A non-targeted metabolomic approach to identify food markers to support discrimination between organic and conventional tomato crops

Journal of Chromatography A, 1546 (2018) 66–76

Contents lists available at ScienceDirect

Journal of Chromatography A
journal homepage: www.elsevier.com/locate/chroma

A non-targeted metabolomic approach to identify food markers to
support discrimination between organic and conventional tomato
crops
María Jesús Martínez Bueno, Francisco José Díaz-Galiano, Łukasz Rajski, Víctor Cutillas,
Amadeo R. Fernández-Alba ∗
University of Almería, Department of Physics and Chemistry, Agrifood Campus of International Excellence (ceiA3), Ctra. Sacramento s/n, La Ca˜
nada de San
Urbano, 04120, Almería, Spain

a r t i c l e

i n f o

Article history:
Received 7 November 2017
Received in revised form 26 February 2018
Accepted 1 March 2018
Available online 3 March 2018
Keywords:
Fingerprint
Authenticity
Natural food components
Pesticides
HRAMS

IRMS

a b s t r a c t
In the last decade, the consumption trend of organic food has increased dramatically worldwide. However, the lack of reliable chemical markers to discriminate between organic and conventional products
makes this market susceptible to food fraud in products labeled as “organic”. Metabolomic fingerprinting
approach has been demonstrated as the best option for a full characterization of metabolome occurring in plants, since their pattern may reflect the impact of both endogenous and exogenous factors.
In the present study, advanced technologies based on high performance liquid chromatography-highresolution accurate mass spectrometry (HPLC-HRAMS) has been used for marker search in organic and
conventional tomatoes grown in greenhouse under controlled agronomic conditions. The screening of
unknown compounds comprised the retrospective analysis of all tomato samples throughout the studied period and data processing using databases (mzCloud, ChemSpider and PubChem). In addition, stable
nitrogen isotope analysis (␦15 N) was assessed as a possible indicator to support discrimination between
both production systems using crop/fertilizer correlations. Pesticide residue analyses were also applied
as a well-established way to evaluate the organic production. Finally, the evaluation by combined chemometric analysis of high-resolution accurate mass spectrometry (HRAMS) and ␦15 N data provided a robust
classification model in accordance with the agricultural practices. Principal component analysis (PCA)
showed a sample clustering according to farming systems and significant differences in the sample profile was observed for six bioactive components (L-tyrosyl-L-isoleucyl-L-threonyl-L-threonine, trilobatin,
phloridzin, tomatine, phloretin and echinenone).
© 2018 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY license
( />
1. Introduction
With a worldwide harvest of over 162 million tons annually,
tomato (Solanum lycopersicum L.) is the second most important vegetable crop in the world next to potato (FAOSTAT Database) [1].
Spain cultivates 25% of the tomatoes produced in Europe (approx.
5 million tons), making the country the second largest producer in
the European Union (EU), behind Italy, according to data released
by the EU statistics office, Eurostat (EUROSTAT Database) [2]. In
the last decade, the production and consumption of organic food
has increased dramatically worldwide. The EU organic food market is the second largest in the world behind the US´ı. In Spain, the
total number of hectares dedicated to organic tomato production in

∗ Corresponding author.
E-mail address: (A.R. Fernández-Alba).

Andalusia reached 350.55 ha in 2014, 66.3% more than the previous
year.
Organic farming in the EU is supported by EU law, Regulations (EC) No 834/2007 [3] and 889/2008/EC [4], with detailed
rules on production, labelling and control via an organic action
plan [5]. Organic farming does not permit the use of synthetic
chemicals, including pesticides and fertilizers. Nonetheless, the
presence of synthetic pesticides in organic food may arise from
environmental pollution (i.e. from neighbouring farms, contaminated soils, etc.). Up to now, little data have been available in
the scientific literature on pesticide residues in organic foods [6].
Thus, several issues within the regulatory framework need to be
improved, as recognised by the Commission [7]. The lack of knowledge and the lack of reliable markers to discriminate between
organic and conventional products make this market susceptible
to foods labeled as “organic” that have, in fact, been produced conventionally. Certainly, instances of fraud published about the sale of

/>0021-9673/© 2018 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY license ( />

M.J. Martínez Bueno et al. / J. Chromatogr. A 1546 (2018) 66–76

organic products are few and far between although the International
Organic Accreditation Service has expressed doubts about whether
this type of fraud is indeed small or whether the low incidence of
fraud detection is due to authorities, or competent agencies, not
having sufficient capability to perform adequate inspections [6].
In the area of analytical chemistry, diode array detector tandem
mass spectrometry (DAD-MS/MS), liquid chromatography tandem mass spectrometry (LC–MS/MS), inductively coupled plasma
mass spectrometry (ICP-MS), mid-infrared spectroscopy (MIR);
nuclear magnetic resonance (NMR) spectroscopy and isotope ratio
mass spectrometry (IRMS) have been the most common analytical
detection technologies investigated as tools for organic tomatoes

authentication [8–10]. With regard to IRMS, the analysis of elements such as nitrogen (N) has been considered as a potential
indicator for food authentication control to support the discrimination between conventionally and organically grown produce
using crop/fertilizer correlations. However, the disadvantages of
this approach are that many factors are known to affect the nutrient
content of crops (e.g. fertilizer use, the vegetables’ growing time,
the use of leguminous plants for enhancing the nitrogen fertility of
soil, water type used for irrigation, etc.) [11,12].
During the last years, non-targeted methods have been the
basis of the discoveries within “omics”, such as metabolomics or
foodomics, all centred on obtaining mass spectra (MS) for a whole
range of masses and the subsequent identification and characterization of molecules responsible for a certain attribute [13]. The
MS profile of a food sample can be regarded as an analytical signature of the food product and thus can help in discriminating
between different production practices, reflecting the impact of
both endogenous and exogenous factors as well as food properties. Primary metabolites are generic components, while secondary
metabolites are food specific components and potentially reliable
markers. Some works have reported a higher content of these
compounds, such as polyphenols, in organic food [9,14–16]. The
different content of bioactive compounds in organic and conventional agricultural practices has been attributed to the absence
of synthetic pesticides in organic farming. It leads plants to synthesize more secondary metabolites to protect themselves against
phytopathogens than tomato plants grown under conventional
conditions [17]. But also, some authors have reported that the type
of fertilizers and the availability of inorganic nitrogen can modulate the plant’s biosynthetic pathways and therefore the levels
of the natural food components [17,18]. High-resolution accurate
mass spectrometry (HRAMS) systems seem to be the best candidates for MS-profiling studies, mainly due to the last advances of
using the full-scan acquisition mode with high sensitivity, along
with high-resolving power (>50,000 FWHM) and accurate mass
measurement (1–5 ppm) [17,19,20]. Nevertheless, the potential of
these tools to achieve a reliable authentication strategy applicable
in routine practice has not been extensively explored for application to date.
On the other hand, given the large quantity of signals detected

in HRAMS, the data sets are extremely complex so, regardless
of whether the goal is group establishment or marker selection,
it is necessary to use statistical tools to extract information. To
date, among the most common approaches are Principal Component Analysis (PCA), Linear Discriminant Analysis (LDA) or
Partial Least Squares-Discriminant Analysis (PLS-DA) [21,22]. Several comparative research studies have been conducted to compare
the composition of various crops originated from conventional and
organic farms, but most had a poor-designed experimental study
[10,17,23,24,25], obtaining variable results that make it impossible
to reach a definitive conclusion on finding significant differences
between crops.
Because of this, the aim of this study was to investigate the
potential of advanced technologies such as HRAMS to identify suit-

67

able markers capable of distinguishing between organically and
conventionally grown tomatoes under controlled agronomic conditions, in a climatic region of leading EU production such as the
Mediterranean. Moreover, another innovative aspect of the present
work concerns the data combination of IRMS and LC-HRAMS analysis by chemometric methods as a tool to get robust classification
models to discriminate between different practices. Finally, the last
objective was to carry out a pesticide residue analysis to evaluate the influence of synthetic pesticides present in the secondary
metabolites content.
2. Experimental section
2.1. Reagents
Water used for LC–MS analysis was obtained from a Milli-Q
water purification system (Direct-QTM 5 Ultrapure Water System
Millipore, Bedford, MA, USA), which provided a specific resistance
of 18.2 M cm. Methanol (MeOH) and acetonitrile (AcN) HPLC–MS
grade were supplied from Merck (Darmstadt, Germany). Anhydrous magnesium sulphate was supplied by Panreac (Barcelona,
Spain). Formic acid (98% purity) was purchased from Fluka (Buchs,

Switzerland).
The standards dimethoate-d6 (CAS 1219794-81-6), dichlorvosd6 (CAS 203645-53-8), tomatine hydrate (CAS 17406-45-0),
tomatidine hydrochloride (CAS 6192-62-7), phloridzin (CAS 6081-1), phloretin (CAS 60-82-2), (±)-catechin (CAS 7295-85-4),
acetanilide (CAS 103-84-4), IAEA-N-1 (CAS 7783-20-2) and IAEAN-2 (CAS 7783-20-2) were purchased from Sigma-Aldrich Quimica
S.A (Madrid, Spain). All were of analytical quality. Individual stock
standard solutions were prepared at about 1 mg/mL in acetonitrile,
and stored at −20 ◦ C. Individual standard solutions of dimethoated6 and dichlorvos-d6 prepared in methanol were used as injection
internal standards for LC analysis in order to ensure quality measurements.
2.2. Sample crops and preparation
Cocktail tomatoes (Solanum lycopersicum var. cerasiforme) were
produced in greenhouses under controlled agronomic conditions
in two different farms, located in Almeria (Southeast Spain), using
organic and conventional cultivation methods. Organic samples
were provided by a local company, expert in crops cultivated
in line with EU organic production legislation (EC 834/2007 and
889/2008). Ecological manure was used for their production without the usage of synthetic fertilizers nor pesticides. Plant protection
against pests was based on herbal extracts. Conventional samples
were provided by a farmer of the same area. In this crop system,
mineral fertilizers containing soluble inorganic nitrogen and other
mineral contents (K, Ca, Na, Mg, Fe, Zn) were applied. In this case,
commercial pesticides were employed for protection against pests.
The tomatoes were produced from September-2016 to May-2017,
which corresponds to a full harvest cycle. To obtain representative
samples, and thus to compensate for possible variability in their
composition, samples were taken from different areas of the field,
over a full week, monthly, and then pooled (approx. 5 kg). Samples
were harvested in full maturity. Additionally, 11 tomato samples
were purchased from different local markets (labeled as organic or
non-organic) in order to check the classification capability of the
developed model. A total of 25 tomato samples were analyzed (14

produced under controlled agronomic conditions and 11 from local
markets). Detailed information on samples, crop system, farms,
fertilisation and plant protection is shown in Table 1.
Tomatoes (skin, flesh, and pips) were blended to obtain a tomato
paste, freeze-dried and ground to a fine powder of an average parti-


68

M.J. Martínez Bueno et al. / J. Chromatogr. A 1546 (2018) 66–76

Table 1
Detailed information on the tomato samples produced under controlled agronomic
conditions, crop system, farms, fertilisation and plant protection carried out in this
study.
Crop system

Manure

Fertilizer

Pest control

# samples

Name

Organic
Conventional

Ecological
Conventional

–
Mineral

Herbal extracts
Pesticides

7
7

O
C

cle size of about 1 mm. Dried tomato powder samples (0.5 ± 0.01 g)
were weighed in a 50 mL polypropylene centrifuge tube and then
2 ␮L of a 10 mg/L methanolic internal standard solution was added
(dichlorvos-d6). Next, the tomatoes were rehydrated by adding
0.5 mL of ultrapure water, and the mixture was vortexed for 30 s.
The tubes were then automatically shaken for 5 min, after the addition 5 mL of methanol. After that, 0.5 g Mg2 SO4 (anhydrous) plus
0.25 g NaCl were added directly to each tube and the mixture was
shaken again (automatically) for 5 more min. Finally, a centrifugation step (3500 rpm/1730 × g, 5 min) was performed, and 0.5 mL of
the supernatant was diluted with 0.5 mL of ultrapure water. Samples were stored at −20 ◦ C until analysis by LC-Q-Orbitrap-MS.
2.3. HRAMS analysis
To evaluate the differences between organic and conventional
production systems, a non-targeted analysis was applied using an
LC-ESI-Q-Orbitrap. For the LC separation, UHPLC DionexTM Ultimate 3000 (Thermo ScientificTM , San Jose, USA) was used. Mobile
phase A was 98% water and 2% methanol whereas mobile phase
B was 98% methanol and 2% water; both mobile phases contained

5 mM of ammonium formate and 0.1% formic acid. Separation was
carried out on a Phenomenex Luna C8 column (mobile phase flow:
350 ␮L/min). The length, diameter and particle size were 100 mm,
2.0 mm and 3 ␮m, respectively. The column was thermostated at
30 ◦ C. The mobile phase gradient started form 100% of mobile phase
A and maintained for 1 min, from 1 to 2 min, the amount of mobile
phase B increased to 30%, from 2 to 3 min to 50%, from 3 to 11 min
to 100%. 100% of B was maintained until 14 min. Following this, the
mobile phase was changed to 100% A and maintained over 3 min for
re-equilibration. The injection volume was 10 ␮L. The autosampler
was thermostated at 10 ◦ C.
A Q-Orbitrap (Thermo Scientific, Bremen, Germany) mass spectrometer was equipped with Heated Electrospray Ionization Source
(HESI II). The HESI parameters in positive polarity were as follows:
sheath gas flow rate: 40; auxiliary gas flow rate: 5; sweep gas
flow rate: 1; spray voltage: 3.00 kV; capillary temperature: 280 ◦ C;
S–lens RF level: 55.0; heater temperature: 350 ◦ C. The instrument
was operated in full scan/all ion fragmentation MS2 (AIF). AIF is an
acquisition mode in which all precursor ions are fragmented without a preselection in the quadrupole. Fragmentation is obtained
with the higher energy collision-induced dissociation (HCD) cell,
located at the far side of the quadrupole ion trap “C-trap”. During filling of the HCD collision cell, the energy can be set to step
between values at specified percent values around the chosen middle energy regardless of the ion’s characteristics. The parameters of
full scan analysis were as follows: scan range 74−1100 m/z, resolution 70,000 FWHM and automatic gain control (AGC) target 1 × 106 .
AGC is used to regulate the total number if ions collected in the
C-trap before being injected into the orbitrap for analysis. The maximum ion injection time (max IT) was set to ‘AUTO’, Parameters of
“All Ion Fragmentation MS2 mode” were as follows: resolution 70
000 FWHM, collision energy CE 30 V, AGC target 1 × 106 , max IT
auto, scan range 74–1100 m/z. An external mass calibration and
quadrupole calibration were carried out daily, using a mixture
of n−butylamine, caffeine, Ultramark 1621 and Met-Arg-Phe-Ala
(MRFA).

2.4. IRMS analysis
Dried tomato powder samples (2 ± 0.1 g) were weighed into
tin capsules. Nitrogen isotope composition was determined using
a Flash EA1112 elemental analyzer coupled to a Finnigan Deltaplus isotope ratio mass spectrometer (Thermo Scientific, Bremen,
Germany). All samples were analyzed in duplicate, and for the
majority of samples the absolute difference between duplicate
measurements was <0.2‰. Each batch of samples included replicate analyses of the in-house standard, IAEA-N-1 (␦15 N = +0.4‰)
and IAEA-N-2 (␦15 N = +20.3‰), which was used for the drift
correction of raw analytical measurement data. The long-term performance of the mass spectrometer was monitored by analysis of a
secondary reference material, acetanilide (␦15 N = ± 0.15‰), which
was included with every batch of samples. Nitrogen isotope ratios
were reported with respect to air nitrogen.

2.5. Data processing and chemometric analysis
Compound Discoverer software version 2.1 (Thermo Scientific)
in combination with library searching was the workflow used to
identify potential marker compounds. This tool allows automatic
analyte detection based on the presence of the exact mass of precursor ions, with a mass tolerance of 5 ppm. To reduce chemical
interferences from the matrix, a mass tolerance for alignment of
5 ppm, an intensity tolerance of 30%, a total intensity threshold
of 1 × 105 , a maximum shift of 0.5 min and a signal-to-noise (S/N)
threshold of 10 were the filters set.
In a second step, the data were manually processed for the
comprehensive identification of chemical markers. The identification of each compound was based on the acquisition of at least 3
diagnostic ions (one precursor ion and two fragment ions) with
a mass accuracy <5 ppm. For that, Xcalibur 4.0,Mass Frontier 7.0
and TraceFinder 4.1 software’s (Thermo Scientific) were further
employed to data review and check the diagnostic ions obtained
in MS and MS2 spectra.

Finally, chemometric tools were employed for the processing of
MS and ␦15 N data Principal component analysis (PCA) was used to
differentiate the statistical significance between farming systems.
The statistical analyses were performed using R software (version
3.3.3).

2.6. Method validation
Rigorous validation assays according to SANTE/11945/2015 and
ISO/IEC 17025:2005 Guidelines were performed to ensure high
quality analytical measurements [26,27]. Therefore, after a preliminary screening of the components in tomato samples, we estimated
the amount of the compounds detected in organic and conventional
tomatoes. Then, analyte quantification of the real samples was performed by the standard addition method. In addition, recoveries,
linear dynamic range, matrix effect, sensitivity (limit of detection
and limit of quantification, LOD and LOQ, respectively) and precision (intra and inter-day) both in solvent and matrix were also
evaluated.
Furthermore, to ensure quality measurements, each day before
analysis, the calibration of the Exactive mass spectrometer was
performed using the calibration mixture. Afterwards, a control
standard mixture (10 ng/mL) was injected with the purpose of
checking the performance of the HPLC, the analytical column and
the Orbitrap MS system. A continuous monitoring of the quality of
the analytical procedure was carried out through the inclusion of a
blank (solvent) during the day-work sequence. Finally, an injection
internal standard (dimethoate-d6) was also added to each sample


M.J. Martínez Bueno et al. / J. Chromatogr. A 1546 (2018) 66–76

before its analysis with the purpose of checking the performance
of each analysis.

3. Results and discussion
3.1. Tentative identification of markers by HRAMS
To date, a reliable characterization and determination of food
markers in terms of natural food components (NFCs) to distinguish
between organic and conventional tomato crops under controlled
agronomic conditions (greenhouses), and in a climatic region of
leading EU production such as the Mediterranean has not been
found [19,25].
In this study, an analytical approach was tested on organic and
conventional tomato samples obtained under controlled field trials in Almeria (Spain). A screening of NFCs was accomplished.
MeOH solvent was able to successfully extract polar and semi-polar
metabolites of great interest. Then, for the identification of food
markers and to reduce chemical interferences from the matrix due
to co-eluting and isobaric compounds, the following filters were
set: a mass tolerance for alignment of 5 ppm, an intensity tolerance
of 30%, a maximum shift of 0.5 min and an S/N threshold of 10.
In a second step, a volcano plot was generated with Compound Discoverer 2.1 software to further characterize significant
differences between both farming systems. The volcano graphical
display plots the significance of the observed changes, named “fold
change”. Fold changes are usually expressed in log2 values to distinguish changes that are two-fold or higher. These changes are
plotted in the context of the confidence with which those changes
were observed, and the confidence is expressed as a negative log10
p-value (smaller log p-value is less confident). The signals of analytes that increase or decrease significantly are therefore high in
confidence. Out of all the probable compounds (approx. 17.000
components), less than 10 compounds displayed a signal intensity
≥105 and a log2 values ≥2 in organic tomato samples. In addition, a
CV% ≤20% between injections (each sample was injected by triplicate, n = 3) was another fixed parameter. Thereby, a significant data
reduction was achieved.
The third step was the structural elucidation of markers based

on the acquisition of at least 3 diagnostic ions with a mass accuracy <5 ppm. The more demanding requirements regarding mass
spectrometric confirmation currently set by EU regulations were
taken into account for confirmation and quantification of each target compound. Commission Decision 2002/657/EC [28] describes a
system of identification to obtain a minimum of four identification
points, in which, detection of transition products in HRAMS yields
two identification points per ion. SANTE/11945/2015 Guideline
[26] recommended the acquisition of ≥2 diagnostic ions (preferably including the quasimolecular ion and at least one fragment
ion) and a mass accuracy <5 ppm. In this way, tentative identification was made by comparing retention times (when standards
were available), MS data (accurate mass and isotopic distribution)
and MS/MS data (fragmentation pattern in positive mode) of the
compounds detected with the tomato compounds reported in literature and searching in the existing on-line public databases such as:
mzCloud <www.mzcloud.org>, ChemSpider com>, PubChem <www.pubchem.ncbi.nlm.nih.gov> and Metlin
<www.metlin.scripps.edu>.
Taking all these considerations into account, some of the peaks
detected could be identified. Table 2 presents all the mass spectrometric criteria used in this study: retention time (Rt), accurate
mass values for the precursor and fragment ions (m/z), elemental composition and mass deviation (ppm) values. As shown in
Table 2, three diagnostic ions (one precursor ion and at least two
fragment ions) with a mass accuracy <5 ppm (even to m/z < 200),

69

were used for a tentative structural elucidation. Accurate mass
measurements of the precursor ions at [M+H]+ m/z 497.2600,
437.1453, 459.1266, 1034.5520, 275.0910 and 551.4253 correlated well to the theoretical exact mass (m/z 497.2606, 437.1442,
459.1261, 1034.5530, 275.0914 and 551.4247, respectively), with
mass deviation or mass errors below 2.5 ppm (−1.3, 2.5, 1.2, −1.0,
−1.5 and 1.2, respectively). The list of possible empirical formulas proposed for the protonated molecules [M+H]+ yielded as
first options: C23 H37 N4 O8 , C21 H25 O10 , C21 H24 O10 Na, C50 H84 NO21 ,
C15 H15 O5 and C40 H55 O, respectively. Moreover, the data obtained

for the fragment ions contributed to a better confidence in the
identification. Mass Frontier 7.0 software was used to match each
fragment obtained in MS/MS mode by LC-Q-Orbitrap-MS in combination with the databases (above-cited). Therefore, a total of
six bioactive components, belonging to various metabolite classes,
including flavonoids, glycoalkaloids, carotenes and oligopeptides
were identified as markers of organic tomato samples. Two compounds ((±)-catechin and tomatidine) were finally removed from
the list because they did not show a constant relation between
organic/conventional throughout a full harvest cycle.
The theoretical exact mass of the precursor ion at [M+H]+
m/z 497.2606 was hypothesized as the oligopeptide L-tyrosyl(L-Tyr-L-Ile-L-Thr-L-Thr)
L-isoleucyl-L-threonyl-L-threonine
(C23 H37 N4 O8 ) and confirmed through accurate mass studies using
a mass resolving power of 70,000 FWHM. In fact, the analytical
standard of this compound was not commercially available. But,
an additional experiment based on a Parallel Reaction Monitoring
(PRM) analysis was employed using the theoretical m/z value for
the [M+H]+ ion of L-Tyr-L-Ile-L-Thr-L-Thr, in order to obtained
a high confident targeted confirmation. PRM methodology uses
the quadrupole of the Q Exactive to isolate a target precursor ion,
then fragments the targeted precursor ion in the collision cell, and
then detects the resulting product ions in the Orbitrap mass analyzer with high resolution and high accuracy. This kind of analysis
provides high selectivity and high sensitivity, eliminating the background interference ions and false positives. Thus, the low mass
error obtained for fragment ions at m/z 378.2023 (C19 H28 N3 O5 ),
277.1547 (C15 H21 N2 O3 ) and 221.1132 (C8 H17 N2 O5 ), with errors of
−2.3, −2.8 and −3.3 ppm, respectively, was considered sufficient
for its tentative identification. This oligopeptide, to the best of our
knowledge, has never been reported in organic tomato samples
until now.
The calculated masses of the fragment ions at m/z
902.5108(C42 H78 O20 ),

740.4580
(C39 H66 NO12 ),
578.4051
(C33 H56 NO7 ) and 416.3523 (C27 H46 NO2 ), with errors of −1.9,
−1.4, −1.1 and −1.5 ppm, respectively, were sufficiently consistent
to determine the elemental composition, and thus to identify
this compound as the glycoalkaloid tomatine (C50 H84 NO21 ). The
ions at m/z 902, 740, 578 and 416 correspond to the loss of a
pentose moiety [M+H–132], a pentose and one hexose unit [M+H
– 132–162], a pentose and two hexose units [M+H–132 − 2 × 162]
and a pentose and three hexose units [M+H – 132 − 3 × 162],
respectively. In addition, analytical standard-grade tomatine was
commercially available and therefore the peak at Rt 6.2 min as well
as its fragment ions could be confirmed.
Following this workflow, three flavonoids (trilobatin, phloridzin
and phloretin) were also identified. An accurate mass value of
437.1453 (theoretical m/z 437.1442), obtained for the presumed
protonated molecule, enabled the determination of the elemental composition C21 H25 O10 with a mass error of 2.5 ppm. The peak
at 5.10 min was hypothesized as either the compounds trilobatin
or pterosupin. PRM mode was again employed to obtain a high
confident targeted confirmation both standard compounds were
not available in laboratory. Fig. 1 shows a comparison of the spectra obtained for trilobatin in organic tomato samples using both
acquisition modes (AIF vs. PRM). As can be seen, PRM analysis


70

M.J. Martínez Bueno et al. / J. Chromatogr. A 1546 (2018) 66–76

Table 2

Natural food components (NFCs) identified as markers in tomatoes by HRAMS.
Precursor ion

Fragment ions

Rt

Accurate mass
value [M+H]+

Proposed formula
[M+H]+

Mass deviationa
(ppm)

Accurate mass
value (m/z)

Proposed formula

Mass deviationa
(ppm)

Identified
compounds

3.95

497.2606

C23 H37 N4 O8

−0.7

5.10

437.1442

C21 H25 O10

2.5

459.1261

C21 H24 O10 Na

0.9

6.20

1034.5530

C50 H84 NO21

−0.8

6.60

275.0914

C15 H15 O5

−0.7

12.74

551.4247

C40 H55 O

1.0

C19 H28 N3 O5
C15 H21 N2 O3
C8 H17 N2 O5
C21 H19 O7
C20 H17 O6
C12 H11 O5
C15 H14 O5 Na
C6 H10 O5 Na
C42 H78 O20
C39 H66 NO12
C33 H56 NO7
C27 H46 NO2
C7 H5 O4
C7 H7 O
C6 H7 O
C26 H33
C19 H25 O

−2.3
−2.8
−3.3
−2.8
−2.6
−3.2
−0.6
−0.2
−1.9
−1.4
−1.1
−1.5
−3.0
−4.7
−4.4
−1.6
−3.8

L-tyrosyl-Lisoleucyl-L-threonylL-threonineb
Trilobatinb

5.45

378.2023
277.1547
221.1132
383.1125
353.1019
235.0600

297.0733
185.0420
902.5108
740.4580
578.4051
416.3523
153.0183
107.0491
95.0491
345.2577
269.1900

Phloridzinc
Tomatinec

Phloretinc

Echinenoneb

Rt: retention time.
a
Values reported in the worst case.
b
marker tentatively identified.
c
marker identified with the injection of the reference standard.

Fig. 1. Comparison of the spectral obtained for trilobatin in an organic tomato sample using both acquisition modes (AIF vs. PRM) by HRAMS.

provided a high selectivity and high sensitivity, eliminating the

background interference ions. Fragment ions at accurate mass m/z
383.1125 (C21 H19 O7 , −2.8 ppm), 353.1019 (C20 H17 O6 , −2.6 ppm)
and 235.0600 (C12 H11 O5 , −3.2 ppm), supported the structural elucidation of trilobatin. This isomer of phloridzin (trilobatin or
phloretin 4 -glucoside), has been to our knowledge reported here
for the first time in tomato. The candidate pterosupin, with the
same formula C21 H19 O7 and m/z 437.1442, was finally discarded
due to the differences found in the fragmentation spectra. Phloridzin was confirmed with the precursor ion and two fragment
ions at m/z 297.0733 corresponding to the loss of a hexose moiety [M+H – 162] and at m/z 185.0420 corresponding to a hexose
moiety plus the sodium ion [162+23], with a mass error of −0.6

and −0.2 ppm, respectively; these values correlate well to the elemental composition C15 H14 O5 Na and C6 H10 O5 Na, respectively. As
for the last flavonoid, the low mass errors found allowed us to
identify the compound phloretin with high security- for the theoretical protonated molecule at m/z 275.0914 (−1.5 ppm) and for
the fragment ions at m/z 153.0183 (C7 H5 O4 , −3.0 ppm), 107.0491
(C7 H7 O, −4.7 ppm) and 95.0491 (C6 H7 O, −4.4 ppm). Available analytical standards of phloridzin and phloretin enabled us to confirm
their Rt at 5.45 and 6.60 min, respectively, as well as their proposed
fragment ions.
Finally, the xanthophyll echinenone at Rt 12.74 min was tentatively identified with the precursor ion and two fragment ions
at m/z 345.2577 and 269.1900, which corresponding with −1.6


M.J. Martínez Bueno et al. / J. Chromatogr. A 1546 (2018) 66–76

71

Table 3
Method validation Results.
Name

Matrix Effect (%)

Inter/intra-day (R.S.D, %)

Rec. (%, ±␴)

LOD (␮g/kg DW)

LOQ (␮g/kg DW)

MDL (␮g/kg FW)

MQL (␮g/kg FW)

±-Catechin
Phloridzin
Tomatine
Phloretin
Tomatidine

50
36
−55a
−15
−53a

14/18
8/15
3/12
7/11
2/8

78 (9)
98 (8)
55 (4)
102 (4)
60 (4)

0.6
1.0
0.5
0.9
0.5

2.0
3.3
1.6
3.0
1.6

1.0
1.6
0.8
1.4
0.8

3.2
5.3
2.6
4.8
2.6

Inter/intra-day: repeatability/reproducibility of the instrumental method (R.S.D,%); Rec: recovery average values (n = 6); ␴: dispersion from the average recovery values;
LOQ: limit of quantification; LOD: limit of detection; DW: dry weight; MQL: method quantification limit; MDL: method detection limit, FW: fresh weight.
a
Matrix effect evaluated using one of the fragment ions (578.4051 for tomatine; 398.3410 for tomatidine).

and −3.8 ppm; these values correlate well to the elemental composition C26 H33 and C19 H25 O, respectively. These fragmentation
results were confirmed by PRM analysis using the theoretical exact
mass at m/z 551.4247. The product ions reported were attributed
to cleavages of the polyene chain by direct retro-ene and vinylally fragmentation reactions, as it has been previously described
by other authors for the case of carotenoids [29].
All these secondary metabolites are attracting considerable
attention from producer and consumers due to their antioxidant
activity and nutritional properties. Detailed information about all
mass fragmentation mechanisms for the six markers identified by
HRAMS has been included as Supplementary Material (Fig. S.1).

3.2. Method validation results
An initial screening of bioactive components was carried out
both in organic and conventional tomato samples, comparing two
different extraction solvents, methanol vs acetonitrile. Methanol
was selected as the extraction solvent because it offered better
performance than acetonitrile (see Fig. S.2 in Supplementary Material). Once the food markers were identified and their concentration
levels estimated (between 5–500 ␮g/kg in fresh weight (FW)), the
validation of the analytical method was performed for quantification purposes, for those analytes for which suitable standards were
available.
Good results were obtained in terms of linearity, precision and
sensitivity. The repeatability/reproducibility of the instrumental
method was estimated by determining the intra- and inter-day relative standard deviation (R.S.D., %) by the repeated analysis (n = 5)
of a spiked conventional tomato extract at 10 ␮g/kg from run-torun over one day and five days, respectively. The precision of the

method was also acceptable for quantification. Inter-day repeatability and intra-day reproducibility were less than 14 and 18%,
respectively, for all analytes, (±)-catechin being the compound
which showed higher RSD values.
The sensitivity of the method was calculated in terms of LOQs
and LODs, which were calculated as the quantity of analyte able to
produce a chromatographic peak 10 and 3 times higher than the
noise of the baseline in a chromatogram (S/N = 10 and 3, respectively) of a non-fortified sample. The LOD and LOQ values ranged
between 0.5–1 ␮g/kg and between 1.6–3.3 ␮g/kg for all the studies
compounds in dry weight (DW) tomato samples (see Table 3). However, method quantification limits (MQLs, ␮g/kg in fresh weight
(FW)) are more appropriate for establishing analysis detection
thresholds, since the calculation must take into account the stages
of dilution or pre-concentration made during the sample preparation. In our case, a 1.6-fold pre-concentration step is applied to
the initial sample (fresh tomato). Therefore, and considering LOQ
values obtained in matrix, MQLs ranged from 2.6 to 5,3 ␮g/kg for
all compounds (see Table 3). Thus, the developed analytical method
allowed the quantification of these bioactive compounds at concentration levels on the order of a few ␮g/kg (ppb) from fresh tomato
samples.

The matrix effects and linearity of the analytical response
were also evaluated using solvent and matrix-matched calibration curves prepared in conventional tomato samples at five
concentration levels covering three orders of magnitude, from
LOQ–500 ␮g/kg, based on linear regression and squared correlation coefficient (r2 ). All the studied compounds presented very
good response of three orders of magnitude, with correlation coefficients higher than 0.997 in all cases. The matrix effect was studied
by comparison of the slopes of the calibration curves in solvent
and in matrix. When the percentage of the difference between
these slopes is positive, then there is signal enhancement, whereas
a negative value indicates signal suppression. According to our
results shown in Table 3, one compound showed no matrix effect
(<20%, because this variation is close to the repeatability values),
while three compounds ((±)-catechin, tomatidine and tomatine)

presented a high matrix effect (≥50%).
Recovery studies were performed at a concentration level of
50 ␮g/kg (n = 6) using conventional tomato samples. The average
recovery values were higher than 75% for all the studied flavonoids
((±)-catechin, phloridzin and phloretin), and higher than 55% for
the glycoalkaloid compounds (tomatidine 60%; tomatine 55%).
Deviation data obtained highlighted the precision of the extraction method (see Table 3). No suitable standards were available for
L-Tyr-L-Ile-L-Thr-L-Thr, trilobatin and echinenone. Therefore, no
information about their behavior during sample extraction could
be obtained for them.
In view of these results (matrix effect and recovery data), analyte
quantification of the real samples was performed by the standard addition method (five points). This procedure is designed to
compensate high matrix effects, low extraction recoveries, especially where isotopically labeled standards are not available or
are too costly, and in cases when no suitable commercial reference standards or blank material is available for the preparation of
matrix-matched standard solutions. For this purpose, five portions
of dehydrated tomato samples (0.5 g) were spiked with the same
amount of the internal standard (dichlorvos-d6). Then, known
amounts of the available analytes were added to the other four
test portions immediately prior to extraction. The initial spiked
level was chosen to increase the original content of bioactive
components in organic and conventional tomatoes by a factor
between 2 and 3. The concentration of bioactive components in
the “unspiked” extract was calculated from the relative responses
of the analyte in the sample extract and the spiked extracts.
The concentration of the analyte was derived by extrapolation.
Those analytes for which no suitable standards were available,
were semi-quantified using (±)-catechin, phloridzin and tomatidine as standards (see Table 4). The strategy of standard selection
was based taking into consideration physicochemical properties
(related with elution times) of each compound. In all cases, recoveries of the internal standards were above 89%, for both dichlorvos-d6
and dimethoate-d6. These standards allowed us to verify that the

extraction method performance and the analysis were satisfactory.


72

M.J. Martínez Bueno et al. / J. Chromatogr. A 1546 (2018) 66–76

Table 4
Estimated concentration ranges and mean concentrations of the markers identified in organic and conventional tomato samples during a full harvest cycle (␮g/kg FW).
Organic

L-tyrosyl-L-isoleucyl-L-threonyl-L-threonine a
Trilobatin b
Phloridzin
Tomatine
Phloretin
Echinenone c
a
b
c

Conventional

Conc.range

MeanConc.

Conc.range

MeanConc.

20–169
75–1014
49–292
6–36
186–1677
4–22

105
337
182
20
557
13

–
14–20
–
3–8
20–73
3–7

–
12
–
5
46
5

compounds semi-quantified with catechin as standard.

compounds semi-quantified with phloridzin as standard.
compounds semi-quantified with tomatidine as standard.

3.3. Determination of markers by HRAMS
The high level of selectivity and sensitivity of the HRAMS
approach allowed the characterization of six food components
present at trace levels. The quantification data allowed us to compare the intensities and the seasonal trend during a complete
farming campaign for all markers identified, both in organic and
conventional crops (see Fig. 2). All components were detected
in all organic tomato samples whereas only four of them were
found in conventional samples (trilobatin, tomatine, echinenone
and phloretin). Therefore, two bioactive compounds (L-Tyr-L-Ile-LThr-L-Thr and phloridzin) were only found in production systems
without the usage of synthetic fertilizers nor pesticides.
Tomatoes are a major source of polyphenolic compounds in
the human diet. In line with previous statements, the number of
polyphenolic compounds occurring in plants varies according to
the agricultural practice, environmental factors, harvest time and
the fertilisation mode [18,30]. However, in this study, a similar
trend could be observed according to each family group. Thus,
two flavonoids (trilobatin and phloretin) showed greater intensity
at the beginning of the harvest. By contrast, phloridzin’s maxima levels occurred during the mid-crop period similarly, to the
oligopeptide L-Tyr-L-Ile-L-Thr-L-Thr, to the glycoalkaloid tomatine
and to the xanthophyll echinenone (see Fig. 2).
Table 4 shows the concentration ranges and mean concentrations of the markers identified in organic and conventional tomato
samples during a full harvest cycle (␮g/kg FW). To date, some
flavonoids (including flavonols, flavanones, flavones and chalcones)
have been identified in different tomato varieties. However, there is
still a limited knowledge about dihydrochalcone contents (such as
phloretin or phloridzin-C-diglycoside) in tomato. In 2008, Slimestad et al. [31] identified the compound phloretin for the first
time in tomatoes. Phloretin was the dihydrochalcone found at the

highest concentration levels, ranging from 186 to 1677 ␮g/kg FW
(557 ␮g/kg FW of mean concentration) in organic tomato samples to 20–73 ␮g/kg FW (46 ␮g/kg FW of mean concentration) in
conventional tomatoes. The higher content of phloretin in organic
tomato crops has also been confirmed by Anton et al. [9]. However, the values found in the conventional grown tomatoes were
lower than those reported in another study carried out with tomatoes from the same geographic zone but of different varieties
(raf, daniela, rambo) [32]. Trilobatin levels were also higher in
organic tomatoes with a mean concentration of 337 ␮g/kg FW in
organic tomatoes (ranging from 75 to 1014 ␮g/kg FW), whereas
12 ␮g/kg FW were the levels found in conventional tomatoes (ranging from 14 to 20 ␮g/kg FW). Phloretin and trilobatin were the
markers which showed the biggest differences between harvest
time. By contrast, phloridzin and L-Tyr-L-Ile-L-Thr-L-Thr were the
markers which showed the biggest differences between agricultural practices. Phloridzin was found at levels ranging from 49
to 292 ␮g/kg FW whereas the oligopeptide content, L-Tyr-L-IleL-Thr-L-Thr, ranged from 20 to 169 ␮g/kg FW, in organic tomato

samples. The results showed that L-Tyr-L-Ile-L-Thr-L-Thr concentrations were low in immature tomatoes and increase rapidly as
the fruit matures. No data on contents of these three bioactive
compounds in tomatoes grown under organically practices (trilobatin, phloridzin and L-Tyr-L-Ile-L-Thr-L-Thr) have been found in
literature.
The smaller differences between organic and conventional
tomatoes happen for the glycoalkaloid tomatine and the xanthophyll echinenone. Organic tomatoes contained a mean concentration of 20 ␮g/kg FW of tomatine, whereas conventional tomatoes
contained 5 ␮g/kg FW. These values are in agreement with data
found in literature. Koh et al. [18] reported that the tomatine content in same variety of tomatoes harvested from plants grown
under “organic” soil and conditions was about twice that plants
grown under conventional conditions.
The concentration of echinenone ranged from 4 to 22 ␮g/kg FW
in organic tomato samples to 3–7 ␮g/kg FW in conventional tomatoes. Organic tomatoes contain an average content of 13 ␮g/kg FW,
meanwhile conventional tomatoes had a lower concentration of
5 ␮g/kg FW. The results found were up to three orders of magnitude lower than those reported for the ␤-carotene in conventional
tomato crops [33].
3.4. Pesticide residues

Pesticide residues were evaluated to determine their concentration in both produced tomato crops. It permitted understanding
in what extension these differences (in organic/conventional samples) can support the authentication as well as to correlate the
effects of the presence of pesticides on the secondary metabolites
content. Up until now, some authors have commented that the different content of bioactive compounds in organic and conventional
agricultural practices can be related with the absence of synthetic
pesticides in organic farming [17]. Therefore, that evaluation can
support this correlation.
The analysis data are presented in Table 5. The results showed
that within the scope of the pesticides analyzed (160 in total),
7 pesticides (chlorantraniliprole, cyromazine, fenhexamid, fenpyroximate, pymetrozine, spiromesifen, triadimenol) were found in
the conventional tomato samples. The insecticide chlorantraniliprole was detected in all conventionally grown tomato samples
analyzed at concentrations of 23 and 358 ␮g/kg. The insecticide
spiromesifen was present in four samples (57% positive samples)
at concentrations between 38 and 536 ␮g/kg FW. The acaricide
fenpyroximate, the insecticide pymetrozine and the fungicide triadimenol were found in three samples (43% positive samples) up
to levels of 179, 177 and 197 ␮g/kg FW, respectively. The triazine
insecticide cyromazine and the fungicide fenhexamid were quantified in one sample at concentrations of 532 and 33 ␮g/kg FW,
respectively. In all cases, the levels were below the maximum
residue limit (MRL) values established in the EU regulation [3–5]
for tomato, in agreement with good agricultural practices (GAP).


M.J. Martínez Bueno et al. / J. Chromatogr. A 1546 (2018) 66–76

73

Fig. 2. Seasonal trend corresponding to a complete farming campaign for all bioactive compounds identified as markers in organic tomato crops by HRAMS.

Table 5
Pesticide residues levels (␮g/kg FW) found in conventional and organic tomato samples, chemical class and main use.

Pesticide

Chemical class

Main use

Sample
type

MRLs in Tomato
(␮g/kg)

Chlorantraniliprole
Spiromesifen
Spinosad
Fenpyroximate
Pymetrozine
Triadimenol
Cyromazine
Fenhexamid

Anthranilic diamide
Tetronic acid
Spinosyn
Pyrazole
Pyridine
Triazole
Triazine
Hydroxyanilide

Insecticide
Insecticide
Insecticide
Acaricide
Insecticide
Fungicide
Insecticide
Fungicide

C
C
O
C
C
C
C
C

600
1.000
700
200
500
1.000
600
2.000

Concentration levels in samples (␮g/kg FW)

Positive

samples

Nov

Dec

Jan

Feb

Mar

Apr

May

57
38
37
–
177
33
–
–

23
323
39
–
94

–
–
–

60
536
–
–
19
–
–
–

135
424
–
–
–
–
–
–

358
–
–
156
–
25
–
33

217
–
100
136
–
–
–
–

301
–
31
179
–
197
532
–

100%
57%
57%
43%
43%
43%
14%
14%

Spinosad: sum of spinosyn A + D; C: conventional tomatoes; O: organic tomatoes; MRL: Maximun residue limit.

By contrast, only one pesticide (spinosad) was detected in the
organically grown tomatoes. Spinosad is a novel mode-of-action
insecticide derived from a family of natural products obtained
from the fermentation of Saccharopolyspora spinosa. According
to the Regulation (EC) No 834/2007, its use in organic production is allowed [4]. The greater concentrations of the insecticide
spinosad were found at the beginning and at the end of the harvest
(<100 ␮g/kg FW) (see Table 5). No pesticide residue was detected
in the mid-period, where the majority of bioactive compounds
showed higher levels (phloridzin, L-Tyr-L-Ile-L-Thr-L-Thr, tomatine
and echinenone). Based on these preliminary results, small variations in the content of these organic markers could be explained
with the spinosad levels determined in the organic tomato production. A 2–5-fold decrease in phloridzin, L-Tyr-L-Ile-L-Thr-L-Thr,
tomatine and echinenone’s concentration ranges took place when
the pesticide spinosad was detected. The residue levels of spinosad
were approx. lower than 10 times its maximum authorized residue
levels (MRLs) in tomato (0.7 mg/kg) [4].
Although the use of synthetic pesticides is not allowed in organic
farming, pesticide residues can occur in organic practices because of

field cross-contamination or contamination during storage. There
are not enough studies to evaluate what levels can be considered
acceptable as consequence of the cross contamination. In general,
low levels of 1–2 residues are typically accepted. High residue
concentrations, the presence of metabolites or a high number of
residues are clearly indicative of illegal use and support the classification in organic and conventional production. But, in any case,
they are not relevant in the evaluation of chemical fertilizers applications.
3.5. IRMS results
Stable nitrogen isotope analysis (␦15 N) was also assessed as a
possible indicator to support the discrimination between both production systems using crop/fertilizer correlations. Table 6 shows
the ␦15 N values found in organic and conventional tomato crops in
greenhouses in Southeastern Spain (Almería), which is a Mediterranean region leading in vegetable production in the EU.

The tomatoes’ nitrogen isotope fingerprint showed a clear difference between tomatoes grown using organic and synthetic


74

M.J. Martínez Bueno et al. / J. Chromatogr. A 1546 (2018) 66–76

Table 6
␦15 N values found in organic and conventional tomato grown in greenhouse under
controlled agronomic conditions during a complete farming campaign.

␦15 N (‰) – Organic
␦15 N (‰) – Conventional

NOV

DEC

JAN

FEB

MAR

APR

MAY

15.7
6.7

15.3
7.5

13.5
4.9

11.8
2.5

11.7
5.2

11.7
5.0

9.8
4.5

fertilizers. In accordance with previously published scientific studies, tomatoes from organic production showed higher ␦15 N values
(+9.8 to +15.7‰) than those tomatoes grown using synthetic fertilizers, such as potash and ammonia (+2.5 to +7.5‰). Therefore,
in view of these results it is possible to think that ␦15 N data are
appropriate for distinguishing the use of organic versus synthetic
fertilizers, and thus provide a linkage to the production system.
However, a reliable threshold could not be established after a full
harvest cycle. Bateman et al. [34] evaluated organic and conventional tomatoes grown in greenhouses within the United Kingdom
over a 2-year period. The authors reported a mean ␦15 N value of
8.1‰ for the organically grown tomatoes compared with a mean
value of −0.1‰ for those grown conventionally. The extent of these
differences can depend on the fertilizer ␦15 N value, the rates of

application over a given period of time, but also other N transformation processes prior to plant uptake, such as biological N2 fixation,
ammonia volatilization, denitrification, nitrification and as well as
plant organs and plant age [35].
In our study, a decrease in ␦15 N values was observed in both
farming systems along the production cycle, although more especially in organic production. The equations of the trend line of
␦15 N data were y = −0,0321x + 1386,4 and y = −0,0143x + 618,69

for organic and conventional tomato farming. This trend can be
explained by the fact that in conventional agriculture, the use of
synthetic fertilizers (NO3 − /NH4 + /urea) was done throughout all the
crop cycle, whereas in the organic crops the soil fertilisation was
mainly done at the beginning of the season.
On the other hand, the influence of the water used for irrigation
on the ␦15 N of the plants was also evaluated in this study, since
the NO3 − content and ␦15 N composition of supply water vary geographically, which can help us to better understanding the results.
High concentrations of nitrate were found in the water used for
irrigation in the studied area (Southeastern Spain: values around
75 mg/L NO3 − ). In this area, the irrigation water used was basically groundwater. ␦15 N values of irrigation water were estimated
between +4‰ and +6‰, which explains the higher ␦15 N levels
found in both production systems in comparison with others study
from different geographic zone [34].
In summary, ␦15 N data can only provide supporting evidence in
suspected fraud cases, but not for discriminating between both production systems. Since it has been commented by previous authors,
the application of a small input of manure or the use of water with
a large concentration of nitrate (NO3 − ) can result in an increase of
the ␦15 N values, close to those obtained in organic grows [12,36].
3.6. Chemometric analysis
Metabolic profiling combined with chemometric techniques is
an emerging analytical tool that is being increasingly applied to differentiate conventional and organic products [17,19]. In our study,
the concentration data obtained for the six markers identified by

Fig 3. PCA graphs employing software R using only markers’ HRAMS data (A) and (C); and including HRAMS & IRMS data (B) and (D). PCs: Number of principal components
used. PCs: Number of principal components used.


M.J. Martínez Bueno et al. / J. Chromatogr. A 1546 (2018) 66–76

HRAMS analysis and ␦15 N data values obtained by IRMS analysis
were selected for the chemometric analysis. Data from these two
sources were concatenated and pre-processed with log scaling [37].
A principal component analysis (PCA) was applied to differentiate
the statistical significance between both farming systems (organic
vs conventional). A PCA graph allowed us the visualisation of multidimensional information in the form of a few principal components
(PCs) retaining the maximum possible variability within the data
set. The results were evaluated in terms of recognition ability,
which represents the percentage of successfully classified samples
in the training set. As it is presented in Fig. 3, the PCA graph using
only HRAMS data showed an efficient separation between organic
and conventional tomato crops (recognition ability 84.6% and 6
PCs). Comparable results between both practices could be observed
when the statistical model included HRAMS and IRMS data (recognition ability 83.7% and 7 PCs). Thus, similar recognition abilities
were obtained in both studies. These results put in evidence that
the MS profiling data had stronger relevance on the sample clustering according to farming systems than ␦15 N data. Moreover, if only
the flavonoid data (trilobatin, phloridzin and phloretin) were used
for the chemometric analysis, the recognition ability increased up
to 95%, achieving an improvement in the differentiation model of
organically and conventionally grown tomatoes (HRAMS data only
and 3 PCs). In a recent study carried out employing Direct Analysis
in Real Time (DART)-TOF-MS, the authors distinguished between
tomato samples from both production system with a recognition

and a prediction ability of 80% and 65%, respectively, using 13 PCs
[17].
Finally, a linear discriminant analysis (LDA) was applied to
establish a predictive model based on the six markers (L-TyrL-Ile-L-Thr-L-Thr, trilobatin, phloridzin, tomatine, phloretin and
echinenone) for sample classification in terms of prediction abilities. This parameter gives information about the percentage of
correctly classified samples in the test set by the model developed during the training set by PCA. A total of 11 tomato samples
acquired from different local markets (8 labeled as “organic” and
3 as “non-organic”) were analyzed in order to check the classification capability of the developed model. Since the results obtained
in the work put in evidence that the MS profiling data had stronger
relevance on the sample clustering according to farming systems
than ␦15 N data, only HRAMS analysis based on the assessment of
the six identified markers was used to classify the samples. The
intensities found for them were used for the sample classification in terms of prediction abilities by LDA. Detailed information
about HRAMS data found for the six markers in market samples
has been included as Supplementary Material (Table S.1). The prediction ability was 73%. Thus, the model concluded that about 27%
of the market tomatoes showed some irregularity. Out of the 11
samples analyzed, 8 were correctly classified. All the conventional
tomato samples purchased (3) were correctly classified, whereas
the model classified 3 of the tomato samples labeled as “organic”
as “non-organic”. Then, in order to understand this classification,
a pesticide residues screening was performed on these 3 tomato
samples labeled as “organic” to correlate the low content of these
organic markers with the possible presence of pesticides in the
“organic” samples. Although they should be free of pesticides, the
presence of 1 or 2 pesticide residues (spinosad and pyrimethanil)
were detected. In all cases, the concentration levels found were
below their MRLs in tomatoes, 0.7 and 1 mg/kg, respectively. As it
was previously commented pesticide residues can occur in organic
practices because of field cross-contamination or contamination
during storage. In any case, although the prediction ability value

may seem relatively low (73%), it can still be considered as acceptable since different origin, irrigation, fertilising and plant protection
systems were considered. A similar value has been reported in
other work (80%) in which organically and conventionally grown

75

tomatoes under controlled agronomic conditions were compared
[17].
4. Conclusions
Metabolomic fingerprinting of tomato samples performed by
HRAMS in full scan and fragmentation analysis has shown to be
a decisive analytical approach for the identification of bioactive
components of interest to distinguish between conventional and
organic tomato production practices. Methanol was selected as
the extraction solvent because it offered better performance than
acetonitrile. The study has provided the tentative identification
of six markers of organic tomato samples, two of them, L-TyrL-Ile-L-Thr-L-Thr and trilobatin, as far as we know, have been
reported and evaluated for the first time in tomato. This study supports the fact that authentication studies should not be based on
IRMS analysis only, since the variation of organic matter and/or
nutrients incorporated during the crop, can modify the ␦15 N threshold data. Seven pesticides were detected in the conventionally
grown tomato samples whereas only one insecticide (spinosad)
was found in organic tomatoes. The combination of multivariate
statistical analysis (MSA) of MS profiling and ␦15 N data is a useful
approach for sample clustering according to farming production
systems. However, MS profiling data are limited to specific crops.
Therefore, continuous buildup of HRAMS databases obtained under
controlled agronomic conditions with different vegetable varieties
and different geographical locations can facilitate to help ensure
the authenticity of organic production as well as the detection of
possible fraud cases.

Acknowledgements
M.J. Martínez Bueno thanks the Marie Skłodowska-Curie
Actions (MSCA) for the individual fellowships (H2020-MSCAIF-2015 #707816 ORGANIC QUAL TRACERS). The authors also
gratefully acknowledge “BIOSABOR S.A.T” for assistance with
tomato sample production and harvest. The authors also would like
to thank Thermo Fisher Scientific for the instrumentation facilities.
Appendix A. Supplementary data
Supplementary data associated with this article can be found,
in the online version, at />002.
References
[1] FAOSTAT Database 2014. />[2] EUROSTAT Database 2014. />[3] EC Regulation 834/2007. Council Regulation 834/2007/EC of 28 June 2007 on
organic production and labelling of organic products. Official Journal of the
European Union L189, 2007, 1-23.
[4] EC Regulation 889/2008. Commission Regulation 889/2008/EC of 5 September
2008 laying down detailed rules for the implementation of Council Regulation
(EC) No 834/2007 on organic production and labelling of organic products
with regard to organic production, labelling and control. Official Journal of the
European Union L250, 2008, 1–84.
[5] Action for the future of Organic Production in E.U, 2014. />en/events/Documents/COM(2014)%20179%20final.pdf.
[6] EFSA, The 2013 European Union report on pesticide residues in food, EFSA J.
13 (2015) 4038.
[7] Working document of the Commission services on official controls in the
organic sector. en.pdf.
[8] A. Ribas-Agustí, P. Bouchagier, E. Skotti, D. Erba, C. Casiraghi, C. Sárraga, M.
Castellari, Effects of different organic anti-fungal treatments on tomato plant
productivity and selected nutritional components of tomato fruit, J. Hortic.
Sci. Biotechnol. 88 (2013) 67–72.
[9] D. Anton, D. Matt, P. Pedastsaar, I. Bender, R. Kazimierczak, M. Roasto, T. Kaart,
A. Luik, T. Püssa, Three-year comparative study of polyphenol contents and
antioxidant capacities in fruits of tomato (Lycopersicon esculentum Mill.)

76

[10]

[11]

[12]

[13]

[14]

[15]
[16]
[17]

[18]

[19]

[20]

[21]

[22]

M.J. Martínez Bueno et al. / J. Chromatogr. A 1546 (2018) 66–76
cultivars grown under organic and conventional conditions, J. Agric. Food

Chem. 62 (2014) 5173–5180.
M. Hohmann, Y. Monakhova, S. Erich, N. Christoph, H. Wachter, U. Holzgrabe,
Differentiation of organically and conventionally grown tomatoes by
chemometric analysis of combined data from proton nuclear magnetic
resonance and mid-infrared spectroscopy and stable isotope analysis, J. Agric.
Food Chem. 63 (2015) 9666–9675.
K.H. Laursen, J.K. Schjoerring, S.D. Kelly, S. Husted, Authentication of
organically grown plants – advantages and limitations of atomic spectroscopy
for multi-element and stable isotope analysis, TrAC 59 (2014) 73–82.
F. Camin, M. Perini, L. Bontempo, S. Fabroni, W. Faedi, S. Magnani, G. Baruzzi,
M. Bonoli, M.R. Tabilio, S. Musmeci, A. Rossmann, S.D. Kelly, P. Rapisarda,
Potential isotopic and chemical markers for characterizing organic fruits,
Food Chem. 125 (2011) 1072–1082.
S. Wagner, K. Scholz, M. Donegan, L. Burton, J. Wingate, W. Völkel,
Metabonomics and biomarker discovery: LC–MS metabolic profiling and
constant neutral loss scanning combined with multivariate data analysis for
mercapturic acid analysis, Anal. Chem. 78 (2006) 1296–1305.
M. Carbonaro, M. Mattera, S. Nicoli, P. Bergamo, M. Cappelloni, Modulation of
antioxidant compounds in organic vs conventional fruit (peach Prunus persica
L., and pear, Pyrus communis L.), J. Agric. Food Chem. 50 (2012) 5458–5462.
E. Rembialkowska, Quality of plant products from organic agriculture, J. Sci.
Food Agric. 87 (2007) 2757–2762.
D. Lairon, Nutritional quality and safety of organic food. A review, Agron.
Sustain. Dev. 30 (2009) 33–41.
˛
H. Novotná, O. Kmiecik, M. Gałazka,
V. Krtková, A. Hurajová, V. Schulzová, E.
Hallmann, E. Rembiałkowska, J. Hajˇslová, Metabolomic fingerprinting
employing DART-TOFMS for authentication of tomatoes and peppers from
organic and conventional farming, Food Addit. Contam. 29 (2012) 1335–1346.

E. Koh, S. Kaffka, E. Mitchell, A long-term comparison of the influence of
organic and conventional crop management practices on the content of the
glycoalkaloid ␣-tomatine in tomatoes, J. Sci. Food Agric. 93 (2013) 1537–1542.
A. Vallverdú-Queralt, O. Jáuregui, A. Medina-Remón, R.M. Lamuela-Raventós,
Evaluation of a method to characterize the phenolic profile of organic and
conventional tomatoes, J. Agric. Food Chem. 60 (2012) 3373–3380.
L. Rajski, M.M. Gómez-Ramos, A.R. Fernández-Alba, Large pesticide
multiresidue screening method by liquid chromatography-Orbitrap mass
spectrometry in full scan mode applied to fruit and vegetables, J. Chromatogr.
A 1360 (2014) 119–127.
A. de Juan, R. Tauler, Factor analysis of hyphenated chromatographic data:
exploration, resolution and quantification of multicomponent systems, J.
Chromatogr. A 1158 (2007) 184–195.
M.D. Gil García, M.J. Culzoni, M.M. De Zan, R. Santiago Valverde, M. Martinez
Galera, H.C. Goicoechea, Solving matrix-effects exploiting the second order
advantage in the resolution and determination of eight tetracycline
antibiotics in effluent wastewater by modelling liquid chromatography data

[23]

[24]

[25]

[26]

[27]
[28]
[29]

[30]

[31]

[32]

[33]

[34]

[35]

[36]

[37]

with multivariate curve resolution-alternating least squares and
unfolded-partial least squares followed by residual bilinearization algorithms,
J. Chromatogr. A 1179 (2008) 106–124.
K. Woese, D. Lange, C. Boess, K.W. Bögl, A comparison of organically and
conventionally grown foods – results of a review of the relevant literature, J.
Sci. Food Agric. 74 (1997) 281–293.
H. Willer, L. Kilcher, The world of organic agriculture. Statistics and
emergingtrends 2011. FiBL-IFOAM Report. Bonn, Germany/Frick, Switzerland:
IFOAM and FiBL. Version 1.3, 2017.
M.D. Raigón, A. Rodríguez-Burruezo, J. Prohens, Effects of organic and
conventional cultivation methods on composition of eggplant fruits, J. Agric.
Food Chem. 58 (2010) 6833–6840.
DG-SANTE, European Commission, Guidance Document on Analytical Quality
Control and Method Validation Procedures for Pesticides Residues Analysis in

Food and Feed Document No. SANTE/11945/2015, 2016.
ISO/IEC 17025, General Requirements for the competence of testing and
calibration laboratories, ISO, Geneva, Switzerland, 2005.
Commission Decision 2002/657/EC, Off. J. Eur. Communities L221 (2002) 8.
F. Carnevale Neto, T. Guaratini, L. Costa-Lotufo, P. Colepicolo, P.J. Gates, N.
Peporine Lopes, Re-investigation of the fragmentation of protonated
carotenoids by electrospray ionization and nanospray tandem mass
spectrometry, Rapid Commun. Mass Spectrom. 30 (2016) 1540–1548.
E. Hallmann, The influence of organic and conventional cultivation systems
on the nutritional value and content of bioactive compounds in selected
tomato types, J. Sci. Food Agric. 92 (2012) 2840–2848.
R. Slimestad, M. Verheul, Review of flavonoids and other phenolics from fruits
of different tomato (Lycopersicon esculentum Mill) cultivars, J. Sci. Food Agric.
89 (2009) 1255–1270.
M. Gómez-Romero, A. Segura-Carretero, A. Fernández-Gutiérrez, Metabolite
profiling and quantification of phenolic compounds in methanol extracts of
tomato fruit, Phytochemistry 71 (2010) 1848–1864.
N. Kozukue, M. Friedman, Tomatine chlorophyll, b-carotene and lycopene
content in tomatoes during growth and maturation, J. Sci. Food Agric. 83
(2003) 195–200.
A.S. Bateman, S.D. Kelly, M. Woolfe, Nitrogen isotope composition of
organically and conventionally grown crops, J. Agric. Food Chem. 55 (2007)
2664–2670.
C.T. Inácio, P.M. Chalk, A.M. Magalhães, Principles and limitations of stable
isotopes in differentiating organic and conventional foodstuffs: 1 Plant
products, Crit. Rev. Food Sci. Nutr. 55 (2015) 1206–1218.
A.S. Bateman, S.D. Kelly, T.D. Jickells, Nitrogen isotope relationships between
crops and fertilizer: implications for using nitrogen isotope analysis as an
indicator of agricultural regime, J. Agric. Food Chem. 53 (2005) 5760–5765.
˜ O. Busto, Data fusion

E. Borràs, J. Ferré, R. Boqué, M. Mestres, L. Acena,
methodologies for food and beverage authentication and quality assessment
– a review, Anal. Chim. Acta 891 (2015) 1–14.