactures spend significant effort on sourcing raw material with specific characteristics
form a particular country. Even that might be the trend, price, trade balances, politics, and the
volume of cocoa beans available determine the choice of bean origin for lower value cocoa
products. The main cocoa bean producing countries are African countries supplying 73.3% of
the global production in the last year, and the leading cocoa bean processing countries are
European countries grinding 38.7% of the 4.6 million tons of cocoa harvested in 2016 (“The
International Cocoa Organization (ICCO) | Cocoa Producing and Cocoa Consuming
Countries,”) The origin can have an impact for several reasons: first, different countries
cultivate botanically different subspecies of Theobroma cacao. There are four main cultivars
of cocoa: Criollo, Forastero, Trinitario, and Nacional (Lima, Almeida, Rob, & Zwietering,
2011; Rusconi & Conti, 2010). Secondly, the practice of producing dried beans, the first stage
of the complex production process of chocolate, cocoa powder and butter, can occur in different
ways. Cocoa beans are submitted to an extensive production chain, starting from harvesting
the fruits on farms and plantations followed by pod opening and bean removal from the pod.
Subsequently, beans are fermented and then dried to less than 8% (Badrie, Bekele, Sikora, &
Sikora, 2015; Beckett, Harding, & Freedman, 2008). Drying is essential, both for stopping the
fermentation process and reducing moisture thus minimizing microbial growth during transport
and storage. Climate conditions at the geographical areas of cultivation have been shown to
significantly influence bean chemical composition and quality (Chaiseri & Dimick, 1989;
Lehrian, Keeney, & Butler, 1980). Finally, different storage and shipping methods have their
contribution to the quality of the dried beans (Beckett, Fowler, & Ziegler, 2017).
Lipids represent approximately half of cocoa beans’ chemical constituents. As in many other
plant seed oils, triacylglycerols (TAGs) are the predominant species, defining the physical
characteristic of cocoa butter (CB). Despite the extensive work made on studying the formation
of lipid-related flavor precursors during the fermentation process (Afoakwa, Paterson, Fowler,
& Ryan, 2008; Kadow, Bohlmann, Phillips, & Lieberei, 2013) and ample progress on
understanding the chemistry of CB, there are few attempts on describing the TAGs in
unfermented (please note that the terms wet beans and raw beans are used as well) cocoa beans
106
Chapter 5
and changes of lipids occurring during fermentation (Liendo, Padilla, & Quintana, 1997; Lima
et al., 2011). Studies on CB classify cocoa butter as soft and hard butter on origin basis with
each butter used in dedicated product applications. It was previously suggested that soft CBs
are characterized by higher 1-palmitoyl-2,3-dioleoyl-glycerol (POO), 1-stearoyl-2,3-dioleoyl-
glycerol (SOO) content (Chaiseri & Dimick, 1989; Lipp & Anklam, 1998) whereas hard CBs
are characterized by increased saturated fatty acid content (Chin, 1989). Therefore, Brazilian
and Malaysian CBs are generally considered as extremes since Brazilian cocoa butter is rich
in di- and tri-unsaturated TAGs (di- and tri- are referred to the number of double bonds in the
fatty acid chain), while Malaysian cocoa butter was reported to contain a lower concentration
of polyunsaturated TAGs and a higher concentration of saturated TAGs (Figueira et al., 1997;
Marty & Marangoni, 2009).
For the description of chemical differences between cocoa beans from varying origin,
several different analytical techniques have been employed to profile the typical metabolite
classes in cocoa beans including polyphenol profiling, protein profiling, and lipid profiling
(Caligiani et al., 2014; Kumari et al., 2016; Marseglia et al., 2016; D’Souza et al., 2017).
HPLC-MS provides most chemical information (Hernandez, Castellote, & Permanyer, 1991;
Lísa, Holčapek, & Boháč, 2009) to profile the chemical composition of CB including minor
lipids and species co-eluting with each other.
Previously we presented data concerning the TAGs diversity in wet unfermented cocoa
beans, identifying 83 TAG species in these beans. To achieve this, we have developed a suitable
analytical procedure, which employs HPLC coupled to mass spectrometry (MS) (Sirbu, Corno,
Ullrich, & Kuhnert, 2018). We now apply this method to assess the differences in the lipid
profile specific to geographic origin and the process status of the samples. In this manuscript,
we profile TAGs in unfermented wet beans and fermented dried beans from six different origins
including Brazil, Ecuador, Ivory Coast, Tanzania, Malaysia, and Indonesia. These countries
were chosen mostly based on their global cocoa production, as well as sourcing opportunities
and geographical diversity. Multivariate statistical analysis has been employed to confirm the
key compositional difference in TAG profiles when comparing unfermented and fermented
beans, as well as to visualize the chemical differences according to origins of the seeds
identifying in the process biomarkers allowing a clear distinction between sample groups. This
is the first study examining the TAG profile in unfermented raw and fermented dried cocoa
beans.
107
Chapter 5
(Rotisolv® HPLC grade) and Tetra-dodecylammonium bromide was purchased from Carl Roth
and formic acid (puriss., ≥ 98% (T) for mass spectrometry) were purchased from Sigma-Aldrich
Chemie (Steinheim, Germany). Ethanol was subjected to distillation prior use.
Frozen fresh unfermented cocoa beans samples and dried fermented cocoa beans samples
from 6 different origins were received in several sets from Barry Callebaut. A total of 24 pairs
of samples coming from Ivory Coast, Indonesia, Ecuador, Tanzania, Malaysia, and Brazil were
analyzed. Firstly, the seed material was deshelled and ground using a grinder with the purpose
of making homogenous powder. Henceforth, the samples were extracted employing an
overnight Soxhlet (Buchi Extraction System B811 instrument, Flawil, Switzerland) method
using dichloromethane as an extraction solvent, 150 mL and 5 g of powdered beans. Extracted
lipids were quantified gravimetrically after evaporation to dryness in a rotavapor. The dry
residue was then stored at -20 °C until further analysis. For HPLC analysis, a concentration of
0.045 mg/mL in chloroform/ethanol (50/50) of cocoa lipid extract was prepared.
5.2.3 Moisture analysis
Moisture content was determined by heating around 2.5 g of cocoa powder in the oven at
105 °C for 24 hours, cooled in the dry atmosphere of a desiccator, and then reweighed. Total
lipid content was normalized according to: = Cocoa Powder (g) − Weight Loss (g)
108
Chapter 5
5.2.4 TLC separation
Total lipid extracts were dissolved in chloroform/ethanol (50/50) at the concentration of 10

mg/mL and applied to the origin of a Silica Gel 60 HPTLC plate 20 cm x 20 cm (Macherey-
Nagel, München, Germany). The dried lipid spots were developed with hexane/diethyl
ether/glacial acetic acid (80:20:1, by vol.). Plates were allowed to dry and run another time
with chloroform/methanol/glacial acetic acid (97:3:1, by vol.). After a second complete
drying time, the plate was sprayed with Cerium Molybdate Stain (Hanessian's Stain) in order
to visualize the separated spots of the lipid extract (Christie & Han, 2010).
consisted of ethanol with 10 mM/L ammonium formate and 0.01% formic acid. The mobile
10 min. The column was equilibrated at 40/60 solvent A/B for 5 min before reuse. Tetra-
dodecyl ammonium bromide at the concentration of 0.001 mg/mL, mass-to-charge (m/z)
690.7850, and retention time (Rt) 4.5 min was used as internal standard.
109
Chapter 5
quadratic mode.
5.2.7 Statistical analysis
The individual TAG lists of the high-resolution masses were combined for peak matching
and retention time alignment using a m/z tolerance of Δm/z = 0.0001 and a Rt tolerance of 0.5
min for elution time <20 min and 1.0 min for an elution time ≥20 min. Extensive manual
curation resulted in a set of 126 peaks whereas for 74 peaks a TAG could be uniquely assigned
with m/z and Rt tolerances of Δm/z = 0.01 and ΔRt = 0.8 min. For further analyses, the peak
areas were normalized to the peak area of the internal standard (Tetra-dodecyl ammonium
bromide). Reproducibility data on the internal standard, are given in the supplementary
material (Appendix, Table C.1). For TAG nomenclature we use the common three-letter
shorthand, for example, POO, SOO, POL or POS to designate the three acyl substituent on the
glycerol moiety (P – palmitic acid, O - oleic acid, S - stearic acid, L linoleic acid). TAG
identification was based on high-resolution mass spectrometry establishing molecular
formulae followed by tandem mass spectrometry using a quadrupole ion trap mass
spectrometer. Fragment spectra allowed the definition of the three TAG acyl moieties as well
as their regiochemistry (Sirbu, Corno, Ullrich, & Kuhnert, 2018). To allow relative
quantification an internal standard was added to each sample and peak areas normalized with
respect to the internal standard.
Each one-way analysis of variance (ANOVA) was conducted for the normalized peak areas
of the 12 major TAGs providing the country of origin classification with a significance level
of p < 0.05. In case of significant effects for multiple comparisons, the ANOVAs are followed
by post-hoc pairwise T-tests with multiple comparisons using Benjamini-Hochberg procedure
(α = 0.1) (Benjamini & Hochberg, 1995, 2000).
The hierarchical clustering of the scaled normalized peak areas excluding the 12 TAGs was
calculated using averaged linkage method and the Hamming distance metric. The peak profiles
in the attached heat map were sorted column-wise in ascending retention time order. The herein
performed principal component analyses (PCA) were applying mean centering and Pareto-
scaling on the respective normalized data (van den Berg, Hoefsloot, Westerhuis, Smilde, & van
der Werf, 2006). For highlighting the data separation, colored two-standard deviational ellipses
were added for each group based on the mean standard deviation, and the covariance of the
110
Chapter 5
corresponding data. Additionally, symbol coloring and symbol shape facilitate the group
assignment concerning country of origin and fermentation status, respectively.
For the softness assessment, the relative normalized peak areas were used as cumulative
values and ratio, respectively. More precisely, the sum of particular TAGs was considered and
the ratio of TAGs with specific properties, i.e., poly-unsaturated, mono-unsaturated and
saturated ones.
For this study, a sample collection was chosen, comprising 24 pairs of unfermented and
fermented dry cocoa beans from six different countries of origin.
5.3.1 Total lipid content
Soxhlet extraction enabled us to obtain an extraction yield ranging from 49% to 63% of total
lipid content in cocoa beans. Results showed that fermentation and drying lead to an increase
of the lipid amount in the beans. Considerable variations in the total lipid content were observed
for several origins, while there were no significant differences when comparing countries (Fig.
5.1). However, African samples appear to have a slightly higher total lipid content in
comparison to South American and South Asian samples. These differences can be mostly
attributed to biological variabilities. Lehrian and Keeney estimate an increase between 4 and
20% of the fat during the last ten days of cocoa fruit ripening (Lehrian & Keeney, 1980).
Figure 5.1 Box plots of total lipid content determined gravimetrically following Soxhlet extraction
for (a) unfermented raw and (b) fermented dried cocoa beans from different countries of origin.
111
Chapter 5
Other factors such as the variety of cocoa, environmental conditions, and the harvesting and
post-harvesting conditions account for the differences observed (Beckett, Fowler, & Ziegler,
2017). Using a much larger sample set compared to these authors our study estimates the
increase of lipid content during fermentation to 3-6%.
5.3.2 Triacylglycerol profiles
Total lipid extracts were firstly investigated by thin layer chromatography (TLC). The TLC
showed a separation of cocoa lipids into five distinct fractions. Direct comparison of fermented
and unfermented beans revealed the absence of a spot at retention factor (Rf) of 5.5 in
fermented dry beans (Fig. 5.2 (a)). Subsequent use of ESI-HPLC (Fig. 5.2 (b)) coupled to a
TOF mass spectrometer has confirmed those differences with unfermented beans displaying a
variety of peaks between 10 and 22 min retention time, which are absent in dried fermented
beans. To appreciate the full complexity of the lipid profile we present 2D chromatograms
displaying retention time on the x-axis and m/z values on the y-axis (Fig. 5.2 (c) and (d)). 117
TAG compounds were detected in unfermented beans and as inferable from the 2D map and
around 70 TAGs in fermented samples. A full list of detected TAGs is provided along with their
MS relative quantification data in the supplementary information (Appendix, Tables C.2-C.4).
Further chromatograms and 2D graphs of samples for the other five origins to allow direct
comparison are presented in the supplementary material (Appendix, Fig. C1-C5). The relative
composition as percentage of total lipids measured as the sum over all LC-MS peak intensities
of the 25 most abundant TAGs in cocoa lipid extract from different geographical sources is
presented in Table 5.1. The relative quantification was carried out from extracted ion
chromatograms (EICs) by normalization with respect to the internal standard, assuming that
the detector response for all TAG-classes was identical. Our findings revealed that the amounts
for the three main TAGs were overestimated whereas values of minor TAGs were
underestimated if compared to literature data (Chaiseri & Dimick, 1995; Podlaha, Toregard, &
Puschl, 1984; Shukla, 2005). This fact is probably due to the limitations of previously
employed TAG detection by refractive index detectors, not accounting for co-eluting minor
TAG species.
As it can be deducted from data presented in Table 5.1, cocoa TAG profiles are fairly
variable across geographical origins. Figure 5.3 displays variations of key TAGs. Key TAGs
denote those TAGs that were considered to show crucial variations by other authors (Chaiseri
112
Chapter 5
& Dimick, 1989; Lipp et al., 2001; Podlaha et al., 1984; Shukla, 2005; Simoneau, Lipp,
Ulberth, & Anklam, 2000; Vieira, Efraim, Van, De, & Dewettinck, 2015). The relative
quantities of the key TAG species show variations of up to ± 20% if unfermented and dried
fermented bean pairs. For instance, 9.5% POP was observed in unfermented Malaysian beans
compared to 10.8% in dried fermented beans or 27.3% SOS in Ecuadorian unfermented beans
compared to 31.3% in the corresponding dried fermented beans. These difference seemed to
result from process variability, in terms of fermentation and drying, but no clear trend was
observed. Considerably higher variations of individual TAGs were observed comparing
samples from different origins. These variations extend up to 2.5 fold difference for certain
TAGs, for example, POP showed a maximum 16.1% and a minimum of 6.9%, whereas SOS
showed a maximum of 42.8% and a minimum of 22.7%. Besides the aspects of variability
discussed above, these large differences might cover a significant biological variability
concerning cocoa varieties and ripeness degree of cocoa pods. Despite what stated above, TAG
lipid profile of cocoa beans from some countries has shown significant differences.
Figure 5.2 Representative chromatography results for fermented dried and unfermented raw cocoa
lipid extract of Ivory Coast origin. (a) TLC plate image and the parallel corresponding (b) HPLC-MS
chromatograms in positive ion mode (P – palmitic acid, O - oleic acid, S – stearic acid, L – linoleic
acid, Ln – linolenic acid, Ma – margaric acid, A – arachidic acid). (c) Two-dimensional (2D) map of
reversed-phase high-resolution MS of unfermented raw and (d) fermented dried cocoa bean samples
from Ivory Coast. The colormap of the 2D map represents the peak intensity with red being the
highest and blue the lowest.
113
Chapter 5
114
Chapter 5
Figure 5.3 Relative abundance of the key 12 TAGs for unfermented and fermented cocoa beans
from different countries of origin. The sample labels denote the countries of origin, namely, B –
Brazil, E – Ecuador, I – Indonesia, M – Malaysia, T – Tanzania, and V – Ivory Coast, as well as the
fermentation status, F – fermented and U – unfermented. The gray painted bars illustrate the
remaining detected TAGs.
5.3.3 Comparison of Triacylglycerol profiles of different origins
To obtain a global overview of significant variations of TAGs within the sample set Principal
Component Analysis (PCA) of the LC-MS data was used as an unsupervised machine learning
method. Both score and loading plot based on 12 most relevant TAGs: 10 most abundant TAGs
plus SLnS and OOO having the same m/z ratio of SLO are shown in Figure 5.4.
No overall separation according to countries of origin can be observed in the score plots
(Fig. 5.4 (a)); however, pairs of countries on a given continent can be distinguished, for
example, Brazil versus Ecuador, Indonesia versus Malaysia, and Tanzania versus Ivory Coast.
Moreover, sample pairs of unfermented and dried fermented beans are located side by side,
indicating that total variance of the key 12 TAGs is not attributed to fermentation status but
115
Chapter 5
rather to the origin. Major contributors to that separation are the high-abundant TAGs POP,
POS, and SOS while the remaining TAGs seemed to have the same influence (Fig. 5.4 (b)).
Interestingly, the total variation of the entire peak profile is very similar to the 12 key TAGs
(Fig. 5.4, Fig. C.6 and Fig. C.7). These observations demonstrated that these key 12 TAGs
alone are sufficiently characterize the overall variations among samples.
Figure 5.4 PCA of the 12 most relevant TAGs for unfermented and fermented cocoa beans from
different countries of origin. The sample labeling and coloring in (a) the scores plot as well as the
coloring of the two-standard deviational ellipses illustrate the different countries of origin, namely, B
– Brazil, E – Ecuador, I – Indonesia, M – Malaysia, T – Tanzania, and V – Ivory Coast. In addition,
the sample label as well as the shape indicate the fermentation status, F – fermented and U –
unfermented. The labelling in (b) the loadings plot denotes the major TAGs.
Therefore, the differences of these 12 key TAGs for unfermented and fermented cocoa beans
were further examined across the different countries of origin and visualized using box plots
(see Fig. 5.5). Significant differences across the countries of origin for POP, POS, SOS, POO,
SLS, and SOA. Using ANOVA, p-values were determined for these 12 key TAGs comparing
pairwise samples from two origin countries. The results are visualized in Figure 5.5, lower-
right panel. The shading in the pairwise comparison indicates the difference of a given country
pair for the 12 TAGs, the brighter the more significant differences were found. Full data are
given in the supplementary information (Appendix, Tables C.5-C.10). Thus, dissimilarities as
for Ecuador and Brazil become clear as well as country couples which seemed to be more
similar. For instance, statistically significant variations of POP, POS SOS, and SLS are
116
Chapter 5
observed if Brazilian and Ivory Coast lipids are compared (bottom left corner) or statistically
significant variations of POP, POS SOS, and SLS for the pair Ecuador and Brazil. TAG profiles
from Tanzania and Indonesia, however, showed no statistically significant difference in any of
the TAG constituents, indicating resemblance between these two countries.
Figure 5.5 Box plots of the 12 TAGs for unfermented and fermented cocoa beans from different
countries of origin. The subplot label denotes the significance value (p) of the respective ANOVA.
The heat map on the lower right illustrates the pairwise T-tests with a name for each TAG, which is
significantly different for the respective country pair. The brightness of the heat map additionally
indicates the number of significantly different TAGs per country couple.
5.3.4 Correlation of triacylglycerol profiles with cocoa butter hardness
It is generally accepted that the softness of CB is affected by the ratio between the sum of
di- and polyunsaturated TAGs and the sum of monounsaturated and saturated TAGs (Beckett
et al., 2008; Chaiseri & Dimick, 1989, 1995; Ribeiro, Correa Basso, et al., 2012; Vieira, Efraim,
Van, De, & Dewettinck, 2015). The softness of butters was correlated to some parameters such
117
Chapter 5
as high iodine value and low solid fat content as well as increased POO and SOO content (Lipp
& Anklam, 1998; Marty & Marangoni, 2009) leading to unsatisfactory crystallization
characteristics (Chaiseri & Dimick, 1995; Ribeiro, Silva, et al., 2012). The main TAGs, namely
POP, POS, SOS are solid at room temperature whereas POO and SOO are more liquid at room
temperature. POO and SOO together with PLS and SLS are the second most abundant species
in cocoa lipid extract, with the last ones not always taken into account by other authors until
now (Table 5.1). Using our statistical analysis alternative biomarkers defining softness and
hardness of cocoa butter can be identified. Such statistical analysis comparing soft and hard
butters with each other reveal according to p-values variations between the ratios of POO/PLS
and SOO/SLS after the fermentation and drying of the beans as key markers. EICs illustrating
these variations are shown in Figure 5.6 (a) and (c). In Ivory Coast, for example, wet and dried
beans showed a different profile of these four TAGs. This phenomenon is also observed in
other countries but to a lower extent. In all the samples, the major change noted was the
increase in TAGs containing linoleic acid in particular PLS and SLS. However, across all
samples the ratios of POO/PLS and SOO/SLS exposed a strong correlation (Pearson
correlation coefficient of 0.967). Following this line of argumentation, it can be determined
that the first stage of production process affects all the members of the two homologs
compounds in combination (Fig. 5.6 (b)).
Figure 5.7 (a) illustrates the assessment of the softness according to the literature. The box
plots suggest a tendency of the distribution of the TAGs according to the distance to the equator,
hard butters are produced closer to the equator, in contrast, soft butters further away from the
equator as highlighted by other authors (Ribeiro et al., 2012; Tucci et al., 1996; Vieira et al.,
2015). The same authors describe that the softness of the butter increase according to the
average ambient temperature of the cultivation location. Only Ecuador does not follow this
trend showing a much larger variability. This variation can be rationalized by the presence of
two cocoa varieties Nacional and Forastero (manly CCN51 hybrid) and the diverse climate in
this region. Geographic coordinates (GPS data) and hybrid name for all samples are provided
in supplementary material (Appendix, Table C.11).
Taking into account the previous reports and the strong correlation between the ratios of
POO/PLS and SOO/SLS, we have summed up their contents. Bar plots representation of the
sum of the 4 TAGs (Fig. 5.7 (b)) shown to be an identical projection to the bar plots
representation of the inferred established way of softness estimation. Therefore, we suggest
using the sum of the quantities of these selected TAGs to predict trends of the softness of CB
118
Chapter 5
from any other source. Thus, a descriptive model of a reduced complexity is proposed. This
new indicator of softness relies only on the abundance/intensity of four specific TAGs and its
determination, accordingly, requires less expenditure. Nevertheless, a standard reference of the
four TAGs would be expected as means of comparison. Hence, a prediction regarding the
relative softness of any given samples could be done in a faster way routinely.
Figure 5.6 (a) Extract ion chromatogram of m/z 876.8015 corresponding to POO (left peak) and
PLS (right peak); (b) Scatter plot on the SLS:SOO ratio and the PLS:POO (of all the samples) and (c)
Extract ion chromatogram of m/z 904.8328 corresponding to SOO (left peak) and SLS (right peak) of
2 Malaysian samples, 2 Ivorian and 2 Ecuadorian samples.
Figure 5.7 (a) Box plots of softness indicators for unfermented and fermented cocoa beans from
different countries of origin. As softness indicator serve the established (a) ratio of polyunsaturated
and monounsaturated plus saturated TAGs and (b) the newly conceived cumulative values of POO,
PLS, SOO and SLS.
119
Chapter 5
5.3.5 Comparison of triacylglycerol format in unfermented and fermented dried beans
Above, we demonstrated that TLC analysis and LC-MS analysis showed marked differences
of TAG profiles when comparing unfermented and dried fermented beans are directly
compared. However, this distinction was not apparent in the PCA using the entire TAG peak
profile (see Fig. C6), illustrating the failure of PCA to identify key minor biomarkers in the
absence of biased normalization or scaling. Since the major 12 TAGs dominate the total
variance of the entire peak profile, we decided to perform a modified PCA on the peak profile
excluding the data for the 12 key TAGs and secondly carry out a complementary multivariate
analysis, in this case, a hierarchical clustering analysis using the LC-MS data. The restricted
PCA showed a clear separation of unfermented and dried fermented bean samples in the score
plot (Fig. 5.8). While the fermented samples are mainly separated along the second principal
component, the unfermented samples showed higher variance across the first two principal
components. These first 2 principal components explain together more than 69% of the total
variance. Moreover, a hierarchical cluster analysis was performed corroborating the findings
of the restricted PCA. In detail, the hierarchical cluster analysis showed a net separation of
unfermented from dried fermented beans (Fig. 5.9). Aligning the underlying TAG profiles to
the clustering illustrated the major differences of the unfermented and fermented samples.
Polar TAGs, here, eluting at retention time from 10 to 20 min, were present in unfermented
beans and absent in dried fermented beans.
Figure 5.8 PCA of the TAGs profiles from

LC-MS data, excluding the major 12 TAGs for
unfermented and fermented cocoa beans from
different countries of origin. The sample labeling
and coloring illustrate the different countries of
origin, namely, B – Brazil, E – Ecuador, I –
Indonesia, M – Malaysia, T – Tanzania, and V –
Ivory Coast. In addition, the sample label, as well
as the shape and the coloring of the two-standard
deviational ellipses indicate the fermentation
status, F – fermented and U – unfermented.
120
Chapter 5
Figure 5.9 Heat map of hierarchical clustering of the peak profiles from LC-MS data, excluding
the major 12 TAGs for unfermented (U) and fermented (F) cocoa beans from different countries of
origin. The detected peaks are given in ascending retention time order.
Our previous work allowed structural identification of these polar TAGs by a combination
of 2D-NMR and mass spectrometry as TAGs with a hydroxyl-allyl fatty acid substituent (Sirbu,
Corno, Ullrich, & Kuhnert, 2018). These compounds are apparently powerful biomarkers of
cocoa fermentation. At this stage, we propose to name these fatty acid moieties as cacaoic acid.
A structure of one of the biomarkers is shown in Fig. 5.10.
Figure 5.10 Proposed structure of a triacylglycerol containing a hydroxyl-allyl fatty acid

substituent, namely cacaoic acid.
121
Chapter 5
The identification of such suitable biomarkers for fermentation status constitutes a challenge
in cocoa quality control. As described above, TLC analysis using Hanessian's stain provides a
simple tool for detecting the presence/absence of TAGs at Rf 5.5 containing cacaoic acid
suitable to aid sourcing in origin countries lacking sophisticated scientific infrastructure. Hence
we propose to use this TLC method in the assessment of cocoa fermentation status.
5.4 Conclusions
In this manuscript, we extended the understanding of TAG profiles in cocoa of different

origins. We have shown, to which degree variations of concentrations of TAGs occur in cocoa
butters from different origins. We have, based on multivariate statistical analysis, and proposed
a new softness indicator by taking into account only four di-unsaturated TAGs relative
quantities. We have revealed that the presence of hydroxyl-allyl fatty acid derived TAGs (now
cacaoic acid) is unique to unfermented beans and allows chemical differentiation between
unfermented and fermented cocoa beans. A simple analytical test suitable for field work and
sourcing, testing for the presence of these biomarkers has been introduced.
References
https://doi.org/10.1080/10408390701719272
https://doi.org/10.1080/10408398.2012.669428
Beckett, Stephen T., Harding, J., & Freedman, B. (2008). The Science of Chocolate (2nd edition). Cambridge,
UK: Royal Society of Chemistry.
Beckett, Steve T., Fowler, M. S., & Ziegler, G. R. (2017). Beckett’s Industrial Chocolate Manufacture and Use.
John Wiley & Sons.
Benjamini, Y., & Hochberg, Y. (1995). Controlling the False Discovery Rate: A Practical and Powerful Approach
to Multiple Testing. Journal of the Royal Statistical Society. Series B (Methodological), 57(1), 289–300. Retrieved
from http://www.jstor.org/stable/2346101
122
Chapter 5
Benjamini, Y., & Hochberg, Y. (2000). On the adaptive control of the false discovery rate in multiple testing with
independent statistics. Journal of Educational and Behavioral Statistics, 25(1), 60–83.
Caligiani, A., Palla, L., Acquotti, D., Marseglia, A., & Palla, G. (2014). Application of 1H NMR for the
characterisation of cocoa beans of different geographical origins and fermentation levels. Food Chemistry, 157
(Supplement C), 94–99. https://doi.org/10.1016/j.foodchem.2014.01.116
Chaiseri, S., & Dimick, P. S. (1989). Lipid and hardness characteristics of cocoa butters from different geographic
regions. Journal of the American Oil Chemists’ Society, 66(12), 1771–1776. https://doi.org/10.1007/BF02660745
Chaiseri, S., & Dimick, P. S. (1995). Dynamic crystallization of cocoa butter. I. characterization of simple lipids
in rapid- and slow-nucleating cocoa butters and their seed crystals. Journal of the American Oil Chemists’ Society,
72(12), 1491–1496. https://doi.org/10.1007/BF02577842
Chin, A. H. G. (mardi. (1989). Cocoa butter characteristics of Malaysian clonal cocoa [Theobroma cacao].
Retrieved from http://agris.upm.edu.my:8080/dspace/handle/0/4247
Dsouza, R., Grimbs, S., Behrends, B., Bernaert, H., Ullrich, M., & Kuhnert, N. (2017). Origin-based polyphenolic
fingerprinting of Theobroma cacao in unfermented and fermented beans. Food Research International.
Figueira, A., Lambert, S., Carpenter, D., Pires, J., Cascardo, J. C. M., & Romanczyk, L. (1997). The similarity of
cocoa flavour of fermented seeds from fingerprinted genotypes of Theobroma cacao L. from Brazil and Malaysia
(Vol. 74).
Kumari, N., Kofi, K. J., Grimbs, S., D’Souza, R. N., Kuhnert, N., Vrancken, G., & Ullrich, M. S. (2016).
Biochemical fate of vicilin storage protein during fermentation and drying of cocoa beans. Food Research
International (Ottawa, Ont.), 90, 53–65. https://doi.org/10.1016/j.foodres.2016.10.033
Lehrian, D. W., & Keeney, P. G. (1980). Changes in lipid components of seeds during growth and ripening of
cacao fruit. Journal of the American Oil Chemists’ Society, 57(2), 61–65. https://doi.org/10.1007/BF02674361
Liendo, R., Padilla, F. C., & Quintana, A. (1997). Characterization of cocoa butter extracted from Criollo cultivars
of Theobroma cacao L. Food Research International, 30(9), 727–731. https://doi.org/10.1016/S0963-
9969(98)00025-8
123
Chapter 5
Lima, L. J. R., Almeida, M. H., Rob, N., & Zwietering, M. H. (2011). Theobroma cacao L., “the food of the gods”:
Quality determinants of commercial cocoa beans, with particular reference to the impact of fermentation. Critical
Reviews in Food Science and Nutrition, 51(8), 731–761. https://doi.org/10.1080/10408391003799913
Lipp, M., Simoneau, C., Ulberth, F., Anklam, E., Crews, C., Brereton, P., … Wiedmaier, C. (2001). Composition
of Genuine Cocoa Butter and Cocoa Butter Equivalents. Journal of Food Composition and Analysis, 14(4), 399–
Lísa, M., Holčapek, M., & Boháč, M. (2009). Statistical Evaluation of Triacylglycerol Composition in Plant Oils
Based on High-Performance Liquid Chromatography−Atmospheric Pressure Chemical Ionization Mass
Spectrometry Data. Journal of Agricultural and Food Chemistry, 57(15), 6888–6898.
https://doi.org/10.1021/jf901189u
Marseglia, A., Acquotti, D., Consonni, R., Cagliani, L. R., Palla, G., & Caligiani, A. (2016). HR MAS 1H NMR
and chemometrics as useful tool to assess the geographical origin of cocoa beans – Comparison with HR 1H
NMR. Food Research International, 85(Supplement C), 273–281. https://doi.org/10.1016/j.foodres.2016.05.001
Marty, S., & Marangoni, A. G. (2009). Effects of Cocoa Butter Origin, Tempering Procedure, and Structure on
Oil Migration Kinetics. Crystal Growth & Design, 9(10), 4415–4423. https://doi.org/10.1021/cg9004505
Ribeiro, A. P. B., Correa Basso, R., Goncalves, L. a. G., Gioielli, L. A., Oliveira dos Santos, A., Pavie Cardoso,
L., & Guenter Kieckbusch, T. (2012). Physico-chemical properties of Brazilian cocoa butter and industrial blends.
Part II - Microstructure, polymorphic behavior and crystallization characteristics. GRASAS Y ACEITES, 63(1),
89–99. https://doi.org/10.3989/gya.069111
Rusconi, M., & Conti, A. (2010). Theobroma cacao L., the Food of the Gods: A scientific approach beyond myths
and claims. Pharmacological Research, 61(1), 5–13. https://doi.org/10.1016/j.phrs.2009.08.008
Shukla, V. (2005). Cocoa Butter, Cocoa Butter Equivalents, and Cocoa Butter Substitutes.
https://doi.org/10.1201/9781420039184.ch12
Simoneau, C., Lipp, M., Ulberth, F., & Anklam, E. (2000). Quantification of cocoa butter equivalents in mixtures
with cocoa butter by chromatographic methods and multivariate data evaluation. European Food Research and
Technology, 211(2), 147–152. https://doi.org/10.1007/s002179900145
124
Chapter 5
Sirbu, D., Corno, M., Ullrich, M. S., & Kuhnert, N. (n.d.). Characterization of triacylglycerols in unfermented
cocoa beans by HPLC-ESI mass spectrometry. Food Chemistry. https://doi.org/10.1016/j.foodchem.2018.01.194
The International Cocoa Organization (ICCO) | Cocoa Producing and Cocoa Consuming Countries. (n.d.).
Retrieved February 1, 2018, from https://www.icco.org/
Tucci, M. L. S., Abreu, M. F. de, Coral, F. J., Futino, A. M., Alfonsi, L. R. R., & Saes, L. A. (1996). Fat and fatty
acids content of cocoa clones under the conditions of Vale do Ribeira, State of São Paulo, Brazil. Bragantia, 55(2),
207–213. https://doi.org/10.1590/S0006-87051996000200001
van den Berg, R. A., Hoefsloot, H. C., Westerhuis, J. A., Smilde, A. K., & van der Werf, M. J. (2006). Centering,
scaling, and transformations: improving the biological information content of metabolomics data. BMC
Genomics, 7, 142. https://doi.org/10.1186/1471-2164-7-142
Vieira, L. R., Efraim, P., Van, D. W., De, C., & Dewettinck, K. (2015). Influence of Brazilian Geographic Region
and Organic Agriculture on the Composition and Crystallization Properties of Cocoa Butter. JAOCS, Journal of
the American Oil Chemists’ Society, 92(11–12), 1579–1592. https://doi.org/10.1007/s11746-015-2728-y
125
Chapter 6
Chapter 6:
6 Fatty acids profiles in unfermented and fermented

dried cocoa beans of different origins and
correlation with the triacylglycerol profile
Abstract
Fatty acid profile, obtained as FAME data, of unfermented wet cocoa beans and fermented
dried cocoa beans from six different origins were analyzed and statistical analysis was applied
to assess the significance of the differences observed. Results showed that the origin and the
fermentation status have an impact on the fatty acid profile. Samples from Ivory Coast showed
a significance value of p˂0.0001 for pentadecylic acid, margaroleic acid, and margaric acid.
Palmitoleic acid and linoleic acid increases after the fermentation whereas gadoleic acid
decreases significantly. Besides, correlation studied between the most abundant fatty acids and
most abundant triacylglycerols were investigated. A high correlation between palmitic acid and
the main triacylglycerols were observed. The relations between fatty acids and triacylglycerols
diminished within the fermentation.
Keywords: fatty acids; GC-FID; unfermented cocoa beans; fermented cocoa beans; p value;
correlations; variance.
126
Chapter 6
6.1 Introduction
The Aztec emperor Montezuma II (1502-1520) has already recognized the nutritional value
of cocoa beans: “The divine drink, which builds up resistance and fights fatigue. A cup of this
precious drink permits a man to walk for a whole day without food” (Coe & Coe, 2007). The
high-energy values of the cocoa are almost entirely derived from the fat content. One gram of
lipid provides around 9 kcal/g, as compared to protein and carbohydrate each yielding around
4.5 kcal/g. The fat content of cocoa consists mostly of palmitic acid, oleic acid, and stearic acid
forming a selection of triacylglycerols. The geographical origin influences the composition
content of the fatty acids. Thus, palmitic acid content ranges in between 24.1-27.9%, stearic
acid between 33.3-37.6%, and oleic acid between 32.7-36.5% (Lipp & Anklam, 1998; Lísa &
Holcapek, 2008; Podlaha, Toregard, & Puschl, 1984). Vieira et al. showed a big variability
within the Brazilian samples, in which, samples from trees growing in warmer regions shown
higher values of saturated fatty acids in contrast to the samples from trees growing in areas of
lower temperatures which shown higher values of unsaturated fatty acids (Luana Reis Vieira,
Efraim, Walle, Clercq, & Dewettinck, 2015). Torres-Moreno et al. have determined the fatty
acid profile in unroasted cocoa beans and chocolates of two different origins, where they have
shown that only the origin had an impact on the fatty acid profile. To date, there are no studies
showing the influence of the fermentation on the fatty acids profile.
Fatty acids constitute the nutritional components of cocoa butter. Although the chocolate
contains large amount of saturated, the health benefits of chocolate are controversial due to
presence of the polyphenols having strong antioxidant activity (Lee, Kim, Lee, & Lee, 2003).
Epidemiological data have shown that unsaturated fatty acids have a hypocholesterolemic
effect, while saturated fatty acids have an opposite effect (Hu et al., 1999). Stearic acid seems
to exert a neutral effect on plasma LDL-cholesterol and shown have fewer implications for the
progression of coronary heart disease (Kris-Etherton et al., 1993). Although the mechanism of
its hypocholesterolemic effect is still not clear, the reason may lay in the limited absorption of
this fatty acid from the intestinal tract in contrast to most other lipids (Weisburger, 2001). A
short-term cocoa consumption significantly reduced blood cholesterol, but the changes were
dependent on the dose of cocoa consumption and the health status of participants (Jia et al.,
2010).
127
Chapter 6
The analysis of fatty acids is performed by converting the triacylglycerols to the

corresponding fatty acids methyl esters (FAME) (Petrović, Kezić, & Bolanča, 2010). The aim
of this study was to evaluate the fatty acid profile of cocoa in unfermented wet beans, and
fermented dried beans of different origins. Moreover, a correlation between the fatty acid and
triacylglycerol profile was attempted to expand the understanding of the chemistry of cocoa
butter.
Dichloromethane 99,8% stabilized with amylene for synthesis, potassium hydroxide,

sulphuric acid, chloroform, heptane, methanol, ethanol, molecular sieves. All chemicals were
purchased from Panreac AppliChem (Darmstadt, Germany), Sigma-Aldrich Chemie (Steinheim,
Germany), Carl Roth (Karlsruhe, Germany).
Frozen fresh unfermented cocoa beans samples and dried fermented cocoa beans samples
from 6 different origins (Ivory Coast, Tanzania, Indonesia, Ecuador, Malaysia, and Brazil) were
received in several sets from Barry Callebaut. Firstly, the seed material was deshelled and a
standardized grinding method were used to grind wet unfermented frozen beans and dried
fermented beans, employing the Grindomix Retsch GM 200 equipment with the purpose to
make a homogenous powder. Henceforth, an overnight Soxhlet (Buchi Extraction System B811
instrument, Flawil, Switzerland) method using dichloromethane as an extraction solvent, 150
mL of 5 g of powdered beans, was used. Extracted lipids were quantified gravimetrically after
evaporation to dryness in a rotavapor. The dry residue was then stored at -20 °C until further
analysis.
6.2.3 FAME synthesis
Optimization of the synthesis of FAME and the limits of detection and quantification was
128
Chapter 6
achieved after by choosing carefully few suitable methods from the literature. The results of
the procedure are presented in the bachelor thesis of Megi Mustafai (May, 2015) under author’s
supervision. The final optimized method was based on the method developed by O’Fallon et
al. with modifications (O’Fallon, Busboom, Nelson, & Gaskins, 2007). The maximum limit of
quantification (LOQ) was found at 300 mg and the minimum at 5 mg. Thus, 100 mg initial
lipid extract mass was applied for each FAME synthesis method. Method 1 was chosen to
extract the bulk samples. More specifically, 0.7 mL 10 N KOH, 5.3 mL methanol and 100 mg
initial lipid extract were stirred continuously at 50 ⁰C for one hour. 500 μL chloroform was
added together with potassium hydroxide and methanol, to allow good dissolvement of lipid
extracts. The samples were cooled down at room temperature and 0.58 mL 24 N H2SO4 was
added. All samples were stirred continuously at 50 ⁰C for one additional hour. FAME were
extracted with 2 mL heptane, vortexed for 5 minutes and centrifuged at 3.500 g for 3 minutes.
Molecular sieves were added for 10 minutes in the final FAME solution to further neutralize
the mixture. Then solution was diluted 1 to 10 and 1 to 20 and 2 vials and nitrogen was blown
on every sample to avoid oxidation and then the samples were directly analysed by gas
chromatography.
6.2.4 GC-FID method
A capillary column VF-5ms, 30 m x 0.25 mm was used, having an injector temperature of

265 °C and detector temperature 300 °C. Helium was used as a carrier gas, at a flow rate of
16.2 mL/min. The split ratio was 1:10 and the sample injection volume 1.0 μL. GC analysis
started at 60 °C for 1 minute, followed by a second step of 10 °C /min temperature rate until
200 °C, with a hold time of 10 minutes, then 7 °C /min rate until 300 °C, at a hold time of 3
minutes and lastly having a temperature rate of 10 °C /min at 300 °C and a final holding time
of 10 minutes. Fatty acids identification was performed comparing with a Supelco 37 comp.
FAME MIX standard and GC-MS determined according to Elvert at al. by using the same
instrument (Elvert, Boetius, Knittel, & Jørgensen, 2003).
Microsoft Office Excel 2016 was used to order and process the data. Known compounds
were quantified based on the calibration curve for FAME MIX standard using serial dilutions,
129
Chapter 6
for which at least 8 points were acquired (Appendix, Table D.1 and Table D.2). Unknown
compounds were relatively quantified using the calibration curve of the nearest lower retention
time fatty acid methyl ester standard reference. A one-factor and a two-factors ANOVA were
used to study the effect of origin and the effect of fermentation status on fatty acid profile. The
analysis were carried out with XLSTAT Evaluation 2018.2 software version 2018 (Addinsoft,
France).
6.3.1 Fatty acids profile
GC-FID chromatograms shown a high diversity of peaks in FAME cocoa extract (Fig. 6.1).
18 compounds have been positively identified by comparison of retention times to FAME
standard mix. Additionally, GC-MS data helped to confirm the identification for the most
abundant peaks. Palmitic acid, oleic acid, and stearic acid are the most abundant fatty acids,
followed by palmitoleic acid, linoleic acid, and arachidic acid as second-level most abundant
fatty acid (Fig. 6.1). The complete fatty acid composition is given in Table D.2 Appendix D,
along with the calibration data for the quantification of fatty acids displayed in Table D.1.
Figure 6.1 Representative GC-FID chromatogram of fatty acids methyl esters of cocoa lipid
extract, Ecuador unfermented beans.
130
Chapter 6
As it can be seen from these Tables, palmitic acids varies between 18 and 25% of the total
fatty acids, oleic acid between 24 and 34% of the total fatty acids, and stearic acid between 25
and 36% of the total fatty acids. These values showed a significantly larger spread than what
is described in the literature to date. A plausible explanation could be a smaller sample number
investigated. Authors described a biased normalisation procedure taking only identified fatty
acids into consideration. Therefore, we would like to emphasize the importance of a correct
evaluation of the fatty acid amount.
Analysis of samples of different origins and fermentation status showed significant
differences in the fatty acids profile (Fig. 6.2). The FAME extract of unfermented wet beans
demonstrated the presence of a higher number of peaks if compared to the FAME extract of
fermented dried beans. Furthermore, already from a first glance, the fermentation process
seems to have in impact on the major peaks as well, since differences in the intensity of the
peaks could be observed. Representative chromatograms of unfermented wet and fermented
dried FAME extract from Ecuador and Indonesia are shown in Figure 6.2.
Figure 6.2 Representative GC-FID chromatogram of FAME extract of (a) and (c) unfermented wet
beans and (b) and (d) fermented dried beans from Ecuador and Indonesia respectively.
131
Chapter 6
6.3.2 Fatty acids distribution
Figure 6.3 displays box plots indicating the distribution of the most abundant fatty acids if
origins are directly compared. From the Figure, it can be deduced that there is an unexpected
variability across origins. Nevertheless, few remarks can be made: the difference between the
amount of palmitic acid and stearic acid showed to be less in Ecuadorian samples, whereas
Brazilian and Ivorian samples indicate a substantial distance between the quantities of these
fatty acids. Brazilian and Ecuadorian samples showed higher values of palmitic acid if
compared to the other 5 countries. Lower values of palmitic acids are present in Ivorian sample
whereas stearic is more abundant in this country with respect to the others. A higher amount of
oleic acid is present in Brazilian samples, while other countries maintain almost the same
profile. Interestingly, Brazilian samples showed a higher amount of stearic acid and oleic acid
and lower amount of palmitic acid if compared to the Ecuadorian samples. Another observation
is that linoleic acid profile follows almost the same pattern of oleic acid.
Bar plots of the most abundant fatty acids in unfermented wet and fermented dried
illustrated in Figure 6.4 (a) and (b) allow some insights on how the fermentation process is
affecting the fatty acids profile. Although all the fatty acids variations are random, linoleic fatty
acid in particular, displays a clear increasing after the fermentation with more than 50%
increment in some samples.
Figure 6.3 Most abundant fatty acid distribution according to the origin.
132
Chapter 6
Figure 6.5 shows the total fatty acids amount the ratio between total unsaturated and total
saturated fatty. Large fluctuations in the total fatty acids amount were observed across all origin
pairs, defined as unfermented wet beans and fermented dried beans, in which no trend could
be reasoned. An increase of the ratio between total unsaturated and total saturated fatty during
the fermentation would be desirable for cocoa manufacturers however, it was only observed in
some samples from Ecuador, Malaysia, Ivory Coast, and Tanzania.
Figure 6.4 (a) Bar plot of the most abundant 3 fatty acids and (b) of the other 3 fatty acids in
unfermented wet and fermented dried cocoa beans of different origins.
Besides the identified peaks, GC-FID analysis of the fatty acids methyl esters derived from
unfermented cocoa beans revealed the occurrence of additional peaks if compared to fermented
dried beans, as it can be inferred from the Figure 6.2. From all unknown peaks, 13 of them
have shown a high difference between the unfermented wet beans and fermented dried beans
(Fig. 6.6). These compounds eluting at retention times between stearic and gadoleic acid were
tentatively assigned as hydroxyl fatty acids, well described in Chapter 4.
133
Chapter 6
Figure 6.5 Bar plot illustrating the total fatty acids* content in unfermented wet and fermented dried
cocoa beans of different origins and their corresponding ratio unsaturated/saturated fatty acids*.
*Only the data of the positively identified peaks were taken into account.
Figure 6.6 Bar plot illustrating the hydroxyl fatty acids content in unfermented wet and fermented
dried cocoa beans of different origins. *the amount was calculated as mg of stearic acid/g of cocoa lipid extract.
134
Chapter 6
Unfortunately, the exact assignment could not be achieved as the main fatty acid dictated the
concentration level, at which these fatty acids were much diluted and therefore complicated to
perform a GC-MS analysis. In all the samples, significant differences between unfermented
wet beans and fermented dried beans could be estimated (Fig. 6.6). The total amount of
hydroxyl fatty acids, determined as stearic acid equivalents, varied between 0.5 to almost 5
mg/g of cocoa lipid extract in unfermented wet beans, dropping down to less than 0.5 mg/g of
cocoa lipid extract in fermented dried beans. These type of fatty acids are described for the first
time in cocoa lipid extract, and these results are in complete accordance with what has been
found in Chapter 5.
One-factor and two-factor ANOVA were used to study the effect of the geographical origin
and the fermentation status on the fatty acids profile. For this purpose quantification date were
employed. Results showed a significant effect on the following fatty acids: pentadecylic acid,
margaroleic acid and margaric acid, p-value ˂0.0001, concerning origins. Indeed, Ivory Coast
has shown higher amounts of these fatty acids. With the same confidence interval at 95%
statistical significant difference for myristic acid, palmitoleic acid, linoleic acid, oleic acid,
elaidic acid, gadoleic acid, eicosanoic acid, henecosylic acid, and lignoceric acid could be
observed. Their p values are given in Table 6.1.
From the analysis of variance, it could be inferred that certainly, the fatty acid profile
changes with the geographical origin. Brazilian and Ecuadorian samples showed significant
differences in the amount of the short chain fatty acid, whereas Malaysian samples contain a
higher amount of long chain fatty acids, such as eicosanoic acid (C20:0), heneicosylic acid
(C21:0), and lignoceric acid (C24:0). These results are in line with the data for triacylglycerol
(TAGs) profiling, Chapter 5, where Malaysian samples show a higher amount of saturated
TAGs and Brazilian and Ecuadorian samples showed a higher amount of unsaturated TAGs.
Vieira et al. have observed significant variance for margaric, linoleic and arachidic acid within
Brazilian samples from different areas in Brazil (L.R. Vieira, Efraim, Van, De, & Dewettinck,
2015); and Torres-Moreno have observed significant variance for myristic, margaric,
margaroleic, palmitic, and stearic acid when comparing samples from Ecuador and Ghana.
Very few authors describe the presence of palmitoleic acid and margaric acid in cocoa butter
(Rossi, Arnoldi, Salvioni, & Schiraldi, 1989; Torres-Moreno, Torrescasana, Salas-Salvadó, &
135
Chapter 6
136
Chapter 6
R2 value of 0.69 and 0.77 between oleic acid and POO, and between oleic acid and SOO
respectively were obtained. Thus, a correlation between the oleic acid amount and the amount
of the TAGs containing two oleic acid moieties was found.
Figure 6.7 (a) Scatter plots illustrating the variation of oleic acid and main TAGs in unfermented wet
cocoa beans and (b) in fermented dried cocoa beans and of different origins E – Ecuador, M –
Malaysia, I– Indonesia, B – Brazil, V – Ivory Coast.
Usually, in statistics, three types of correlations are measured: Pearson correlation, Kendall
rank correlation, and Spearman correlation to evaluate the strength of the association between
two variables and the direction of the relationship where a value of ± 1 indicates a perfect
degree of association between the two set of variables. Thus, to confirm what observed
correlation coefficient of Pearson, Kendall, and Spearman were calculated. Results are shown
in Table 6.2. The correlation coefficients showed a high negative correlation between palmitic
and stearic acid. Palmitic acid showed a high positive correlation between POP and POS and a
negative relationship with SOS. On the other hand, stearic acid showed a strong association
with SOS and a negative correlation with POP. Linoleic acid showed a positive correlation with
all 4 diunsaturated TAGs, namely POO, PLS, SOO, and SLS. From these results, it could be
inferred that the profile of the most abundant fatty acids is reflected by the occurrences of the
most abundant TAGs. Moreover, the values may give an indication until which degree such
reflection is covered. The best correlation was observed between stearic and SOS showing a
Pearson correlation coefficient value of 0.945, whereas palmitic acid showed a Pearson
137
Chapter 6
correlation coefficient value of 0.889. This indicates that the contribution of SOS to the total
amount of stearic fatty acid is higher than the contribution of POP to the total amount of
palmitic acid.
Table 6.2 Correlation matrixes of fatty acid and TAGs in unfermented wet and dried fermented cocoa
beans.
138
Chapter 6
Remarkably, a decrease in the value of the correlation coefficient in fermented dried beans
if compared to unfermented wet beans was observed. The greatest difference was observed for
linoleic acid grieved, which showed no correlation with POO and SOO in fermented dried
cocoa beans. The reduction in the value of the correlation coefficient after the fermentation and
drying of the beans suggest that the amount of the total fatty acids was subjected to alterations
from an external source by adding or subtracting fatty acids. Another hypothesis is that the
fermentation process implies chemical modification. Consequently, the fatty acids undergo
other chemical transformation pathways. Besides, during the fermentation, as well as during
the transportation, the temperature rises, which can entail melting of the low melting point
TAGs that consequently could leak outside the beans, therefore, changing the total profile of
fatty acids.
6.4 Conclusions
Although fermentation has shown to have a significant impact on the solids of cocoa,
evaluation of the changed of the fatty acids profile before and after fermentation was not yet
completed. This study shows clear differences supported by statistical analysis between
geographical origins and fermentation status of the beans. From the results, key features can
be selected to help the quality assessment of cocoa beans. Herein, the quantity of palmitoleic
acid, linoleic acid, and unusual fatty acids could be considered as markers of the fermentation
status, whereas the quantity of specific saturated fatty acids could indicate the geographical
origin. The distinct fatty acids fingerprint may be used in the future to qualify the beans
according to their lipid profile. Moreover, it provides insight into how to optimize and tailor
fatty acid profile with respect to innovative applications and health products.
References
Hudson.
Elvert, M., Boetius, A., Knittel, K., & Jørgensen, B. B. (2003). Characterization of Specific Membrane Fatty Acids
as Chemotaxonomic Markers for Sulfate-Reducing Bacteria Involved in Anaerobic Oxidation of Methane.
Geomicrobiology Journal, 20(4), 403–419. https://doi.org/10.1080/01490450303894
139
Chapter 6
Hu, F. B., Stampfer, M. J., Manson, J. E., Ascherio, A., Colditz, G. A., Speizer, F. E., … Willett, W. C. (1999).
Dietary saturated fats and their food sources in relation to the risk of coronary heart disease in women. The
American Journal of Clinical Nutrition, 70(6), 1001–1008. https://doi.org/10.1093/ajcn/70.6.1001
Lee, K. W., Kim, Y. J., Lee, H. J., & Lee, C. Y. (2003). Cocoa Has More Phenolic Phytochemicals and a Higher
Antioxidant Capacity than Teas and Red Wine. Journal of Agricultural and Food Chemistry, 51(25), 7292–7295.
https://doi.org/10.1021/jf0344385
Lísa, M., & Holcapek, M. (2008). Triacylglycerols profiling in plant oils important in food industry, dietetics and
cosmetics using high-performance liquid chromatography-atmospheric pressure chemical ionization mass
spectrometry. Journal of Chromatography. A, 1198–1199, 115–130.
O’Fallon, J. V., Busboom, J. R., Nelson, M. L., & Gaskins, C. T. (2007). A direct method for fatty acid methyl
ester synthesis: application to wet meat tissues, oils, and feedstuffs. Journal of Animal Science, 85(6), 1511–1521.
https://doi.org/10.2527/jas.2006-491
Petrović, M., Kezić, N., & Bolanča, V. (2010). Optimization of the GC method for routine analysis of the fatty
acid profile in several food samples. Food Chemistry, 122(1), 285–291.
https://doi.org/10.1016/j.foodchem.2010.02.018
Vieira, L.R., Efraim, P., Van, D. W., De, C., & Dewettinck, K. (2015). Influence of Brazilian Geographic Region
and Organic Agriculture on the Composition and Crystallization Properties of Cocoa Butter. JAOCS, Journal of
the American Oil Chemists’ Society, 92(11–12), 1579–1592. https://doi.org/10.1007/s11746-015-2728-y
140
Chapter 6
Vieira, Luana Reis, Efraim, P., Walle, D. V. de, Clercq, N. D., & Dewettinck, K. (2015). Influence of Brazilian
Geographic Region and Organic Agriculture on the Composition and Crystallization Properties of Cocoa Butter.
Journal of the American Oil Chemists’ Society, 92(11–12), 1579–1592. https://doi.org/10.1007/s11746-015-2728-
y
Weisburger, J. H. (2001). Chemopreventive Effects of Cocoa Polyphenols on Chronic Diseases. Experimental

Biology and Medicine, 226(10), 891–897. https://doi.org/10.1177/153537020122601003
141
Chapter 7
Chapter 7:
7 Monitoring lipids of Theobroma cacao during the

fermentation
Abstract
Within the frame of understanding the impact of the fermentation on the lipid profile in cocoa, 6
fermentation trials from 3 different origins were analyzed. Triacylglycerols and fatty acids content
showed significant variations between 72 and 96 h. Hydroxyl derivatives of the triacylglycerols
decrease entirely after 96 h of fermentation whereas linoleic acid showed a considerable increment
throughout the process. The lipid profile of minor components indicated that the microorganisms
involved in the fermentation process give their contribution in determining the profile of the less
abundant lipid compounds in cocoa.
Keywords: fermentation; triacylglycerols; fatty acids; bacteria; p-value; linoleic acid; 2D map.
142
Chapter 7
7.1 Introduction
The fermentation process is the most important contributing factor to the quality of flavour
and final composition of cocoa. The beans of cocoa are surrounded by a mucilaginous pulp
rich in sugars and citric acid, which provide an excellent medium for the growth of the yeast
and bacteria (Badrie, Bekele, Sikora, & Sikora, 2015; Ho, Zhao, & Fleet, 2014). The
fermentation process consists of two phases: an anaerobic phase in the first 24 to 36 h after
harvesting and opening the pods; and an aerobic phase after 48 to 96 h, when the beans are
turned to allow aeration. In the first phase, yeasts exhibit maximum development after 24 h,
after which increased alcohol concentration and pH activates the growth of lactic acid bacteria.
During this phase lactic acid, acetic acid, ethanol, and carbon dioxide are created from the pulp
sugars. The second phase is characterized by a dynamic development of acetic acid bacteria
responsible for the oxidation of ethanol to acetic acid, and the exothermic process leading to a
temperature increase in the cocoa mass up to 45-52 °C (Aprotosoaie, Luca, & Miron, 2016;
Lima, Almeida, Rob, & Zwietering, 2011). This phase is considered essential for flavour
development because of the diffusion of metabolites in the cotyledons causing the death of the
embryo and activation of endogenous enzymes (Afoakwa, Paterson, Fowler, & Ryan, 2008;
Ho et al., 2014; Schwan & Wheals, 2004). The duration of the fermentation varied between 96
and 168 h in a range of different methods depending on the country traditional common
practices (Lima et al., 2011). During the fermentation, major biochemical changes occur
(Kadow, Bohlmann, Phillips, & Lieberei, 2013; Kongor et al., 2016). Thus, the
physicochemical transformation is the result of the following concatenation of micro-effects:
acidification, subcellular decompartimentalization, enzymatic degradation, and components
leaching (Lima et al., 2011; Schwan & Wheals, 2004; Ziegleder Gottfried, 2017). Most of the
cocoa bean fermentations in the world are performed in a spontaneous process, which shows
vast variations due to the inconstant environmental microorganisms. Hitherto, the fermentation
can occur under controlled condition using specific culture strains (De Vuyst L. & Weckx S.,
2016).
Cocoa pods are harvested at maturation, the time point when the germination of seed plant
commences. Therefore, mobilization of the storage lipids, primary carbon source in oil seeds,
takes place to offer energy for the germinating seedling (Feussner, Kühn, & Wasternack, 2001).
This process is catalysed by Lipoxygenases (LOXs, linoleate:oxygen oxidoreductase, EC
143
Chapter 7
1.13.11.12), a non-heme iron containing dioxygenases that catalyse the formation of

hydroperoxy derivatives of polyunsaturated fatty acids (Rosahl, 1996). Moreover, these
enzymes monitor the production of multiple other growth regulators, antimicrobial
compounds, flavours and odours as well as signal molecules. One of the compounds regulated
by LOX is jasmonic acid, a phytohormone, known to be involved growth processes and
responses to abiotic and biotic stress (Ghasemi Pirbalouti, Sajjadi, & Parang, 2014; Turner,
Ellis, & Devoto, 2002). Patras et al. have identified the presence of jasmonic acid derivative,
12-hydroxyjasmonic acid-O-sulfate (Patras, Milev, Vrancken, & Kuhnert, 2014).
Microbial attack during the fermentation constitutes an enormous stressed induced process
that triggers stress response mechanisms in cocoa seeds. A recent study revealed an increase of
the antioxidant capacity in fermented versus unfermented beans, extract of which contained
besides the polyphenols 12-hydroxyjasmonic acid-O-sulfate (Dsouza et al., 2017). Much of
work has been done in understanding how the post harvesting procedures affect the formation
of flavour precursors, sugars, and proteins, and antioxidant capacity of the polyphenols,
extensively reviewed by several authors (Afoakwa et al., 2008; Aprotosoaie et al., 2016).
However, not much is known about the fate of the lipids during the fermentation. Above and
beyond, authors have drawn attention towards the need of elucidation of the fermentation
impact on the triacylglycerol composition (Lima et al., 2011).
Herein, the present study aimed at covering the knowledge gap by observing the changes
occurring in the triacylglycerol profile, fatty acids, and other minor components during the
fermentation. Within this aim, a lipidomics approach via a recently established method based
on reversed-phase high-performance liquid chromatography (HPLC) coupled with high
resolution mass spectrometry was used to analyze the lipid composition in cocoa. Besides,
fractionation of the lipid extract, described in Chapter 3, was applied for better visualization of
the minor lipid components.
144
Chapter 7
(Rotisolv® HPLC grade) and Tetra-dodecylammonium bromide, heptane, methanol, ethyl

acetate, was purchased from Carl Roth (Karlsruhe, Germany), dichloromethane 99,8%
stabilized with amylene for synthesis was purchased from Panreac AppliChem (Darmstadt,
Germany), ammonium formate LC-MS Ultra and formic acid (puriss., ≥ 98% (T) for mass
spectrometry), and acetic acid were purchased from Sigma-Aldrich Chemie (Steinheim,
Germany). Ethanol was subjected to distillation prior use.
Cocoa beans samples from 6 spontaneous fermentation trials and 1 controlled fermentation
trial from 3 different origins, including Ecuador, Malaysia, and Ivory Coast, were received in
several sets from Barry Callebaut. A fermentation trial represents a series of samples collected
at specific time points during the fermentation process of a batch of cocoa beans. Usually, the
time length of a fermentation process is between 5 and 7 days. Firstly, the seed material was
deshelled and ground using a grinder with the purpose of making homogenous powder.
Henceforth, an overnight Soxhlet (Buchi Extraction System B811 instrument, Flawil,
Switzerland) method using dichloromethane as an extraction solvent, 150 mL of 5 g of
powdered beans, was used. Extracted lipids were quantified gravimetrically after evaporation
to dryness in a rotavapor. The dry residue was then stored at -20 °C until further analysis. For
HPLC analysis, a concentration of 0.045 mg/mL in chloroform/ethanol (50/50) of cocoa lipid
extract was prepared.
7.2.3 Moisture analysis
Moisture content was determined by heating around 2.5 g of cocoa powder in the oven at
105 °C for 24, cooled in the dry atmosphere of a desiccator, and then reweighed. Total lipid
content was normalized according to: = .
Cocoa Powder (g) − Weight Loss (g)

145
Chapter 7
the SPE fractionation. The cartridges were first activated with 20 mL heptane/ethylacetate
(80/20, v/v), then, 20 ml heptane. Then, 10 mg of the lipid extract dissolved in 1 mL of heptane
were applied, after which the fractions were eluted with 5 different solvent mixtures: 1) 20 mL
heptane, first fraction collected ; 2) 20 mL heptane/ethyl acetate (97/3, v/v), for the second
fraction; 3) 20 mL heptane/ethyl acetate (85/15, v/v), third fraction collected; 4) 20 mL
heptane/ethyl acetate (85/15, v/v) plus 1% acetic acid, for the forth fraction; and 5) 20 mL
methanol. Extracted lipids were evaporated to dryness in a rotavapor and stored at -20 °C until
further analysis.
phase was pumped through the column at a flow rate of 0.6 mL/min. Two different elution
gradients were used. One for the total lipid separation and another to separate minor lipid
components in cocoa. The first gradient elution program consisted of holding solvent steady
A/B (40/60) for 3 min; followed by a linear gradient to solvent B (100) for 33 min, and ending
with isocratic elution at solvent B (100) for 10 min. The column was equilibrated at 40/60
solvent A/B for 5 min before reuse. The second gradient elution program consisted of holding
solvent steady A (100) for 5 min; followed by a linear gradient to solvent A/B (70/30) for
another 5; then by a linear gradient to solvent B (100) for 90 min, and ending with isocratic
elution at solvent B (100) for 10 min. The column was equilibrated at 100 solvent A for 5 min
before reuse.
and retention time alignment using a mass-to-charge (m/z) tolerance of Δm/z = 0.0001 and a
retention time (Rt) tolerance of 0.5 min for elution time <20 min and 1.0 min for an elution
time ≥20 min. Extensive manual curation of the TAGs peaks was performed to be uniquely
assigned with m/z and retention time (Rt) tolerances of Δm/z = 0.01 and ΔRt = 0.8 min
respectively. For further analyses, the peak areas were normalized to the internal standard
(Tetra-dodecylammonium bromide) at Rt 4.4 min.
146
Chapter 7
spectrometer (Bruker Daltonics UHT Ultra, Bremen, Germany) fitted with an ESI source used
as the detector with the following parameter settings: capillary voltage of 4.5 kV; nebulizing
gas pressure of 2 Ba; drying gas flow rate of 10 L/min; drying gas temperature of 220 °C. ESI
mass spectra were measured in the range of m/z 200–1200 in the positive ion mode. Internal
quadratic mode.
7.2.7 FAME synthesis
The method was based on the method developed by O’Fallon et al. with modifications
(O’Fallon, Busboom, Nelson, & Gaskins, 2007). Thus, 100 mg initial lipid extract mass was
applied for each FAME synthesis method. Method 1 was chosen to extract the bulk samples.
More specifically, 0.7 mL 10 N KOH, 5.3 mL methanol and 100 mg initial lipid extract were
stirred continuously at 50 ⁰C for one hour. 500 μL chloroform was added together with
potassium hydroxide and methanol, to allow good dissolvement of lipid extracts. The samples
were cooled down at room temperature and 0.58 mL 24 N H2SO4 was added. All samples were
stirred continuously at 50 ⁰C for one additional hour. FAME were extracted with 2 mL heptane,
vortexed for 5 minutes and centrifuged at 3.500 g for 3 minutes. Molecular sieves were added
for 10 minutes in the final FAME solution to further neutralize the mixture. Nitrogen was blown
on every sample to avoid oxidation and then the samples were directly analysed by gas
chromatography.
7.2.8 GC-FID method
A capillary column VF-5ms, 30 m x 0.25 mm was used, having an injector temperature of

265 ⁰C and detector temperature 300 ⁰C. Helium was used as a carrier gas, at a flow rate of
16.2 mL/min. The split ratio was 1:10 and the sample injection volume 1.0 μL. GC analysis
started at 60 ⁰C for 1 minute, followed by a second step of 10 ⁰C/min temperature rate until
200⁰C, with a hold time of 10 minutes, then 7 ⁰C/min rate until 300 ⁰C, at a hold time of 3
147
Chapter 7
minutes and lastly having a temperature rate of 10 ⁰C/min at 300 ⁰C and a final holding time
of 10 minutes. Fatty acids identification was performed comparing with a Supelco 37 comp.
FAME MIX standard and GC-MS determined according to Elvert at al. by using the same
instrument (Elvert, Boetius, Knittel, & Jørgensen, 2003).
7.2.9 Spectrophotometric analysis
Color analysis of cocoa lipid extract was measured by absorption. Spectra were recorded on
a Cary 4000 UV-Vis spectrophotometer (Varian, now Agilent, Waldbronn, Germany). For this
purpose, 50 mg of cocoa lipid extract were dissolved in 3 ml chloroform/methanol (50/50).
7.2.10 Lipid extraction from the pulp
Three mL of pulp were extracted with 6 mL chloroform ethanol (50/50, v/v) in a

liquid/liquid extraction. The mixture was placed in the ultrasound bath for 15 min and then
vigorously shacked with using a vortex. After that, the mixture was centrifuged at 3000 rpm
for 5 min and the organic phase was collected and evaporated using Turbovap. The lipid
content was redissolved in 1 ml chloroform ethanol (50/50, v/v) and after adding the internal
standard it was analyzed by LC-MS.
Microsoft Office Excel 2016 was used to order and process the data. A one-factor, two-
factors ANOVA and MANOVA were used to study the effect of origin and the effect of
fermentation status on the TAG profile. The analysis were carried out with XLSTAT Evaluation
2018.2 software version 2018 (Addinsoft, France).
148
Chapter 7
7.3 Results
7.3.1 Lipid amount
Fluctuations in the lipid amount were observed in all fermentation trials (Fig. 7.1), defined
as series of samples collected at specific time points. Although no definite pattern was found,
from the graph it seems that the profile shows turning point at 72 hours (h), 96 h and 120 h.
Only Malaysian samples, in both trials, showed a continuous increase during the fermentation
process.
Figure 7.1 Lipid amount as the percentage in dry mass of six fermentation trials from 6 different
origins.
7.3.2 TAGs profile during the fermentation
Individual TAGs have been analysed and compared in all fermentation trials. Figure 7.2
illustrates the relative amount of the main TAGs in cocoa beans, namely POP, POS, and SOS,
and the overall monounsaturated TAGs in the lipid extract. For all the trials it appears that 72
h (in Ecuadorian and Malaysian samples) and 96 h (in Ivorian beans) are crucial time points
for the monounsaturated TAGs during the fermentation. Ecuadorian sample showed fluctuation
but not an overall change in the total amount of monounsaturated TAGs, whereas the other two
origins exhibit a decrease during the fermentation.
149
Chapter 7
Figure 7.2 Monounsaturated TAGs relative amount in 3 fermentation trials from 3 different
origins.
A second observation could be made over the variation content of the three main
triacylglycerols during the second phase of the fermentation process. The variation pattern of
POP, POS, and SOS have shown differences when comparing the fermentation trials in the
three countries. In Ecuadorian samples, POS and SOS have shown the same kinetics whereas
in Malaysian beans the ratio between SOS and POS is lower during the second half of the
fermentation process. This ratio in Ivorian samples is close to one.
Surprisingly, POO and SOO have shown an increase at 72 h in all fermentation series (Fig.
7.3), which then drops at 96 h showing an opposite trend concerning the main TAGs described
above. Ecuador and Ivory Coast origins exposed a second increase at 120 h. In contrast, PLS
and SLS have shown a completely different behavior. In Ecuadorian and Ivorian samples the
amount of these 2 TAGs increase at 48 h and decrease at 72 h, followed by a second increase;
whereas in Malaysian samples, PLS and SLS increase until 96 h and then lower.
During the fermentation, Malaysian samples double their amount of saturated TAGs (Fig.
7.4), characteristic for this origin as already known from our previous work, Chapter 5. No
significant variations were observed in Ecuadorian samples. The lipid extract of the beans from
Ivory Coast showed a decrease of saturated TAGs during the fermentation.
150
Chapter 7
Figure 7.3 Di- und triunsaturated TAGs relative amount in 3 fermentation trials from 3 different
origins.
Figure 7.4 Saturated TAGs relative amount in 3 fermentation trials from 3 different origins.
Figure 7.5 illustrates the evolution of the monounsaturated, di- and triunsaturated, and
saturated TAGs in the Ecuadorian hybrid CCN51. All three series of TAGs showed a trend
disturbance at 72 h, where an increase was observed.
151
Chapter 7
Figure 7.5 Evolution of TAGs profiles in CCN51 hybrid during the fermentation.
In Chapter 4 we have described the presence of the unusual TAGs presenting hydroxyl fatty
acids esterified to the glycerol backbone, and in Chapter 5 the disappearance of this unusual
TAGs suggesting to consider these compounds as markers for unfermented cocoa beans. These
compounds decrease rapidly throughout the first 3 days of fermentation (Fig. 7.6). Two batched
of the same hybrid were compared and showed large variability (Fig. 7.7 (a)), yet with the same
decreasing development. Identical results were observed when comparing two different
hybrids (Fig. 7.7 (b)).
Figure 7.6 Unusual TAGs during the fermentation process.
152
Chapter 7
Figure 7.7 Unusual TAGs during the fermentation in (a) 2 batched of the same hybrid and (b) in 2
batched of two hybrids of the same origin.
Figure 7.8 Unusual TAGs during the fermentation in samples from Malaysia.
Ivory Coast hybrids showed a quite singular zigzag decrease of the unusual TAGs. Same it
was observed in the samples from Malaysia but more thickly than in Ivorian samples. It appears
that during the first 18 h the profile stays constant and after a series of increasing and decreasing
values follows reaching an almost complete decrease at 96 h.
153
Chapter 7
7.3.3 Statistical analysis on TAGs
Analysis of variance has shown that there are statistical differences in the number of
different TAGs during the fermentation. Figure 7.9 illustrates the mean values of 12 TAGs
subdivided as being part of a homologues series. A homologous series is a series of compounds
that differ in a defined mass increment corresponding to addition or subtraction of a fixed
moiety, in this case or +C2H4. As it can be observed, all the members of a homologues series
follow the same trend.
Figure 7.9 Means of the distribution of 4 homologues series TAGs in Malaysian samples.
154
Chapter 7
Only AOO did not show the same trend. As for POS, the analysis of variance provide ˂0.05
p values, presented in Table 7.1, indicating statistical difference between the first days of
fermentation and the last days. With the aim of encompassing all usual TAGs, a multivariate
analysis of variance (MANOVA) was conducted to assess if mean differences exist.
Multivariate F value (Wilks' λ) gave 1.643, p-value 0.005 for monounsaturated TAGs; F 1.49
and p-value 0.0001 for saturated TAGs; F 1.49 and p-value ˂0.0001 for diunsaturated POO
type; F 1.784 and p-value 0.08 for diunsaturated PLS type; F 1.508 and p-value ˂0.0001 for
triunsaturated TAGs; and F 1.946 and p-value ˂0.0001 for tetraunsaturated TAGs.
Table 7.1 Analysis of variance p-value for POS in two Malaysian fermentation trials.
POS Q1-144h Q1-120h Q1-96h Q1-72h Q1-54h Q1-48h Q1-42h Q1-36h Q1-30h Q1-24h Q1-18h Q1-12h Q1-6h
Q1-6h 0.002 0.002 0.008 0.006 0.973 0.502 0.703 0.471 0.686 0.861 0.926 0.474
Q1-12h 0.000 0.000 0.002 0.002 0.468 0.176 0.745 0.164 0.303 0.298 0.453
Q1-18h 0.002 0.002 0.010 0.007 0.892 0.565 0.635 0.533 0.752 0.958
Q1-24h 0.001 0.001 0.006 0.005 0.845 0.559 0.574 0.516 0.761
Q1-30h 0.005 0.005 0.020 0.012 0.621 0.817 0.425 0.782
Q1-36h 0.006 0.005 0.022 0.013 0.400 0.963 0.257
Q1-42h 0.001 0.001 0.004 0.003 0.704 0.295
Q1-48h 0.009 0.007 0.031 0.017 0.466
Q1-54h 0.002 0.002 0.008 0.006
Q1-72h 0.650 0.810 0.441
Q1-96h 0.654 0.503
Q1-120h 0.802
Q1-144h
7.3.4 Fatty acid profile during the fermentation
FAME analysis has shown an overall increase of the fatty acids during the fermentation
(Fig. 7.11). Herein, the turning points are also present at 72 and 96 h, as observed for TAGs.
Figure 7.12 illustrates this occurrence in another hybrid from Ivory Coast which was fermented
in spontaneous and controlled fermentation.
Interestingly, the same batch of sample fermented in 2 different conditions showed a
considerably different profile. The spontaneous fermentation of the hybrid G11UTA402XT413
showed an overall slight increase of the total fatty acids at the end of the fermentation with a
significant increase of all fatty acids 96 h that started decreasing at 120 h, reaching the initial
values at 144 h. This fact is almost not observed in the batch fermented under controlled
conditions. Moreover, spontaneous fermentation showed a significant increase of linoleic acid,
started at 48 h, and of palmitoleic acid, started at 72 h. The content of these 2 fatty acid return
at the average level at 144 h.
155
Chapter 7
Figure 7.11 Total fatty acids and main fatty acids profile of a fermentation series from Ivory
Coast.
Figure 7.12 Total fatty acids and main fatty acids profile of a spontaneous and controlled
fermentation series from Ivory Coast.
7.3.5 Colour of the lipid extract during the fermentation
Upon extraction, the lipid obtained from different time points have shown a varying range
of colors going through pale yellow to a brownish color. Therefore, the lipid extract was
analyzed by spectrophotometry. Results showed that the absorbance at 400 nm increased within
the fermentation reaching the maximum point at 96 and 120 h (Fig. 7.13). Similar events were
observed in other trials.
156
Chapter 7
Figure 7.13 Evolution of the color during the fermentation of cocoa beans.
7.3.6 Minor lipid classes during the fermentation
Solid phase extraction was performed with the aim of separating the TAGs fraction from the
remaining part of the lipid extract. Minor components were further analysed by HPLC-MS.
Figure 7.14 and 7.15 illustrates the evolution of the minor components during the fermentation
as two-dimensional (2D) map of the chromatograms with retention time on the x axes and m/z
ratios on the y axes. The interrupted blue circle at 24 h indicated the presence of the unusual
TAGs, which decrease in intensity and number at 48 h and then increase at 72 h. These results
are following what observed when analyzing the total lipid profile (Fig. 7.1).
At 96 h these compound disappear almost entirely and another class of compounds appears.
This class was identified as phospholipids, mainly phosphatidylcholines and
phosphatidylethanolamines. This class of compound persists until the last day of the
fermentation. In Ecuadorian samples, the loss of the unusual TAGs and the joint appearance of
the phospholipidic class can be detected at 72h.
Lipid extract from the pulp at 24 and 48 hours illustrate a different profile than the lipid
extract from cocoa (Fig. 7.16). This statement indicates that the pulp lipid reflects the lipid
profile of bacteria present in the pulp during the fermentation. The green circle shows the TAGs
class whereas the red circle indicates the phospholipid class. It can be noticed that the intensity
157
Chapter 7
158
Chapter 7
159
Chapter 7
of the spots is increasing passing from 24 h to 48 h. It can be assumed that due to the earlier
retention time, the TAGs in the pulp extract are represented by more unsaturated species. A
second observation can be made on the overall lipid profile of the pulp of different origins. The
pulp from the German Forastero showed a different profile from the pulp of PBC 140 and PBC
159.
Figure 7.16 2D map of the LC-MS chromatogram in positive ion mode of the lipid extract from
pulp at 24 and 48h from 3 different hybrids from 2 origins: Ivory Coast and Malaysia. At the upper
part, same 2D map* left and right side of a reference chromatogram from Ivory Coast. *these 2D maps are
inserted for comparison purposes.
7.4 Discussion
Oscillations of the lipid amount during the fermentation were observed in all the
fermentation trials, and this phenomenon was not reported previously. As presented in Chapter
160
Chapter 7
5, the lipid amount showed an increase in fermented dried beans compared to unfermented wet
cocoa. Although that appeared as a trend, considerable variability among the samples was
present. However, in this study, such a conclusion can be drawn only for Malaysian
fermentation trials. Contradictory to our findings, Afoakwa et al. revealed that the fat content
of the samples decreased significantly with fermentation and pod storage in samples from
Ghana and Aremu et al. also found a decrease from 62.9% to 55.7% by the sixth day of
fermentation in Nigerian samples (Afoakwa, Quao, Takrama, Budu, & Saalia, 2013; Aremu,
Agiang, & Ayatse, 1995).
During the fermentation, the beans were turned at 48 h and 96 h in trials from Malaysia and
Ivory Coast and the beans from Ecuador were turned every day. The decrease of all
monounsaturated TAGs at 72 h in almost all fermentation trials, could suggest a consequence
of the aerobic phase. Only the trials from Ivory Coast showed the decrease happening at 96 h,
indicating perhaps a one-day delay in the fermentation process. At this point of fermentation,
the acetic acid bacteria started to convert ethanol to acetic acid and to penetrate into bean,
causing biochemical reactions. De Britto et al. observed maximum alterations of the cell wall
and reduction in the cytoplasmic content of phenolic compounds and protein bodies at 72 h (de
Brito Edy S et al., 2000). The decrease of the most abundant TAG profile could happen as a
response of the beans against the exponential increase of the bacterial reproduction by using
intensively the fat resources to produce secondary metabolite aimed at defense. Due to the
increased amount of ethanol, lactic acid, acetic acid, and temperature increment, it could be
speculated that the main TAGs undergo hydrolysis, however this will not explain the increase
of diunsaturated TAGs containing oleic acid observed in all fermentation trials. Therefore, it
could be deduced that the bacteria involved in the fermentation use the lipid from cocoa as
substrate for building their own lipids. On the other hand, microorganisms and germinating oil
seeds convert lipid to carbohydrate via gluconeogenesis, employing the glyoxylate cycle
(Itamar Luís Gonçalves, Albanin Aparecida Mielniczki-Pereira, Ana Claudia Piovezan Borges,
& Alice Teresa Valduga, 2016). At 96 h, a second turning of the beans occurs which could
explain the variation in the TAG profile observed in the second part of the fermentation. It
appears that the turning of the beans perturbs the trend of rising or declining of the di- and
triunsaturated TAGs. CCN 51 Ecuadorian hybrid has shown an increase of the mono, di, tri,
and saturated fatty acids only at 72 h, with POO and SOO continuing at this level for another
24 h. This hybrid become one of the most planted cultivars in Ecuador, mainly because of its
high productivity and disease resistance (Boza et al., 2014).
161
Chapter 7
A rapid decrease of unusual TAGs during the first 4 days of fermentation was observed.
Hydroxyl-containing TAGs are known to be produced during the germination (Weichert,
Kolbe, Kraus, Wasternack, & Feussner, 2002) catalyzed by lipoxygenases. Moreover, this
compounds are involved in the synthesis of cutin derived mostly from oleic fatty acid (Blée &
Schuber, 1993). Since the fermentation process inhibits many of the endogenous enzymes
(Hansen Carl E, del Olmo Margarita, & Burri Christine, 1999) it could inhibit LOX; thus, no
production of this enzyme occurs after the second phase of the fermentation. Besides, Sadik et
al. showed that flavonoids of cocoa inhibited mammalian LOX (Sadik, Sies, & Schewe, 2003).
Unusual TAGs were consumed which might indicate that they rapidly undergo degradation or
these compounds are redirected into Jasmonic acid synthesis pathway to the increase the
defense response (Ghasemi Pirbalouti Abdollah, Sajjadi Seyed Ebrahim, & Parang Keykavous,
2014; Mosblech, Feussner, & Heilmann, 2009). Two batched of the same hybrid from Ecuador
showed different amount of the hydroxyl TAGs but the same decreasing trend during the
fermentation. A control experiment, applying a thermal treatment of the beans at 45 °C for 48
h in aerobic condition, has shown no decrease in the amount of these TAGs. Therefore, a higher
amount in the usual TAGs might indicate the germination status of the beans. The zigzag
shaped reduction of the unusual TAGs might be explained by the exposure of the beans at
different types of bacteria during the first phase of the fermentation or sampling inhomogeneity.
Statistical analysis supported the trend of all the TAGs.
FAME analysis has shown an overall increase of the fatty acids during the fermentation. A
remarkable rise in linoleic acid starting at 48 h in spontaneous fermentation and at 24 h in
controlled fermentation followed by a decrease until reaching the initial levels in the last day
of the fermentation in Ivorian samples. In other studies, fermentation of castor oil showed that
the fermented seeds contained more unsaturated fatty acids than the unfermented seeds
(Anosike & Egwuatu, 1980). Whereas during the fermentation of must of Vitis vinifera fatty
acids were consumed, by the time that 50% of the sugar was used up, with the high
consumption of large-chain saturated and unsaturated fatty acids (Ancín, Ayestarán, García, &
Garrido, 1998).
The 2D map visualization of the chromatographic data, a tool embedded in the analytical
software, offered a direct visual display of the changes occurring in the lipid profile of the
cocoa and pulp extracts. 2D map of minor components showed correlations with the
quantification data. Key differences in the minor lipid profile were observed at the same time
when significant changes in the main TAGs occurred. The evolution of minor components is
162
Chapter 7
described by the loss of hydroxyl derivative of TAGs and the appearance at 72 h, and 96 h of
phospholipidic class, which is most probably of a different provenience. 2D map of
spontaneous fermentation illustrated the presence of the phospholipid class already at 24 h
(Appendix Fig. E.1). Additional analysis of the pulp showed a very different lipid profile if
compared to cocoa bean lipid profile. The major lipid class in the pulp is represented by more
unsaturated and shorter chain fatty acids TAGs, characteristic of yeast and bacterial origin
(Botha & Kock, 1993; O’Leary, 1962). From the 2D map of the pulp lipid profile, it can be
noted a prominent variation in the bacterial strains from two different origins. Badrie et al.
reviewed the microbiological diversity related to the origin of the beans (Badrie et al., 2015).
The profile of TAGs going up during the first 48 h and then decreasing, follows the profile of
Yeasts and Lactobacillus species which, increase reaching the max amount at around 36 h
(Schwan & Wheals, 2004). Results indicate that the major transformation in the lipid profile
of cocoa samples occurred when acetic acid bacteria and Bacillus species became the dominant
microorganisms (Ardhana & Fleet, 2003; Lima et al., 2011). They remained at high population
until the fermentations were considered complete. Studies showed that acetic acid bacteria
promote lipid accumulation in yeast which is mainly constituted by palmitic acid (C16:0),
stearic acid (C18:0), oleic acid (C18:1), and linoleic acid (C18:2). Lipids produced in media
containing a high concentration of acetic acid had 19% C16:0, 16% C18:0, 44% C18:1, and
10% C18:2 (Huang et al., 2016). Moreover, a special class of lipids found in acetic acid
bacteria, dihydroceramides, showed to improve cognitive function in dementia model rats
(Fukami, Tachimoto, Kishi, Kaga, & Tanaka, 2010). Other studies showed that during storage
the amount of linoleic and α-linolenic acids increased in the probiotic cream fermented with
B. lactis compared to the control cream (Yilmaz-Ersan, 2013). Moreover, many studies
reported the capacity of the yeasts to synthetize cocoa lipid-like compound (Papanikolaou,
Muniglia, Chevalot, Aggelis, & Marc, 2003; Wei, Gossing, Bergenholm, Siewers, & Nielsen,
2017). The fatty acids composition of the lipids in yeast changes with temperature: the lower
the temperature, the more unsaturated the membrane fatty-acyl composition; while the higher
the temperature, the more medium-chain fatty acids (C8:0-C12:0) are produced influencing the
volatile compounds produced during the fermentation (Torija et al., 2003). Giving this fact, no
correlation between the temperature during the fermentation and the amount of the
polyunsaturated TAGs could be perceived. Although, substantial variations occurring in the
lipid profile of all fermentation trials have happened at 72 and 96 h time points when the batch
temperature reaches its maximum (see Appendix, Fig. E.5).
163
Chapter 7
Nevertheless, further investigations are needed to elucidate the lipid profile of the pulp at
different stages of the fermentation process and to correlate the values with the occurring
fermentative microbiota. An important research question is to consider and evaluate up to
which rate the bacteria involved in the fermentation causes the remodeling of the fatty acid
profile and up to which degree environmental factor and hybrid type. The results showed that
the ecological conditions highly dictate the lipid profile. Nonetheless, understanding the factors
contributing to variations in the chemical composition and physical qualities of cocoa beans
during the fermentation processes would have significant commercial implications.
7.5 Conclusions
This study showed the impact of the fermentation on the lipid profile of cocoa. The results
of this chapter could put the basis for further understanding of the effects of the microbial
activity in cocoa during fermentation since as it could be seen, already from the first hours of
fermentation changes of the lipid profile occurred. Although not substantial alterations happen
in the main TAGs profile at the end of the fermentation, the microorganisms involved in the
fermentation shaped the profile of the less abundant lipid compounds. This fact helps to explain
the findings of Chapter 5 regarding the POO, PLS, SOO, and SLS irregular trend. Therefore,
a part of the temperature at the origin during fruit ripening a second microbiological factor
must be considered.
References
https://doi.org/10.1080/10408390701719272
Afoakwa, E. O., Quao, J., Takrama, J., Budu, A. S., & Saalia, F. K. (2013). Chemical composition and physical
quality characteristics of Ghanaian cocoa beans as affected by pulp pre-conditioning and fermentation. Journal
of Food Science and Technology, 50(6), 1097–1105. https://doi.org/10.1007/s13197-011-0446-5
Anosike, E. O., & Egwuatu, C. K. (1980). Biochemical changes during the fermentation of castor oil (Ricinus
communis) seeds for use as a seasoning agent. Plant Foods for Human Nutrition, 30(3–4), 181–185.
https://doi.org/10.1007/BF01094021
164
Chapter 7
4337.12180
Ardhana, M. M., & Fleet, G. H. (2003). The microbial ecology of cocoa bean fermentations in Indonesia.
International Journal of Food Microbiology, 86(1–2), 87–99. https://doi.org/10.1016/S0168-1605(03)00081-3
Aremu, C. Y., Agiang, M. A., & Ayatse, J. O. I. (1995). Nutrient and antinutrient profiles of raw and fermented
cocoa beans. Plant Foods for Human Nutrition, 48(3), 217–223. https://doi.org/10.1007/BF01088443
https://doi.org/10.1080/10408398.2012.669428
Blée, E., & Schuber, F. (1993). Biosynthesis of cutin monomers: Involvement of a lipoxygenase/peroxygenase
pathway. Plant Journal, 4(1), 113–123.
Botha, A., & Kock, J. L. F. (1993). Application of fatty acid profiles in the identification of yeasts. International
Journal of Food Microbiology, 19(1), 39–51. https://doi.org/10.1016/0168-1605(93)90122-W
de Brito Edy S, García Nelson H Pezoa, Gallão M I, Cortelazzo Angelo L, Fevereiro Pedro S, & Braga Márcia R.
(2000). Structural and chemical changes in cocoa (Theobroma cacao L) during fermentation, drying and roasting.
Journal of the Science of Food and Agriculture, 81(2), 281–288. https://doi.org/10.1002/1097-
0010(20010115)81:2<281::AID-JSFA808>3.0.CO;2-B
De Vuyst L., & Weckx S. (2016). The cocoa bean fermentation process: from ecosystem analysis to starter culture
development. Journal of Applied Microbiology, 121(1), 5–17. https://doi.org/10.1111/jam.13045
Dsouza, R., Grimbs, S., Behrends, B., Bernaert, H., Ullrich, M., & Kuhnert, N. (2017). Origin-based polyphenolic
fingerprinting of Theobroma cacao in unfermented and fermented beans. Food Research International.
Elvert, M., Boetius, A., Knittel, K., & Jørgensen, B. B. (2003). Characterization of Specific Membrane Fatty Acids
as Chemotaxonomic Markers for Sulfate-Reducing Bacteria Involved in Anaerobic Oxidation of Methane.
Geomicrobiology Journal, 20(4), 403–419. https://doi.org/10.1080/01490450303894
Feussner, I., Kühn, H., & Wasternack, C. (2001). Lipoxygenase-dependent degradation of storage lipids. Trends
in Plant Science, 6(6), 268–273. https://doi.org/10.1016/S1360-1385(01)01950-1
Fukami, H., Tachimoto, H., Kishi, M., Kaga, T., & Tanaka, Y. (2010). Acetic Acid Bacterial Lipids Improve
Cognitive Function in Dementia Model Rats. Journal of Agricultural and Food Chemistry, 58(7), 4084–4089.
https://doi.org/10.1021/jf9045842
165
Chapter 7
Ghasemi Pirbalouti, A., Sajjadi, S. E., & Parang, K. (2014). A Review (Research and Patents) on Jasmonic Acid
and Its Derivatives. Archiv Der Pharmazie, 347(4), 229–239. https://doi.org/10.1002/ardp.201300287
Ghasemi Pirbalouti Abdollah, Sajjadi Seyed Ebrahim, & Parang Keykavous. (2014). A Review (Research and
Patents) on Jasmonic Acid and Its Derivatives. Archiv Der Pharmazie, 347(4), 229–239.
https://doi.org/10.1002/ardp.201300287
Hansen Carl E, del Olmo Margarita, & Burri Christine. (1999). Enzyme activities in cocoa beans during
fermentation. Journal of the Science of Food and Agriculture, 77(2), 273–281.
https://doi.org/10.1002/(SICI)1097-0010(199806)77:2<273::AID-JSFA40>3.0.CO;2-M
Ho, V. T. T., Zhao, J., & Fleet, G. (2014). Yeasts are essential for cocoa bean fermentation. International Journal
of Food Microbiology, 174, 72–87. https://doi.org/10.1016/j.ijfoodmicro.2013.12.014
Huang, X.-F., Liu, J.-N., Lu, L.-J., Peng, K.-M., Yang, G.-X., & Liu, J. (2016). Culture strategies for lipid
production using acetic acid as sole carbon source by Rhodosporidium toruloides. Bioresource Technology, 206,
141–149. https://doi.org/10.1016/j.biortech.2016.01.073
Itamar Luís Gonçalves, Albanin Aparecida Mielniczki-Pereira, Ana Claudia Piovezan Borges, & Alice Teresa
Valduga. (2016). &60;b&62;Metabolic modeling and comparative biochemistry in glyoxylate cycle. Acta
Scientiarum &58; Biological Sciences, 38(1), 1–6. https://doi.org/10.4025/actascibiolsci.v38i1.24597
International, 82(Supplement C), 44–52. https://doi.org/10.1016/j.foodres.2016.01.012
Lima, L. J. R., Almeida, M. H., Rob, N., & Zwietering, M. H. (2011). Theobroma cacao L., “the food of the gods”:
Quality determinants of commercial cocoa beans, with particular reference to the impact of fermentation. Critical
Reviews in Food Science and Nutrition, 51(8), 731–761. https://doi.org/10.1080/10408391003799913
Mosblech, A., Feussner, I., & Heilmann, I. (2009). Oxylipins: Structurally diverse metabolites from fatty acid
oxidation. Plant Physiology and Biochemistry, 47(6), 511–517. https://doi.org/10.1016/j.plaphy.2008.12.011
O’Fallon, J. V., Busboom, J. R., Nelson, M. L., & Gaskins, C. T. (2007). A direct method for fatty acid methyl
ester synthesis: application to wet meat tissues, oils, and feedstuffs. Journal of Animal Science, 85(6), 1511–1521.
https://doi.org/10.2527/jas.2006-491
O’Leary, W. M. (1962). THE FATTY ACIDS OF BACTERIA. Bacteriological Reviews, 26(4), 421–447.
Retrieved from https://www.ncbi.nlm.nih.gov/pmc/articles/PMC441163/
166
Chapter 7
Papanikolaou, S., Muniglia, L., Chevalot, I., Aggelis, G., & Marc, I. (2003). Accumulation of a Cocoa-Butter-
Like Lipid by <Emphasis Type="Italic">Yarrowia lipolytica</Emphasis> Cultivated on Agro-Industrial Residues.
Current Microbiology, 46(2), 0124–0130. https://doi.org/10.1007/s00284-002-3833-3
Patras, M. A., Milev, B. P., Vrancken, G., & Kuhnert, N. (2014). Identification of novel cocoa flavonoids from
raw fermented cocoa beans by HPLC–MSn. Food Research International, 63(Part C), 353–359.
Rosahl, S. (2014). Lipoxygenases in Plants -Their Role in Development and Stress Response. Zeitschrift Für
Naturforschung C, 51(3–4), 123–138. https://doi.org/10.1515/znc-1996-3-401
Sadik, C. D., Sies, H., & Schewe, T. (2003). Inhibition of 15-lipoxygenases by flavonoids: structure–activity
relations and mode of action. Biochemical Pharmacology, 65(5), 773–781. https://doi.org/10.1016/S0006-
2952(02)01621-0
https://doi.org/10.1080/10408690490464104
Torija, M. J., Beltran, G., Novo, M., Poblet, M., Guillamón, J. M., Mas, A., & Rozès, N. (2003). Effects of
fermentation temperature and Saccharomyces species on the cell fatty acid composition and presence of volatile
compounds in wine. International Journal of Food Microbiology, 85(1), 127–136. https://doi.org/10.1016/S0168-
1605(02)00506-8
Turner, J. G., Ellis, C., & Devoto, A. (2002). The Jasmonate Signal Pathway. The Plant Cell, 14(Suppl), s153–
s164. https://doi.org/10.1105/tpc.000679
Wei, Y., Gossing, M., Bergenholm, D., Siewers, V., & Nielsen, J. (2017). Increasing cocoa butter-like lipid
production of Saccharomyces cerevisiae by expression of selected cocoa genes. AMB Express, 7(1), 34.
https://doi.org/10.1186/s13568-017-0333-1
Weichert, H., Kolbe, A., Kraus, A., Wasternack, C., & Feussner, I. (2002). Metabolic profiling of oxylipins in
germinating cucumber seedlings – lipoxygenase-dependent degradation of triacylglycerols and biosynthesis of
volatile aldehydes. Planta, 215(4), 612–619. https://doi.org/10.1007/s00425-002-0779-4
Yilmaz-Ersan, L. (2013). Sastav masnih kiselina vrhnja fermentiranog probiotičkim bakterijama. Mljekarstvo :
Časopis Za Unaprjeđenje Proizvodnje i Prerade Mlijeka, 63(3), 132–139. Retrieved from
https://hrcak.srce.hr/106561
Ziegleder Gottfried. (2017). Flavour development in cocoa and chocolate. Beckett’s Industrial Chocolate
Manufacture and Use. https://doi.org/10.1002/9781118923597.ch8
167
Chapter 8
Chapter 8:
8 Evolution of cocoa lipids across factory production

process
8.1 Introduction
After the fermentation and drying, the manufacture of cocoa liquor involves roasting,
another important stage of the cocoa processing. During roasting, evaporation of volatile acids
occurs, and the moisture content reduces (Afoakwa, Paterson, Fowler, & Ryan, 2008;
Aprotosoaie, Luca, & Miron, 2016; Beckett, Harding, & Freedman, 2008). The degree of
roasting is a time/temperature dependent process, and it ranges between 5 to 120 min (usually
10 to 35) and 110 to 150 °C (Afoakwa et al., 2008; Aprotosoaie et al., 2016; Kongor et al.,
2016). Roasting process provide ideal conditions for the Maillard reactions, where small
molecules as amino acids, short chain peptides, and reducing sugars, produced during the
fermentation and drying, undergo important chemical reactions having an impact on the final
flavour of cocoa (Counet, Ouwerx, Rosoux, & Collin, 2004; Frauendorfer & Schieberle, 2008).
Roasting can be achieved by 3 methods: whole bean roasting, nib roasting, and liquor roasting
(Afoakwa et al., 2008; Aprotosoaie et al., 2016).
In 1828 Coenraad Van Houten combined alkaline salts with cocoa powder (Fig. 8.1), which
resulted in reduction of acidity (natural cocoa is typically 5.0 –6.0 pH) reducing the bitterness
and sourness of the chocolate, making the drink more palatable. It improved the dispersibility
of cocoa powder in liquid and enhancement of the, giving it the appearance of a more intense
flavor and darker color. Coenraad Van Houten patented his cocoa alkalizing discovery in
Netherlands in 1828, hence the phrases dutched cocoa or alkalized cocoa are now used
synonymously in today’s chocolate industry (Beckett, 2011; Coe & Coe, 2007). Nowadays,
alkalization is carried out primarily to change the colour and to improve the flavour of cocoa
powder. It reduces astringency by polymerization of flavonoids during alkali treatment
(Afoakwa et al., 2008). Alkalization occur at 3 different process stages: nib alkalization, cocoa
168
Chapter 8
cake alkalization and cocoa liquor alkalization. Each process has distinct advantages and
disadvantages. Nib and cocoa liquor alkalization will have an effect on cocoa butter.
Figure 8.1 Alkalized ingredients used in alkalization of cocoa adapted from (“Published Articles |
Blommer Chocolate Company,” ).
In the first part of this chapter, the triacylglycerol profile of dried beans and their
correspondent cocoa liquor were analyzed. The aim of this study was to determine the impact
of the alkalization on the lipid profile. The second part of this chapter describes structure
elucidation of markers of alkalization process. Quantification data of these compounds can be
found in Audrey’s Bergounhou master thesis, achieved under author’s supervision. Correlation
studies between the markers of dutching process and physical properties, and samples
specifications are described in Appendix I.
(Rotisolv® HPLC grade) and Tetra-dodecylammonium bromide was purchased from Carl Roth
and formic acid (puriss., ≥ 98% (T) for mass spectrometry) were purchased from Sigma-Aldrich
Chemie (Steinheim, Germany). Ethanol was subjected to distillation prior use.
169
Chapter 8
Dried cocoa beans samples and cocoa liquor sample from 5 different origins were received
in several sets from Barry Callebaut. A total of 7 pairs of samples coming from Cuba, San
Thome, Madagascar, Tanzania, and Venezuela were analyzed. Firstly, the seed material was
deshelled and ground using a grinder with the purpose of making homogenous powder.
Henceforth, an overnight Soxhlet (Buchi Extraction System B811 instrument, Flawil,
Switzerland) method using dichloromethane as an extraction solvent, 150 mL of 5 g of
powdered beans, was used. Cocoa liquors were treated in the same way. Extracted lipids were
stored at -20 °C until further analysis. For HPLC analysis, a concentration of 0.045 mg/mL in
chloroform/ethanol (50/50) of cocoa lipid extract was prepared.
spectrometer (Bruker Daltonics UHT Ultra, Bremen, Germany) fitted with an ESI source used
as the detector with the following parameter settings: capillary voltage of 4.5 kV; nebulizing
gas pressure of 2 Ba; drying gas flow rate of 10 L/min; drying gas temperature of 220 °C. ESI
mass spectra were measured in the range of m/z 200-1200 in the positive ion mode. Internal
170
Chapter 8
quadratic mode.
8.2.5 DSC analysis
Thermal measurements were made using a Mettler Toledo DSC 822 instrument (Barcelona,
Spain). A sealed aluminum pan containing n-eicosane was used as reference, and temperature
and enthalpy were calibrated using cyclohexane and indium. The lipid samples (10 mg) were
sealed in aluminum pans. Measurements were performed as follows: initially, the temperature
was increased from 20 °C to 60 °C at the rate of 20 K/min to erase any thermal memory; then
holding at 60 °C for 3 min; after this period, crystallization was carried out by cooling from 60
°C to -10 °C at a rate of -10 K/min; then, after holding for 3 min at -10 °C melting thermograms
were obtained by heating the samples to 60 °C at 10 K/min. Results were analysed using
Mettler Toledo Star software.
and retention time alignment using a mass-to-charge (m/z) tolerance of Δm/z = 0.0001 and a
retention time (Rt) tolerance of 0.5 min for elution time <20 min and 1.0 min for an elution
time ≥20 min. Extensive manual curation of the peaks could allow the assignment with m/z and
Rt tolerances of Δm/z = 0.01 and ΔRt = 0.8 min respectively. For further analyses, the peak
areas were normalized to the internal standard (Tetra-dodecylammonium bromide) at Rt 4.4
min and consequently all the ratio were normalized to 100. Thus, data are presented as
percentage.
Microsoft Office Excel 2016 was used to order and process the data. A one-factor and a two-
factors ANOVA were used to study the effect of origin and the effect of fermentation status on
fatty acid profile. The analysis were carried out with XLSTAT Evaluation 2018.2 software
version 2018 (Addinsoft, France).
171
Chapter 8
8.3.1 Lipid content in dried beans and cocoa liquor
The roasting process have shown to affect the total lipid amount (Fig. 8.2). Only the samples
from San Thome did not show variation in the lipid content after roasting. In all other samples,
the total lipid amount decreases from bean to liquor. Two explanation can be drawn, first, the
roasting affect the extractability of the lipid amount. Studies showed that the roasting process
increase the extractability of the butter (Asep et al., 2008). Secondly, a loss of lipid content
could occur due to some chemical reactions of an unknown nature involving the lipids that
might take place during the roasting and grinding processes.
Figure 8.2 Lipid amount in cocoa beans and cocoa liquors from different origin.
8.3.2 TAGs profile of dried beans versus roasted beans in form of cocoa liquor
Monounsaturated TAGs showed an overall decrease with the roasting of the beans (Fig. 8.3,
Table F.3 Appendix). However, not all samples exhibit the same behavior. The most abundant
TAG present a variable behavior. No trend could be delineated, 5 of 7 samples displayed a
decrease of POS.
172
Chapter 8
Figure 8.3 Profile amount of monounsaturated TAGs in cocoa beans and cocoa liquors from
different origin.
Di- and triunsaturated TAGs total amount revealed an increasing trend after roasting.
Notwithstanding no distribution pattern of the individual unsaturated TAGs could be observed
(Fig. 8.4, Table F.4 Appendix). The ratio between di- and triunsaturated/monounsaturated
TAGs seems to increase after roasting, whereas the ratio saturated/monounsaturated TAGs has
a random comportment (Fig. 8.5). It is worth remarking that form the graph we can grasp the
similarities and differences between the samples. Thus, samples from Cuba have higher amount
of saturated TAGs, while Tanzanian samples own higher amount of di- and triunsaturated
TAGs.
Figure 8.4 Profile amount of di- and triunsaturated TAGs in cocoa beans and cocoa liquors from
different origin.
173
Chapter 8
Figure 8.5 Ratios between di- and triunsaturated TAGs and ratios between saturated and
monounsaturated TAGs in cocoa beans and cocoa liquors from different origin.
Hydroxyl-TAGs were found to be markers of the fermentation status of the beans (Chapter
7). Consequently, the presence of these compounds in the cocoa samples might indicate that
incomplete fermentation of the beans (Fig. 8.6, Table F.5 Appendix). Only the samples from
Cuba and San Thome have shown lower amount of hydroxyl-TAGs. Figure 8.7 illustrate a
complete random nature of saturated TAGs (Table F.1 Appendix).
Figure 8.6 Profile amount of hydroxyl-TAGs in cocoa beans and cocoa liquors from different
origin.
174
Chapter 8
Figure 8.7 Profile amount of saturated TAGs in cocoa beans and cocoa liquors from different
origin.
8.3.3 DAGs profile of dried beans versus roasted beans in form of cocoa liquor
In the same order, DAGs did not show a particular demeanor. In 4 origins out of 7 the
amount decreases with the bean roasting, whereas in 2 samples DAG increases with roasting
(Fig. 8.8, Table F.2 Appendix).
Figure 8.8 Profile amount of DAGs in cocoa beans and cocoa liquors from different origin.
175
Chapter 8
8.3.4 Correlations studies on TAG and DAG profiles
Correlation studies have shown that there is a significant negative correlation between the
amount of monounsaturated TAGs and the amount of di and tri-unsaturated TAGs, which
indicated that there is an inverse proportion of these 2 subclasses of TAGs. Saturated TAGs
have shown a positive correlation with monounsaturated TAGs and a negative correlation with
hydroxyl- TAGs and di- and triunsaturated TAGs.
Table 8.1 Correlation values of the TAG lipid classes, DAG, and total lipid amount.
Assuming that the sampling error is very low, the outcome of the results can be rationalized
as following: a random acyl-rearrangement, or interesterification, on the glycerol molecule
occurred under the high temperature conditions of the roasting process. Interesterification has
long been used to modify modify the physico-chemical and nutritional characteristics of oils
and fats (Willis, Lencki, & Marangoni, 1998). Two types of interesterification are recognised:
chemical and enzymatic (Sreenivasan, 1978). Many studies were conducted to produce cocoa
butter-like fats through interesterification (Abigor et al., 2003; Chang, Abraham, & John,
1990).
176
Chapter 8
8.3.5 Elucidation of the fatty acids amides as markers of alkalization
Separation of lipid compounds from normal and alkalized cocoa butter (CB) by HPLC-MS
(Fig. 8.9) revealed the presence of significant differences in the lipid profile of the less
abundant components. To be noted that the butter underwent SPE fractionation prior analysis.
These compounds were identified as the amide form of the most abundant fatty acids present
in cocoa, namely palmitamide, oleamide, and stearamide (Fig. 8.10). Figure 8.11 illustrates the
tandem MS spectra of oleamide as example.
Figure 8.9 Positive ion mode HPLC-MS chromatogram of normal and alkalized lipid profile.
Figure 8.10 Positive ion mode HPLC-MS chromatogram (zoomed) of alkalized butter and mass
spectra of the new identified compounds.
177
Chapter 8
Fragmentation of oleamide showed the loss of 17 Da corresponding to ammonia and

subsequent water loss. MS3 revealed the typical fatty acid fragmentation, each 14 Da
corresponds to a CH2.
Figure 8.11 Tandem MS spectra of oleamide.
8.3.6 Influence of the fatty acid amides on the crystallization behavior of cocoa butter
To gain more insight into the impact of fatty acids amides (FAA) on the crystallization of
cocoa butter thermal measurements using DSC were performed. Figure 8.12 and 8.13 displays
the thermograms acquired for natural cocoa butter (CB), natural CB mixed with 0.1% fatty
acid amides mix, natural CB plus 1% FAA mix, and natural CB plus 2% FAA mix. The shape
of the DSC curves (Fig. 8.12) indicates a two-step crystallization process when natural butter
was mixed with 1 and 2% of FAA. The complete crystallization of cocoa butter in the presence
of FAA occurred at a lower temperature suggesting that FAA affects crystallization to lower
polymorphic forms such as γ and α. Concerning the melting peaks (Fig. 8.13), the absolute
values were lower for cocoa butter containing the highest amount of FAA. The solid fat content
(SFC), defined as a parameter that expresses the solid/liquid mass relation of a fat at different
temperatures, have shown that an amount between 0.1 and 1% FAA can be related to an
increase of approximately 3% in SFC value at 25 °C (Fig. 8.14). In conclusion, DSC analysis
revealed that FAA has a measurable effect on the crystallization of cocoa butter.
178
Chapter 8
Figure 8.12 DSC crystallization thermograms recorded on cooling of natural CB, natural cocoa
butter (CB), natural CB plus 0.1% fatty acid amides (FAA) mixture, natural CB plus 1% FAA, and
natural CB plus 2% FAA mixture.
Figure 8.13 DSC crystallization melting curves of natural CB, of natural cocoa butter (CB),
natural CB plus 0.1% fatty acid amides (FAA) mixture, natural CB plus 1% FAA, and natural CB plus
2% FAA mixture.
179
Chapter 8
Figure 8.14 Solid fat content (in %) as a function of temperature (°C) of natural cocoa butter (CB),
natural CB plus 0.1% fatty acid amides (FAA) mixture, and natural CB plus 1% FAA mixture.
8.4 Conclusions
Analysis of dried beans and their respective cocoa liquor did not show any pattern on TAGs
and DAGs distribution. However, saturated TAGs have shown a positive correlation with
monounsaturated TAGs. Therefore, a conclusion could be that an interesterification of the fat
or other chemical reactions might occur and this process happens completely random.
Successful elucidation of the compounds responsible for the differences observed between
normal and alkalized cocoa butter, namely fatty acids amides, was achieved. These compounds
were consequently designated as markers of the alkalization process. FAA affect the
crystallization of cocoa butter by promoting the crystallization of lower polymorphic forms
from the melt. Further work is necessary to complete the picture on the physical properties of
cocoa butter containing FAA such as crystallization under isothermal conditions.
180
Chapter 8
References
Abigor, R. D., Marmer, W. N., Foglia, T. A., Jones, K. C., DiCiccio, R. J., Ashby, R., & Uadia, P. O. (2003).
Production of cocoa butter-like fats by the lipase-catalyzed interesterification of palm oil and hydrogenated
soybean oil. Journal of the American Oil Chemists’ Society, 80(12), 1193–1196. https://doi.org/10.1007/s11746-
003-0841-7
https://doi.org/10.1080/10408390701719272
4337.12180
Beckett, Stephen T., Harding, J., & Freedman, B. (2008). The Science of Chocolate (2nd edition). Cambridge,
UK: Royal Society of Chemistry.
Beckett, Steve T. (2011). Industrial Chocolate Manufacture and Use. John Wiley & Sons.
Chang, M.-K., Abraham, G., & John, V. T. (1990). Production of cocoa butter-like fat from interesterification of
vegetable oils. Journal of the American Oil Chemists’ Society, 67(11), 832–834.
https://doi.org/10.1007/BF02540501
Hudson.
Counet, C., Ouwerx, C., Rosoux, D., & Collin, S. (2004). Relationship between Procyanidin and Flavor Contents
of Cocoa Liquors from Different Origins. Journal of Agricultural and Food Chemistry, 52(20), 6243–6249.
https://doi.org/10.1021/jf040105b
Frauendorfer, F., & Schieberle, P. (2008). Changes in Key Aroma Compounds of Criollo Cocoa Beans During
Roasting. Journal of Agricultural and Food Chemistry, 56(21), 10244–10251. https://doi.org/10.1021/jf802098f
International, 82(Supplement C), 44–52. https://doi.org/10.1016/j.foodres.2016.01.012
181
Chapter 8
Published Articles | Blommer Chocolate Company. (n.d.). Retrieved March 23, 2018, from
http://www.blommer.com/articles.php
Sreenivasan, B. (1978). Interesterification of fats. Journal of the American Oil Chemists’ Society, 55(11), 796–
805. https://doi.org/10.1007/BF02682651
Willis, W. M., Lencki, R. W., & Marangoni, A. G. (1998). Lipid Modification Strategies in the Production of
Nutritionally Functional Fats and Oils. Critical Reviews in Food Science and Nutrition, 38(8), 639–674.
https://doi.org/10.1080/10408699891274336
182
Chapter 9
Chapter 9:
9 Minor lipid compounds in cocoa
9.1 Introduction
Cocoa butter is mainly constituted by triacylglycerols (TAGs) (95-98%) and a complex

mixture of minor components (2-5%) of a diverse chemical nature. These compounds show a
wide qualitative and quantitative composition, depending on the species, climatic conditions,
pre- and the post-fermentation process of the beans (Cert, Moreda, & Pérez-Camino, 2000;
Lipp et al., 2001; Lipp & Anklam, 1998).
The main groups of minor constituents present in a vegetable oil and animal fat are: long-
chain aliphatic and triterpenic alcohols, methyl sterols, sterols, waxes, sterol esters,
tocopherols, tocotrienols, pigments, diacylglycerols (DAGs), monoacylglycerols (MAGs),
glycolipids, phospholipids and triterpenic acids. This group encloses numerous species with a
broad range of polarities, concentration and chemical structures and functions; therefore,
characterization of these compounds is quite tricky. The required methods imply isolation and
analysis of minor constituents from a lipid extract through several steps of separation,
identification, and quantification (Christie & Han, 2010).
Until this moment, the research studies of this thesis were mostly focused on the analysis of
samples coming from the post-harvesting process, wet unfermented and dried fermented beans
with a major attention on triacylglycerols identification. Identification of the minor lipid
component is still at the early stage and the results shown here must be considered as
preliminary and not yet comprehensive.
To analyses the minor components, a preliminary qualitative and quantitative isolation step
from the TAGs matrix was required. Solid phase extraction (SPE) was applied, as it is a quicker
technique and saves solvent volumes (Kim & Salem, 1990). SPE allowed the separation of 5
fractions (see Chapter 3).
Identification and quantification of individual molecular species required further
chromatography separation and since lipids lack chromophores mass spectrometry (microTOF
183
Chapter 9
MS, Bruker) is used to obtain molecular mass data, structural information, and identification
of compounds. Separation of the individual components that belong to one constituent class by
reversed-phase HPLC using Ethanol as a polar solvent was achieved. Each lipid class has a
unique fragmentation pattern, which aids unambiguous identification of the species eluted from
a column through product-ion MS analysis (IonTrap MS, Bruker).
(Rotisolv® HPLC grade), acetonitrile (Rotisolv® HPLC ultra gradient grade) and chloroform
(Rotisolv® HPLC grade) was purchased from Carl Roth (Karlsruhe, Germany), dichloromethane
99,8% stabilized with amylene for synthesis was purchased from Panreac AppliChem
(Darmstadt, Germany), ammonium formate LC-MS Ultra and formic acid (puriss., ≥ 98% (T)
for mass spectrometry) were purchased from Sigma-Aldrich Chemie (Steinheim, Germany).
Ethanol was subjected to distillation prior use. Standard DAG 1-palmitoyl-2-oleoyl-sn-glycerol
and cholesterol all at ≥ 99% were purchased from Sigma-Aldrich Chemie (Steinheim, Germany).
Frozen unfermented wet cocoa bean samples and fermented dried cocoa beans from
different origins were received in several sets from Barry Callebaut (Lebbeke-Wieze,
Belgium). Firstly, the seed material was crushed using a grinder Retsch Grindomix GM200
knife mill (Haan, Germany) with the purpose of making homogenous powder. Henceforth, an
overnight Soxhlet (Buchi Extraction System B811 instrument) method using dichloromethane
as an extraction solvent, 150 mL of 5 g of crushed beans, was used. Extracted lipids were
stored at -20 °C until further analysis. For HPLC analysis, a concentration of 0.045 mg/mL in
chloroform/ethanol (50/50) of cocoa lipid extract was prepared.
184
Chapter 9

the SPE fractionation. The cartridges were first activated with 20 mL heptane/ethylacetate
(80/20, v/v), then, 20 ml heptane. Then, 10 mg of the lipid extract dissolved in 1 mL of heptane
were applied, after which the fractions were eluted with 5 different solvent mixtures: 1) 20 mL
heptane, first fraction collected ; 2) 20 mL heptane/ethyl acetate (97/3, v/v), for the second
fraction; 3) 20 mL heptane/ethyl acetate (85/15, v/v), third fraction collected; 4) 20 mL
heptane/ethyl acetate (85/15, v/v) plus 1% acetic acid, for the forth fraction; and 5) 20 mL
methanol. Extracted lipids were evaporated to dryness in a rotavapor and stored at -20 °C until
further analysis.
phase was pumped through the column at a flow rate of 0.6 mL/min. Two different elution
gradients were used. One for the total lipid separation and another to separate minor lipid
components in cocoa. The first gradient elution program consisted of holding solvent steady
A/B (40/60) for 3 min; followed by a linear gradient to solvent B (100) for 33 min, and ending
with isocratic elution at solvent B (100) for 10 min. The column was equilibrated at 40/60
solvent A/B for 5 min before reuse. The second gradient elution program consisted of holding
solvent steady A (100) for 5 min; followed by a linear gradient to solvent A/B (70/30) for
another 5; then by a linear gradient to solvent B (100) for 90 min, and ending with isocratic
elution at solvent B (100) for 10 min. The column was equilibrated at 100 solvent A for 5 min
before reuse.
185
Chapter 9
High-resolution masses were acquired using a time of flightMicrOTOF Focus mass

spectra were measured in the range of m/z 200-1200 in the positive ion mode. Internal
quadratic mode. Molecular formulae suggestions were accepted if the mass error was below 5
ppm.
9.2.6 Tandem Mass spectrometry conditions
Minor components have been separated from the major components in 4 fractions: fraction
1, fraction 3, fraction 4 and fraction 5. Fraction 2 represents the TAGs (see Chapter 3, Fig. 3.6),
and they constitute major components, the usual TAGs found in cocoa butter and most
abundant ones. These TAGs constitute the macro components of cocoa butter, and they have
been extensively characterized in Chapter 4.
186
Chapter 9
9.3.1 Molecular species fraction 1
Highly lipophilic compounds characterize fraction 1 (Fig. 9.1). So far, one class of
compounds could be identified in this fraction, which are sterol esters. Assignment of
individual sterol ester structures was based on ESI-tandem MS (MSn) measurements.
components of the SPE lipid extract from fermented dried beans. Red circle indicates the sterol esters
class.
Collision induced dissociation of [M+NH4]+ precursor ions resulted in the neutral loss of
297.3 Da, acyl group and ammonium adduct, generating a sterol fragment ion at m/z 383.4 and
m/z 397.4 for the 2 sterol esters respectively (Fig. 9.2 (a) and (b)). In the absence of sterol ester
standard, standard cholesterol was used as surrogate standard for comparison (Fig. 9.2 (c)).
MS3 of the two illustrated sterol esters, 18:2 Campesteryl ester and 18:2 Sitosteryl ester,
showed very similar fragmentation pattern if compared to MS2 of the standard. Herein, 5
compounds were identified.
187
Chapter 9
Figure 9.2: Positive ion mode tandem mass spectrum of sterol esters (a) 18:2 Campesteryl ester (m/z
680.7) ammonia adducts, (b) 18:2 Sitosteryl ester (m/z 694.6) ammonia adducts, (c) cholesterol
standard (m/z 369.2) corresponding to [M-OH]+ molecular ion.
A different profile was observed when comparing the chromatograms of the fraction 3 from
unfermented wet cocoa beans and fermented dried cocoa beans (Fig. 9.3). The difference is
attributable to the unusual TAGs in unfermented wet beans, which have been described as
hydroxyl TAG derivatives and thus, labeled as markers of the fermentation status of the beans
(see Chapter 4 and 5). Other classes were identified in fraction 3 and show no substantial
variations between the two types of cocoa beans, which are diacylglycerols (DAGs) and waxes.
188
Chapter 9
cocoa beans. Blue circle indicates the hydroxyl-TAGs, green circle indicates the diacylglycerol class,
and magenta circle indicates waxes class.
Concerning DAGs, two peaks at different retention time were observed (Fig. 9.4 (a)) thus,
indicating the occurrence of two regioisomers 1,2-acyl-sn-glycerol and 1,3-acyl-sn-glycerol.
In order to determine in which sequence the two regioisomers elute on the reversed-phase
column, 1-palmitoyl-2-oleoyl-sn-glycerol standard was injected and tandem MS of the
standard and the DAGs containing palmitic and oleic acid esterified to the glycerol backbone,
m/z 577.5, were studied (Fig. 9.4 (b)). The m/z 577.5 corresponds to [M-H2O-NH4]+ molecular
ion and showed higher intensity than the precursor ion. By taking into account the retention
time of the standard, it could be deduced that the first peak eluting at 27.6 min corresponds to
1-palmitoyl-2-oleoyl-sn-glycerol. However, such conclusion could not be inferred from MS2
fragments since the two regioisomers displayed same fragmentation pattern. Notably, adduct
ions of 1-palmitoyl-2-oleoyl-sn-glycerol and 1-palmitoyl-3-oleoyl-sn-glycerol have shown
different stability at given MS conditions (Fig. 9.5). From the MS spectrum, it can be supposed
that 1-palmitoyl-3-oleoyl-sn-glycerol is more stable because the molecular ion [M+NH4]+ is
present in higher intensity but entirely absent for 1-palmitoyl-2-oleoyl-sn-glycerol. Sodium
adduct is present for both regioisomers.
189
Chapter 9
Identification of DAGs was carried out using a high-resolution ESI-TOF mass spectrometer
operating in the positive ion mode leading to assignment of molecular formulae. Molecular
formulae were accepted if only the mass error was found to be below 5 ppm. LC-ESI-TOF MS
method gave two pseudomolecular ions, including [M-H2O-NH4]+, and [M+Na]+, as the most
common base peak with [M+Na]+ being the most abundant. Figure 9.6 and Table 9.1 illustrates
the HPLC-MS extract ion chromatogram in positive ion mode of all diacylglycerols identified
in cocoa lipid extract. Moreover, lipidmaps database was used to compare experimental m/z
values with database entries to confirm the masses assigned.
Besides DAGs, wax lipid compounds were identified in fraction 3 by tandem MS. As an
example, heptadecyl-palmitate, m/z 512.5 [M+NH4]+, produced MS2 base peak at m/z 283.07
and m/z 256.06, showing a neutral loss of 230 and 256 Da respectively (Fig. 9.7). 256 Da
corresponds to the loss of palmitic acid.
Figure 9.4 (a) HPLC-MS extract ion chromatogram of m/z 577.5 in positive ion mode corresponding
to palmitoyl-oleoyl-sn-glycerol and (b) tandem mass spectrum displaying the fragmentation of 2
regioisomers and tandem MS spectrum of the standard 1-palmitoyl-2-oleoyl-sn-glycerol.
190
Chapter 9
Figure 9.5 MS spectrum illustrating the adduct ions of 1-palmitoyl-2-oleoyl-sn-glycerol, 1-palmitoyl-

3-oleoyl-sn-glycerol from cocoa lipid extract, and of 1-palmitoyl-2-oleoyl-sn-glycerol standard
compound.
Figure 9.6 HPLC-MS extract ion chromatogram in positive mode of diacylglycerols identified in
cocoa lipid extract.
191
Chapter 9
Table 9.1 Diacylglycerol molecular species in cocoa beans lipid extract.
R. Theoretical
Experimental Error Molecular Short
No. time mass Common Name
mass [M+Na]+ ([ppm]) formulae Name
(min) [M+Na]+
1 22.5 641.5124 641.5115 1.4 C39H70NaO5 DAG(18:1/18:2/0:0) 1,2-OL

2 23.1 641.513 641.5115 2.3 C39H70NaO5 DAG(18:1/0:0/18:2) 1,3-OL
3 23 615.4973 615.4959 2.2 C37H68NaO5 DAG(16:0/18:2/0:0) 1,2-PL
4 23.7 615.4969 615.4959 1.6 C37H68NaO5 DAG(16:0/0:0/18:2) 1,3-PL
5 27.1 643.5293 643.5272 3.2 C39H72NaO5 DAG(18:1/18:1/0:0) 1,2-OO
6 27.9 643.528 643.5272 1.3 C39H72NaO5 DAG(18:1/0:0/18:1) 1,3-OO
7 27.7 617.5131 617.5115 2.5 C37H70NaO5 DAG(16:0/18:1/0:0) 1,2-PO
8 28.6 617.5133 617.5115 2.8 C37H70NaO5 DAG(16:0/0:0/18:1) 1,3-PO
9 27.9 643.5267 643.5272 0.8 C39H72NaO5 DAG(18:0/18:2/0:0) 1,2-SL
10 28.8 643.5274 643.5272 0.4 C39H72NaO5 DAG(18:0/0:0/18:2) 1,3-SL
11 28.4 591.4976 591.4959 3 C35H68NaO5 DAG(16:0/16:0/0:0) 1,2-PP
12 29.4 591.4976 591.4959 2.8 C35H68NaO5 DAG(16:0/0:0/16:0) 1,3-PP
13 33.3 645.5449 645.5428 3.2 C39H74NaO5 DAG(18:0/18:1/0:0) 1,2-SO
14 34.2 645.5442 645.5428 2.1 C39H74NaO5 DAG(18:0/0:0/18:1) 1,3-SO
15 34.1 619.5285 619.5272 2.2 C37H72NaO5 DAG(18:0/16:0/0:0) 1,2-SP
16 35.3 619.5276 619.5272 0.6 C37H72NaO5 DAG(18:0/0:0/16:0) 1,3-SP
17 40.3 647.5579 647.5585 1 C39H76NaO5 DAG(18:0/18:0/0:0) 1,2-SS
18 41.4 647.5584 647.5585 0.2 C39H76NaO5 DAG(18:0/0:0/18:0) 1,3-SS
Figure 9.7 Tandem mass spectrum heptadecyl-palmitate, m/z 512.5, as representative of wax class.
HPLC-MS analysis of fraction 4 in negative ion mode displays the occurrence of free fatty
acids (Fig. 9.8) and other unknown compounds present with very low intensity. On the other
hand, positive ion mode shows the presence of other classes of compounds, so far unidentified,
192
Chapter 9
one of which is present in unfermented wet cocoa beans but subsequently absent in dried
fermented cocoa bean (Fig. 9.9).
Figure 9.8 HPLC-MS extract ion chromatogram in negative mode of free fatty identified in cocoa
lipid extract.
cocoa beans. Blue circle indicates other unusual TAGs, light brown circle indicates other unknown
lipid classes.
Sterol esters glucosides and phospholipids are the most representative classes in fraction 5
(Fig. 9.10). Several molecules could be identified based on tandem MS data. Figure 9.11
193
Chapter 9
illustrates the tandem mass spectrum in positive ion mode of sterol esters glucosides 16:0-
Glc-Sitosterol (m/z 832.6) ammonia adducts, 18:0-Glc-Stigmasterol (m/z 858.7) ammonia
adducts. As for sterol esters, the tandem MS of [M+NH4]+ precursor ions generated a MS2
sterol fragment ion at m/z 383.4 and m/z 397.4 for the 2 sterol esters glucosides respectively
(Fig. 9.11 (a) and (b)). The resulted neutral losses correspond to acyl group bonded to the
glucose and ammonium adduct.
Two major subclasses of phospholipids were identified: phosphocholines, and
phosphatidylethanolamines. Figure 9.12 (a) and (b) show the MS and MS2 spectra of 1-
palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine and 1,2-dioleoyl-sn-
phosphatidylethanolamine under both positive ionization modes. Major fragment ion for
phosphatidylcholine is derived from the loss of the polar head group choline [C5H13NPO4]+,
m/z 184, which is characteristic for this class of phospholipids. Whereas for
phosphatidylethanolamines, the MS2 base peak at m/z 603.5, corresponding to [OO]+ product
ion, indicated that the polar head group 141 Da was lost as neutral loss. According to these
fragmentation parameters, 17 phospholipids were identified. It worth mentioning that
phospholipids were present only in dried fermented beans.
cocoa beans. Red circle indicated sterol esters glucosides class and light blue circle indicates
phospholipid class in cocoa.
194
Chapter 9
Figure 9.11 Positive ion mode tandem mass spectrum of sterol esters glucosides (a) 16:0-Glc-
Sitosterol (m/z 832.6) ammonia adducts and (b) 18:0-Glc-Stigmasterol (m/z 858.7) ammonia adducts.
Figure 9.12 Positive ion mode tandem mass spectrum of phospholipids (a) 1-palmitoyl-2-oleoyl-sn-
glycero-3-phosphocholine, (b) 1,2-dioleoyl-sn-phosphatidylethanolamine, (c) and (d) their
corresponding chemical structures.
195
Chapter 9
9.4 Conclusions
In this chapter, structure elucidation of 62 minor components subdivided by lipid classes

was achieved. Thus, fractionation allowed separation of minor lipid components from the
major TAGs and therefore concentration enhancement, whereas HPLC-MS methods endorsed
detection and identification of sterol esters, diacylglycerols, waxes, free fatty acids, sterol
esters glucosides and phospholipids. Minor components have shown variation in occurrence in
unfermented wet cocoa beans if compared to dried fermented cocoa beans. However, relative
and absolute quantification of such differences is required and has to be considered as a focus
of future work.
References
Cert, A., Moreda, W., & Pérez-Camino, M. C. (2000). Chromatographic analysis of minor constituents in
vegetable oils. Journal of Chromatography A, 881(1–2), 131–148. https://doi.org/10.1016/S0021-
9673(00)00389-7
Kim, H. Y., & Salem, N. (1990). Separation of lipid classes by solid phase extraction. Journal of Lipid Research,
31(12), 2285–2289.
Lipp, M., Simoneau, C., Ulberth, F., Anklam, E., Crews, C., Brereton, P., Wiedmaier, C. (2001). Composition of
Genuine Cocoa Butter and Cocoa Butter Equivalents. Journal of Food Composition and Analysis, 14(4), 399–
196
Chapter 10
Chapter 10:
10 Comparative lipidomics of Cylindrotheca

closterium and Scenedesmus sp. reveals differences
in lipid production under nitrogen starvation
Song Wanga, Diana Sirbub, Laurenz Thomsena, Nikolai Kuhnertb, Matthias S. Ullrichb
1
Department of Physics and Earth Sciences, Jacobs University Bremen, Campus Ring 1, 28759 Bremen,
Germany
2
Department of Life Science and Chemistry, Jacobs University Bremen, Campus Ring 1, 28759 Bremen,
Germany
*Correspondence to:
Song Wang, [email protected]
Submitted to: Journal of Lipid Research.
Supplementary material is available on request. E-mail: [email protected],

[email protected]
197
Chapter 10
Abstract
Microalgae-derived biofuel is gaining increasing interests. However, factors, influencing

lipid production from oleaginous algae, still require investigation. In the present study, the lipid
production of two promising candidates—Scenedesmus sp. (Chlorophyceae) and
Cylindrotheca closterium (Bacillariophyceae) under nitrogen starvation, was investigated by a
method using liquid chromatography-electrospray ionization mass spectrometry techniques.
By combining the alterations of total lipids, growth kinetics, fatty acids composition, and
glycerolipids profiles, a comprehensive analysis of lipid synthesis in response to nitrogen stress
were carried out at a molecular level. The cell concentration of Cylindrotheca closterium
doubled within 24 h, whereas Scenedesmus sp. increased 10% at 24 h of nitrogen starvation.
Triacylglycerols (TAGs) in Scenedesmus sp. increased continuously during nitrogen starvation
and represented 11% of the dry biomass. In contrast, TAGs in C. closterium decreased
significantly within 12 h nitrogen starvation and started to increase after 12 h, which accounted
for 17% at 96 h. An increased digalactosyldiacylglycerols concentration in Scenedesmus sp.
showed a more important role in keeping the integrity of photosynthetic systems in green algae
than diatom. A rapid decrease of diacylglycerolstrimethylhomoserine concentration in
Scenedesmus sp. within 12 h suggested its role as the first response to nitrogen starvation. 60
and 72 TAG molecules were identified in Scenedesmus sp. and C. closterium respectively
providing the most abundant TAG profiles in green algae and diatom. The de novo synthesis
of the lipids in Scenedesmus sp. under nitrogen starvation showed a preference in a “eukaryotic
pathway”, while C. closterium relied on “prokaryotic pathway” to enhance TAG production.
Keywords: diatom; green algae; lipid; fatty acid; triacylglycerol; digalactosyldiacylglycerol.
198
Chapter 10
10.1 Introduction
Due to the dramatically expanding global population, energy consumption will increase by
56% between 2010 and 2040 leading to rapid exhaustion of the limited fossil fuel stock within
30 years (Abomohra et al., 2016). The overexploitation of fossil fuels caused severe
environmental concerns due to the emission of carbon dioxide, sulfur and nitrous oxides and
poly-aromatic hydrocarbons. Biofuel, derived from biomass, is suggested as a potentially
sustainable alternative to fossil fuels (Sharma and Singh, 2017). Microalgae represent a large
group of photosynthetically active microorganisms distributed in various aquatic
environments, such as freshwater, seawater, brackish water or even wastewater and readily
adaptive to a wide range of abiotic parameters, such as temperature, light intensity, pH, and
nutrients availability (Hu Q et al., 2008). Microalgae, are receiving increasing attention as a
promising feedstock for biofuels production, owing to the following advantages over the 1st
and 2nd generation of biomass sources: 1) the production of microalgae biomass requires
neither arable land nor freshwater (Michael Hannon et al., 2010); 2) microalgae, due to a higher
photosynthetic efficiency and CO2 concentration mechanism, yield higher biomass than
terrestrial plants (Sayre, 2010) (Spalding, 2008); 3) microalgae can accumulate a higher lipid
content under stress conditions (Gouveia and Oliveira, 2009); 4) microalgae, cultivated in
wastewater, can remove nutrients and harmful compounds to produce a clean effluent to offset
the cost of biofuel production (Randrianarison and Ashraf, 2017).
However, the high cost of algal biofuel production has, to date, industrialization inhibited.
The selection of algae species with an increased lipid production could cut down the cost and
make exploitation feasible (Chu, 2017). Stress strategies, such as modification of nutrient
supply and environmental parameters, offer promising approaches, which were applied to algae
cultivation to induce lipid synthesis and have been successfully demonstrated as a proof of
principle (Benavente-Valdes et al., 2016). Among all the stress approaches, nitrogen starvation
is most investigated, causing boosting of the lipid synthesis and offering insights into the lipid
biosynthesis pathway in algae. For instance, higher efficiency of lipid production in
Phaeodactylum tricornutum was obtained when nitrogen availability was limited in media
(Frada et al., 2013). Furthermore, the biomass of green algae N. oleoabundans, with higher
lipid content, was produced after the depletion of nitrogen, accompanied with the static cell
division and decrement of chlorophyll a (Li et al., 2008). It has been reported that the lipid
199
Chapter 10
content of Scenedesmus obliquus attained 22.4% on the dry biomass basis by 5-day nitrogen
starvation (Ho et al., 2012).
Nevertheless, the lipid metabolisms of algae from different classes in response to nitrogen
starvation have not yet been well understood. In most of the cases, researchers focused on total
lipids content and fatty acids profile, neglecting detailed analysis at a molecular level.
Comprehensive analysis of the changes of intact lipids, including storage lipids and membrane
lipids under nitrogen starvation has only been carried out in several model species, such as
Phaeodactylum tricornutum (Levitan et al., 2015), Nannochloropsis oceanica (Li et al., 2014),
and Chlamydomonas reinhardtii (Yang et al., 2015). It has been reported that under nitrogen
starvation nitrogen in half of the Phaeodactylum tricornutum proteins was scavenged to
strengthen the nitrogen incorporation enzymes as well as photosynthetic carbon was
reallocated into lipid biosynthesis pathway (Levitan et al., 2015). Moreover, varied fatty acid
chains in triacylglycerols (TAGs) and diacylglycerols-trimethyl-homoserine (DGTS) were
found in Chlamydomonas reinhardtii under nitrogen deprivation (Yang et al., 2015).
Combination of transcriptomics and lipidomics analysis revealed the suppression of the de
novo fatty acid biosynthesis pathway of Nannochloropsis oceanica, yet the pathway that
redirected the carbon flux from protein to lipid and carbohydrate was upregulated in a nitrogen-
free medium (Li et al., 2014). However, more studies on other oleaginous algae are necessary
to reveal the lipid metabolism in algae and guide a rational manipulation of lipid biosynthesis.
In the present study, the changes of lipids in a freshwater green algae Scenedesmus sp.
(Chlorophyceae) and a benthic diatom Cylindrotheca closterium (Bacillariophyceae) were
determined in response to the nitrogen starvation. The Scenedesmus sp. was isolated from a
municipal sewage water plant, which is highly tolerant to heavy metals, organic pollutants and
wide thermal variation. Cylindrotheca closterium in this experiment, featuring a high growth
rate and lipid content, was considered as a promising candidate for biofuel and high-value
pigment production (Wang et al., 2015). On another note, both species could biosynthesize a
wide arrange of poly-unsaturated fatty acids, essential to human health, such as α-Linolenic
acid by Scenedesmus obliquus (Makulla, 2000) and eicosapentaenoic acid by C. closterium
(Wang et al., 2015). The current experiment is designated to elucidate the variance of lipid
metabolisms, with special emphasis on triacylglycerol, between marine diatom and freshwater
green algae triggered by nitrogen stress at a molecular level.
200
Chapter 10
10.2.1 Algae and culture condition
The stock solutions of Scenedesmus sp. and C. closterium were maintained in freshwater with
Basal Bold Medium (BBM) and seawater enriched with f/2 medium nutrients (Guillard, 1975)
respectively at 12/12 light/darkness circle and 40 μmol/m2/s light intensity. Algae used in
nitrogen starvation experiment were harvested from a culture in a bag photobioreactor (PBR)
patented by Phytolutions GmbH at 18/6 light regime, 100 μmol/m2/s. C. closterium in the bag
PBR was cultured in f medium at 20±1 °C, whereas Scenedesmus sp. in BBM 25±2 °C. Algae,
reaching the late exponential phase, were harvested by centrifugation at 2000 g. C. closterium
was rinsed with seawater, while Scenedesmus sp. by ultra-pure water for three times to remove
nitrogen. The algae were re-suspended in nitrogen-free medium bottle PBRs at the same
environmental parameters to the bag PBRs. Algal biomass was harvested at time points of 0 h,
12 h, 24 h, 48 h and 96 h after inoculation with three biological replicates. Dry biomass was
obtained by lyophilization for 48 h before lipid extraction. The growth performance was
monitored by counting the cell concentration daily.
Ethanol gradient grade was purchased from Merck (Darmstadt, Germany); isopropanol
(Rotisolv® HPLC grade), acetonitrile (Rotisolv® HPLC ultra gradient grade), methanol
(Rotisolv® HPLC ultra gradient grade), chloroform (Rotisolv® HPLC grade) and Tetra-
dodecyl ammonium bromide were purchased from Carl Roth (Karlsruhe, Germany);
dichloromethane 99,8% stabilized with amylene for synthesis was purchased from Panreac
AppliChem (Darmstadt, Germany); ammonium formate LC-MS Ultra and formic acid (puriss.,
≥ 98% (T) for mass spectrometry), potassium hydroxide (KOH), sulfuric acid (H2SO4), heptane
(synthesis grade) and potassium chloride (KCl) were ordered from Carl Roth (Karlsruhe,
Germany); molecular sieves pellets (AW-300, 1.6 mm) and Spelco Supelco 37 Component
fatty acids methyl esterification (FAME) mix 1x1 mL were purchased from Sigma-Aldrich
Chemie (Steinheim, Germany).
201
Chapter 10
10.2.3 Total lipids extraction
Total lipids were extracted by a modified gravitational method. Approximately 100 mg of

every sample was extracted by 9 mL of dichloromethane/methanol at the ratio of 1:1 in a glass
vial. Extraction was assisted by a magnetic stirrer for 1h at room temperature. The mixture was
centrifuged at 3000 g for 10 min. The extraction procedure was repeated once. The supernatant
was combined in fresh vials. 4.5 mL KCl solution (0.88%) was added to the combined
supernatant. An organic and an aqueous phase were separated by centrifugation at 3000 g for
5 min. The separation procedure was repeated once more with KCl solution. 3 mL organic
phase was transferred to a pre-weighed test vial and evaporated by N2. Total lipids content was
calculated by subtracting the mass of the empty vials from that of vials with lipids. Chloroform
was added to dissolve the lipids at a concentration of 5 mg/mL, which were stored at -20 °C
for further analysis.
10.2.4 FAME preparation and GC analysis
500 µL of the lipid extract, dissolved in chloroform, was transferred in a glass vial for Fatty
acid methyl ester (FAME) synthesis. 700 µL 10 N KOH and 5.3 mL methanol were added to
the extract, stirred continuously at 50 °C for one hour. Subsequently, 580 µL 24 N H2SO4 was
added when samples were cooled down at room temperature. The samples were stirred
continuously at 50 °C for one hour. FAME was isolated with 2 mL heptane by vortex for 5
minutes and centrifugation at 3000 g for 3 minutes. Molecular sieves were added to remove
the aqueous phase. FAME was subject to GC-FID for analysis. A capillary column VF-5ms,
30 m x 0.25 mm with an injector temperature of 265 °C and detector temperature of 300 °C
was integrated to the GC. The GC used helium as a carrier gas at a flow rate of 40 mL/min.
The split ratio was 1:50 with an injection volume of 1.0 μL. GC program started at 60 °C for 1
minute, followed by a second step of 17 °C /min temperature rate until 160 °C, a third step of
7 °C/min rate until 240 °C and a temperature rate of 5 °C/min until 300 °C with a hold time of
10 min. Fatty acids were assigned by Supelco 37 compounds fatty acid methyl ester (FAME)
MIX standard and quantified by the calibration curves of the standard.
202
Chapter 10
10.2.5 High-performance liquid chromatography
Lipids were fractionated by high-performance liquid chromatography (Agilent 1100 series,

Waldbronn, Germany) with our previously established method (Sirbu et al., 2018) to which
slight modifications on elution gradient were applied. One column Pursuit XRs C18 (250 mm
× 3 mm i.d., 5 μm particles) was integrated to the HPLC. The temperature of the column oven
was set to 35 °C. Each extract was diluted at the ratio of 1:22 in chloroform/ethanol (1:1) with
3 µL of sample were injected. Solvent A consisted of acetonitrile with 0.01% formic acid and
B of ethanol with 10 mmol/L ammonium formate and 0.01% formic acid. The mobile phase
was pumped through the column at a flow rate of 0.6 mL/min. The gradient elution program
started with A/B (90/10%) increased linearly to A/B (65/35%) in 4 min; followed by a second
linear increasing to solvent A/B (50/50%) in 16 min; a third linear increase to solvent A/B
(37/63%) in 2 min; a forth linear increase to solvent A/B (20/80%) in 23 min; a fifth step, linear
increase to solvent B (100%) in 2 min; ending with isocratic elution at solvent B (100) for 10
min. The column was equilibrated at A/B (90/10%) solvent for 10 min before reuse. Tetra-
dodecyl ammonium bromide at the concentration of 0.0002 mg/mL, mass-to-charge (m/z)
690.7850, and retention time (Rt) 4.5 min was used as internal standard.
10.2.6 High-resolution mass spectrometry
High-resolution masses were acquired by a time of flight MicrOTOF Focus mass

spectrometer (Bruker Daltonics, Bremen, Germany) fitted with an ESI source as the detector
with the following parameter settings: capillary voltage of 4.5 kV; nebulizing gas pressure of
2 Ba drying gas flow rate of 10 L/min; drying gas temperature of 220 °C. ESI mass spectra
were measured in the range of m/z 200–1200 in the positive ion mode. Internal calibration was
achieved with 10 mL of 0.1 M sodium formate solution injected through a six-port valve prior
to each chromatographic run. Calibration was carried out by the enhanced quadratic mode.
10.2.7 Tandem mass spectrometry
LC-tandem MS was carried out by using an Ion-Trap detector in positive ion mode equipped
with an ESI source (Bruker Daltonics, Bremen, Germany). The full scan mass spectra were
recorded in the range m/z 200-1200 operating in positive ion mode. Capillary temperature was
203
Chapter 10
set to 350 °C, drying gas flow rate 10 L/min and nebulizer pressure 10 psi. Tandem mass spectra
were acquired in Auto MSn (smart fragmentation) using a ramping of the collision energy.
The individual lipid lists of the high-resolution masses were combined for peak matching
and retention time (Rt) alignment using a m/z tolerance of Δm/z = 0.0001 and a Rt tolerance of
0.5 min for elution time <20 min and 1.0 min for an elution time ≥20 min. Extensive manual
curation resulted in a set of 102 peaks for C. closterium and 93 peaks for Scenedesmus sp..
Lipid compounds could be uniquely assigned with m/z and Rt tolerances of Δm/z = 0.01 and
ΔRt = 0.8 min. For further analyses, the peak areas were normalized to the peak area of the
internal standard (Tetra-dodecyl ammonium bromide). Heat maps, as graphical representations
for visualizing attribute values by class in a two-way matrix, as well as principal component
analyses (PCA) were performed by Orange software. Microsoft Office Excel 2016 was used
to process data and perform ANOVA single-factor analysis.
10.3.1 Total lipid content and growth kinetics under nitrogen starvation
Our previous study indicated that Scenedesmus sp. could grow fast in wastewater, recover
the nutrients producing clean effluent and display excellent thermo-tolerant property.
Cylindrotheca closterium features for high specific growth rate and rapid cell sedimentation.
Consequently, Scenedesmus sp. and C closterium were chosen in this experiment for lipids
production owing to the favorable characteristics in cultivation or harvesting. Algae,
accumulating biomass in bag PBRs, were transferred to bottles PBRs to induce lipid synthesis
at the same abiotic parameters. Total lipids were determined gravimetrically following
dichloromethane extraction. As indicated in Fig. 10.1 (a), a lipid content of 13% in dry biomass
was synthesized in Scenedesmus sp. at 0 h. During nitrogen starvation, the total lipids content
of Scenedesmus sp. increased gradually and culminated at 96 h, while no prominent changes
were observed between 0 h and 48 h. C. closterium accumulated 20% lipid before the onset of
nitrogen starvation, consistent with our previous results and indicating a stable lipid content
204
Chapter 10
(Wang et al., 2015). Surprisingly, the lipid content of C. closterium decreased significantly
within 12 h of nitrogen starvation and maintained at this level till 48 h. C. closterium showed
a remarkable increment of lipids between 48 and 96 h, but no significant difference was found
between the total lipids amount at 0 h as a negative control and 96 h of starvation. The growth
kinetics in Fig. 10.1 (b) suggested the cell concentration of Scenedesmus sp. is 5-fold higher
than C. closterium at the late exponential phase. By 24 h of nitrogen stress, the cell
concentration of Scenedesmus sp. increased by approximately 12%, whereas C. closterium
increased by more than 64% at 12 h, nearly doubled the concentration at 24 h and stopped
growing at 48 h. Both Scenedesmus sp. and C. closterium continued growing in nitrogen-free
media. However, the growth rate varied remarkably.
Figure 10.1 (a) Total lipids and (b) growth kinetics of Scenedesmus sp. and C. closterium under
nitrogen starvation. Results were shown as meansSD, n=3.
Despite the absence of nitrogen, both species continued proliferating, especially C.

closterium doubling the cell concentration at 24 h. This might be explained by assuming that
algae in nutrients-sufficient environment tend to assimilate more nutrients, including nitrogen,
to form an internal reservoir in vacuole which could support the growth in a nitrogen-scarce or
even free environment (Lavín and Lourenço2, 2005). The existence of internal pool was
confirmed in Chlorella sp. and Nannochloropsis oculata (Paes et al., 2016). It has been
previously reported Scenedesmus obliquus could accumulate as much as 20% lipid of dry
biomass after 5-day nitrogen starvation compared to 11% in a negative control, accompanied
by an increase of carbohydrate and decrease of protein content (Ho et al., 2012). This literature
prediction is in agreement with the lipid content of the Scenedesmus sp. after nitrogen
starvation in this study. This could be interpreted by the fact that nitrogen was intimately
205
Chapter 10
connected to the carbon metabolism. The reduced availability of nitrogen led to the re-
allocation of photosynthetic carbon flux and proteins scavenged to increase the lipid synthesis
(Alipanah et al., 2015). In response to nitrogen-starvation, Phaeodactylum tricornutum, as the
model diatom, produced significantly more lipid after 48 h cultivation in nitrogen-free medium,
but the cell growth was suppressed if compared to control after 48 h (Alipanah et al., 2015),
which contrasted with the lipid metabolism displayed by C. closterium in the present study.
This distinct lipid metabolism could possibly be explained by the rapid cellular reproduction
within 12 h, which required a significant amount of energy. Therefore, lipids, compounds
storage of energy, were consumed to provide energy due to the deficiency of photosynthetic
ATP production.
10.3.2 Fatty acids profile under nitrogen starvation.
FAME analysis of lipid extraction was carried out and identified compounds were shown in
Fig. 10.2 (a-d). It revealed that C16:3, C16:0 and C18:1 are the major fatty acids species in
Scenedesmus sp. with C18:1 being the most abundant one and long-chain fatty acids (>20 C)
in a minor amount at 0 h before nitrogen starvation (Fig. 10.2 (a)). Saturated fatty acids, such
as C14:0, C16:0 and C18:0, increased significantly during nitrogen starvation with C16:0 and
C18:0 increased by around 4-fold and 10-fold, respectively at 96 h, if compared to 0 h.
Monounsaturated fatty acids C20:1 showed a 3-fold increase at 24 h, while C18:1 only
displayed noticeable increase at 96 h. Polyunsaturated fatty acids, such as C16:3 and C18:3,
and odd number fatty acids C15:0 and C17:0 decreased significantly after 12 h starvation with
the exception of C18:2. C16:3 and C18:3 decreased to 67% and 37%, respectively at 96 h as
opposed to 0 h. C18:2 increased immediately after the onset of nitrogen starvation. After 96 h
nitrogen starvation, C16:0 and C18:1 accounted for 70% of the total fatty acids.
Five most abundant fatty acids (C14:0, C16:1, C16:0, C20:4 and C20:5) were synthesized
by C. closterium with palmitoleic (C16:1) and palmitic acids (C16:0) as the most dominant
species at 0 h (Fig. 10.2 (c)). Under nitrogen starvation, a 26% increase of saturated fatty acid
C14:0 occurred at 96 h compared to 0 h, while C16:0 underwent a decrease within first 12 h
and an increase after 12 h, following the tendency of total lipids. Monounsaturated fatty acids
C15:1 and C16:1 increased in connection during the whole time-course, yet remarkable
increments were only observed at 96 h (p<0.05). Polyunsaturated fatty acids (PUFA) and long-
chain PUFA maintained a constant level till 48 h and decreased significantly at 96 h. C20:4 and
206
Chapter 10
C20:5 reduced by 14% and 19%, respectively at 96 h compared to the negative control at 0 h.
The odd number saturated fatty acids (C15:0 and C17:0), rarely reported previously, in C.
closterium followed a decreasing trend. At 96 h, the two dominant fatty acids—C16:1 and
C16:0 represented 65% of the total fatty acids.
Figure 10.2 Fatty acids composition of Scenedesmus sp. and C. closterium under nitrogen
starvation. (a) Major fatty acids in Scenedesmus sp.; (b) Minor fatty acids in Scenedesmus sp.; (c)
Major fatty acids in C. closterium; (d) Minor fatty acids in C. closterium. Results were shown as
meansSD, n=3. Compounds could be assigned by standards were named by the retention time.
Although Scenedesmus sp. and C. closterium were characterized by different fatty acids
composition, the percentage of polyunsaturated fatty acids decreased in both species at 96 h of
nitrogen starvation, but the ratio of saturated and monounsaturated fatty acids increased in
response to a nitrogen-free environment. Similar alterations of fatty acids happened in
207
Chapter 10
Chlorella zofingiensis under nitrogen starvation with an increased saturated and

monounsaturated but a diminishing polyunsaturated fatty acid ratio (Zhu et al., 2015).
Likewise, in another previous study, the synthesis of saturated fatty acid of P. tricornutum was
elevated in a nitrogen-depleted medium at 96 h compared to nitrogen-replete medium
(Yodsuwan et al., 2017). The decrease of polyunsaturated fatty could be considered as the
consequence of the inhibition of the photosynthesis and degradation of the photosynthetic
systems where PUFA are usually present. Since saturated/monounsaturated acyl chains are
beneficial for biodiesel production, both species featuring high saturated/monounsaturated
fatty acids contents, induced by nitrogen starvation, are suitable for biodiesel production (Islam
et al., 2013).
10.3.3 Relative quantitative analysis of glycerolipids and pigments
Identification of lipid compounds was achieved by a high-resolution MicrOTOF mass

spectrometry (MS) operating in positive mode. The study of the compounds fragmentation was
conducted in an Ion-Trap MS. Moreover, the tandem MS data were compared with
fragmentation pathways in literature and the database LIPID MAPS
(http://www.lipidmaps.org). For high-resolution MS, calculation of molecular formulae with
mass error < 4 ppm was accepted. Tandem MS data were interpreted as described by (Sirbu et
al., 2018) and (Su et al., 2013). Relative quantification of all the identified compounds was
performed in positive ion mode with respect to an internal standard (Tetra-dodecyl ammonium
bromide). Spectra describing the fragmentation pattern of one example from each lipid class
and tables (Appendix Table G.1 and G.2) with identified compounds of the individual species
are given in Appendix. Additionally, representative chromatograms and their two-dimensional
maps at 0 and 96 h for each species are in the Appendix (Fig. G.1).
Thanks to the development of lipidomics based on mass spectrometry, much more
knowledge could be obtained on different groups of lipids in algae. 99 molecules, listed in
Appendix Table G.1, were successfully annotated in Scenedesmus sp. total lipids and assigned
to 9 classes: triacylglycerol (TAG) 3, TAG >3 double-bond, monogalactosyldiacylglycerol
(MGDG), digalactosyldiacylglycerol (DGDG), phosphatidylglycerol (PG), diacylglyceryl-N,
N, N- trimethylhomoserine (DGTS), lyso-DGTS and phosphatidylethanolamine (PE) in Fig.
10.3 (a). Before the start of nitrogen starvation, TAG (3 and >3) and plastidial lipids (MGDG,
DGDG and PG) accounted for 1.8% and 5.1% of the dry biomass, respectively. After 96 h of
208
Chapter 10
nitrogen starvation, the overall amount of TAG increased more than 6-fold more in
Scenedesmus sp. with TAG 3 and >3 double-bonds increased 4-fold and 28-fold, respectively
with respect to the counterparts at 0 h. Total TAGs accounted for 68% and 11% of the total
lipids and biomass respectively at 96 h in Scenedesmus sp.. Pigments including chlorophyll a/b
and pheophytin decreased significantly to one quarter at 96 h in response to nitrogen starvation
if compared to 0 h, concomitant with the decrements of MGDG and PG. MGDG and PG
decreased to one third and one fifth respectively at 96 h as a comparison to 0 h. Another
essential plastidial lipid DGDG reached the maximum at 48 h and dropped at 96 h. The
concentration of DGDG at 48 and 96 h was 4 and 3 times more than the negative control at 0
h, respectively. The plastidial membrane lipids (MGDG, DGDG and PG) comprised 39% of
the total lipids at 0 h, as judged by the peak area, which dropped to 15% at 96 h. DGTS, lyso-
DGTS and PE, as N-containing lipids, are crucial extra-plastidial membrane lipids. A
significant decrease of DGTS and PE occurred at 12 and 24 h, respectively, while lyso-DGTS
first increased at 24 h and subsequently reduced till 48 h.
92 lipid molecules, listed in Appendix Table G.2, were identified in C. closterium allocated
into 6 groups: TAG 3 and >3 double-bond, MGDG, DGDG, sulfoquinovosyl diacylglycerol
(SQDG) and pigments (Fig. 10.3 (b)). The TAG (3 and >3) and plastidial lipids (MGDG,
DGDG and SQDG) represented 14% and 2% of the dry biomass respectively in C. closterium
before subject to nitrogen-free media. Under nitrogen starvation, pigments decreased to one
third at 96 h if compared to 0 h. The abundance of both TAGs classes decreased dramatically
at 12 h and gradually increased after 12 h, with a significant enhancement observed at 96 h.
Total TAG content increased by 20% at 96 h, with TAGs ( 3 double-bond) and TAGs (> 3
double-bond) increasing by 10% and 49% in comparison to 0 h. The TAG represented up to
90% and 17% of the total lipids and dry biomass at 96 h. Among the three plastidial membrane
lipid classes (MGDG, DGDG and SQDG), the dominant class was MGDG decreased to one
fourth at 96 h as opposed to 0 h. Concomitantly, DGDG dropped to half of the amount, while
SQDG rise at 24 h and reduced at the 96 h than 0 h. Plastidial membrane lipids accounted for
10% and 4% of the total lipids at 0 and 96 h, respectively.
209
Chapter 10
Figure 10.3 Changes of glycerolipids and pigments in Scenedesmus sp. and C. closterium under
nitrogen starvation: (a) and (b) Glycerolipids and pigments in Scenedesmus sp.; (c) and (d)
Glycerolipids and pigments in C. closterium. Results were shown as meansSD, n=3.
The results demonstrated that photosynthesis of Scenedesmus sp. and C. closterium was
suppressed due to the reduction of photosynthetic pigments and decrease of plastidial
membrane lipids in nitrogen-free media. The decrease of pigments was due to the degradation
of the essential enzymes and structural protein in the photosynthetic pathway and meanwhile,
the N in pigments, as N-rich compounds, were presumably redirected to protein synthesis under
nitrogen starvation condition (Levitan et al., 2015).
In comparison to Scenedesmus sp., C. closterium adopted a different lipid adaptation
triggered by nitrogen starvation. C. closterium accumulated a higher TAG content at 0 h and
as the proceeding of nitrogen starvation, TAG decreased to provide energy for the rapid cell
division. On the contrary, Scenedesmus sp. accumulated very low amount TAGs at 0 h and
210
Chapter 10
increased gradually. TAG only accounted for 1.8% of the dry biomass of Scenedesmus sp., in
a nutrients-replete environment, which indicated most of the photosynthetically fixed carbon
was partitioned into protein and carbohydrate to enhance biomass. Neither C. closterium nor
Scenedesmus sp. showed a slowdown in TAG accumulation at 96 h. Thus, further TAG increase
could be expected by extending the stress process. In the previous study, Coccomyxa
subellipsoidea could accumulate 13% TAG in dry biomass after 10-day nitrogen starvation
(Allen et al., 2015). Scenedesmus sp. in the present study exhibited similar TAG content, which
C. closterium displayed higher TAG content, and both species were characterized by
favourable fatty acids composition (high saturated/monounsaturated fatty acids ratio), whereby
this enabled them to be potential candidates suitable for biofuel production. Moreover,
Scenedesmus sp. was more abundant in plastidial membrane lipids than C. closterium. As most
of MGDG, DGDG and SQDG are mainly present in thylakoids membrane and have a
substantial impact on the photosynthesis and growth, photosynthetic apparatus in Scenedesmus
sp.is more developed in C. closterium (Peter Dormann et al., 1995). DGDG in Scenedesmus
sp. increased within 48 h but decreased during the whole process in C. closterium. The ratio of
MGDG/DGDG in both species decreased significantly. Decreased MGDG, along with
increased DGDG, was reported previously in Arabidopsis (Gaude N et al., 2007) and P.
tricornutum (Abida et al., 2015), yet in model algae, Chlamydomonas reinhardtii both MDGD
and DGDG decreased under nitrogen starvation (Siaut et al., 2011). This could be explained
DGDG, a bilayer-lipid, showed superiority over MGDG, a non-bilayer-lipid, in maintaining
the integrity of thylakoids under nitrogen stress (Webb and Green, 1991). However, as the
proceeding of nitrogen starvation, the more photosynthetic machinery was degraded, whereby
the DGDG started to decrease at 48 h of nitrogen starvation. The explanation was supported
by the declining MGDG/DGDG ratio. Consequently, we can hypothesize that the structure of
the photosynthetic system in C. closterium is more subject to nitrogen stress than Scenedesmus
sp., due to the immediate reduction of DGDG.
10.3.4 Lipid and pigment molecules alteration
Two principal components analysis of LC-MS data of lipid extract taken at different time
points were performed for each of two species based on the lipids and pigments identified (Fig.
10.4). PCA on Scenedesmus sp. showed clear separation of samples taken at each time series.
The first principal component in the score plot is strongly correlated with the amount of TAG.
211
Chapter 10
Significant changes occurred after 12 h of nitrogen starvation in Scenedesmus sp.. In contrast,

C. closterium did not show a net separation between all the time series, but distinguishable
changes of samples at 0 h and 96 h in comparison to the other 3 could be observed. Moreover,
the grouping shown on the plots reflected the reproducibility of data and how close are the
samples to each other in term of lipids profile.
Figure 10.4 Principal component analysis of Scenedesmus sp. and C. closterium under nitrogen
starvation: (a) Samples from Scenedesmus sp.; (b) Samples from C. closterium. Three biological
replicates were performed at each time point.
The heat maps illustrated the alteration of the lipid and pigment molecules in Scenedesmus
sp. (Fig. 10.5 (a)) and C. closterium (Fig. 10.5 (b)) in the time-course of nitrogen starvation.
All the molecules are listed in Table G.1 for Scenedesmus sp. and G.2 for C. closterium in the
Appendix. 60 and 72 TAG molecules were identified in Scenedesmus sp. and C. closterium
respectively which give the most abundant TAG profiles in both green algae and diatom,
although not comprehensive. Generally, TAGs with more than 3 double bonds were absent at
0 h and only present at 12 h of nitrogen starvation. Among the 60 TAGs in Scenedesmus sp.,
TAG 50:1(16:0/16:0/18:1), 52:1 (18:0/16:0/18:1), 50:2 (14:0/18:1/18:1), 52:2
(16:0/18:1/18:1), 50:3 (16:0/16:0/18:3), 52:3(16:0/18:1/18:2), 54:3 (18:0/18:0/18:3), 52:4
(18:1/16:0/18:3) and 52:4 (16:0/18:2/18:2) are the dominant TAG species at 96 h. The
abundance of the major TAGs (3 double bonds), increased by 8-20 folds at 96 h if compared
to the counterparts at 0 h, except 52:1 (18:0/16:0/18:1) and 50:3 (16:0/16:0/18:3). 52:1
(18:0/16:0/18:1) increased 3-fold at 96 h compared to the negative control, while no 50:3
212
Chapter 10
(16:0/16:0/18:3) was synthesized at 0 h. 52:4 (18:1/16:0/18:3) and 52:4 (16:0/18:2/18:2),

containing >3 double-bond by approximately 60-fold at 96 h, if compared to TAGs at 0 h. With
respect to plastidial membrane lipids, a more complex pattern was adopted. MGDG 34:7
(16:4/18:3), as the most abundant MGDG, decreased significantly at 12 h under nitrogen
starvation. On the contrary, the other three MGDGs 34:6 (16:3/18:3), MGDG 34:6 (16:4/18:2)
and MGDG 34:5 (16:3/18:2) started to decrease at 48 h. All DGDGs displayed the same
tendency as the total DGDG amount culminating at 48 h and decreasing at 96 h. Of special
note is that the dominant DGDG 34:3 (16:0/18:3) at 96 h increased during the whole process.
All PG molecules underwent noteworthy reduction at 24 h. DGTS and PE are the major
constituents of extra-plastidial membrane. PE 36:6 (C18:3/18:3), as the major PE at 0 h,
decreased significantly at 12 h, while the other PEs started to decrease at 24 h. In contrast, most
DGTS lipids with PUFA chains decremented dramatically at 12 h with an exception that no
remarkable difference was found in the dominant DGTS 34:3 (16:0/18:3). The primary lyso-
DGTS (18:3) fluctuated in a certain range during the whole process. DGTSs as a kind of N-
containing lipids tended to be the first cellular reaction coping with nitrogen starvation, which
is also confirmed in Nannochloropsis (Jing Li). DGTS 34:3 (16:0/18:3) and lyso-DGTS (18:3)
might serve as the house-keeping lipid to maintain the whole integrity of cell membrane. Figure
10.6 illustrates the chemical structure of the most abundant lipid compounds in Scenedesmus
sp.
72 TAG molecules have been assigned at 96 h based on high-resolution and tandem MS
data. 8 TAG molecules 46:1 (14:0/16:1/16:0), 48:1 (16:0/16:1/16:0&16:0/18:1/14:0), 52:1
(16:0/18:1/18:0&14:0/18:1/20:0), 48:2 (16:0/16:1/16:1&14:0/18:1/16:1), 50:2
(16:0/18:1/16:1&14:0/18:1/18:1), 52:5 (16:0/16:0/20:5), 52:5 (16:0/16:1/20:4) and 52:6
(16:0/16:1/20:5) were found to be predominant species in C. closterium. All the major TAGs
started to decrease in concentration at 12 h, while significant increase of each major TAG only
took place at 96 h, if compared to the 0 h counterparts. 46:1 (14:0/16:1/16:0), 48:2
(16:0/16:1/16:1&14:0/18:1/16:1), 52:5 (16:0/16:0/20:5) and 52:6 (16:0/16:1/20:5) increased
40-60% at 96 h. MGDG species followed the same trend of the major MGDG 36:8 (16:3/20:5),
which decreased significantly at 12 h. The two primary DGDG 36:7 (C16:2/C20:5) and 36:6
(C16:1/C20:5) stayed unchanged until 12 h and decreased after 12 h. The SQDG 32:1
(16:0/16:1) increased within 24 h before it decreased at 48 h. Figure 10.7 illustrates the
chemical structure of the most abundant lipid compounds in C. closterium.
213
Chapter 10
Figure 10.5 Heat map of glycerolipid and pigment molecules in Scenedesmus sp. and C.
closterium under nitrogen starvation: (a) glycerolipid and pigment molecules from C. closterium; (b)
glycerolipid and pigment molecules from Scenedesmus sp. All the identified molecules were primarily
clustered in 4 groups: molecules in group a' and c' displayed a decreasing trend during the time
course; molecules in group b' decreased at the beginning, then increased till 96 h; group d' exhibited
an increasing trend. compounds were tentatively identified due to the insufficient fragment information.
214
Chapter 10
Figure 10.6 Chemical structures of the most abundant lipid compounds in Scenedesmus sp.
215
Chapter 10
Figure 10.7 Chemical structures of the most abundant lipid compounds in C. closterium.
216
Chapter 10
As shown, the PUFA chains were mainly present in membrane lipids at 0 h in both species,
while TAG with PUFA acyl chain increased significantly at 96 h. PUFA concentration was
observed to decrease in FAME analysis, hence we suggested that PUFA acyl chains are
recycled by TAGs in both algae, which was witnessed in a previous study (Goncalves et al.,
2016). As the precursor of TAG de novo synthesis, DAG was synthesized by the acylation of
the sn-2 position of lysophosphatidic acid catalyzed by lysophosphatidic acid acyltransferases
(LPAAT) (Nobusawa et al., 2017). LPAAT localized in different organelles showed varied
preference. LPAAT in chloroplast showed preference to C16 moiety substrate – the
“prokaryotic” pathway, while LPAAT in endoplasmic reticulum preferred C18 – the
“eukaryotic” pathway (Allen et al., 2014). As shown in data of Table 10.1, although TAGs were
synthesized in both pathways, the ratio of lipids from the two pathways differed. Before the
start of nitrogen starvation, more TAGs were synthesized in “eukaryotic” pathway than
“prokaryotic” pathway in Scenedesmus sp., but the ratio of TAG with C16/18 in sn-2 position
decreased significantly at 24 h in response to nitrogen starvation. In contrast, most of the TAGs
were synthesized by “prokaryotic” pathway at 0 h in C. closterium. Moreover, the increased
TAGs induced by nitrogen starvation were mainly contributed by “prokaryotic” pathway.
Compared to the rapid increase of TAGs in Scenedesmus sp. significant increase only happened
at 96 h in C. closterium revealing Scenedesmus sp. can switch to lipids production faster.
DGDG molecules increased within 48 h in Scenedesmus sp., while all DGDG molecules
decreased significantly at 24 h in C. closterium. Interestingly, the amount of DGDG 34:3
(16:0/18:3) was elevated at 96 h in Scenedesmus sp. indicating its essential function in
photosystems of green algae (Sozer et al., 2011).
Table 10.1 The content of TAGs (% dry biomass) with C16 or C18 FA in the sn-2 position and ratio of
TAG with C16 and C18 in the sn-2 position.
Time (h) Scenedesmus sp. C. closterium
% C16 % C18 Ratio of C16/C18 % C16 % C18 Ratio of C16/C18
sn-2 sn-2 sn-2 sn-2 sn-2 sn-2
0 0.750.06 0.990.15 0.760.1 11.890.1 1.750.06 6.790.16
12 1.080.16 1.460.15 0.740.04 8.660.63 1.280.08 6.770.21
24 1.190.06 1.880.17 0.640.04 9.250.20 1.420.04 6.530.21
48 2.000.11 4.110.18 0.490.01 10.220.10 1.010.05 10.090.57
96 3.490.17 7.530.23 0.460.01 14.841.36 1.410.04 10.531.17
If two or three molecules shared one mass/charge ratio, they were represented by the dominant one
based on the fragment information in the calculation. Data were shown as meanSD, n=3.
217
Chapter 10
DGTS molecules in Scenedesmus sp. dropped immediately after nitrogen starvation indicating
the sensitivity of DGTS to nitrogen stress.
From the results, we can draw a conclusion that Scenedesmus sp. and C. closterium adopted
a different strategy to cope with nitrogen stress, specifically the TAG accumulation mechanism.
The TAG content of C. closterium decreased at 12 h which might be induced by the rapid cell
reproduction. This unique lipid metabolism that we discovered in C. closterium, is presumably
due to its benthic characteristics. Diatoms are more competitive in absorbing nutrients than
other phytoplankton. Large diatoms, in particular, tend to have a relatively larger vacuole than
small diatom where nitrogen could be stored (Litchman, 2007) (Sicko-Goad et al., 1984)
(Raven, 1987). The growth of Scenedesmus sp. in this study slowed down, but C. closterium
duplicated rapidly in response to nitrogen starvation. More small cells of C. closterium were
present at 96 h compared to algae as seen at 0 h from the observation under a microscope. C.
closterium is a kind of large diatom, 30-40 um in length. It could be beneficial for ciliates
defense, but at the same time, they will have a lower area/volume, compared to small cells,
which will not be harmful in a nutrient-rich environment. Nevertheless, in nutrients limited
environment, smaller cells are more favourable for nutrients uptake (Raven, 1998) (Banse,
1976). Therefore, C. closterium improved the capability of nitrogen assimilation by
reproducing smaller cells. Likewise, algae under nitrogen stress tended to strengthen
assimilation apparatus were noted down in several studies, which supported the hypothesis in
principle (Levitan et al., 2015).
10.4 Conclusions
Freshwater green algae and marine benthic diatom exhibited different lipid metabolism in
response to nitrogen starvation. A detailed comparative lipidomic analysis by LC-MS of both
species, yielding information at the molecular level, was carried out for the first time. In
contrast to a constant increase lipid content in Scenedesmus sp., the lipids decreased to support
a rapid duplication of C. closterium within 12 h but increased at 96 h. Nitrogen starvation
induced an increased level of saturated/monounsaturated fatty acids but decreased the
production of polyunsaturated fatty acids, which is beneficial for biofuel production. By
providing the most abundant TAG profiles in green algae and diatom, more insights have been
brought into the different metabolisms of lipid production. TAG increased by 6-fold and 20%
218
Chapter 10
in Scenedesmus sp. and C. closterium at 96 h if compared to samples before nitrogen starvation,

respectively. In comparison to the concomitant decrease of MGDG and DGDG in and C.
closterium, DGDG in Scenedesmus sp. increased at 48 h indicating an essential role in
maintaining the integrity of photosystems in green algae. C. closterium grew rapid at 12,
resulting in a significant TAG decrease to provide energy for reproduction. The TAG of C.
closterium attaining 17% of the dry biomass. The TAG synthesized by “eukaryotic pathway”
in Scenedesmus sp. were strengthened, in contrast, increased TAG was synthesized by
“prokaryotic,” indicating the discrepancy of de novo TAG biosynthesis pathway.
References
Abida, H., Dolch, L.J., Mei, C., Villanova, V., Conte, M., Block, M.A., Finazzi, G., Bastien, O., Tirichine, L.,
Bowler, C., Rebeille, F., Petroutsos, D., Jouhet, J., Marechal, E., 2015. Membrane glycerolipid remodeling
triggered by nitrogen and phosphorus starvation in Phaeodactylum tricornutum. Plant Physiol. 167(1), 118-136.
Abomohra, A.E.-F., Jin, W., Tu, R., Han, S.-F., Eid, M., Eladel, H., 2016. Microalgal biomass production as a
sustainable feedstock for biodiesel: Current status and perspectives. Renewable and Sustainable Energy Reviews.
64, 596-606.
Alipanah, L., Rohloff, J., Winge, P., Bones, A.M., Brembu, T., 2015. Whole-cell response to nitrogen deprivation
in the diatom Phaeodactylum tricornutum. J Exp Bot. 66(20), 6281-6296.
Allen, J.W., DiRusso, C.C., Black, P.N., 2015. Triacylglycerol synthesis during nitrogen stress involves the
prokaryotic lipid synthesis pathway and acyl chain remodeling in the microalgae Coccomyxa subellipsoidea.
Algal Research. 10, 110-120.
Allen, J.W., DiRusso, C.C., Black, P.N., 2014. Triglyceride quantification by catalytic saturation and LC–MS/MS
reveals an evolutionary divergence in regioisometry among green microalgae. Algal Research. 5, 23-31.
Banse, K., 1976. Rates of growth, respiration and photosynthesis of unicellular algae as related to cell size - a
review. J. Plycol. . 12, 135-140.
Benavente-Valdes, J.R., Aguilar, C., Contreras-Esquivel, J.C., Mendez-Zavala, A., Montanez, J., 2016. Strategies
to enhance the production of photosynthetic pigments and lipids in chlorophycae species. Biotechnol Rep (Amst).
10, 117-125.
Chu, W.-L., 2017. Strategies to enhance production of microalgal biomass and lipids for biofuel feedstock.
European Journal of Phycology. 52(4), 419-437.
219
Chapter 10
Frada, M.J., Burrows, E.H., Wyman, K.D., Falkowski, P.G., 2013. Quantum requirements for growth and fatty
acid biosynthesis in the marine diatom Phaeodactylum tricornutum (Bacillariophyceae) in nitrogen replete and
limited conditions. Journal of Phycology. 49(2), 381-388.
Gaude N, Bréhélin C, Tischendorf G, Kessler F, P., D., 2007. Nitrogen deficiency in Arabidopsis affects
galactolipidcomposition and gene expression and results inaccumulation of fatty acid phytyl esters. Plant J. 49(4),
729-739.
Goncalves, E.C., Wilkie, A.C., Kirst, M., Rathinasabapathi, B., 2016. Metabolic regulation of triacylglycerol
accumulation in the green algae: identification of potential targets for engineering to improve oil yield. Plant
Biotechnol J. 14(8), 1649-1660.
Gouveia, L., Oliveira, A.C., 2009. Microalgae as a raw material for biofuels production. J Ind Microbiol
Biotechnol. 36(2), 269-274.
Guillard, R.R.L., 1975. Culture of phytoplankton for feeding marine invertebrates. . In: Smith, W.L. and Chanley,
M.H., Eds., Culture of Marine Invertebrates Animals, Plenum Publishing, New York,, 29-60.
Ho, S.H., Chen, C.Y., Chang, J.S., 2012. Effect of light intensity and nitrogen starvation on CO2 fixation and
lipid/carbohydrate production of an indigenous microalga Scenedesmus obliquus CNW-N. Bioresour Technol.
113, 244-252.
Hu Q, Sommerfeld M, Jarvis E, Ghirardi M, Posewitz M, Seibert M, A., D., 2008. Microalgal triacylglycerols as
feedstocks for biofuelproduction- perspectives and advances. Plant J. . 54(4), 621-639.
Islam, M., Magnusson, M., Brown, R., Ayoko, G., Nabi, M., Heimann, K., 2013. Microalgal Species Selection
for Biodiesel Production Based on Fuel Properties Derived from Fatty Acid Profiles. Energies. 6(11), 5676-5702.
Lavín, P.L., Lourenço2, S.O., 2005. AN EVALUATION OF THE ACCUMULATION OF INTRACELLULAR

INORGANIC NITROGEN POOLS BY MARINE MICROALGAE IN BATCH CULTURES. BRAZILIAN
JOURNAL OF OCEANOGRAPHY. 53(1/2), 55-68.
Levitan, O., Dinamarca, J., Zelzion, E., Lun, D.S., Guerra, L.T., Kim, M.K., Kim, J., Van Mooy, B.A.,
Bhattacharya, D., Falkowski, P.G., 2015. Remodeling of intermediate metabolism in the diatom Phaeodactylum
tricornutum under nitrogen stress. Proc Natl Acad Sci U S A. 112(2), 412-417.
Li, J., Han, D., Wang, D., Ning, K., Jia, J., Wei, L., Jing, X., Huang, S., Chen, J., Li, Y., Hu, Q., Xu, J., 2014.
Choreography of Transcriptomes and Lipidomes of Nannochloropsis Reveals the Mechanisms of Oil Synthesis
in Microalgae. Plant Cell. 26(4), 1645-1665.
Li, Y., Horsman, M., Wang, B., Wu, N., Lan, C.Q., 2008. Effects of nitrogen sources on cell growth and lipid
accumulation of green alga Neochloris oleoabundans. Appl Microbiol Biotechnol. 81(4), 629-636.
220
Chapter 10
Litchman, E., 2007. CHAPTER 16 – Resource Competition and the Ecological Success of
Phytoplankton.Evolution of Primary Producers in the Sea. 351-375.
Makulla, A., 2000. Fatty Acid Composition of Scenedesmus obliquus- Correlation to Dilution Rates.
Limnologica. 30, 162-168.
Michael Hannon, Javier Gimpel, Miller Tran, Beth Rasala, Mayfield, S., 2010. Biofuels from algae- challenges
and potential. Biofuels. 1(5), 763-784.
Nobusawa, T., Hori, K., Mori, H., Kurokawa, K., Ohta, H., 2017. Differently localized lysophosphatidic acid
acyltransferasescrucial for triacylglycerol biosynthesis in the oleaginous alga Nannochloropsis. Plant J. 90(3),
547-559.
Paes, C.R.P.S., Faria, G.R., Tinoco, N.A.B., Castro, D.J.F.A., Barbarino, E., Lourenco, S.O., 2016. Growth,
nutrient uptake and chemical composition of Chlorella sp. and Nannochloropsis oculata under nitrogen starvation.
Latin American Journal of Aquatic Research. 44(2), 275-292.
Peter Dormann, Susanne Hoffmann-Benning, l.B., Benning’, C., 1995. lsolation and Characterization of an
Arabidopsis Mutant Deficient in the Thylakoid Lipid Digalactosyl Diacylglycero. the Plant Cell. 7, 1801-1810.
Randrianarison, G., Ashraf, M.A., 2017. Microalgae: a potential plant for energy production. Geology, Ecology,
and Landscapes. 1(2), 104-120.
Raven, J.A., 1987. The role of vacuoles. New Phytol. . 106, 357–422.
Raven, J.A., 1998. The twelfth Tansley Lecture. Small is beautiful: the picophytoplankton. Functional Ecology
12, 503-513.
Sayre, R., 2010. Microalgae: The Potential for Carbon Capture. BioScience. 60(9), 722-727.
Sharma, Y.C., Singh, V., 2017. Microalgal biodiesel: A possible solution for India’s energy security. Renewable
and Sustainable Energy Reviews. 67, 72-88.
Siaut, M., Cuiné, S., Cagnon, C., Fessler, B., Nguyen, M., Carrier, P., Beyly, A., Beisson, F., Triantaphylidès, C.,
Li-Beisson, Y., Peltier, G., 2011. Oil accumulation in the model green alga Chlamydomonas reinhardtii-
characterization, variability between common laboratory strains and relationship with starch reserves. BMC
Biotechnology. 11, 7.
Sicko-Goad, L.M., Schelske, C.L., Stoermer, E.F., 1984. Estimation of intracellular carbon and silica content of
diatoms from natural assemblages using morphometric techniques. Limnol. Oceanogr. . 29, 1170-1178.
Sirbu, D., Corno, M., Ullrich, M.S., Kuhnert, N., 2018. Characterization of triacylglycerols in unfermented cocoa
beans by HPLC-ESI mass spectrometry. Food Chem. 254, 232-240.
221
Chapter 10
Sozer, O., Kis, M., Gombos, Z., Ughy, B., 2011. Proteins, glycerolipids and carotenoids in the functional
photosystem II architecture. Front Biosci (Landmark Ed). . 16, 619-643.
Spalding, M.H., 2008. Microalgal carbon-dioxide-concentrating mechanisms: Chlamydomonas inorganic carbon

transporters. J Exp Bot. 59(7), 1463-1473.
Wang, S., Zhang, L., Yang, G., Zhu, B., Pan, K., 2015. Purification of a diatom and its identification to
Cylindrotheca closterium. J. Ocean Univ. China. 14(2), 357-361.
Webb, M.S., Green, B.R., 1991. Biochemical and biophysical properties of thylakoid acyl lipids. Biochimica et
Biophysica Acta. 1060, 133-158.
Yang, D., Song, D., Kind, T., Ma, Y., Hoefkens, J., Fiehn, O., 2015. Lipidomic Analysis of Chlamydomonas
reinhardtii under Nitrogen and Sulfur Deprivation. PLoS One. 10(9), e0137948.
Yodsuwan, N., Sawayama, S., Sirisansaneeyakul, S., 2017. Effect of nitrogen concentration on growth, lipid
production and fatty acid profiles of the marine diatom Phaeodactylum tricornutum. Agriculture and Natural
Resources. 51(3), 190-197.
Zhu, S., Wang, Y., Shang, C., Wang, Z., Xu, J., Yuan, Z., 2015. Characterization of lipid and fatty acids
composition of Chlorella zofingiensis in response to nitrogen starvation. J Biosci Bioeng. 120(2), 205-209.
222
General Conclusions
General Conclusions
Unraveling the chemistry of cocoa butter constitutes the basis for understanding the
crystallization properties of this fat. Nowadays, lipidomics offers a broad array of methods, but
not all are appropriate for the analysis of cocoa lipids. Most, if not all previous studies on cocoa
chemistry have used the standardized official methods of analysis and very little to no method
optimization. Consequently, only few lipid compounds were described. Some studies have
reported that minor components influence the physical properties of cocoa butter; for this
reason, an adequate method for the analysis and evaluation of these compounds was necessary.
One of the main objectives of this thesis was to develop the most suitable methods for the
analysis of lipids in a large number of cocoa bean samples from all the different stages of the
industrial processing. In this regard, intense optimization was performed at the extraction,
separation and detection levels. As a consequence, a long list of methods and tools for
comprehensive chemical analysis of cocoa fat is presented in this thesis. Methods such as
Soxhlet extraction method using dichloromethane as extraction solvent, thin layer
chromatography separation and solid phase extraction methods for the separation of different
lipid classes, high-resolution and tandem HPLC-MS method for lipid investigation at the
molecular level, fatty acids methyl esters synthesis and GC-FID identification and
spectrophotometric assays for color assessment and glycerol quantification. Furthermore, this
lipidomics approach was successfully used to investigate the lipids in other seeds or oil
containing plant material, like algae. The advantages of such methods are that they are easy to
apply and they offer a comprehensive picture of all the compounds present in a lipid extract.
Furthermore, these methods were applied to other plant material such as green algae and marine
benthic diatom. Thus, this lipidomic approach helped to elucidate the lipids in algae and to
observe their lipid metabolism in response to nitrogen starvation.
HPLC coupled with high resolution-MS and tandem MS enabled the characterization of 83
different TAGs in unfermented cocoa beans, 77 of which were positively dereplicated and 58
of which were reported for the first time in cocoa. The identification and characterization were
based on retention time behavior, tandem MS fragmentation and two-dimensional m/z retention
time maps as powerful visual representation tools for the identification of entire homologous
series of TAGs. Moreover, the results showed that reversed-phase C18 column packing is not
able to separate regioisomers of TAGs and therefore additional analytical tools, such as chiral
223
General Conclusions
column packings, are required to fully elucidate the entire TAGs profile. At the same time,
further studies need to be carried out in order to completely understand the influences of
different regioisomeric TAGs compositions on the physical properties of cocoa butter.
Among the newly identified compounds, TAGs carrying a hydroxyl-containing side chain
were characterized by tandem MS, derivatization, NMR, and comparison with surrogate
standards. In view of that, many different methods were required to confirm the results
obtained. These compounds are reported for the first time in cocoa and they showed to be
present in high amount in unfermented fresh beans. Hence, further research is required to
appraise their biological functions, physical characteristics, and potential industrial
valorization.
The chemical composition showed an enormous variability of the TAGs composition within
samples from different origins. However, statistical analysis results have shown that the
softness/hardness can be predicted by taking into account only four di-unsaturated TAGs
relative quantities. Therefore, a novel softness/hardness indicator tool was developed, which
could be used by industry to predict the softness/hardness of cocoa butter in a faster and more
affordable way. Nevertheless, a standard reference of the four TAGs would be expected as
means of comparison, and further physical parameters as solid fat content are required to
confirm the functionality of this tool.
A detailed analysis of fatty acids methyl esters showed significant differences among the
samples of different origins and fermentation status. Fermented dried cocoa beans have
exhibited higher levels of linoleic and palmitoleic acid. These findings were not reported before
in cocoa, and they could be used in the future to optimize and tailor cocoa lipid profile with
respect to innovative applications and health products.
The comprehensive analysis has shown that the fermentation process has a significant
impact on the lipid profile of cocoa and the present contribution is reporting this observation
for the first time in literature. There is a lot of potential to be exploited from this findings,
although it is still challenging to elucidate the mechanisms operating behind these chemical
modifications. Moreover, this study draws the attention towards the importance of
understanding the impact of fermentation microbiota on the final chemical composition of
cocoa butter. Consequently, the crystallization behavior of cocoa butter in different stages of
the fermentation process is still to be investigated and correlated with the composition of TAGs
and fermentative microorganisms.
224
General Conclusions
Elucidation of fatty acids amides as markers of the alkalization provided more

understandings into how the factory processing affects the lipid profile of the final product.
Moreover, these compounds were associated with the poor crystallization of cocoa butter.
Concerning minor components, 62 compounds were positively identified. The prominent
compounds among the minor constituents are diacylglycerols. Two regioisomers of
diacylglycerols were detected, 1,2-acyl-sn-glycerol and 1,3-acyl-sn-glycerol; 1,2-acyl-sn-
glycerol regioisomer was the most abundant molecular species. Other minor components could
be identified such as sterol esters, waxes, free fatty acids, sterol esters glucosides, and
phospholipids. Notwithstanding, there are still numerous minor components in the cocoa lipid
extract, which require future identification and quantification. With the help of the improved
tools presented in this thesis, a complete chemical elucidation of all the lipid classes is now
possible, which will lead to a deeper understanding of the crystallization of cocoa butter.
To sum up, the results of this thesis point out towards two main factors influencing the
chemical composition of CB: origin and environmental growing conditions determine the
profile, which is subsequently shaped by the processing conditions. Future work is required to
connect the findings of this thesis to the physical properties of cocoa butter and to allow
substantial industrial and commercial impact.
225

Dissertation Diana Sirbu

Uploaded by

Copyright:

Available Formats

Dissertation Diana Sirbu

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Dissertation Diana Sirbu

Uploaded by

Copyright:

Available Formats

Comprehensive Analysis of

Cocoa Lipidomics – unraveling the unknown

a Thesis submitted in partial fulfillment

Approved Dissertation Committee

Prof. Dr. Nikolai Kuhnert, Jacobs University

Prof. Dr. Matthias Ullrich, Jacobs University

Dr. Frédéric Depypere, Barry Callebaut

Date of Defense: June 1st, 2018

Department of Life Sciences and Chemistry

MOAH mineral oil aromatic hydrocarbons

Figure 1.1 Lipid classes and prominent examples.

Table 1.1 The number of possible TAGs.

1.1.1 General concept of Lipidomics

Lipidomics is an emerging field of metabolomics centered on the analysis of all lipids

1.1.2 Lipid classification and functions

Fatty acids and conjugates

Sterols and derivatives

Figure 1.1 (continued) Lipid classes and prominent examples.

Table 1.1 The number of possible TAGs.

maintenance of electrochemical gradients; and a recent described function as cofactors in

1.1.3 Experimental approaches in lipidomics research

1.1.4 Plant lipidomics

1.1.5 Lipidomics and food science

1.2 Theobroma cacao – from bean to bar

1.2.1 History, origin, and cocoa varieties

International Cocoa Germplasm Database (“ICGD Homepage”) stores records on about

1.2.3 Post-harvesting treatments

1.2.4 Factory production process

1.2.5 Economic impact and health benefits

activities of cocoa constituents on diet-induced obesity and insulin resistance was

Beckett, S. T. (2008). The Science of Chocolate. Royal Society of Chemistry.

ICGD Homepage. (n.d.). Retrieved April 26, 2018, from http://www.icgd.reading.ac.uk/

Niederer, M. (2014). 1.1. 1 Lippenpflegeprodukte/Mineralparaffine (MOSH/MOAH) und allergene Duftstoffe.

Update, C. M. (2014). World Cocoa Foundation. Retrieved March.

World Cocoa Foundation. Retrieved February 26, 2018, from http://www.worldcocoafoundation.org/

2 Review on cocoa lipidomics – state of knowledge

Cocoa butter is an essential ingredient in many commercial products. The chemical

2.2 From cocoa tree to cocoa butter

2.3 Quality criteria of cocoa butter

2.4 Chemical composition background and significance

2.4.1 Cocoa fat content

2.4.2 Fatty acids composition

2.4.3 Triacylglycerol composition

oleoyl-sn-glycerol (rac-POS, >90% regioisomeric purity) and, to a lesser extent, 1,3-

Figure 2.4 Triacylglycerol structure of a 1,3-disaturated-2-unsaturated-sn-glycerol represented by

Theobroma cacao genome contains 84 orthologues genes potentially involved in lipid

2.4.4 Minor components

2.5 Correlation of chemical composition with the physical

2.5.1 Cocoa butter polymorphism

2.5.2 Crystallization and melting behavior

Tempering is a precrystallization technique used to control cocoa butter crystallization. This

2.5.3 Influence of the origin – softness assessment

2.6 Cocoa butter alternatives

2.8 Aim of the thesis

Beckett, S. T. (2008). The Science of Chocolate. Royal Society of Chemistry.

Wood, G. a. R. (2008). From Harvest to Store. In Cocoa (pp. 444–504). Wiley-Blackwell.

3 Exploring cocoa lipid classes

Supplementary material is available on request. E-mail: [email protected]

3.2 Material and Methods

3.2.1 Chemicals and reagents